2023-04-26 10:07:06,675 INFO [finetune.py:1046] (1/7) Training started 2023-04-26 10:07:06,675 INFO [finetune.py:1056] (1/7) Device: cuda:1 2023-04-26 10:07:06,678 INFO [finetune.py:1065] (1/7) {'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': '62e404dd3f3a811d73e424199b3408e309c06e1a', 'k2-git-date': 'Mon Jan 30 02:26:16 2023', 'lhotse-version': '1.12.0.dev+git.3ccfeb7.clean', 'torch-version': '1.13.0', 'torch-cuda-available': True, 'torch-cuda-version': '11.7', 'python-version': '3.8', 'icefall-git-branch': 'master', 'icefall-git-sha1': 'd74822d-dirty', 'icefall-git-date': 'Tue Mar 21 21:35:32 2023', 'icefall-path': '/home/lishaojie/icefall', 'k2-path': '/home/lishaojie/.conda/envs/env_lishaojie/lib/python3.8/site-packages/k2/__init__.py', 'lhotse-path': '/home/lishaojie/.conda/envs/env_lishaojie/lib/python3.8/site-packages/lhotse/__init__.py', 'hostname': 'cnc533', 'IP address': '127.0.1.1'}, 'world_size': 7, 'master_port': 18181, 'tensorboard': True, 'num_epochs': 30, 'start_epoch': 1, 'start_batch': 0, 'exp_dir': PosixPath('pruned_transducer_stateless7_streaming/exp2'), 'bpe_model': 'data/lang_bpe_500/bpe.model', 'base_lr': 0.004, 'lr_batches': 100000.0, 'lr_epochs': 100.0, '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, '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, 'do_finetune': True, 'init_modules': 'encoder', 'finetune_ckpt': '/home/lishaojie/icefall/egs/commonvoice_fr/ASR/pruned_transducer_stateless7_streaming/exp/english_pretrain/pretrained.pt', 'manifest_dir': PosixPath('data/fbank'), 'max_duration': 200, '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} 2023-04-26 10:07:06,678 INFO [finetune.py:1067] (1/7) About to create model 2023-04-26 10:07:07,019 INFO [zipformer.py:405] (1/7) 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. 2023-04-26 10:07:07,028 INFO [finetune.py:1071] (1/7) Number of model parameters: 70369391 2023-04-26 10:07:07,028 INFO [finetune.py:626] (1/7) Loading checkpoint from /home/lishaojie/icefall/egs/commonvoice_fr/ASR/pruned_transducer_stateless7_streaming/exp/english_pretrain/pretrained.pt 2023-04-26 10:07:07,158 INFO [finetune.py:647] (1/7) Loading parameters starting with prefix encoder 2023-04-26 10:07:08,505 INFO [finetune.py:1093] (1/7) Using DDP 2023-04-26 10:07:09,252 INFO [commonvoice_fr.py:392] (1/7) About to get train cuts 2023-04-26 10:07:09,254 INFO [commonvoice_fr.py:218] (1/7) Enable MUSAN 2023-04-26 10:07:09,254 INFO [commonvoice_fr.py:219] (1/7) About to get Musan cuts 2023-04-26 10:07:10,740 INFO [commonvoice_fr.py:243] (1/7) Enable SpecAugment 2023-04-26 10:07:10,740 INFO [commonvoice_fr.py:244] (1/7) Time warp factor: 80 2023-04-26 10:07:10,740 INFO [commonvoice_fr.py:254] (1/7) Num frame mask: 10 2023-04-26 10:07:10,740 INFO [commonvoice_fr.py:267] (1/7) About to create train dataset 2023-04-26 10:07:10,741 INFO [commonvoice_fr.py:294] (1/7) Using DynamicBucketingSampler. 2023-04-26 10:07:13,440 INFO [commonvoice_fr.py:309] (1/7) About to create train dataloader 2023-04-26 10:07:13,441 INFO [commonvoice_fr.py:399] (1/7) About to get dev cuts 2023-04-26 10:07:13,442 INFO [commonvoice_fr.py:340] (1/7) About to create dev dataset 2023-04-26 10:07:13,848 INFO [commonvoice_fr.py:357] (1/7) About to create dev dataloader 2023-04-26 10:07:13,848 INFO [finetune.py:1289] (1/7) Sanity check -- see if any of the batches in epoch 1 would cause OOM. 2023-04-26 10:11:06,931 INFO [finetune.py:1317] (1/7) Maximum memory allocated so far is 4472MB 2023-04-26 10:11:07,628 INFO [finetune.py:1317] (1/7) Maximum memory allocated so far is 5381MB 2023-04-26 10:11:08,317 INFO [finetune.py:1317] (1/7) Maximum memory allocated so far is 5381MB 2023-04-26 10:11:08,989 INFO [finetune.py:1317] (1/7) Maximum memory allocated so far is 5381MB 2023-04-26 10:11:09,669 INFO [finetune.py:1317] (1/7) Maximum memory allocated so far is 5381MB 2023-04-26 10:11:10,371 INFO [finetune.py:1317] (1/7) Maximum memory allocated so far is 5381MB 2023-04-26 10:11:19,553 INFO [finetune.py:976] (1/7) Epoch 1, batch 0, loss[loss=7.459, simple_loss=6.764, pruned_loss=6.933, over 4810.00 frames. ], tot_loss[loss=7.459, simple_loss=6.764, pruned_loss=6.933, over 4810.00 frames. ], batch size: 38, lr: 2.00e-03, grad_scale: 2.0 2023-04-26 10:11:19,553 INFO [finetune.py:1001] (1/7) Computing validation loss 2023-04-26 10:11:40,202 INFO [finetune.py:1010] (1/7) Epoch 1, validation: loss=7.31, simple_loss=6.623, pruned_loss=6.857, over 2265189.00 frames. 2023-04-26 10:11:40,203 INFO [finetune.py:1011] (1/7) Maximum memory allocated so far is 5381MB 2023-04-26 10:11:48,671 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=5.0, num_to_drop=2, layers_to_drop={2, 3} 2023-04-26 10:11:59,753 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=28.16 vs. limit=5.0 2023-04-26 10:12:02,443 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=2.02 vs. limit=2.0 2023-04-26 10:12:11,181 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=23.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 10:12:31,701 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.3848, 2.0785, 1.6094, 1.9407, 2.0741, 1.8749, 2.0387, 1.6705], device='cuda:1'), covar=tensor([0.0484, 0.0312, 0.0496, 0.0409, 0.0372, 0.0402, 0.0381, 0.0498], device='cuda:1'), in_proj_covar=tensor([0.0380, 0.0392, 0.0472, 0.0406, 0.0454, 0.0426, 0.0443, 0.0455], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 10:12:43,279 INFO [finetune.py:976] (1/7) Epoch 1, batch 50, loss[loss=2.529, simple_loss=2.408, pruned_loss=1.234, over 4739.00 frames. ], tot_loss[loss=4.406, simple_loss=3.983, pruned_loss=4.092, over 215351.28 frames. ], batch size: 54, lr: 2.20e-03, grad_scale: 0.00390625 2023-04-26 10:12:44,914 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=30.27 vs. limit=5.0 2023-04-26 10:13:18,325 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=35.70 vs. limit=5.0 2023-04-26 10:13:19,828 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=83.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 10:13:29,969 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.65 vs. limit=2.0 2023-04-26 10:13:40,194 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=7.31 vs. limit=5.0 2023-04-26 10:13:40,724 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=22.85 vs. limit=5.0 2023-04-26 10:13:41,117 WARNING [finetune.py:966] (1/7) Grad scale is small: 6.103515625e-05 2023-04-26 10:13:41,118 INFO [finetune.py:976] (1/7) Epoch 1, batch 100, loss[loss=2.229, simple_loss=2.111, pruned_loss=1.18, over 4871.00 frames. ], tot_loss[loss=3.398, simple_loss=3.138, pruned_loss=2.545, over 377845.48 frames. ], batch size: 34, lr: 2.40e-03, grad_scale: 0.0001220703125 2023-04-26 10:14:02,043 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.3203, 1.7153, 1.2645, 2.1646, 1.8643, 1.9245, 1.6932, 3.5307], device='cuda:1'), covar=tensor([0.0211, 0.0286, 0.0302, 0.0342, 0.0227, 0.0196, 0.0286, 0.0185], device='cuda:1'), in_proj_covar=tensor([0.0039, 0.0042, 0.0042, 0.0047, 0.0044, 0.0041, 0.0042, 0.0067], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0013, 0.0013, 0.0015, 0.0014, 0.0013, 0.0014, 0.0018], device='cuda:1') 2023-04-26 10:14:03,496 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 4.683e+02 1.507e+03 7.833e+03 2.572e+04 3.214e+07, threshold=1.567e+04, percent-clipped=0.0 2023-04-26 10:14:06,180 WARNING [optim.py:389] (1/7) Scaling gradients by 0.014711554162204266, model_norm_threshold=15666.9306640625 2023-04-26 10:14:06,253 INFO [optim.py:451] (1/7) Parameter Dominanting tot_sumsq module.encoder.encoder_embed.conv.0.weight with proportion 0.88, where dominant_sumsq=(grad_sumsq*orig_rms_sq)=1.001e+12, grad_sumsq = 2.310e+12, orig_rms_sq=4.331e-01 2023-04-26 10:14:23,410 WARNING [optim.py:389] (1/7) Scaling gradients by 0.00018281130178365856, model_norm_threshold=15666.9306640625 2023-04-26 10:14:23,483 INFO [optim.py:451] (1/7) Parameter Dominanting tot_sumsq module.encoder.encoder_embed.conv.0.weight with proportion 0.29, where dominant_sumsq=(grad_sumsq*orig_rms_sq)=2.155e+15, grad_sumsq = 4.978e+15, orig_rms_sq=4.329e-01 2023-04-26 10:14:24,109 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=144.0, num_to_drop=2, layers_to_drop={1, 3} 2023-04-26 10:14:27,646 INFO [finetune.py:976] (1/7) Epoch 1, batch 150, loss[loss=1.63, simple_loss=1.483, pruned_loss=1.206, over 4824.00 frames. ], tot_loss[loss=2.887, simple_loss=2.674, pruned_loss=2.043, over 508581.58 frames. ], batch size: 38, lr: 2.60e-03, grad_scale: 3.0517578125e-05 2023-04-26 10:14:28,153 WARNING [optim.py:389] (1/7) Scaling gradients by 0.00022292081848718226, model_norm_threshold=15666.9306640625 2023-04-26 10:14:28,226 INFO [optim.py:451] (1/7) Parameter Dominanting tot_sumsq module.encoder.encoder_embed.conv.0.weight with proportion 0.45, where dominant_sumsq=(grad_sumsq*orig_rms_sq)=2.213e+15, grad_sumsq = 5.111e+15, orig_rms_sq=4.330e-01 2023-04-26 10:14:40,642 WARNING [optim.py:389] (1/7) Scaling gradients by 0.05655747279524803, model_norm_threshold=15666.9306640625 2023-04-26 10:14:40,715 INFO [optim.py:451] (1/7) Parameter Dominanting tot_sumsq module.encoder.encoder_embed.conv.0.weight with proportion 0.84, where dominant_sumsq=(grad_sumsq*orig_rms_sq)=6.482e+10, grad_sumsq = 1.497e+11, orig_rms_sq=4.330e-01 2023-04-26 10:14:56,531 WARNING [finetune.py:966] (1/7) Grad scale is small: 3.0517578125e-05 2023-04-26 10:14:56,531 INFO [finetune.py:976] (1/7) Epoch 1, batch 200, loss[loss=1.258, simple_loss=1.085, pruned_loss=1.193, over 4772.00 frames. ], tot_loss[loss=2.395, simple_loss=2.198, pruned_loss=1.769, over 609045.99 frames. ], batch size: 26, lr: 2.80e-03, grad_scale: 6.103515625e-05 2023-04-26 10:14:58,284 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=2.77 vs. limit=2.0 2023-04-26 10:15:00,261 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0671, 1.0180, 1.3890, 1.1749, 1.2898, 1.6330, 1.6307, 1.0339], device='cuda:1'), covar=tensor([0.0346, 0.0463, 0.0397, 0.0318, 0.0413, 0.0358, 0.0381, 0.0499], device='cuda:1'), in_proj_covar=tensor([0.0396, 0.0401, 0.0387, 0.0358, 0.0420, 0.0446, 0.0385, 0.0425], device='cuda:1'), out_proj_covar=tensor([8.7271e-05, 8.6458e-05, 8.3478e-05, 7.5048e-05, 9.0278e-05, 9.8009e-05, 8.4518e-05, 9.2069e-05], device='cuda:1') 2023-04-26 10:15:06,944 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=2.13 vs. limit=2.0 2023-04-26 10:15:07,790 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 6.387e+01 5.300e+02 1.840e+03 7.547e+03 8.570e+07, threshold=3.680e+03, percent-clipped=20.0 2023-04-26 10:15:12,965 WARNING [optim.py:389] (1/7) Scaling gradients by 0.011872046627104282, model_norm_threshold=3679.54541015625 2023-04-26 10:15:13,039 INFO [optim.py:451] (1/7) Parameter Dominanting tot_sumsq module.encoder.encoder_embed.conv.0.weight with proportion 0.57, where dominant_sumsq=(grad_sumsq*orig_rms_sq)=5.451e+10, grad_sumsq = 1.259e+11, orig_rms_sq=4.329e-01 2023-04-26 10:15:16,158 WARNING [optim.py:389] (1/7) Scaling gradients by 0.08515117317438126, model_norm_threshold=3679.54541015625 2023-04-26 10:15:16,231 INFO [optim.py:451] (1/7) Parameter Dominanting tot_sumsq module.encoder.encoder_embed.conv.0.weight with proportion 0.79, where dominant_sumsq=(grad_sumsq*orig_rms_sq)=1.483e+09, grad_sumsq = 3.425e+09, orig_rms_sq=4.329e-01 2023-04-26 10:15:16,776 WARNING [optim.py:389] (1/7) Scaling gradients by 0.04552413150668144, model_norm_threshold=3679.54541015625 2023-04-26 10:15:16,848 INFO [optim.py:451] (1/7) Parameter Dominanting tot_sumsq module.encoder.encoder_embed.conv.0.weight with proportion 0.84, where dominant_sumsq=(grad_sumsq*orig_rms_sq)=5.493e+09, grad_sumsq = 1.269e+10, orig_rms_sq=4.329e-01 2023-04-26 10:15:25,654 INFO [finetune.py:976] (1/7) Epoch 1, batch 250, loss[loss=1.427, simple_loss=1.22, pruned_loss=1.331, over 4739.00 frames. ], tot_loss[loss=2.071, simple_loss=1.874, pruned_loss=1.619, over 687891.71 frames. ], batch size: 59, lr: 3.00e-03, grad_scale: 6.103515625e-05 2023-04-26 10:15:27,810 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([4.0397, 4.3123, 2.8805, 4.4199, 4.3637, 4.1032, 2.5441, 4.0997], device='cuda:1'), covar=tensor([0.2587, 0.0892, 0.2315, 0.2758, 0.1682, 0.2599, 0.2944, 0.2214], device='cuda:1'), in_proj_covar=tensor([0.0264, 0.0236, 0.0289, 0.0332, 0.0329, 0.0272, 0.0292, 0.0286], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-26 10:15:34,606 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=2.21 vs. limit=2.0 2023-04-26 10:15:50,811 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=296.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 10:15:52,836 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=300.0, num_to_drop=2, layers_to_drop={2, 3} 2023-04-26 10:15:53,283 WARNING [finetune.py:966] (1/7) Grad scale is small: 6.103515625e-05 2023-04-26 10:15:53,283 INFO [finetune.py:976] (1/7) Epoch 1, batch 300, loss[loss=1.263, simple_loss=1.054, pruned_loss=1.225, over 4866.00 frames. ], tot_loss[loss=1.875, simple_loss=1.673, pruned_loss=1.535, over 748059.87 frames. ], batch size: 31, lr: 3.20e-03, grad_scale: 0.0001220703125 2023-04-26 10:15:58,656 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=2.40 vs. limit=2.0 2023-04-26 10:16:02,183 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=4.94 vs. limit=2.0 2023-04-26 10:16:02,526 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 2.002e+01 7.343e+01 2.302e+02 1.070e+03 3.099e+05, threshold=4.604e+02, percent-clipped=16.0 2023-04-26 10:16:05,686 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=3.78 vs. limit=2.0 2023-04-26 10:16:17,489 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=30.20 vs. limit=5.0 2023-04-26 10:16:21,499 INFO [finetune.py:976] (1/7) Epoch 1, batch 350, loss[loss=1.26, simple_loss=1.045, pruned_loss=1.19, over 4720.00 frames. ], tot_loss[loss=1.724, simple_loss=1.516, pruned_loss=1.461, over 794782.32 frames. ], batch size: 59, lr: 3.40e-03, grad_scale: 0.0001220703125 2023-04-26 10:16:24,672 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=357.0, num_to_drop=2, layers_to_drop={0, 1} 2023-04-26 10:16:24,774 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=3.43 vs. limit=2.0 2023-04-26 10:16:42,471 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=387.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 10:16:47,567 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.8443, 3.3242, 1.1158, 1.7182, 1.6021, 1.7173, 4.7696, 2.2855], device='cuda:1'), covar=tensor([0.0120, 0.0225, 0.0194, 0.0249, 0.0134, 0.0192, 0.0114, 0.0176], device='cuda:1'), in_proj_covar=tensor([0.0058, 0.0076, 0.0054, 0.0052, 0.0057, 0.0057, 0.0094, 0.0054], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0011, 0.0008, 0.0008, 0.0008, 0.0008, 0.0012, 0.0008], device='cuda:1') 2023-04-26 10:16:56,312 WARNING [finetune.py:966] (1/7) Grad scale is small: 0.0001220703125 2023-04-26 10:16:56,312 INFO [finetune.py:976] (1/7) Epoch 1, batch 400, loss[loss=1.228, simple_loss=0.9818, pruned_loss=1.23, over 4903.00 frames. ], tot_loss[loss=1.611, simple_loss=1.395, pruned_loss=1.408, over 831316.69 frames. ], batch size: 36, lr: 3.60e-03, grad_scale: 0.000244140625 2023-04-26 10:17:16,619 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.830e+01 2.642e+01 7.119e+01 5.229e+02 3.680e+03, threshold=1.424e+02, percent-clipped=26.0 2023-04-26 10:17:27,885 WARNING [optim.py:389] (1/7) Scaling gradients by 0.02133115753531456, model_norm_threshold=142.37583923339844 2023-04-26 10:17:27,959 INFO [optim.py:451] (1/7) Parameter Dominanting tot_sumsq module.encoder.encoder_embed.conv.0.weight with proportion 0.81, where dominant_sumsq=(grad_sumsq*orig_rms_sq)=3.619e+07, grad_sumsq = 8.362e+07, orig_rms_sq=4.328e-01 2023-04-26 10:17:41,636 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=439.0, num_to_drop=2, layers_to_drop={2, 3} 2023-04-26 10:17:52,431 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=448.0, num_to_drop=2, layers_to_drop={1, 2} 2023-04-26 10:17:53,954 INFO [finetune.py:976] (1/7) Epoch 1, batch 450, loss[loss=1.085, simple_loss=0.857, pruned_loss=1.071, over 4903.00 frames. ], tot_loss[loss=1.511, simple_loss=1.286, pruned_loss=1.358, over 860333.91 frames. ], batch size: 43, lr: 3.80e-03, grad_scale: 0.000244140625 2023-04-26 10:17:56,233 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=11.05 vs. limit=5.0 2023-04-26 10:18:13,739 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=2.06 vs. limit=2.0 2023-04-26 10:18:14,583 WARNING [optim.py:389] (1/7) Scaling gradients by 0.06225070729851723, model_norm_threshold=142.37583923339844 2023-04-26 10:18:14,655 INFO [optim.py:451] (1/7) Parameter Dominanting tot_sumsq module.encoder.encoder_embed.conv.0.weight with proportion 0.68, where dominant_sumsq=(grad_sumsq*orig_rms_sq)=3.535e+06, grad_sumsq = 8.167e+06, orig_rms_sq=4.328e-01 2023-04-26 10:18:21,402 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4994, 1.5350, 1.4823, 0.2914, 1.5281, 1.8196, 1.1424, 1.1309], device='cuda:1'), covar=tensor([0.0456, 0.0698, 0.0524, 0.0699, 0.0442, 0.0540, 0.0636, 0.0964], device='cuda:1'), in_proj_covar=tensor([0.0396, 0.0401, 0.0387, 0.0358, 0.0420, 0.0446, 0.0385, 0.0425], device='cuda:1'), out_proj_covar=tensor([8.7257e-05, 8.6448e-05, 8.3473e-05, 7.5036e-05, 9.0264e-05, 9.7995e-05, 8.4506e-05, 9.2053e-05], device='cuda:1') 2023-04-26 10:18:21,437 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=15.26 vs. limit=5.0 2023-04-26 10:18:22,849 WARNING [finetune.py:966] (1/7) Grad scale is small: 0.000244140625 2023-04-26 10:18:22,849 INFO [finetune.py:976] (1/7) Epoch 1, batch 500, loss[loss=1.147, simple_loss=0.8847, pruned_loss=1.146, over 4851.00 frames. ], tot_loss[loss=1.421, simple_loss=1.189, pruned_loss=1.307, over 880509.27 frames. ], batch size: 49, lr: 4.00e-03, grad_scale: 0.00048828125 2023-04-26 10:18:31,737 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.709e+01 2.273e+01 3.115e+01 1.071e+02 6.675e+03, threshold=6.230e+01, percent-clipped=18.0 2023-04-26 10:18:33,454 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=4.57 vs. limit=2.0 2023-04-26 10:18:47,429 WARNING [optim.py:389] (1/7) Scaling gradients by 0.017591100186109543, model_norm_threshold=62.30100631713867 2023-04-26 10:18:47,507 INFO [optim.py:451] (1/7) Parameter Dominanting tot_sumsq module.encoder.encoder_embed.conv.0.weight with proportion 0.35, where dominant_sumsq=(grad_sumsq*orig_rms_sq)=4.348e+06, grad_sumsq = 1.005e+07, orig_rms_sq=4.327e-01 2023-04-26 10:18:56,301 WARNING [optim.py:389] (1/7) Scaling gradients by 0.005508477333933115, model_norm_threshold=62.30100631713867 2023-04-26 10:18:56,373 INFO [optim.py:451] (1/7) Parameter Dominanting tot_sumsq module.encoder.encoder_embed.conv.0.weight with proportion 0.80, where dominant_sumsq=(grad_sumsq*orig_rms_sq)=1.024e+08, grad_sumsq = 2.367e+08, orig_rms_sq=4.327e-01 2023-04-26 10:18:58,474 INFO [finetune.py:976] (1/7) Epoch 1, batch 550, loss[loss=1.154, simple_loss=0.86, pruned_loss=1.187, over 4871.00 frames. ], tot_loss[loss=1.341, simple_loss=1.101, pruned_loss=1.258, over 898230.18 frames. ], batch size: 34, lr: 4.00e-03, grad_scale: 0.00048828125 2023-04-26 10:19:04,283 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=562.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 10:19:13,090 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=568.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 10:19:35,142 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=590.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 10:19:45,896 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=600.0, num_to_drop=2, layers_to_drop={0, 2} 2023-04-26 10:19:46,333 WARNING [finetune.py:966] (1/7) Grad scale is small: 0.00048828125 2023-04-26 10:19:46,333 INFO [finetune.py:976] (1/7) Epoch 1, batch 600, loss[loss=1.26, simple_loss=0.9363, pruned_loss=1.253, over 4892.00 frames. ], tot_loss[loss=1.283, simple_loss=1.034, pruned_loss=1.223, over 910732.04 frames. ], batch size: 43, lr: 4.00e-03, grad_scale: 0.0009765625 2023-04-26 10:20:06,709 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.706e+01 2.227e+01 2.629e+01 6.305e+01 1.131e+04, threshold=5.258e+01, percent-clipped=26.0 2023-04-26 10:20:09,409 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=623.0, num_to_drop=2, layers_to_drop={0, 2} 2023-04-26 10:20:18,141 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=629.0, num_to_drop=2, layers_to_drop={1, 2} 2023-04-26 10:20:29,916 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=648.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 10:20:31,452 INFO [finetune.py:976] (1/7) Epoch 1, batch 650, loss[loss=0.9591, simple_loss=0.6985, pruned_loss=0.9526, over 4832.00 frames. ], tot_loss[loss=1.248, simple_loss=0.9872, pruned_loss=1.201, over 920248.32 frames. ], batch size: 25, lr: 4.00e-03, grad_scale: 0.0009765625 2023-04-26 10:20:31,538 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=651.0, num_to_drop=2, layers_to_drop={0, 1} 2023-04-26 10:20:32,028 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=652.0, num_to_drop=2, layers_to_drop={0, 1} 2023-04-26 10:20:41,812 WARNING [optim.py:389] (1/7) Scaling gradients by 0.07653743773698807, model_norm_threshold=52.5806770324707 2023-04-26 10:20:41,885 INFO [optim.py:451] (1/7) Parameter Dominanting tot_sumsq module.encoder.encoder_embed.conv.0.weight with proportion 0.93, where dominant_sumsq=(grad_sumsq*orig_rms_sq)=4.392e+05, grad_sumsq = 1.015e+06, orig_rms_sq=4.327e-01 2023-04-26 10:20:50,905 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=17.26 vs. limit=5.0 2023-04-26 10:20:59,955 WARNING [finetune.py:966] (1/7) Grad scale is small: 0.0009765625 2023-04-26 10:20:59,955 INFO [finetune.py:976] (1/7) Epoch 1, batch 700, loss[loss=1.174, simple_loss=0.8472, pruned_loss=1.144, over 4914.00 frames. ], tot_loss[loss=1.219, simple_loss=0.9461, pruned_loss=1.179, over 926274.30 frames. ], batch size: 36, lr: 4.00e-03, grad_scale: 0.001953125 2023-04-26 10:21:08,936 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=10.43 vs. limit=5.0 2023-04-26 10:21:09,245 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.790e+01 2.165e+01 2.490e+01 3.193e+01 6.870e+02, threshold=4.979e+01, percent-clipped=6.0 2023-04-26 10:21:18,125 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=14.09 vs. limit=5.0 2023-04-26 10:21:20,045 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=739.0, num_to_drop=2, layers_to_drop={1, 2} 2023-04-26 10:21:22,090 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=743.0, num_to_drop=2, layers_to_drop={0, 2} 2023-04-26 10:21:28,230 INFO [finetune.py:976] (1/7) Epoch 1, batch 750, loss[loss=1, simple_loss=0.7178, pruned_loss=0.9526, over 4804.00 frames. ], tot_loss[loss=1.192, simple_loss=0.9108, pruned_loss=1.153, over 934284.38 frames. ], batch size: 51, lr: 4.00e-03, grad_scale: 0.001953125 2023-04-26 10:21:29,381 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=16.39 vs. limit=5.0 2023-04-26 10:21:31,462 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.85 vs. limit=2.0 2023-04-26 10:21:48,060 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=787.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 10:21:53,224 WARNING [optim.py:389] (1/7) Scaling gradients by 0.039711207151412964, model_norm_threshold=49.79251480102539 2023-04-26 10:21:53,297 INFO [optim.py:451] (1/7) Parameter Dominanting tot_sumsq module.encoder.encoder_embed.conv.0.weight with proportion 0.81, where dominant_sumsq=(grad_sumsq*orig_rms_sq)=1.271e+06, grad_sumsq = 2.939e+06, orig_rms_sq=4.325e-01 2023-04-26 10:21:55,912 WARNING [finetune.py:966] (1/7) Grad scale is small: 0.001953125 2023-04-26 10:21:55,912 INFO [finetune.py:976] (1/7) Epoch 1, batch 800, loss[loss=1.092, simple_loss=0.7759, pruned_loss=1.023, over 4879.00 frames. ], tot_loss[loss=1.164, simple_loss=0.8769, pruned_loss=1.122, over 938842.32 frames. ], batch size: 32, lr: 4.00e-03, grad_scale: 0.00390625 2023-04-26 10:22:11,121 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 2.049e+01 2.360e+01 2.808e+01 3.468e+01 1.254e+03, threshold=5.615e+01, percent-clipped=6.0 2023-04-26 10:22:14,903 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=2.77 vs. limit=2.0 2023-04-26 10:22:18,487 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=20.89 vs. limit=5.0 2023-04-26 10:22:26,859 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=847.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 10:22:28,866 INFO [finetune.py:976] (1/7) Epoch 1, batch 850, loss[loss=0.9704, simple_loss=0.7045, pruned_loss=0.8609, over 4727.00 frames. ], tot_loss[loss=1.13, simple_loss=0.8419, pruned_loss=1.079, over 943661.21 frames. ], batch size: 54, lr: 4.00e-03, grad_scale: 0.00390625 2023-04-26 10:22:41,363 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=11.59 vs. limit=5.0 2023-04-26 10:23:15,928 WARNING [finetune.py:966] (1/7) Grad scale is small: 0.00390625 2023-04-26 10:23:15,929 INFO [finetune.py:976] (1/7) Epoch 1, batch 900, loss[loss=1.048, simple_loss=0.7447, pruned_loss=0.9311, over 4828.00 frames. ], tot_loss[loss=1.094, simple_loss=0.8081, pruned_loss=1.033, over 945192.72 frames. ], batch size: 33, lr: 4.00e-03, grad_scale: 0.0078125 2023-04-26 10:23:20,201 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=908.0, num_to_drop=2, layers_to_drop={0, 1} 2023-04-26 10:23:26,895 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 2.132e+01 2.534e+01 2.883e+01 3.456e+01 6.931e+01, threshold=5.766e+01, percent-clipped=4.0 2023-04-26 10:23:26,975 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=918.0, num_to_drop=2, layers_to_drop={0, 1} 2023-04-26 10:23:30,106 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=924.0, num_to_drop=2, layers_to_drop={0, 1} 2023-04-26 10:23:31,732 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=13.06 vs. limit=5.0 2023-04-26 10:23:36,472 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2023-04-26 10:23:42,110 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=946.0, num_to_drop=2, layers_to_drop={1, 2} 2023-04-26 10:23:44,693 INFO [finetune.py:976] (1/7) Epoch 1, batch 950, loss[loss=0.9913, simple_loss=0.7096, pruned_loss=0.8531, over 4741.00 frames. ], tot_loss[loss=1.07, simple_loss=0.7845, pruned_loss=0.9953, over 948190.05 frames. ], batch size: 54, lr: 4.00e-03, grad_scale: 0.0078125 2023-04-26 10:23:45,270 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=952.0, num_to_drop=2, layers_to_drop={0, 1} 2023-04-26 10:24:30,922 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=12.03 vs. limit=5.0 2023-04-26 10:24:41,986 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=1000.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 10:24:42,450 WARNING [finetune.py:966] (1/7) Grad scale is small: 0.0078125 2023-04-26 10:24:42,451 INFO [finetune.py:976] (1/7) Epoch 1, batch 1000, loss[loss=1.061, simple_loss=0.7444, pruned_loss=0.913, over 4905.00 frames. ], tot_loss[loss=1.067, simple_loss=0.7766, pruned_loss=0.9744, over 951134.73 frames. ], batch size: 43, lr: 4.00e-03, grad_scale: 0.015625 2023-04-26 10:25:00,173 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 2.486e+01 3.090e+01 3.594e+01 4.276e+01 7.170e+01, threshold=7.188e+01, percent-clipped=7.0 2023-04-26 10:25:14,570 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=1043.0, num_to_drop=2, layers_to_drop={0, 2} 2023-04-26 10:25:18,746 INFO [finetune.py:976] (1/7) Epoch 1, batch 1050, loss[loss=1.136, simple_loss=0.8119, pruned_loss=0.9379, over 4897.00 frames. ], tot_loss[loss=1.068, simple_loss=0.7729, pruned_loss=0.958, over 953012.42 frames. ], batch size: 35, lr: 4.00e-03, grad_scale: 0.015625 2023-04-26 10:25:42,484 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=1091.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 10:25:44,667 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=2.79 vs. limit=2.0 2023-04-26 10:25:48,291 INFO [finetune.py:976] (1/7) Epoch 1, batch 1100, loss[loss=0.9976, simple_loss=0.7187, pruned_loss=0.8006, over 4117.00 frames. ], tot_loss[loss=1.056, simple_loss=0.7608, pruned_loss=0.9319, over 953372.90 frames. ], batch size: 65, lr: 4.00e-03, grad_scale: 0.03125 2023-04-26 10:25:57,275 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 2.855e+01 3.563e+01 4.845e+01 6.131e+01 1.271e+02, threshold=9.690e+01, percent-clipped=11.0 2023-04-26 10:25:57,422 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=114.45 vs. limit=5.0 2023-04-26 10:25:58,922 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6572, 0.9619, 1.4560, 2.0914, 1.4739, 1.8200, 1.8177, 1.8473], device='cuda:1'), covar=tensor([0.1464, 0.1443, 0.1572, 0.0986, 0.1657, 0.1503, 0.1485, 0.2606], device='cuda:1'), in_proj_covar=tensor([0.0495, 0.0562, 0.0633, 0.0613, 0.0544, 0.0598, 0.0629, 0.0619], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002], device='cuda:1') 2023-04-26 10:26:07,214 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6326, 1.4278, 0.7988, 1.2812, 1.4842, 1.5071, 1.2929, 1.3751], device='cuda:1'), covar=tensor([0.0201, 0.0142, 0.0123, 0.0204, 0.0102, 0.0189, 0.0226, 0.0271], device='cuda:1'), in_proj_covar=tensor([0.0035, 0.0028, 0.0026, 0.0035, 0.0023, 0.0035, 0.0035, 0.0037], device='cuda:1'), out_proj_covar=tensor([0.0053, 0.0045, 0.0039, 0.0056, 0.0039, 0.0053, 0.0054, 0.0057], device='cuda:1') 2023-04-26 10:26:10,381 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.2832, 1.8392, 0.9845, 0.7370, 1.0780, 0.7900, 2.7278, 0.8262], device='cuda:1'), covar=tensor([0.0346, 0.0496, 0.0285, 0.0440, 0.0323, 0.0466, 0.0287, 0.0513], device='cuda:1'), in_proj_covar=tensor([0.0058, 0.0076, 0.0054, 0.0052, 0.0057, 0.0057, 0.0094, 0.0054], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0011, 0.0008, 0.0008, 0.0008, 0.0008, 0.0012, 0.0008], device='cuda:1') 2023-04-26 10:26:17,721 INFO [finetune.py:976] (1/7) Epoch 1, batch 1150, loss[loss=1.108, simple_loss=0.7879, pruned_loss=0.8842, over 4823.00 frames. ], tot_loss[loss=1.049, simple_loss=0.7521, pruned_loss=0.91, over 955216.45 frames. ], batch size: 39, lr: 4.00e-03, grad_scale: 0.03125 2023-04-26 10:26:18,917 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=1153.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 10:26:24,756 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=21.94 vs. limit=5.0 2023-04-26 10:26:31,611 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=14.43 vs. limit=5.0 2023-04-26 10:26:43,725 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=2.00 vs. limit=2.0 2023-04-26 10:26:44,844 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=9.92 vs. limit=5.0 2023-04-26 10:26:46,648 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.98 vs. limit=2.0 2023-04-26 10:26:46,978 INFO [finetune.py:976] (1/7) Epoch 1, batch 1200, loss[loss=0.9676, simple_loss=0.6816, pruned_loss=0.7661, over 4825.00 frames. ], tot_loss[loss=1.037, simple_loss=0.7399, pruned_loss=0.8849, over 954083.75 frames. ], batch size: 39, lr: 4.00e-03, grad_scale: 0.0625 2023-04-26 10:26:48,592 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=1203.0, num_to_drop=2, layers_to_drop={1, 2} 2023-04-26 10:26:54,330 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=1214.0, num_to_drop=2, layers_to_drop={0, 2} 2023-04-26 10:26:56,332 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 2.920e+01 3.405e+01 3.864e+01 4.670e+01 1.171e+02, threshold=7.728e+01, percent-clipped=2.0 2023-04-26 10:26:56,425 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=1218.0, num_to_drop=2, layers_to_drop={1, 2} 2023-04-26 10:26:59,566 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=1224.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 10:27:19,079 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=1246.0, num_to_drop=1, layers_to_drop={2} 2023-04-26 10:27:21,697 INFO [finetune.py:976] (1/7) Epoch 1, batch 1250, loss[loss=0.9648, simple_loss=0.6662, pruned_loss=0.7646, over 4753.00 frames. ], tot_loss[loss=1.024, simple_loss=0.7276, pruned_loss=0.8601, over 955859.73 frames. ], batch size: 23, lr: 4.00e-03, grad_scale: 0.0625 2023-04-26 10:27:40,176 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=1266.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 10:27:40,731 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.2527, 0.9109, 4.3916, 4.0141, 4.2921, 4.2487, 4.2633, 3.9428], device='cuda:1'), covar=tensor([0.8872, 0.9030, 0.1786, 0.2712, 0.1458, 0.1997, 0.1708, 0.2594], device='cuda:1'), in_proj_covar=tensor([0.0346, 0.0316, 0.0455, 0.0464, 0.0372, 0.0433, 0.0354, 0.0401], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0002], device='cuda:1') 2023-04-26 10:27:49,523 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=1272.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 10:27:49,660 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=5.88 vs. limit=5.0 2023-04-26 10:28:15,538 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=1294.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 10:28:24,919 INFO [finetune.py:976] (1/7) Epoch 1, batch 1300, loss[loss=0.8943, simple_loss=0.6186, pruned_loss=0.6957, over 4866.00 frames. ], tot_loss[loss=1.009, simple_loss=0.7135, pruned_loss=0.8346, over 956568.44 frames. ], batch size: 31, lr: 4.00e-03, grad_scale: 0.125 2023-04-26 10:28:25,005 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5188, 1.6051, 0.9510, 1.3207, 2.1413, 1.4220, 1.4167, 1.5023], device='cuda:1'), covar=tensor([0.0324, 0.0249, 0.0220, 0.0421, 0.0119, 0.0361, 0.0316, 0.0416], device='cuda:1'), in_proj_covar=tensor([0.0036, 0.0029, 0.0027, 0.0036, 0.0023, 0.0036, 0.0036, 0.0038], device='cuda:1'), out_proj_covar=tensor([0.0055, 0.0047, 0.0040, 0.0058, 0.0040, 0.0055, 0.0056, 0.0059], device='cuda:1') 2023-04-26 10:28:40,246 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 3.024e+01 3.776e+01 4.391e+01 5.836e+01 1.199e+02, threshold=8.782e+01, percent-clipped=8.0 2023-04-26 10:29:00,694 INFO [finetune.py:976] (1/7) Epoch 1, batch 1350, loss[loss=1.014, simple_loss=0.7144, pruned_loss=0.7643, over 4096.00 frames. ], tot_loss[loss=1.008, simple_loss=0.7093, pruned_loss=0.8217, over 955526.42 frames. ], batch size: 65, lr: 4.00e-03, grad_scale: 0.125 2023-04-26 10:29:18,947 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=5.97 vs. limit=5.0 2023-04-26 10:29:19,483 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.86 vs. limit=2.0 2023-04-26 10:29:52,994 INFO [finetune.py:976] (1/7) Epoch 1, batch 1400, loss[loss=1.117, simple_loss=0.7599, pruned_loss=0.8529, over 4796.00 frames. ], tot_loss[loss=1.016, simple_loss=0.7109, pruned_loss=0.8159, over 956628.50 frames. ], batch size: 29, lr: 4.00e-03, grad_scale: 0.25 2023-04-26 10:30:14,312 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 3.571e+01 4.693e+01 6.414e+01 7.905e+01 1.778e+02, threshold=1.283e+02, percent-clipped=17.0 2023-04-26 10:30:15,078 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=6.33 vs. limit=5.0 2023-04-26 10:30:34,168 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=1434.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 10:30:55,950 INFO [finetune.py:976] (1/7) Epoch 1, batch 1450, loss[loss=1.107, simple_loss=0.7681, pruned_loss=0.8195, over 4092.00 frames. ], tot_loss[loss=1.016, simple_loss=0.7075, pruned_loss=0.8051, over 954153.55 frames. ], batch size: 65, lr: 4.00e-03, grad_scale: 0.25 2023-04-26 10:31:42,618 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=1495.0, num_to_drop=2, layers_to_drop={0, 2} 2023-04-26 10:31:53,102 INFO [finetune.py:976] (1/7) Epoch 1, batch 1500, loss[loss=1.011, simple_loss=0.6946, pruned_loss=0.7443, over 4742.00 frames. ], tot_loss[loss=1.013, simple_loss=0.7032, pruned_loss=0.791, over 953678.09 frames. ], batch size: 54, lr: 4.00e-03, grad_scale: 0.5 2023-04-26 10:31:59,747 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=1503.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 10:32:03,095 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=1509.0, num_to_drop=2, layers_to_drop={2, 3} 2023-04-26 10:32:13,078 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 3.937e+01 5.249e+01 6.518e+01 8.001e+01 1.317e+02, threshold=1.304e+02, percent-clipped=3.0 2023-04-26 10:32:15,388 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.0041, 1.3189, 1.1790, 1.6158, 1.3965, 1.3013, 1.2605, 2.4849], device='cuda:1'), covar=tensor([0.0674, 0.0963, 0.1006, 0.1646, 0.0840, 0.0779, 0.0897, 0.0344], device='cuda:1'), in_proj_covar=tensor([0.0041, 0.0045, 0.0045, 0.0051, 0.0046, 0.0043, 0.0045, 0.0071], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0014, 0.0014, 0.0016, 0.0015, 0.0013, 0.0015, 0.0019], device='cuda:1') 2023-04-26 10:32:48,628 INFO [finetune.py:976] (1/7) Epoch 1, batch 1550, loss[loss=0.982, simple_loss=0.6833, pruned_loss=0.7063, over 4813.00 frames. ], tot_loss[loss=1.007, simple_loss=0.6995, pruned_loss=0.7748, over 953976.77 frames. ], batch size: 45, lr: 4.00e-03, grad_scale: 0.5 2023-04-26 10:32:48,699 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=1551.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 10:33:08,371 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=1566.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 10:33:50,408 INFO [finetune.py:976] (1/7) Epoch 1, batch 1600, loss[loss=0.9429, simple_loss=0.677, pruned_loss=0.6543, over 4744.00 frames. ], tot_loss[loss=0.9877, simple_loss=0.6881, pruned_loss=0.7471, over 952921.38 frames. ], batch size: 27, lr: 4.00e-03, grad_scale: 1.0 2023-04-26 10:34:04,225 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=1611.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 10:34:13,528 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 5.628e+01 9.315e+01 1.324e+02 1.713e+02 3.757e+02, threshold=2.648e+02, percent-clipped=49.0 2023-04-26 10:34:24,064 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=1627.0, num_to_drop=2, layers_to_drop={0, 2} 2023-04-26 10:34:48,569 INFO [finetune.py:976] (1/7) Epoch 1, batch 1650, loss[loss=0.8684, simple_loss=0.6324, pruned_loss=0.5896, over 4866.00 frames. ], tot_loss[loss=0.9626, simple_loss=0.6744, pruned_loss=0.7153, over 953470.02 frames. ], batch size: 34, lr: 4.00e-03, grad_scale: 1.0 2023-04-26 10:35:08,285 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=1672.0, num_to_drop=2, layers_to_drop={1, 3} 2023-04-26 10:35:11,088 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.2540, 0.8766, 0.9178, 1.3367, 0.8787, 0.9129, 0.8821, 0.8685], device='cuda:1'), covar=tensor([14.4704, 16.1208, 9.1040, 18.1514, 29.3317, 15.2588, 16.4841, 17.4821], device='cuda:1'), in_proj_covar=tensor([0.0302, 0.0316, 0.0252, 0.0408, 0.0275, 0.0263, 0.0312, 0.0258], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-26 10:35:36,516 INFO [finetune.py:976] (1/7) Epoch 1, batch 1700, loss[loss=0.8099, simple_loss=0.6065, pruned_loss=0.5331, over 4745.00 frames. ], tot_loss[loss=0.9263, simple_loss=0.6552, pruned_loss=0.6754, over 954039.15 frames. ], batch size: 59, lr: 4.00e-03, grad_scale: 1.0 2023-04-26 10:35:59,304 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.556e+01 1.723e+02 2.061e+02 2.467e+02 4.417e+02, threshold=4.123e+02, percent-clipped=15.0 2023-04-26 10:36:12,241 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.8154, 2.6060, 2.2695, 3.2033, 2.6797, 2.7670, 1.2022, 2.6551], device='cuda:1'), covar=tensor([0.2132, 0.1688, 0.2772, 0.1695, 0.2675, 0.1999, 0.6165, 0.2354], device='cuda:1'), in_proj_covar=tensor([0.0264, 0.0239, 0.0292, 0.0332, 0.0330, 0.0273, 0.0290, 0.0287], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-26 10:36:24,060 INFO [finetune.py:976] (1/7) Epoch 1, batch 1750, loss[loss=0.8854, simple_loss=0.6701, pruned_loss=0.5727, over 4811.00 frames. ], tot_loss[loss=0.9037, simple_loss=0.6472, pruned_loss=0.645, over 955638.77 frames. ], batch size: 41, lr: 4.00e-03, grad_scale: 1.0 2023-04-26 10:36:49,399 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=12.96 vs. limit=5.0 2023-04-26 10:36:56,468 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=6.62 vs. limit=5.0 2023-04-26 10:36:58,582 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=1790.0, num_to_drop=1, layers_to_drop={2} 2023-04-26 10:37:11,178 INFO [finetune.py:976] (1/7) Epoch 1, batch 1800, loss[loss=0.7034, simple_loss=0.5452, pruned_loss=0.4434, over 4778.00 frames. ], tot_loss[loss=0.8813, simple_loss=0.641, pruned_loss=0.6151, over 954798.51 frames. ], batch size: 26, lr: 4.00e-03, grad_scale: 1.0 2023-04-26 10:37:21,913 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=1809.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 10:37:28,278 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.499e+02 2.323e+02 2.803e+02 3.399e+02 5.478e+02, threshold=5.607e+02, percent-clipped=9.0 2023-04-26 10:37:28,538 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-04-26 10:37:40,482 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=1838.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 10:37:47,698 INFO [finetune.py:976] (1/7) Epoch 1, batch 1850, loss[loss=0.5887, simple_loss=0.4768, pruned_loss=0.3565, over 4715.00 frames. ], tot_loss[loss=0.8452, simple_loss=0.6251, pruned_loss=0.5771, over 954917.18 frames. ], batch size: 23, lr: 4.00e-03, grad_scale: 1.0 2023-04-26 10:37:51,631 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=1857.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 10:37:51,660 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=1857.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 10:37:59,990 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=1870.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 10:38:15,485 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([4.5259, 4.4076, 2.9446, 5.1031, 4.3930, 4.3931, 2.1667, 4.3512], device='cuda:1'), covar=tensor([0.1383, 0.1024, 0.3663, 0.0908, 0.2432, 0.1552, 0.5266, 0.1934], device='cuda:1'), in_proj_covar=tensor([0.0264, 0.0239, 0.0292, 0.0332, 0.0332, 0.0273, 0.0289, 0.0288], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-26 10:38:17,744 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=1899.0, num_to_drop=2, layers_to_drop={2, 3} 2023-04-26 10:38:18,739 INFO [finetune.py:976] (1/7) Epoch 1, batch 1900, loss[loss=0.6146, simple_loss=0.4921, pruned_loss=0.373, over 4143.00 frames. ], tot_loss[loss=0.8086, simple_loss=0.6085, pruned_loss=0.5403, over 953635.38 frames. ], batch size: 18, lr: 4.00e-03, grad_scale: 1.0 2023-04-26 10:38:28,761 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.629e+02 2.504e+02 3.040e+02 3.671e+02 6.110e+02, threshold=6.080e+02, percent-clipped=2.0 2023-04-26 10:38:28,885 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=1918.0, num_to_drop=2, layers_to_drop={1, 2} 2023-04-26 10:38:32,055 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=1922.0, num_to_drop=1, layers_to_drop={3} 2023-04-26 10:38:39,000 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=1931.0, num_to_drop=2, layers_to_drop={0, 3} 2023-04-26 10:38:50,101 INFO [finetune.py:976] (1/7) Epoch 1, batch 1950, loss[loss=0.5556, simple_loss=0.4586, pruned_loss=0.3279, over 4798.00 frames. ], tot_loss[loss=0.7668, simple_loss=0.5867, pruned_loss=0.5021, over 951722.10 frames. ], batch size: 29, lr: 4.00e-03, grad_scale: 1.0 2023-04-26 10:38:59,734 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=1967.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 10:39:12,316 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.3798, 1.2219, 1.8845, 2.0272, 1.3077, 0.9529, 1.4574, 0.8491], device='cuda:1'), covar=tensor([0.1311, 0.1067, 0.0634, 0.0485, 0.1212, 0.2062, 0.0934, 0.1776], device='cuda:1'), in_proj_covar=tensor([0.0081, 0.0089, 0.0083, 0.0087, 0.0103, 0.0105, 0.0102, 0.0090], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-26 10:39:22,529 INFO [finetune.py:976] (1/7) Epoch 1, batch 2000, loss[loss=0.5761, simple_loss=0.4926, pruned_loss=0.3298, over 4828.00 frames. ], tot_loss[loss=0.7282, simple_loss=0.5664, pruned_loss=0.4674, over 951201.81 frames. ], batch size: 40, lr: 4.00e-03, grad_scale: 2.0 2023-04-26 10:39:39,762 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.592e+02 2.794e+02 3.334e+02 3.918e+02 6.535e+02, threshold=6.668e+02, percent-clipped=3.0 2023-04-26 10:39:47,341 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.1418, 2.0290, 1.6787, 1.6983, 1.7152, 2.0767, 1.9088, 3.6002], device='cuda:1'), covar=tensor([0.0977, 0.0809, 0.0957, 0.1803, 0.0938, 0.0601, 0.0800, 0.0129], device='cuda:1'), in_proj_covar=tensor([0.0040, 0.0043, 0.0043, 0.0049, 0.0044, 0.0042, 0.0043, 0.0068], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0013, 0.0014, 0.0016, 0.0014, 0.0013, 0.0014, 0.0018], device='cuda:1') 2023-04-26 10:40:01,844 INFO [finetune.py:976] (1/7) Epoch 1, batch 2050, loss[loss=0.4861, simple_loss=0.4308, pruned_loss=0.2707, over 4804.00 frames. ], tot_loss[loss=0.6876, simple_loss=0.545, pruned_loss=0.4325, over 953439.92 frames. ], batch size: 29, lr: 4.00e-03, grad_scale: 2.0 2023-04-26 10:40:37,619 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=2090.0, num_to_drop=2, layers_to_drop={0, 2} 2023-04-26 10:40:50,239 INFO [finetune.py:976] (1/7) Epoch 1, batch 2100, loss[loss=0.6226, simple_loss=0.5388, pruned_loss=0.3532, over 4903.00 frames. ], tot_loss[loss=0.6568, simple_loss=0.5304, pruned_loss=0.4051, over 953370.60 frames. ], batch size: 36, lr: 4.00e-03, grad_scale: 2.0 2023-04-26 10:41:11,501 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.614e+02 2.534e+02 2.914e+02 3.246e+02 6.149e+02, threshold=5.827e+02, percent-clipped=0.0 2023-04-26 10:41:31,945 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=2138.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 10:41:41,043 INFO [finetune.py:976] (1/7) Epoch 1, batch 2150, loss[loss=0.6627, simple_loss=0.5774, pruned_loss=0.374, over 4866.00 frames. ], tot_loss[loss=0.6377, simple_loss=0.5244, pruned_loss=0.386, over 953104.69 frames. ], batch size: 34, lr: 4.00e-03, grad_scale: 4.0 2023-04-26 10:42:16,512 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=2.10 vs. limit=2.0 2023-04-26 10:42:25,849 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=2194.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 10:42:30,521 INFO [finetune.py:976] (1/7) Epoch 1, batch 2200, loss[loss=0.5222, simple_loss=0.4762, pruned_loss=0.2841, over 4921.00 frames. ], tot_loss[loss=0.6143, simple_loss=0.515, pruned_loss=0.3651, over 954297.24 frames. ], batch size: 38, lr: 4.00e-03, grad_scale: 4.0 2023-04-26 10:42:37,628 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=2213.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 10:42:40,428 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.685e+02 2.565e+02 3.033e+02 3.519e+02 5.393e+02, threshold=6.065e+02, percent-clipped=0.0 2023-04-26 10:42:42,829 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=2222.0, num_to_drop=2, layers_to_drop={1, 2} 2023-04-26 10:42:45,138 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=2226.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 10:42:53,169 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6730, 2.3994, 1.8141, 2.2958, 1.6693, 1.6489, 2.4253, 1.6213], device='cuda:1'), covar=tensor([0.2384, 0.1688, 0.1647, 0.1605, 0.3054, 0.1932, 0.1625, 0.2523], device='cuda:1'), in_proj_covar=tensor([0.0277, 0.0290, 0.0213, 0.0270, 0.0286, 0.0247, 0.0248, 0.0259], device='cuda:1'), out_proj_covar=tensor([1.1307e-04, 1.1870e-04, 8.7352e-05, 1.0932e-04, 1.1890e-04, 9.9580e-05, 1.0309e-04, 1.0557e-04], device='cuda:1') 2023-04-26 10:43:02,985 INFO [finetune.py:976] (1/7) Epoch 1, batch 2250, loss[loss=0.5463, simple_loss=0.4825, pruned_loss=0.305, over 4804.00 frames. ], tot_loss[loss=0.5925, simple_loss=0.5046, pruned_loss=0.3466, over 953954.04 frames. ], batch size: 25, lr: 4.00e-03, grad_scale: 4.0 2023-04-26 10:43:12,372 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=2267.0, num_to_drop=1, layers_to_drop={2} 2023-04-26 10:43:14,135 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=2270.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 10:43:57,381 INFO [finetune.py:976] (1/7) Epoch 1, batch 2300, loss[loss=0.4601, simple_loss=0.4246, pruned_loss=0.2478, over 4754.00 frames. ], tot_loss[loss=0.5726, simple_loss=0.4954, pruned_loss=0.3298, over 953866.73 frames. ], batch size: 28, lr: 4.00e-03, grad_scale: 4.0 2023-04-26 10:44:17,314 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=2315.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 10:44:19,009 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.752e+02 2.423e+02 2.781e+02 3.392e+02 7.688e+02, threshold=5.562e+02, percent-clipped=1.0 2023-04-26 10:44:43,394 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=2340.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 10:44:51,821 INFO [finetune.py:976] (1/7) Epoch 1, batch 2350, loss[loss=0.4788, simple_loss=0.4373, pruned_loss=0.2602, over 4739.00 frames. ], tot_loss[loss=0.5498, simple_loss=0.4827, pruned_loss=0.3124, over 954496.75 frames. ], batch size: 54, lr: 4.00e-03, grad_scale: 4.0 2023-04-26 10:44:56,673 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=5.02 vs. limit=5.0 2023-04-26 10:45:34,702 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-04-26 10:45:40,788 INFO [finetune.py:976] (1/7) Epoch 1, batch 2400, loss[loss=0.4638, simple_loss=0.4295, pruned_loss=0.2491, over 4911.00 frames. ], tot_loss[loss=0.5271, simple_loss=0.4691, pruned_loss=0.2956, over 956363.56 frames. ], batch size: 43, lr: 4.00e-03, grad_scale: 8.0 2023-04-26 10:45:40,915 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=2401.0, num_to_drop=2, layers_to_drop={0, 3} 2023-04-26 10:45:51,695 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.659e+02 2.375e+02 2.859e+02 3.332e+02 6.118e+02, threshold=5.719e+02, percent-clipped=1.0 2023-04-26 10:46:11,802 INFO [finetune.py:976] (1/7) Epoch 1, batch 2450, loss[loss=0.5537, simple_loss=0.493, pruned_loss=0.3072, over 4757.00 frames. ], tot_loss[loss=0.5065, simple_loss=0.4567, pruned_loss=0.2805, over 957318.58 frames. ], batch size: 59, lr: 4.00e-03, grad_scale: 8.0 2023-04-26 10:46:39,130 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=2494.0, num_to_drop=1, layers_to_drop={2} 2023-04-26 10:46:44,088 INFO [finetune.py:976] (1/7) Epoch 1, batch 2500, loss[loss=0.422, simple_loss=0.3905, pruned_loss=0.2268, over 4100.00 frames. ], tot_loss[loss=0.4969, simple_loss=0.4523, pruned_loss=0.2726, over 953523.73 frames. ], batch size: 18, lr: 4.00e-03, grad_scale: 8.0 2023-04-26 10:46:52,650 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=2513.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 10:46:56,028 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.542e+02 2.485e+02 2.798e+02 3.305e+02 6.030e+02, threshold=5.597e+02, percent-clipped=1.0 2023-04-26 10:47:00,940 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=2526.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 10:47:10,434 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=2542.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 10:47:16,166 INFO [finetune.py:976] (1/7) Epoch 1, batch 2550, loss[loss=0.4969, simple_loss=0.4791, pruned_loss=0.2573, over 4826.00 frames. ], tot_loss[loss=0.4895, simple_loss=0.4505, pruned_loss=0.2657, over 953753.90 frames. ], batch size: 47, lr: 4.00e-03, grad_scale: 8.0 2023-04-26 10:47:34,871 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=2561.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 10:47:50,044 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=2574.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 10:48:01,471 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-04-26 10:48:06,451 INFO [finetune.py:976] (1/7) Epoch 1, batch 2600, loss[loss=0.444, simple_loss=0.4244, pruned_loss=0.2317, over 4064.00 frames. ], tot_loss[loss=0.4796, simple_loss=0.4462, pruned_loss=0.2576, over 953324.53 frames. ], batch size: 65, lr: 4.00e-03, grad_scale: 8.0 2023-04-26 10:48:07,832 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=2.16 vs. limit=2.0 2023-04-26 10:48:09,465 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5190, 1.4183, 1.6675, 1.7013, 1.9433, 1.3283, 0.9687, 1.5318], device='cuda:1'), covar=tensor([0.1446, 0.1516, 0.1056, 0.1215, 0.0833, 0.1547, 0.1664, 0.1181], device='cuda:1'), in_proj_covar=tensor([0.0204, 0.0211, 0.0191, 0.0173, 0.0172, 0.0189, 0.0167, 0.0184], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 10:48:18,465 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.395e+02 2.465e+02 2.854e+02 3.439e+02 6.010e+02, threshold=5.707e+02, percent-clipped=1.0 2023-04-26 10:48:38,254 INFO [finetune.py:976] (1/7) Epoch 1, batch 2650, loss[loss=0.3733, simple_loss=0.3699, pruned_loss=0.1884, over 4731.00 frames. ], tot_loss[loss=0.4706, simple_loss=0.4427, pruned_loss=0.2501, over 955484.84 frames. ], batch size: 23, lr: 4.00e-03, grad_scale: 8.0 2023-04-26 10:48:48,261 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.48 vs. limit=2.0 2023-04-26 10:49:01,786 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=2685.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 10:49:06,006 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.2697, 1.7348, 1.0821, 0.9405, 0.9999, 0.9640, 1.0356, 0.9014], device='cuda:1'), covar=tensor([0.3966, 0.3386, 0.5552, 0.6116, 0.5516, 0.4939, 0.4065, 0.5570], device='cuda:1'), in_proj_covar=tensor([0.0208, 0.0232, 0.0212, 0.0228, 0.0246, 0.0206, 0.0202, 0.0220], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-26 10:49:14,009 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=2696.0, num_to_drop=1, layers_to_drop={3} 2023-04-26 10:49:16,819 INFO [finetune.py:976] (1/7) Epoch 1, batch 2700, loss[loss=0.4531, simple_loss=0.4399, pruned_loss=0.2332, over 4897.00 frames. ], tot_loss[loss=0.4611, simple_loss=0.4379, pruned_loss=0.2428, over 956896.96 frames. ], batch size: 36, lr: 4.00e-03, grad_scale: 8.0 2023-04-26 10:49:39,495 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.764e+02 2.585e+02 2.984e+02 3.485e+02 4.746e+02, threshold=5.968e+02, percent-clipped=0.0 2023-04-26 10:50:02,443 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.98 vs. limit=5.0 2023-04-26 10:50:04,159 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=2746.0, num_to_drop=2, layers_to_drop={1, 3} 2023-04-26 10:50:13,029 INFO [finetune.py:976] (1/7) Epoch 1, batch 2750, loss[loss=0.4133, simple_loss=0.4068, pruned_loss=0.2099, over 4743.00 frames. ], tot_loss[loss=0.4477, simple_loss=0.4286, pruned_loss=0.2339, over 954938.84 frames. ], batch size: 26, lr: 4.00e-03, grad_scale: 8.0 2023-04-26 10:51:16,383 INFO [finetune.py:976] (1/7) Epoch 1, batch 2800, loss[loss=0.3363, simple_loss=0.3523, pruned_loss=0.1602, over 4880.00 frames. ], tot_loss[loss=0.4349, simple_loss=0.4197, pruned_loss=0.2255, over 956169.47 frames. ], batch size: 31, lr: 4.00e-03, grad_scale: 8.0 2023-04-26 10:51:38,598 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.584e+02 2.321e+02 2.765e+02 3.400e+02 8.366e+02, threshold=5.531e+02, percent-clipped=2.0 2023-04-26 10:51:49,422 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=2824.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 10:52:01,826 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.61 vs. limit=2.0 2023-04-26 10:52:23,172 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([0.7601, 1.7323, 1.6950, 1.3946, 1.8152, 1.4154, 2.3870, 1.4278], device='cuda:1'), covar=tensor([0.3184, 0.1036, 0.2949, 0.1877, 0.1250, 0.1782, 0.0651, 0.2860], device='cuda:1'), in_proj_covar=tensor([0.0304, 0.0314, 0.0382, 0.0326, 0.0363, 0.0339, 0.0351, 0.0366], device='cuda:1'), out_proj_covar=tensor([9.4219e-05, 9.7030e-05, 1.1819e-04, 1.0173e-04, 1.1130e-04, 1.0368e-04, 1.0615e-04, 1.1362e-04], device='cuda:1') 2023-04-26 10:52:24,247 INFO [finetune.py:976] (1/7) Epoch 1, batch 2850, loss[loss=0.4449, simple_loss=0.4105, pruned_loss=0.2397, over 4213.00 frames. ], tot_loss[loss=0.4264, simple_loss=0.4143, pruned_loss=0.2196, over 954899.69 frames. ], batch size: 65, lr: 4.00e-03, grad_scale: 8.0 2023-04-26 10:53:07,939 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=2885.0, num_to_drop=2, layers_to_drop={1, 2} 2023-04-26 10:53:29,242 INFO [finetune.py:976] (1/7) Epoch 1, batch 2900, loss[loss=0.4283, simple_loss=0.4339, pruned_loss=0.2113, over 4907.00 frames. ], tot_loss[loss=0.4263, simple_loss=0.4151, pruned_loss=0.219, over 951368.71 frames. ], batch size: 36, lr: 4.00e-03, grad_scale: 8.0 2023-04-26 10:53:46,011 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.613e+02 2.467e+02 2.930e+02 3.474e+02 6.951e+02, threshold=5.860e+02, percent-clipped=1.0 2023-04-26 10:53:53,065 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=7.91 vs. limit=5.0 2023-04-26 10:53:59,778 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=2937.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 10:54:05,654 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=5.45 vs. limit=5.0 2023-04-26 10:54:08,443 INFO [finetune.py:976] (1/7) Epoch 1, batch 2950, loss[loss=0.3584, simple_loss=0.3735, pruned_loss=0.1717, over 4270.00 frames. ], tot_loss[loss=0.4254, simple_loss=0.4167, pruned_loss=0.2173, over 950889.81 frames. ], batch size: 65, lr: 4.00e-03, grad_scale: 8.0 2023-04-26 10:54:28,962 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5024, 1.5780, 1.7590, 1.7490, 1.8051, 1.3197, 0.9405, 1.6460], device='cuda:1'), covar=tensor([0.1297, 0.1476, 0.0972, 0.1112, 0.0945, 0.1444, 0.1671, 0.0846], device='cuda:1'), in_proj_covar=tensor([0.0203, 0.0209, 0.0189, 0.0173, 0.0172, 0.0188, 0.0167, 0.0184], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 10:54:37,933 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=2996.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 10:54:39,143 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=2998.0, num_to_drop=2, layers_to_drop={0, 3} 2023-04-26 10:54:39,657 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([3.5796, 3.5510, 2.9466, 4.0742, 3.4764, 3.4755, 1.9138, 3.5237], device='cuda:1'), covar=tensor([0.1677, 0.1073, 0.3406, 0.1100, 0.2149, 0.1817, 0.4375, 0.1832], device='cuda:1'), in_proj_covar=tensor([0.0261, 0.0232, 0.0283, 0.0326, 0.0320, 0.0271, 0.0283, 0.0283], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-26 10:54:40,894 INFO [finetune.py:976] (1/7) Epoch 1, batch 3000, loss[loss=0.3257, simple_loss=0.3254, pruned_loss=0.163, over 4124.00 frames. ], tot_loss[loss=0.4216, simple_loss=0.4151, pruned_loss=0.2142, over 950111.07 frames. ], batch size: 18, lr: 4.00e-03, grad_scale: 8.0 2023-04-26 10:54:40,894 INFO [finetune.py:1001] (1/7) Computing validation loss 2023-04-26 10:54:46,140 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.0648, 0.7545, 0.8352, 0.9024, 0.7783, 0.7404, 0.5190, 0.6025], device='cuda:1'), covar=tensor([ 7.3023, 9.4050, 4.1677, 13.5464, 11.0526, 8.2225, 11.0452, 11.0487], device='cuda:1'), in_proj_covar=tensor([0.0267, 0.0286, 0.0226, 0.0354, 0.0247, 0.0236, 0.0281, 0.0233], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-26 10:54:46,330 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.1972, 2.3037, 1.2202, 1.6108, 1.8078, 1.5165, 2.9188, 1.7498], device='cuda:1'), covar=tensor([0.0582, 0.0739, 0.0918, 0.0946, 0.0483, 0.0778, 0.0196, 0.0533], device='cuda:1'), in_proj_covar=tensor([0.0057, 0.0072, 0.0052, 0.0049, 0.0054, 0.0055, 0.0088, 0.0053], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0011, 0.0007], device='cuda:1') 2023-04-26 10:54:51,382 INFO [finetune.py:1010] (1/7) Epoch 1, validation: loss=0.4217, simple_loss=0.4614, pruned_loss=0.191, over 2265189.00 frames. 2023-04-26 10:54:51,382 INFO [finetune.py:1011] (1/7) Maximum memory allocated so far is 5441MB 2023-04-26 10:55:01,571 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.562e+02 2.517e+02 2.904e+02 3.818e+02 1.122e+03, threshold=5.808e+02, percent-clipped=2.0 2023-04-26 10:55:02,263 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=3019.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 10:55:04,631 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8299, 2.2544, 1.7838, 2.3521, 1.8894, 1.6544, 2.0615, 1.4929], device='cuda:1'), covar=tensor([0.1989, 0.1834, 0.1454, 0.1376, 0.2879, 0.1780, 0.1722, 0.2864], device='cuda:1'), in_proj_covar=tensor([0.0292, 0.0307, 0.0227, 0.0286, 0.0300, 0.0262, 0.0262, 0.0276], device='cuda:1'), out_proj_covar=tensor([1.1973e-04, 1.2583e-04, 9.3045e-05, 1.1594e-04, 1.2466e-04, 1.0599e-04, 1.0875e-04, 1.1266e-04], device='cuda:1') 2023-04-26 10:55:16,252 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=3041.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 10:55:18,040 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=3044.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 10:55:23,631 INFO [finetune.py:976] (1/7) Epoch 1, batch 3050, loss[loss=0.376, simple_loss=0.3963, pruned_loss=0.1778, over 4795.00 frames. ], tot_loss[loss=0.4177, simple_loss=0.4138, pruned_loss=0.2109, over 950420.23 frames. ], batch size: 51, lr: 4.00e-03, grad_scale: 8.0 2023-04-26 10:55:29,090 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-04-26 10:55:42,238 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=3080.0, num_to_drop=2, layers_to_drop={1, 2} 2023-04-26 10:55:56,457 INFO [finetune.py:976] (1/7) Epoch 1, batch 3100, loss[loss=0.4561, simple_loss=0.4277, pruned_loss=0.2422, over 4775.00 frames. ], tot_loss[loss=0.4088, simple_loss=0.4075, pruned_loss=0.2052, over 951486.89 frames. ], batch size: 29, lr: 4.00e-03, grad_scale: 8.0 2023-04-26 10:56:12,962 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.494e+02 2.339e+02 2.765e+02 3.275e+02 7.045e+02, threshold=5.531e+02, percent-clipped=2.0 2023-04-26 10:56:25,676 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5116, 1.3198, 2.0063, 2.0935, 1.3257, 1.1439, 1.5219, 1.1201], device='cuda:1'), covar=tensor([0.1698, 0.1085, 0.0544, 0.0525, 0.1546, 0.1859, 0.1095, 0.1439], device='cuda:1'), in_proj_covar=tensor([0.0079, 0.0085, 0.0079, 0.0082, 0.0099, 0.0102, 0.0099, 0.0086], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-04-26 10:56:55,557 INFO [finetune.py:976] (1/7) Epoch 1, batch 3150, loss[loss=0.3217, simple_loss=0.3286, pruned_loss=0.1574, over 3995.00 frames. ], tot_loss[loss=0.3995, simple_loss=0.4001, pruned_loss=0.1995, over 951002.75 frames. ], batch size: 17, lr: 4.00e-03, grad_scale: 8.0 2023-04-26 10:57:07,258 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=7.16 vs. limit=5.0 2023-04-26 10:57:09,949 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.0973, 1.1289, 1.2119, 1.4776, 1.4094, 1.2123, 1.2179, 1.1849], device='cuda:1'), covar=tensor([19.1105, 30.5114, 29.4275, 19.3914, 31.0561, 31.8898, 34.8417, 22.3434], device='cuda:1'), in_proj_covar=tensor([0.0473, 0.0541, 0.0615, 0.0584, 0.0512, 0.0580, 0.0591, 0.0591], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 10:57:18,528 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=6.30 vs. limit=5.0 2023-04-26 10:57:32,972 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=3180.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 10:57:46,515 INFO [finetune.py:976] (1/7) Epoch 1, batch 3200, loss[loss=0.4218, simple_loss=0.4181, pruned_loss=0.2128, over 4808.00 frames. ], tot_loss[loss=0.3907, simple_loss=0.3932, pruned_loss=0.1941, over 951481.11 frames. ], batch size: 41, lr: 4.00e-03, grad_scale: 8.0 2023-04-26 10:58:09,340 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7792, 1.3446, 1.1695, 1.2748, 1.8656, 1.6867, 1.3609, 1.2353], device='cuda:1'), covar=tensor([0.1224, 0.1500, 0.2650, 0.1385, 0.0739, 0.1072, 0.1638, 0.1767], device='cuda:1'), in_proj_covar=tensor([0.0340, 0.0352, 0.0351, 0.0313, 0.0364, 0.0386, 0.0335, 0.0372], device='cuda:1'), out_proj_covar=tensor([7.4560e-05, 7.5866e-05, 7.5761e-05, 6.5408e-05, 7.7627e-05, 8.4497e-05, 7.3042e-05, 8.0538e-05], device='cuda:1') 2023-04-26 10:58:16,459 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.736e+02 2.453e+02 2.807e+02 3.222e+02 7.994e+02, threshold=5.615e+02, percent-clipped=2.0 2023-04-26 10:58:27,261 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.0850, 1.0315, 1.2412, 1.6117, 1.3517, 1.2462, 1.2615, 1.2169], device='cuda:1'), covar=tensor([17.8866, 28.1780, 25.7030, 17.9785, 24.4871, 29.4024, 32.2844, 17.8694], device='cuda:1'), in_proj_covar=tensor([0.0472, 0.0539, 0.0614, 0.0584, 0.0510, 0.0578, 0.0589, 0.0589], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 10:58:52,918 INFO [finetune.py:976] (1/7) Epoch 1, batch 3250, loss[loss=0.4057, simple_loss=0.4064, pruned_loss=0.2025, over 4075.00 frames. ], tot_loss[loss=0.3877, simple_loss=0.3919, pruned_loss=0.1918, over 951284.09 frames. ], batch size: 65, lr: 4.00e-03, grad_scale: 8.0 2023-04-26 10:59:39,024 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.2936, 0.8217, 0.9812, 1.1938, 0.8756, 0.8929, 0.7019, 0.7148], device='cuda:1'), covar=tensor([4.7943, 6.5721, 3.1895, 9.4073, 8.4780, 5.1914, 7.9737, 8.4802], device='cuda:1'), in_proj_covar=tensor([0.0264, 0.0285, 0.0224, 0.0350, 0.0245, 0.0235, 0.0280, 0.0233], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-26 10:59:46,820 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.77 vs. limit=2.0 2023-04-26 10:59:47,317 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=3288.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 10:59:50,299 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=3293.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:00:00,686 INFO [finetune.py:976] (1/7) Epoch 1, batch 3300, loss[loss=0.3654, simple_loss=0.3802, pruned_loss=0.1753, over 4794.00 frames. ], tot_loss[loss=0.3908, simple_loss=0.396, pruned_loss=0.1928, over 953190.37 frames. ], batch size: 29, lr: 4.00e-03, grad_scale: 8.0 2023-04-26 11:00:11,743 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=5.09 vs. limit=5.0 2023-04-26 11:00:16,383 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=6.53 vs. limit=5.0 2023-04-26 11:00:19,087 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.582e+02 2.400e+02 2.649e+02 3.193e+02 6.501e+02, threshold=5.297e+02, percent-clipped=1.0 2023-04-26 11:00:33,912 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=3341.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:00:38,746 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=3349.0, num_to_drop=2, layers_to_drop={1, 3} 2023-04-26 11:00:39,793 INFO [finetune.py:976] (1/7) Epoch 1, batch 3350, loss[loss=0.4104, simple_loss=0.4237, pruned_loss=0.1985, over 4898.00 frames. ], tot_loss[loss=0.3905, simple_loss=0.3976, pruned_loss=0.1918, over 953397.11 frames. ], batch size: 46, lr: 4.00e-03, grad_scale: 8.0 2023-04-26 11:00:57,404 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=3375.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:01:05,768 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=3389.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:01:13,016 INFO [finetune.py:976] (1/7) Epoch 1, batch 3400, loss[loss=0.3743, simple_loss=0.396, pruned_loss=0.1763, over 4866.00 frames. ], tot_loss[loss=0.3893, simple_loss=0.3974, pruned_loss=0.1906, over 953890.57 frames. ], batch size: 31, lr: 4.00e-03, grad_scale: 8.0 2023-04-26 11:01:23,590 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.622e+02 2.298e+02 2.697e+02 3.198e+02 5.981e+02, threshold=5.394e+02, percent-clipped=4.0 2023-04-26 11:01:53,827 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=6.95 vs. limit=5.0 2023-04-26 11:01:57,865 INFO [finetune.py:976] (1/7) Epoch 1, batch 3450, loss[loss=0.3665, simple_loss=0.3835, pruned_loss=0.1748, over 4922.00 frames. ], tot_loss[loss=0.3855, simple_loss=0.3951, pruned_loss=0.188, over 953797.61 frames. ], batch size: 42, lr: 4.00e-03, grad_scale: 8.0 2023-04-26 11:02:36,002 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=3480.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:02:49,059 INFO [finetune.py:976] (1/7) Epoch 1, batch 3500, loss[loss=0.3686, simple_loss=0.3728, pruned_loss=0.1822, over 4801.00 frames. ], tot_loss[loss=0.3814, simple_loss=0.3913, pruned_loss=0.1858, over 954872.93 frames. ], batch size: 51, lr: 4.00e-03, grad_scale: 8.0 2023-04-26 11:02:51,666 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([0.8938, 1.9852, 1.6820, 1.8639, 2.0435, 1.5999, 2.6972, 1.3265], device='cuda:1'), covar=tensor([0.3477, 0.1296, 0.3471, 0.2007, 0.1598, 0.2277, 0.0722, 0.3588], device='cuda:1'), in_proj_covar=tensor([0.0310, 0.0320, 0.0392, 0.0333, 0.0367, 0.0343, 0.0360, 0.0373], device='cuda:1'), out_proj_covar=tensor([9.5815e-05, 9.8893e-05, 1.2139e-04, 1.0409e-04, 1.1247e-04, 1.0487e-04, 1.0894e-04, 1.1555e-04], device='cuda:1') 2023-04-26 11:03:00,464 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.505e+02 2.494e+02 2.871e+02 4.319e+02 1.287e+03, threshold=5.742e+02, percent-clipped=13.0 2023-04-26 11:03:07,522 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=3528.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:03:09,944 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([0.4983, 0.4930, 0.6170, 0.7992, 0.7906, 0.5364, 0.5741, 0.6194], device='cuda:1'), covar=tensor([18.0427, 27.2974, 27.0559, 21.0594, 22.2759, 32.0192, 29.9747, 18.8514], device='cuda:1'), in_proj_covar=tensor([0.0464, 0.0532, 0.0606, 0.0578, 0.0502, 0.0567, 0.0576, 0.0579], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 11:03:24,388 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=3546.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 11:03:27,265 INFO [finetune.py:976] (1/7) Epoch 1, batch 3550, loss[loss=0.3116, simple_loss=0.3441, pruned_loss=0.1395, over 4818.00 frames. ], tot_loss[loss=0.3756, simple_loss=0.3863, pruned_loss=0.1825, over 957106.93 frames. ], batch size: 41, lr: 4.00e-03, grad_scale: 8.0 2023-04-26 11:03:38,135 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-04-26 11:03:55,570 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-04-26 11:04:10,789 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-04-26 11:04:11,236 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=3593.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:04:16,040 INFO [finetune.py:976] (1/7) Epoch 1, batch 3600, loss[loss=0.3372, simple_loss=0.3571, pruned_loss=0.1587, over 4863.00 frames. ], tot_loss[loss=0.37, simple_loss=0.3816, pruned_loss=0.1792, over 957634.45 frames. ], batch size: 31, lr: 4.00e-03, grad_scale: 8.0 2023-04-26 11:04:18,036 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([0.4041, 0.3852, 0.4508, 0.5981, 0.6106, 0.4233, 0.4338, 0.4765], device='cuda:1'), covar=tensor([20.8054, 29.2013, 27.8891, 21.9833, 25.7780, 35.7007, 31.7989, 18.4283], device='cuda:1'), in_proj_covar=tensor([0.0457, 0.0525, 0.0597, 0.0569, 0.0493, 0.0558, 0.0566, 0.0570], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 11:04:19,839 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=3607.0, num_to_drop=2, layers_to_drop={1, 3} 2023-04-26 11:04:26,346 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.597e+02 2.647e+02 3.095e+02 3.748e+02 7.550e+02, threshold=6.190e+02, percent-clipped=4.0 2023-04-26 11:04:42,906 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=3641.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:04:44,770 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=3644.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 11:04:48,954 INFO [finetune.py:976] (1/7) Epoch 1, batch 3650, loss[loss=0.3654, simple_loss=0.381, pruned_loss=0.1749, over 4918.00 frames. ], tot_loss[loss=0.3717, simple_loss=0.3839, pruned_loss=0.1797, over 956383.97 frames. ], batch size: 38, lr: 4.00e-03, grad_scale: 8.0 2023-04-26 11:04:51,055 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.94 vs. limit=5.0 2023-04-26 11:05:04,832 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=3675.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:05:05,421 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=3676.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 11:05:23,021 INFO [finetune.py:976] (1/7) Epoch 1, batch 3700, loss[loss=0.4361, simple_loss=0.4341, pruned_loss=0.219, over 4883.00 frames. ], tot_loss[loss=0.3753, simple_loss=0.389, pruned_loss=0.1808, over 956334.77 frames. ], batch size: 32, lr: 4.00e-03, grad_scale: 8.0 2023-04-26 11:05:44,462 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.676e+02 2.590e+02 3.040e+02 3.768e+02 5.314e+02, threshold=6.080e+02, percent-clipped=0.0 2023-04-26 11:05:53,054 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=3723.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:05:55,391 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([0.7854, 0.4780, 0.5227, 0.5085, 0.5391, 0.6121, 0.4581, 0.5117], device='cuda:1'), covar=tensor([ 7.6514, 23.4597, 15.9977, 12.7028, 12.9521, 15.6235, 21.7355, 16.1200], device='cuda:1'), in_proj_covar=tensor([0.0269, 0.0385, 0.0314, 0.0311, 0.0340, 0.0350, 0.0377, 0.0342], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-26 11:06:12,569 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=3737.0, num_to_drop=2, layers_to_drop={0, 2} 2023-04-26 11:06:21,956 INFO [finetune.py:976] (1/7) Epoch 1, batch 3750, loss[loss=0.3326, simple_loss=0.363, pruned_loss=0.1511, over 4936.00 frames. ], tot_loss[loss=0.3725, simple_loss=0.3876, pruned_loss=0.1787, over 955940.25 frames. ], batch size: 29, lr: 4.00e-03, grad_scale: 8.0 2023-04-26 11:06:38,531 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-04-26 11:07:07,655 INFO [finetune.py:976] (1/7) Epoch 1, batch 3800, loss[loss=0.4107, simple_loss=0.4197, pruned_loss=0.2008, over 4817.00 frames. ], tot_loss[loss=0.3703, simple_loss=0.3872, pruned_loss=0.1768, over 956536.31 frames. ], batch size: 25, lr: 4.00e-03, grad_scale: 8.0 2023-04-26 11:07:29,603 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.740e+02 2.601e+02 3.098e+02 3.867e+02 7.221e+02, threshold=6.196e+02, percent-clipped=5.0 2023-04-26 11:08:14,573 INFO [finetune.py:976] (1/7) Epoch 1, batch 3850, loss[loss=0.3388, simple_loss=0.3658, pruned_loss=0.1559, over 4902.00 frames. ], tot_loss[loss=0.3649, simple_loss=0.3833, pruned_loss=0.1733, over 956999.56 frames. ], batch size: 43, lr: 4.00e-03, grad_scale: 8.0 2023-04-26 11:08:32,931 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8746, 2.3075, 1.7895, 2.2431, 1.7252, 1.7655, 2.0975, 1.6278], device='cuda:1'), covar=tensor([0.2152, 0.1594, 0.1513, 0.1518, 0.3357, 0.1714, 0.1895, 0.3185], device='cuda:1'), in_proj_covar=tensor([0.0298, 0.0314, 0.0232, 0.0293, 0.0301, 0.0269, 0.0267, 0.0285], device='cuda:1'), out_proj_covar=tensor([1.2195e-04, 1.2857e-04, 9.5185e-05, 1.1857e-04, 1.2479e-04, 1.0867e-04, 1.1069e-04, 1.1621e-04], device='cuda:1') 2023-04-26 11:09:10,880 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4846, 0.6776, 0.7552, 0.7581, 0.8380, 0.9261, 0.6814, 0.7769], device='cuda:1'), covar=tensor([ 6.0836, 21.3449, 14.1860, 11.7720, 11.3839, 15.5738, 19.9086, 15.4546], device='cuda:1'), in_proj_covar=tensor([0.0258, 0.0369, 0.0300, 0.0298, 0.0327, 0.0337, 0.0360, 0.0328], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-26 11:09:22,875 INFO [finetune.py:976] (1/7) Epoch 1, batch 3900, loss[loss=0.2731, simple_loss=0.2898, pruned_loss=0.1282, over 3987.00 frames. ], tot_loss[loss=0.3597, simple_loss=0.3779, pruned_loss=0.1707, over 956534.15 frames. ], batch size: 16, lr: 4.00e-03, grad_scale: 8.0 2023-04-26 11:09:29,481 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=3902.0, num_to_drop=1, layers_to_drop={3} 2023-04-26 11:09:44,795 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.635e+02 2.527e+02 3.007e+02 3.749e+02 8.787e+02, threshold=6.015e+02, percent-clipped=2.0 2023-04-26 11:09:49,804 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=3926.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:10:02,238 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=3944.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 11:10:07,405 INFO [finetune.py:976] (1/7) Epoch 1, batch 3950, loss[loss=0.3596, simple_loss=0.3697, pruned_loss=0.1748, over 4917.00 frames. ], tot_loss[loss=0.3561, simple_loss=0.3739, pruned_loss=0.1691, over 953991.66 frames. ], batch size: 43, lr: 4.00e-03, grad_scale: 8.0 2023-04-26 11:10:25,617 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8921, 2.2730, 0.9155, 1.2439, 1.3767, 1.1892, 2.5412, 1.4257], device='cuda:1'), covar=tensor([0.0741, 0.0498, 0.0668, 0.1118, 0.0514, 0.0985, 0.0271, 0.0673], device='cuda:1'), in_proj_covar=tensor([0.0057, 0.0073, 0.0053, 0.0050, 0.0055, 0.0056, 0.0088, 0.0054], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0008, 0.0008, 0.0008, 0.0008, 0.0012, 0.0008], device='cuda:1') 2023-04-26 11:10:30,551 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=3987.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:10:34,042 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=3992.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 11:10:42,151 INFO [finetune.py:976] (1/7) Epoch 1, batch 4000, loss[loss=0.3264, simple_loss=0.3597, pruned_loss=0.1465, over 4825.00 frames. ], tot_loss[loss=0.3512, simple_loss=0.3694, pruned_loss=0.1665, over 952957.78 frames. ], batch size: 33, lr: 4.00e-03, grad_scale: 8.0 2023-04-26 11:10:49,862 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([3.5015, 3.4860, 2.6502, 3.9587, 3.4375, 3.5364, 1.3789, 3.4012], device='cuda:1'), covar=tensor([0.1746, 0.1122, 0.2871, 0.1961, 0.2849, 0.1860, 0.5447, 0.2244], device='cuda:1'), in_proj_covar=tensor([0.0262, 0.0232, 0.0284, 0.0328, 0.0322, 0.0271, 0.0286, 0.0286], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-26 11:10:51,861 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.99 vs. limit=5.0 2023-04-26 11:10:54,137 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.677e+02 2.386e+02 2.816e+02 3.337e+02 7.046e+02, threshold=5.633e+02, percent-clipped=3.0 2023-04-26 11:11:02,812 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=4032.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 11:11:15,866 INFO [finetune.py:976] (1/7) Epoch 1, batch 4050, loss[loss=0.3191, simple_loss=0.3466, pruned_loss=0.1458, over 4769.00 frames. ], tot_loss[loss=0.3513, simple_loss=0.3705, pruned_loss=0.166, over 951949.93 frames. ], batch size: 28, lr: 4.00e-03, grad_scale: 8.0 2023-04-26 11:11:49,877 INFO [finetune.py:976] (1/7) Epoch 1, batch 4100, loss[loss=0.3872, simple_loss=0.4243, pruned_loss=0.1751, over 4916.00 frames. ], tot_loss[loss=0.3528, simple_loss=0.3737, pruned_loss=0.166, over 953702.77 frames. ], batch size: 42, lr: 4.00e-03, grad_scale: 8.0 2023-04-26 11:12:08,255 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.683e+02 2.322e+02 2.767e+02 3.338e+02 6.077e+02, threshold=5.534e+02, percent-clipped=1.0 2023-04-26 11:12:32,907 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8534, 2.2061, 1.0478, 1.2519, 1.4459, 1.1095, 2.5717, 1.4334], device='cuda:1'), covar=tensor([0.0717, 0.0528, 0.0651, 0.1186, 0.0500, 0.1039, 0.0252, 0.0661], device='cuda:1'), in_proj_covar=tensor([0.0057, 0.0073, 0.0053, 0.0050, 0.0055, 0.0056, 0.0088, 0.0054], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0008, 0.0008, 0.0008, 0.0008, 0.0012, 0.0008], device='cuda:1') 2023-04-26 11:12:34,741 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=4141.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:12:41,352 INFO [finetune.py:976] (1/7) Epoch 1, batch 4150, loss[loss=0.3667, simple_loss=0.389, pruned_loss=0.1721, over 4896.00 frames. ], tot_loss[loss=0.3514, simple_loss=0.3734, pruned_loss=0.1647, over 954608.54 frames. ], batch size: 35, lr: 4.00e-03, grad_scale: 16.0 2023-04-26 11:13:06,604 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=2.04 vs. limit=2.0 2023-04-26 11:13:09,758 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2023-04-26 11:13:15,084 INFO [finetune.py:976] (1/7) Epoch 1, batch 4200, loss[loss=0.2826, simple_loss=0.3353, pruned_loss=0.1149, over 4737.00 frames. ], tot_loss[loss=0.3492, simple_loss=0.3729, pruned_loss=0.1628, over 955591.76 frames. ], batch size: 54, lr: 4.00e-03, grad_scale: 16.0 2023-04-26 11:13:16,194 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=4202.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 11:13:16,220 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=4202.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:13:25,257 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.1107, 1.6959, 1.4953, 1.7487, 1.6691, 2.0929, 1.5831, 3.5401], device='cuda:1'), covar=tensor([0.0763, 0.0702, 0.0716, 0.1258, 0.0649, 0.0539, 0.0688, 0.0148], device='cuda:1'), in_proj_covar=tensor([0.0039, 0.0040, 0.0041, 0.0046, 0.0041, 0.0040, 0.0041, 0.0064], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0017], device='cuda:1') 2023-04-26 11:13:28,645 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.716e+02 2.286e+02 2.811e+02 3.257e+02 1.063e+03, threshold=5.622e+02, percent-clipped=1.0 2023-04-26 11:14:05,468 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=4250.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 11:14:05,986 INFO [finetune.py:976] (1/7) Epoch 1, batch 4250, loss[loss=0.3886, simple_loss=0.3977, pruned_loss=0.1897, over 4897.00 frames. ], tot_loss[loss=0.343, simple_loss=0.3677, pruned_loss=0.1591, over 956251.18 frames. ], batch size: 32, lr: 4.00e-03, grad_scale: 16.0 2023-04-26 11:14:17,531 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=5.24 vs. limit=5.0 2023-04-26 11:14:51,200 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=4282.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:14:59,299 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=4288.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:15:07,280 INFO [finetune.py:976] (1/7) Epoch 1, batch 4300, loss[loss=0.3234, simple_loss=0.346, pruned_loss=0.1504, over 4799.00 frames. ], tot_loss[loss=0.3367, simple_loss=0.3619, pruned_loss=0.1558, over 955178.87 frames. ], batch size: 25, lr: 4.00e-03, grad_scale: 16.0 2023-04-26 11:15:19,787 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.630e+02 2.200e+02 2.676e+02 3.122e+02 6.239e+02, threshold=5.353e+02, percent-clipped=1.0 2023-04-26 11:15:25,292 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2023-04-26 11:15:29,470 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=4332.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 11:15:39,951 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=4349.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:15:41,058 INFO [finetune.py:976] (1/7) Epoch 1, batch 4350, loss[loss=0.3311, simple_loss=0.3586, pruned_loss=0.1518, over 4912.00 frames. ], tot_loss[loss=0.3306, simple_loss=0.3566, pruned_loss=0.1524, over 953922.60 frames. ], batch size: 37, lr: 4.00e-03, grad_scale: 16.0 2023-04-26 11:16:02,409 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=4380.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 11:16:15,320 INFO [finetune.py:976] (1/7) Epoch 1, batch 4400, loss[loss=0.3704, simple_loss=0.391, pruned_loss=0.1749, over 4806.00 frames. ], tot_loss[loss=0.3304, simple_loss=0.3566, pruned_loss=0.1521, over 954204.70 frames. ], batch size: 45, lr: 4.00e-03, grad_scale: 16.0 2023-04-26 11:16:17,432 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.60 vs. limit=5.0 2023-04-26 11:16:26,704 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.735e+02 2.527e+02 2.839e+02 3.555e+02 1.567e+03, threshold=5.678e+02, percent-clipped=5.0 2023-04-26 11:16:27,985 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6605, 0.9616, 1.2206, 1.7512, 1.2983, 0.9991, 0.8943, 1.1875], device='cuda:1'), covar=tensor([0.7809, 1.1239, 0.5172, 1.2250, 1.1815, 0.8209, 1.8640, 1.1440], device='cuda:1'), in_proj_covar=tensor([0.0250, 0.0270, 0.0212, 0.0328, 0.0230, 0.0223, 0.0265, 0.0218], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-26 11:16:48,901 INFO [finetune.py:976] (1/7) Epoch 1, batch 4450, loss[loss=0.3171, simple_loss=0.3592, pruned_loss=0.1375, over 4921.00 frames. ], tot_loss[loss=0.3344, simple_loss=0.3604, pruned_loss=0.1541, over 950761.78 frames. ], batch size: 38, lr: 4.00e-03, grad_scale: 16.0 2023-04-26 11:17:19,613 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=4497.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:17:22,031 INFO [finetune.py:976] (1/7) Epoch 1, batch 4500, loss[loss=0.351, simple_loss=0.3742, pruned_loss=0.1639, over 4899.00 frames. ], tot_loss[loss=0.3343, simple_loss=0.3617, pruned_loss=0.1534, over 953242.60 frames. ], batch size: 36, lr: 4.00e-03, grad_scale: 16.0 2023-04-26 11:17:22,736 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5864, 1.7083, 0.9846, 1.1982, 2.0557, 1.4704, 1.3188, 1.4691], device='cuda:1'), covar=tensor([0.0658, 0.0530, 0.0537, 0.0694, 0.0382, 0.0663, 0.0657, 0.0781], device='cuda:1'), in_proj_covar=tensor([0.0032, 0.0027, 0.0025, 0.0031, 0.0022, 0.0031, 0.0030, 0.0033], device='cuda:1'), out_proj_covar=tensor([0.0048, 0.0043, 0.0038, 0.0049, 0.0037, 0.0048, 0.0047, 0.0051], device='cuda:1') 2023-04-26 11:17:22,845 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=2.02 vs. limit=2.0 2023-04-26 11:17:27,918 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-04-26 11:17:32,938 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.661e+02 2.259e+02 2.800e+02 3.318e+02 7.116e+02, threshold=5.601e+02, percent-clipped=2.0 2023-04-26 11:18:15,792 INFO [finetune.py:976] (1/7) Epoch 1, batch 4550, loss[loss=0.3571, simple_loss=0.3794, pruned_loss=0.1674, over 4881.00 frames. ], tot_loss[loss=0.3346, simple_loss=0.3634, pruned_loss=0.1529, over 954092.81 frames. ], batch size: 35, lr: 4.00e-03, grad_scale: 16.0 2023-04-26 11:19:01,281 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=4582.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:19:13,032 INFO [finetune.py:976] (1/7) Epoch 1, batch 4600, loss[loss=0.3053, simple_loss=0.3351, pruned_loss=0.1377, over 4830.00 frames. ], tot_loss[loss=0.3328, simple_loss=0.3618, pruned_loss=0.1519, over 955699.93 frames. ], batch size: 47, lr: 4.00e-03, grad_scale: 16.0 2023-04-26 11:19:23,471 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.373e+02 2.361e+02 2.771e+02 3.480e+02 8.913e+02, threshold=5.542e+02, percent-clipped=3.0 2023-04-26 11:19:33,260 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=4630.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:19:40,011 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-04-26 11:19:42,839 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=4644.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:19:47,128 INFO [finetune.py:976] (1/7) Epoch 1, batch 4650, loss[loss=0.3509, simple_loss=0.3589, pruned_loss=0.1715, over 4132.00 frames. ], tot_loss[loss=0.3284, simple_loss=0.3572, pruned_loss=0.1498, over 954367.20 frames. ], batch size: 65, lr: 4.00e-03, grad_scale: 16.0 2023-04-26 11:20:42,427 INFO [finetune.py:976] (1/7) Epoch 1, batch 4700, loss[loss=0.2819, simple_loss=0.3109, pruned_loss=0.1265, over 4910.00 frames. ], tot_loss[loss=0.3218, simple_loss=0.3511, pruned_loss=0.1463, over 953803.45 frames. ], batch size: 32, lr: 4.00e-03, grad_scale: 16.0 2023-04-26 11:21:03,816 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.132e+02 1.981e+02 2.474e+02 3.057e+02 6.452e+02, threshold=4.948e+02, percent-clipped=2.0 2023-04-26 11:21:05,915 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.53 vs. limit=2.0 2023-04-26 11:21:39,261 INFO [finetune.py:976] (1/7) Epoch 1, batch 4750, loss[loss=0.2688, simple_loss=0.3153, pruned_loss=0.1112, over 4816.00 frames. ], tot_loss[loss=0.3184, simple_loss=0.3484, pruned_loss=0.1442, over 955011.04 frames. ], batch size: 51, lr: 4.00e-03, grad_scale: 16.0 2023-04-26 11:21:59,464 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7464, 1.8622, 2.1434, 2.1401, 2.1405, 1.5904, 1.2675, 1.8432], device='cuda:1'), covar=tensor([0.1313, 0.1219, 0.0662, 0.0898, 0.0785, 0.1324, 0.1423, 0.0905], device='cuda:1'), in_proj_covar=tensor([0.0205, 0.0207, 0.0188, 0.0178, 0.0175, 0.0192, 0.0171, 0.0188], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 11:22:01,844 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6820, 1.9809, 1.5251, 1.9573, 1.5086, 1.4959, 1.7215, 1.3491], device='cuda:1'), covar=tensor([0.2276, 0.1720, 0.1625, 0.1586, 0.3306, 0.1817, 0.2100, 0.2992], device='cuda:1'), in_proj_covar=tensor([0.0308, 0.0326, 0.0241, 0.0303, 0.0310, 0.0278, 0.0275, 0.0296], device='cuda:1'), out_proj_covar=tensor([1.2610e-04, 1.3370e-04, 9.8578e-05, 1.2292e-04, 1.2820e-04, 1.1262e-04, 1.1394e-04, 1.2052e-04], device='cuda:1') 2023-04-26 11:22:08,690 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4472, 2.0380, 1.3205, 1.3285, 0.9703, 1.0607, 1.3187, 0.8987], device='cuda:1'), covar=tensor([0.2432, 0.2537, 0.3169, 0.3813, 0.4383, 0.3100, 0.2387, 0.3482], device='cuda:1'), in_proj_covar=tensor([0.0198, 0.0221, 0.0200, 0.0217, 0.0237, 0.0198, 0.0193, 0.0209], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-26 11:22:10,388 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5830, 1.3459, 4.2973, 3.9465, 3.7805, 4.0487, 3.9958, 3.8033], device='cuda:1'), covar=tensor([0.6931, 0.5998, 0.1024, 0.1774, 0.1157, 0.1835, 0.1568, 0.1597], device='cuda:1'), in_proj_covar=tensor([0.0343, 0.0319, 0.0455, 0.0463, 0.0383, 0.0438, 0.0346, 0.0409], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0002], device='cuda:1') 2023-04-26 11:22:11,037 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=4797.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:22:13,545 INFO [finetune.py:976] (1/7) Epoch 1, batch 4800, loss[loss=0.3812, simple_loss=0.3971, pruned_loss=0.1827, over 4904.00 frames. ], tot_loss[loss=0.321, simple_loss=0.351, pruned_loss=0.1455, over 953722.83 frames. ], batch size: 37, lr: 4.00e-03, grad_scale: 16.0 2023-04-26 11:22:16,767 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6933, 1.7036, 1.8552, 1.9833, 1.9407, 1.6384, 1.3444, 1.7841], device='cuda:1'), covar=tensor([0.1217, 0.1115, 0.0679, 0.0853, 0.0836, 0.1186, 0.1387, 0.0816], device='cuda:1'), in_proj_covar=tensor([0.0206, 0.0208, 0.0188, 0.0179, 0.0176, 0.0193, 0.0172, 0.0189], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 11:22:19,915 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=5.47 vs. limit=5.0 2023-04-26 11:22:24,010 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.522e+02 2.397e+02 2.799e+02 3.310e+02 5.685e+02, threshold=5.598e+02, percent-clipped=2.0 2023-04-26 11:22:43,637 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=4845.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:22:44,747 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4997, 1.5373, 0.7470, 1.2295, 1.8007, 1.4202, 1.2907, 1.3638], device='cuda:1'), covar=tensor([0.0691, 0.0535, 0.0600, 0.0693, 0.0383, 0.0698, 0.0670, 0.0807], device='cuda:1'), in_proj_covar=tensor([0.0032, 0.0027, 0.0025, 0.0031, 0.0022, 0.0031, 0.0030, 0.0033], device='cuda:1'), out_proj_covar=tensor([0.0049, 0.0043, 0.0038, 0.0049, 0.0037, 0.0048, 0.0047, 0.0051], device='cuda:1') 2023-04-26 11:22:47,739 INFO [finetune.py:976] (1/7) Epoch 1, batch 4850, loss[loss=0.2888, simple_loss=0.3297, pruned_loss=0.1239, over 4868.00 frames. ], tot_loss[loss=0.3217, simple_loss=0.3534, pruned_loss=0.145, over 953244.59 frames. ], batch size: 31, lr: 4.00e-03, grad_scale: 16.0 2023-04-26 11:23:24,932 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9507, 2.3849, 1.8125, 2.3583, 1.8874, 1.7870, 2.0073, 1.5325], device='cuda:1'), covar=tensor([0.2171, 0.1493, 0.1532, 0.1305, 0.2954, 0.1608, 0.1941, 0.3151], device='cuda:1'), in_proj_covar=tensor([0.0308, 0.0327, 0.0241, 0.0304, 0.0309, 0.0278, 0.0275, 0.0296], device='cuda:1'), out_proj_covar=tensor([1.2613e-04, 1.3397e-04, 9.8616e-05, 1.2309e-04, 1.2815e-04, 1.1255e-04, 1.1415e-04, 1.2060e-04], device='cuda:1') 2023-04-26 11:23:33,606 INFO [finetune.py:976] (1/7) Epoch 1, batch 4900, loss[loss=0.2951, simple_loss=0.3449, pruned_loss=0.1226, over 4733.00 frames. ], tot_loss[loss=0.3229, simple_loss=0.3553, pruned_loss=0.1453, over 955041.57 frames. ], batch size: 59, lr: 4.00e-03, grad_scale: 16.0 2023-04-26 11:23:49,557 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.488e+02 2.175e+02 2.572e+02 3.151e+02 5.494e+02, threshold=5.143e+02, percent-clipped=0.0 2023-04-26 11:24:29,556 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=4944.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:24:30,760 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-04-26 11:24:41,257 INFO [finetune.py:976] (1/7) Epoch 1, batch 4950, loss[loss=0.3135, simple_loss=0.341, pruned_loss=0.143, over 4750.00 frames. ], tot_loss[loss=0.3212, simple_loss=0.3544, pruned_loss=0.1441, over 955072.35 frames. ], batch size: 28, lr: 4.00e-03, grad_scale: 16.0 2023-04-26 11:25:18,982 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=4992.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:25:26,586 INFO [finetune.py:976] (1/7) Epoch 1, batch 5000, loss[loss=0.3494, simple_loss=0.3773, pruned_loss=0.1608, over 4826.00 frames. ], tot_loss[loss=0.3153, simple_loss=0.3494, pruned_loss=0.1406, over 955465.18 frames. ], batch size: 40, lr: 4.00e-03, grad_scale: 16.0 2023-04-26 11:25:37,462 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.472e+02 2.232e+02 2.673e+02 3.232e+02 7.029e+02, threshold=5.346e+02, percent-clipped=5.0 2023-04-26 11:26:10,645 INFO [finetune.py:976] (1/7) Epoch 1, batch 5050, loss[loss=0.2973, simple_loss=0.3292, pruned_loss=0.1327, over 4927.00 frames. ], tot_loss[loss=0.3105, simple_loss=0.3446, pruned_loss=0.1382, over 956940.47 frames. ], batch size: 33, lr: 4.00e-03, grad_scale: 16.0 2023-04-26 11:27:11,307 INFO [finetune.py:976] (1/7) Epoch 1, batch 5100, loss[loss=0.2552, simple_loss=0.304, pruned_loss=0.1032, over 4790.00 frames. ], tot_loss[loss=0.3059, simple_loss=0.3402, pruned_loss=0.1358, over 957117.22 frames. ], batch size: 29, lr: 4.00e-03, grad_scale: 16.0 2023-04-26 11:27:40,805 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.449e+02 2.109e+02 2.481e+02 2.814e+02 6.642e+02, threshold=4.963e+02, percent-clipped=1.0 2023-04-26 11:28:17,910 INFO [finetune.py:976] (1/7) Epoch 1, batch 5150, loss[loss=0.3118, simple_loss=0.349, pruned_loss=0.1373, over 4820.00 frames. ], tot_loss[loss=0.3059, simple_loss=0.3398, pruned_loss=0.136, over 957164.64 frames. ], batch size: 38, lr: 4.00e-03, grad_scale: 16.0 2023-04-26 11:28:55,637 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6165, 1.8415, 1.9830, 1.9757, 1.9701, 1.5712, 1.2944, 1.7481], device='cuda:1'), covar=tensor([0.1078, 0.1087, 0.0665, 0.0799, 0.0774, 0.1169, 0.1230, 0.0724], device='cuda:1'), in_proj_covar=tensor([0.0207, 0.0208, 0.0188, 0.0180, 0.0176, 0.0194, 0.0173, 0.0189], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 11:29:06,099 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8303, 1.1375, 4.3400, 3.7154, 3.8906, 4.0580, 3.9238, 3.6704], device='cuda:1'), covar=tensor([0.8430, 0.9504, 0.1477, 0.3464, 0.2261, 0.3427, 0.2310, 0.2537], device='cuda:1'), in_proj_covar=tensor([0.0340, 0.0319, 0.0452, 0.0462, 0.0380, 0.0437, 0.0344, 0.0406], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0002], device='cuda:1') 2023-04-26 11:29:07,360 INFO [finetune.py:976] (1/7) Epoch 1, batch 5200, loss[loss=0.3371, simple_loss=0.3903, pruned_loss=0.1419, over 4851.00 frames. ], tot_loss[loss=0.3119, simple_loss=0.3461, pruned_loss=0.1388, over 956229.83 frames. ], batch size: 44, lr: 4.00e-03, grad_scale: 16.0 2023-04-26 11:29:21,206 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.499e+02 2.511e+02 2.846e+02 3.705e+02 8.558e+02, threshold=5.692e+02, percent-clipped=9.0 2023-04-26 11:29:41,592 INFO [finetune.py:976] (1/7) Epoch 1, batch 5250, loss[loss=0.2778, simple_loss=0.3234, pruned_loss=0.1161, over 4857.00 frames. ], tot_loss[loss=0.312, simple_loss=0.3474, pruned_loss=0.1383, over 956092.16 frames. ], batch size: 31, lr: 4.00e-03, grad_scale: 16.0 2023-04-26 11:30:05,441 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-04-26 11:30:15,015 INFO [finetune.py:976] (1/7) Epoch 1, batch 5300, loss[loss=0.3059, simple_loss=0.3415, pruned_loss=0.1351, over 4828.00 frames. ], tot_loss[loss=0.3102, simple_loss=0.3468, pruned_loss=0.1367, over 956020.65 frames. ], batch size: 49, lr: 4.00e-03, grad_scale: 16.0 2023-04-26 11:30:27,273 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.471e+02 2.212e+02 2.624e+02 3.137e+02 5.522e+02, threshold=5.248e+02, percent-clipped=0.0 2023-04-26 11:30:49,039 INFO [finetune.py:976] (1/7) Epoch 1, batch 5350, loss[loss=0.3565, simple_loss=0.3843, pruned_loss=0.1644, over 4918.00 frames. ], tot_loss[loss=0.3088, simple_loss=0.3459, pruned_loss=0.1359, over 953418.33 frames. ], batch size: 38, lr: 4.00e-03, grad_scale: 16.0 2023-04-26 11:31:32,459 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6861, 2.0457, 1.5673, 2.0473, 1.5118, 1.4838, 1.7958, 1.2910], device='cuda:1'), covar=tensor([0.2128, 0.1528, 0.1502, 0.1398, 0.3535, 0.1831, 0.1843, 0.3023], device='cuda:1'), in_proj_covar=tensor([0.0308, 0.0327, 0.0241, 0.0303, 0.0309, 0.0279, 0.0275, 0.0294], device='cuda:1'), out_proj_covar=tensor([1.2622e-04, 1.3429e-04, 9.8716e-05, 1.2263e-04, 1.2782e-04, 1.1296e-04, 1.1392e-04, 1.1987e-04], device='cuda:1') 2023-04-26 11:31:35,564 INFO [finetune.py:976] (1/7) Epoch 1, batch 5400, loss[loss=0.2901, simple_loss=0.3232, pruned_loss=0.1285, over 4893.00 frames. ], tot_loss[loss=0.3043, simple_loss=0.3412, pruned_loss=0.1337, over 954473.89 frames. ], batch size: 32, lr: 4.00e-03, grad_scale: 16.0 2023-04-26 11:31:51,771 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([0.3665, 1.4398, 1.4003, 1.1837, 1.4979, 1.1462, 1.8097, 1.3165], device='cuda:1'), covar=tensor([0.4076, 0.1537, 0.4954, 0.2299, 0.1542, 0.2187, 0.1378, 0.4382], device='cuda:1'), in_proj_covar=tensor([0.0329, 0.0340, 0.0421, 0.0355, 0.0389, 0.0362, 0.0386, 0.0397], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 11:31:52,845 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.437e+02 2.058e+02 2.416e+02 3.015e+02 5.928e+02, threshold=4.831e+02, percent-clipped=1.0 2023-04-26 11:32:15,535 INFO [finetune.py:976] (1/7) Epoch 1, batch 5450, loss[loss=0.3278, simple_loss=0.3634, pruned_loss=0.1461, over 4918.00 frames. ], tot_loss[loss=0.2982, simple_loss=0.3353, pruned_loss=0.1306, over 953360.30 frames. ], batch size: 37, lr: 4.00e-03, grad_scale: 16.0 2023-04-26 11:32:24,226 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4541, 0.8739, 0.3547, 1.1854, 0.9729, 1.3449, 1.2336, 1.2794], device='cuda:1'), covar=tensor([0.0646, 0.0551, 0.0580, 0.0655, 0.0420, 0.0649, 0.0628, 0.0755], device='cuda:1'), in_proj_covar=tensor([0.0032, 0.0027, 0.0025, 0.0031, 0.0022, 0.0031, 0.0030, 0.0033], device='cuda:1'), out_proj_covar=tensor([0.0048, 0.0043, 0.0038, 0.0049, 0.0037, 0.0048, 0.0047, 0.0051], device='cuda:1') 2023-04-26 11:32:38,887 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5858, 2.1755, 1.3286, 1.2738, 1.1209, 1.1520, 1.3508, 1.0001], device='cuda:1'), covar=tensor([0.2673, 0.2375, 0.3164, 0.3740, 0.4222, 0.3348, 0.2414, 0.3567], device='cuda:1'), in_proj_covar=tensor([0.0198, 0.0221, 0.0199, 0.0217, 0.0236, 0.0197, 0.0193, 0.0209], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-26 11:32:49,243 INFO [finetune.py:976] (1/7) Epoch 1, batch 5500, loss[loss=0.2864, simple_loss=0.3318, pruned_loss=0.1205, over 4772.00 frames. ], tot_loss[loss=0.2917, simple_loss=0.3293, pruned_loss=0.127, over 952479.86 frames. ], batch size: 51, lr: 4.00e-03, grad_scale: 16.0 2023-04-26 11:32:59,728 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.389e+02 2.158e+02 2.515e+02 3.075e+02 5.534e+02, threshold=5.030e+02, percent-clipped=1.0 2023-04-26 11:33:46,363 INFO [finetune.py:976] (1/7) Epoch 1, batch 5550, loss[loss=0.2426, simple_loss=0.2847, pruned_loss=0.1002, over 4722.00 frames. ], tot_loss[loss=0.2945, simple_loss=0.332, pruned_loss=0.1286, over 953441.15 frames. ], batch size: 23, lr: 4.00e-03, grad_scale: 16.0 2023-04-26 11:34:07,132 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([0.9439, 0.9096, 1.1168, 1.1072, 1.0199, 0.9222, 0.9682, 0.9406], device='cuda:1'), covar=tensor([ 9.5040, 14.5747, 15.6645, 14.5582, 12.2385, 17.4352, 17.0640, 11.3346], device='cuda:1'), in_proj_covar=tensor([0.0430, 0.0495, 0.0570, 0.0546, 0.0457, 0.0517, 0.0523, 0.0532], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 11:34:53,183 INFO [finetune.py:976] (1/7) Epoch 1, batch 5600, loss[loss=0.3108, simple_loss=0.3597, pruned_loss=0.131, over 4845.00 frames. ], tot_loss[loss=0.2986, simple_loss=0.3371, pruned_loss=0.1301, over 954391.03 frames. ], batch size: 49, lr: 4.00e-03, grad_scale: 16.0 2023-04-26 11:35:13,912 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.486e+02 2.261e+02 2.635e+02 3.434e+02 7.043e+02, threshold=5.269e+02, percent-clipped=4.0 2023-04-26 11:35:23,317 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.3758, 1.4229, 1.6274, 1.6799, 1.7567, 1.3149, 1.0889, 1.5204], device='cuda:1'), covar=tensor([0.1258, 0.1315, 0.0875, 0.0900, 0.0804, 0.1308, 0.1442, 0.0845], device='cuda:1'), in_proj_covar=tensor([0.0209, 0.0210, 0.0190, 0.0181, 0.0178, 0.0196, 0.0174, 0.0191], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 11:35:44,937 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=5650.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 11:35:45,423 INFO [finetune.py:976] (1/7) Epoch 1, batch 5650, loss[loss=0.3819, simple_loss=0.3845, pruned_loss=0.1897, over 4338.00 frames. ], tot_loss[loss=0.3038, simple_loss=0.3419, pruned_loss=0.1328, over 952079.75 frames. ], batch size: 19, lr: 4.00e-03, grad_scale: 16.0 2023-04-26 11:35:56,199 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.2017, 1.0700, 1.2056, 1.4499, 1.3423, 1.2169, 1.1906, 1.2047], device='cuda:1'), covar=tensor([10.1331, 15.1171, 14.5640, 15.1107, 16.2160, 15.7568, 14.8183, 9.3897], device='cuda:1'), in_proj_covar=tensor([0.0432, 0.0497, 0.0572, 0.0549, 0.0459, 0.0519, 0.0525, 0.0534], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 11:36:15,236 INFO [finetune.py:976] (1/7) Epoch 1, batch 5700, loss[loss=0.2324, simple_loss=0.2736, pruned_loss=0.09565, over 4372.00 frames. ], tot_loss[loss=0.2991, simple_loss=0.3364, pruned_loss=0.1309, over 934810.70 frames. ], batch size: 19, lr: 4.00e-03, grad_scale: 16.0 2023-04-26 11:36:16,592 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=5.18 vs. limit=5.0 2023-04-26 11:36:21,270 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=5711.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 11:36:31,944 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.325e+02 1.970e+02 2.507e+02 3.043e+02 6.051e+02, threshold=5.014e+02, percent-clipped=1.0 2023-04-26 11:37:01,400 INFO [finetune.py:976] (1/7) Epoch 2, batch 0, loss[loss=0.3151, simple_loss=0.3592, pruned_loss=0.1355, over 4809.00 frames. ], tot_loss[loss=0.3151, simple_loss=0.3592, pruned_loss=0.1355, over 4809.00 frames. ], batch size: 33, lr: 4.00e-03, grad_scale: 16.0 2023-04-26 11:37:01,400 INFO [finetune.py:1001] (1/7) Computing validation loss 2023-04-26 11:37:23,938 INFO [finetune.py:1010] (1/7) Epoch 2, validation: loss=0.2101, simple_loss=0.2777, pruned_loss=0.0712, over 2265189.00 frames. 2023-04-26 11:37:23,939 INFO [finetune.py:1011] (1/7) Maximum memory allocated so far is 6069MB 2023-04-26 11:37:46,568 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=5755.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:38:02,075 INFO [finetune.py:976] (1/7) Epoch 2, batch 50, loss[loss=0.2817, simple_loss=0.3312, pruned_loss=0.1161, over 4913.00 frames. ], tot_loss[loss=0.2955, simple_loss=0.3348, pruned_loss=0.1281, over 214359.06 frames. ], batch size: 46, lr: 4.00e-03, grad_scale: 16.0 2023-04-26 11:38:04,572 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-04-26 11:38:14,017 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.1571, 2.6251, 1.0662, 1.4585, 1.9722, 1.3032, 3.5016, 1.6294], device='cuda:1'), covar=tensor([0.0686, 0.0638, 0.0905, 0.1201, 0.0568, 0.1026, 0.0190, 0.0704], device='cuda:1'), in_proj_covar=tensor([0.0058, 0.0074, 0.0054, 0.0051, 0.0056, 0.0056, 0.0088, 0.0054], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0008, 0.0008, 0.0008, 0.0008, 0.0012, 0.0008], device='cuda:1') 2023-04-26 11:38:22,292 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.76 vs. limit=2.0 2023-04-26 11:38:27,681 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=5816.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:38:29,284 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.431e+02 2.066e+02 2.410e+02 2.890e+02 4.929e+02, threshold=4.819e+02, percent-clipped=0.0 2023-04-26 11:38:35,780 INFO [finetune.py:976] (1/7) Epoch 2, batch 100, loss[loss=0.2446, simple_loss=0.2953, pruned_loss=0.09698, over 4759.00 frames. ], tot_loss[loss=0.2863, simple_loss=0.3265, pruned_loss=0.1231, over 378971.60 frames. ], batch size: 28, lr: 4.00e-03, grad_scale: 16.0 2023-04-26 11:38:50,879 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.3677, 1.1630, 1.5891, 1.5019, 1.2825, 1.0558, 1.2390, 0.8409], device='cuda:1'), covar=tensor([0.0925, 0.1060, 0.0694, 0.1026, 0.1297, 0.1718, 0.0915, 0.1480], device='cuda:1'), in_proj_covar=tensor([0.0072, 0.0079, 0.0076, 0.0076, 0.0090, 0.0095, 0.0091, 0.0081], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003], device='cuda:1') 2023-04-26 11:38:54,553 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([0.7861, 1.8668, 1.7832, 1.5041, 1.9845, 1.6665, 2.5559, 1.4938], device='cuda:1'), covar=tensor([0.4173, 0.1473, 0.4360, 0.2573, 0.1756, 0.2209, 0.1179, 0.4083], device='cuda:1'), in_proj_covar=tensor([0.0331, 0.0342, 0.0424, 0.0357, 0.0392, 0.0363, 0.0388, 0.0398], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 11:38:55,147 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=5858.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:39:01,393 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([0.9240, 0.7687, 0.9630, 1.1711, 1.1070, 0.9069, 0.9139, 0.9416], device='cuda:1'), covar=tensor([ 9.6284, 14.4303, 12.7324, 13.1797, 12.2140, 14.4004, 15.4099, 9.6208], device='cuda:1'), in_proj_covar=tensor([0.0435, 0.0500, 0.0577, 0.0555, 0.0463, 0.0522, 0.0528, 0.0537], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 11:39:09,190 INFO [finetune.py:976] (1/7) Epoch 2, batch 150, loss[loss=0.2771, simple_loss=0.3228, pruned_loss=0.1157, over 4921.00 frames. ], tot_loss[loss=0.2801, simple_loss=0.3201, pruned_loss=0.12, over 507198.37 frames. ], batch size: 33, lr: 4.00e-03, grad_scale: 16.0 2023-04-26 11:39:31,364 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8476, 1.6652, 2.3807, 2.3986, 1.6546, 1.2924, 1.8573, 1.1445], device='cuda:1'), covar=tensor([0.1101, 0.1161, 0.0636, 0.0951, 0.1590, 0.1813, 0.1071, 0.1773], device='cuda:1'), in_proj_covar=tensor([0.0071, 0.0079, 0.0075, 0.0075, 0.0089, 0.0094, 0.0090, 0.0080], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003], device='cuda:1') 2023-04-26 11:39:34,932 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.317e+02 2.049e+02 2.408e+02 2.939e+02 5.453e+02, threshold=4.816e+02, percent-clipped=1.0 2023-04-26 11:39:35,641 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7715, 4.2371, 0.9006, 2.2820, 2.3941, 2.7492, 2.9170, 1.0186], device='cuda:1'), covar=tensor([0.1311, 0.1047, 0.2183, 0.1278, 0.0961, 0.1111, 0.1146, 0.1976], device='cuda:1'), in_proj_covar=tensor([0.0123, 0.0271, 0.0150, 0.0132, 0.0142, 0.0165, 0.0129, 0.0132], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-26 11:39:35,679 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=5919.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:39:39,693 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0727, 1.2906, 5.1612, 4.7702, 4.5723, 4.8014, 4.5939, 4.5552], device='cuda:1'), covar=tensor([0.6038, 0.6264, 0.0904, 0.1858, 0.1005, 0.1180, 0.1086, 0.1255], device='cuda:1'), in_proj_covar=tensor([0.0337, 0.0314, 0.0450, 0.0459, 0.0378, 0.0436, 0.0343, 0.0402], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0002], device='cuda:1') 2023-04-26 11:39:48,162 INFO [finetune.py:976] (1/7) Epoch 2, batch 200, loss[loss=0.2306, simple_loss=0.28, pruned_loss=0.09058, over 4718.00 frames. ], tot_loss[loss=0.2806, simple_loss=0.3199, pruned_loss=0.1207, over 608413.57 frames. ], batch size: 23, lr: 4.00e-03, grad_scale: 16.0 2023-04-26 11:40:40,476 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.49 vs. limit=2.0 2023-04-26 11:40:52,009 INFO [finetune.py:976] (1/7) Epoch 2, batch 250, loss[loss=0.2664, simple_loss=0.3115, pruned_loss=0.1107, over 4768.00 frames. ], tot_loss[loss=0.2859, simple_loss=0.3248, pruned_loss=0.1235, over 682701.16 frames. ], batch size: 26, lr: 4.00e-03, grad_scale: 16.0 2023-04-26 11:41:06,861 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=5990.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:41:09,885 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-04-26 11:41:18,234 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6006.0, num_to_drop=1, layers_to_drop={2} 2023-04-26 11:41:25,491 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.356e+02 2.171e+02 2.695e+02 3.137e+02 1.432e+03, threshold=5.390e+02, percent-clipped=4.0 2023-04-26 11:41:33,100 INFO [finetune.py:976] (1/7) Epoch 2, batch 300, loss[loss=0.2605, simple_loss=0.3021, pruned_loss=0.1094, over 4737.00 frames. ], tot_loss[loss=0.2889, simple_loss=0.3286, pruned_loss=0.1246, over 743590.51 frames. ], batch size: 23, lr: 4.00e-03, grad_scale: 16.0 2023-04-26 11:41:49,306 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=6043.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:41:54,642 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=6051.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:42:22,355 INFO [finetune.py:976] (1/7) Epoch 2, batch 350, loss[loss=0.2992, simple_loss=0.348, pruned_loss=0.1252, over 4841.00 frames. ], tot_loss[loss=0.2917, simple_loss=0.332, pruned_loss=0.1256, over 792056.90 frames. ], batch size: 47, lr: 4.00e-03, grad_scale: 16.0 2023-04-26 11:42:47,470 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=6104.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:42:57,936 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6111.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:43:02,078 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.403e+02 2.085e+02 2.458e+02 2.985e+02 5.837e+02, threshold=4.917e+02, percent-clipped=2.0 2023-04-26 11:43:12,559 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.4653, 1.4972, 1.4541, 1.9432, 2.3239, 2.0185, 1.8045, 1.7766], device='cuda:1'), covar=tensor([0.1378, 0.2495, 0.2588, 0.2809, 0.1361, 0.2528, 0.2866, 0.2214], device='cuda:1'), in_proj_covar=tensor([0.0329, 0.0345, 0.0348, 0.0316, 0.0354, 0.0380, 0.0327, 0.0360], device='cuda:1'), out_proj_covar=tensor([7.1982e-05, 7.4310e-05, 7.5090e-05, 6.6269e-05, 7.5522e-05, 8.3172e-05, 7.1251e-05, 7.7858e-05], device='cuda:1') 2023-04-26 11:43:14,250 INFO [finetune.py:976] (1/7) Epoch 2, batch 400, loss[loss=0.326, simple_loss=0.3602, pruned_loss=0.1459, over 4833.00 frames. ], tot_loss[loss=0.292, simple_loss=0.3328, pruned_loss=0.1256, over 829581.03 frames. ], batch size: 30, lr: 4.00e-03, grad_scale: 16.0 2023-04-26 11:43:16,914 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-04-26 11:43:20,196 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=6135.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:43:43,647 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=6170.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:43:48,408 INFO [finetune.py:976] (1/7) Epoch 2, batch 450, loss[loss=0.2996, simple_loss=0.3425, pruned_loss=0.1284, over 4833.00 frames. ], tot_loss[loss=0.2905, simple_loss=0.3315, pruned_loss=0.1247, over 859187.07 frames. ], batch size: 33, lr: 4.00e-03, grad_scale: 32.0 2023-04-26 11:43:49,720 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.2407, 2.5430, 0.9577, 1.5613, 1.5621, 1.9766, 1.7608, 0.8822], device='cuda:1'), covar=tensor([0.1218, 0.0938, 0.1599, 0.1220, 0.0958, 0.0800, 0.1375, 0.1514], device='cuda:1'), in_proj_covar=tensor([0.0124, 0.0273, 0.0151, 0.0133, 0.0144, 0.0166, 0.0130, 0.0133], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-26 11:44:00,945 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=6196.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:44:06,748 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=6203.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:44:11,646 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=6211.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:44:13,463 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6214.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:44:15,817 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.476e+02 2.104e+02 2.551e+02 3.094e+02 4.755e+02, threshold=5.103e+02, percent-clipped=0.0 2023-04-26 11:44:17,756 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.1025, 1.8022, 4.7274, 4.4506, 4.2560, 4.3853, 4.3654, 4.2267], device='cuda:1'), covar=tensor([0.5727, 0.5333, 0.0990, 0.1812, 0.0991, 0.1743, 0.1018, 0.1475], device='cuda:1'), in_proj_covar=tensor([0.0341, 0.0318, 0.0456, 0.0464, 0.0382, 0.0441, 0.0347, 0.0405], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0002], device='cuda:1') 2023-04-26 11:44:21,877 INFO [finetune.py:976] (1/7) Epoch 2, batch 500, loss[loss=0.2173, simple_loss=0.2724, pruned_loss=0.0811, over 4770.00 frames. ], tot_loss[loss=0.2868, simple_loss=0.3283, pruned_loss=0.1227, over 880831.55 frames. ], batch size: 28, lr: 4.00e-03, grad_scale: 32.0 2023-04-26 11:44:23,790 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=6231.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:44:46,820 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=6264.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:44:48,006 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=6266.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:44:51,700 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=6272.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:44:55,246 INFO [finetune.py:976] (1/7) Epoch 2, batch 550, loss[loss=0.3348, simple_loss=0.3494, pruned_loss=0.1601, over 4898.00 frames. ], tot_loss[loss=0.284, simple_loss=0.3248, pruned_loss=0.1216, over 896591.67 frames. ], batch size: 43, lr: 4.00e-03, grad_scale: 32.0 2023-04-26 11:44:57,218 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5100, 1.7895, 1.4026, 1.7580, 1.4970, 1.4134, 1.7111, 1.1760], device='cuda:1'), covar=tensor([0.1849, 0.1469, 0.1311, 0.1365, 0.2903, 0.1530, 0.1691, 0.2596], device='cuda:1'), in_proj_covar=tensor([0.0312, 0.0334, 0.0244, 0.0309, 0.0311, 0.0283, 0.0279, 0.0300], device='cuda:1'), out_proj_covar=tensor([1.2769e-04, 1.3705e-04, 9.9974e-05, 1.2517e-04, 1.2892e-04, 1.1456e-04, 1.1552e-04, 1.2221e-04], device='cuda:1') 2023-04-26 11:45:15,017 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=6306.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 11:45:28,233 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.445e+02 2.269e+02 2.692e+02 3.237e+02 6.108e+02, threshold=5.385e+02, percent-clipped=2.0 2023-04-26 11:45:37,843 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=6327.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:45:38,337 INFO [finetune.py:976] (1/7) Epoch 2, batch 600, loss[loss=0.2814, simple_loss=0.3122, pruned_loss=0.1253, over 4769.00 frames. ], tot_loss[loss=0.2843, simple_loss=0.325, pruned_loss=0.1218, over 906783.92 frames. ], batch size: 26, lr: 4.00e-03, grad_scale: 32.0 2023-04-26 11:45:40,959 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4360, 2.4285, 2.0308, 2.2419, 2.6585, 1.9792, 3.2532, 1.8921], device='cuda:1'), covar=tensor([0.4652, 0.1839, 0.4541, 0.3181, 0.1940, 0.2891, 0.1378, 0.4347], device='cuda:1'), in_proj_covar=tensor([0.0335, 0.0343, 0.0425, 0.0358, 0.0395, 0.0365, 0.0389, 0.0399], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 11:46:00,720 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6346.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:46:10,183 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5115, 1.6366, 0.9177, 1.2037, 1.7802, 1.4295, 1.2984, 1.4162], device='cuda:1'), covar=tensor([0.0658, 0.0493, 0.0538, 0.0678, 0.0380, 0.0662, 0.0657, 0.0730], device='cuda:1'), in_proj_covar=tensor([0.0032, 0.0027, 0.0025, 0.0031, 0.0022, 0.0031, 0.0030, 0.0033], device='cuda:1'), out_proj_covar=tensor([0.0049, 0.0044, 0.0038, 0.0049, 0.0038, 0.0048, 0.0047, 0.0051], device='cuda:1') 2023-04-26 11:46:12,539 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=6354.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 11:46:39,285 INFO [finetune.py:976] (1/7) Epoch 2, batch 650, loss[loss=0.2978, simple_loss=0.3588, pruned_loss=0.1184, over 4921.00 frames. ], tot_loss[loss=0.2883, simple_loss=0.3292, pruned_loss=0.1238, over 918024.41 frames. ], batch size: 42, lr: 4.00e-03, grad_scale: 32.0 2023-04-26 11:46:51,864 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5211, 1.1438, 1.1468, 1.0629, 1.7047, 1.4718, 1.1026, 1.0785], device='cuda:1'), covar=tensor([0.1518, 0.1778, 0.2151, 0.1857, 0.0821, 0.1749, 0.2169, 0.2342], device='cuda:1'), in_proj_covar=tensor([0.0328, 0.0344, 0.0347, 0.0316, 0.0354, 0.0379, 0.0327, 0.0361], device='cuda:1'), out_proj_covar=tensor([7.1835e-05, 7.4028e-05, 7.5035e-05, 6.6300e-05, 7.5451e-05, 8.3038e-05, 7.1300e-05, 7.8013e-05], device='cuda:1') 2023-04-26 11:46:53,615 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6399.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:47:02,013 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5332, 1.7930, 1.3795, 1.7889, 1.4894, 1.4326, 1.7053, 1.1733], device='cuda:1'), covar=tensor([0.2121, 0.1703, 0.1415, 0.1566, 0.3157, 0.1696, 0.2047, 0.3008], device='cuda:1'), in_proj_covar=tensor([0.0313, 0.0334, 0.0244, 0.0309, 0.0312, 0.0283, 0.0280, 0.0300], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 11:47:02,618 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=6411.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:47:04,441 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-04-26 11:47:07,253 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.650e+02 2.377e+02 2.867e+02 3.488e+02 9.630e+02, threshold=5.734e+02, percent-clipped=2.0 2023-04-26 11:47:13,364 INFO [finetune.py:976] (1/7) Epoch 2, batch 700, loss[loss=0.3154, simple_loss=0.3494, pruned_loss=0.1407, over 4727.00 frames. ], tot_loss[loss=0.2913, simple_loss=0.3324, pruned_loss=0.1251, over 928328.13 frames. ], batch size: 59, lr: 4.00e-03, grad_scale: 32.0 2023-04-26 11:47:40,533 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([3.7585, 3.6989, 2.7896, 4.2889, 3.7387, 3.6779, 1.6102, 3.7720], device='cuda:1'), covar=tensor([0.1637, 0.1003, 0.2997, 0.1441, 0.2473, 0.1743, 0.5443, 0.1941], device='cuda:1'), in_proj_covar=tensor([0.0260, 0.0230, 0.0280, 0.0327, 0.0322, 0.0271, 0.0286, 0.0286], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-26 11:47:46,403 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=6459.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:47:46,484 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5594, 2.1803, 1.3724, 1.2886, 1.1644, 1.1636, 1.3213, 1.0865], device='cuda:1'), covar=tensor([0.2599, 0.2333, 0.3118, 0.3755, 0.4263, 0.3270, 0.2491, 0.3322], device='cuda:1'), in_proj_covar=tensor([0.0199, 0.0224, 0.0200, 0.0219, 0.0237, 0.0199, 0.0194, 0.0210], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-26 11:47:48,747 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=6462.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:47:51,711 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=6467.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:47:55,876 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4233, 2.3367, 1.8290, 2.0613, 2.3245, 1.9478, 2.9801, 1.6600], device='cuda:1'), covar=tensor([0.3862, 0.1549, 0.4367, 0.2807, 0.1877, 0.2416, 0.1334, 0.3620], device='cuda:1'), in_proj_covar=tensor([0.0336, 0.0345, 0.0428, 0.0361, 0.0397, 0.0367, 0.0392, 0.0402], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 11:47:58,778 INFO [finetune.py:976] (1/7) Epoch 2, batch 750, loss[loss=0.3021, simple_loss=0.3358, pruned_loss=0.1342, over 4838.00 frames. ], tot_loss[loss=0.2927, simple_loss=0.3338, pruned_loss=0.1258, over 934075.91 frames. ], batch size: 47, lr: 4.00e-03, grad_scale: 32.0 2023-04-26 11:48:18,127 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6491.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:48:20,023 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-04-26 11:48:43,006 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0026, 2.1233, 2.3302, 2.3886, 2.4544, 1.8239, 1.5208, 2.1115], device='cuda:1'), covar=tensor([0.1232, 0.0963, 0.0627, 0.0828, 0.0766, 0.1223, 0.1314, 0.0830], device='cuda:1'), in_proj_covar=tensor([0.0212, 0.0211, 0.0191, 0.0183, 0.0180, 0.0199, 0.0177, 0.0194], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 11:48:50,403 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.74 vs. limit=5.0 2023-04-26 11:48:50,919 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=6514.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:48:53,698 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.458e+02 2.088e+02 2.383e+02 2.881e+02 6.100e+02, threshold=4.765e+02, percent-clipped=1.0 2023-04-26 11:49:03,652 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=6523.0, num_to_drop=1, layers_to_drop={2} 2023-04-26 11:49:05,428 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6526.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:49:06,607 INFO [finetune.py:976] (1/7) Epoch 2, batch 800, loss[loss=0.2305, simple_loss=0.2824, pruned_loss=0.08927, over 4733.00 frames. ], tot_loss[loss=0.2917, simple_loss=0.3335, pruned_loss=0.125, over 939816.58 frames. ], batch size: 59, lr: 4.00e-03, grad_scale: 32.0 2023-04-26 11:49:06,733 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=6528.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:49:26,687 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6559.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:49:29,548 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=6562.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:49:33,149 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6567.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:49:35,011 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.1508, 2.5212, 0.8437, 1.4443, 1.4892, 1.9797, 1.6925, 0.9454], device='cuda:1'), covar=tensor([0.1536, 0.1163, 0.1969, 0.1420, 0.1210, 0.0891, 0.1559, 0.1574], device='cuda:1'), in_proj_covar=tensor([0.0125, 0.0276, 0.0153, 0.0134, 0.0145, 0.0167, 0.0131, 0.0134], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-26 11:49:40,229 INFO [finetune.py:976] (1/7) Epoch 2, batch 850, loss[loss=0.2774, simple_loss=0.3183, pruned_loss=0.1183, over 4813.00 frames. ], tot_loss[loss=0.2878, simple_loss=0.3299, pruned_loss=0.1228, over 943075.26 frames. ], batch size: 39, lr: 4.00e-03, grad_scale: 32.0 2023-04-26 11:50:18,614 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.341e+02 2.044e+02 2.405e+02 3.038e+02 6.612e+02, threshold=4.810e+02, percent-clipped=4.0 2023-04-26 11:50:21,183 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6622.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:50:25,267 INFO [finetune.py:976] (1/7) Epoch 2, batch 900, loss[loss=0.2163, simple_loss=0.2713, pruned_loss=0.08062, over 4764.00 frames. ], tot_loss[loss=0.2842, simple_loss=0.3264, pruned_loss=0.121, over 946403.07 frames. ], batch size: 27, lr: 4.00e-03, grad_scale: 32.0 2023-04-26 11:50:25,990 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=6629.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:50:36,415 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=6646.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:50:42,596 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.74 vs. limit=5.0 2023-04-26 11:50:58,408 INFO [finetune.py:976] (1/7) Epoch 2, batch 950, loss[loss=0.3217, simple_loss=0.3546, pruned_loss=0.1444, over 4932.00 frames. ], tot_loss[loss=0.2809, simple_loss=0.3232, pruned_loss=0.1193, over 948765.51 frames. ], batch size: 38, lr: 4.00e-03, grad_scale: 32.0 2023-04-26 11:51:06,474 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=6690.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:51:08,882 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=6694.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:51:11,964 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=6699.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:51:41,855 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.328e+02 2.213e+02 2.518e+02 3.022e+02 8.936e+02, threshold=5.037e+02, percent-clipped=4.0 2023-04-26 11:51:48,517 INFO [finetune.py:976] (1/7) Epoch 2, batch 1000, loss[loss=0.2351, simple_loss=0.2829, pruned_loss=0.09364, over 4912.00 frames. ], tot_loss[loss=0.2824, simple_loss=0.3242, pruned_loss=0.1203, over 952119.33 frames. ], batch size: 32, lr: 4.00e-03, grad_scale: 32.0 2023-04-26 11:51:48,614 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5557, 1.4062, 1.5232, 1.8397, 2.0023, 1.4109, 0.9302, 1.6506], device='cuda:1'), covar=tensor([0.1074, 0.1424, 0.1014, 0.0737, 0.0637, 0.1133, 0.1320, 0.0727], device='cuda:1'), in_proj_covar=tensor([0.0209, 0.0208, 0.0190, 0.0182, 0.0179, 0.0197, 0.0175, 0.0193], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 11:52:11,812 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=6747.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:52:55,093 INFO [finetune.py:976] (1/7) Epoch 2, batch 1050, loss[loss=0.2754, simple_loss=0.3203, pruned_loss=0.1153, over 4758.00 frames. ], tot_loss[loss=0.2848, simple_loss=0.3273, pruned_loss=0.1212, over 952955.90 frames. ], batch size: 59, lr: 4.00e-03, grad_scale: 32.0 2023-04-26 11:53:03,725 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=6791.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:53:04,450 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-04-26 11:53:33,222 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.642e+02 2.078e+02 2.516e+02 2.901e+02 4.864e+02, threshold=5.033e+02, percent-clipped=0.0 2023-04-26 11:53:33,306 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6818.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 11:53:42,744 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6823.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:53:44,644 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=6826.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:53:51,480 INFO [finetune.py:976] (1/7) Epoch 2, batch 1100, loss[loss=0.3049, simple_loss=0.3535, pruned_loss=0.1282, over 4723.00 frames. ], tot_loss[loss=0.2854, simple_loss=0.3287, pruned_loss=0.121, over 954321.02 frames. ], batch size: 59, lr: 4.00e-03, grad_scale: 32.0 2023-04-26 11:54:04,438 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=6839.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:54:06,404 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.0167, 0.7998, 1.1244, 1.2996, 1.2359, 1.0247, 1.0978, 1.0769], device='cuda:1'), covar=tensor([5.2059, 8.0106, 8.7176, 8.8701, 5.8615, 9.4318, 9.6791, 6.5614], device='cuda:1'), in_proj_covar=tensor([0.0439, 0.0504, 0.0586, 0.0569, 0.0469, 0.0525, 0.0532, 0.0542], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 11:54:16,610 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=6859.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:54:22,447 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=6867.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:54:27,626 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=6874.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:54:30,495 INFO [finetune.py:976] (1/7) Epoch 2, batch 1150, loss[loss=0.2642, simple_loss=0.3117, pruned_loss=0.1084, over 4891.00 frames. ], tot_loss[loss=0.2841, simple_loss=0.3277, pruned_loss=0.1202, over 954902.35 frames. ], batch size: 32, lr: 4.00e-03, grad_scale: 32.0 2023-04-26 11:54:30,617 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=6878.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 11:54:49,243 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=6907.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:54:50,547 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=6909.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:54:54,155 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=6915.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:54:56,914 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.447e+02 2.231e+02 2.530e+02 3.126e+02 7.008e+02, threshold=5.060e+02, percent-clipped=3.0 2023-04-26 11:54:59,919 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=6922.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:55:04,519 INFO [finetune.py:976] (1/7) Epoch 2, batch 1200, loss[loss=0.2533, simple_loss=0.2993, pruned_loss=0.1037, over 4741.00 frames. ], tot_loss[loss=0.2802, simple_loss=0.3244, pruned_loss=0.1181, over 956013.38 frames. ], batch size: 23, lr: 4.00e-03, grad_scale: 32.0 2023-04-26 11:55:11,765 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=6939.0, num_to_drop=1, layers_to_drop={3} 2023-04-26 11:55:31,686 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=6970.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:55:31,757 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=6970.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:55:38,016 INFO [finetune.py:976] (1/7) Epoch 2, batch 1250, loss[loss=0.2656, simple_loss=0.3111, pruned_loss=0.11, over 4823.00 frames. ], tot_loss[loss=0.2769, simple_loss=0.3209, pruned_loss=0.1165, over 957171.28 frames. ], batch size: 41, lr: 4.00e-03, grad_scale: 32.0 2023-04-26 11:55:41,700 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.1125, 2.2339, 2.0130, 1.9523, 2.4255, 1.8772, 2.9596, 1.7627], device='cuda:1'), covar=tensor([0.4788, 0.1665, 0.4421, 0.2935, 0.1776, 0.2594, 0.1128, 0.4043], device='cuda:1'), in_proj_covar=tensor([0.0336, 0.0345, 0.0429, 0.0361, 0.0396, 0.0367, 0.0392, 0.0403], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 11:55:43,399 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6985.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:55:52,567 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8411, 1.1763, 1.5297, 1.8792, 1.3956, 1.2028, 0.9178, 1.3242], device='cuda:1'), covar=tensor([0.7303, 0.8450, 0.4057, 0.9264, 0.8371, 0.6793, 1.1745, 0.8011], device='cuda:1'), in_proj_covar=tensor([0.0263, 0.0278, 0.0221, 0.0345, 0.0236, 0.0234, 0.0273, 0.0222], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-26 11:56:04,201 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.519e+02 2.106e+02 2.592e+02 3.021e+02 5.805e+02, threshold=5.184e+02, percent-clipped=1.0 2023-04-26 11:56:11,776 INFO [finetune.py:976] (1/7) Epoch 2, batch 1300, loss[loss=0.3104, simple_loss=0.3381, pruned_loss=0.1414, over 4821.00 frames. ], tot_loss[loss=0.2736, simple_loss=0.3172, pruned_loss=0.115, over 957692.07 frames. ], batch size: 38, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 11:56:21,376 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.46 vs. limit=5.0 2023-04-26 11:56:25,522 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=7048.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 11:56:48,864 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.1195, 1.0013, 1.1800, 1.1537, 1.0469, 0.8911, 0.9639, 0.6927], device='cuda:1'), covar=tensor([0.0866, 0.0892, 0.0810, 0.0887, 0.1017, 0.1655, 0.0688, 0.1338], device='cuda:1'), in_proj_covar=tensor([0.0072, 0.0080, 0.0076, 0.0075, 0.0088, 0.0096, 0.0090, 0.0081], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003], device='cuda:1') 2023-04-26 11:56:55,868 INFO [finetune.py:976] (1/7) Epoch 2, batch 1350, loss[loss=0.2221, simple_loss=0.2809, pruned_loss=0.08165, over 4735.00 frames. ], tot_loss[loss=0.2736, simple_loss=0.3168, pruned_loss=0.1151, over 957094.25 frames. ], batch size: 23, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 11:57:03,757 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.69 vs. limit=5.0 2023-04-26 11:57:28,638 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=7109.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 11:57:39,841 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.388e+02 2.079e+02 2.494e+02 3.022e+02 7.754e+02, threshold=4.988e+02, percent-clipped=2.0 2023-04-26 11:57:39,945 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=7118.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 11:57:48,851 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=7123.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:57:51,916 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.49 vs. limit=2.0 2023-04-26 11:57:56,136 INFO [finetune.py:976] (1/7) Epoch 2, batch 1400, loss[loss=0.2521, simple_loss=0.3086, pruned_loss=0.09774, over 4899.00 frames. ], tot_loss[loss=0.2781, simple_loss=0.3216, pruned_loss=0.1173, over 958664.79 frames. ], batch size: 36, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 11:58:36,327 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=7163.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:58:38,105 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=7166.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:58:47,202 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=7171.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:58:51,417 INFO [finetune.py:976] (1/7) Epoch 2, batch 1450, loss[loss=0.3256, simple_loss=0.3641, pruned_loss=0.1435, over 4855.00 frames. ], tot_loss[loss=0.2807, simple_loss=0.3247, pruned_loss=0.1183, over 959074.06 frames. ], batch size: 44, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 11:59:18,333 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.1680, 0.9747, 1.1857, 1.4143, 1.3469, 1.0915, 1.1290, 1.0934], device='cuda:1'), covar=tensor([4.9229, 7.6340, 8.2896, 8.6058, 5.5410, 8.9564, 9.0908, 6.6420], device='cuda:1'), in_proj_covar=tensor([0.0445, 0.0509, 0.0593, 0.0579, 0.0475, 0.0531, 0.0538, 0.0548], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 11:59:19,364 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.461e+02 2.205e+02 2.539e+02 2.992e+02 8.732e+02, threshold=5.079e+02, percent-clipped=2.0 2023-04-26 11:59:23,175 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=7224.0, num_to_drop=1, layers_to_drop={3} 2023-04-26 11:59:25,497 INFO [finetune.py:976] (1/7) Epoch 2, batch 1500, loss[loss=0.2707, simple_loss=0.3182, pruned_loss=0.1116, over 4901.00 frames. ], tot_loss[loss=0.2808, simple_loss=0.3256, pruned_loss=0.118, over 959752.86 frames. ], batch size: 36, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 11:59:29,688 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=7234.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 11:59:51,523 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=7265.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:59:57,742 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=7275.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:59:59,532 INFO [finetune.py:976] (1/7) Epoch 2, batch 1550, loss[loss=0.2586, simple_loss=0.3111, pruned_loss=0.1031, over 4783.00 frames. ], tot_loss[loss=0.2807, simple_loss=0.3258, pruned_loss=0.1178, over 958394.28 frames. ], batch size: 26, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 12:00:04,483 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=7285.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:00:11,019 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5398, 1.2866, 1.7704, 1.6292, 1.3527, 1.1766, 1.3719, 0.8748], device='cuda:1'), covar=tensor([0.0841, 0.1068, 0.0615, 0.1089, 0.1348, 0.1640, 0.0978, 0.1406], device='cuda:1'), in_proj_covar=tensor([0.0072, 0.0080, 0.0076, 0.0076, 0.0089, 0.0096, 0.0090, 0.0081], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003], device='cuda:1') 2023-04-26 12:00:27,651 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.588e+02 2.228e+02 2.487e+02 3.016e+02 5.047e+02, threshold=4.974e+02, percent-clipped=0.0 2023-04-26 12:00:33,829 INFO [finetune.py:976] (1/7) Epoch 2, batch 1600, loss[loss=0.2666, simple_loss=0.2985, pruned_loss=0.1174, over 4277.00 frames. ], tot_loss[loss=0.277, simple_loss=0.3217, pruned_loss=0.1162, over 956933.92 frames. ], batch size: 18, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 12:00:36,965 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=7333.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:00:38,886 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=7336.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:00:57,644 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8144, 2.4328, 1.6695, 1.5746, 1.3322, 1.3932, 1.6567, 1.2515], device='cuda:1'), covar=tensor([0.2567, 0.2339, 0.2833, 0.3583, 0.4001, 0.3009, 0.2259, 0.3406], device='cuda:1'), in_proj_covar=tensor([0.0200, 0.0224, 0.0199, 0.0220, 0.0237, 0.0198, 0.0194, 0.0211], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-26 12:01:06,050 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.1136, 0.9035, 1.1289, 1.3657, 1.2793, 1.0635, 1.1292, 1.1256], device='cuda:1'), covar=tensor([ 5.5275, 8.4501, 9.4021, 9.5122, 6.3424, 10.1724, 9.7214, 7.1119], device='cuda:1'), in_proj_covar=tensor([0.0443, 0.0508, 0.0592, 0.0579, 0.0475, 0.0529, 0.0536, 0.0547], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 12:01:07,694 INFO [finetune.py:976] (1/7) Epoch 2, batch 1650, loss[loss=0.2478, simple_loss=0.3009, pruned_loss=0.09736, over 4922.00 frames. ], tot_loss[loss=0.2747, simple_loss=0.3187, pruned_loss=0.1153, over 957856.54 frames. ], batch size: 43, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 12:01:18,398 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.1479, 1.6399, 1.6770, 1.5612, 2.2093, 1.9864, 1.5983, 1.6886], device='cuda:1'), covar=tensor([0.1059, 0.1516, 0.1948, 0.1630, 0.0908, 0.1409, 0.1820, 0.1954], device='cuda:1'), in_proj_covar=tensor([0.0327, 0.0341, 0.0346, 0.0316, 0.0352, 0.0374, 0.0326, 0.0357], device='cuda:1'), out_proj_covar=tensor([7.1664e-05, 7.3398e-05, 7.4671e-05, 6.6257e-05, 7.4997e-05, 8.1965e-05, 7.0939e-05, 7.7144e-05], device='cuda:1') 2023-04-26 12:01:25,937 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=7404.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 12:01:28,257 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=7407.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:01:34,835 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.570e+02 2.142e+02 2.529e+02 2.973e+02 5.158e+02, threshold=5.058e+02, percent-clipped=1.0 2023-04-26 12:01:41,457 INFO [finetune.py:976] (1/7) Epoch 2, batch 1700, loss[loss=0.2875, simple_loss=0.3236, pruned_loss=0.1257, over 4755.00 frames. ], tot_loss[loss=0.2717, simple_loss=0.3151, pruned_loss=0.1141, over 954579.98 frames. ], batch size: 27, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 12:01:57,269 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.78 vs. limit=5.0 2023-04-26 12:02:09,779 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=7468.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:02:22,154 INFO [finetune.py:976] (1/7) Epoch 2, batch 1750, loss[loss=0.3196, simple_loss=0.3609, pruned_loss=0.1392, over 4809.00 frames. ], tot_loss[loss=0.2734, simple_loss=0.3172, pruned_loss=0.1148, over 953327.55 frames. ], batch size: 45, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 12:02:33,562 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6372, 1.1363, 4.4290, 4.1181, 3.9043, 4.1294, 4.1068, 3.8865], device='cuda:1'), covar=tensor([0.6985, 0.6157, 0.0958, 0.1842, 0.1101, 0.1563, 0.1597, 0.1518], device='cuda:1'), in_proj_covar=tensor([0.0338, 0.0316, 0.0453, 0.0460, 0.0379, 0.0437, 0.0346, 0.0403], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0002], device='cuda:1') 2023-04-26 12:02:45,342 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.2111, 2.2219, 2.4228, 2.7493, 2.5453, 2.0010, 1.4994, 2.1226], device='cuda:1'), covar=tensor([0.1361, 0.0987, 0.0655, 0.0826, 0.0791, 0.1290, 0.1454, 0.0923], device='cuda:1'), in_proj_covar=tensor([0.0210, 0.0208, 0.0189, 0.0182, 0.0179, 0.0197, 0.0175, 0.0193], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 12:02:54,657 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5700, 2.3520, 1.4122, 1.4088, 1.1278, 1.2095, 1.4709, 1.0272], device='cuda:1'), covar=tensor([0.2698, 0.2306, 0.3014, 0.3713, 0.4339, 0.3103, 0.2469, 0.3485], device='cuda:1'), in_proj_covar=tensor([0.0200, 0.0225, 0.0199, 0.0220, 0.0237, 0.0198, 0.0194, 0.0210], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-26 12:03:04,396 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5805, 2.0300, 1.3400, 1.1972, 1.1547, 1.1911, 1.3334, 1.0523], device='cuda:1'), covar=tensor([0.2488, 0.2124, 0.2923, 0.3492, 0.4021, 0.2948, 0.2318, 0.3319], device='cuda:1'), in_proj_covar=tensor([0.0200, 0.0225, 0.0199, 0.0220, 0.0237, 0.0198, 0.0194, 0.0210], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-26 12:03:11,384 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.401e+02 2.126e+02 2.582e+02 2.994e+02 4.895e+02, threshold=5.165e+02, percent-clipped=0.0 2023-04-26 12:03:12,084 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=7519.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 12:03:23,644 INFO [finetune.py:976] (1/7) Epoch 2, batch 1800, loss[loss=0.2607, simple_loss=0.3036, pruned_loss=0.1089, over 4767.00 frames. ], tot_loss[loss=0.2743, simple_loss=0.3191, pruned_loss=0.1147, over 953022.48 frames. ], batch size: 28, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 12:03:26,823 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=2.00 vs. limit=2.0 2023-04-26 12:03:27,309 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=7534.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 12:04:12,445 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=7565.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:04:31,078 INFO [finetune.py:976] (1/7) Epoch 2, batch 1850, loss[loss=0.2545, simple_loss=0.3101, pruned_loss=0.09941, over 4921.00 frames. ], tot_loss[loss=0.2768, simple_loss=0.3219, pruned_loss=0.1158, over 952969.69 frames. ], batch size: 33, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 12:04:31,896 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-04-26 12:04:34,064 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=7582.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 12:05:07,100 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=7613.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:05:10,536 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.515e+02 2.096e+02 2.728e+02 3.385e+02 6.441e+02, threshold=5.455e+02, percent-clipped=3.0 2023-04-26 12:05:14,368 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=7624.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:05:16,722 INFO [finetune.py:976] (1/7) Epoch 2, batch 1900, loss[loss=0.3323, simple_loss=0.3766, pruned_loss=0.144, over 4899.00 frames. ], tot_loss[loss=0.2765, simple_loss=0.3224, pruned_loss=0.1153, over 951495.80 frames. ], batch size: 37, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 12:05:18,626 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=7631.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:05:50,037 INFO [finetune.py:976] (1/7) Epoch 2, batch 1950, loss[loss=0.2992, simple_loss=0.323, pruned_loss=0.1377, over 4835.00 frames. ], tot_loss[loss=0.2749, simple_loss=0.3212, pruned_loss=0.1143, over 954095.79 frames. ], batch size: 47, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 12:05:54,909 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=7685.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:06:06,517 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=7704.0, num_to_drop=1, layers_to_drop={2} 2023-04-26 12:06:17,408 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.610e+02 2.113e+02 2.438e+02 2.790e+02 7.164e+02, threshold=4.875e+02, percent-clipped=1.0 2023-04-26 12:06:23,982 INFO [finetune.py:976] (1/7) Epoch 2, batch 2000, loss[loss=0.2628, simple_loss=0.2984, pruned_loss=0.1136, over 4774.00 frames. ], tot_loss[loss=0.2727, simple_loss=0.3183, pruned_loss=0.1135, over 953340.39 frames. ], batch size: 28, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 12:06:24,232 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 2023-04-26 12:06:33,721 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=7743.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:06:39,112 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=7752.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 12:06:46,793 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=7763.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:06:57,780 INFO [finetune.py:976] (1/7) Epoch 2, batch 2050, loss[loss=0.2342, simple_loss=0.2924, pruned_loss=0.08795, over 4774.00 frames. ], tot_loss[loss=0.2691, simple_loss=0.3147, pruned_loss=0.1118, over 956019.02 frames. ], batch size: 28, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 12:07:11,311 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.47 vs. limit=2.0 2023-04-26 12:07:13,069 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=7802.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:07:14,344 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=7804.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:07:24,635 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.397e+02 2.038e+02 2.548e+02 3.001e+02 7.131e+02, threshold=5.096e+02, percent-clipped=2.0 2023-04-26 12:07:25,351 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=7819.0, num_to_drop=1, layers_to_drop={2} 2023-04-26 12:07:31,147 INFO [finetune.py:976] (1/7) Epoch 2, batch 2100, loss[loss=0.2423, simple_loss=0.2906, pruned_loss=0.09698, over 4828.00 frames. ], tot_loss[loss=0.2698, simple_loss=0.3141, pruned_loss=0.1127, over 955366.85 frames. ], batch size: 25, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 12:07:34,381 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6633, 2.0567, 1.5910, 1.9278, 1.5064, 1.4529, 1.7374, 1.3885], device='cuda:1'), covar=tensor([0.2077, 0.1751, 0.1498, 0.1705, 0.3420, 0.1986, 0.2131, 0.3008], device='cuda:1'), in_proj_covar=tensor([0.0313, 0.0333, 0.0244, 0.0310, 0.0312, 0.0284, 0.0278, 0.0300], device='cuda:1'), out_proj_covar=tensor([1.2802e-04, 1.3666e-04, 9.9988e-05, 1.2557e-04, 1.2927e-04, 1.1464e-04, 1.1512e-04, 1.2201e-04], device='cuda:1') 2023-04-26 12:07:54,052 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=7863.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:07:56,431 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=7867.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:08:15,064 INFO [finetune.py:976] (1/7) Epoch 2, batch 2150, loss[loss=0.2749, simple_loss=0.3308, pruned_loss=0.1095, over 4837.00 frames. ], tot_loss[loss=0.2721, simple_loss=0.3172, pruned_loss=0.1135, over 953036.35 frames. ], batch size: 47, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 12:09:00,309 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6716, 2.0844, 1.6240, 1.9667, 1.4944, 1.5869, 1.7340, 1.4170], device='cuda:1'), covar=tensor([0.2082, 0.1475, 0.1382, 0.1515, 0.3199, 0.1704, 0.2127, 0.2873], device='cuda:1'), in_proj_covar=tensor([0.0317, 0.0337, 0.0247, 0.0313, 0.0316, 0.0287, 0.0281, 0.0304], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 12:09:02,672 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.204e+02 2.251e+02 2.703e+02 3.381e+02 5.777e+02, threshold=5.406e+02, percent-clipped=2.0 2023-04-26 12:09:20,982 INFO [finetune.py:976] (1/7) Epoch 2, batch 2200, loss[loss=0.2274, simple_loss=0.2773, pruned_loss=0.08869, over 4777.00 frames. ], tot_loss[loss=0.274, simple_loss=0.3195, pruned_loss=0.1142, over 952634.62 frames. ], batch size: 26, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 12:09:22,953 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=7931.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:09:24,325 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.49 vs. limit=2.0 2023-04-26 12:10:05,324 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4005, 3.3548, 0.8391, 1.8564, 1.8158, 2.4327, 1.9770, 1.0182], device='cuda:1'), covar=tensor([0.1374, 0.0908, 0.2179, 0.1281, 0.1158, 0.1126, 0.1605, 0.2161], device='cuda:1'), in_proj_covar=tensor([0.0126, 0.0276, 0.0154, 0.0134, 0.0146, 0.0169, 0.0131, 0.0135], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-26 12:10:21,348 INFO [finetune.py:976] (1/7) Epoch 2, batch 2250, loss[loss=0.2989, simple_loss=0.3395, pruned_loss=0.1291, over 4268.00 frames. ], tot_loss[loss=0.2754, simple_loss=0.321, pruned_loss=0.1148, over 949815.76 frames. ], batch size: 65, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 12:10:22,014 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=7979.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:10:23,120 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=7980.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:10:47,701 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.449e+02 2.028e+02 2.420e+02 2.875e+02 5.113e+02, threshold=4.840e+02, percent-clipped=0.0 2023-04-26 12:10:55,295 INFO [finetune.py:976] (1/7) Epoch 2, batch 2300, loss[loss=0.2871, simple_loss=0.3201, pruned_loss=0.1271, over 4198.00 frames. ], tot_loss[loss=0.2743, simple_loss=0.3208, pruned_loss=0.1138, over 948298.35 frames. ], batch size: 65, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 12:11:00,445 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.1426, 1.6010, 1.3787, 1.8070, 1.7217, 2.1322, 1.4856, 3.6052], device='cuda:1'), covar=tensor([0.0746, 0.0772, 0.0842, 0.1268, 0.0656, 0.0547, 0.0752, 0.0151], device='cuda:1'), in_proj_covar=tensor([0.0041, 0.0041, 0.0042, 0.0046, 0.0042, 0.0041, 0.0041, 0.0066], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0013, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0017], device='cuda:1') 2023-04-26 12:11:01,703 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.2137, 1.4678, 1.0517, 1.4300, 1.3692, 1.0286, 1.2630, 0.8889], device='cuda:1'), covar=tensor([0.2296, 0.1650, 0.1701, 0.1649, 0.3646, 0.1935, 0.2339, 0.3091], device='cuda:1'), in_proj_covar=tensor([0.0318, 0.0338, 0.0248, 0.0313, 0.0316, 0.0287, 0.0282, 0.0304], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 12:11:01,732 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5652, 1.1025, 1.3092, 1.4172, 1.1634, 1.0224, 0.6517, 0.9551], device='cuda:1'), covar=tensor([0.7220, 0.8203, 0.3880, 0.6891, 0.8537, 0.5970, 1.1048, 0.8148], device='cuda:1'), in_proj_covar=tensor([0.0271, 0.0283, 0.0226, 0.0352, 0.0241, 0.0239, 0.0277, 0.0225], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-26 12:11:18,730 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=8063.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:11:28,370 INFO [finetune.py:976] (1/7) Epoch 2, batch 2350, loss[loss=0.2495, simple_loss=0.3005, pruned_loss=0.09921, over 4846.00 frames. ], tot_loss[loss=0.2729, simple_loss=0.3193, pruned_loss=0.1133, over 952191.20 frames. ], batch size: 49, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 12:11:42,662 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=8099.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:11:50,387 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=8111.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:11:51,701 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.2175, 2.7764, 1.1539, 1.3631, 2.1750, 1.2742, 3.5579, 1.8393], device='cuda:1'), covar=tensor([0.0711, 0.0603, 0.0889, 0.1364, 0.0550, 0.1078, 0.0200, 0.0698], device='cuda:1'), in_proj_covar=tensor([0.0058, 0.0074, 0.0055, 0.0051, 0.0056, 0.0057, 0.0088, 0.0055], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0008, 0.0008, 0.0008, 0.0008, 0.0012, 0.0008], device='cuda:1') 2023-04-26 12:11:54,656 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.190e+02 2.063e+02 2.510e+02 2.746e+02 4.758e+02, threshold=5.020e+02, percent-clipped=0.0 2023-04-26 12:12:01,654 INFO [finetune.py:976] (1/7) Epoch 2, batch 2400, loss[loss=0.2373, simple_loss=0.296, pruned_loss=0.08934, over 4866.00 frames. ], tot_loss[loss=0.2668, simple_loss=0.313, pruned_loss=0.1103, over 950694.76 frames. ], batch size: 31, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 12:12:21,842 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=8158.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:12:34,354 INFO [finetune.py:976] (1/7) Epoch 2, batch 2450, loss[loss=0.23, simple_loss=0.2786, pruned_loss=0.09064, over 4814.00 frames. ], tot_loss[loss=0.2617, simple_loss=0.308, pruned_loss=0.1077, over 951223.33 frames. ], batch size: 51, lr: 3.99e-03, grad_scale: 64.0 2023-04-26 12:12:57,201 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.1016, 1.3155, 1.2560, 1.6083, 1.4446, 1.6174, 1.3140, 2.5066], device='cuda:1'), covar=tensor([0.0624, 0.0757, 0.0800, 0.1258, 0.0668, 0.0509, 0.0730, 0.0249], device='cuda:1'), in_proj_covar=tensor([0.0040, 0.0041, 0.0042, 0.0046, 0.0042, 0.0041, 0.0041, 0.0065], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0017], device='cuda:1') 2023-04-26 12:13:01,853 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.405e+02 2.093e+02 2.421e+02 2.814e+02 5.190e+02, threshold=4.843e+02, percent-clipped=1.0 2023-04-26 12:13:07,998 INFO [finetune.py:976] (1/7) Epoch 2, batch 2500, loss[loss=0.2808, simple_loss=0.3429, pruned_loss=0.1093, over 4896.00 frames. ], tot_loss[loss=0.2655, simple_loss=0.311, pruned_loss=0.11, over 952774.81 frames. ], batch size: 43, lr: 3.99e-03, grad_scale: 64.0 2023-04-26 12:13:49,849 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=8265.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:14:03,521 INFO [finetune.py:976] (1/7) Epoch 2, batch 2550, loss[loss=0.2758, simple_loss=0.3015, pruned_loss=0.125, over 4104.00 frames. ], tot_loss[loss=0.2677, simple_loss=0.3142, pruned_loss=0.1106, over 951446.64 frames. ], batch size: 17, lr: 3.99e-03, grad_scale: 64.0 2023-04-26 12:14:04,876 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=8280.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:14:15,602 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-04-26 12:14:27,239 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 2023-04-26 12:14:46,726 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-04-26 12:14:58,607 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4353, 1.1772, 1.7900, 1.5930, 1.2776, 1.0321, 1.3342, 1.0371], device='cuda:1'), covar=tensor([0.0883, 0.0935, 0.0526, 0.0852, 0.1004, 0.1482, 0.0810, 0.1023], device='cuda:1'), in_proj_covar=tensor([0.0071, 0.0079, 0.0075, 0.0074, 0.0087, 0.0095, 0.0089, 0.0080], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003], device='cuda:1') 2023-04-26 12:15:02,135 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.269e+02 2.032e+02 2.420e+02 2.910e+02 5.283e+02, threshold=4.841e+02, percent-clipped=1.0 2023-04-26 12:15:13,540 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=8326.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:15:20,195 INFO [finetune.py:976] (1/7) Epoch 2, batch 2600, loss[loss=0.3007, simple_loss=0.3538, pruned_loss=0.1238, over 4827.00 frames. ], tot_loss[loss=0.2698, simple_loss=0.3168, pruned_loss=0.1114, over 950605.13 frames. ], batch size: 39, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 12:15:20,257 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=8328.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:15:52,667 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8634, 1.5681, 1.8194, 2.2385, 2.3281, 1.7818, 1.4438, 1.9030], device='cuda:1'), covar=tensor([0.1225, 0.1505, 0.0979, 0.0761, 0.0640, 0.1160, 0.1277, 0.0849], device='cuda:1'), in_proj_covar=tensor([0.0213, 0.0211, 0.0191, 0.0184, 0.0181, 0.0200, 0.0178, 0.0195], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 12:15:54,661 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-04-26 12:15:57,060 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-04-26 12:16:23,380 INFO [finetune.py:976] (1/7) Epoch 2, batch 2650, loss[loss=0.2357, simple_loss=0.3036, pruned_loss=0.08394, over 4838.00 frames. ], tot_loss[loss=0.2709, simple_loss=0.3185, pruned_loss=0.1116, over 953198.65 frames. ], batch size: 49, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 12:16:37,902 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6390, 1.5640, 1.6993, 1.9154, 1.8936, 1.6284, 1.3169, 1.7653], device='cuda:1'), covar=tensor([0.0990, 0.1085, 0.0657, 0.0600, 0.0592, 0.0866, 0.1038, 0.0558], device='cuda:1'), in_proj_covar=tensor([0.0213, 0.0212, 0.0191, 0.0184, 0.0181, 0.0200, 0.0178, 0.0195], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 12:16:49,116 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=8399.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:17:08,779 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.2953, 1.1053, 1.3744, 1.4842, 1.4498, 1.2468, 1.3061, 1.2541], device='cuda:1'), covar=tensor([4.4619, 6.5338, 7.4639, 7.0582, 4.4275, 8.0751, 7.9719, 6.3599], device='cuda:1'), in_proj_covar=tensor([0.0445, 0.0509, 0.0594, 0.0585, 0.0477, 0.0527, 0.0535, 0.0547], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 12:17:11,316 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-04-26 12:17:13,447 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.592e+02 2.071e+02 2.411e+02 3.038e+02 9.260e+02, threshold=4.823e+02, percent-clipped=4.0 2023-04-26 12:17:19,410 INFO [finetune.py:976] (1/7) Epoch 2, batch 2700, loss[loss=0.2461, simple_loss=0.2934, pruned_loss=0.09942, over 4175.00 frames. ], tot_loss[loss=0.2685, simple_loss=0.3163, pruned_loss=0.1104, over 952028.81 frames. ], batch size: 65, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 12:17:25,737 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.74 vs. limit=5.0 2023-04-26 12:17:29,035 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4899, 1.3832, 4.1543, 3.8804, 3.7234, 3.9733, 3.9313, 3.7504], device='cuda:1'), covar=tensor([0.6702, 0.5408, 0.0929, 0.1592, 0.0985, 0.1624, 0.1300, 0.1268], device='cuda:1'), in_proj_covar=tensor([0.0338, 0.0317, 0.0451, 0.0460, 0.0380, 0.0437, 0.0344, 0.0402], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0002], device='cuda:1') 2023-04-26 12:17:32,538 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=8447.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:17:40,753 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=8458.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:17:53,352 INFO [finetune.py:976] (1/7) Epoch 2, batch 2750, loss[loss=0.2719, simple_loss=0.3093, pruned_loss=0.1172, over 4874.00 frames. ], tot_loss[loss=0.2665, simple_loss=0.3132, pruned_loss=0.1099, over 951723.57 frames. ], batch size: 34, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 12:18:05,286 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5773, 1.7446, 1.8347, 1.9030, 1.9602, 1.5604, 1.1267, 1.7230], device='cuda:1'), covar=tensor([0.1155, 0.1136, 0.0744, 0.0742, 0.0624, 0.0971, 0.1175, 0.0676], device='cuda:1'), in_proj_covar=tensor([0.0211, 0.0211, 0.0190, 0.0184, 0.0180, 0.0199, 0.0177, 0.0194], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 12:18:12,775 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=8506.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:18:16,324 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8398, 2.0613, 1.1206, 1.4706, 2.3606, 1.7592, 1.6258, 1.7700], device='cuda:1'), covar=tensor([0.0619, 0.0444, 0.0441, 0.0651, 0.0275, 0.0608, 0.0598, 0.0715], device='cuda:1'), in_proj_covar=tensor([0.0032, 0.0027, 0.0025, 0.0031, 0.0022, 0.0031, 0.0030, 0.0033], device='cuda:1'), out_proj_covar=tensor([0.0048, 0.0044, 0.0039, 0.0049, 0.0037, 0.0048, 0.0047, 0.0051], device='cuda:1') 2023-04-26 12:18:21,016 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.402e+02 1.950e+02 2.341e+02 2.831e+02 5.180e+02, threshold=4.681e+02, percent-clipped=1.0 2023-04-26 12:18:26,542 INFO [finetune.py:976] (1/7) Epoch 2, batch 2800, loss[loss=0.2163, simple_loss=0.2756, pruned_loss=0.07846, over 4826.00 frames. ], tot_loss[loss=0.2624, simple_loss=0.3089, pruned_loss=0.1079, over 951275.26 frames. ], batch size: 41, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 12:19:00,280 INFO [finetune.py:976] (1/7) Epoch 2, batch 2850, loss[loss=0.2812, simple_loss=0.3213, pruned_loss=0.1206, over 4912.00 frames. ], tot_loss[loss=0.2606, simple_loss=0.3072, pruned_loss=0.107, over 952259.09 frames. ], batch size: 37, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 12:19:45,300 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.328e+02 2.166e+02 2.478e+02 2.838e+02 4.479e+02, threshold=4.956e+02, percent-clipped=0.0 2023-04-26 12:19:52,097 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=8621.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:19:56,364 INFO [finetune.py:976] (1/7) Epoch 2, batch 2900, loss[loss=0.2199, simple_loss=0.2618, pruned_loss=0.08896, over 3853.00 frames. ], tot_loss[loss=0.263, simple_loss=0.3098, pruned_loss=0.1081, over 952256.87 frames. ], batch size: 16, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 12:21:07,306 INFO [finetune.py:976] (1/7) Epoch 2, batch 2950, loss[loss=0.3098, simple_loss=0.3637, pruned_loss=0.1279, over 4813.00 frames. ], tot_loss[loss=0.2676, simple_loss=0.3151, pruned_loss=0.11, over 953986.90 frames. ], batch size: 40, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 12:22:03,514 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.340e+02 2.015e+02 2.424e+02 2.946e+02 5.788e+02, threshold=4.848e+02, percent-clipped=1.0 2023-04-26 12:22:15,450 INFO [finetune.py:976] (1/7) Epoch 2, batch 3000, loss[loss=0.2932, simple_loss=0.3294, pruned_loss=0.1285, over 4163.00 frames. ], tot_loss[loss=0.2679, simple_loss=0.3158, pruned_loss=0.11, over 953409.47 frames. ], batch size: 65, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 12:22:15,450 INFO [finetune.py:1001] (1/7) Computing validation loss 2023-04-26 12:22:22,972 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4465, 1.0300, 1.1963, 1.0758, 1.6430, 1.3609, 1.0385, 1.1977], device='cuda:1'), covar=tensor([0.1921, 0.1881, 0.2754, 0.2057, 0.1058, 0.1870, 0.2125, 0.2276], device='cuda:1'), in_proj_covar=tensor([0.0328, 0.0342, 0.0351, 0.0318, 0.0355, 0.0376, 0.0326, 0.0360], device='cuda:1'), out_proj_covar=tensor([7.1600e-05, 7.3557e-05, 7.5896e-05, 6.6750e-05, 7.5650e-05, 8.2187e-05, 7.0993e-05, 7.7677e-05], device='cuda:1') 2023-04-26 12:22:24,533 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4688, 1.0305, 1.1736, 1.0788, 1.6365, 1.3648, 1.0351, 1.1755], device='cuda:1'), covar=tensor([0.1850, 0.1876, 0.2604, 0.1939, 0.1278, 0.1779, 0.2765, 0.2292], device='cuda:1'), in_proj_covar=tensor([0.0328, 0.0342, 0.0351, 0.0318, 0.0355, 0.0376, 0.0326, 0.0360], device='cuda:1'), out_proj_covar=tensor([7.1600e-05, 7.3557e-05, 7.5896e-05, 6.6750e-05, 7.5650e-05, 8.2187e-05, 7.0993e-05, 7.7677e-05], device='cuda:1') 2023-04-26 12:22:32,547 INFO [finetune.py:1010] (1/7) Epoch 2, validation: loss=0.1863, simple_loss=0.2571, pruned_loss=0.0578, over 2265189.00 frames. 2023-04-26 12:22:32,547 INFO [finetune.py:1011] (1/7) Maximum memory allocated so far is 6069MB 2023-04-26 12:22:42,126 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9428, 1.9194, 1.3488, 1.6656, 2.0476, 1.8471, 1.7428, 1.8529], device='cuda:1'), covar=tensor([0.0519, 0.0400, 0.0409, 0.0570, 0.0281, 0.0555, 0.0524, 0.0594], device='cuda:1'), in_proj_covar=tensor([0.0032, 0.0027, 0.0025, 0.0031, 0.0022, 0.0031, 0.0030, 0.0032], device='cuda:1'), out_proj_covar=tensor([0.0048, 0.0044, 0.0039, 0.0049, 0.0037, 0.0048, 0.0047, 0.0051], device='cuda:1') 2023-04-26 12:23:09,086 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=8760.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:23:21,913 INFO [finetune.py:976] (1/7) Epoch 2, batch 3050, loss[loss=0.2588, simple_loss=0.3146, pruned_loss=0.1015, over 4825.00 frames. ], tot_loss[loss=0.2687, simple_loss=0.3169, pruned_loss=0.1102, over 951607.20 frames. ], batch size: 39, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 12:23:36,752 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4644, 2.3126, 2.0324, 2.2000, 2.4367, 1.9275, 3.2859, 1.6717], device='cuda:1'), covar=tensor([0.4702, 0.2404, 0.4954, 0.3934, 0.2436, 0.3270, 0.1406, 0.4826], device='cuda:1'), in_proj_covar=tensor([0.0347, 0.0353, 0.0438, 0.0371, 0.0406, 0.0377, 0.0401, 0.0413], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 12:23:37,889 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-04-26 12:23:49,150 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.298e+02 2.073e+02 2.370e+02 3.091e+02 5.908e+02, threshold=4.740e+02, percent-clipped=3.0 2023-04-26 12:23:51,000 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=8821.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:23:56,176 INFO [finetune.py:976] (1/7) Epoch 2, batch 3100, loss[loss=0.2478, simple_loss=0.2982, pruned_loss=0.09867, over 4771.00 frames. ], tot_loss[loss=0.2653, simple_loss=0.314, pruned_loss=0.1084, over 953597.02 frames. ], batch size: 29, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 12:24:29,315 INFO [finetune.py:976] (1/7) Epoch 2, batch 3150, loss[loss=0.2448, simple_loss=0.3012, pruned_loss=0.09414, over 4797.00 frames. ], tot_loss[loss=0.2633, simple_loss=0.3109, pruned_loss=0.1078, over 954678.78 frames. ], batch size: 29, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 12:24:47,494 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.66 vs. limit=5.0 2023-04-26 12:24:56,261 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.408e+02 2.116e+02 2.518e+02 3.010e+02 6.088e+02, threshold=5.037e+02, percent-clipped=1.0 2023-04-26 12:24:57,598 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=8921.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:25:01,165 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0850, 1.5049, 1.3266, 1.8468, 2.1896, 1.8254, 1.7081, 1.4919], device='cuda:1'), covar=tensor([0.2327, 0.2289, 0.2748, 0.2641, 0.1309, 0.2354, 0.2831, 0.2507], device='cuda:1'), in_proj_covar=tensor([0.0326, 0.0340, 0.0348, 0.0317, 0.0353, 0.0373, 0.0325, 0.0357], device='cuda:1'), out_proj_covar=tensor([7.1291e-05, 7.3127e-05, 7.5273e-05, 6.6416e-05, 7.5144e-05, 8.1642e-05, 7.0839e-05, 7.7204e-05], device='cuda:1') 2023-04-26 12:25:01,743 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=8927.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:25:02,221 INFO [finetune.py:976] (1/7) Epoch 2, batch 3200, loss[loss=0.28, simple_loss=0.3238, pruned_loss=0.1181, over 4842.00 frames. ], tot_loss[loss=0.2596, simple_loss=0.3071, pruned_loss=0.1061, over 955371.89 frames. ], batch size: 47, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 12:25:09,929 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.1665, 1.3450, 1.3066, 1.4020, 1.4334, 1.5803, 1.4101, 1.4071], device='cuda:1'), covar=tensor([1.8604, 4.3278, 3.3341, 2.8183, 3.0086, 4.9471, 4.1709, 3.4413], device='cuda:1'), in_proj_covar=tensor([0.0279, 0.0379, 0.0303, 0.0306, 0.0334, 0.0369, 0.0366, 0.0332], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-26 12:25:13,535 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4524, 1.6041, 1.1850, 0.9024, 1.1524, 1.1212, 1.1488, 1.0960], device='cuda:1'), covar=tensor([0.2345, 0.2005, 0.2834, 0.3030, 0.3869, 0.2902, 0.2080, 0.3093], device='cuda:1'), in_proj_covar=tensor([0.0200, 0.0226, 0.0197, 0.0220, 0.0235, 0.0198, 0.0193, 0.0211], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-26 12:25:21,848 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5672, 1.9815, 1.4861, 1.8775, 1.5817, 1.4820, 1.6270, 1.3430], device='cuda:1'), covar=tensor([0.2147, 0.1453, 0.1325, 0.1615, 0.3718, 0.1651, 0.1955, 0.2845], device='cuda:1'), in_proj_covar=tensor([0.0318, 0.0338, 0.0248, 0.0313, 0.0318, 0.0290, 0.0281, 0.0305], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 12:25:30,260 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=8969.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:25:32,765 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6664, 1.3650, 1.2491, 1.4044, 1.9573, 1.6450, 1.3226, 1.1811], device='cuda:1'), covar=tensor([0.1557, 0.1607, 0.2020, 0.1652, 0.0743, 0.1549, 0.2225, 0.1930], device='cuda:1'), in_proj_covar=tensor([0.0326, 0.0340, 0.0348, 0.0317, 0.0353, 0.0373, 0.0325, 0.0357], device='cuda:1'), out_proj_covar=tensor([7.1197e-05, 7.3126e-05, 7.5234e-05, 6.6486e-05, 7.5118e-05, 8.1530e-05, 7.0796e-05, 7.7010e-05], device='cuda:1') 2023-04-26 12:25:41,812 INFO [finetune.py:976] (1/7) Epoch 2, batch 3250, loss[loss=0.3945, simple_loss=0.4094, pruned_loss=0.1899, over 4828.00 frames. ], tot_loss[loss=0.2596, simple_loss=0.3066, pruned_loss=0.1064, over 954307.88 frames. ], batch size: 40, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 12:25:53,823 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=8986.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:25:54,451 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4795, 1.3346, 0.4964, 1.1547, 1.4099, 1.3541, 1.2452, 1.2937], device='cuda:1'), covar=tensor([0.0616, 0.0512, 0.0557, 0.0650, 0.0352, 0.0608, 0.0598, 0.0744], device='cuda:1'), in_proj_covar=tensor([0.0032, 0.0027, 0.0025, 0.0031, 0.0022, 0.0031, 0.0030, 0.0032], device='cuda:1'), out_proj_covar=tensor([0.0048, 0.0044, 0.0038, 0.0049, 0.0037, 0.0048, 0.0047, 0.0050], device='cuda:1') 2023-04-26 12:25:55,588 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=8988.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:26:27,444 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=9013.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:26:37,526 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.210e+02 2.159e+02 2.478e+02 3.003e+02 4.851e+02, threshold=4.955e+02, percent-clipped=0.0 2023-04-26 12:26:38,900 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5116, 1.2178, 1.1870, 1.0514, 1.7503, 1.4442, 1.0908, 1.1372], device='cuda:1'), covar=tensor([0.1539, 0.1489, 0.1981, 0.1753, 0.0800, 0.1405, 0.2127, 0.1810], device='cuda:1'), in_proj_covar=tensor([0.0327, 0.0341, 0.0349, 0.0318, 0.0354, 0.0374, 0.0327, 0.0358], device='cuda:1'), out_proj_covar=tensor([7.1371e-05, 7.3292e-05, 7.5463e-05, 6.6746e-05, 7.5474e-05, 8.1902e-05, 7.1234e-05, 7.7229e-05], device='cuda:1') 2023-04-26 12:26:48,553 INFO [finetune.py:976] (1/7) Epoch 2, batch 3300, loss[loss=0.2571, simple_loss=0.3054, pruned_loss=0.1044, over 4815.00 frames. ], tot_loss[loss=0.2641, simple_loss=0.3118, pruned_loss=0.1082, over 955120.90 frames. ], batch size: 51, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 12:27:08,797 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=9047.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:27:25,102 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-04-26 12:27:26,251 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=9074.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:27:28,528 INFO [finetune.py:976] (1/7) Epoch 2, batch 3350, loss[loss=0.2779, simple_loss=0.3257, pruned_loss=0.115, over 4890.00 frames. ], tot_loss[loss=0.2658, simple_loss=0.3142, pruned_loss=0.1087, over 956326.66 frames. ], batch size: 43, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 12:28:00,084 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=9116.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:28:07,186 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.418e+02 2.168e+02 2.587e+02 3.094e+02 5.996e+02, threshold=5.175e+02, percent-clipped=1.0 2023-04-26 12:28:17,532 INFO [finetune.py:976] (1/7) Epoch 2, batch 3400, loss[loss=0.2794, simple_loss=0.3283, pruned_loss=0.1152, over 4738.00 frames. ], tot_loss[loss=0.2672, simple_loss=0.3159, pruned_loss=0.1092, over 955897.36 frames. ], batch size: 54, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 12:28:43,472 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.3224, 2.8879, 0.9542, 1.4045, 1.9819, 1.4247, 3.7824, 1.5528], device='cuda:1'), covar=tensor([0.0655, 0.0974, 0.1022, 0.1241, 0.0595, 0.1013, 0.0198, 0.0751], device='cuda:1'), in_proj_covar=tensor([0.0057, 0.0073, 0.0055, 0.0051, 0.0056, 0.0056, 0.0087, 0.0055], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0008, 0.0008, 0.0008, 0.0008, 0.0011, 0.0008], device='cuda:1') 2023-04-26 12:28:43,532 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.3033, 1.2726, 1.4841, 1.5458, 1.4833, 1.2988, 1.3899, 1.3671], device='cuda:1'), covar=tensor([4.1396, 5.3557, 6.8893, 6.7471, 4.1834, 7.6604, 7.7381, 5.4454], device='cuda:1'), in_proj_covar=tensor([0.0448, 0.0508, 0.0597, 0.0591, 0.0479, 0.0527, 0.0536, 0.0547], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 12:29:01,795 INFO [finetune.py:976] (1/7) Epoch 2, batch 3450, loss[loss=0.2162, simple_loss=0.2697, pruned_loss=0.08134, over 4757.00 frames. ], tot_loss[loss=0.2659, simple_loss=0.315, pruned_loss=0.1084, over 955980.04 frames. ], batch size: 26, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 12:29:23,768 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-04-26 12:29:29,510 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.432e+02 2.133e+02 2.511e+02 2.934e+02 6.196e+02, threshold=5.021e+02, percent-clipped=2.0 2023-04-26 12:29:35,077 INFO [finetune.py:976] (1/7) Epoch 2, batch 3500, loss[loss=0.2349, simple_loss=0.2795, pruned_loss=0.09519, over 4808.00 frames. ], tot_loss[loss=0.2628, simple_loss=0.3114, pruned_loss=0.1071, over 954841.62 frames. ], batch size: 25, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 12:29:46,508 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-04-26 12:30:08,949 INFO [finetune.py:976] (1/7) Epoch 2, batch 3550, loss[loss=0.2354, simple_loss=0.2818, pruned_loss=0.09453, over 4769.00 frames. ], tot_loss[loss=0.2586, simple_loss=0.307, pruned_loss=0.1051, over 954683.12 frames. ], batch size: 26, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 12:30:12,101 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=9283.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:30:13,875 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([0.5580, 1.7690, 1.7277, 1.2697, 1.7378, 1.3265, 2.3018, 1.3838], device='cuda:1'), covar=tensor([0.3640, 0.1498, 0.4501, 0.2735, 0.1620, 0.2448, 0.1269, 0.4434], device='cuda:1'), in_proj_covar=tensor([0.0346, 0.0351, 0.0434, 0.0368, 0.0403, 0.0373, 0.0397, 0.0410], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 12:30:28,094 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8891, 2.5964, 1.7462, 1.7353, 1.2964, 1.3759, 1.7899, 1.3065], device='cuda:1'), covar=tensor([0.2347, 0.2355, 0.2542, 0.3276, 0.3792, 0.2890, 0.2133, 0.3171], device='cuda:1'), in_proj_covar=tensor([0.0200, 0.0227, 0.0197, 0.0218, 0.0235, 0.0198, 0.0192, 0.0211], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-26 12:30:34,722 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=5.01 vs. limit=5.0 2023-04-26 12:30:37,458 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.026e+02 2.002e+02 2.435e+02 2.997e+02 6.904e+02, threshold=4.871e+02, percent-clipped=3.0 2023-04-26 12:30:48,982 INFO [finetune.py:976] (1/7) Epoch 2, batch 3600, loss[loss=0.2546, simple_loss=0.3042, pruned_loss=0.1025, over 4871.00 frames. ], tot_loss[loss=0.2555, simple_loss=0.3038, pruned_loss=0.1037, over 956074.54 frames. ], batch size: 34, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 12:31:03,416 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=9342.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:31:44,877 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=9369.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:31:55,586 INFO [finetune.py:976] (1/7) Epoch 2, batch 3650, loss[loss=0.2759, simple_loss=0.3194, pruned_loss=0.1162, over 4820.00 frames. ], tot_loss[loss=0.2585, simple_loss=0.3065, pruned_loss=0.1053, over 954798.46 frames. ], batch size: 51, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 12:31:58,911 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-04-26 12:32:20,718 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.1700, 0.6760, 0.9974, 1.5498, 1.4180, 1.1118, 1.0683, 0.9827], device='cuda:1'), covar=tensor([3.7147, 5.6936, 6.4753, 7.1913, 3.8926, 6.1685, 6.4019, 4.5306], device='cuda:1'), in_proj_covar=tensor([0.0448, 0.0510, 0.0600, 0.0592, 0.0480, 0.0528, 0.0537, 0.0549], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 12:32:25,432 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=9414.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:32:27,084 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=9416.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:32:28,803 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.507e+02 2.166e+02 2.483e+02 2.970e+02 6.745e+02, threshold=4.965e+02, percent-clipped=1.0 2023-04-26 12:32:34,828 INFO [finetune.py:976] (1/7) Epoch 2, batch 3700, loss[loss=0.2471, simple_loss=0.3036, pruned_loss=0.09534, over 4828.00 frames. ], tot_loss[loss=0.2609, simple_loss=0.3097, pruned_loss=0.1061, over 953878.07 frames. ], batch size: 30, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 12:32:45,512 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.53 vs. limit=2.0 2023-04-26 12:32:58,812 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=9464.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:33:07,014 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=9475.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:33:08,728 INFO [finetune.py:976] (1/7) Epoch 2, batch 3750, loss[loss=0.2774, simple_loss=0.3304, pruned_loss=0.1122, over 4811.00 frames. ], tot_loss[loss=0.2638, simple_loss=0.313, pruned_loss=0.1073, over 956385.78 frames. ], batch size: 33, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 12:33:47,552 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.577e+02 2.140e+02 2.529e+02 3.098e+02 5.505e+02, threshold=5.058e+02, percent-clipped=1.0 2023-04-26 12:33:59,195 INFO [finetune.py:976] (1/7) Epoch 2, batch 3800, loss[loss=0.2437, simple_loss=0.2952, pruned_loss=0.09611, over 4901.00 frames. ], tot_loss[loss=0.2631, simple_loss=0.3126, pruned_loss=0.1068, over 956092.72 frames. ], batch size: 36, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 12:34:21,437 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.3168, 1.6762, 1.4742, 2.1352, 1.9951, 2.2729, 1.5531, 4.4834], device='cuda:1'), covar=tensor([0.0734, 0.0844, 0.0867, 0.1257, 0.0685, 0.0575, 0.0818, 0.0116], device='cuda:1'), in_proj_covar=tensor([0.0040, 0.0041, 0.0042, 0.0046, 0.0042, 0.0041, 0.0041, 0.0066], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0013, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0017], device='cuda:1') 2023-04-26 12:34:55,620 INFO [finetune.py:976] (1/7) Epoch 2, batch 3850, loss[loss=0.2346, simple_loss=0.2839, pruned_loss=0.09271, over 4930.00 frames. ], tot_loss[loss=0.2606, simple_loss=0.3103, pruned_loss=0.1055, over 955681.97 frames. ], batch size: 33, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 12:34:58,832 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=9583.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:34:59,460 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=9584.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:35:05,049 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.1948, 0.5385, 0.8983, 1.5093, 1.3982, 1.0562, 1.0420, 0.9699], device='cuda:1'), covar=tensor([2.8141, 3.7780, 4.0560, 5.0237, 2.8536, 4.4003, 4.7522, 3.2690], device='cuda:1'), in_proj_covar=tensor([0.0449, 0.0508, 0.0600, 0.0594, 0.0481, 0.0526, 0.0537, 0.0548], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 12:35:22,631 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.190e+02 2.119e+02 2.508e+02 2.913e+02 5.368e+02, threshold=5.017e+02, percent-clipped=1.0 2023-04-26 12:35:29,153 INFO [finetune.py:976] (1/7) Epoch 2, batch 3900, loss[loss=0.3034, simple_loss=0.3416, pruned_loss=0.1325, over 4919.00 frames. ], tot_loss[loss=0.2572, simple_loss=0.3064, pruned_loss=0.104, over 956709.25 frames. ], batch size: 38, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 12:35:29,274 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.1582, 2.1349, 1.8574, 1.8396, 2.3063, 1.7685, 2.8781, 1.5991], device='cuda:1'), covar=tensor([0.4403, 0.1777, 0.4790, 0.3203, 0.1706, 0.2758, 0.1310, 0.4405], device='cuda:1'), in_proj_covar=tensor([0.0346, 0.0349, 0.0436, 0.0369, 0.0402, 0.0372, 0.0398, 0.0412], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 12:35:29,443 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 2023-04-26 12:35:38,163 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=9631.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:35:50,797 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=9642.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:35:52,644 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=9645.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:36:12,998 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=9658.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:36:26,739 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=9669.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:36:38,402 INFO [finetune.py:976] (1/7) Epoch 2, batch 3950, loss[loss=0.1959, simple_loss=0.2585, pruned_loss=0.06664, over 4763.00 frames. ], tot_loss[loss=0.2526, simple_loss=0.3016, pruned_loss=0.1018, over 956213.89 frames. ], batch size: 28, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 12:36:55,187 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=9690.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:37:29,747 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=9717.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:37:36,041 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.382e+02 1.910e+02 2.412e+02 2.849e+02 4.927e+02, threshold=4.824e+02, percent-clipped=0.0 2023-04-26 12:37:36,194 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=9719.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:37:39,127 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.76 vs. limit=5.0 2023-04-26 12:37:47,956 INFO [finetune.py:976] (1/7) Epoch 2, batch 4000, loss[loss=0.2853, simple_loss=0.3, pruned_loss=0.1353, over 4726.00 frames. ], tot_loss[loss=0.253, simple_loss=0.3014, pruned_loss=0.1023, over 954201.90 frames. ], batch size: 23, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 12:37:58,328 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=9743.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:38:13,936 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-04-26 12:38:15,579 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=9770.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:38:20,826 INFO [finetune.py:976] (1/7) Epoch 2, batch 4050, loss[loss=0.2447, simple_loss=0.2816, pruned_loss=0.1039, over 4086.00 frames. ], tot_loss[loss=0.2567, simple_loss=0.3052, pruned_loss=0.104, over 953059.80 frames. ], batch size: 17, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 12:38:31,582 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.4521, 1.5015, 1.5854, 2.0802, 2.3701, 2.1448, 1.9214, 1.7560], device='cuda:1'), covar=tensor([0.1850, 0.2728, 0.2519, 0.2875, 0.1779, 0.2496, 0.2656, 0.2218], device='cuda:1'), in_proj_covar=tensor([0.0327, 0.0344, 0.0351, 0.0320, 0.0357, 0.0375, 0.0325, 0.0358], device='cuda:1'), out_proj_covar=tensor([7.1536e-05, 7.3896e-05, 7.6049e-05, 6.7137e-05, 7.6093e-05, 8.2061e-05, 7.0839e-05, 7.7367e-05], device='cuda:1') 2023-04-26 12:38:38,418 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=9804.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:38:47,820 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.265e+02 2.123e+02 2.473e+02 2.970e+02 7.758e+02, threshold=4.946e+02, percent-clipped=3.0 2023-04-26 12:38:54,759 INFO [finetune.py:976] (1/7) Epoch 2, batch 4100, loss[loss=0.2055, simple_loss=0.2583, pruned_loss=0.07634, over 4734.00 frames. ], tot_loss[loss=0.259, simple_loss=0.3076, pruned_loss=0.1052, over 948816.53 frames. ], batch size: 23, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 12:39:19,595 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.1318, 1.7836, 2.3522, 2.5572, 1.7514, 1.4352, 2.0333, 1.3427], device='cuda:1'), covar=tensor([0.0948, 0.1332, 0.0629, 0.1091, 0.1242, 0.1581, 0.1228, 0.1533], device='cuda:1'), in_proj_covar=tensor([0.0071, 0.0079, 0.0076, 0.0073, 0.0086, 0.0095, 0.0089, 0.0080], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003], device='cuda:1') 2023-04-26 12:39:34,295 INFO [finetune.py:976] (1/7) Epoch 2, batch 4150, loss[loss=0.2912, simple_loss=0.3361, pruned_loss=0.1232, over 4757.00 frames. ], tot_loss[loss=0.2595, simple_loss=0.3091, pruned_loss=0.1049, over 952267.13 frames. ], batch size: 28, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 12:39:41,055 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5711, 1.2072, 1.2087, 1.3136, 1.8301, 1.5425, 1.2419, 1.1557], device='cuda:1'), covar=tensor([0.1578, 0.1800, 0.2147, 0.1534, 0.0901, 0.1597, 0.2052, 0.1930], device='cuda:1'), in_proj_covar=tensor([0.0327, 0.0342, 0.0350, 0.0318, 0.0356, 0.0373, 0.0325, 0.0357], device='cuda:1'), out_proj_covar=tensor([7.1378e-05, 7.3445e-05, 7.5892e-05, 6.6826e-05, 7.5869e-05, 8.1486e-05, 7.0769e-05, 7.7291e-05], device='cuda:1') 2023-04-26 12:39:45,729 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.2046, 1.5461, 1.3216, 1.9141, 2.2725, 1.8972, 1.7501, 1.5626], device='cuda:1'), covar=tensor([0.1910, 0.2418, 0.2892, 0.2450, 0.1426, 0.2515, 0.2466, 0.2216], device='cuda:1'), in_proj_covar=tensor([0.0326, 0.0341, 0.0350, 0.0318, 0.0355, 0.0372, 0.0325, 0.0357], device='cuda:1'), out_proj_covar=tensor([7.1328e-05, 7.3328e-05, 7.5818e-05, 6.6735e-05, 7.5772e-05, 8.1374e-05, 7.0685e-05, 7.7185e-05], device='cuda:1') 2023-04-26 12:40:10,100 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5009, 1.4486, 0.5200, 1.2132, 1.4639, 1.3876, 1.2987, 1.3259], device='cuda:1'), covar=tensor([0.0611, 0.0482, 0.0556, 0.0656, 0.0373, 0.0613, 0.0630, 0.0749], device='cuda:1'), in_proj_covar=tensor([0.0032, 0.0027, 0.0025, 0.0031, 0.0022, 0.0031, 0.0030, 0.0032], device='cuda:1'), out_proj_covar=tensor([0.0049, 0.0044, 0.0038, 0.0049, 0.0038, 0.0048, 0.0047, 0.0051], device='cuda:1') 2023-04-26 12:40:18,083 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.356e+02 2.059e+02 2.438e+02 3.061e+02 5.791e+02, threshold=4.877e+02, percent-clipped=3.0 2023-04-26 12:40:29,031 INFO [finetune.py:976] (1/7) Epoch 2, batch 4200, loss[loss=0.2971, simple_loss=0.3402, pruned_loss=0.127, over 4820.00 frames. ], tot_loss[loss=0.2578, simple_loss=0.3087, pruned_loss=0.1035, over 955273.88 frames. ], batch size: 38, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 12:40:43,340 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=9940.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:41:36,131 INFO [finetune.py:976] (1/7) Epoch 2, batch 4250, loss[loss=0.2201, simple_loss=0.2874, pruned_loss=0.07646, over 4890.00 frames. ], tot_loss[loss=0.2553, simple_loss=0.3064, pruned_loss=0.1021, over 956738.98 frames. ], batch size: 32, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 12:42:30,333 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=10014.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:42:33,325 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.150e+02 1.966e+02 2.336e+02 2.838e+02 4.911e+02, threshold=4.671e+02, percent-clipped=1.0 2023-04-26 12:42:43,920 INFO [finetune.py:976] (1/7) Epoch 2, batch 4300, loss[loss=0.237, simple_loss=0.2831, pruned_loss=0.09546, over 4735.00 frames. ], tot_loss[loss=0.251, simple_loss=0.3018, pruned_loss=0.1001, over 958297.39 frames. ], batch size: 59, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 12:43:02,768 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=10039.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:43:18,142 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=10062.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 12:43:23,021 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=10070.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:43:24,370 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.45 vs. limit=5.0 2023-04-26 12:43:27,740 INFO [finetune.py:976] (1/7) Epoch 2, batch 4350, loss[loss=0.2715, simple_loss=0.3138, pruned_loss=0.1146, over 4869.00 frames. ], tot_loss[loss=0.248, simple_loss=0.2985, pruned_loss=0.09878, over 958347.43 frames. ], batch size: 34, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 12:43:41,996 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=10099.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:43:42,650 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6532, 1.7952, 1.0042, 1.4032, 2.1240, 1.5255, 1.4305, 1.5900], device='cuda:1'), covar=tensor([0.0587, 0.0439, 0.0436, 0.0604, 0.0286, 0.0570, 0.0563, 0.0710], device='cuda:1'), in_proj_covar=tensor([0.0032, 0.0027, 0.0025, 0.0031, 0.0022, 0.0031, 0.0030, 0.0032], device='cuda:1'), out_proj_covar=tensor([0.0049, 0.0044, 0.0038, 0.0049, 0.0038, 0.0048, 0.0047, 0.0051], device='cuda:1') 2023-04-26 12:43:42,674 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=10100.0, num_to_drop=1, layers_to_drop={3} 2023-04-26 12:43:54,964 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=10118.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:43:55,494 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.384e+02 2.038e+02 2.484e+02 2.915e+02 5.390e+02, threshold=4.968e+02, percent-clipped=2.0 2023-04-26 12:43:58,068 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=10123.0, num_to_drop=1, layers_to_drop={3} 2023-04-26 12:44:00,629 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-04-26 12:44:00,983 INFO [finetune.py:976] (1/7) Epoch 2, batch 4400, loss[loss=0.2729, simple_loss=0.306, pruned_loss=0.1199, over 4080.00 frames. ], tot_loss[loss=0.2502, simple_loss=0.3001, pruned_loss=0.1001, over 955696.56 frames. ], batch size: 17, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 12:44:34,795 INFO [finetune.py:976] (1/7) Epoch 2, batch 4450, loss[loss=0.3278, simple_loss=0.3719, pruned_loss=0.1418, over 4906.00 frames. ], tot_loss[loss=0.2553, simple_loss=0.3055, pruned_loss=0.1025, over 954622.10 frames. ], batch size: 43, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 12:45:03,115 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.522e+02 2.117e+02 2.495e+02 3.123e+02 6.793e+02, threshold=4.991e+02, percent-clipped=1.0 2023-04-26 12:45:08,100 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0844, 2.4950, 1.1756, 1.2899, 1.8806, 1.2132, 3.5303, 1.5569], device='cuda:1'), covar=tensor([0.0674, 0.1033, 0.0928, 0.1263, 0.0551, 0.1041, 0.0255, 0.0733], device='cuda:1'), in_proj_covar=tensor([0.0056, 0.0073, 0.0054, 0.0050, 0.0055, 0.0056, 0.0086, 0.0054], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0008, 0.0008, 0.0008, 0.0008, 0.0011, 0.0008], device='cuda:1') 2023-04-26 12:45:08,646 INFO [finetune.py:976] (1/7) Epoch 2, batch 4500, loss[loss=0.2716, simple_loss=0.3148, pruned_loss=0.1143, over 4898.00 frames. ], tot_loss[loss=0.2583, simple_loss=0.3086, pruned_loss=0.104, over 955588.95 frames. ], batch size: 37, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 12:45:15,983 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=10240.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:45:42,229 INFO [finetune.py:976] (1/7) Epoch 2, batch 4550, loss[loss=0.2572, simple_loss=0.3123, pruned_loss=0.1011, over 4727.00 frames. ], tot_loss[loss=0.2579, simple_loss=0.3085, pruned_loss=0.1037, over 955161.22 frames. ], batch size: 59, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 12:45:48,412 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=10288.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:46:08,743 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-04-26 12:46:21,464 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=10314.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:46:29,829 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.123e+02 1.925e+02 2.197e+02 2.697e+02 5.130e+02, threshold=4.395e+02, percent-clipped=1.0 2023-04-26 12:46:41,911 INFO [finetune.py:976] (1/7) Epoch 2, batch 4600, loss[loss=0.2519, simple_loss=0.3024, pruned_loss=0.1007, over 4881.00 frames. ], tot_loss[loss=0.2557, simple_loss=0.3065, pruned_loss=0.1025, over 956515.77 frames. ], batch size: 35, lr: 3.99e-03, grad_scale: 64.0 2023-04-26 12:47:09,969 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=10362.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:47:29,153 INFO [finetune.py:976] (1/7) Epoch 2, batch 4650, loss[loss=0.2621, simple_loss=0.3126, pruned_loss=0.1058, over 4893.00 frames. ], tot_loss[loss=0.2548, simple_loss=0.3046, pruned_loss=0.1025, over 956131.89 frames. ], batch size: 35, lr: 3.99e-03, grad_scale: 64.0 2023-04-26 12:47:50,958 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=10395.0, num_to_drop=1, layers_to_drop={3} 2023-04-26 12:47:58,935 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=10399.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:48:12,474 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=10409.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:48:23,999 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=10418.0, num_to_drop=1, layers_to_drop={3} 2023-04-26 12:48:25,137 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.154e+02 2.119e+02 2.411e+02 2.762e+02 5.438e+02, threshold=4.823e+02, percent-clipped=2.0 2023-04-26 12:48:35,916 INFO [finetune.py:976] (1/7) Epoch 2, batch 4700, loss[loss=0.2292, simple_loss=0.269, pruned_loss=0.09471, over 4435.00 frames. ], tot_loss[loss=0.2509, simple_loss=0.3006, pruned_loss=0.1006, over 955747.18 frames. ], batch size: 19, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 12:48:53,256 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=10447.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:48:53,303 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.2008, 1.5128, 1.5098, 1.7938, 1.6225, 2.0410, 1.4232, 3.7410], device='cuda:1'), covar=tensor([0.0723, 0.0821, 0.0813, 0.1319, 0.0723, 0.0605, 0.0810, 0.0166], device='cuda:1'), in_proj_covar=tensor([0.0041, 0.0041, 0.0042, 0.0047, 0.0042, 0.0041, 0.0041, 0.0066], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0013, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0017], device='cuda:1') 2023-04-26 12:48:57,666 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=10454.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:49:09,530 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=10470.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:49:15,191 INFO [finetune.py:976] (1/7) Epoch 2, batch 4750, loss[loss=0.2183, simple_loss=0.28, pruned_loss=0.07827, over 4910.00 frames. ], tot_loss[loss=0.2482, simple_loss=0.2982, pruned_loss=0.09913, over 956351.77 frames. ], batch size: 32, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 12:49:38,678 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5403, 3.4615, 0.9913, 1.9593, 1.7783, 2.4473, 1.9945, 1.0250], device='cuda:1'), covar=tensor([0.1309, 0.0915, 0.2172, 0.1249, 0.1133, 0.1072, 0.1480, 0.2178], device='cuda:1'), in_proj_covar=tensor([0.0125, 0.0272, 0.0152, 0.0132, 0.0144, 0.0166, 0.0130, 0.0134], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-26 12:49:40,438 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=10515.0, num_to_drop=1, layers_to_drop={2} 2023-04-26 12:49:43,424 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.363e+02 1.957e+02 2.292e+02 2.912e+02 6.222e+02, threshold=4.583e+02, percent-clipped=3.0 2023-04-26 12:49:49,341 INFO [finetune.py:976] (1/7) Epoch 2, batch 4800, loss[loss=0.2782, simple_loss=0.3297, pruned_loss=0.1134, over 4927.00 frames. ], tot_loss[loss=0.2513, simple_loss=0.3017, pruned_loss=0.1005, over 956372.68 frames. ], batch size: 38, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 12:50:23,189 INFO [finetune.py:976] (1/7) Epoch 2, batch 4850, loss[loss=0.2101, simple_loss=0.2763, pruned_loss=0.07195, over 4888.00 frames. ], tot_loss[loss=0.2538, simple_loss=0.3049, pruned_loss=0.1013, over 956090.51 frames. ], batch size: 32, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 12:50:44,399 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.1888, 2.2187, 2.4386, 2.7582, 2.7373, 1.9829, 1.6421, 2.1943], device='cuda:1'), covar=tensor([0.1047, 0.1020, 0.0642, 0.0681, 0.0608, 0.1066, 0.1207, 0.0681], device='cuda:1'), in_proj_covar=tensor([0.0214, 0.0213, 0.0191, 0.0186, 0.0183, 0.0202, 0.0179, 0.0197], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 12:50:50,776 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.143e+02 2.100e+02 2.315e+02 2.802e+02 4.090e+02, threshold=4.629e+02, percent-clipped=1.0 2023-04-26 12:50:52,706 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.1776, 1.6556, 2.1617, 2.5884, 1.6449, 1.1648, 1.9293, 1.1434], device='cuda:1'), covar=tensor([0.0917, 0.1125, 0.0756, 0.0825, 0.1268, 0.2333, 0.1094, 0.1662], device='cuda:1'), in_proj_covar=tensor([0.0071, 0.0080, 0.0076, 0.0073, 0.0086, 0.0096, 0.0089, 0.0080], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003], device='cuda:1') 2023-04-26 12:50:56,074 INFO [finetune.py:976] (1/7) Epoch 2, batch 4900, loss[loss=0.3016, simple_loss=0.3393, pruned_loss=0.132, over 4895.00 frames. ], tot_loss[loss=0.2555, simple_loss=0.3066, pruned_loss=0.1021, over 957271.43 frames. ], batch size: 46, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 12:51:46,566 INFO [finetune.py:976] (1/7) Epoch 2, batch 4950, loss[loss=0.234, simple_loss=0.2944, pruned_loss=0.08685, over 4933.00 frames. ], tot_loss[loss=0.2551, simple_loss=0.3068, pruned_loss=0.1018, over 956321.39 frames. ], batch size: 42, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 12:52:10,655 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=10695.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:52:32,452 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6297, 2.4432, 1.5016, 1.5227, 1.2383, 1.2883, 1.5496, 1.2133], device='cuda:1'), covar=tensor([0.2576, 0.1989, 0.2711, 0.3082, 0.3993, 0.3163, 0.2230, 0.3208], device='cuda:1'), in_proj_covar=tensor([0.0198, 0.0225, 0.0193, 0.0216, 0.0232, 0.0196, 0.0189, 0.0208], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-26 12:52:36,492 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=10718.0, num_to_drop=1, layers_to_drop={2} 2023-04-26 12:52:43,790 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.438e+02 1.913e+02 2.320e+02 2.882e+02 5.075e+02, threshold=4.640e+02, percent-clipped=3.0 2023-04-26 12:52:54,984 INFO [finetune.py:976] (1/7) Epoch 2, batch 5000, loss[loss=0.2678, simple_loss=0.3066, pruned_loss=0.1145, over 4896.00 frames. ], tot_loss[loss=0.2523, simple_loss=0.3044, pruned_loss=0.1001, over 958368.94 frames. ], batch size: 35, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 12:53:06,410 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.61 vs. limit=5.0 2023-04-26 12:53:16,590 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=10743.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:53:17,242 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=10744.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:53:29,935 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=10765.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:53:30,543 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=10766.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 12:53:42,296 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=10774.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:53:45,584 INFO [finetune.py:976] (1/7) Epoch 2, batch 5050, loss[loss=0.2523, simple_loss=0.3018, pruned_loss=0.1014, over 4861.00 frames. ], tot_loss[loss=0.2495, simple_loss=0.3009, pruned_loss=0.09906, over 958613.57 frames. ], batch size: 49, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 12:54:25,674 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=10805.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:54:28,748 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=10810.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 12:54:30,009 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([0.9355, 2.0299, 1.7172, 1.7211, 1.9517, 1.3994, 2.5736, 1.2746], device='cuda:1'), covar=tensor([0.4413, 0.1526, 0.4614, 0.2736, 0.2121, 0.3076, 0.1259, 0.4591], device='cuda:1'), in_proj_covar=tensor([0.0349, 0.0352, 0.0438, 0.0370, 0.0406, 0.0377, 0.0404, 0.0417], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 12:54:40,510 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.434e+02 1.942e+02 2.267e+02 2.719e+02 4.388e+02, threshold=4.533e+02, percent-clipped=0.0 2023-04-26 12:54:51,196 INFO [finetune.py:976] (1/7) Epoch 2, batch 5100, loss[loss=0.2876, simple_loss=0.2952, pruned_loss=0.14, over 3759.00 frames. ], tot_loss[loss=0.2466, simple_loss=0.2975, pruned_loss=0.09784, over 955896.27 frames. ], batch size: 16, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 12:55:01,989 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=10835.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:55:12,405 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.69 vs. limit=5.0 2023-04-26 12:55:22,578 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2023-04-26 12:55:41,027 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.5342, 3.0106, 1.0692, 1.5947, 2.4257, 1.5664, 4.2081, 2.1972], device='cuda:1'), covar=tensor([0.0642, 0.0874, 0.0999, 0.1322, 0.0561, 0.1010, 0.0216, 0.0636], device='cuda:1'), in_proj_covar=tensor([0.0056, 0.0072, 0.0054, 0.0050, 0.0055, 0.0056, 0.0085, 0.0054], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0008, 0.0008, 0.0008, 0.0008, 0.0011, 0.0008], device='cuda:1') 2023-04-26 12:55:45,710 INFO [finetune.py:976] (1/7) Epoch 2, batch 5150, loss[loss=0.3021, simple_loss=0.3487, pruned_loss=0.1277, over 4384.00 frames. ], tot_loss[loss=0.2477, simple_loss=0.2981, pruned_loss=0.09863, over 955740.57 frames. ], batch size: 65, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 12:55:47,215 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.49 vs. limit=2.0 2023-04-26 12:55:50,599 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=10885.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 12:56:13,137 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.326e+02 2.074e+02 2.408e+02 2.882e+02 5.966e+02, threshold=4.815e+02, percent-clipped=3.0 2023-04-26 12:56:18,495 INFO [finetune.py:976] (1/7) Epoch 2, batch 5200, loss[loss=0.3082, simple_loss=0.3429, pruned_loss=0.1368, over 4897.00 frames. ], tot_loss[loss=0.2506, simple_loss=0.3015, pruned_loss=0.09987, over 955119.54 frames. ], batch size: 37, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 12:56:21,046 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([3.7928, 3.7244, 2.8063, 4.3081, 3.7387, 3.7463, 1.8365, 3.7377], device='cuda:1'), covar=tensor([0.1588, 0.1092, 0.3262, 0.1677, 0.3594, 0.1699, 0.4993, 0.2192], device='cuda:1'), in_proj_covar=tensor([0.0254, 0.0228, 0.0270, 0.0321, 0.0315, 0.0266, 0.0279, 0.0282], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-26 12:56:31,978 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=10946.0, num_to_drop=1, layers_to_drop={2} 2023-04-26 12:56:57,990 INFO [finetune.py:976] (1/7) Epoch 2, batch 5250, loss[loss=0.2725, simple_loss=0.3289, pruned_loss=0.1081, over 4865.00 frames. ], tot_loss[loss=0.253, simple_loss=0.3045, pruned_loss=0.1008, over 954282.90 frames. ], batch size: 34, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 12:57:39,781 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5676, 1.2535, 1.4702, 1.8061, 1.7342, 1.4802, 1.5219, 1.4540], device='cuda:1'), covar=tensor([2.4256, 3.2990, 3.8933, 4.6160, 2.7960, 4.2867, 3.8893, 2.9644], device='cuda:1'), in_proj_covar=tensor([0.0447, 0.0505, 0.0599, 0.0597, 0.0480, 0.0523, 0.0533, 0.0544], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 12:57:40,209 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.533e+02 2.106e+02 2.454e+02 2.928e+02 5.017e+02, threshold=4.909e+02, percent-clipped=1.0 2023-04-26 12:57:51,419 INFO [finetune.py:976] (1/7) Epoch 2, batch 5300, loss[loss=0.2155, simple_loss=0.282, pruned_loss=0.07445, over 4814.00 frames. ], tot_loss[loss=0.2536, simple_loss=0.3054, pruned_loss=0.1009, over 954588.52 frames. ], batch size: 33, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 12:58:45,117 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=11065.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:58:57,624 INFO [finetune.py:976] (1/7) Epoch 2, batch 5350, loss[loss=0.2532, simple_loss=0.308, pruned_loss=0.09922, over 4836.00 frames. ], tot_loss[loss=0.2536, simple_loss=0.3058, pruned_loss=0.1007, over 954651.85 frames. ], batch size: 30, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 12:59:29,512 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=11100.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:59:47,702 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=11110.0, num_to_drop=1, layers_to_drop={2} 2023-04-26 12:59:50,021 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=11113.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:00:00,706 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.273e+02 1.923e+02 2.240e+02 2.858e+02 6.419e+02, threshold=4.480e+02, percent-clipped=4.0 2023-04-26 13:00:11,119 INFO [finetune.py:976] (1/7) Epoch 2, batch 5400, loss[loss=0.2171, simple_loss=0.2654, pruned_loss=0.08438, over 4850.00 frames. ], tot_loss[loss=0.2515, simple_loss=0.3028, pruned_loss=0.1001, over 951020.06 frames. ], batch size: 49, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 13:00:12,405 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=11130.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:00:32,453 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=11158.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:00:42,744 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=11174.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:00:45,046 INFO [finetune.py:976] (1/7) Epoch 2, batch 5450, loss[loss=0.1985, simple_loss=0.2562, pruned_loss=0.07037, over 4797.00 frames. ], tot_loss[loss=0.2474, simple_loss=0.2985, pruned_loss=0.09814, over 952655.95 frames. ], batch size: 25, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 13:01:18,382 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.436e+02 1.901e+02 2.286e+02 2.895e+02 5.867e+02, threshold=4.573e+02, percent-clipped=4.0 2023-04-26 13:01:23,718 INFO [finetune.py:976] (1/7) Epoch 2, batch 5500, loss[loss=0.244, simple_loss=0.2947, pruned_loss=0.09666, over 4825.00 frames. ], tot_loss[loss=0.2436, simple_loss=0.2945, pruned_loss=0.09641, over 952533.03 frames. ], batch size: 39, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 13:01:28,092 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=11235.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:01:31,711 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=11241.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 13:01:45,822 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6884, 3.6661, 0.9494, 1.9553, 2.0754, 2.5482, 2.1686, 1.0269], device='cuda:1'), covar=tensor([0.1329, 0.1011, 0.2274, 0.1460, 0.1035, 0.1033, 0.1452, 0.1878], device='cuda:1'), in_proj_covar=tensor([0.0126, 0.0273, 0.0153, 0.0133, 0.0144, 0.0166, 0.0130, 0.0134], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-26 13:01:57,533 INFO [finetune.py:976] (1/7) Epoch 2, batch 5550, loss[loss=0.2157, simple_loss=0.2649, pruned_loss=0.08329, over 4778.00 frames. ], tot_loss[loss=0.2438, simple_loss=0.2949, pruned_loss=0.09636, over 953206.45 frames. ], batch size: 26, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 13:02:34,877 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.61 vs. limit=2.0 2023-04-26 13:02:35,264 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.528e+02 2.026e+02 2.364e+02 2.674e+02 3.741e+02, threshold=4.728e+02, percent-clipped=0.0 2023-04-26 13:02:39,949 INFO [finetune.py:976] (1/7) Epoch 2, batch 5600, loss[loss=0.241, simple_loss=0.2924, pruned_loss=0.09481, over 4725.00 frames. ], tot_loss[loss=0.2449, simple_loss=0.2974, pruned_loss=0.09626, over 953032.32 frames. ], batch size: 23, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 13:03:42,986 INFO [finetune.py:976] (1/7) Epoch 2, batch 5650, loss[loss=0.2281, simple_loss=0.2775, pruned_loss=0.08937, over 4695.00 frames. ], tot_loss[loss=0.2484, simple_loss=0.3013, pruned_loss=0.0977, over 953443.84 frames. ], batch size: 23, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 13:03:43,070 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.3077, 1.7650, 1.6819, 2.2202, 2.0568, 2.0972, 1.6675, 4.6891], device='cuda:1'), covar=tensor([0.0714, 0.0806, 0.0882, 0.1313, 0.0655, 0.0593, 0.0799, 0.0122], device='cuda:1'), in_proj_covar=tensor([0.0040, 0.0041, 0.0042, 0.0046, 0.0041, 0.0041, 0.0041, 0.0065], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0017], device='cuda:1') 2023-04-26 13:04:08,639 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=11400.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:04:31,185 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.082e+02 1.867e+02 2.284e+02 2.863e+02 5.268e+02, threshold=4.568e+02, percent-clipped=1.0 2023-04-26 13:04:35,382 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8192, 2.2170, 2.2527, 2.6798, 2.3957, 2.4763, 2.1436, 4.8742], device='cuda:1'), covar=tensor([0.0607, 0.0689, 0.0713, 0.1066, 0.0600, 0.0532, 0.0661, 0.0143], device='cuda:1'), in_proj_covar=tensor([0.0040, 0.0040, 0.0042, 0.0046, 0.0041, 0.0041, 0.0041, 0.0065], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0017], device='cuda:1') 2023-04-26 13:04:35,906 INFO [finetune.py:976] (1/7) Epoch 2, batch 5700, loss[loss=0.1882, simple_loss=0.2321, pruned_loss=0.0722, over 4178.00 frames. ], tot_loss[loss=0.2466, simple_loss=0.2978, pruned_loss=0.09768, over 936983.82 frames. ], batch size: 18, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 13:04:37,207 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=11430.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:04:47,961 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=11448.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:05:07,855 INFO [finetune.py:976] (1/7) Epoch 3, batch 0, loss[loss=0.2043, simple_loss=0.2575, pruned_loss=0.07558, over 4811.00 frames. ], tot_loss[loss=0.2043, simple_loss=0.2575, pruned_loss=0.07558, over 4811.00 frames. ], batch size: 25, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 13:05:07,855 INFO [finetune.py:1001] (1/7) Computing validation loss 2023-04-26 13:05:24,857 INFO [finetune.py:1010] (1/7) Epoch 3, validation: loss=0.1779, simple_loss=0.251, pruned_loss=0.05243, over 2265189.00 frames. 2023-04-26 13:05:24,857 INFO [finetune.py:1011] (1/7) Maximum memory allocated so far is 6391MB 2023-04-26 13:05:42,226 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=11478.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:06:01,720 INFO [finetune.py:976] (1/7) Epoch 3, batch 50, loss[loss=0.284, simple_loss=0.3324, pruned_loss=0.1178, over 4862.00 frames. ], tot_loss[loss=0.2514, simple_loss=0.3033, pruned_loss=0.09973, over 215570.01 frames. ], batch size: 34, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 13:06:11,129 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.469e+02 1.955e+02 2.272e+02 2.719e+02 4.720e+02, threshold=4.545e+02, percent-clipped=1.0 2023-04-26 13:06:17,274 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=11530.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:06:18,525 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=11532.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:06:24,381 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=11541.0, num_to_drop=1, layers_to_drop={2} 2023-04-26 13:06:26,353 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-04-26 13:06:34,900 INFO [finetune.py:976] (1/7) Epoch 3, batch 100, loss[loss=0.1684, simple_loss=0.2271, pruned_loss=0.05488, over 4828.00 frames. ], tot_loss[loss=0.2425, simple_loss=0.2943, pruned_loss=0.09534, over 381514.96 frames. ], batch size: 25, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 13:06:55,969 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=11589.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 13:06:58,873 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.2990, 1.5672, 1.5179, 1.8510, 1.6072, 2.0904, 1.4399, 3.7413], device='cuda:1'), covar=tensor([0.0748, 0.0829, 0.0877, 0.1309, 0.0721, 0.0572, 0.0783, 0.0166], device='cuda:1'), in_proj_covar=tensor([0.0040, 0.0041, 0.0042, 0.0047, 0.0042, 0.0041, 0.0041, 0.0065], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0013, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0017], device='cuda:1') 2023-04-26 13:06:58,897 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=11593.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:07:08,541 INFO [finetune.py:976] (1/7) Epoch 3, batch 150, loss[loss=0.2387, simple_loss=0.2916, pruned_loss=0.09292, over 4820.00 frames. ], tot_loss[loss=0.2366, simple_loss=0.2883, pruned_loss=0.09245, over 508981.64 frames. ], batch size: 38, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 13:07:18,014 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.434e+02 1.865e+02 2.222e+02 2.672e+02 4.139e+02, threshold=4.445e+02, percent-clipped=0.0 2023-04-26 13:07:27,311 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=11635.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:07:42,007 INFO [finetune.py:976] (1/7) Epoch 3, batch 200, loss[loss=0.2692, simple_loss=0.3109, pruned_loss=0.1138, over 4108.00 frames. ], tot_loss[loss=0.2359, simple_loss=0.2868, pruned_loss=0.09252, over 608843.80 frames. ], batch size: 65, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 13:07:46,361 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-04-26 13:08:29,866 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=11696.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:08:31,704 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=11699.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:08:42,574 INFO [finetune.py:976] (1/7) Epoch 3, batch 250, loss[loss=0.2404, simple_loss=0.3039, pruned_loss=0.08846, over 4867.00 frames. ], tot_loss[loss=0.2433, simple_loss=0.2937, pruned_loss=0.09644, over 685494.73 frames. ], batch size: 34, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 13:09:02,114 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.384e+02 2.012e+02 2.336e+02 2.910e+02 4.662e+02, threshold=4.672e+02, percent-clipped=2.0 2023-04-26 13:09:36,496 INFO [finetune.py:976] (1/7) Epoch 3, batch 300, loss[loss=0.2859, simple_loss=0.3327, pruned_loss=0.1195, over 4215.00 frames. ], tot_loss[loss=0.2454, simple_loss=0.2972, pruned_loss=0.09676, over 744437.43 frames. ], batch size: 65, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 13:09:40,539 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=11760.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:09:57,763 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8834, 2.3901, 0.9524, 1.1284, 1.7502, 1.0303, 2.8592, 1.2842], device='cuda:1'), covar=tensor([0.0905, 0.0912, 0.1108, 0.1511, 0.0648, 0.1363, 0.0347, 0.0951], device='cuda:1'), in_proj_covar=tensor([0.0056, 0.0074, 0.0054, 0.0051, 0.0056, 0.0057, 0.0086, 0.0054], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0008, 0.0008, 0.0008, 0.0008, 0.0011, 0.0008], device='cuda:1') 2023-04-26 13:10:09,487 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5388, 3.7179, 0.9616, 1.8076, 2.0429, 2.4841, 2.1070, 1.0395], device='cuda:1'), covar=tensor([0.1433, 0.1033, 0.2182, 0.1523, 0.1168, 0.1219, 0.1516, 0.2060], device='cuda:1'), in_proj_covar=tensor([0.0125, 0.0270, 0.0151, 0.0131, 0.0143, 0.0165, 0.0128, 0.0133], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-26 13:10:10,602 INFO [finetune.py:976] (1/7) Epoch 3, batch 350, loss[loss=0.2634, simple_loss=0.3276, pruned_loss=0.09959, over 4800.00 frames. ], tot_loss[loss=0.248, simple_loss=0.3003, pruned_loss=0.09781, over 791170.11 frames. ], batch size: 51, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 13:10:18,376 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=11816.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:10:20,685 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6891, 1.6469, 0.8646, 1.3109, 1.8583, 1.5464, 1.4279, 1.5321], device='cuda:1'), covar=tensor([0.0577, 0.0455, 0.0473, 0.0648, 0.0322, 0.0618, 0.0580, 0.0727], device='cuda:1'), in_proj_covar=tensor([0.0032, 0.0027, 0.0025, 0.0031, 0.0022, 0.0031, 0.0030, 0.0032], device='cuda:1'), out_proj_covar=tensor([0.0048, 0.0044, 0.0038, 0.0049, 0.0037, 0.0048, 0.0047, 0.0050], device='cuda:1') 2023-04-26 13:10:21,168 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.314e+02 2.028e+02 2.307e+02 2.756e+02 7.160e+02, threshold=4.615e+02, percent-clipped=2.0 2023-04-26 13:10:28,357 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=11830.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:10:38,186 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=11846.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:10:50,157 INFO [finetune.py:976] (1/7) Epoch 3, batch 400, loss[loss=0.2429, simple_loss=0.28, pruned_loss=0.1029, over 4703.00 frames. ], tot_loss[loss=0.2493, simple_loss=0.3019, pruned_loss=0.0984, over 826335.95 frames. ], batch size: 23, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 13:11:20,874 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=11877.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:11:21,407 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=11878.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:11:33,328 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=11888.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:11:44,732 INFO [finetune.py:976] (1/7) Epoch 3, batch 450, loss[loss=0.2506, simple_loss=0.297, pruned_loss=0.1021, over 4817.00 frames. ], tot_loss[loss=0.2468, simple_loss=0.2996, pruned_loss=0.09705, over 855068.99 frames. ], batch size: 38, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 13:11:45,490 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=11907.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:11:54,737 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.367e+02 1.957e+02 2.306e+02 2.695e+02 4.680e+02, threshold=4.613e+02, percent-clipped=1.0 2023-04-26 13:11:54,860 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=11920.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:12:06,632 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.4181, 2.2772, 2.5806, 2.8612, 2.6169, 2.2235, 1.7927, 2.3874], device='cuda:1'), covar=tensor([0.1083, 0.1095, 0.0580, 0.0750, 0.0726, 0.1054, 0.1223, 0.0707], device='cuda:1'), in_proj_covar=tensor([0.0210, 0.0211, 0.0190, 0.0185, 0.0181, 0.0201, 0.0177, 0.0196], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 13:12:13,318 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=11948.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:12:18,146 INFO [finetune.py:976] (1/7) Epoch 3, batch 500, loss[loss=0.2349, simple_loss=0.2914, pruned_loss=0.08923, over 4819.00 frames. ], tot_loss[loss=0.2455, simple_loss=0.2974, pruned_loss=0.09678, over 876685.54 frames. ], batch size: 41, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 13:12:18,871 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2023-04-26 13:12:36,385 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=11981.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:12:38,122 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=11984.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:12:42,324 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=11991.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:12:44,791 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4380, 1.2484, 1.6768, 1.6004, 1.3108, 1.0488, 1.4313, 1.0676], device='cuda:1'), covar=tensor([0.0823, 0.0707, 0.0516, 0.0672, 0.0907, 0.1239, 0.0668, 0.0954], device='cuda:1'), in_proj_covar=tensor([0.0070, 0.0079, 0.0075, 0.0072, 0.0085, 0.0095, 0.0088, 0.0080], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003], device='cuda:1') 2023-04-26 13:12:52,573 INFO [finetune.py:976] (1/7) Epoch 3, batch 550, loss[loss=0.2416, simple_loss=0.3017, pruned_loss=0.09081, over 4897.00 frames. ], tot_loss[loss=0.2434, simple_loss=0.295, pruned_loss=0.09587, over 893947.53 frames. ], batch size: 35, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 13:12:55,031 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=12009.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:13:02,613 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.347e+02 1.896e+02 2.113e+02 2.653e+02 4.840e+02, threshold=4.226e+02, percent-clipped=1.0 2023-04-26 13:13:19,313 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=12045.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:13:30,908 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=12055.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:13:31,437 INFO [finetune.py:976] (1/7) Epoch 3, batch 600, loss[loss=0.199, simple_loss=0.2519, pruned_loss=0.07308, over 4102.00 frames. ], tot_loss[loss=0.2423, simple_loss=0.2944, pruned_loss=0.09509, over 906692.91 frames. ], batch size: 18, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 13:14:01,480 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=12076.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:14:02,108 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0856, 1.2585, 1.3126, 1.4611, 1.4091, 1.5301, 1.3611, 1.3952], device='cuda:1'), covar=tensor([1.6594, 2.7617, 2.2174, 1.8541, 2.1961, 3.5802, 2.6685, 2.2892], device='cuda:1'), in_proj_covar=tensor([0.0295, 0.0396, 0.0313, 0.0319, 0.0346, 0.0393, 0.0380, 0.0342], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-26 13:14:38,091 INFO [finetune.py:976] (1/7) Epoch 3, batch 650, loss[loss=0.2315, simple_loss=0.2983, pruned_loss=0.08234, over 4850.00 frames. ], tot_loss[loss=0.2451, simple_loss=0.2979, pruned_loss=0.09611, over 918142.89 frames. ], batch size: 44, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 13:14:44,593 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([3.4564, 3.2678, 2.5551, 3.9049, 3.3687, 3.4019, 1.5040, 3.3356], device='cuda:1'), covar=tensor([0.1675, 0.1312, 0.3195, 0.1937, 0.3716, 0.1953, 0.5845, 0.2363], device='cuda:1'), in_proj_covar=tensor([0.0253, 0.0228, 0.0269, 0.0321, 0.0313, 0.0265, 0.0280, 0.0281], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-26 13:14:55,287 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.228e+02 1.956e+02 2.261e+02 2.773e+02 5.813e+02, threshold=4.521e+02, percent-clipped=3.0 2023-04-26 13:15:18,801 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=12137.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:15:22,672 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.49 vs. limit=2.0 2023-04-26 13:15:30,234 INFO [finetune.py:976] (1/7) Epoch 3, batch 700, loss[loss=0.2643, simple_loss=0.318, pruned_loss=0.1053, over 4813.00 frames. ], tot_loss[loss=0.2459, simple_loss=0.2995, pruned_loss=0.09611, over 927046.52 frames. ], batch size: 39, lr: 3.98e-03, grad_scale: 32.0 2023-04-26 13:15:40,470 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=12172.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:15:52,210 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=12188.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:16:00,832 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=12202.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:16:03,160 INFO [finetune.py:976] (1/7) Epoch 3, batch 750, loss[loss=0.2428, simple_loss=0.3057, pruned_loss=0.08992, over 4909.00 frames. ], tot_loss[loss=0.2466, simple_loss=0.3004, pruned_loss=0.0964, over 933749.58 frames. ], batch size: 36, lr: 3.98e-03, grad_scale: 32.0 2023-04-26 13:16:11,580 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.101e+02 2.083e+02 2.562e+02 2.927e+02 7.910e+02, threshold=5.125e+02, percent-clipped=5.0 2023-04-26 13:16:22,130 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5513, 2.1600, 1.5615, 1.3310, 1.1768, 1.2109, 1.5151, 1.1877], device='cuda:1'), covar=tensor([0.2343, 0.1951, 0.2372, 0.2814, 0.3585, 0.2850, 0.1950, 0.2943], device='cuda:1'), in_proj_covar=tensor([0.0198, 0.0225, 0.0192, 0.0216, 0.0230, 0.0196, 0.0188, 0.0206], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-26 13:16:23,271 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=12236.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:16:42,544 INFO [finetune.py:976] (1/7) Epoch 3, batch 800, loss[loss=0.2285, simple_loss=0.2926, pruned_loss=0.08221, over 4811.00 frames. ], tot_loss[loss=0.2449, simple_loss=0.2993, pruned_loss=0.09525, over 937459.88 frames. ], batch size: 39, lr: 3.98e-03, grad_scale: 32.0 2023-04-26 13:17:00,569 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=12276.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:17:11,619 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=12291.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:17:19,897 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=12304.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:17:21,019 INFO [finetune.py:976] (1/7) Epoch 3, batch 850, loss[loss=0.2233, simple_loss=0.2667, pruned_loss=0.08999, over 4222.00 frames. ], tot_loss[loss=0.2457, simple_loss=0.2987, pruned_loss=0.09633, over 942247.77 frames. ], batch size: 18, lr: 3.98e-03, grad_scale: 32.0 2023-04-26 13:17:29,559 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.293e+02 1.971e+02 2.371e+02 2.612e+02 4.896e+02, threshold=4.741e+02, percent-clipped=0.0 2023-04-26 13:17:43,089 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=12339.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:17:43,712 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=12340.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:17:54,289 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=12355.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:17:54,806 INFO [finetune.py:976] (1/7) Epoch 3, batch 900, loss[loss=0.24, simple_loss=0.2778, pruned_loss=0.1011, over 4352.00 frames. ], tot_loss[loss=0.2412, simple_loss=0.2943, pruned_loss=0.0941, over 943636.44 frames. ], batch size: 65, lr: 3.98e-03, grad_scale: 32.0 2023-04-26 13:18:09,316 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([3.7603, 3.5825, 2.8623, 4.3263, 3.6979, 3.7816, 1.9011, 3.7445], device='cuda:1'), covar=tensor([0.1524, 0.1350, 0.3413, 0.1480, 0.3092, 0.1849, 0.5252, 0.2188], device='cuda:1'), in_proj_covar=tensor([0.0254, 0.0228, 0.0269, 0.0321, 0.0314, 0.0265, 0.0280, 0.0281], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-26 13:18:26,968 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=12403.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:18:28,771 INFO [finetune.py:976] (1/7) Epoch 3, batch 950, loss[loss=0.222, simple_loss=0.2749, pruned_loss=0.08457, over 4721.00 frames. ], tot_loss[loss=0.2395, simple_loss=0.2922, pruned_loss=0.09345, over 947592.29 frames. ], batch size: 23, lr: 3.98e-03, grad_scale: 64.0 2023-04-26 13:18:37,308 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.210e+02 1.930e+02 2.195e+02 2.798e+02 5.251e+02, threshold=4.389e+02, percent-clipped=2.0 2023-04-26 13:18:50,720 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=12432.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:19:13,290 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=12446.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:19:20,773 INFO [finetune.py:976] (1/7) Epoch 3, batch 1000, loss[loss=0.2619, simple_loss=0.304, pruned_loss=0.1099, over 4762.00 frames. ], tot_loss[loss=0.2425, simple_loss=0.2952, pruned_loss=0.0949, over 950598.17 frames. ], batch size: 26, lr: 3.98e-03, grad_scale: 64.0 2023-04-26 13:19:29,364 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.3911, 1.3367, 4.0409, 3.7371, 3.5921, 3.8660, 3.7252, 3.5491], device='cuda:1'), covar=tensor([0.7335, 0.5916, 0.1006, 0.1710, 0.1133, 0.1600, 0.2129, 0.1384], device='cuda:1'), in_proj_covar=tensor([0.0327, 0.0310, 0.0438, 0.0446, 0.0375, 0.0425, 0.0336, 0.0394], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0002], device='cuda:1') 2023-04-26 13:19:30,610 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=12472.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:19:57,862 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=12502.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:20:06,938 INFO [finetune.py:976] (1/7) Epoch 3, batch 1050, loss[loss=0.2382, simple_loss=0.3021, pruned_loss=0.0871, over 4904.00 frames. ], tot_loss[loss=0.2448, simple_loss=0.298, pruned_loss=0.09582, over 950577.81 frames. ], batch size: 36, lr: 3.98e-03, grad_scale: 64.0 2023-04-26 13:20:07,657 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=12507.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:20:25,976 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.380e+02 2.009e+02 2.317e+02 2.704e+02 5.500e+02, threshold=4.634e+02, percent-clipped=2.0 2023-04-26 13:20:26,048 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=12520.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:21:02,673 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=12550.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:21:12,628 INFO [finetune.py:976] (1/7) Epoch 3, batch 1100, loss[loss=0.2482, simple_loss=0.2973, pruned_loss=0.09959, over 4136.00 frames. ], tot_loss[loss=0.2447, simple_loss=0.2983, pruned_loss=0.09557, over 950919.40 frames. ], batch size: 65, lr: 3.98e-03, grad_scale: 64.0 2023-04-26 13:21:25,842 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=12576.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:21:44,455 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=12604.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:21:46,055 INFO [finetune.py:976] (1/7) Epoch 3, batch 1150, loss[loss=0.266, simple_loss=0.3212, pruned_loss=0.1055, over 4759.00 frames. ], tot_loss[loss=0.2452, simple_loss=0.2992, pruned_loss=0.09565, over 951088.66 frames. ], batch size: 54, lr: 3.98e-03, grad_scale: 64.0 2023-04-26 13:21:55,572 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.367e+02 1.845e+02 2.160e+02 2.613e+02 5.486e+02, threshold=4.320e+02, percent-clipped=1.0 2023-04-26 13:21:58,098 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=12624.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:22:19,251 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=12640.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:22:34,206 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=12652.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:22:36,635 INFO [finetune.py:976] (1/7) Epoch 3, batch 1200, loss[loss=0.1643, simple_loss=0.2147, pruned_loss=0.05693, over 4213.00 frames. ], tot_loss[loss=0.2433, simple_loss=0.2969, pruned_loss=0.09486, over 951874.52 frames. ], batch size: 18, lr: 3.98e-03, grad_scale: 64.0 2023-04-26 13:22:55,147 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5626, 1.8888, 1.8683, 2.1683, 2.0028, 2.1603, 1.7396, 3.6660], device='cuda:1'), covar=tensor([0.0649, 0.0696, 0.0747, 0.1085, 0.0594, 0.0476, 0.0685, 0.0221], device='cuda:1'), in_proj_covar=tensor([0.0041, 0.0041, 0.0042, 0.0047, 0.0042, 0.0042, 0.0041, 0.0065], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0013, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0017], device='cuda:1') 2023-04-26 13:22:57,552 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=12688.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:23:09,934 INFO [finetune.py:976] (1/7) Epoch 3, batch 1250, loss[loss=0.2473, simple_loss=0.2901, pruned_loss=0.1023, over 4672.00 frames. ], tot_loss[loss=0.2413, simple_loss=0.2946, pruned_loss=0.09403, over 952752.29 frames. ], batch size: 23, lr: 3.98e-03, grad_scale: 64.0 2023-04-26 13:23:19,412 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.393e+02 1.929e+02 2.277e+02 2.741e+02 5.638e+02, threshold=4.554e+02, percent-clipped=3.0 2023-04-26 13:23:26,757 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=12732.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:23:42,758 INFO [finetune.py:976] (1/7) Epoch 3, batch 1300, loss[loss=0.2045, simple_loss=0.2606, pruned_loss=0.07419, over 4905.00 frames. ], tot_loss[loss=0.2375, simple_loss=0.2906, pruned_loss=0.09222, over 955236.56 frames. ], batch size: 46, lr: 3.98e-03, grad_scale: 64.0 2023-04-26 13:23:47,595 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=12763.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:24:09,813 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=12780.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:24:35,997 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=12802.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:24:39,386 INFO [finetune.py:976] (1/7) Epoch 3, batch 1350, loss[loss=0.2284, simple_loss=0.2925, pruned_loss=0.08218, over 4816.00 frames. ], tot_loss[loss=0.2393, simple_loss=0.292, pruned_loss=0.09326, over 955377.74 frames. ], batch size: 38, lr: 3.98e-03, grad_scale: 64.0 2023-04-26 13:24:48,915 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.232e+02 1.880e+02 2.141e+02 2.633e+02 4.297e+02, threshold=4.283e+02, percent-clipped=1.0 2023-04-26 13:24:51,945 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=12824.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 13:25:09,587 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.54 vs. limit=2.0 2023-04-26 13:25:12,234 INFO [finetune.py:976] (1/7) Epoch 3, batch 1400, loss[loss=0.2026, simple_loss=0.2577, pruned_loss=0.07379, over 4788.00 frames. ], tot_loss[loss=0.2402, simple_loss=0.2938, pruned_loss=0.09329, over 955558.42 frames. ], batch size: 25, lr: 3.98e-03, grad_scale: 64.0 2023-04-26 13:25:56,166 INFO [finetune.py:976] (1/7) Epoch 3, batch 1450, loss[loss=0.2754, simple_loss=0.3267, pruned_loss=0.1121, over 4821.00 frames. ], tot_loss[loss=0.2419, simple_loss=0.2961, pruned_loss=0.09381, over 954476.20 frames. ], batch size: 38, lr: 3.98e-03, grad_scale: 64.0 2023-04-26 13:26:17,005 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.305e+02 1.925e+02 2.434e+02 2.951e+02 7.906e+02, threshold=4.868e+02, percent-clipped=4.0 2023-04-26 13:27:02,917 INFO [finetune.py:976] (1/7) Epoch 3, batch 1500, loss[loss=0.2307, simple_loss=0.2928, pruned_loss=0.08433, over 4717.00 frames. ], tot_loss[loss=0.2454, simple_loss=0.2992, pruned_loss=0.09575, over 956456.91 frames. ], batch size: 54, lr: 3.98e-03, grad_scale: 64.0 2023-04-26 13:27:05,543 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.5125, 1.6597, 1.6186, 2.0936, 2.3369, 2.1131, 1.9263, 1.8136], device='cuda:1'), covar=tensor([0.2079, 0.2335, 0.2739, 0.2103, 0.1695, 0.2356, 0.2998, 0.2235], device='cuda:1'), in_proj_covar=tensor([0.0321, 0.0337, 0.0349, 0.0313, 0.0348, 0.0362, 0.0320, 0.0352], device='cuda:1'), out_proj_covar=tensor([6.9832e-05, 7.2333e-05, 7.5535e-05, 6.5644e-05, 7.3993e-05, 7.9110e-05, 6.9704e-05, 7.6081e-05], device='cuda:1') 2023-04-26 13:27:40,500 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.0820, 0.7309, 0.8859, 0.7053, 1.2625, 0.9450, 0.7850, 0.9901], device='cuda:1'), covar=tensor([0.1998, 0.1859, 0.2456, 0.1867, 0.1201, 0.1802, 0.2292, 0.2338], device='cuda:1'), in_proj_covar=tensor([0.0320, 0.0336, 0.0348, 0.0311, 0.0347, 0.0361, 0.0319, 0.0351], device='cuda:1'), out_proj_covar=tensor([6.9679e-05, 7.2143e-05, 7.5337e-05, 6.5307e-05, 7.3843e-05, 7.8795e-05, 6.9507e-05, 7.5821e-05], device='cuda:1') 2023-04-26 13:27:41,571 INFO [finetune.py:976] (1/7) Epoch 3, batch 1550, loss[loss=0.2224, simple_loss=0.2775, pruned_loss=0.0836, over 4751.00 frames. ], tot_loss[loss=0.2448, simple_loss=0.2989, pruned_loss=0.09531, over 957893.52 frames. ], batch size: 27, lr: 3.98e-03, grad_scale: 64.0 2023-04-26 13:28:02,727 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.270e+02 2.021e+02 2.296e+02 2.726e+02 4.661e+02, threshold=4.592e+02, percent-clipped=0.0 2023-04-26 13:28:47,302 INFO [finetune.py:976] (1/7) Epoch 3, batch 1600, loss[loss=0.1608, simple_loss=0.2307, pruned_loss=0.04546, over 4892.00 frames. ], tot_loss[loss=0.241, simple_loss=0.2947, pruned_loss=0.09363, over 958913.71 frames. ], batch size: 35, lr: 3.98e-03, grad_scale: 64.0 2023-04-26 13:29:17,236 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9641, 2.2554, 0.9448, 1.2743, 1.5965, 1.2100, 2.4073, 1.3867], device='cuda:1'), covar=tensor([0.0624, 0.0655, 0.0660, 0.1049, 0.0411, 0.0894, 0.0269, 0.0593], device='cuda:1'), in_proj_covar=tensor([0.0055, 0.0073, 0.0053, 0.0050, 0.0055, 0.0056, 0.0085, 0.0054], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0008, 0.0008, 0.0008, 0.0008, 0.0011, 0.0008], device='cuda:1') 2023-04-26 13:29:23,838 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=13102.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:29:26,215 INFO [finetune.py:976] (1/7) Epoch 3, batch 1650, loss[loss=0.1962, simple_loss=0.2528, pruned_loss=0.06977, over 4910.00 frames. ], tot_loss[loss=0.2398, simple_loss=0.2926, pruned_loss=0.09349, over 957072.73 frames. ], batch size: 43, lr: 3.98e-03, grad_scale: 64.0 2023-04-26 13:29:34,729 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=13119.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 13:29:35,229 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.220e+02 1.899e+02 2.180e+02 2.609e+02 5.027e+02, threshold=4.359e+02, percent-clipped=1.0 2023-04-26 13:29:41,086 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.5470, 1.7889, 1.6203, 1.8060, 1.6641, 1.9013, 1.6888, 1.6471], device='cuda:1'), covar=tensor([1.6594, 2.7854, 2.1800, 1.7898, 2.0865, 3.1866, 2.9262, 2.4491], device='cuda:1'), in_proj_covar=tensor([0.0300, 0.0400, 0.0316, 0.0323, 0.0350, 0.0403, 0.0386, 0.0345], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-26 13:29:55,913 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=13150.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:29:59,490 INFO [finetune.py:976] (1/7) Epoch 3, batch 1700, loss[loss=0.2498, simple_loss=0.2918, pruned_loss=0.1039, over 4860.00 frames. ], tot_loss[loss=0.2381, simple_loss=0.291, pruned_loss=0.09261, over 959561.25 frames. ], batch size: 49, lr: 3.98e-03, grad_scale: 64.0 2023-04-26 13:30:10,979 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.6396, 1.8930, 1.6863, 1.8740, 1.7300, 1.9882, 1.8037, 1.7580], device='cuda:1'), covar=tensor([1.3893, 2.6833, 2.1355, 1.7957, 2.0810, 2.9722, 2.7297, 2.3101], device='cuda:1'), in_proj_covar=tensor([0.0299, 0.0400, 0.0316, 0.0323, 0.0350, 0.0402, 0.0385, 0.0345], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-26 13:30:33,505 INFO [finetune.py:976] (1/7) Epoch 3, batch 1750, loss[loss=0.2706, simple_loss=0.3281, pruned_loss=0.1065, over 4821.00 frames. ], tot_loss[loss=0.2398, simple_loss=0.2927, pruned_loss=0.09342, over 959476.05 frames. ], batch size: 33, lr: 3.98e-03, grad_scale: 64.0 2023-04-26 13:30:43,120 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.228e+02 1.834e+02 2.276e+02 2.863e+02 4.766e+02, threshold=4.553e+02, percent-clipped=3.0 2023-04-26 13:30:49,003 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.3225, 1.0182, 1.5606, 1.4113, 1.1546, 0.9409, 1.2039, 0.8926], device='cuda:1'), covar=tensor([0.0957, 0.1133, 0.0692, 0.0989, 0.1143, 0.1931, 0.0841, 0.1135], device='cuda:1'), in_proj_covar=tensor([0.0071, 0.0080, 0.0077, 0.0072, 0.0086, 0.0098, 0.0089, 0.0081], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003], device='cuda:1') 2023-04-26 13:31:02,753 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-04-26 13:31:07,404 INFO [finetune.py:976] (1/7) Epoch 3, batch 1800, loss[loss=0.2291, simple_loss=0.296, pruned_loss=0.0811, over 4852.00 frames. ], tot_loss[loss=0.2426, simple_loss=0.296, pruned_loss=0.09458, over 956193.02 frames. ], batch size: 47, lr: 3.98e-03, grad_scale: 32.0 2023-04-26 13:31:23,292 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.52 vs. limit=2.0 2023-04-26 13:31:47,121 INFO [finetune.py:976] (1/7) Epoch 3, batch 1850, loss[loss=0.2512, simple_loss=0.3152, pruned_loss=0.09355, over 4882.00 frames. ], tot_loss[loss=0.2434, simple_loss=0.297, pruned_loss=0.09488, over 953989.26 frames. ], batch size: 43, lr: 3.98e-03, grad_scale: 16.0 2023-04-26 13:31:56,813 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.349e+02 1.904e+02 2.289e+02 2.801e+02 9.204e+02, threshold=4.579e+02, percent-clipped=2.0 2023-04-26 13:32:28,657 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-04-26 13:32:30,328 INFO [finetune.py:976] (1/7) Epoch 3, batch 1900, loss[loss=0.2127, simple_loss=0.2713, pruned_loss=0.07707, over 4860.00 frames. ], tot_loss[loss=0.243, simple_loss=0.2975, pruned_loss=0.09421, over 956581.80 frames. ], batch size: 44, lr: 3.98e-03, grad_scale: 16.0 2023-04-26 13:32:49,470 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([0.9168, 1.1759, 3.2939, 3.0539, 2.9916, 3.2065, 3.2174, 2.9295], device='cuda:1'), covar=tensor([0.7149, 0.5374, 0.1487, 0.2135, 0.1285, 0.1995, 0.1605, 0.1642], device='cuda:1'), in_proj_covar=tensor([0.0327, 0.0311, 0.0439, 0.0444, 0.0373, 0.0425, 0.0336, 0.0393], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0002], device='cuda:1') 2023-04-26 13:33:26,630 INFO [finetune.py:976] (1/7) Epoch 3, batch 1950, loss[loss=0.2431, simple_loss=0.2891, pruned_loss=0.09855, over 4747.00 frames. ], tot_loss[loss=0.2415, simple_loss=0.2956, pruned_loss=0.09367, over 956124.96 frames. ], batch size: 27, lr: 3.98e-03, grad_scale: 16.0 2023-04-26 13:33:39,985 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=13419.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:33:41,755 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.072e+02 1.766e+02 2.138e+02 2.602e+02 4.545e+02, threshold=4.277e+02, percent-clipped=0.0 2023-04-26 13:34:17,442 INFO [finetune.py:976] (1/7) Epoch 3, batch 2000, loss[loss=0.1695, simple_loss=0.229, pruned_loss=0.05497, over 4671.00 frames. ], tot_loss[loss=0.2377, simple_loss=0.2914, pruned_loss=0.092, over 955132.46 frames. ], batch size: 23, lr: 3.98e-03, grad_scale: 16.0 2023-04-26 13:34:20,383 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.2361, 3.4165, 1.2697, 1.6047, 1.7777, 2.4249, 2.0644, 1.0101], device='cuda:1'), covar=tensor([0.2029, 0.1708, 0.2402, 0.2165, 0.1522, 0.1566, 0.1807, 0.2239], device='cuda:1'), in_proj_covar=tensor([0.0124, 0.0269, 0.0150, 0.0131, 0.0142, 0.0165, 0.0128, 0.0132], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-26 13:34:24,601 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=13467.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:34:46,549 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.39 vs. limit=5.0 2023-04-26 13:34:51,090 INFO [finetune.py:976] (1/7) Epoch 3, batch 2050, loss[loss=0.2503, simple_loss=0.29, pruned_loss=0.1053, over 4676.00 frames. ], tot_loss[loss=0.2338, simple_loss=0.2875, pruned_loss=0.09006, over 956001.73 frames. ], batch size: 23, lr: 3.98e-03, grad_scale: 16.0 2023-04-26 13:35:01,245 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.294e+02 1.882e+02 2.225e+02 2.584e+02 5.744e+02, threshold=4.451e+02, percent-clipped=3.0 2023-04-26 13:35:15,792 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=5.12 vs. limit=5.0 2023-04-26 13:35:24,229 INFO [finetune.py:976] (1/7) Epoch 3, batch 2100, loss[loss=0.2303, simple_loss=0.2796, pruned_loss=0.0905, over 4766.00 frames. ], tot_loss[loss=0.2343, simple_loss=0.2878, pruned_loss=0.09038, over 957261.22 frames. ], batch size: 26, lr: 3.98e-03, grad_scale: 16.0 2023-04-26 13:35:25,040 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.62 vs. limit=5.0 2023-04-26 13:35:43,763 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6382, 2.2299, 1.5362, 1.4883, 1.1819, 1.2690, 1.6183, 1.1768], device='cuda:1'), covar=tensor([0.2188, 0.2021, 0.2295, 0.2712, 0.3429, 0.2548, 0.1737, 0.2789], device='cuda:1'), in_proj_covar=tensor([0.0198, 0.0226, 0.0191, 0.0216, 0.0230, 0.0195, 0.0186, 0.0207], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-26 13:35:47,653 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.3448, 1.0011, 0.3189, 1.0696, 1.0934, 1.2605, 1.1308, 1.1351], device='cuda:1'), covar=tensor([0.0613, 0.0504, 0.0544, 0.0652, 0.0339, 0.0599, 0.0593, 0.0708], device='cuda:1'), in_proj_covar=tensor([0.0032, 0.0027, 0.0025, 0.0031, 0.0022, 0.0031, 0.0030, 0.0032], device='cuda:1'), out_proj_covar=tensor([0.0049, 0.0044, 0.0038, 0.0049, 0.0038, 0.0048, 0.0047, 0.0050], device='cuda:1') 2023-04-26 13:35:50,103 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.2368, 2.6426, 1.1364, 1.2119, 2.2810, 1.1572, 3.5497, 1.6701], device='cuda:1'), covar=tensor([0.0698, 0.0615, 0.0870, 0.1410, 0.0468, 0.1073, 0.0247, 0.0686], device='cuda:1'), in_proj_covar=tensor([0.0056, 0.0073, 0.0054, 0.0050, 0.0055, 0.0056, 0.0085, 0.0053], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0008, 0.0008, 0.0008, 0.0008, 0.0011, 0.0008], device='cuda:1') 2023-04-26 13:35:51,964 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.71 vs. limit=2.0 2023-04-26 13:35:57,671 INFO [finetune.py:976] (1/7) Epoch 3, batch 2150, loss[loss=0.3187, simple_loss=0.3503, pruned_loss=0.1436, over 4166.00 frames. ], tot_loss[loss=0.2388, simple_loss=0.2926, pruned_loss=0.09256, over 954855.20 frames. ], batch size: 66, lr: 3.98e-03, grad_scale: 16.0 2023-04-26 13:36:07,907 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.340e+02 1.924e+02 2.294e+02 2.856e+02 4.572e+02, threshold=4.587e+02, percent-clipped=1.0 2023-04-26 13:36:30,854 INFO [finetune.py:976] (1/7) Epoch 3, batch 2200, loss[loss=0.2437, simple_loss=0.3105, pruned_loss=0.08843, over 4898.00 frames. ], tot_loss[loss=0.2404, simple_loss=0.2941, pruned_loss=0.09333, over 952128.98 frames. ], batch size: 35, lr: 3.98e-03, grad_scale: 16.0 2023-04-26 13:36:30,995 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.2372, 1.9955, 2.4627, 2.7118, 1.9000, 1.5549, 2.1937, 1.2497], device='cuda:1'), covar=tensor([0.0766, 0.1058, 0.0714, 0.0772, 0.0951, 0.1573, 0.0985, 0.1494], device='cuda:1'), in_proj_covar=tensor([0.0070, 0.0079, 0.0076, 0.0071, 0.0084, 0.0096, 0.0088, 0.0081], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003], device='cuda:1') 2023-04-26 13:37:08,555 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9230, 1.8639, 2.0290, 2.2683, 2.3032, 1.8041, 1.4342, 1.9871], device='cuda:1'), covar=tensor([0.1041, 0.1026, 0.0644, 0.0615, 0.0583, 0.1062, 0.1089, 0.0663], device='cuda:1'), in_proj_covar=tensor([0.0209, 0.0209, 0.0188, 0.0183, 0.0182, 0.0199, 0.0175, 0.0195], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 13:37:21,693 INFO [finetune.py:976] (1/7) Epoch 3, batch 2250, loss[loss=0.2257, simple_loss=0.2662, pruned_loss=0.09255, over 3961.00 frames. ], tot_loss[loss=0.2428, simple_loss=0.2968, pruned_loss=0.09441, over 953768.87 frames. ], batch size: 17, lr: 3.98e-03, grad_scale: 16.0 2023-04-26 13:37:31,818 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.241e+02 1.916e+02 2.386e+02 2.867e+02 7.645e+02, threshold=4.772e+02, percent-clipped=4.0 2023-04-26 13:37:43,787 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=13737.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:38:02,156 INFO [finetune.py:976] (1/7) Epoch 3, batch 2300, loss[loss=0.207, simple_loss=0.2614, pruned_loss=0.0763, over 4755.00 frames. ], tot_loss[loss=0.2409, simple_loss=0.2956, pruned_loss=0.09312, over 951970.71 frames. ], batch size: 28, lr: 3.98e-03, grad_scale: 16.0 2023-04-26 13:38:57,794 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=13798.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 13:39:05,806 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4643, 1.8579, 1.6574, 2.3358, 2.0201, 2.1293, 1.6588, 4.5513], device='cuda:1'), covar=tensor([0.0651, 0.0750, 0.0827, 0.1163, 0.0634, 0.0536, 0.0751, 0.0142], device='cuda:1'), in_proj_covar=tensor([0.0040, 0.0040, 0.0042, 0.0046, 0.0041, 0.0041, 0.0041, 0.0064], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0017], device='cuda:1') 2023-04-26 13:39:09,165 INFO [finetune.py:976] (1/7) Epoch 3, batch 2350, loss[loss=0.2328, simple_loss=0.2764, pruned_loss=0.09464, over 4871.00 frames. ], tot_loss[loss=0.2388, simple_loss=0.2929, pruned_loss=0.09231, over 951744.90 frames. ], batch size: 34, lr: 3.98e-03, grad_scale: 16.0 2023-04-26 13:39:37,984 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.317e+02 2.025e+02 2.480e+02 2.975e+02 5.089e+02, threshold=4.960e+02, percent-clipped=2.0 2023-04-26 13:40:20,272 INFO [finetune.py:976] (1/7) Epoch 3, batch 2400, loss[loss=0.2413, simple_loss=0.2729, pruned_loss=0.1048, over 4247.00 frames. ], tot_loss[loss=0.2349, simple_loss=0.2893, pruned_loss=0.09026, over 951831.81 frames. ], batch size: 65, lr: 3.98e-03, grad_scale: 16.0 2023-04-26 13:40:37,906 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.47 vs. limit=2.0 2023-04-26 13:40:51,165 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5487, 1.3034, 1.7316, 1.6803, 1.3864, 1.1011, 1.4688, 0.9579], device='cuda:1'), covar=tensor([0.0817, 0.1069, 0.0658, 0.0929, 0.1094, 0.1723, 0.1000, 0.1224], device='cuda:1'), in_proj_covar=tensor([0.0070, 0.0079, 0.0076, 0.0071, 0.0084, 0.0097, 0.0088, 0.0081], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003], device='cuda:1') 2023-04-26 13:40:57,681 INFO [finetune.py:976] (1/7) Epoch 3, batch 2450, loss[loss=0.1861, simple_loss=0.2534, pruned_loss=0.05943, over 4832.00 frames. ], tot_loss[loss=0.2342, simple_loss=0.2881, pruned_loss=0.09017, over 952842.69 frames. ], batch size: 47, lr: 3.98e-03, grad_scale: 16.0 2023-04-26 13:41:07,653 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([0.9808, 2.0722, 1.7327, 1.6480, 2.1604, 1.6799, 2.7046, 1.4640], device='cuda:1'), covar=tensor([0.4837, 0.1701, 0.6130, 0.3266, 0.1887, 0.2981, 0.1388, 0.5206], device='cuda:1'), in_proj_covar=tensor([0.0353, 0.0355, 0.0438, 0.0373, 0.0405, 0.0383, 0.0400, 0.0418], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 13:41:09,355 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.373e+02 2.035e+02 2.305e+02 2.759e+02 5.668e+02, threshold=4.609e+02, percent-clipped=2.0 2023-04-26 13:41:31,085 INFO [finetune.py:976] (1/7) Epoch 3, batch 2500, loss[loss=0.2264, simple_loss=0.2908, pruned_loss=0.081, over 4827.00 frames. ], tot_loss[loss=0.2369, simple_loss=0.2905, pruned_loss=0.0916, over 954076.33 frames. ], batch size: 33, lr: 3.98e-03, grad_scale: 16.0 2023-04-26 13:41:31,200 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0832, 1.5551, 1.8211, 2.4139, 2.5069, 1.8231, 1.5447, 2.0075], device='cuda:1'), covar=tensor([0.1170, 0.1510, 0.0957, 0.0702, 0.0661, 0.1311, 0.1247, 0.0850], device='cuda:1'), in_proj_covar=tensor([0.0212, 0.0212, 0.0191, 0.0185, 0.0185, 0.0202, 0.0177, 0.0198], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 13:41:44,336 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.58 vs. limit=2.0 2023-04-26 13:42:05,882 INFO [finetune.py:976] (1/7) Epoch 3, batch 2550, loss[loss=0.2294, simple_loss=0.294, pruned_loss=0.08241, over 4826.00 frames. ], tot_loss[loss=0.2395, simple_loss=0.2935, pruned_loss=0.09271, over 953562.59 frames. ], batch size: 51, lr: 3.98e-03, grad_scale: 16.0 2023-04-26 13:42:28,112 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.240e+02 1.992e+02 2.434e+02 2.841e+02 4.921e+02, threshold=4.868e+02, percent-clipped=1.0 2023-04-26 13:42:47,343 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9263, 1.8202, 2.0089, 2.2518, 2.1870, 1.7619, 1.5234, 1.9309], device='cuda:1'), covar=tensor([0.0903, 0.0956, 0.0609, 0.0532, 0.0637, 0.1071, 0.1064, 0.0605], device='cuda:1'), in_proj_covar=tensor([0.0210, 0.0211, 0.0190, 0.0184, 0.0184, 0.0201, 0.0175, 0.0196], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 13:43:10,324 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4192, 0.9142, 1.1753, 1.6988, 1.5496, 1.2692, 1.2918, 1.2643], device='cuda:1'), covar=tensor([2.5545, 3.4088, 3.8379, 4.6060, 2.7606, 4.1156, 4.0483, 3.1558], device='cuda:1'), in_proj_covar=tensor([0.0446, 0.0496, 0.0588, 0.0599, 0.0477, 0.0511, 0.0522, 0.0534], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 13:43:13,251 INFO [finetune.py:976] (1/7) Epoch 3, batch 2600, loss[loss=0.2225, simple_loss=0.2943, pruned_loss=0.07536, over 4806.00 frames. ], tot_loss[loss=0.2412, simple_loss=0.2955, pruned_loss=0.09343, over 954954.70 frames. ], batch size: 40, lr: 3.98e-03, grad_scale: 16.0 2023-04-26 13:43:37,583 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=14072.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:44:06,886 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=14093.0, num_to_drop=1, layers_to_drop={3} 2023-04-26 13:44:19,008 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=14103.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:44:25,130 INFO [finetune.py:976] (1/7) Epoch 3, batch 2650, loss[loss=0.2401, simple_loss=0.2971, pruned_loss=0.09161, over 4810.00 frames. ], tot_loss[loss=0.2414, simple_loss=0.2966, pruned_loss=0.09309, over 956262.59 frames. ], batch size: 40, lr: 3.98e-03, grad_scale: 16.0 2023-04-26 13:44:39,920 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=14120.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 13:44:47,141 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.256e+02 1.867e+02 2.323e+02 2.768e+02 1.192e+03, threshold=4.647e+02, percent-clipped=2.0 2023-04-26 13:45:00,931 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=14133.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:45:30,809 INFO [finetune.py:976] (1/7) Epoch 3, batch 2700, loss[loss=0.2387, simple_loss=0.289, pruned_loss=0.09421, over 4798.00 frames. ], tot_loss[loss=0.2391, simple_loss=0.2944, pruned_loss=0.0919, over 955757.49 frames. ], batch size: 29, lr: 3.98e-03, grad_scale: 16.0 2023-04-26 13:45:41,228 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=14164.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:45:51,238 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.1031, 2.1714, 1.8744, 1.8471, 2.3034, 1.7863, 2.8790, 1.6117], device='cuda:1'), covar=tensor([0.4536, 0.2063, 0.5085, 0.3374, 0.1978, 0.2866, 0.1423, 0.4792], device='cuda:1'), in_proj_covar=tensor([0.0355, 0.0358, 0.0439, 0.0375, 0.0406, 0.0384, 0.0401, 0.0420], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 13:46:03,643 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=14181.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 13:46:18,591 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4429, 1.7668, 1.5651, 2.0942, 1.8472, 2.1660, 1.5395, 4.3573], device='cuda:1'), covar=tensor([0.0609, 0.0759, 0.0766, 0.1178, 0.0642, 0.0546, 0.0773, 0.0107], device='cuda:1'), in_proj_covar=tensor([0.0040, 0.0041, 0.0042, 0.0046, 0.0042, 0.0041, 0.0041, 0.0064], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0017], device='cuda:1') 2023-04-26 13:46:30,714 INFO [finetune.py:976] (1/7) Epoch 3, batch 2750, loss[loss=0.187, simple_loss=0.2506, pruned_loss=0.06172, over 4928.00 frames. ], tot_loss[loss=0.2375, simple_loss=0.2919, pruned_loss=0.09159, over 955067.11 frames. ], batch size: 33, lr: 3.98e-03, grad_scale: 16.0 2023-04-26 13:46:40,385 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.352e+02 1.823e+02 2.218e+02 2.622e+02 5.684e+02, threshold=4.435e+02, percent-clipped=2.0 2023-04-26 13:46:40,483 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.2141, 1.1880, 5.0840, 4.7085, 4.4008, 4.8200, 4.4611, 4.4956], device='cuda:1'), covar=tensor([0.6335, 0.6501, 0.0979, 0.1730, 0.0986, 0.0965, 0.1343, 0.1478], device='cuda:1'), in_proj_covar=tensor([0.0325, 0.0309, 0.0435, 0.0441, 0.0369, 0.0421, 0.0333, 0.0389], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-26 13:47:03,042 INFO [finetune.py:976] (1/7) Epoch 3, batch 2800, loss[loss=0.2233, simple_loss=0.2775, pruned_loss=0.08458, over 4807.00 frames. ], tot_loss[loss=0.2354, simple_loss=0.289, pruned_loss=0.09089, over 954747.93 frames. ], batch size: 25, lr: 3.98e-03, grad_scale: 16.0 2023-04-26 13:47:24,645 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9873, 1.3143, 1.8058, 2.2234, 1.6929, 1.2817, 1.0788, 1.5581], device='cuda:1'), covar=tensor([0.5071, 0.6046, 0.2685, 0.4562, 0.5577, 0.4542, 0.7407, 0.5265], device='cuda:1'), in_proj_covar=tensor([0.0269, 0.0271, 0.0222, 0.0340, 0.0230, 0.0233, 0.0261, 0.0208], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-26 13:47:35,420 INFO [finetune.py:976] (1/7) Epoch 3, batch 2850, loss[loss=0.2307, simple_loss=0.2767, pruned_loss=0.09238, over 4240.00 frames. ], tot_loss[loss=0.2334, simple_loss=0.2872, pruned_loss=0.0898, over 955318.82 frames. ], batch size: 18, lr: 3.98e-03, grad_scale: 16.0 2023-04-26 13:47:40,132 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8860, 1.3313, 1.7364, 1.9775, 1.6007, 1.3148, 0.9415, 1.4311], device='cuda:1'), covar=tensor([0.4704, 0.5622, 0.2637, 0.4425, 0.5358, 0.4399, 0.7768, 0.4963], device='cuda:1'), in_proj_covar=tensor([0.0271, 0.0272, 0.0224, 0.0342, 0.0231, 0.0235, 0.0262, 0.0209], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-26 13:47:45,429 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.273e+02 1.919e+02 2.285e+02 2.703e+02 8.098e+02, threshold=4.570e+02, percent-clipped=1.0 2023-04-26 13:48:06,253 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.14 vs. limit=2.0 2023-04-26 13:48:08,420 INFO [finetune.py:976] (1/7) Epoch 3, batch 2900, loss[loss=0.2592, simple_loss=0.3326, pruned_loss=0.09293, over 4819.00 frames. ], tot_loss[loss=0.2379, simple_loss=0.2918, pruned_loss=0.09201, over 953703.36 frames. ], batch size: 38, lr: 3.98e-03, grad_scale: 16.0 2023-04-26 13:48:33,678 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=14393.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:48:41,650 INFO [finetune.py:976] (1/7) Epoch 3, batch 2950, loss[loss=0.2928, simple_loss=0.342, pruned_loss=0.1218, over 4820.00 frames. ], tot_loss[loss=0.2423, simple_loss=0.2967, pruned_loss=0.09396, over 954291.86 frames. ], batch size: 39, lr: 3.98e-03, grad_scale: 16.0 2023-04-26 13:48:46,046 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=14413.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:48:50,181 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.2835, 2.6639, 2.6511, 3.1000, 2.9315, 3.1726, 2.6746, 4.9334], device='cuda:1'), covar=tensor([0.0469, 0.0524, 0.0582, 0.0871, 0.0441, 0.0346, 0.0549, 0.0141], device='cuda:1'), in_proj_covar=tensor([0.0040, 0.0041, 0.0042, 0.0046, 0.0042, 0.0041, 0.0041, 0.0064], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0013, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0017], device='cuda:1') 2023-04-26 13:48:51,910 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.499e+02 2.004e+02 2.390e+02 2.807e+02 6.198e+02, threshold=4.780e+02, percent-clipped=3.0 2023-04-26 13:48:55,646 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=14428.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:48:56,368 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.76 vs. limit=5.0 2023-04-26 13:49:04,518 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=14441.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:49:14,493 INFO [finetune.py:976] (1/7) Epoch 3, batch 3000, loss[loss=0.2648, simple_loss=0.3216, pruned_loss=0.104, over 4917.00 frames. ], tot_loss[loss=0.241, simple_loss=0.2956, pruned_loss=0.09324, over 953937.11 frames. ], batch size: 38, lr: 3.98e-03, grad_scale: 16.0 2023-04-26 13:49:14,494 INFO [finetune.py:1001] (1/7) Computing validation loss 2023-04-26 13:49:19,632 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4699, 1.1509, 1.2354, 1.1458, 1.7086, 1.3468, 1.1179, 1.2160], device='cuda:1'), covar=tensor([0.1895, 0.1516, 0.2243, 0.1821, 0.0921, 0.1536, 0.2307, 0.2069], device='cuda:1'), in_proj_covar=tensor([0.0318, 0.0335, 0.0346, 0.0309, 0.0346, 0.0357, 0.0317, 0.0349], device='cuda:1'), out_proj_covar=tensor([6.9186e-05, 7.1943e-05, 7.4821e-05, 6.4723e-05, 7.3581e-05, 7.7871e-05, 6.8976e-05, 7.5434e-05], device='cuda:1') 2023-04-26 13:49:25,023 INFO [finetune.py:1010] (1/7) Epoch 3, validation: loss=0.1699, simple_loss=0.2433, pruned_loss=0.04821, over 2265189.00 frames. 2023-04-26 13:49:25,023 INFO [finetune.py:1011] (1/7) Maximum memory allocated so far is 6391MB 2023-04-26 13:49:26,925 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=14459.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:49:27,543 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.1507, 4.5794, 1.4936, 2.3738, 2.6014, 3.1527, 2.8917, 1.4269], device='cuda:1'), covar=tensor([0.1455, 0.1854, 0.2187, 0.1669, 0.1132, 0.1360, 0.1444, 0.1969], device='cuda:1'), in_proj_covar=tensor([0.0123, 0.0267, 0.0149, 0.0131, 0.0141, 0.0164, 0.0127, 0.0130], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-26 13:49:36,030 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=14474.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:49:37,146 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=14476.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 13:49:44,310 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=14487.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:50:13,201 INFO [finetune.py:976] (1/7) Epoch 3, batch 3050, loss[loss=0.2049, simple_loss=0.2721, pruned_loss=0.06885, over 4773.00 frames. ], tot_loss[loss=0.2418, simple_loss=0.2967, pruned_loss=0.09349, over 953819.98 frames. ], batch size: 28, lr: 3.98e-03, grad_scale: 16.0 2023-04-26 13:50:24,171 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.447e+02 2.059e+02 2.417e+02 2.833e+02 5.136e+02, threshold=4.834e+02, percent-clipped=1.0 2023-04-26 13:50:52,309 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=14548.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:50:53,739 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-04-26 13:51:03,204 INFO [finetune.py:976] (1/7) Epoch 3, batch 3100, loss[loss=0.1871, simple_loss=0.2538, pruned_loss=0.06019, over 4885.00 frames. ], tot_loss[loss=0.239, simple_loss=0.2939, pruned_loss=0.09206, over 954784.27 frames. ], batch size: 32, lr: 3.98e-03, grad_scale: 16.0 2023-04-26 13:52:09,681 INFO [finetune.py:976] (1/7) Epoch 3, batch 3150, loss[loss=0.2388, simple_loss=0.3013, pruned_loss=0.08812, over 4768.00 frames. ], tot_loss[loss=0.2362, simple_loss=0.291, pruned_loss=0.09075, over 955396.71 frames. ], batch size: 26, lr: 3.98e-03, grad_scale: 16.0 2023-04-26 13:52:30,357 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.439e+02 2.002e+02 2.381e+02 3.061e+02 6.553e+02, threshold=4.761e+02, percent-clipped=3.0 2023-04-26 13:52:51,772 INFO [finetune.py:976] (1/7) Epoch 3, batch 3200, loss[loss=0.2425, simple_loss=0.2697, pruned_loss=0.1076, over 4826.00 frames. ], tot_loss[loss=0.2326, simple_loss=0.287, pruned_loss=0.08913, over 955431.33 frames. ], batch size: 30, lr: 3.98e-03, grad_scale: 16.0 2023-04-26 13:53:25,517 INFO [finetune.py:976] (1/7) Epoch 3, batch 3250, loss[loss=0.1903, simple_loss=0.247, pruned_loss=0.06679, over 4746.00 frames. ], tot_loss[loss=0.2329, simple_loss=0.2872, pruned_loss=0.08932, over 956518.78 frames. ], batch size: 26, lr: 3.98e-03, grad_scale: 16.0 2023-04-26 13:53:37,682 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.274e+02 1.942e+02 2.271e+02 2.763e+02 5.504e+02, threshold=4.542e+02, percent-clipped=2.0 2023-04-26 13:53:40,157 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4725, 1.6157, 1.4125, 1.9654, 1.7985, 2.1796, 1.4754, 4.4460], device='cuda:1'), covar=tensor([0.0648, 0.0796, 0.0864, 0.1274, 0.0697, 0.0573, 0.0822, 0.0140], device='cuda:1'), in_proj_covar=tensor([0.0040, 0.0041, 0.0042, 0.0046, 0.0041, 0.0041, 0.0041, 0.0064], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0013, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0017], device='cuda:1') 2023-04-26 13:53:41,980 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=14728.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:53:49,359 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=14740.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:53:59,447 INFO [finetune.py:976] (1/7) Epoch 3, batch 3300, loss[loss=0.2258, simple_loss=0.2814, pruned_loss=0.08509, over 4832.00 frames. ], tot_loss[loss=0.2361, simple_loss=0.2909, pruned_loss=0.0906, over 953880.98 frames. ], batch size: 30, lr: 3.98e-03, grad_scale: 16.0 2023-04-26 13:54:01,389 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=14759.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:54:09,431 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=14769.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:54:14,673 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=14776.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:54:14,721 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=14776.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 13:54:30,604 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=14801.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:54:33,538 INFO [finetune.py:976] (1/7) Epoch 3, batch 3350, loss[loss=0.2507, simple_loss=0.3061, pruned_loss=0.09762, over 4831.00 frames. ], tot_loss[loss=0.2379, simple_loss=0.2933, pruned_loss=0.09124, over 955370.54 frames. ], batch size: 47, lr: 3.98e-03, grad_scale: 16.0 2023-04-26 13:54:34,180 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=14807.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:54:41,264 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([3.9453, 3.7546, 2.8215, 4.4902, 3.9230, 3.9540, 1.6884, 3.7750], device='cuda:1'), covar=tensor([0.1601, 0.1244, 0.2665, 0.1478, 0.3232, 0.1573, 0.6012, 0.2151], device='cuda:1'), in_proj_covar=tensor([0.0256, 0.0228, 0.0269, 0.0320, 0.0314, 0.0264, 0.0282, 0.0281], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-26 13:54:44,575 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.421e+02 1.946e+02 2.255e+02 2.721e+02 6.180e+02, threshold=4.510e+02, percent-clipped=3.0 2023-04-26 13:54:46,352 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=14824.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 13:54:58,940 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=14843.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:55:13,043 INFO [finetune.py:976] (1/7) Epoch 3, batch 3400, loss[loss=0.2398, simple_loss=0.291, pruned_loss=0.09434, over 4808.00 frames. ], tot_loss[loss=0.2405, simple_loss=0.2953, pruned_loss=0.09286, over 955468.50 frames. ], batch size: 40, lr: 3.98e-03, grad_scale: 16.0 2023-04-26 13:55:45,799 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7130, 2.4618, 1.6261, 1.6096, 1.2344, 1.2613, 1.7905, 1.1934], device='cuda:1'), covar=tensor([0.2210, 0.2155, 0.2430, 0.2955, 0.3588, 0.2660, 0.1791, 0.2853], device='cuda:1'), in_proj_covar=tensor([0.0196, 0.0223, 0.0187, 0.0213, 0.0226, 0.0192, 0.0182, 0.0202], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-26 13:56:06,887 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-04-26 13:56:08,499 INFO [finetune.py:976] (1/7) Epoch 3, batch 3450, loss[loss=0.1884, simple_loss=0.2428, pruned_loss=0.06698, over 4673.00 frames. ], tot_loss[loss=0.2383, simple_loss=0.2934, pruned_loss=0.09156, over 954877.21 frames. ], batch size: 23, lr: 3.98e-03, grad_scale: 16.0 2023-04-26 13:56:18,850 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.286e+02 1.992e+02 2.282e+02 2.746e+02 5.093e+02, threshold=4.564e+02, percent-clipped=2.0 2023-04-26 13:56:20,659 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0318, 2.4657, 1.0577, 1.2855, 1.8622, 1.2081, 3.0551, 1.5251], device='cuda:1'), covar=tensor([0.0680, 0.0568, 0.0777, 0.1241, 0.0506, 0.1020, 0.0295, 0.0670], device='cuda:1'), in_proj_covar=tensor([0.0055, 0.0072, 0.0053, 0.0050, 0.0055, 0.0056, 0.0084, 0.0053], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0008, 0.0008, 0.0008, 0.0008, 0.0011, 0.0008], device='cuda:1') 2023-04-26 13:56:42,394 INFO [finetune.py:976] (1/7) Epoch 3, batch 3500, loss[loss=0.2238, simple_loss=0.2834, pruned_loss=0.08213, over 4755.00 frames. ], tot_loss[loss=0.2354, simple_loss=0.2902, pruned_loss=0.09027, over 953985.79 frames. ], batch size: 59, lr: 3.98e-03, grad_scale: 16.0 2023-04-26 13:56:52,733 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.0817, 1.4412, 1.1492, 1.3999, 1.2630, 1.1944, 1.2635, 1.0689], device='cuda:1'), covar=tensor([0.2578, 0.2117, 0.1747, 0.1947, 0.3541, 0.2028, 0.2155, 0.2809], device='cuda:1'), in_proj_covar=tensor([0.0318, 0.0336, 0.0246, 0.0308, 0.0319, 0.0289, 0.0279, 0.0301], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 13:57:21,305 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 2023-04-26 13:57:38,374 INFO [finetune.py:976] (1/7) Epoch 3, batch 3550, loss[loss=0.2244, simple_loss=0.2783, pruned_loss=0.08528, over 4815.00 frames. ], tot_loss[loss=0.2338, simple_loss=0.2877, pruned_loss=0.08989, over 954318.09 frames. ], batch size: 25, lr: 3.98e-03, grad_scale: 16.0 2023-04-26 13:57:54,035 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.153e+02 1.782e+02 2.249e+02 2.630e+02 4.772e+02, threshold=4.498e+02, percent-clipped=1.0 2023-04-26 13:58:28,220 INFO [finetune.py:976] (1/7) Epoch 3, batch 3600, loss[loss=0.2707, simple_loss=0.3189, pruned_loss=0.1113, over 4739.00 frames. ], tot_loss[loss=0.2326, simple_loss=0.2861, pruned_loss=0.08954, over 954986.34 frames. ], batch size: 59, lr: 3.98e-03, grad_scale: 16.0 2023-04-26 13:58:36,829 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=15069.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:58:50,339 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([0.6465, 1.8066, 1.6021, 1.4421, 1.8257, 1.3892, 2.2722, 1.3193], device='cuda:1'), covar=tensor([0.4047, 0.1560, 0.5359, 0.2898, 0.1744, 0.2534, 0.1400, 0.4670], device='cuda:1'), in_proj_covar=tensor([0.0356, 0.0359, 0.0440, 0.0374, 0.0408, 0.0385, 0.0403, 0.0421], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 13:58:50,417 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.53 vs. limit=2.0 2023-04-26 13:58:56,108 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=15096.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:58:56,196 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6653, 1.5439, 1.8939, 1.8627, 1.7545, 1.6100, 1.7838, 1.7470], device='cuda:1'), covar=tensor([2.8997, 3.7138, 4.6848, 5.0191, 3.0199, 4.8066, 4.9911, 3.8265], device='cuda:1'), in_proj_covar=tensor([0.0446, 0.0496, 0.0588, 0.0597, 0.0479, 0.0510, 0.0524, 0.0532], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 13:59:02,216 INFO [finetune.py:976] (1/7) Epoch 3, batch 3650, loss[loss=0.1927, simple_loss=0.2604, pruned_loss=0.06252, over 4867.00 frames. ], tot_loss[loss=0.2359, simple_loss=0.2895, pruned_loss=0.0911, over 953124.29 frames. ], batch size: 31, lr: 3.98e-03, grad_scale: 16.0 2023-04-26 13:59:09,046 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=15117.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:59:12,569 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.417e+02 2.035e+02 2.405e+02 3.001e+02 5.155e+02, threshold=4.810e+02, percent-clipped=2.0 2023-04-26 13:59:28,039 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=15143.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:59:36,245 INFO [finetune.py:976] (1/7) Epoch 3, batch 3700, loss[loss=0.2288, simple_loss=0.2841, pruned_loss=0.08673, over 4895.00 frames. ], tot_loss[loss=0.2393, simple_loss=0.294, pruned_loss=0.09233, over 953612.59 frames. ], batch size: 35, lr: 3.98e-03, grad_scale: 16.0 2023-04-26 13:59:43,137 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-04-26 13:59:47,144 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9386, 1.2595, 1.8635, 2.3559, 1.6846, 1.3410, 1.1176, 1.5948], device='cuda:1'), covar=tensor([0.5417, 0.6718, 0.3104, 0.4826, 0.5892, 0.4813, 0.7701, 0.5295], device='cuda:1'), in_proj_covar=tensor([0.0272, 0.0272, 0.0223, 0.0341, 0.0230, 0.0234, 0.0260, 0.0208], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-26 13:59:59,374 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=15191.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 14:00:02,739 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=15195.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 14:00:10,003 INFO [finetune.py:976] (1/7) Epoch 3, batch 3750, loss[loss=0.2231, simple_loss=0.2877, pruned_loss=0.0793, over 4722.00 frames. ], tot_loss[loss=0.2373, simple_loss=0.2925, pruned_loss=0.0911, over 951015.69 frames. ], batch size: 54, lr: 3.98e-03, grad_scale: 16.0 2023-04-26 14:00:10,817 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.14 vs. limit=2.0 2023-04-26 14:00:24,836 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.429e+02 1.976e+02 2.277e+02 2.735e+02 5.558e+02, threshold=4.554e+02, percent-clipped=2.0 2023-04-26 14:00:36,577 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6839, 2.0034, 1.1378, 1.2724, 2.2015, 1.5198, 1.3571, 1.4636], device='cuda:1'), covar=tensor([0.0565, 0.0386, 0.0411, 0.0611, 0.0262, 0.0565, 0.0572, 0.0663], device='cuda:1'), in_proj_covar=tensor([0.0031, 0.0027, 0.0025, 0.0031, 0.0022, 0.0031, 0.0030, 0.0032], device='cuda:1'), out_proj_covar=tensor([0.0048, 0.0044, 0.0038, 0.0049, 0.0037, 0.0048, 0.0047, 0.0050], device='cuda:1') 2023-04-26 14:00:40,863 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4834, 0.9256, 0.5985, 1.1409, 1.1428, 1.3971, 1.2496, 1.2276], device='cuda:1'), covar=tensor([0.0575, 0.0464, 0.0461, 0.0628, 0.0349, 0.0566, 0.0547, 0.0698], device='cuda:1'), in_proj_covar=tensor([0.0031, 0.0027, 0.0025, 0.0031, 0.0022, 0.0031, 0.0030, 0.0032], device='cuda:1'), out_proj_covar=tensor([0.0048, 0.0044, 0.0038, 0.0049, 0.0037, 0.0048, 0.0047, 0.0050], device='cuda:1') 2023-04-26 14:00:55,185 INFO [finetune.py:976] (1/7) Epoch 3, batch 3800, loss[loss=0.2049, simple_loss=0.2631, pruned_loss=0.07334, over 4807.00 frames. ], tot_loss[loss=0.2374, simple_loss=0.2929, pruned_loss=0.09091, over 951134.30 frames. ], batch size: 25, lr: 3.98e-03, grad_scale: 16.0 2023-04-26 14:00:55,294 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=15256.0, num_to_drop=1, layers_to_drop={3} 2023-04-26 14:01:06,965 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-04-26 14:01:25,034 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5561, 2.0870, 1.8048, 1.9565, 1.6492, 1.6931, 1.8200, 1.4731], device='cuda:1'), covar=tensor([0.2533, 0.1514, 0.1079, 0.1524, 0.3404, 0.1539, 0.2048, 0.2684], device='cuda:1'), in_proj_covar=tensor([0.0318, 0.0336, 0.0246, 0.0309, 0.0320, 0.0290, 0.0278, 0.0301], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 14:01:29,009 INFO [finetune.py:976] (1/7) Epoch 3, batch 3850, loss[loss=0.2631, simple_loss=0.3089, pruned_loss=0.1087, over 4808.00 frames. ], tot_loss[loss=0.2345, simple_loss=0.2906, pruned_loss=0.08923, over 952852.31 frames. ], batch size: 40, lr: 3.98e-03, grad_scale: 32.0 2023-04-26 14:01:38,750 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.117e+02 1.876e+02 2.244e+02 2.688e+02 1.462e+03, threshold=4.488e+02, percent-clipped=2.0 2023-04-26 14:01:40,704 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6171, 1.5029, 0.7112, 1.1838, 1.8159, 1.4640, 1.3123, 1.3428], device='cuda:1'), covar=tensor([0.0580, 0.0443, 0.0469, 0.0615, 0.0282, 0.0585, 0.0564, 0.0669], device='cuda:1'), in_proj_covar=tensor([0.0031, 0.0027, 0.0025, 0.0031, 0.0022, 0.0031, 0.0030, 0.0032], device='cuda:1'), out_proj_covar=tensor([0.0049, 0.0044, 0.0038, 0.0049, 0.0037, 0.0048, 0.0047, 0.0050], device='cuda:1') 2023-04-26 14:02:01,731 INFO [finetune.py:976] (1/7) Epoch 3, batch 3900, loss[loss=0.2587, simple_loss=0.3022, pruned_loss=0.1076, over 4908.00 frames. ], tot_loss[loss=0.2333, simple_loss=0.2886, pruned_loss=0.08899, over 954862.05 frames. ], batch size: 37, lr: 3.98e-03, grad_scale: 32.0 2023-04-26 14:02:50,622 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=15396.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 14:02:51,444 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.59 vs. limit=5.0 2023-04-26 14:03:03,039 INFO [finetune.py:976] (1/7) Epoch 3, batch 3950, loss[loss=0.1731, simple_loss=0.2324, pruned_loss=0.05685, over 4751.00 frames. ], tot_loss[loss=0.2283, simple_loss=0.284, pruned_loss=0.08629, over 956573.24 frames. ], batch size: 27, lr: 3.98e-03, grad_scale: 32.0 2023-04-26 14:03:24,417 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.293e+02 1.749e+02 2.086e+02 2.518e+02 3.780e+02, threshold=4.171e+02, percent-clipped=0.0 2023-04-26 14:03:54,146 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=15444.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 14:04:08,310 INFO [finetune.py:976] (1/7) Epoch 3, batch 4000, loss[loss=0.2198, simple_loss=0.2818, pruned_loss=0.07892, over 4839.00 frames. ], tot_loss[loss=0.2296, simple_loss=0.2842, pruned_loss=0.08748, over 953774.66 frames. ], batch size: 47, lr: 3.98e-03, grad_scale: 32.0 2023-04-26 14:05:04,587 INFO [finetune.py:976] (1/7) Epoch 3, batch 4050, loss[loss=0.2232, simple_loss=0.2793, pruned_loss=0.08352, over 4896.00 frames. ], tot_loss[loss=0.2325, simple_loss=0.2881, pruned_loss=0.08844, over 956328.51 frames. ], batch size: 35, lr: 3.98e-03, grad_scale: 32.0 2023-04-26 14:05:08,775 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-04-26 14:05:15,729 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.134e+02 1.986e+02 2.201e+02 2.641e+02 4.376e+02, threshold=4.402e+02, percent-clipped=3.0 2023-04-26 14:05:33,867 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=15551.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 14:05:36,855 INFO [finetune.py:976] (1/7) Epoch 3, batch 4100, loss[loss=0.2669, simple_loss=0.3139, pruned_loss=0.11, over 4182.00 frames. ], tot_loss[loss=0.2356, simple_loss=0.2913, pruned_loss=0.08988, over 954156.99 frames. ], batch size: 66, lr: 3.98e-03, grad_scale: 32.0 2023-04-26 14:05:49,418 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9485, 1.6891, 2.2744, 2.3255, 1.7158, 1.4466, 1.9718, 1.0428], device='cuda:1'), covar=tensor([0.1116, 0.1298, 0.0707, 0.1170, 0.1496, 0.1778, 0.1017, 0.1551], device='cuda:1'), in_proj_covar=tensor([0.0069, 0.0079, 0.0076, 0.0071, 0.0084, 0.0097, 0.0087, 0.0081], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003], device='cuda:1') 2023-04-26 14:05:52,052 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-04-26 14:06:10,673 INFO [finetune.py:976] (1/7) Epoch 3, batch 4150, loss[loss=0.2274, simple_loss=0.2853, pruned_loss=0.08477, over 4748.00 frames. ], tot_loss[loss=0.2361, simple_loss=0.2923, pruned_loss=0.08994, over 954989.64 frames. ], batch size: 27, lr: 3.98e-03, grad_scale: 32.0 2023-04-26 14:06:14,841 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.74 vs. limit=5.0 2023-04-26 14:06:23,083 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.264e+02 1.923e+02 2.257e+02 2.741e+02 5.009e+02, threshold=4.514e+02, percent-clipped=1.0 2023-04-26 14:06:44,251 INFO [finetune.py:976] (1/7) Epoch 3, batch 4200, loss[loss=0.2282, simple_loss=0.2892, pruned_loss=0.08361, over 4788.00 frames. ], tot_loss[loss=0.2361, simple_loss=0.2927, pruned_loss=0.08974, over 954415.66 frames. ], batch size: 51, lr: 3.98e-03, grad_scale: 32.0 2023-04-26 14:07:17,986 INFO [finetune.py:976] (1/7) Epoch 3, batch 4250, loss[loss=0.2404, simple_loss=0.2957, pruned_loss=0.09251, over 4748.00 frames. ], tot_loss[loss=0.2334, simple_loss=0.2898, pruned_loss=0.08853, over 955942.87 frames. ], batch size: 23, lr: 3.98e-03, grad_scale: 32.0 2023-04-26 14:07:27,346 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6614, 1.2493, 4.5759, 4.2304, 3.9620, 4.2956, 4.1978, 3.9829], device='cuda:1'), covar=tensor([0.6917, 0.6483, 0.1038, 0.1828, 0.1127, 0.1677, 0.1235, 0.1571], device='cuda:1'), in_proj_covar=tensor([0.0325, 0.0310, 0.0434, 0.0439, 0.0369, 0.0424, 0.0332, 0.0390], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-26 14:07:29,536 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.151e+02 1.926e+02 2.195e+02 2.674e+02 5.290e+02, threshold=4.390e+02, percent-clipped=1.0 2023-04-26 14:07:50,581 INFO [finetune.py:976] (1/7) Epoch 3, batch 4300, loss[loss=0.1938, simple_loss=0.2524, pruned_loss=0.06757, over 4833.00 frames. ], tot_loss[loss=0.2299, simple_loss=0.286, pruned_loss=0.08686, over 955944.40 frames. ], batch size: 33, lr: 3.98e-03, grad_scale: 32.0 2023-04-26 14:08:39,474 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.79 vs. limit=2.0 2023-04-26 14:08:59,709 INFO [finetune.py:976] (1/7) Epoch 3, batch 4350, loss[loss=0.2375, simple_loss=0.2964, pruned_loss=0.08931, over 4837.00 frames. ], tot_loss[loss=0.2273, simple_loss=0.2832, pruned_loss=0.08573, over 955107.80 frames. ], batch size: 47, lr: 3.98e-03, grad_scale: 32.0 2023-04-26 14:09:07,907 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.33 vs. limit=5.0 2023-04-26 14:09:21,815 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.118e+02 1.851e+02 2.117e+02 2.553e+02 4.240e+02, threshold=4.233e+02, percent-clipped=0.0 2023-04-26 14:10:01,712 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=15851.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 14:10:04,717 INFO [finetune.py:976] (1/7) Epoch 3, batch 4400, loss[loss=0.2264, simple_loss=0.2827, pruned_loss=0.08508, over 4754.00 frames. ], tot_loss[loss=0.2283, simple_loss=0.2839, pruned_loss=0.08639, over 953782.63 frames. ], batch size: 27, lr: 3.97e-03, grad_scale: 32.0 2023-04-26 14:10:14,161 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9088, 1.6236, 1.9089, 2.1226, 2.1058, 1.8088, 1.9139, 1.8050], device='cuda:1'), covar=tensor([1.7225, 2.1990, 2.7173, 2.9557, 1.6737, 2.7519, 2.5457, 2.0394], device='cuda:1'), in_proj_covar=tensor([0.0446, 0.0494, 0.0585, 0.0596, 0.0476, 0.0507, 0.0520, 0.0530], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 14:10:39,618 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=15883.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 14:10:48,472 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.1252, 1.4737, 1.4283, 1.7315, 1.5011, 1.8435, 1.4967, 3.5612], device='cuda:1'), covar=tensor([0.0738, 0.0856, 0.0842, 0.1307, 0.0738, 0.0580, 0.0788, 0.0163], device='cuda:1'), in_proj_covar=tensor([0.0040, 0.0040, 0.0042, 0.0046, 0.0042, 0.0041, 0.0040, 0.0064], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0017], device='cuda:1') 2023-04-26 14:10:49,609 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=15899.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 14:10:53,840 INFO [finetune.py:976] (1/7) Epoch 3, batch 4450, loss[loss=0.2446, simple_loss=0.324, pruned_loss=0.08256, over 4822.00 frames. ], tot_loss[loss=0.2325, simple_loss=0.2884, pruned_loss=0.08829, over 951756.16 frames. ], batch size: 40, lr: 3.97e-03, grad_scale: 32.0 2023-04-26 14:10:59,477 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.55 vs. limit=2.0 2023-04-26 14:11:04,060 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.361e+02 2.045e+02 2.492e+02 3.065e+02 6.089e+02, threshold=4.985e+02, percent-clipped=5.0 2023-04-26 14:11:20,092 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=15944.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 14:11:22,477 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0920, 2.6870, 1.1868, 1.3137, 2.2355, 1.1204, 3.5610, 1.5943], device='cuda:1'), covar=tensor([0.0726, 0.0717, 0.0898, 0.1338, 0.0496, 0.1103, 0.0234, 0.0679], device='cuda:1'), in_proj_covar=tensor([0.0055, 0.0072, 0.0053, 0.0050, 0.0055, 0.0055, 0.0084, 0.0053], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0008, 0.0008, 0.0008, 0.0008, 0.0011, 0.0008], device='cuda:1') 2023-04-26 14:11:27,286 INFO [finetune.py:976] (1/7) Epoch 3, batch 4500, loss[loss=0.2752, simple_loss=0.3159, pruned_loss=0.1172, over 4822.00 frames. ], tot_loss[loss=0.2324, simple_loss=0.2887, pruned_loss=0.08801, over 952422.55 frames. ], batch size: 49, lr: 3.97e-03, grad_scale: 32.0 2023-04-26 14:11:34,157 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.1010, 0.6437, 0.9185, 0.7104, 1.2649, 0.9680, 0.7934, 1.0359], device='cuda:1'), covar=tensor([0.2056, 0.1840, 0.2363, 0.1946, 0.1080, 0.1669, 0.2185, 0.2207], device='cuda:1'), in_proj_covar=tensor([0.0319, 0.0338, 0.0350, 0.0312, 0.0350, 0.0361, 0.0320, 0.0353], device='cuda:1'), out_proj_covar=tensor([6.9300e-05, 7.2568e-05, 7.5973e-05, 6.5314e-05, 7.4260e-05, 7.8730e-05, 6.9619e-05, 7.6228e-05], device='cuda:1') 2023-04-26 14:11:34,750 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.1581, 1.3170, 1.4776, 1.5485, 1.5774, 1.2063, 0.8110, 1.3398], device='cuda:1'), covar=tensor([0.1009, 0.1357, 0.0781, 0.0687, 0.0672, 0.1009, 0.1184, 0.0742], device='cuda:1'), in_proj_covar=tensor([0.0209, 0.0210, 0.0188, 0.0184, 0.0183, 0.0199, 0.0174, 0.0194], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 14:11:43,098 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.3732, 2.2339, 2.6043, 2.8561, 2.6579, 2.0205, 1.7182, 2.2495], device='cuda:1'), covar=tensor([0.1098, 0.1064, 0.0598, 0.0744, 0.0840, 0.1392, 0.1315, 0.0817], device='cuda:1'), in_proj_covar=tensor([0.0210, 0.0210, 0.0188, 0.0184, 0.0183, 0.0200, 0.0174, 0.0194], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 14:11:44,349 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7371, 1.3419, 1.3043, 1.4525, 2.0727, 1.6197, 1.3038, 1.2384], device='cuda:1'), covar=tensor([0.1633, 0.1748, 0.2204, 0.1557, 0.0793, 0.1907, 0.2436, 0.1830], device='cuda:1'), in_proj_covar=tensor([0.0320, 0.0339, 0.0351, 0.0313, 0.0351, 0.0361, 0.0321, 0.0354], device='cuda:1'), out_proj_covar=tensor([6.9410e-05, 7.2734e-05, 7.6134e-05, 6.5532e-05, 7.4458e-05, 7.8822e-05, 6.9783e-05, 7.6383e-05], device='cuda:1') 2023-04-26 14:11:53,586 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.7668, 1.8534, 1.6919, 1.9340, 1.7610, 2.0025, 1.7897, 1.7841], device='cuda:1'), covar=tensor([1.2700, 2.1673, 1.9263, 1.4600, 1.7073, 2.5471, 2.5631, 1.8909], device='cuda:1'), in_proj_covar=tensor([0.0309, 0.0405, 0.0320, 0.0326, 0.0353, 0.0411, 0.0390, 0.0346], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-26 14:11:59,086 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=16001.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 14:12:02,031 INFO [finetune.py:976] (1/7) Epoch 3, batch 4550, loss[loss=0.2511, simple_loss=0.3142, pruned_loss=0.09404, over 4833.00 frames. ], tot_loss[loss=0.2346, simple_loss=0.291, pruned_loss=0.0891, over 953166.09 frames. ], batch size: 47, lr: 3.97e-03, grad_scale: 32.0 2023-04-26 14:12:05,825 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4980, 1.0923, 1.3280, 1.1271, 1.7200, 1.3763, 1.0655, 1.2484], device='cuda:1'), covar=tensor([0.1993, 0.1688, 0.2241, 0.1830, 0.1006, 0.1675, 0.2584, 0.2185], device='cuda:1'), in_proj_covar=tensor([0.0318, 0.0338, 0.0350, 0.0311, 0.0349, 0.0359, 0.0318, 0.0352], device='cuda:1'), out_proj_covar=tensor([6.9024e-05, 7.2375e-05, 7.5804e-05, 6.5219e-05, 7.4148e-05, 7.8310e-05, 6.9219e-05, 7.6037e-05], device='cuda:1') 2023-04-26 14:12:11,702 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.282e+02 1.831e+02 2.180e+02 2.572e+02 4.192e+02, threshold=4.359e+02, percent-clipped=0.0 2023-04-26 14:12:18,133 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9170, 2.2614, 1.9306, 2.1872, 1.6650, 1.8388, 2.0287, 1.5565], device='cuda:1'), covar=tensor([0.2432, 0.1609, 0.1077, 0.1489, 0.3714, 0.1510, 0.2348, 0.3169], device='cuda:1'), in_proj_covar=tensor([0.0319, 0.0337, 0.0248, 0.0312, 0.0323, 0.0290, 0.0280, 0.0302], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 14:12:35,166 INFO [finetune.py:976] (1/7) Epoch 3, batch 4600, loss[loss=0.242, simple_loss=0.2858, pruned_loss=0.09912, over 4374.00 frames. ], tot_loss[loss=0.2323, simple_loss=0.2889, pruned_loss=0.0878, over 952343.62 frames. ], batch size: 66, lr: 3.97e-03, grad_scale: 32.0 2023-04-26 14:12:38,923 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=16062.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 14:12:50,362 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8626, 1.2987, 1.4462, 1.4544, 2.0104, 1.7077, 1.3269, 1.3761], device='cuda:1'), covar=tensor([0.1571, 0.1755, 0.2253, 0.1518, 0.0957, 0.1699, 0.2248, 0.2113], device='cuda:1'), in_proj_covar=tensor([0.0317, 0.0336, 0.0349, 0.0310, 0.0348, 0.0357, 0.0317, 0.0352], device='cuda:1'), out_proj_covar=tensor([6.8734e-05, 7.2072e-05, 7.5568e-05, 6.4903e-05, 7.3870e-05, 7.7861e-05, 6.9064e-05, 7.6054e-05], device='cuda:1') 2023-04-26 14:12:53,917 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=16085.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 14:13:06,149 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.70 vs. limit=2.0 2023-04-26 14:13:08,933 INFO [finetune.py:976] (1/7) Epoch 3, batch 4650, loss[loss=0.2441, simple_loss=0.2905, pruned_loss=0.09885, over 4847.00 frames. ], tot_loss[loss=0.2313, simple_loss=0.287, pruned_loss=0.08784, over 952640.04 frames. ], batch size: 49, lr: 3.97e-03, grad_scale: 32.0 2023-04-26 14:13:18,702 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.295e+02 1.883e+02 2.189e+02 2.586e+02 3.625e+02, threshold=4.377e+02, percent-clipped=0.0 2023-04-26 14:13:47,582 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=16146.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 14:13:49,899 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=16149.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 14:13:53,995 INFO [finetune.py:976] (1/7) Epoch 3, batch 4700, loss[loss=0.1835, simple_loss=0.2391, pruned_loss=0.06401, over 4707.00 frames. ], tot_loss[loss=0.2274, simple_loss=0.2832, pruned_loss=0.08582, over 953574.78 frames. ], batch size: 23, lr: 3.97e-03, grad_scale: 32.0 2023-04-26 14:14:05,674 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8352, 1.6546, 1.8794, 2.1037, 1.6645, 1.2837, 1.8494, 1.1966], device='cuda:1'), covar=tensor([0.0938, 0.0717, 0.0806, 0.0765, 0.0952, 0.1339, 0.0680, 0.1267], device='cuda:1'), in_proj_covar=tensor([0.0070, 0.0079, 0.0078, 0.0071, 0.0084, 0.0099, 0.0088, 0.0081], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003], device='cuda:1') 2023-04-26 14:14:38,783 INFO [finetune.py:976] (1/7) Epoch 3, batch 4750, loss[loss=0.2124, simple_loss=0.2768, pruned_loss=0.07406, over 4842.00 frames. ], tot_loss[loss=0.225, simple_loss=0.2807, pruned_loss=0.08466, over 952648.98 frames. ], batch size: 44, lr: 3.97e-03, grad_scale: 32.0 2023-04-26 14:14:47,176 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=16210.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 14:15:00,398 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.300e+02 1.819e+02 2.117e+02 2.422e+02 6.375e+02, threshold=4.235e+02, percent-clipped=3.0 2023-04-26 14:15:10,808 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=16239.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 14:15:25,743 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9990, 2.4873, 2.1241, 2.2568, 1.8845, 2.0628, 2.2078, 1.7344], device='cuda:1'), covar=tensor([0.2198, 0.1426, 0.1069, 0.1618, 0.2818, 0.1425, 0.2032, 0.2962], device='cuda:1'), in_proj_covar=tensor([0.0318, 0.0337, 0.0247, 0.0310, 0.0322, 0.0289, 0.0278, 0.0301], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 14:15:35,035 INFO [finetune.py:976] (1/7) Epoch 3, batch 4800, loss[loss=0.2798, simple_loss=0.3259, pruned_loss=0.1168, over 4902.00 frames. ], tot_loss[loss=0.2278, simple_loss=0.2835, pruned_loss=0.086, over 954320.23 frames. ], batch size: 37, lr: 3.97e-03, grad_scale: 32.0 2023-04-26 14:15:48,509 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0827, 1.4322, 1.6368, 1.7200, 2.2633, 1.9732, 1.6689, 1.5223], device='cuda:1'), covar=tensor([0.1632, 0.2166, 0.2179, 0.1677, 0.1029, 0.1477, 0.2451, 0.2006], device='cuda:1'), in_proj_covar=tensor([0.0318, 0.0338, 0.0351, 0.0312, 0.0350, 0.0360, 0.0319, 0.0353], device='cuda:1'), out_proj_covar=tensor([6.8919e-05, 7.2465e-05, 7.6009e-05, 6.5415e-05, 7.4386e-05, 7.8456e-05, 6.9501e-05, 7.6278e-05], device='cuda:1') 2023-04-26 14:15:57,166 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2023-04-26 14:16:42,027 INFO [finetune.py:976] (1/7) Epoch 3, batch 4850, loss[loss=0.2494, simple_loss=0.3038, pruned_loss=0.09746, over 4824.00 frames. ], tot_loss[loss=0.2307, simple_loss=0.287, pruned_loss=0.08719, over 952179.48 frames. ], batch size: 33, lr: 3.97e-03, grad_scale: 32.0 2023-04-26 14:17:03,614 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.365e+02 1.946e+02 2.319e+02 2.886e+02 4.382e+02, threshold=4.637e+02, percent-clipped=1.0 2023-04-26 14:17:31,160 INFO [finetune.py:976] (1/7) Epoch 3, batch 4900, loss[loss=0.2419, simple_loss=0.305, pruned_loss=0.08943, over 4916.00 frames. ], tot_loss[loss=0.2336, simple_loss=0.2895, pruned_loss=0.08887, over 953522.44 frames. ], batch size: 42, lr: 3.97e-03, grad_scale: 32.0 2023-04-26 14:17:32,327 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=16357.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 14:17:42,340 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=16371.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 14:18:04,092 INFO [finetune.py:976] (1/7) Epoch 3, batch 4950, loss[loss=0.2062, simple_loss=0.2701, pruned_loss=0.0711, over 4135.00 frames. ], tot_loss[loss=0.2354, simple_loss=0.2922, pruned_loss=0.08937, over 955034.13 frames. ], batch size: 66, lr: 3.97e-03, grad_scale: 16.0 2023-04-26 14:18:16,362 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.286e+02 1.926e+02 2.230e+02 2.725e+02 6.217e+02, threshold=4.460e+02, percent-clipped=3.0 2023-04-26 14:18:21,897 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=16432.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 14:18:25,025 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.62 vs. limit=5.0 2023-04-26 14:18:27,290 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=16441.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 14:18:37,160 INFO [finetune.py:976] (1/7) Epoch 3, batch 5000, loss[loss=0.2669, simple_loss=0.3132, pruned_loss=0.1103, over 4879.00 frames. ], tot_loss[loss=0.2342, simple_loss=0.291, pruned_loss=0.08867, over 953616.87 frames. ], batch size: 32, lr: 3.97e-03, grad_scale: 16.0 2023-04-26 14:18:51,241 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9302, 2.2473, 1.0050, 1.2865, 1.4783, 1.2180, 2.4943, 1.4559], device='cuda:1'), covar=tensor([0.0703, 0.0613, 0.0689, 0.1217, 0.0505, 0.1007, 0.0355, 0.0643], device='cuda:1'), in_proj_covar=tensor([0.0054, 0.0072, 0.0053, 0.0050, 0.0055, 0.0055, 0.0084, 0.0053], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0008, 0.0008, 0.0008, 0.0008, 0.0011, 0.0008], device='cuda:1') 2023-04-26 14:19:09,883 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=16505.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 14:19:10,435 INFO [finetune.py:976] (1/7) Epoch 3, batch 5050, loss[loss=0.1834, simple_loss=0.2365, pruned_loss=0.06513, over 4752.00 frames. ], tot_loss[loss=0.2312, simple_loss=0.2871, pruned_loss=0.08768, over 951687.31 frames. ], batch size: 23, lr: 3.97e-03, grad_scale: 16.0 2023-04-26 14:19:14,779 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.53 vs. limit=2.0 2023-04-26 14:19:23,710 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.221e+02 1.760e+02 2.084e+02 2.475e+02 5.733e+02, threshold=4.169e+02, percent-clipped=2.0 2023-04-26 14:19:29,916 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9353, 2.5696, 1.8897, 1.6781, 1.4160, 1.4314, 1.9467, 1.4114], device='cuda:1'), covar=tensor([0.2239, 0.2098, 0.2186, 0.2855, 0.3268, 0.2546, 0.1596, 0.2734], device='cuda:1'), in_proj_covar=tensor([0.0194, 0.0222, 0.0185, 0.0211, 0.0223, 0.0191, 0.0180, 0.0201], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-26 14:19:33,539 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=16539.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 14:19:43,695 INFO [finetune.py:976] (1/7) Epoch 3, batch 5100, loss[loss=0.2534, simple_loss=0.2849, pruned_loss=0.1109, over 4761.00 frames. ], tot_loss[loss=0.2278, simple_loss=0.2833, pruned_loss=0.08612, over 952584.90 frames. ], batch size: 27, lr: 3.97e-03, grad_scale: 16.0 2023-04-26 14:19:57,651 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5313, 1.0115, 0.3674, 1.1987, 1.2754, 1.4242, 1.2750, 1.2499], device='cuda:1'), covar=tensor([0.0599, 0.0469, 0.0515, 0.0641, 0.0304, 0.0600, 0.0562, 0.0681], device='cuda:1'), in_proj_covar=tensor([0.0031, 0.0027, 0.0025, 0.0031, 0.0022, 0.0031, 0.0030, 0.0032], device='cuda:1'), out_proj_covar=tensor([0.0049, 0.0045, 0.0038, 0.0049, 0.0038, 0.0048, 0.0047, 0.0050], device='cuda:1') 2023-04-26 14:20:10,601 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=16587.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 14:20:11,412 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2023-04-26 14:20:33,131 INFO [finetune.py:976] (1/7) Epoch 3, batch 5150, loss[loss=0.2291, simple_loss=0.2982, pruned_loss=0.07994, over 4902.00 frames. ], tot_loss[loss=0.227, simple_loss=0.2828, pruned_loss=0.08562, over 953594.67 frames. ], batch size: 32, lr: 3.97e-03, grad_scale: 16.0 2023-04-26 14:20:34,478 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5822, 3.2000, 0.9332, 1.8299, 1.6996, 2.1749, 1.8367, 1.0083], device='cuda:1'), covar=tensor([0.1398, 0.1068, 0.2161, 0.1405, 0.1302, 0.1214, 0.1635, 0.2142], device='cuda:1'), in_proj_covar=tensor([0.0123, 0.0267, 0.0149, 0.0131, 0.0141, 0.0164, 0.0128, 0.0131], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-26 14:20:56,490 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.263e+02 1.945e+02 2.270e+02 2.709e+02 5.209e+02, threshold=4.540e+02, percent-clipped=2.0 2023-04-26 14:21:30,489 INFO [finetune.py:976] (1/7) Epoch 3, batch 5200, loss[loss=0.1948, simple_loss=0.2559, pruned_loss=0.06682, over 4749.00 frames. ], tot_loss[loss=0.2291, simple_loss=0.2863, pruned_loss=0.08602, over 953573.09 frames. ], batch size: 27, lr: 3.97e-03, grad_scale: 16.0 2023-04-26 14:21:31,187 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=16657.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 14:22:02,899 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=16705.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 14:22:03,452 INFO [finetune.py:976] (1/7) Epoch 3, batch 5250, loss[loss=0.1932, simple_loss=0.2652, pruned_loss=0.06054, over 4794.00 frames. ], tot_loss[loss=0.2304, simple_loss=0.2879, pruned_loss=0.08644, over 954987.58 frames. ], batch size: 29, lr: 3.97e-03, grad_scale: 16.0 2023-04-26 14:22:14,742 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.228e+02 1.971e+02 2.303e+02 2.786e+02 5.327e+02, threshold=4.606e+02, percent-clipped=1.0 2023-04-26 14:22:18,693 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=16727.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 14:22:21,626 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.2313, 2.1020, 2.4398, 2.8229, 2.7224, 2.0913, 1.9126, 2.2045], device='cuda:1'), covar=tensor([0.1044, 0.1082, 0.0650, 0.0686, 0.0658, 0.1175, 0.1138, 0.0765], device='cuda:1'), in_proj_covar=tensor([0.0209, 0.0212, 0.0188, 0.0184, 0.0183, 0.0200, 0.0174, 0.0194], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 14:22:27,619 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=16741.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 14:22:47,370 INFO [finetune.py:976] (1/7) Epoch 3, batch 5300, loss[loss=0.2601, simple_loss=0.3084, pruned_loss=0.1059, over 4758.00 frames. ], tot_loss[loss=0.2317, simple_loss=0.289, pruned_loss=0.08718, over 956184.15 frames. ], batch size: 54, lr: 3.97e-03, grad_scale: 16.0 2023-04-26 14:22:49,430 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.2431, 1.3520, 1.5422, 1.6549, 1.5928, 1.7482, 1.5824, 1.6518], device='cuda:1'), covar=tensor([0.9665, 1.5988, 1.4062, 1.1911, 1.4192, 2.2835, 1.7521, 1.4870], device='cuda:1'), in_proj_covar=tensor([0.0308, 0.0402, 0.0321, 0.0326, 0.0352, 0.0412, 0.0389, 0.0346], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-26 14:23:10,425 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=16789.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 14:23:20,214 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=16805.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 14:23:20,766 INFO [finetune.py:976] (1/7) Epoch 3, batch 5350, loss[loss=0.2161, simple_loss=0.273, pruned_loss=0.07963, over 4767.00 frames. ], tot_loss[loss=0.2312, simple_loss=0.2888, pruned_loss=0.08678, over 955073.27 frames. ], batch size: 26, lr: 3.97e-03, grad_scale: 16.0 2023-04-26 14:23:31,440 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.249e+02 1.831e+02 2.243e+02 2.677e+02 3.612e+02, threshold=4.486e+02, percent-clipped=0.0 2023-04-26 14:23:33,848 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.1958, 1.8517, 2.3899, 2.7489, 1.9082, 1.3748, 2.0022, 1.0278], device='cuda:1'), covar=tensor([0.0928, 0.0916, 0.0649, 0.0735, 0.1093, 0.2399, 0.1276, 0.1775], device='cuda:1'), in_proj_covar=tensor([0.0069, 0.0079, 0.0077, 0.0072, 0.0084, 0.0098, 0.0087, 0.0081], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003], device='cuda:1') 2023-04-26 14:23:52,043 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=16853.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 14:23:53,793 INFO [finetune.py:976] (1/7) Epoch 3, batch 5400, loss[loss=0.218, simple_loss=0.2666, pruned_loss=0.08469, over 4863.00 frames. ], tot_loss[loss=0.2287, simple_loss=0.2859, pruned_loss=0.08575, over 954750.41 frames. ], batch size: 31, lr: 3.97e-03, grad_scale: 16.0 2023-04-26 14:24:01,177 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=16868.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 14:24:22,070 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=3.95 vs. limit=5.0 2023-04-26 14:24:27,334 INFO [finetune.py:976] (1/7) Epoch 3, batch 5450, loss[loss=0.2148, simple_loss=0.2679, pruned_loss=0.08086, over 4829.00 frames. ], tot_loss[loss=0.2259, simple_loss=0.2825, pruned_loss=0.08467, over 954605.60 frames. ], batch size: 39, lr: 3.97e-03, grad_scale: 16.0 2023-04-26 14:24:27,440 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=16906.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 14:24:37,663 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.358e+02 1.908e+02 2.254e+02 2.695e+02 6.070e+02, threshold=4.507e+02, percent-clipped=3.0 2023-04-26 14:24:37,850 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-04-26 14:24:40,099 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.1020, 2.4143, 2.0796, 2.3191, 1.7976, 2.0965, 2.3083, 1.8324], device='cuda:1'), covar=tensor([0.1463, 0.0952, 0.0889, 0.0955, 0.2589, 0.1019, 0.1334, 0.2184], device='cuda:1'), in_proj_covar=tensor([0.0313, 0.0330, 0.0244, 0.0305, 0.0316, 0.0284, 0.0275, 0.0297], device='cuda:1'), out_proj_covar=tensor([1.2788e-04, 1.3470e-04, 9.9568e-05, 1.2291e-04, 1.3051e-04, 1.1479e-04, 1.1346e-04, 1.2040e-04], device='cuda:1') 2023-04-26 14:24:41,872 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=16929.0, num_to_drop=1, layers_to_drop={2} 2023-04-26 14:24:56,966 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8641, 1.6873, 4.4305, 4.1670, 3.9954, 4.2176, 4.1819, 3.9667], device='cuda:1'), covar=tensor([0.6394, 0.5600, 0.1085, 0.1682, 0.0915, 0.1658, 0.1110, 0.1566], device='cuda:1'), in_proj_covar=tensor([0.0323, 0.0311, 0.0432, 0.0436, 0.0368, 0.0421, 0.0329, 0.0390], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-26 14:25:00,558 INFO [finetune.py:976] (1/7) Epoch 3, batch 5500, loss[loss=0.164, simple_loss=0.2252, pruned_loss=0.0514, over 4695.00 frames. ], tot_loss[loss=0.2221, simple_loss=0.2783, pruned_loss=0.08297, over 952605.52 frames. ], batch size: 23, lr: 3.97e-03, grad_scale: 16.0 2023-04-26 14:25:03,659 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.1784, 1.9223, 2.4006, 2.4651, 1.8605, 1.5545, 2.0746, 1.1160], device='cuda:1'), covar=tensor([0.0781, 0.1092, 0.0698, 0.1246, 0.1231, 0.1788, 0.0999, 0.1489], device='cuda:1'), in_proj_covar=tensor([0.0069, 0.0078, 0.0076, 0.0071, 0.0083, 0.0097, 0.0086, 0.0080], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:1') 2023-04-26 14:25:07,413 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=16967.0, num_to_drop=1, layers_to_drop={2} 2023-04-26 14:25:33,761 INFO [finetune.py:976] (1/7) Epoch 3, batch 5550, loss[loss=0.1974, simple_loss=0.2736, pruned_loss=0.06059, over 4746.00 frames. ], tot_loss[loss=0.227, simple_loss=0.2828, pruned_loss=0.08565, over 953248.77 frames. ], batch size: 27, lr: 3.97e-03, grad_scale: 16.0 2023-04-26 14:25:54,585 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.318e+02 1.795e+02 2.241e+02 2.790e+02 4.152e+02, threshold=4.483e+02, percent-clipped=0.0 2023-04-26 14:26:02,724 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=17027.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 14:26:03,344 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8748, 1.2074, 4.9356, 4.5865, 4.3112, 4.6309, 4.3224, 4.3491], device='cuda:1'), covar=tensor([0.7151, 0.6419, 0.0972, 0.1774, 0.1152, 0.1558, 0.1548, 0.1612], device='cuda:1'), in_proj_covar=tensor([0.0324, 0.0311, 0.0432, 0.0437, 0.0368, 0.0421, 0.0329, 0.0390], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-26 14:26:35,595 INFO [finetune.py:976] (1/7) Epoch 3, batch 5600, loss[loss=0.1985, simple_loss=0.2548, pruned_loss=0.07108, over 4785.00 frames. ], tot_loss[loss=0.2292, simple_loss=0.2864, pruned_loss=0.086, over 954471.28 frames. ], batch size: 23, lr: 3.97e-03, grad_scale: 16.0 2023-04-26 14:26:59,003 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=17075.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 14:27:05,429 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([3.5421, 3.4945, 2.6790, 4.0986, 3.4879, 3.5775, 1.7053, 3.5028], device='cuda:1'), covar=tensor([0.1473, 0.1223, 0.3068, 0.1679, 0.2865, 0.1660, 0.5192, 0.2112], device='cuda:1'), in_proj_covar=tensor([0.0256, 0.0227, 0.0269, 0.0323, 0.0317, 0.0265, 0.0282, 0.0285], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-26 14:27:38,803 INFO [finetune.py:976] (1/7) Epoch 3, batch 5650, loss[loss=0.303, simple_loss=0.3554, pruned_loss=0.1253, over 4262.00 frames. ], tot_loss[loss=0.2314, simple_loss=0.2894, pruned_loss=0.08674, over 953722.93 frames. ], batch size: 65, lr: 3.97e-03, grad_scale: 16.0 2023-04-26 14:27:55,215 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.353e+02 1.900e+02 2.198e+02 2.594e+02 5.346e+02, threshold=4.396e+02, percent-clipped=1.0 2023-04-26 14:27:56,077 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=2.05 vs. limit=2.0 2023-04-26 14:28:15,588 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0640, 1.6835, 2.1741, 2.2386, 1.7768, 1.5494, 1.9031, 1.3090], device='cuda:1'), covar=tensor([0.0640, 0.1032, 0.0564, 0.0902, 0.0968, 0.1344, 0.0762, 0.1083], device='cuda:1'), in_proj_covar=tensor([0.0070, 0.0079, 0.0077, 0.0071, 0.0084, 0.0098, 0.0087, 0.0081], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003], device='cuda:1') 2023-04-26 14:28:26,722 INFO [finetune.py:976] (1/7) Epoch 3, batch 5700, loss[loss=0.2009, simple_loss=0.2431, pruned_loss=0.07936, over 4175.00 frames. ], tot_loss[loss=0.2276, simple_loss=0.2838, pruned_loss=0.08572, over 933475.09 frames. ], batch size: 18, lr: 3.97e-03, grad_scale: 16.0 2023-04-26 14:28:37,576 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([3.1076, 3.5437, 1.6910, 2.3201, 2.9108, 2.2399, 4.8357, 3.1004], device='cuda:1'), covar=tensor([0.0508, 0.0654, 0.0731, 0.1120, 0.0467, 0.0869, 0.0179, 0.0449], device='cuda:1'), in_proj_covar=tensor([0.0054, 0.0071, 0.0053, 0.0050, 0.0055, 0.0055, 0.0084, 0.0053], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0008, 0.0008, 0.0008, 0.0008, 0.0011, 0.0008], device='cuda:1') 2023-04-26 14:29:04,044 INFO [finetune.py:976] (1/7) Epoch 4, batch 0, loss[loss=0.2382, simple_loss=0.2976, pruned_loss=0.08945, over 4819.00 frames. ], tot_loss[loss=0.2382, simple_loss=0.2976, pruned_loss=0.08945, over 4819.00 frames. ], batch size: 33, lr: 3.97e-03, grad_scale: 16.0 2023-04-26 14:29:04,045 INFO [finetune.py:1001] (1/7) Computing validation loss 2023-04-26 14:29:11,661 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5547, 1.0843, 1.2894, 1.1773, 1.7602, 1.4118, 1.1175, 1.3192], device='cuda:1'), covar=tensor([0.1734, 0.1746, 0.2641, 0.1682, 0.1050, 0.1880, 0.3009, 0.2515], device='cuda:1'), in_proj_covar=tensor([0.0313, 0.0334, 0.0348, 0.0309, 0.0345, 0.0353, 0.0314, 0.0351], device='cuda:1'), out_proj_covar=tensor([6.7823e-05, 7.1689e-05, 7.5527e-05, 6.4718e-05, 7.3169e-05, 7.7032e-05, 6.8408e-05, 7.5909e-05], device='cuda:1') 2023-04-26 14:29:26,720 INFO [finetune.py:1010] (1/7) Epoch 4, validation: loss=0.1686, simple_loss=0.2415, pruned_loss=0.04788, over 2265189.00 frames. 2023-04-26 14:29:26,721 INFO [finetune.py:1011] (1/7) Maximum memory allocated so far is 6391MB 2023-04-26 14:29:31,433 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=17189.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 14:29:33,446 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.51 vs. limit=2.0 2023-04-26 14:29:52,773 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.151e+02 1.687e+02 2.032e+02 2.492e+02 4.364e+02, threshold=4.064e+02, percent-clipped=0.0 2023-04-26 14:29:53,450 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=17224.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 14:29:58,842 INFO [finetune.py:976] (1/7) Epoch 4, batch 50, loss[loss=0.2386, simple_loss=0.2826, pruned_loss=0.09732, over 4363.00 frames. ], tot_loss[loss=0.2327, simple_loss=0.29, pruned_loss=0.0877, over 217407.66 frames. ], batch size: 19, lr: 3.97e-03, grad_scale: 16.0 2023-04-26 14:30:01,112 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([3.3423, 3.2463, 2.4988, 3.7803, 3.2126, 3.2392, 1.3604, 3.1753], device='cuda:1'), covar=tensor([0.1562, 0.1210, 0.3324, 0.2326, 0.2240, 0.1774, 0.5268, 0.2357], device='cuda:1'), in_proj_covar=tensor([0.0251, 0.0223, 0.0264, 0.0317, 0.0311, 0.0261, 0.0277, 0.0279], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-26 14:30:11,276 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=17250.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 14:30:18,603 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=17262.0, num_to_drop=1, layers_to_drop={3} 2023-04-26 14:30:31,853 INFO [finetune.py:976] (1/7) Epoch 4, batch 100, loss[loss=0.2842, simple_loss=0.296, pruned_loss=0.1362, over 4305.00 frames. ], tot_loss[loss=0.2286, simple_loss=0.2844, pruned_loss=0.08634, over 380214.67 frames. ], batch size: 65, lr: 3.97e-03, grad_scale: 16.0 2023-04-26 14:30:43,184 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6321, 1.2596, 1.2704, 1.3015, 1.8721, 1.4664, 1.1741, 1.2534], device='cuda:1'), covar=tensor([0.1613, 0.1544, 0.2367, 0.1494, 0.0856, 0.1703, 0.2276, 0.2020], device='cuda:1'), in_proj_covar=tensor([0.0315, 0.0337, 0.0350, 0.0312, 0.0347, 0.0356, 0.0317, 0.0354], device='cuda:1'), out_proj_covar=tensor([6.8232e-05, 7.2207e-05, 7.5963e-05, 6.5354e-05, 7.3570e-05, 7.7620e-05, 6.8979e-05, 7.6482e-05], device='cuda:1') 2023-04-26 14:30:58,842 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.312e+02 1.837e+02 2.342e+02 2.858e+02 5.078e+02, threshold=4.684e+02, percent-clipped=3.0 2023-04-26 14:31:04,966 INFO [finetune.py:976] (1/7) Epoch 4, batch 150, loss[loss=0.212, simple_loss=0.2733, pruned_loss=0.0754, over 4846.00 frames. ], tot_loss[loss=0.223, simple_loss=0.2782, pruned_loss=0.08384, over 505676.51 frames. ], batch size: 44, lr: 3.97e-03, grad_scale: 16.0 2023-04-26 14:31:24,773 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=17362.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 14:31:28,972 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([3.4001, 3.4160, 2.4858, 3.8509, 3.3201, 3.3387, 1.3295, 3.2887], device='cuda:1'), covar=tensor([0.1972, 0.1290, 0.3326, 0.2392, 0.3700, 0.2131, 0.5873, 0.2530], device='cuda:1'), in_proj_covar=tensor([0.0251, 0.0223, 0.0263, 0.0317, 0.0310, 0.0261, 0.0277, 0.0279], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-26 14:31:36,402 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-04-26 14:31:38,033 INFO [finetune.py:976] (1/7) Epoch 4, batch 200, loss[loss=0.2416, simple_loss=0.2935, pruned_loss=0.0949, over 4757.00 frames. ], tot_loss[loss=0.2243, simple_loss=0.2783, pruned_loss=0.08514, over 605477.26 frames. ], batch size: 54, lr: 3.97e-03, grad_scale: 16.0 2023-04-26 14:31:38,165 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=17383.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 14:31:54,470 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.44 vs. limit=5.0 2023-04-26 14:32:05,046 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.188e+02 2.027e+02 2.314e+02 2.814e+02 1.019e+03, threshold=4.629e+02, percent-clipped=4.0 2023-04-26 14:32:05,198 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=17423.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 14:32:11,142 INFO [finetune.py:976] (1/7) Epoch 4, batch 250, loss[loss=0.2055, simple_loss=0.2727, pruned_loss=0.06911, over 4822.00 frames. ], tot_loss[loss=0.2259, simple_loss=0.2816, pruned_loss=0.08512, over 682639.04 frames. ], batch size: 33, lr: 3.97e-03, grad_scale: 16.0 2023-04-26 14:32:22,355 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9878, 1.3623, 1.7727, 2.0488, 1.6485, 1.3182, 0.9856, 1.4562], device='cuda:1'), covar=tensor([0.4834, 0.6059, 0.2716, 0.4583, 0.5554, 0.4446, 0.7318, 0.4933], device='cuda:1'), in_proj_covar=tensor([0.0275, 0.0271, 0.0225, 0.0342, 0.0229, 0.0235, 0.0258, 0.0205], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-26 14:32:25,191 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=17444.0, num_to_drop=1, layers_to_drop={3} 2023-04-26 14:33:09,712 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.31 vs. limit=5.0 2023-04-26 14:33:16,612 INFO [finetune.py:976] (1/7) Epoch 4, batch 300, loss[loss=0.2133, simple_loss=0.2762, pruned_loss=0.0752, over 4921.00 frames. ], tot_loss[loss=0.2302, simple_loss=0.2858, pruned_loss=0.08725, over 740468.39 frames. ], batch size: 33, lr: 3.97e-03, grad_scale: 16.0 2023-04-26 14:33:19,138 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0745, 1.4401, 1.9451, 2.2841, 1.7676, 1.3943, 1.2248, 1.6378], device='cuda:1'), covar=tensor([0.5019, 0.5867, 0.2761, 0.4731, 0.5389, 0.4449, 0.7031, 0.4971], device='cuda:1'), in_proj_covar=tensor([0.0276, 0.0272, 0.0226, 0.0343, 0.0230, 0.0236, 0.0259, 0.0206], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-26 14:33:50,255 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.75 vs. limit=2.0 2023-04-26 14:34:04,407 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.014e+02 1.938e+02 2.299e+02 2.709e+02 4.777e+02, threshold=4.598e+02, percent-clipped=1.0 2023-04-26 14:34:10,661 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=17524.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 14:34:22,587 INFO [finetune.py:976] (1/7) Epoch 4, batch 350, loss[loss=0.2778, simple_loss=0.3177, pruned_loss=0.1189, over 4885.00 frames. ], tot_loss[loss=0.2327, simple_loss=0.2884, pruned_loss=0.08852, over 787942.62 frames. ], batch size: 32, lr: 3.97e-03, grad_scale: 16.0 2023-04-26 14:34:36,263 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=17545.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 14:34:38,034 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.1212, 1.8695, 2.0549, 2.4338, 2.4341, 1.9606, 1.5347, 2.1262], device='cuda:1'), covar=tensor([0.0838, 0.1159, 0.0748, 0.0593, 0.0579, 0.0973, 0.1091, 0.0579], device='cuda:1'), in_proj_covar=tensor([0.0208, 0.0210, 0.0187, 0.0183, 0.0183, 0.0199, 0.0173, 0.0192], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 14:34:48,916 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6079, 1.2297, 4.2027, 3.9279, 3.7076, 3.8989, 3.8231, 3.6946], device='cuda:1'), covar=tensor([0.6863, 0.5791, 0.1002, 0.1688, 0.1107, 0.1326, 0.1962, 0.1380], device='cuda:1'), in_proj_covar=tensor([0.0319, 0.0310, 0.0429, 0.0432, 0.0366, 0.0417, 0.0327, 0.0386], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-26 14:34:59,104 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=17562.0, num_to_drop=1, layers_to_drop={2} 2023-04-26 14:35:10,345 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=17572.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 14:35:17,455 INFO [finetune.py:976] (1/7) Epoch 4, batch 400, loss[loss=0.2073, simple_loss=0.2696, pruned_loss=0.07254, over 4730.00 frames. ], tot_loss[loss=0.2325, simple_loss=0.2891, pruned_loss=0.08799, over 822844.98 frames. ], batch size: 54, lr: 3.97e-03, grad_scale: 16.0 2023-04-26 14:35:37,142 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=17610.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 14:35:45,056 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.309e+02 1.823e+02 2.162e+02 2.599e+02 8.047e+02, threshold=4.324e+02, percent-clipped=1.0 2023-04-26 14:35:45,248 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 2023-04-26 14:35:51,211 INFO [finetune.py:976] (1/7) Epoch 4, batch 450, loss[loss=0.2316, simple_loss=0.2863, pruned_loss=0.08847, over 4813.00 frames. ], tot_loss[loss=0.2291, simple_loss=0.2858, pruned_loss=0.08622, over 852287.37 frames. ], batch size: 30, lr: 3.97e-03, grad_scale: 16.0 2023-04-26 14:36:25,014 INFO [finetune.py:976] (1/7) Epoch 4, batch 500, loss[loss=0.2152, simple_loss=0.2719, pruned_loss=0.07926, over 4903.00 frames. ], tot_loss[loss=0.2276, simple_loss=0.2837, pruned_loss=0.08573, over 875983.89 frames. ], batch size: 43, lr: 3.97e-03, grad_scale: 16.0 2023-04-26 14:36:25,129 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.1276, 1.8902, 2.0889, 2.4379, 2.3815, 1.9268, 1.5770, 2.0591], device='cuda:1'), covar=tensor([0.0962, 0.1117, 0.0661, 0.0649, 0.0667, 0.1001, 0.1157, 0.0628], device='cuda:1'), in_proj_covar=tensor([0.0209, 0.0211, 0.0188, 0.0183, 0.0183, 0.0200, 0.0173, 0.0193], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 14:36:35,722 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=17700.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 14:36:45,505 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-04-26 14:36:49,493 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=17718.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 14:36:52,485 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.156e+02 1.778e+02 2.181e+02 2.733e+02 7.061e+02, threshold=4.362e+02, percent-clipped=3.0 2023-04-26 14:36:56,855 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=17730.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 14:36:58,589 INFO [finetune.py:976] (1/7) Epoch 4, batch 550, loss[loss=0.2958, simple_loss=0.3149, pruned_loss=0.1383, over 4056.00 frames. ], tot_loss[loss=0.2245, simple_loss=0.2802, pruned_loss=0.08438, over 893486.06 frames. ], batch size: 65, lr: 3.97e-03, grad_scale: 16.0 2023-04-26 14:37:02,831 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=17739.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 14:37:18,661 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=17761.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 14:37:43,915 INFO [finetune.py:976] (1/7) Epoch 4, batch 600, loss[loss=0.2021, simple_loss=0.2569, pruned_loss=0.07367, over 4763.00 frames. ], tot_loss[loss=0.2246, simple_loss=0.2805, pruned_loss=0.08434, over 904788.09 frames. ], batch size: 26, lr: 3.97e-03, grad_scale: 16.0 2023-04-26 14:37:49,321 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=17791.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 14:38:23,099 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.336e+02 1.937e+02 2.249e+02 2.763e+02 6.989e+02, threshold=4.497e+02, percent-clipped=2.0 2023-04-26 14:38:34,870 INFO [finetune.py:976] (1/7) Epoch 4, batch 650, loss[loss=0.2363, simple_loss=0.2854, pruned_loss=0.09362, over 4814.00 frames. ], tot_loss[loss=0.2286, simple_loss=0.2855, pruned_loss=0.08587, over 917274.60 frames. ], batch size: 30, lr: 3.97e-03, grad_scale: 16.0 2023-04-26 14:38:52,802 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=17845.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 14:39:22,791 INFO [finetune.py:976] (1/7) Epoch 4, batch 700, loss[loss=0.2372, simple_loss=0.2944, pruned_loss=0.08999, over 4737.00 frames. ], tot_loss[loss=0.2304, simple_loss=0.2874, pruned_loss=0.08668, over 926016.21 frames. ], batch size: 27, lr: 3.97e-03, grad_scale: 16.0 2023-04-26 14:39:28,941 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=17893.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 14:40:07,831 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.332e+02 1.898e+02 2.217e+02 2.664e+02 7.094e+02, threshold=4.434e+02, percent-clipped=3.0 2023-04-26 14:40:19,660 INFO [finetune.py:976] (1/7) Epoch 4, batch 750, loss[loss=0.2891, simple_loss=0.3305, pruned_loss=0.1239, over 4859.00 frames. ], tot_loss[loss=0.2303, simple_loss=0.2876, pruned_loss=0.08646, over 929257.26 frames. ], batch size: 31, lr: 3.97e-03, grad_scale: 16.0 2023-04-26 14:40:43,546 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9602, 1.1531, 1.2561, 1.4137, 1.3678, 1.5728, 1.3540, 1.3723], device='cuda:1'), covar=tensor([0.9838, 1.6249, 1.3323, 1.2390, 1.3870, 2.1680, 1.5909, 1.3631], device='cuda:1'), in_proj_covar=tensor([0.0307, 0.0403, 0.0321, 0.0326, 0.0352, 0.0413, 0.0388, 0.0343], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-26 14:40:53,858 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=17960.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 14:41:24,413 INFO [finetune.py:976] (1/7) Epoch 4, batch 800, loss[loss=0.1781, simple_loss=0.2471, pruned_loss=0.05461, over 4831.00 frames. ], tot_loss[loss=0.2289, simple_loss=0.2868, pruned_loss=0.08544, over 932809.60 frames. ], batch size: 33, lr: 3.97e-03, grad_scale: 16.0 2023-04-26 14:41:48,924 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=18018.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 14:41:51,294 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=18021.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 14:41:52,410 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.136e+02 1.866e+02 2.231e+02 2.895e+02 6.246e+02, threshold=4.462e+02, percent-clipped=2.0 2023-04-26 14:41:58,981 INFO [finetune.py:976] (1/7) Epoch 4, batch 850, loss[loss=0.2432, simple_loss=0.295, pruned_loss=0.09565, over 4930.00 frames. ], tot_loss[loss=0.2266, simple_loss=0.2844, pruned_loss=0.08443, over 937620.66 frames. ], batch size: 38, lr: 3.97e-03, grad_scale: 16.0 2023-04-26 14:41:59,852 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-04-26 14:42:02,670 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=18039.0, num_to_drop=1, layers_to_drop={2} 2023-04-26 14:42:07,591 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8158, 1.2840, 1.3259, 1.5100, 2.0493, 1.6500, 1.3987, 1.3287], device='cuda:1'), covar=tensor([0.1667, 0.1901, 0.2332, 0.1554, 0.0847, 0.2013, 0.2287, 0.1950], device='cuda:1'), in_proj_covar=tensor([0.0316, 0.0336, 0.0350, 0.0309, 0.0346, 0.0355, 0.0316, 0.0351], device='cuda:1'), out_proj_covar=tensor([6.8481e-05, 7.2017e-05, 7.5896e-05, 6.4644e-05, 7.3500e-05, 7.7404e-05, 6.8644e-05, 7.5854e-05], device='cuda:1') 2023-04-26 14:42:12,386 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.3275, 1.5432, 1.3449, 1.4640, 1.4335, 1.2880, 1.4092, 1.0030], device='cuda:1'), covar=tensor([0.1960, 0.1541, 0.1168, 0.1522, 0.3456, 0.1538, 0.1843, 0.2635], device='cuda:1'), in_proj_covar=tensor([0.0317, 0.0336, 0.0246, 0.0309, 0.0323, 0.0288, 0.0277, 0.0301], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 14:42:13,525 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=18056.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 14:42:20,042 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=18066.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 14:42:32,279 INFO [finetune.py:976] (1/7) Epoch 4, batch 900, loss[loss=0.1642, simple_loss=0.2324, pruned_loss=0.04798, over 4844.00 frames. ], tot_loss[loss=0.2247, simple_loss=0.2816, pruned_loss=0.08386, over 943103.37 frames. ], batch size: 33, lr: 3.97e-03, grad_scale: 16.0 2023-04-26 14:42:34,142 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=18086.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 14:42:34,719 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=18087.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 14:42:58,117 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.220e+02 1.952e+02 2.266e+02 2.638e+02 7.997e+02, threshold=4.531e+02, percent-clipped=4.0 2023-04-26 14:43:05,176 INFO [finetune.py:976] (1/7) Epoch 4, batch 950, loss[loss=0.1753, simple_loss=0.2502, pruned_loss=0.05023, over 4787.00 frames. ], tot_loss[loss=0.2227, simple_loss=0.2797, pruned_loss=0.08285, over 948177.18 frames. ], batch size: 25, lr: 3.97e-03, grad_scale: 16.0 2023-04-26 14:43:11,631 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-04-26 14:43:44,162 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=18174.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 14:43:44,768 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=18175.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 14:43:53,136 INFO [finetune.py:976] (1/7) Epoch 4, batch 1000, loss[loss=0.2001, simple_loss=0.2554, pruned_loss=0.07237, over 4764.00 frames. ], tot_loss[loss=0.226, simple_loss=0.2829, pruned_loss=0.08452, over 949435.25 frames. ], batch size: 27, lr: 3.97e-03, grad_scale: 16.0 2023-04-26 14:44:29,365 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7130, 1.2877, 1.2911, 1.4838, 1.9608, 1.6217, 1.3504, 1.2884], device='cuda:1'), covar=tensor([0.2075, 0.2221, 0.2621, 0.1881, 0.1092, 0.1727, 0.2710, 0.2138], device='cuda:1'), in_proj_covar=tensor([0.0316, 0.0337, 0.0351, 0.0310, 0.0347, 0.0354, 0.0316, 0.0351], device='cuda:1'), out_proj_covar=tensor([6.8392e-05, 7.2230e-05, 7.6017e-05, 6.4813e-05, 7.3599e-05, 7.7231e-05, 6.8635e-05, 7.5812e-05], device='cuda:1') 2023-04-26 14:44:30,497 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.196e+02 1.905e+02 2.183e+02 2.540e+02 5.721e+02, threshold=4.367e+02, percent-clipped=1.0 2023-04-26 14:44:38,563 INFO [finetune.py:976] (1/7) Epoch 4, batch 1050, loss[loss=0.2569, simple_loss=0.3277, pruned_loss=0.09311, over 4837.00 frames. ], tot_loss[loss=0.2268, simple_loss=0.2847, pruned_loss=0.08444, over 952181.02 frames. ], batch size: 47, lr: 3.97e-03, grad_scale: 16.0 2023-04-26 14:44:39,859 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=18235.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 14:44:40,467 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=18236.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 14:45:04,112 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.2081, 1.4657, 1.3805, 2.0580, 2.2907, 1.8998, 1.7985, 1.5811], device='cuda:1'), covar=tensor([0.2398, 0.2904, 0.2933, 0.1955, 0.1583, 0.2610, 0.3527, 0.2546], device='cuda:1'), in_proj_covar=tensor([0.0316, 0.0337, 0.0351, 0.0310, 0.0347, 0.0354, 0.0316, 0.0351], device='cuda:1'), out_proj_covar=tensor([6.8448e-05, 7.2335e-05, 7.6114e-05, 6.4899e-05, 7.3753e-05, 7.7256e-05, 6.8685e-05, 7.5870e-05], device='cuda:1') 2023-04-26 14:45:04,737 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0296, 1.4594, 1.9358, 2.2246, 1.8128, 1.4357, 1.0308, 1.6515], device='cuda:1'), covar=tensor([0.4615, 0.5402, 0.2372, 0.4621, 0.5031, 0.4021, 0.7076, 0.4402], device='cuda:1'), in_proj_covar=tensor([0.0274, 0.0269, 0.0223, 0.0340, 0.0227, 0.0234, 0.0254, 0.0203], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-26 14:45:12,450 INFO [finetune.py:976] (1/7) Epoch 4, batch 1100, loss[loss=0.2469, simple_loss=0.3078, pruned_loss=0.09295, over 4805.00 frames. ], tot_loss[loss=0.2292, simple_loss=0.2869, pruned_loss=0.0858, over 953622.13 frames. ], batch size: 41, lr: 3.97e-03, grad_scale: 16.0 2023-04-26 14:45:30,404 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-04-26 14:45:50,513 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=18316.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 14:46:00,501 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.371e+02 2.039e+02 2.386e+02 2.843e+02 4.542e+02, threshold=4.773e+02, percent-clipped=1.0 2023-04-26 14:46:08,158 INFO [finetune.py:976] (1/7) Epoch 4, batch 1150, loss[loss=0.2277, simple_loss=0.283, pruned_loss=0.08627, over 4816.00 frames. ], tot_loss[loss=0.2308, simple_loss=0.288, pruned_loss=0.08674, over 954685.47 frames. ], batch size: 40, lr: 3.97e-03, grad_scale: 16.0 2023-04-26 14:46:23,841 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=18356.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 14:46:52,607 INFO [finetune.py:976] (1/7) Epoch 4, batch 1200, loss[loss=0.2271, simple_loss=0.2846, pruned_loss=0.08481, over 4812.00 frames. ], tot_loss[loss=0.2291, simple_loss=0.2864, pruned_loss=0.08588, over 955543.81 frames. ], batch size: 40, lr: 3.97e-03, grad_scale: 16.0 2023-04-26 14:46:54,498 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-04-26 14:46:55,525 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=18386.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 14:47:18,677 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=18404.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 14:47:36,011 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.337e+02 1.888e+02 2.199e+02 2.566e+02 6.503e+02, threshold=4.398e+02, percent-clipped=1.0 2023-04-26 14:47:41,678 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-04-26 14:47:42,651 INFO [finetune.py:976] (1/7) Epoch 4, batch 1250, loss[loss=0.2003, simple_loss=0.2585, pruned_loss=0.07104, over 4899.00 frames. ], tot_loss[loss=0.2252, simple_loss=0.2825, pruned_loss=0.0839, over 956962.54 frames. ], batch size: 32, lr: 3.97e-03, grad_scale: 32.0 2023-04-26 14:47:43,325 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=18434.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 14:47:49,148 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 2023-04-26 14:48:01,610 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8649, 1.7166, 1.8630, 2.0203, 1.6856, 1.3094, 1.8068, 1.1669], device='cuda:1'), covar=tensor([0.1037, 0.0741, 0.0753, 0.0859, 0.0963, 0.1249, 0.0819, 0.1056], device='cuda:1'), in_proj_covar=tensor([0.0069, 0.0078, 0.0076, 0.0071, 0.0083, 0.0098, 0.0087, 0.0080], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:1') 2023-04-26 14:48:12,029 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.44 vs. limit=5.0 2023-04-26 14:48:15,942 INFO [finetune.py:976] (1/7) Epoch 4, batch 1300, loss[loss=0.2055, simple_loss=0.2524, pruned_loss=0.07933, over 3990.00 frames. ], tot_loss[loss=0.2217, simple_loss=0.2786, pruned_loss=0.08238, over 954963.15 frames. ], batch size: 17, lr: 3.97e-03, grad_scale: 32.0 2023-04-26 14:48:21,224 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=18490.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 14:48:34,019 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=18508.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 14:48:43,001 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.221e+02 1.895e+02 2.151e+02 2.623e+02 4.474e+02, threshold=4.301e+02, percent-clipped=1.0 2023-04-26 14:48:47,801 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=18530.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 14:48:48,425 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=18531.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 14:48:49,575 INFO [finetune.py:976] (1/7) Epoch 4, batch 1350, loss[loss=0.2703, simple_loss=0.328, pruned_loss=0.1063, over 4733.00 frames. ], tot_loss[loss=0.223, simple_loss=0.2795, pruned_loss=0.0833, over 952836.14 frames. ], batch size: 59, lr: 3.97e-03, grad_scale: 32.0 2023-04-26 14:49:08,061 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=18551.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 14:49:32,000 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=18569.0, num_to_drop=1, layers_to_drop={2} 2023-04-26 14:49:45,978 INFO [finetune.py:976] (1/7) Epoch 4, batch 1400, loss[loss=0.2672, simple_loss=0.3086, pruned_loss=0.1129, over 4831.00 frames. ], tot_loss[loss=0.2273, simple_loss=0.2843, pruned_loss=0.08516, over 952759.71 frames. ], batch size: 30, lr: 3.97e-03, grad_scale: 32.0 2023-04-26 14:49:46,671 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6717, 2.3020, 1.9453, 2.1611, 1.6250, 1.7812, 2.0837, 1.4749], device='cuda:1'), covar=tensor([0.2987, 0.2239, 0.1444, 0.1991, 0.3821, 0.2000, 0.2267, 0.3409], device='cuda:1'), in_proj_covar=tensor([0.0316, 0.0334, 0.0245, 0.0308, 0.0323, 0.0287, 0.0276, 0.0299], device='cuda:1'), out_proj_covar=tensor([1.2922e-04, 1.3618e-04, 9.9834e-05, 1.2428e-04, 1.3308e-04, 1.1593e-04, 1.1373e-04, 1.2110e-04], device='cuda:1') 2023-04-26 14:49:51,015 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-04-26 14:50:09,125 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=18616.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 14:50:13,226 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.222e+02 1.882e+02 2.176e+02 2.822e+02 8.396e+02, threshold=4.353e+02, percent-clipped=2.0 2023-04-26 14:50:19,725 INFO [finetune.py:976] (1/7) Epoch 4, batch 1450, loss[loss=0.2122, simple_loss=0.2684, pruned_loss=0.07803, over 4825.00 frames. ], tot_loss[loss=0.2277, simple_loss=0.2853, pruned_loss=0.0851, over 953298.54 frames. ], batch size: 25, lr: 3.97e-03, grad_scale: 32.0 2023-04-26 14:50:35,558 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([0.3131, 1.2435, 1.3900, 0.9793, 1.3629, 1.0669, 1.8090, 1.2540], device='cuda:1'), covar=tensor([0.4059, 0.1790, 0.5076, 0.3022, 0.1663, 0.2423, 0.1594, 0.4871], device='cuda:1'), in_proj_covar=tensor([0.0359, 0.0360, 0.0445, 0.0376, 0.0410, 0.0387, 0.0402, 0.0425], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 14:50:40,786 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=18664.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 14:50:50,781 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7876, 2.5288, 2.0703, 2.2473, 1.8505, 1.9111, 2.1764, 1.6445], device='cuda:1'), covar=tensor([0.2726, 0.1654, 0.1222, 0.1748, 0.3331, 0.1910, 0.2245, 0.3223], device='cuda:1'), in_proj_covar=tensor([0.0317, 0.0335, 0.0245, 0.0309, 0.0323, 0.0288, 0.0277, 0.0300], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 14:50:51,510 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-04-26 14:50:52,623 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=18674.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 14:51:03,065 INFO [finetune.py:976] (1/7) Epoch 4, batch 1500, loss[loss=0.2114, simple_loss=0.2662, pruned_loss=0.07834, over 4744.00 frames. ], tot_loss[loss=0.2285, simple_loss=0.2865, pruned_loss=0.08527, over 954443.29 frames. ], batch size: 28, lr: 3.96e-03, grad_scale: 32.0 2023-04-26 14:51:57,996 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.134e+02 1.902e+02 2.274e+02 2.776e+02 4.497e+02, threshold=4.548e+02, percent-clipped=1.0 2023-04-26 14:52:09,746 INFO [finetune.py:976] (1/7) Epoch 4, batch 1550, loss[loss=0.2712, simple_loss=0.3133, pruned_loss=0.1145, over 4793.00 frames. ], tot_loss[loss=0.2271, simple_loss=0.2853, pruned_loss=0.08446, over 952857.26 frames. ], batch size: 51, lr: 3.96e-03, grad_scale: 32.0 2023-04-26 14:52:11,083 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=18735.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 14:52:20,640 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9637, 1.2407, 1.5951, 1.5589, 2.0917, 1.7119, 1.3481, 1.5017], device='cuda:1'), covar=tensor([0.1470, 0.1677, 0.1704, 0.1446, 0.0813, 0.1521, 0.2366, 0.1834], device='cuda:1'), in_proj_covar=tensor([0.0312, 0.0332, 0.0346, 0.0305, 0.0342, 0.0349, 0.0312, 0.0348], device='cuda:1'), out_proj_covar=tensor([6.7566e-05, 7.1223e-05, 7.4945e-05, 6.3851e-05, 7.2653e-05, 7.6113e-05, 6.7736e-05, 7.5173e-05], device='cuda:1') 2023-04-26 14:52:53,632 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=2.00 vs. limit=2.0 2023-04-26 14:53:12,336 INFO [finetune.py:976] (1/7) Epoch 4, batch 1600, loss[loss=0.2267, simple_loss=0.2764, pruned_loss=0.08852, over 4870.00 frames. ], tot_loss[loss=0.2245, simple_loss=0.2823, pruned_loss=0.08329, over 953507.51 frames. ], batch size: 31, lr: 3.96e-03, grad_scale: 32.0 2023-04-26 14:53:12,477 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7552, 2.5208, 1.6675, 1.6064, 1.2645, 1.3459, 1.7258, 1.2445], device='cuda:1'), covar=tensor([0.1938, 0.1642, 0.1984, 0.2423, 0.3203, 0.2364, 0.1506, 0.2578], device='cuda:1'), in_proj_covar=tensor([0.0193, 0.0220, 0.0183, 0.0209, 0.0220, 0.0189, 0.0176, 0.0199], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-26 14:53:25,285 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([0.4543, 1.3135, 1.3983, 1.0615, 1.4750, 1.1314, 1.8466, 1.2574], device='cuda:1'), covar=tensor([0.3592, 0.1852, 0.5516, 0.2912, 0.1569, 0.2265, 0.1599, 0.4855], device='cuda:1'), in_proj_covar=tensor([0.0356, 0.0358, 0.0443, 0.0374, 0.0407, 0.0385, 0.0401, 0.0423], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 14:53:44,152 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.258e+02 2.049e+02 2.445e+02 2.756e+02 4.210e+02, threshold=4.889e+02, percent-clipped=0.0 2023-04-26 14:53:48,501 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=18830.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 14:53:49,087 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=18831.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 14:53:50,210 INFO [finetune.py:976] (1/7) Epoch 4, batch 1650, loss[loss=0.1946, simple_loss=0.2574, pruned_loss=0.06586, over 4732.00 frames. ], tot_loss[loss=0.2219, simple_loss=0.2793, pruned_loss=0.08221, over 952958.37 frames. ], batch size: 59, lr: 3.96e-03, grad_scale: 32.0 2023-04-26 14:53:50,922 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5589, 1.9984, 1.6727, 1.8891, 1.4993, 1.5450, 1.6902, 1.3367], device='cuda:1'), covar=tensor([0.2204, 0.1382, 0.1041, 0.1399, 0.3779, 0.1596, 0.2190, 0.2893], device='cuda:1'), in_proj_covar=tensor([0.0317, 0.0335, 0.0246, 0.0310, 0.0324, 0.0289, 0.0277, 0.0301], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 14:53:58,600 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=18846.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 14:54:01,775 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.57 vs. limit=2.0 2023-04-26 14:54:01,798 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.76 vs. limit=2.0 2023-04-26 14:54:11,008 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=18864.0, num_to_drop=1, layers_to_drop={2} 2023-04-26 14:54:12,764 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.1444, 4.5188, 0.9562, 2.6107, 2.8110, 2.9921, 2.7964, 1.2024], device='cuda:1'), covar=tensor([0.1209, 0.0805, 0.2272, 0.1150, 0.0831, 0.1090, 0.1311, 0.2028], device='cuda:1'), in_proj_covar=tensor([0.0122, 0.0264, 0.0148, 0.0129, 0.0139, 0.0161, 0.0126, 0.0129], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-26 14:54:20,473 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=18878.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 14:54:21,067 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=18879.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 14:54:23,389 INFO [finetune.py:976] (1/7) Epoch 4, batch 1700, loss[loss=0.1857, simple_loss=0.2475, pruned_loss=0.06193, over 4758.00 frames. ], tot_loss[loss=0.2185, simple_loss=0.2759, pruned_loss=0.0806, over 952054.33 frames. ], batch size: 26, lr: 3.96e-03, grad_scale: 32.0 2023-04-26 14:55:17,077 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.323e+02 1.925e+02 2.287e+02 2.733e+02 5.363e+02, threshold=4.574e+02, percent-clipped=1.0 2023-04-26 14:55:21,947 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=18930.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 14:55:23,645 INFO [finetune.py:976] (1/7) Epoch 4, batch 1750, loss[loss=0.283, simple_loss=0.3419, pruned_loss=0.1121, over 4814.00 frames. ], tot_loss[loss=0.222, simple_loss=0.2792, pruned_loss=0.08238, over 951224.51 frames. ], batch size: 45, lr: 3.96e-03, grad_scale: 32.0 2023-04-26 14:55:39,386 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.5266, 1.7520, 1.4160, 1.6637, 1.4923, 1.8241, 1.6127, 1.5190], device='cuda:1'), covar=tensor([1.0180, 1.7249, 1.7636, 1.3257, 1.5272, 2.2918, 1.9381, 1.7283], device='cuda:1'), in_proj_covar=tensor([0.0307, 0.0400, 0.0320, 0.0325, 0.0350, 0.0411, 0.0384, 0.0340], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-26 14:55:52,911 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5736, 1.2153, 1.4868, 1.8578, 1.6739, 1.4731, 1.5228, 1.5675], device='cuda:1'), covar=tensor([1.5194, 1.9189, 2.3054, 2.4158, 1.7615, 2.5277, 2.4376, 1.9179], device='cuda:1'), in_proj_covar=tensor([0.0433, 0.0478, 0.0567, 0.0584, 0.0464, 0.0492, 0.0505, 0.0512], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 14:55:56,819 INFO [finetune.py:976] (1/7) Epoch 4, batch 1800, loss[loss=0.2245, simple_loss=0.2806, pruned_loss=0.08422, over 4826.00 frames. ], tot_loss[loss=0.2238, simple_loss=0.2815, pruned_loss=0.08312, over 952754.18 frames. ], batch size: 30, lr: 3.96e-03, grad_scale: 32.0 2023-04-26 14:56:01,890 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=18991.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 14:56:12,512 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6942, 2.3568, 1.7314, 1.7104, 1.3864, 1.4184, 1.7356, 1.3664], device='cuda:1'), covar=tensor([0.1611, 0.1466, 0.1739, 0.1961, 0.2636, 0.1895, 0.1238, 0.2102], device='cuda:1'), in_proj_covar=tensor([0.0194, 0.0221, 0.0183, 0.0210, 0.0221, 0.0189, 0.0177, 0.0199], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-26 14:56:24,507 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=5.20 vs. limit=5.0 2023-04-26 14:56:35,175 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0959, 1.8746, 2.2371, 2.3275, 1.8665, 1.5567, 2.0024, 1.3696], device='cuda:1'), covar=tensor([0.0666, 0.1174, 0.0640, 0.0850, 0.1190, 0.1414, 0.0825, 0.1082], device='cuda:1'), in_proj_covar=tensor([0.0069, 0.0078, 0.0076, 0.0070, 0.0082, 0.0098, 0.0086, 0.0080], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:1') 2023-04-26 14:56:44,255 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.520e+02 1.971e+02 2.444e+02 3.087e+02 4.882e+02, threshold=4.887e+02, percent-clipped=3.0 2023-04-26 14:56:45,011 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9151, 2.4205, 2.0615, 2.2937, 1.6945, 1.9154, 2.0875, 1.5510], device='cuda:1'), covar=tensor([0.2190, 0.1347, 0.1039, 0.1293, 0.3148, 0.1447, 0.1989, 0.3002], device='cuda:1'), in_proj_covar=tensor([0.0314, 0.0333, 0.0243, 0.0307, 0.0321, 0.0286, 0.0275, 0.0297], device='cuda:1'), out_proj_covar=tensor([1.2823e-04, 1.3555e-04, 9.9207e-05, 1.2375e-04, 1.3244e-04, 1.1545e-04, 1.1361e-04, 1.2035e-04], device='cuda:1') 2023-04-26 14:56:49,158 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=19030.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 14:56:50,954 INFO [finetune.py:976] (1/7) Epoch 4, batch 1850, loss[loss=0.2162, simple_loss=0.2807, pruned_loss=0.07581, over 4898.00 frames. ], tot_loss[loss=0.223, simple_loss=0.2813, pruned_loss=0.08233, over 952610.31 frames. ], batch size: 43, lr: 3.96e-03, grad_scale: 32.0 2023-04-26 14:57:13,071 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.2410, 1.5537, 1.4854, 1.9706, 1.7128, 1.9952, 1.4458, 3.9864], device='cuda:1'), covar=tensor([0.0736, 0.0815, 0.0875, 0.1268, 0.0689, 0.0688, 0.0831, 0.0188], device='cuda:1'), in_proj_covar=tensor([0.0040, 0.0040, 0.0042, 0.0046, 0.0041, 0.0041, 0.0040, 0.0063], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0017], device='cuda:1') 2023-04-26 14:57:24,082 INFO [finetune.py:976] (1/7) Epoch 4, batch 1900, loss[loss=0.1995, simple_loss=0.2617, pruned_loss=0.06869, over 4748.00 frames. ], tot_loss[loss=0.2248, simple_loss=0.2835, pruned_loss=0.08303, over 953486.32 frames. ], batch size: 26, lr: 3.96e-03, grad_scale: 32.0 2023-04-26 14:57:50,668 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.323e+02 1.880e+02 2.172e+02 2.548e+02 4.187e+02, threshold=4.344e+02, percent-clipped=0.0 2023-04-26 14:57:54,922 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=19129.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 14:57:57,175 INFO [finetune.py:976] (1/7) Epoch 4, batch 1950, loss[loss=0.2021, simple_loss=0.2352, pruned_loss=0.08446, over 4284.00 frames. ], tot_loss[loss=0.2235, simple_loss=0.2821, pruned_loss=0.08242, over 953150.18 frames. ], batch size: 18, lr: 3.96e-03, grad_scale: 32.0 2023-04-26 14:58:05,510 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=19146.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 14:58:17,022 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=19164.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 14:58:30,343 INFO [finetune.py:976] (1/7) Epoch 4, batch 2000, loss[loss=0.2113, simple_loss=0.2674, pruned_loss=0.07756, over 4908.00 frames. ], tot_loss[loss=0.2217, simple_loss=0.2797, pruned_loss=0.0818, over 954523.18 frames. ], batch size: 37, lr: 3.96e-03, grad_scale: 32.0 2023-04-26 14:58:40,473 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=19190.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 14:58:48,413 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=19194.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 14:58:59,216 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 2023-04-26 14:59:10,842 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=19212.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 14:59:24,209 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.131e+02 1.742e+02 2.062e+02 2.571e+02 4.374e+02, threshold=4.123e+02, percent-clipped=1.0 2023-04-26 14:59:25,102 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-04-26 14:59:36,725 INFO [finetune.py:976] (1/7) Epoch 4, batch 2050, loss[loss=0.1944, simple_loss=0.2476, pruned_loss=0.07063, over 4810.00 frames. ], tot_loss[loss=0.2187, simple_loss=0.2761, pruned_loss=0.08064, over 954026.05 frames. ], batch size: 25, lr: 3.96e-03, grad_scale: 32.0 2023-04-26 14:59:53,470 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.1744, 4.5836, 1.0588, 2.3638, 2.7662, 2.7886, 2.8262, 1.2357], device='cuda:1'), covar=tensor([0.1216, 0.0863, 0.2155, 0.1332, 0.0899, 0.1247, 0.1207, 0.1958], device='cuda:1'), in_proj_covar=tensor([0.0122, 0.0264, 0.0148, 0.0129, 0.0139, 0.0161, 0.0126, 0.0129], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-26 15:00:03,703 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8916, 1.3707, 1.7349, 1.7814, 1.5406, 1.2810, 0.8689, 1.3809], device='cuda:1'), covar=tensor([0.4443, 0.5179, 0.2519, 0.3899, 0.4614, 0.3841, 0.6696, 0.3814], device='cuda:1'), in_proj_covar=tensor([0.0274, 0.0268, 0.0223, 0.0338, 0.0226, 0.0232, 0.0253, 0.0202], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-26 15:00:15,254 INFO [finetune.py:976] (1/7) Epoch 4, batch 2100, loss[loss=0.2056, simple_loss=0.2663, pruned_loss=0.07246, over 4871.00 frames. ], tot_loss[loss=0.2194, simple_loss=0.2763, pruned_loss=0.08123, over 955507.67 frames. ], batch size: 34, lr: 3.96e-03, grad_scale: 32.0 2023-04-26 15:00:17,154 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=19286.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:00:36,849 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.2681, 1.5134, 1.4051, 1.5793, 1.4384, 1.6880, 1.5984, 1.4888], device='cuda:1'), covar=tensor([1.0377, 1.6030, 1.4823, 1.2193, 1.4192, 2.2538, 1.7381, 1.5101], device='cuda:1'), in_proj_covar=tensor([0.0311, 0.0404, 0.0322, 0.0328, 0.0352, 0.0415, 0.0387, 0.0343], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-26 15:00:45,250 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6192, 1.9045, 1.0837, 1.3823, 2.1086, 1.5038, 1.4603, 1.5077], device='cuda:1'), covar=tensor([0.0559, 0.0408, 0.0373, 0.0582, 0.0268, 0.0582, 0.0541, 0.0620], device='cuda:1'), in_proj_covar=tensor([0.0031, 0.0027, 0.0025, 0.0031, 0.0022, 0.0031, 0.0030, 0.0032], device='cuda:1'), out_proj_covar=tensor([0.0049, 0.0045, 0.0038, 0.0049, 0.0038, 0.0048, 0.0047, 0.0050], device='cuda:1') 2023-04-26 15:01:01,584 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.066e+02 1.821e+02 2.202e+02 2.671e+02 7.978e+02, threshold=4.404e+02, percent-clipped=3.0 2023-04-26 15:01:06,053 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2023-04-26 15:01:06,999 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=19330.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:01:14,462 INFO [finetune.py:976] (1/7) Epoch 4, batch 2150, loss[loss=0.1969, simple_loss=0.2686, pruned_loss=0.06257, over 4795.00 frames. ], tot_loss[loss=0.2235, simple_loss=0.2809, pruned_loss=0.08302, over 955284.50 frames. ], batch size: 26, lr: 3.96e-03, grad_scale: 32.0 2023-04-26 15:01:15,849 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5523, 1.3281, 1.5824, 1.8355, 1.7649, 1.4643, 1.5322, 1.5312], device='cuda:1'), covar=tensor([1.2973, 1.7009, 1.9193, 2.0906, 1.4471, 2.1112, 2.0603, 1.8543], device='cuda:1'), in_proj_covar=tensor([0.0432, 0.0477, 0.0564, 0.0584, 0.0463, 0.0491, 0.0502, 0.0511], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 15:01:16,388 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([3.2418, 3.2733, 2.3805, 3.7855, 3.3031, 3.2267, 1.4648, 3.2070], device='cuda:1'), covar=tensor([0.2295, 0.1426, 0.4006, 0.2405, 0.3190, 0.2292, 0.5986, 0.2920], device='cuda:1'), in_proj_covar=tensor([0.0253, 0.0225, 0.0264, 0.0316, 0.0311, 0.0261, 0.0278, 0.0282], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-26 15:02:07,705 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=19378.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:02:16,423 INFO [finetune.py:976] (1/7) Epoch 4, batch 2200, loss[loss=0.2484, simple_loss=0.3022, pruned_loss=0.09726, over 4738.00 frames. ], tot_loss[loss=0.2265, simple_loss=0.2838, pruned_loss=0.08457, over 954851.61 frames. ], batch size: 59, lr: 3.96e-03, grad_scale: 32.0 2023-04-26 15:02:27,626 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7714, 0.9431, 1.2075, 1.3882, 1.3596, 1.5555, 1.2704, 1.2708], device='cuda:1'), covar=tensor([0.8675, 1.2231, 1.0806, 0.9738, 1.1937, 1.8269, 1.2698, 1.2034], device='cuda:1'), in_proj_covar=tensor([0.0310, 0.0401, 0.0320, 0.0326, 0.0350, 0.0413, 0.0385, 0.0341], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-26 15:02:30,488 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=19394.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:03:06,440 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.252e+02 1.815e+02 2.166e+02 2.683e+02 4.330e+02, threshold=4.331e+02, percent-clipped=0.0 2023-04-26 15:03:09,559 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9319, 1.4156, 1.8348, 2.0827, 1.6976, 1.3523, 1.0439, 1.5587], device='cuda:1'), covar=tensor([0.4465, 0.5093, 0.2273, 0.3825, 0.4640, 0.3741, 0.6478, 0.4270], device='cuda:1'), in_proj_covar=tensor([0.0274, 0.0268, 0.0223, 0.0339, 0.0226, 0.0233, 0.0253, 0.0203], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-26 15:03:18,527 INFO [finetune.py:976] (1/7) Epoch 4, batch 2250, loss[loss=0.2181, simple_loss=0.2871, pruned_loss=0.07452, over 4748.00 frames. ], tot_loss[loss=0.2276, simple_loss=0.2853, pruned_loss=0.08493, over 954959.47 frames. ], batch size: 28, lr: 3.96e-03, grad_scale: 32.0 2023-04-26 15:03:51,636 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=19455.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:03:55,186 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9804, 2.3519, 1.0801, 1.2892, 1.9210, 1.2483, 2.9493, 1.6128], device='cuda:1'), covar=tensor([0.0686, 0.0688, 0.0863, 0.1270, 0.0479, 0.1000, 0.0283, 0.0676], device='cuda:1'), in_proj_covar=tensor([0.0055, 0.0072, 0.0053, 0.0050, 0.0055, 0.0055, 0.0084, 0.0053], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0008, 0.0008, 0.0008, 0.0008, 0.0011, 0.0008], device='cuda:1') 2023-04-26 15:04:21,506 INFO [finetune.py:976] (1/7) Epoch 4, batch 2300, loss[loss=0.2377, simple_loss=0.2877, pruned_loss=0.09383, over 4274.00 frames. ], tot_loss[loss=0.2255, simple_loss=0.2842, pruned_loss=0.08344, over 953858.55 frames. ], batch size: 66, lr: 3.96e-03, grad_scale: 32.0 2023-04-26 15:04:23,292 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=19485.0, num_to_drop=1, layers_to_drop={2} 2023-04-26 15:04:58,953 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.086e+02 1.751e+02 2.123e+02 2.573e+02 6.035e+02, threshold=4.246e+02, percent-clipped=1.0 2023-04-26 15:05:05,527 INFO [finetune.py:976] (1/7) Epoch 4, batch 2350, loss[loss=0.2193, simple_loss=0.2706, pruned_loss=0.08401, over 4801.00 frames. ], tot_loss[loss=0.2235, simple_loss=0.2815, pruned_loss=0.08274, over 954199.93 frames. ], batch size: 25, lr: 3.96e-03, grad_scale: 32.0 2023-04-26 15:05:29,137 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.74 vs. limit=5.0 2023-04-26 15:05:32,813 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.53 vs. limit=2.0 2023-04-26 15:05:39,105 INFO [finetune.py:976] (1/7) Epoch 4, batch 2400, loss[loss=0.1862, simple_loss=0.2466, pruned_loss=0.06296, over 4757.00 frames. ], tot_loss[loss=0.2227, simple_loss=0.2799, pruned_loss=0.08275, over 954752.41 frames. ], batch size: 26, lr: 3.96e-03, grad_scale: 32.0 2023-04-26 15:05:41,022 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=19586.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:06:03,611 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([3.3500, 3.2975, 2.5360, 3.8837, 3.4050, 3.4447, 1.3225, 3.2513], device='cuda:1'), covar=tensor([0.2081, 0.1481, 0.3348, 0.2599, 0.4007, 0.2051, 0.6256, 0.3072], device='cuda:1'), in_proj_covar=tensor([0.0255, 0.0226, 0.0264, 0.0318, 0.0312, 0.0262, 0.0279, 0.0283], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-26 15:06:03,661 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=19618.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:06:04,898 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7552, 1.1647, 1.4070, 1.4210, 2.0345, 1.6355, 1.2696, 1.3510], device='cuda:1'), covar=tensor([0.2133, 0.1855, 0.2353, 0.1619, 0.0893, 0.1606, 0.2459, 0.2096], device='cuda:1'), in_proj_covar=tensor([0.0319, 0.0340, 0.0355, 0.0312, 0.0352, 0.0357, 0.0317, 0.0356], device='cuda:1'), out_proj_covar=tensor([6.9049e-05, 7.3066e-05, 7.6853e-05, 6.5455e-05, 7.4642e-05, 7.7862e-05, 6.8957e-05, 7.6970e-05], device='cuda:1') 2023-04-26 15:06:06,614 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.143e+02 1.798e+02 2.118e+02 2.540e+02 5.281e+02, threshold=4.235e+02, percent-clipped=1.0 2023-04-26 15:06:12,782 INFO [finetune.py:976] (1/7) Epoch 4, batch 2450, loss[loss=0.2024, simple_loss=0.258, pruned_loss=0.07347, over 4790.00 frames. ], tot_loss[loss=0.2198, simple_loss=0.277, pruned_loss=0.08135, over 955912.06 frames. ], batch size: 26, lr: 3.96e-03, grad_scale: 32.0 2023-04-26 15:06:13,448 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=19634.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:06:34,153 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.53 vs. limit=2.0 2023-04-26 15:06:37,043 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([4.6765, 4.5339, 3.2161, 5.2940, 4.7348, 4.6253, 2.1130, 4.5013], device='cuda:1'), covar=tensor([0.1568, 0.1060, 0.3042, 0.1034, 0.3989, 0.1649, 0.5977, 0.2233], device='cuda:1'), in_proj_covar=tensor([0.0253, 0.0225, 0.0262, 0.0316, 0.0311, 0.0261, 0.0278, 0.0281], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-26 15:06:44,220 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=19679.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:06:46,564 INFO [finetune.py:976] (1/7) Epoch 4, batch 2500, loss[loss=0.189, simple_loss=0.2574, pruned_loss=0.06026, over 4827.00 frames. ], tot_loss[loss=0.2211, simple_loss=0.2781, pruned_loss=0.08199, over 956230.79 frames. ], batch size: 25, lr: 3.96e-03, grad_scale: 32.0 2023-04-26 15:07:19,858 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([4.7333, 4.7465, 3.3036, 5.4787, 4.8612, 4.7798, 2.4333, 4.6751], device='cuda:1'), covar=tensor([0.1369, 0.0911, 0.2649, 0.0838, 0.4052, 0.1549, 0.4820, 0.1869], device='cuda:1'), in_proj_covar=tensor([0.0252, 0.0224, 0.0262, 0.0315, 0.0310, 0.0260, 0.0277, 0.0280], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-26 15:07:31,980 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.364e+02 1.985e+02 2.362e+02 2.892e+02 4.639e+02, threshold=4.724e+02, percent-clipped=3.0 2023-04-26 15:07:43,942 INFO [finetune.py:976] (1/7) Epoch 4, batch 2550, loss[loss=0.2035, simple_loss=0.2626, pruned_loss=0.07224, over 4699.00 frames. ], tot_loss[loss=0.2214, simple_loss=0.2794, pruned_loss=0.08169, over 953845.67 frames. ], batch size: 23, lr: 3.96e-03, grad_scale: 32.0 2023-04-26 15:08:07,132 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=19750.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:08:50,445 INFO [finetune.py:976] (1/7) Epoch 4, batch 2600, loss[loss=0.2616, simple_loss=0.3092, pruned_loss=0.107, over 4820.00 frames. ], tot_loss[loss=0.2257, simple_loss=0.2839, pruned_loss=0.08375, over 955526.55 frames. ], batch size: 30, lr: 3.96e-03, grad_scale: 32.0 2023-04-26 15:08:57,321 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=19785.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 15:09:03,897 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9142, 1.3874, 1.7625, 2.0184, 1.6428, 1.3135, 0.8687, 1.4166], device='cuda:1'), covar=tensor([0.4446, 0.5307, 0.2446, 0.3537, 0.4632, 0.3901, 0.6425, 0.3932], device='cuda:1'), in_proj_covar=tensor([0.0272, 0.0265, 0.0221, 0.0335, 0.0224, 0.0231, 0.0251, 0.0201], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-26 15:09:23,390 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.319e+02 1.754e+02 2.126e+02 2.494e+02 3.908e+02, threshold=4.253e+02, percent-clipped=0.0 2023-04-26 15:09:29,904 INFO [finetune.py:976] (1/7) Epoch 4, batch 2650, loss[loss=0.2142, simple_loss=0.2829, pruned_loss=0.07277, over 4783.00 frames. ], tot_loss[loss=0.2257, simple_loss=0.2842, pruned_loss=0.08364, over 953004.18 frames. ], batch size: 51, lr: 3.96e-03, grad_scale: 32.0 2023-04-26 15:09:29,969 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=19833.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 15:10:24,177 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.1269, 1.4998, 1.3807, 1.8527, 1.7063, 2.0052, 1.4098, 3.7056], device='cuda:1'), covar=tensor([0.0732, 0.0771, 0.0805, 0.1167, 0.0617, 0.0615, 0.0787, 0.0140], device='cuda:1'), in_proj_covar=tensor([0.0040, 0.0040, 0.0042, 0.0046, 0.0041, 0.0041, 0.0040, 0.0063], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0017], device='cuda:1') 2023-04-26 15:10:35,941 INFO [finetune.py:976] (1/7) Epoch 4, batch 2700, loss[loss=0.1881, simple_loss=0.2422, pruned_loss=0.06697, over 4888.00 frames. ], tot_loss[loss=0.2249, simple_loss=0.2827, pruned_loss=0.08349, over 953235.16 frames. ], batch size: 32, lr: 3.96e-03, grad_scale: 32.0 2023-04-26 15:11:15,310 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.121e+02 1.807e+02 2.180e+02 2.734e+02 4.615e+02, threshold=4.361e+02, percent-clipped=4.0 2023-04-26 15:11:21,340 INFO [finetune.py:976] (1/7) Epoch 4, batch 2750, loss[loss=0.228, simple_loss=0.2831, pruned_loss=0.08647, over 4772.00 frames. ], tot_loss[loss=0.2241, simple_loss=0.281, pruned_loss=0.08358, over 954006.04 frames. ], batch size: 28, lr: 3.96e-03, grad_scale: 32.0 2023-04-26 15:11:42,957 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5911, 2.0454, 1.7731, 2.0091, 1.5549, 1.7771, 1.6793, 1.4069], device='cuda:1'), covar=tensor([0.2069, 0.1447, 0.0971, 0.1192, 0.3282, 0.1258, 0.1996, 0.2500], device='cuda:1'), in_proj_covar=tensor([0.0313, 0.0331, 0.0241, 0.0305, 0.0320, 0.0284, 0.0273, 0.0295], device='cuda:1'), out_proj_covar=tensor([1.2755e-04, 1.3472e-04, 9.8238e-05, 1.2304e-04, 1.3214e-04, 1.1466e-04, 1.1268e-04, 1.1912e-04], device='cuda:1') 2023-04-26 15:11:49,852 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=19974.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:11:55,248 INFO [finetune.py:976] (1/7) Epoch 4, batch 2800, loss[loss=0.1457, simple_loss=0.212, pruned_loss=0.03969, over 4926.00 frames. ], tot_loss[loss=0.2193, simple_loss=0.2764, pruned_loss=0.08103, over 955088.15 frames. ], batch size: 37, lr: 3.96e-03, grad_scale: 32.0 2023-04-26 15:12:20,803 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=20020.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:12:23,470 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.089e+02 1.778e+02 2.116e+02 2.453e+02 4.080e+02, threshold=4.233e+02, percent-clipped=0.0 2023-04-26 15:12:28,416 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-04-26 15:12:30,049 INFO [finetune.py:976] (1/7) Epoch 4, batch 2850, loss[loss=0.22, simple_loss=0.2771, pruned_loss=0.08142, over 4844.00 frames. ], tot_loss[loss=0.2167, simple_loss=0.2739, pruned_loss=0.07976, over 954915.08 frames. ], batch size: 25, lr: 3.96e-03, grad_scale: 32.0 2023-04-26 15:12:40,866 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=20050.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:12:52,095 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.0053, 1.2634, 3.2384, 2.9984, 2.9245, 3.0732, 3.0920, 2.8819], device='cuda:1'), covar=tensor([0.6948, 0.5214, 0.1276, 0.1908, 0.1390, 0.1923, 0.2251, 0.1731], device='cuda:1'), in_proj_covar=tensor([0.0318, 0.0308, 0.0427, 0.0433, 0.0365, 0.0420, 0.0327, 0.0387], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-26 15:13:02,626 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=20081.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:13:03,753 INFO [finetune.py:976] (1/7) Epoch 4, batch 2900, loss[loss=0.3223, simple_loss=0.3712, pruned_loss=0.1367, over 4738.00 frames. ], tot_loss[loss=0.2219, simple_loss=0.2794, pruned_loss=0.08218, over 955012.32 frames. ], batch size: 59, lr: 3.96e-03, grad_scale: 32.0 2023-04-26 15:13:07,538 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=20089.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:13:08,250 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.62 vs. limit=5.0 2023-04-26 15:13:13,376 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=20098.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:13:19,610 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=5.07 vs. limit=5.0 2023-04-26 15:13:21,882 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=20112.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:13:29,802 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.335e+02 1.818e+02 2.275e+02 2.799e+02 4.560e+02, threshold=4.551e+02, percent-clipped=2.0 2023-04-26 15:13:47,512 INFO [finetune.py:976] (1/7) Epoch 4, batch 2950, loss[loss=0.2179, simple_loss=0.2785, pruned_loss=0.07861, over 4844.00 frames. ], tot_loss[loss=0.2241, simple_loss=0.2824, pruned_loss=0.08294, over 956041.34 frames. ], batch size: 47, lr: 3.96e-03, grad_scale: 32.0 2023-04-26 15:14:09,949 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=20150.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:14:35,343 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=20173.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:14:54,003 INFO [finetune.py:976] (1/7) Epoch 4, batch 3000, loss[loss=0.2083, simple_loss=0.2715, pruned_loss=0.07258, over 4753.00 frames. ], tot_loss[loss=0.225, simple_loss=0.2836, pruned_loss=0.08321, over 957535.95 frames. ], batch size: 27, lr: 3.96e-03, grad_scale: 32.0 2023-04-26 15:14:54,003 INFO [finetune.py:1001] (1/7) Computing validation loss 2023-04-26 15:15:01,785 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([0.7076, 1.7384, 1.6912, 1.3447, 1.8852, 1.5192, 2.3007, 1.5203], device='cuda:1'), covar=tensor([0.4360, 0.1991, 0.6084, 0.3509, 0.1887, 0.2684, 0.1813, 0.5658], device='cuda:1'), in_proj_covar=tensor([0.0354, 0.0359, 0.0444, 0.0373, 0.0405, 0.0382, 0.0398, 0.0423], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 15:15:03,075 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([3.5611, 3.5142, 2.6275, 3.9817, 3.5445, 3.6457, 1.5467, 3.4551], device='cuda:1'), covar=tensor([0.1440, 0.1194, 0.3289, 0.1810, 0.2917, 0.1562, 0.5241, 0.2367], device='cuda:1'), in_proj_covar=tensor([0.0251, 0.0222, 0.0260, 0.0313, 0.0309, 0.0258, 0.0276, 0.0279], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-26 15:15:10,814 INFO [finetune.py:1010] (1/7) Epoch 4, validation: loss=0.1632, simple_loss=0.2363, pruned_loss=0.04509, over 2265189.00 frames. 2023-04-26 15:15:10,815 INFO [finetune.py:1011] (1/7) Maximum memory allocated so far is 6391MB 2023-04-26 15:15:35,621 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=20204.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:15:47,604 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.225e+02 1.833e+02 2.218e+02 2.696e+02 4.809e+02, threshold=4.435e+02, percent-clipped=1.0 2023-04-26 15:16:05,215 INFO [finetune.py:976] (1/7) Epoch 4, batch 3050, loss[loss=0.1852, simple_loss=0.2605, pruned_loss=0.05497, over 4745.00 frames. ], tot_loss[loss=0.2248, simple_loss=0.2835, pruned_loss=0.08308, over 957777.35 frames. ], batch size: 54, lr: 3.96e-03, grad_scale: 32.0 2023-04-26 15:16:43,930 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=20265.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:16:49,430 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=20274.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:16:55,395 INFO [finetune.py:976] (1/7) Epoch 4, batch 3100, loss[loss=0.227, simple_loss=0.2771, pruned_loss=0.08841, over 4907.00 frames. ], tot_loss[loss=0.2234, simple_loss=0.2817, pruned_loss=0.08251, over 956376.41 frames. ], batch size: 35, lr: 3.96e-03, grad_scale: 32.0 2023-04-26 15:17:21,578 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=20322.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:17:22,123 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.364e+02 1.832e+02 2.083e+02 2.367e+02 4.071e+02, threshold=4.166e+02, percent-clipped=0.0 2023-04-26 15:17:28,233 INFO [finetune.py:976] (1/7) Epoch 4, batch 3150, loss[loss=0.1809, simple_loss=0.2364, pruned_loss=0.06273, over 4822.00 frames. ], tot_loss[loss=0.2205, simple_loss=0.2785, pruned_loss=0.0813, over 957757.07 frames. ], batch size: 40, lr: 3.96e-03, grad_scale: 32.0 2023-04-26 15:17:39,868 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.2106, 2.3515, 1.9731, 1.9193, 2.1952, 1.8315, 2.8792, 1.4968], device='cuda:1'), covar=tensor([0.4579, 0.1555, 0.4335, 0.3422, 0.2253, 0.2958, 0.1630, 0.5251], device='cuda:1'), in_proj_covar=tensor([0.0359, 0.0363, 0.0448, 0.0376, 0.0409, 0.0386, 0.0403, 0.0428], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 15:17:39,892 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.1144, 1.5289, 2.0609, 2.4374, 1.9343, 1.5362, 1.1610, 1.8311], device='cuda:1'), covar=tensor([0.4388, 0.5443, 0.2385, 0.4028, 0.4480, 0.3964, 0.7079, 0.4017], device='cuda:1'), in_proj_covar=tensor([0.0272, 0.0265, 0.0222, 0.0335, 0.0225, 0.0231, 0.0251, 0.0201], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-26 15:17:54,202 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4925, 3.4507, 0.8646, 1.7683, 1.8683, 2.2948, 2.0474, 1.0622], device='cuda:1'), covar=tensor([0.1544, 0.1314, 0.2343, 0.1575, 0.1318, 0.1328, 0.1589, 0.2039], device='cuda:1'), in_proj_covar=tensor([0.0121, 0.0263, 0.0147, 0.0129, 0.0138, 0.0161, 0.0125, 0.0128], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-26 15:17:57,215 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=20376.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:18:01,415 INFO [finetune.py:976] (1/7) Epoch 4, batch 3200, loss[loss=0.2117, simple_loss=0.2703, pruned_loss=0.07658, over 4722.00 frames. ], tot_loss[loss=0.2173, simple_loss=0.2752, pruned_loss=0.07974, over 958503.23 frames. ], batch size: 23, lr: 3.96e-03, grad_scale: 32.0 2023-04-26 15:18:04,538 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=20388.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:18:20,934 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.28 vs. limit=5.0 2023-04-26 15:18:25,071 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=20418.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:18:28,510 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.209e+02 1.902e+02 2.131e+02 2.743e+02 4.123e+02, threshold=4.262e+02, percent-clipped=0.0 2023-04-26 15:18:34,645 INFO [finetune.py:976] (1/7) Epoch 4, batch 3250, loss[loss=0.1588, simple_loss=0.2322, pruned_loss=0.04268, over 4797.00 frames. ], tot_loss[loss=0.2183, simple_loss=0.276, pruned_loss=0.08033, over 956970.46 frames. ], batch size: 29, lr: 3.96e-03, grad_scale: 64.0 2023-04-26 15:18:37,726 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4971, 1.4911, 0.5349, 1.1973, 1.4665, 1.3843, 1.2760, 1.2937], device='cuda:1'), covar=tensor([0.0583, 0.0416, 0.0495, 0.0604, 0.0350, 0.0586, 0.0537, 0.0685], device='cuda:1'), in_proj_covar=tensor([0.0031, 0.0027, 0.0025, 0.0031, 0.0022, 0.0031, 0.0030, 0.0032], device='cuda:1'), out_proj_covar=tensor([0.0049, 0.0045, 0.0038, 0.0049, 0.0038, 0.0048, 0.0048, 0.0050], device='cuda:1') 2023-04-26 15:18:42,471 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=20445.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:18:45,876 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=20449.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:19:04,255 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=20468.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:19:11,503 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=20479.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:19:13,809 INFO [finetune.py:976] (1/7) Epoch 4, batch 3300, loss[loss=0.1957, simple_loss=0.2557, pruned_loss=0.0679, over 4741.00 frames. ], tot_loss[loss=0.2212, simple_loss=0.2792, pruned_loss=0.0816, over 954333.30 frames. ], batch size: 27, lr: 3.96e-03, grad_scale: 32.0 2023-04-26 15:19:26,832 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.2704, 1.4749, 1.3007, 1.8174, 1.6017, 1.9379, 1.3665, 3.3750], device='cuda:1'), covar=tensor([0.0701, 0.0793, 0.0859, 0.1217, 0.0662, 0.0573, 0.0773, 0.0176], device='cuda:1'), in_proj_covar=tensor([0.0040, 0.0040, 0.0042, 0.0046, 0.0041, 0.0041, 0.0040, 0.0063], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0017], device='cuda:1') 2023-04-26 15:19:45,655 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8873, 3.8509, 1.2823, 2.1981, 2.2714, 2.7899, 2.3926, 1.4585], device='cuda:1'), covar=tensor([0.1429, 0.1589, 0.2272, 0.1442, 0.1142, 0.1178, 0.1464, 0.1933], device='cuda:1'), in_proj_covar=tensor([0.0121, 0.0262, 0.0147, 0.0129, 0.0138, 0.0160, 0.0125, 0.0128], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-26 15:19:47,324 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.231e+02 1.894e+02 2.274e+02 2.897e+02 7.817e+02, threshold=4.548e+02, percent-clipped=5.0 2023-04-26 15:19:58,662 INFO [finetune.py:976] (1/7) Epoch 4, batch 3350, loss[loss=0.2151, simple_loss=0.2959, pruned_loss=0.06718, over 4891.00 frames. ], tot_loss[loss=0.2238, simple_loss=0.2824, pruned_loss=0.08263, over 955558.96 frames. ], batch size: 43, lr: 3.96e-03, grad_scale: 16.0 2023-04-26 15:20:17,999 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.81 vs. limit=5.0 2023-04-26 15:20:39,484 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=20560.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:20:55,812 INFO [finetune.py:976] (1/7) Epoch 4, batch 3400, loss[loss=0.213, simple_loss=0.282, pruned_loss=0.07201, over 4766.00 frames. ], tot_loss[loss=0.2246, simple_loss=0.2834, pruned_loss=0.08285, over 957223.17 frames. ], batch size: 26, lr: 3.96e-03, grad_scale: 16.0 2023-04-26 15:21:04,541 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=20588.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:21:29,472 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.140e+02 1.806e+02 2.102e+02 2.435e+02 3.817e+02, threshold=4.203e+02, percent-clipped=0.0 2023-04-26 15:21:30,776 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.7372, 4.7936, 1.6688, 2.7171, 3.2376, 3.4388, 3.1047, 1.7133], device='cuda:1'), covar=tensor([0.1016, 0.0890, 0.1906, 0.1216, 0.0788, 0.0963, 0.1165, 0.1762], device='cuda:1'), in_proj_covar=tensor([0.0121, 0.0262, 0.0147, 0.0128, 0.0138, 0.0161, 0.0125, 0.0128], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-26 15:21:34,360 INFO [finetune.py:976] (1/7) Epoch 4, batch 3450, loss[loss=0.2236, simple_loss=0.274, pruned_loss=0.08658, over 4001.00 frames. ], tot_loss[loss=0.2231, simple_loss=0.2822, pruned_loss=0.08197, over 953591.82 frames. ], batch size: 17, lr: 3.96e-03, grad_scale: 16.0 2023-04-26 15:21:44,649 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=20649.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:22:10,203 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5429, 0.8504, 1.2503, 1.1185, 1.6283, 1.3268, 1.0526, 1.1935], device='cuda:1'), covar=tensor([0.1422, 0.1787, 0.2139, 0.1456, 0.0877, 0.1445, 0.1819, 0.1888], device='cuda:1'), in_proj_covar=tensor([0.0317, 0.0337, 0.0352, 0.0310, 0.0351, 0.0353, 0.0314, 0.0353], device='cuda:1'), out_proj_covar=tensor([6.8435e-05, 7.2229e-05, 7.6239e-05, 6.4890e-05, 7.4479e-05, 7.6842e-05, 6.8329e-05, 7.6272e-05], device='cuda:1') 2023-04-26 15:22:19,739 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=20676.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:22:23,907 INFO [finetune.py:976] (1/7) Epoch 4, batch 3500, loss[loss=0.2653, simple_loss=0.3084, pruned_loss=0.1111, over 4822.00 frames. ], tot_loss[loss=0.2212, simple_loss=0.2798, pruned_loss=0.08134, over 955067.07 frames. ], batch size: 33, lr: 3.96e-03, grad_scale: 16.0 2023-04-26 15:22:58,015 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=20724.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:22:58,547 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.140e+02 1.751e+02 2.119e+02 2.772e+02 4.921e+02, threshold=4.238e+02, percent-clipped=1.0 2023-04-26 15:23:03,439 INFO [finetune.py:976] (1/7) Epoch 4, batch 3550, loss[loss=0.182, simple_loss=0.2434, pruned_loss=0.06031, over 4913.00 frames. ], tot_loss[loss=0.2192, simple_loss=0.2769, pruned_loss=0.08076, over 954895.07 frames. ], batch size: 46, lr: 3.96e-03, grad_scale: 16.0 2023-04-26 15:23:15,945 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=20744.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:23:16,613 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=20745.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:23:26,463 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5288, 1.4276, 1.8587, 1.7830, 1.4153, 1.2107, 1.5714, 1.2226], device='cuda:1'), covar=tensor([0.0760, 0.0802, 0.0505, 0.1008, 0.1157, 0.1440, 0.0821, 0.0917], device='cuda:1'), in_proj_covar=tensor([0.0069, 0.0077, 0.0076, 0.0070, 0.0082, 0.0097, 0.0086, 0.0079], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:1') 2023-04-26 15:23:48,059 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=20768.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:23:58,621 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=20774.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:24:04,022 INFO [finetune.py:976] (1/7) Epoch 4, batch 3600, loss[loss=0.2422, simple_loss=0.298, pruned_loss=0.09323, over 4740.00 frames. ], tot_loss[loss=0.2155, simple_loss=0.2737, pruned_loss=0.07869, over 955127.77 frames. ], batch size: 54, lr: 3.96e-03, grad_scale: 16.0 2023-04-26 15:24:10,253 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=20793.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:24:31,211 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=20816.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:24:38,615 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.149e+02 1.726e+02 2.072e+02 2.545e+02 4.959e+02, threshold=4.144e+02, percent-clipped=1.0 2023-04-26 15:24:43,526 INFO [finetune.py:976] (1/7) Epoch 4, batch 3650, loss[loss=0.2919, simple_loss=0.3334, pruned_loss=0.1252, over 4818.00 frames. ], tot_loss[loss=0.2197, simple_loss=0.2777, pruned_loss=0.0808, over 953413.63 frames. ], batch size: 30, lr: 3.96e-03, grad_scale: 16.0 2023-04-26 15:24:58,146 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.1124, 0.6131, 0.9253, 0.6867, 1.2738, 0.9638, 0.7500, 1.0333], device='cuda:1'), covar=tensor([0.2034, 0.1957, 0.2467, 0.1877, 0.1225, 0.1792, 0.2277, 0.2378], device='cuda:1'), in_proj_covar=tensor([0.0317, 0.0339, 0.0355, 0.0311, 0.0352, 0.0354, 0.0317, 0.0355], device='cuda:1'), out_proj_covar=tensor([6.8490e-05, 7.2733e-05, 7.6919e-05, 6.5171e-05, 7.4799e-05, 7.7148e-05, 6.8799e-05, 7.6707e-05], device='cuda:1') 2023-04-26 15:25:00,515 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=20860.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:25:16,797 INFO [finetune.py:976] (1/7) Epoch 4, batch 3700, loss[loss=0.2438, simple_loss=0.29, pruned_loss=0.09883, over 4719.00 frames. ], tot_loss[loss=0.2234, simple_loss=0.2823, pruned_loss=0.08221, over 953945.91 frames. ], batch size: 23, lr: 3.96e-03, grad_scale: 16.0 2023-04-26 15:25:31,275 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([0.7819, 1.9290, 1.8687, 1.5507, 1.9619, 1.5649, 2.6207, 1.4265], device='cuda:1'), covar=tensor([0.4181, 0.1687, 0.4118, 0.3154, 0.1816, 0.2683, 0.1199, 0.4838], device='cuda:1'), in_proj_covar=tensor([0.0354, 0.0356, 0.0440, 0.0371, 0.0402, 0.0381, 0.0397, 0.0423], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 15:25:31,311 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5651, 1.0655, 1.4377, 1.8826, 1.6497, 1.4432, 1.5052, 1.5383], device='cuda:1'), covar=tensor([1.4596, 1.9805, 2.0817, 2.2931, 1.6386, 2.4646, 2.4497, 1.7531], device='cuda:1'), in_proj_covar=tensor([0.0430, 0.0474, 0.0562, 0.0580, 0.0463, 0.0487, 0.0500, 0.0505], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 15:25:32,430 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=20908.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:25:49,806 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.59 vs. limit=2.0 2023-04-26 15:25:50,198 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.325e+02 1.979e+02 2.308e+02 2.768e+02 4.690e+02, threshold=4.617e+02, percent-clipped=4.0 2023-04-26 15:26:01,984 INFO [finetune.py:976] (1/7) Epoch 4, batch 3750, loss[loss=0.2277, simple_loss=0.2862, pruned_loss=0.08459, over 4803.00 frames. ], tot_loss[loss=0.2237, simple_loss=0.2823, pruned_loss=0.08249, over 953778.03 frames. ], batch size: 51, lr: 3.96e-03, grad_scale: 16.0 2023-04-26 15:26:14,232 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=20944.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:26:45,380 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5398, 1.4479, 0.6110, 1.2561, 1.6907, 1.4485, 1.3591, 1.3458], device='cuda:1'), covar=tensor([0.0518, 0.0411, 0.0461, 0.0582, 0.0305, 0.0576, 0.0553, 0.0600], device='cuda:1'), in_proj_covar=tensor([0.0031, 0.0027, 0.0024, 0.0031, 0.0022, 0.0030, 0.0029, 0.0031], device='cuda:1'), out_proj_covar=tensor([0.0048, 0.0044, 0.0038, 0.0049, 0.0038, 0.0048, 0.0047, 0.0050], device='cuda:1') 2023-04-26 15:27:02,460 INFO [finetune.py:976] (1/7) Epoch 4, batch 3800, loss[loss=0.1772, simple_loss=0.2425, pruned_loss=0.05596, over 4812.00 frames. ], tot_loss[loss=0.2236, simple_loss=0.2832, pruned_loss=0.08205, over 955612.89 frames. ], batch size: 25, lr: 3.96e-03, grad_scale: 16.0 2023-04-26 15:27:10,586 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.4017, 1.6296, 1.6988, 2.2574, 2.5143, 2.0950, 1.9854, 1.9061], device='cuda:1'), covar=tensor([0.1939, 0.2097, 0.2395, 0.2064, 0.1376, 0.2089, 0.2546, 0.2040], device='cuda:1'), in_proj_covar=tensor([0.0315, 0.0337, 0.0353, 0.0310, 0.0350, 0.0352, 0.0314, 0.0353], device='cuda:1'), out_proj_covar=tensor([6.8012e-05, 7.2231e-05, 7.6551e-05, 6.4914e-05, 7.4369e-05, 7.6667e-05, 6.8224e-05, 7.6176e-05], device='cuda:1') 2023-04-26 15:27:34,614 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-04-26 15:27:46,381 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4725, 2.5538, 1.9185, 2.1859, 2.2970, 1.8145, 3.1724, 1.4807], device='cuda:1'), covar=tensor([0.4550, 0.1433, 0.4060, 0.3244, 0.2369, 0.3043, 0.1696, 0.5303], device='cuda:1'), in_proj_covar=tensor([0.0356, 0.0357, 0.0443, 0.0372, 0.0405, 0.0382, 0.0399, 0.0425], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 15:27:46,873 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.129e+02 1.703e+02 2.110e+02 2.594e+02 5.214e+02, threshold=4.219e+02, percent-clipped=1.0 2023-04-26 15:28:05,809 INFO [finetune.py:976] (1/7) Epoch 4, batch 3850, loss[loss=0.1813, simple_loss=0.2302, pruned_loss=0.06621, over 4703.00 frames. ], tot_loss[loss=0.2221, simple_loss=0.2813, pruned_loss=0.08143, over 954934.71 frames. ], batch size: 23, lr: 3.96e-03, grad_scale: 16.0 2023-04-26 15:28:18,862 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=21044.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:28:29,640 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=21052.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:28:54,707 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=21074.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:29:11,948 INFO [finetune.py:976] (1/7) Epoch 4, batch 3900, loss[loss=0.2625, simple_loss=0.3038, pruned_loss=0.1106, over 4745.00 frames. ], tot_loss[loss=0.2207, simple_loss=0.279, pruned_loss=0.08117, over 955402.92 frames. ], batch size: 27, lr: 3.96e-03, grad_scale: 16.0 2023-04-26 15:29:16,765 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0370, 1.2069, 1.3620, 1.4782, 1.4547, 1.6554, 1.3918, 1.4386], device='cuda:1'), covar=tensor([0.7950, 1.2676, 1.1335, 0.9654, 1.1448, 1.8428, 1.3468, 1.1698], device='cuda:1'), in_proj_covar=tensor([0.0312, 0.0401, 0.0322, 0.0328, 0.0351, 0.0415, 0.0387, 0.0342], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-26 15:29:18,469 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=21092.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:29:31,310 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=21113.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:29:32,533 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4446, 1.1361, 1.2409, 1.1057, 1.6461, 1.3252, 1.0614, 1.2566], device='cuda:1'), covar=tensor([0.1434, 0.1246, 0.1661, 0.1441, 0.0678, 0.1358, 0.1848, 0.1541], device='cuda:1'), in_proj_covar=tensor([0.0312, 0.0334, 0.0350, 0.0307, 0.0347, 0.0349, 0.0311, 0.0349], device='cuda:1'), out_proj_covar=tensor([6.7165e-05, 7.1708e-05, 7.5931e-05, 6.4215e-05, 7.3786e-05, 7.5992e-05, 6.7604e-05, 7.5439e-05], device='cuda:1') 2023-04-26 15:29:34,791 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=21118.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:29:37,129 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=21122.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:29:38,882 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.064e+02 1.817e+02 2.053e+02 2.461e+02 7.075e+02, threshold=4.105e+02, percent-clipped=2.0 2023-04-26 15:29:43,729 INFO [finetune.py:976] (1/7) Epoch 4, batch 3950, loss[loss=0.1926, simple_loss=0.2644, pruned_loss=0.06039, over 4696.00 frames. ], tot_loss[loss=0.2164, simple_loss=0.2744, pruned_loss=0.07918, over 954902.60 frames. ], batch size: 23, lr: 3.96e-03, grad_scale: 16.0 2023-04-26 15:30:14,116 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=21179.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:30:16,453 INFO [finetune.py:976] (1/7) Epoch 4, batch 4000, loss[loss=0.2303, simple_loss=0.2997, pruned_loss=0.08042, over 4915.00 frames. ], tot_loss[loss=0.2154, simple_loss=0.2733, pruned_loss=0.07875, over 953532.59 frames. ], batch size: 36, lr: 3.96e-03, grad_scale: 16.0 2023-04-26 15:30:22,235 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8379, 2.6912, 2.0701, 2.3787, 2.1394, 2.1050, 2.4318, 1.7003], device='cuda:1'), covar=tensor([0.3039, 0.2398, 0.1633, 0.2102, 0.3299, 0.2135, 0.2053, 0.3205], device='cuda:1'), in_proj_covar=tensor([0.0312, 0.0334, 0.0242, 0.0306, 0.0321, 0.0285, 0.0274, 0.0297], device='cuda:1'), out_proj_covar=tensor([1.2729e-04, 1.3590e-04, 9.8734e-05, 1.2355e-04, 1.3240e-04, 1.1510e-04, 1.1318e-04, 1.1993e-04], device='cuda:1') 2023-04-26 15:30:45,238 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.385e+02 1.961e+02 2.328e+02 2.721e+02 5.358e+02, threshold=4.656e+02, percent-clipped=3.0 2023-04-26 15:30:50,170 INFO [finetune.py:976] (1/7) Epoch 4, batch 4050, loss[loss=0.2028, simple_loss=0.2636, pruned_loss=0.071, over 4714.00 frames. ], tot_loss[loss=0.2207, simple_loss=0.2784, pruned_loss=0.08146, over 953006.28 frames. ], batch size: 23, lr: 3.96e-03, grad_scale: 16.0 2023-04-26 15:30:53,317 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([0.9117, 1.9469, 1.7547, 1.5351, 2.0687, 1.6027, 2.5880, 1.5348], device='cuda:1'), covar=tensor([0.4000, 0.1747, 0.4916, 0.3210, 0.1659, 0.2451, 0.1658, 0.4762], device='cuda:1'), in_proj_covar=tensor([0.0353, 0.0355, 0.0439, 0.0370, 0.0401, 0.0380, 0.0397, 0.0422], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 15:30:58,866 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=21244.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:31:01,154 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.1922, 1.4698, 1.3818, 1.6873, 1.5538, 1.8194, 1.3906, 3.3306], device='cuda:1'), covar=tensor([0.0749, 0.0811, 0.0792, 0.1254, 0.0657, 0.0745, 0.0775, 0.0164], device='cuda:1'), in_proj_covar=tensor([0.0040, 0.0040, 0.0042, 0.0046, 0.0041, 0.0041, 0.0040, 0.0063], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0017], device='cuda:1') 2023-04-26 15:31:23,491 INFO [finetune.py:976] (1/7) Epoch 4, batch 4100, loss[loss=0.2451, simple_loss=0.301, pruned_loss=0.09465, over 4859.00 frames. ], tot_loss[loss=0.2237, simple_loss=0.2821, pruned_loss=0.08264, over 952770.29 frames. ], batch size: 44, lr: 3.96e-03, grad_scale: 16.0 2023-04-26 15:31:29,565 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=21292.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:31:43,173 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=21312.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:31:49,254 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([0.3508, 1.3689, 1.3670, 0.9894, 1.3736, 1.1919, 1.7348, 1.2384], device='cuda:1'), covar=tensor([0.4281, 0.1792, 0.5705, 0.3100, 0.1854, 0.2167, 0.1917, 0.5384], device='cuda:1'), in_proj_covar=tensor([0.0356, 0.0357, 0.0442, 0.0371, 0.0404, 0.0383, 0.0398, 0.0425], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 15:31:50,917 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.243e+02 1.937e+02 2.313e+02 2.737e+02 5.004e+02, threshold=4.625e+02, percent-clipped=1.0 2023-04-26 15:31:52,897 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=21328.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:31:56,304 INFO [finetune.py:976] (1/7) Epoch 4, batch 4150, loss[loss=0.2477, simple_loss=0.3059, pruned_loss=0.09477, over 4929.00 frames. ], tot_loss[loss=0.2232, simple_loss=0.282, pruned_loss=0.08223, over 953835.06 frames. ], batch size: 33, lr: 3.95e-03, grad_scale: 16.0 2023-04-26 15:32:20,039 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.7740, 1.9402, 2.2585, 3.0955, 2.8298, 2.2228, 1.8735, 2.5120], device='cuda:1'), covar=tensor([0.0920, 0.1540, 0.0890, 0.0702, 0.0652, 0.1233, 0.1201, 0.0791], device='cuda:1'), in_proj_covar=tensor([0.0207, 0.0207, 0.0185, 0.0180, 0.0180, 0.0197, 0.0169, 0.0191], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 15:32:40,033 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=21373.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:32:44,262 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6297, 3.8928, 0.7942, 1.9377, 2.1723, 2.3832, 2.3194, 0.9773], device='cuda:1'), covar=tensor([0.1324, 0.0819, 0.2109, 0.1413, 0.1007, 0.1194, 0.1345, 0.2335], device='cuda:1'), in_proj_covar=tensor([0.0121, 0.0261, 0.0146, 0.0128, 0.0138, 0.0159, 0.0124, 0.0128], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-26 15:32:46,015 INFO [finetune.py:976] (1/7) Epoch 4, batch 4200, loss[loss=0.2541, simple_loss=0.3088, pruned_loss=0.09967, over 4894.00 frames. ], tot_loss[loss=0.2221, simple_loss=0.2816, pruned_loss=0.08136, over 953449.19 frames. ], batch size: 35, lr: 3.95e-03, grad_scale: 16.0 2023-04-26 15:32:47,837 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9248, 2.5431, 2.1815, 2.3898, 1.9379, 2.0597, 2.2119, 1.6503], device='cuda:1'), covar=tensor([0.2533, 0.1553, 0.0975, 0.1643, 0.3350, 0.1880, 0.2021, 0.3102], device='cuda:1'), in_proj_covar=tensor([0.0313, 0.0334, 0.0242, 0.0306, 0.0322, 0.0286, 0.0274, 0.0297], device='cuda:1'), out_proj_covar=tensor([1.2752e-04, 1.3588e-04, 9.8440e-05, 1.2341e-04, 1.3298e-04, 1.1554e-04, 1.1305e-04, 1.2009e-04], device='cuda:1') 2023-04-26 15:32:50,242 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=21389.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:33:21,601 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=21408.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:33:42,773 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.105e+02 1.624e+02 1.995e+02 2.490e+02 4.928e+02, threshold=3.989e+02, percent-clipped=1.0 2023-04-26 15:33:52,191 INFO [finetune.py:976] (1/7) Epoch 4, batch 4250, loss[loss=0.2042, simple_loss=0.2624, pruned_loss=0.07295, over 4815.00 frames. ], tot_loss[loss=0.2195, simple_loss=0.2785, pruned_loss=0.0803, over 953139.69 frames. ], batch size: 40, lr: 3.95e-03, grad_scale: 16.0 2023-04-26 15:34:03,304 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-04-26 15:34:30,940 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=21474.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:34:41,941 INFO [finetune.py:976] (1/7) Epoch 4, batch 4300, loss[loss=0.1764, simple_loss=0.2428, pruned_loss=0.05496, over 4717.00 frames. ], tot_loss[loss=0.2184, simple_loss=0.2767, pruned_loss=0.08003, over 953397.75 frames. ], batch size: 23, lr: 3.95e-03, grad_scale: 16.0 2023-04-26 15:35:24,193 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7610, 1.6252, 1.9488, 1.9284, 1.9521, 1.6438, 1.7650, 1.8376], device='cuda:1'), covar=tensor([1.2675, 1.6906, 2.0248, 1.9814, 1.2899, 2.2684, 2.3259, 1.8030], device='cuda:1'), in_proj_covar=tensor([0.0427, 0.0470, 0.0557, 0.0577, 0.0459, 0.0484, 0.0498, 0.0501], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 15:35:27,095 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.101e+02 1.707e+02 2.018e+02 2.403e+02 4.975e+02, threshold=4.035e+02, percent-clipped=3.0 2023-04-26 15:35:31,900 INFO [finetune.py:976] (1/7) Epoch 4, batch 4350, loss[loss=0.1723, simple_loss=0.2199, pruned_loss=0.06239, over 4177.00 frames. ], tot_loss[loss=0.2157, simple_loss=0.2736, pruned_loss=0.07894, over 952242.62 frames. ], batch size: 17, lr: 3.95e-03, grad_scale: 16.0 2023-04-26 15:35:54,105 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.36 vs. limit=5.0 2023-04-26 15:36:01,880 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-04-26 15:36:16,125 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.3351, 1.3600, 1.4052, 1.6367, 1.5918, 1.3001, 0.9348, 1.4491], device='cuda:1'), covar=tensor([0.0985, 0.1336, 0.0839, 0.0652, 0.0664, 0.0942, 0.1052, 0.0642], device='cuda:1'), in_proj_covar=tensor([0.0206, 0.0206, 0.0183, 0.0179, 0.0179, 0.0195, 0.0168, 0.0189], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 15:36:37,729 INFO [finetune.py:976] (1/7) Epoch 4, batch 4400, loss[loss=0.2848, simple_loss=0.3455, pruned_loss=0.1121, over 4742.00 frames. ], tot_loss[loss=0.2178, simple_loss=0.2753, pruned_loss=0.08013, over 951743.14 frames. ], batch size: 54, lr: 3.95e-03, grad_scale: 16.0 2023-04-26 15:36:51,049 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=21595.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:37:09,353 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=21613.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:37:17,113 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.164e+02 1.848e+02 2.167e+02 2.564e+02 6.946e+02, threshold=4.335e+02, percent-clipped=3.0 2023-04-26 15:37:22,022 INFO [finetune.py:976] (1/7) Epoch 4, batch 4450, loss[loss=0.2111, simple_loss=0.2863, pruned_loss=0.068, over 4922.00 frames. ], tot_loss[loss=0.22, simple_loss=0.2782, pruned_loss=0.08087, over 951943.67 frames. ], batch size: 42, lr: 3.95e-03, grad_scale: 16.0 2023-04-26 15:37:36,748 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=21656.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:37:46,471 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=21668.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:37:50,677 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=21674.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:37:56,116 INFO [finetune.py:976] (1/7) Epoch 4, batch 4500, loss[loss=0.2157, simple_loss=0.2804, pruned_loss=0.07548, over 4904.00 frames. ], tot_loss[loss=0.2213, simple_loss=0.2799, pruned_loss=0.08132, over 951648.32 frames. ], batch size: 37, lr: 3.95e-03, grad_scale: 16.0 2023-04-26 15:37:56,777 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=21684.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:38:11,958 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=21708.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:38:16,092 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.1573, 2.0221, 2.5000, 2.7445, 1.8478, 1.3757, 2.0745, 1.1267], device='cuda:1'), covar=tensor([0.1163, 0.1002, 0.0645, 0.0798, 0.1104, 0.2372, 0.1083, 0.1713], device='cuda:1'), in_proj_covar=tensor([0.0068, 0.0078, 0.0075, 0.0070, 0.0081, 0.0097, 0.0085, 0.0079], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:1') 2023-04-26 15:38:24,949 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.135e+02 1.777e+02 2.074e+02 2.624e+02 7.543e+02, threshold=4.148e+02, percent-clipped=2.0 2023-04-26 15:38:29,903 INFO [finetune.py:976] (1/7) Epoch 4, batch 4550, loss[loss=0.2223, simple_loss=0.2825, pruned_loss=0.08103, over 4120.00 frames. ], tot_loss[loss=0.2237, simple_loss=0.2825, pruned_loss=0.08249, over 951296.14 frames. ], batch size: 65, lr: 3.95e-03, grad_scale: 16.0 2023-04-26 15:38:35,737 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.99 vs. limit=2.0 2023-04-26 15:38:40,466 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5519, 1.8543, 1.7470, 2.0915, 1.8641, 2.2464, 1.6643, 3.6673], device='cuda:1'), covar=tensor([0.0690, 0.0754, 0.0810, 0.1171, 0.0653, 0.0502, 0.0746, 0.0178], device='cuda:1'), in_proj_covar=tensor([0.0041, 0.0040, 0.0042, 0.0046, 0.0041, 0.0041, 0.0041, 0.0063], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0017], device='cuda:1') 2023-04-26 15:38:44,652 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=21756.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:38:49,058 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=21763.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:38:58,613 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=21774.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:39:04,081 INFO [finetune.py:976] (1/7) Epoch 4, batch 4600, loss[loss=0.1478, simple_loss=0.2193, pruned_loss=0.03811, over 4775.00 frames. ], tot_loss[loss=0.2212, simple_loss=0.2806, pruned_loss=0.08094, over 952881.38 frames. ], batch size: 29, lr: 3.95e-03, grad_scale: 16.0 2023-04-26 15:39:22,027 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.46 vs. limit=5.0 2023-04-26 15:39:29,986 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=21822.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:39:31,723 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=21824.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:39:37,419 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.091e+02 1.786e+02 2.120e+02 2.552e+02 5.015e+02, threshold=4.240e+02, percent-clipped=1.0 2023-04-26 15:39:48,032 INFO [finetune.py:976] (1/7) Epoch 4, batch 4650, loss[loss=0.25, simple_loss=0.3019, pruned_loss=0.09909, over 4903.00 frames. ], tot_loss[loss=0.2192, simple_loss=0.278, pruned_loss=0.08023, over 954231.38 frames. ], batch size: 36, lr: 3.95e-03, grad_scale: 16.0 2023-04-26 15:40:01,626 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6124, 1.1299, 1.2695, 1.2869, 1.8076, 1.4548, 1.1479, 1.1873], device='cuda:1'), covar=tensor([0.1743, 0.1710, 0.2268, 0.1672, 0.0879, 0.1713, 0.2363, 0.2079], device='cuda:1'), in_proj_covar=tensor([0.0317, 0.0338, 0.0355, 0.0310, 0.0351, 0.0352, 0.0315, 0.0355], device='cuda:1'), out_proj_covar=tensor([6.8333e-05, 7.2413e-05, 7.6983e-05, 6.4970e-05, 7.4401e-05, 7.6634e-05, 6.8364e-05, 7.6531e-05], device='cuda:1') 2023-04-26 15:40:04,066 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=21850.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:40:49,527 INFO [finetune.py:976] (1/7) Epoch 4, batch 4700, loss[loss=0.1896, simple_loss=0.2542, pruned_loss=0.06249, over 4770.00 frames. ], tot_loss[loss=0.2158, simple_loss=0.2745, pruned_loss=0.07851, over 956159.93 frames. ], batch size: 28, lr: 3.95e-03, grad_scale: 16.0 2023-04-26 15:41:12,784 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=21911.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:41:21,258 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([0.4526, 1.3897, 1.3950, 0.9921, 1.4081, 1.1226, 1.7066, 1.2256], device='cuda:1'), covar=tensor([0.3548, 0.1542, 0.4404, 0.2688, 0.1611, 0.2228, 0.1758, 0.4714], device='cuda:1'), in_proj_covar=tensor([0.0357, 0.0358, 0.0441, 0.0372, 0.0403, 0.0386, 0.0398, 0.0424], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 15:41:21,740 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.201e+02 1.822e+02 2.098e+02 2.414e+02 4.975e+02, threshold=4.197e+02, percent-clipped=3.0 2023-04-26 15:41:28,677 INFO [finetune.py:976] (1/7) Epoch 4, batch 4750, loss[loss=0.2298, simple_loss=0.2825, pruned_loss=0.08852, over 4767.00 frames. ], tot_loss[loss=0.2147, simple_loss=0.2728, pruned_loss=0.07826, over 956376.28 frames. ], batch size: 26, lr: 3.95e-03, grad_scale: 16.0 2023-04-26 15:41:39,625 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([3.7331, 3.6265, 2.7629, 4.3989, 3.8066, 3.7842, 1.7773, 3.6682], device='cuda:1'), covar=tensor([0.1520, 0.1145, 0.3572, 0.1411, 0.3325, 0.1744, 0.5131, 0.2377], device='cuda:1'), in_proj_covar=tensor([0.0250, 0.0224, 0.0262, 0.0315, 0.0311, 0.0259, 0.0277, 0.0280], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-26 15:41:40,251 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=21951.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:41:42,272 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.76 vs. limit=5.0 2023-04-26 15:41:44,478 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([3.4328, 3.2989, 2.6143, 4.0552, 3.4691, 3.4654, 1.5779, 3.4360], device='cuda:1'), covar=tensor([0.1665, 0.1621, 0.3632, 0.1797, 0.3800, 0.1892, 0.5341, 0.2403], device='cuda:1'), in_proj_covar=tensor([0.0250, 0.0224, 0.0261, 0.0315, 0.0311, 0.0259, 0.0277, 0.0280], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-26 15:41:50,659 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=21968.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:41:51,755 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=21969.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:41:52,416 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=21970.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:42:02,272 INFO [finetune.py:976] (1/7) Epoch 4, batch 4800, loss[loss=0.2355, simple_loss=0.3024, pruned_loss=0.08427, over 4833.00 frames. ], tot_loss[loss=0.2185, simple_loss=0.277, pruned_loss=0.07999, over 956534.56 frames. ], batch size: 47, lr: 3.95e-03, grad_scale: 16.0 2023-04-26 15:42:03,019 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=21984.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:42:39,321 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=22016.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:42:50,413 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.187e+02 1.864e+02 2.140e+02 2.595e+02 5.123e+02, threshold=4.279e+02, percent-clipped=1.0 2023-04-26 15:42:54,222 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=22031.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:42:55,317 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=22032.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:42:56,373 INFO [finetune.py:976] (1/7) Epoch 4, batch 4850, loss[loss=0.226, simple_loss=0.2913, pruned_loss=0.0804, over 4855.00 frames. ], tot_loss[loss=0.2202, simple_loss=0.2788, pruned_loss=0.0808, over 955290.87 frames. ], batch size: 31, lr: 3.95e-03, grad_scale: 16.0 2023-04-26 15:43:12,688 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.1716, 1.4731, 1.3224, 1.6820, 1.5254, 1.6870, 1.3819, 2.9481], device='cuda:1'), covar=tensor([0.0684, 0.0703, 0.0784, 0.1148, 0.0610, 0.0618, 0.0703, 0.0192], device='cuda:1'), in_proj_covar=tensor([0.0040, 0.0040, 0.0042, 0.0046, 0.0041, 0.0041, 0.0041, 0.0063], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0017], device='cuda:1') 2023-04-26 15:43:45,552 INFO [finetune.py:976] (1/7) Epoch 4, batch 4900, loss[loss=0.2075, simple_loss=0.2817, pruned_loss=0.06665, over 4902.00 frames. ], tot_loss[loss=0.222, simple_loss=0.2807, pruned_loss=0.08163, over 955173.69 frames. ], batch size: 35, lr: 3.95e-03, grad_scale: 16.0 2023-04-26 15:44:08,989 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8249, 2.2552, 1.8899, 2.2038, 1.7171, 1.8633, 2.0347, 1.5420], device='cuda:1'), covar=tensor([0.2242, 0.1375, 0.1040, 0.1227, 0.3021, 0.1390, 0.1838, 0.2686], device='cuda:1'), in_proj_covar=tensor([0.0310, 0.0332, 0.0241, 0.0305, 0.0322, 0.0286, 0.0274, 0.0296], device='cuda:1'), out_proj_covar=tensor([1.2643e-04, 1.3501e-04, 9.8100e-05, 1.2267e-04, 1.3299e-04, 1.1569e-04, 1.1284e-04, 1.1956e-04], device='cuda:1') 2023-04-26 15:44:15,744 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=22119.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:44:19,291 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.258e+02 1.858e+02 2.268e+02 2.601e+02 5.071e+02, threshold=4.535e+02, percent-clipped=2.0 2023-04-26 15:44:22,973 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=22130.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:44:24,755 INFO [finetune.py:976] (1/7) Epoch 4, batch 4950, loss[loss=0.217, simple_loss=0.2734, pruned_loss=0.08029, over 4850.00 frames. ], tot_loss[loss=0.2225, simple_loss=0.282, pruned_loss=0.08155, over 955509.26 frames. ], batch size: 44, lr: 3.95e-03, grad_scale: 16.0 2023-04-26 15:44:41,584 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.70 vs. limit=2.0 2023-04-26 15:44:51,208 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([0.9822, 2.4133, 0.7949, 1.2639, 1.4461, 1.8497, 1.6084, 0.7777], device='cuda:1'), covar=tensor([0.2011, 0.1824, 0.2331, 0.2130, 0.1403, 0.1393, 0.1793, 0.1981], device='cuda:1'), in_proj_covar=tensor([0.0121, 0.0261, 0.0147, 0.0128, 0.0138, 0.0160, 0.0124, 0.0127], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-26 15:44:53,684 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=22176.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:44:58,356 INFO [finetune.py:976] (1/7) Epoch 4, batch 5000, loss[loss=0.2721, simple_loss=0.3098, pruned_loss=0.1172, over 4830.00 frames. ], tot_loss[loss=0.2195, simple_loss=0.2792, pruned_loss=0.07991, over 955742.79 frames. ], batch size: 47, lr: 3.95e-03, grad_scale: 16.0 2023-04-26 15:45:03,887 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=22191.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:45:14,451 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.1411, 2.2048, 1.9482, 1.8368, 2.2749, 1.9121, 2.8101, 1.7148], device='cuda:1'), covar=tensor([0.4188, 0.1706, 0.4531, 0.3512, 0.1952, 0.2357, 0.1542, 0.4292], device='cuda:1'), in_proj_covar=tensor([0.0357, 0.0358, 0.0440, 0.0372, 0.0403, 0.0384, 0.0399, 0.0423], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 15:45:15,014 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=22206.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:45:20,508 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=22215.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:45:28,877 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-04-26 15:45:31,709 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.348e+02 1.793e+02 1.980e+02 2.407e+02 3.890e+02, threshold=3.961e+02, percent-clipped=0.0 2023-04-26 15:45:38,435 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.1958, 2.6763, 0.9892, 1.4203, 1.9621, 1.3009, 3.4301, 1.7820], device='cuda:1'), covar=tensor([0.0596, 0.0671, 0.0844, 0.1135, 0.0531, 0.0916, 0.0211, 0.0612], device='cuda:1'), in_proj_covar=tensor([0.0054, 0.0070, 0.0053, 0.0049, 0.0054, 0.0055, 0.0082, 0.0052], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0011, 0.0008], device='cuda:1') 2023-04-26 15:45:42,549 INFO [finetune.py:976] (1/7) Epoch 4, batch 5050, loss[loss=0.1777, simple_loss=0.2393, pruned_loss=0.05804, over 4769.00 frames. ], tot_loss[loss=0.2166, simple_loss=0.2757, pruned_loss=0.07869, over 954757.33 frames. ], batch size: 54, lr: 3.95e-03, grad_scale: 16.0 2023-04-26 15:45:45,107 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=22237.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:45:45,118 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=22237.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:46:06,550 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=22251.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:46:28,174 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=22269.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:46:32,413 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=22276.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:46:42,366 INFO [finetune.py:976] (1/7) Epoch 4, batch 5100, loss[loss=0.2082, simple_loss=0.2652, pruned_loss=0.07562, over 4826.00 frames. ], tot_loss[loss=0.2147, simple_loss=0.2732, pruned_loss=0.07813, over 955259.89 frames. ], batch size: 40, lr: 3.95e-03, grad_scale: 16.0 2023-04-26 15:46:50,207 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.0133, 2.1158, 1.7708, 1.6602, 2.1000, 1.7323, 2.6024, 1.5072], device='cuda:1'), covar=tensor([0.3912, 0.1704, 0.4700, 0.3012, 0.1814, 0.2565, 0.1466, 0.4655], device='cuda:1'), in_proj_covar=tensor([0.0355, 0.0356, 0.0438, 0.0370, 0.0402, 0.0384, 0.0398, 0.0422], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 15:46:52,065 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=22298.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:46:53,600 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=22299.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:46:57,726 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-04-26 15:47:06,006 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=22317.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:47:10,807 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.090e+02 1.713e+02 2.102e+02 2.590e+02 4.893e+02, threshold=4.205e+02, percent-clipped=2.0 2023-04-26 15:47:11,504 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=22326.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:47:15,715 INFO [finetune.py:976] (1/7) Epoch 4, batch 5150, loss[loss=0.245, simple_loss=0.3002, pruned_loss=0.09492, over 4777.00 frames. ], tot_loss[loss=0.2151, simple_loss=0.2731, pruned_loss=0.07853, over 954135.60 frames. ], batch size: 28, lr: 3.95e-03, grad_scale: 16.0 2023-04-26 15:47:16,583 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.72 vs. limit=5.0 2023-04-26 15:47:52,651 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9615, 1.4613, 1.8248, 2.1395, 1.7549, 1.4141, 1.1026, 1.5276], device='cuda:1'), covar=tensor([0.4321, 0.4848, 0.2222, 0.3861, 0.4042, 0.3547, 0.6338, 0.3760], device='cuda:1'), in_proj_covar=tensor([0.0273, 0.0264, 0.0221, 0.0333, 0.0223, 0.0230, 0.0249, 0.0199], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-26 15:47:54,331 INFO [finetune.py:976] (1/7) Epoch 4, batch 5200, loss[loss=0.2706, simple_loss=0.3198, pruned_loss=0.1107, over 4809.00 frames. ], tot_loss[loss=0.2202, simple_loss=0.2783, pruned_loss=0.08104, over 954348.19 frames. ], batch size: 41, lr: 3.95e-03, grad_scale: 16.0 2023-04-26 15:48:01,222 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=22385.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:48:05,962 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6724, 1.5126, 1.6751, 1.9965, 2.0584, 1.6274, 1.4125, 1.7023], device='cuda:1'), covar=tensor([0.0995, 0.1213, 0.0764, 0.0637, 0.0625, 0.1029, 0.1022, 0.0665], device='cuda:1'), in_proj_covar=tensor([0.0207, 0.0208, 0.0186, 0.0182, 0.0181, 0.0197, 0.0170, 0.0193], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 15:48:22,642 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4792, 0.9626, 1.2862, 1.1153, 1.6383, 1.3827, 1.0466, 1.2387], device='cuda:1'), covar=tensor([0.1579, 0.1392, 0.1908, 0.1499, 0.0873, 0.1386, 0.2053, 0.2012], device='cuda:1'), in_proj_covar=tensor([0.0316, 0.0335, 0.0353, 0.0310, 0.0349, 0.0350, 0.0316, 0.0354], device='cuda:1'), out_proj_covar=tensor([6.8212e-05, 7.1884e-05, 7.6489e-05, 6.4778e-05, 7.3894e-05, 7.6299e-05, 6.8505e-05, 7.6373e-05], device='cuda:1') 2023-04-26 15:48:30,904 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=22419.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:48:34,446 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.341e+02 1.984e+02 2.270e+02 2.624e+02 8.595e+02, threshold=4.540e+02, percent-clipped=4.0 2023-04-26 15:48:39,404 INFO [finetune.py:976] (1/7) Epoch 4, batch 5250, loss[loss=0.2383, simple_loss=0.2843, pruned_loss=0.09611, over 4845.00 frames. ], tot_loss[loss=0.2221, simple_loss=0.2806, pruned_loss=0.08181, over 953712.58 frames. ], batch size: 49, lr: 3.95e-03, grad_scale: 16.0 2023-04-26 15:48:47,971 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=22446.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:49:01,866 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2023-04-26 15:49:09,417 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=22467.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:49:25,137 INFO [finetune.py:976] (1/7) Epoch 4, batch 5300, loss[loss=0.2261, simple_loss=0.291, pruned_loss=0.08053, over 4751.00 frames. ], tot_loss[loss=0.222, simple_loss=0.2812, pruned_loss=0.08141, over 954674.90 frames. ], batch size: 27, lr: 3.95e-03, grad_scale: 16.0 2023-04-26 15:49:32,901 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=22486.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:49:56,902 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=22506.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:50:15,636 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.302e+02 1.836e+02 2.188e+02 2.701e+02 5.070e+02, threshold=4.376e+02, percent-clipped=2.0 2023-04-26 15:50:19,965 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=22532.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:50:20,503 INFO [finetune.py:976] (1/7) Epoch 4, batch 5350, loss[loss=0.1893, simple_loss=0.266, pruned_loss=0.05634, over 4837.00 frames. ], tot_loss[loss=0.2203, simple_loss=0.2802, pruned_loss=0.08019, over 954413.21 frames. ], batch size: 47, lr: 3.95e-03, grad_scale: 32.0 2023-04-26 15:50:33,844 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=22554.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:50:46,633 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=22571.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:50:53,879 INFO [finetune.py:976] (1/7) Epoch 4, batch 5400, loss[loss=0.1913, simple_loss=0.2461, pruned_loss=0.06827, over 4800.00 frames. ], tot_loss[loss=0.218, simple_loss=0.2778, pruned_loss=0.07903, over 954589.62 frames. ], batch size: 25, lr: 3.95e-03, grad_scale: 32.0 2023-04-26 15:50:59,985 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=22593.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:51:06,188 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.3500, 0.5614, 1.0826, 1.7342, 1.5530, 1.2359, 1.1981, 1.2640], device='cuda:1'), covar=tensor([1.0938, 1.4151, 1.5808, 1.8763, 1.1985, 1.7170, 1.8225, 1.4924], device='cuda:1'), in_proj_covar=tensor([0.0427, 0.0468, 0.0555, 0.0575, 0.0459, 0.0485, 0.0497, 0.0499], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 15:51:20,352 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.3153, 1.6384, 1.5523, 2.0292, 1.7117, 2.1603, 1.4717, 3.8222], device='cuda:1'), covar=tensor([0.0703, 0.0765, 0.0849, 0.1185, 0.0676, 0.0510, 0.0797, 0.0173], device='cuda:1'), in_proj_covar=tensor([0.0040, 0.0040, 0.0042, 0.0046, 0.0041, 0.0041, 0.0040, 0.0063], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0017], device='cuda:1') 2023-04-26 15:51:38,541 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.141e+02 1.846e+02 2.198e+02 2.575e+02 4.196e+02, threshold=4.395e+02, percent-clipped=1.0 2023-04-26 15:51:39,233 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=22626.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:51:47,649 INFO [finetune.py:976] (1/7) Epoch 4, batch 5450, loss[loss=0.1924, simple_loss=0.2536, pruned_loss=0.0656, over 4813.00 frames. ], tot_loss[loss=0.2152, simple_loss=0.2745, pruned_loss=0.07795, over 952987.95 frames. ], batch size: 30, lr: 3.95e-03, grad_scale: 32.0 2023-04-26 15:51:51,436 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=22639.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:52:39,603 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=22674.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:52:50,912 INFO [finetune.py:976] (1/7) Epoch 4, batch 5500, loss[loss=0.2316, simple_loss=0.2806, pruned_loss=0.09131, over 4862.00 frames. ], tot_loss[loss=0.2133, simple_loss=0.2718, pruned_loss=0.07737, over 954034.98 frames. ], batch size: 49, lr: 3.95e-03, grad_scale: 32.0 2023-04-26 15:53:12,597 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=22700.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 15:53:25,984 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.2316, 1.4849, 1.5065, 1.6590, 1.5836, 1.7601, 1.6250, 1.5634], device='cuda:1'), covar=tensor([0.7403, 1.2412, 1.0023, 0.8891, 1.1074, 1.6688, 1.2427, 1.0907], device='cuda:1'), in_proj_covar=tensor([0.0311, 0.0398, 0.0319, 0.0325, 0.0350, 0.0413, 0.0385, 0.0339], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-26 15:53:30,817 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6698, 1.3046, 1.7293, 1.9756, 1.8034, 1.6449, 1.7468, 1.7262], device='cuda:1'), covar=tensor([1.2965, 1.7653, 1.9502, 2.2572, 1.5027, 2.1874, 2.0411, 1.7286], device='cuda:1'), in_proj_covar=tensor([0.0426, 0.0467, 0.0554, 0.0574, 0.0458, 0.0484, 0.0496, 0.0498], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 15:53:34,589 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.236e+02 2.045e+02 2.332e+02 2.799e+02 5.603e+02, threshold=4.665e+02, percent-clipped=2.0 2023-04-26 15:53:37,066 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6442, 2.4102, 1.7168, 1.7160, 1.2401, 1.3232, 1.8487, 1.2185], device='cuda:1'), covar=tensor([0.1980, 0.1590, 0.1825, 0.2204, 0.3031, 0.2313, 0.1372, 0.2445], device='cuda:1'), in_proj_covar=tensor([0.0194, 0.0220, 0.0179, 0.0209, 0.0218, 0.0188, 0.0174, 0.0196], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-26 15:53:38,792 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5354, 1.4940, 0.5957, 1.2476, 1.6708, 1.4153, 1.2893, 1.3535], device='cuda:1'), covar=tensor([0.0549, 0.0405, 0.0459, 0.0594, 0.0332, 0.0564, 0.0520, 0.0632], device='cuda:1'), in_proj_covar=tensor([0.0031, 0.0027, 0.0025, 0.0031, 0.0022, 0.0031, 0.0030, 0.0031], device='cuda:1'), out_proj_covar=tensor([0.0049, 0.0045, 0.0038, 0.0049, 0.0038, 0.0048, 0.0047, 0.0050], device='cuda:1') 2023-04-26 15:53:40,510 INFO [finetune.py:976] (1/7) Epoch 4, batch 5550, loss[loss=0.2232, simple_loss=0.2789, pruned_loss=0.08373, over 4817.00 frames. ], tot_loss[loss=0.2169, simple_loss=0.2756, pruned_loss=0.07908, over 956112.27 frames. ], batch size: 38, lr: 3.95e-03, grad_scale: 32.0 2023-04-26 15:53:45,457 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=22741.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:54:11,073 INFO [finetune.py:976] (1/7) Epoch 4, batch 5600, loss[loss=0.2711, simple_loss=0.3346, pruned_loss=0.1038, over 4899.00 frames. ], tot_loss[loss=0.2217, simple_loss=0.2805, pruned_loss=0.08147, over 954560.43 frames. ], batch size: 35, lr: 3.95e-03, grad_scale: 32.0 2023-04-26 15:54:12,891 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=22786.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:54:18,058 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=22794.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:54:37,300 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.226e+02 1.792e+02 2.075e+02 2.605e+02 4.713e+02, threshold=4.150e+02, percent-clipped=1.0 2023-04-26 15:54:37,405 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=22825.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:54:37,415 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.5441, 1.6547, 1.7792, 2.4317, 2.6293, 2.2527, 1.9945, 1.9325], device='cuda:1'), covar=tensor([0.2130, 0.2256, 0.2096, 0.1951, 0.1445, 0.1933, 0.2731, 0.2139], device='cuda:1'), in_proj_covar=tensor([0.0316, 0.0336, 0.0353, 0.0309, 0.0349, 0.0351, 0.0316, 0.0355], device='cuda:1'), out_proj_covar=tensor([6.8184e-05, 7.1996e-05, 7.6556e-05, 6.4657e-05, 7.3892e-05, 7.6391e-05, 6.8539e-05, 7.6683e-05], device='cuda:1') 2023-04-26 15:54:41,470 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=22832.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:54:42,026 INFO [finetune.py:976] (1/7) Epoch 4, batch 5650, loss[loss=0.1997, simple_loss=0.2746, pruned_loss=0.06243, over 4762.00 frames. ], tot_loss[loss=0.2232, simple_loss=0.2823, pruned_loss=0.08207, over 953323.42 frames. ], batch size: 28, lr: 3.95e-03, grad_scale: 32.0 2023-04-26 15:54:42,644 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=22834.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:54:42,859 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.73 vs. limit=2.0 2023-04-26 15:54:55,187 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=22855.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:55:15,924 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=22871.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:55:27,439 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=22880.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:55:29,216 INFO [finetune.py:976] (1/7) Epoch 4, batch 5700, loss[loss=0.1953, simple_loss=0.2409, pruned_loss=0.07485, over 4151.00 frames. ], tot_loss[loss=0.2199, simple_loss=0.2773, pruned_loss=0.08125, over 932645.48 frames. ], batch size: 18, lr: 3.95e-03, grad_scale: 32.0 2023-04-26 15:55:29,911 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([0.6326, 1.6787, 1.8309, 1.3663, 1.8251, 1.4145, 2.2763, 1.5364], device='cuda:1'), covar=tensor([0.3820, 0.1706, 0.4881, 0.2853, 0.1432, 0.2362, 0.1584, 0.4365], device='cuda:1'), in_proj_covar=tensor([0.0356, 0.0358, 0.0441, 0.0370, 0.0403, 0.0385, 0.0399, 0.0425], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 15:55:31,128 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=22886.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:55:40,637 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=22893.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:56:19,324 INFO [finetune.py:976] (1/7) Epoch 5, batch 0, loss[loss=0.2574, simple_loss=0.3143, pruned_loss=0.1003, over 4920.00 frames. ], tot_loss[loss=0.2574, simple_loss=0.3143, pruned_loss=0.1003, over 4920.00 frames. ], batch size: 42, lr: 3.95e-03, grad_scale: 32.0 2023-04-26 15:56:19,324 INFO [finetune.py:1001] (1/7) Computing validation loss 2023-04-26 15:56:29,192 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4731, 1.2589, 1.6967, 1.5743, 1.3587, 1.2115, 1.3931, 0.8903], device='cuda:1'), covar=tensor([0.0776, 0.1219, 0.0747, 0.1156, 0.1084, 0.1698, 0.0903, 0.1247], device='cuda:1'), in_proj_covar=tensor([0.0068, 0.0077, 0.0075, 0.0069, 0.0080, 0.0096, 0.0083, 0.0078], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-26 15:56:30,075 INFO [finetune.py:1010] (1/7) Epoch 5, validation: loss=0.1632, simple_loss=0.2369, pruned_loss=0.04473, over 2265189.00 frames. 2023-04-26 15:56:30,075 INFO [finetune.py:1011] (1/7) Maximum memory allocated so far is 6391MB 2023-04-26 15:56:33,899 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6682, 1.4905, 1.9055, 2.0412, 1.9557, 1.5729, 1.7350, 1.7686], device='cuda:1'), covar=tensor([1.2500, 1.6993, 1.8666, 1.8188, 1.2788, 2.1046, 2.1814, 1.7671], device='cuda:1'), in_proj_covar=tensor([0.0426, 0.0466, 0.0554, 0.0573, 0.0458, 0.0483, 0.0495, 0.0498], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 15:56:35,032 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=22919.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:56:38,536 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.121e+02 1.609e+02 1.975e+02 2.369e+02 5.506e+02, threshold=3.950e+02, percent-clipped=1.0 2023-04-26 15:56:48,359 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=22941.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:57:02,117 INFO [finetune.py:976] (1/7) Epoch 5, batch 50, loss[loss=0.2725, simple_loss=0.3065, pruned_loss=0.1192, over 4918.00 frames. ], tot_loss[loss=0.2222, simple_loss=0.2818, pruned_loss=0.08127, over 216906.53 frames. ], batch size: 33, lr: 3.95e-03, grad_scale: 32.0 2023-04-26 15:57:30,148 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=22995.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 15:57:52,330 INFO [finetune.py:976] (1/7) Epoch 5, batch 100, loss[loss=0.1813, simple_loss=0.2499, pruned_loss=0.05631, over 4791.00 frames. ], tot_loss[loss=0.2173, simple_loss=0.2762, pruned_loss=0.07919, over 380200.18 frames. ], batch size: 29, lr: 3.95e-03, grad_scale: 32.0 2023-04-26 15:58:02,669 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.276e+01 1.854e+02 2.138e+02 2.662e+02 6.636e+02, threshold=4.277e+02, percent-clipped=4.0 2023-04-26 15:58:10,112 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-04-26 15:58:12,551 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=23041.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:58:20,376 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.0847, 1.4433, 1.4312, 1.7313, 1.5616, 1.8620, 1.3084, 3.3722], device='cuda:1'), covar=tensor([0.0729, 0.0811, 0.0795, 0.1284, 0.0688, 0.0541, 0.0815, 0.0165], device='cuda:1'), in_proj_covar=tensor([0.0040, 0.0040, 0.0042, 0.0046, 0.0041, 0.0041, 0.0040, 0.0063], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0017], device='cuda:1') 2023-04-26 15:58:25,553 INFO [finetune.py:976] (1/7) Epoch 5, batch 150, loss[loss=0.225, simple_loss=0.2808, pruned_loss=0.08456, over 4821.00 frames. ], tot_loss[loss=0.2117, simple_loss=0.2696, pruned_loss=0.0769, over 506857.62 frames. ], batch size: 39, lr: 3.95e-03, grad_scale: 32.0 2023-04-26 15:58:37,397 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.0074, 1.3316, 1.2484, 1.6091, 1.4880, 1.4899, 1.3150, 2.4377], device='cuda:1'), covar=tensor([0.0702, 0.0831, 0.0872, 0.1307, 0.0683, 0.0585, 0.0797, 0.0269], device='cuda:1'), in_proj_covar=tensor([0.0040, 0.0040, 0.0042, 0.0046, 0.0041, 0.0041, 0.0040, 0.0063], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0017], device='cuda:1') 2023-04-26 15:58:44,621 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=23089.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:58:57,338 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=23108.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:58:59,550 INFO [finetune.py:976] (1/7) Epoch 5, batch 200, loss[loss=0.2182, simple_loss=0.2855, pruned_loss=0.07543, over 4902.00 frames. ], tot_loss[loss=0.2107, simple_loss=0.2684, pruned_loss=0.07656, over 606799.80 frames. ], batch size: 43, lr: 3.95e-03, grad_scale: 32.0 2023-04-26 15:59:09,623 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.3474, 1.5730, 1.5135, 1.6830, 1.5642, 1.7224, 1.6980, 1.6870], device='cuda:1'), covar=tensor([0.8172, 1.2171, 1.0949, 0.9425, 1.1862, 1.8125, 1.2912, 1.1227], device='cuda:1'), in_proj_covar=tensor([0.0311, 0.0397, 0.0320, 0.0326, 0.0349, 0.0413, 0.0384, 0.0339], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-26 15:59:10,064 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.278e+02 1.706e+02 2.083e+02 2.458e+02 7.322e+02, threshold=4.166e+02, percent-clipped=1.0 2023-04-26 15:59:23,672 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.48 vs. limit=2.0 2023-04-26 15:59:25,393 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=23150.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:59:33,030 INFO [finetune.py:976] (1/7) Epoch 5, batch 250, loss[loss=0.2332, simple_loss=0.2999, pruned_loss=0.08325, over 4909.00 frames. ], tot_loss[loss=0.2146, simple_loss=0.273, pruned_loss=0.07812, over 682613.28 frames. ], batch size: 37, lr: 3.95e-03, grad_scale: 32.0 2023-04-26 15:59:38,540 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=23169.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:59:47,771 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=23181.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 16:00:06,064 INFO [finetune.py:976] (1/7) Epoch 5, batch 300, loss[loss=0.2149, simple_loss=0.2833, pruned_loss=0.07321, over 4911.00 frames. ], tot_loss[loss=0.2162, simple_loss=0.2759, pruned_loss=0.07827, over 744116.40 frames. ], batch size: 37, lr: 3.95e-03, grad_scale: 32.0 2023-04-26 16:00:10,761 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=23217.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 16:00:15,563 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.275e+02 1.925e+02 2.300e+02 2.815e+02 4.609e+02, threshold=4.600e+02, percent-clipped=3.0 2023-04-26 16:00:18,578 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=23229.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 16:00:37,773 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5147, 1.3528, 4.3430, 4.0523, 3.7613, 4.0886, 4.0267, 3.8173], device='cuda:1'), covar=tensor([0.7209, 0.5730, 0.1020, 0.1793, 0.1279, 0.1355, 0.1525, 0.1304], device='cuda:1'), in_proj_covar=tensor([0.0318, 0.0307, 0.0422, 0.0428, 0.0362, 0.0416, 0.0324, 0.0383], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-26 16:00:55,702 INFO [finetune.py:976] (1/7) Epoch 5, batch 350, loss[loss=0.2119, simple_loss=0.2831, pruned_loss=0.07032, over 4752.00 frames. ], tot_loss[loss=0.2183, simple_loss=0.2784, pruned_loss=0.07906, over 791718.68 frames. ], batch size: 26, lr: 3.95e-03, grad_scale: 32.0 2023-04-26 16:01:18,169 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=23278.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 16:01:31,691 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=23290.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 16:01:40,978 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=23295.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 16:01:53,627 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=3.93 vs. limit=5.0 2023-04-26 16:01:55,882 INFO [finetune.py:976] (1/7) Epoch 5, batch 400, loss[loss=0.1878, simple_loss=0.2574, pruned_loss=0.05908, over 4826.00 frames. ], tot_loss[loss=0.2202, simple_loss=0.2804, pruned_loss=0.08, over 825397.82 frames. ], batch size: 49, lr: 3.95e-03, grad_scale: 32.0 2023-04-26 16:02:03,711 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8264, 1.7397, 2.1293, 2.1810, 2.0592, 1.7058, 1.8877, 1.9572], device='cuda:1'), covar=tensor([1.2518, 1.6333, 1.7334, 1.9375, 1.3953, 2.2277, 2.0836, 1.6534], device='cuda:1'), in_proj_covar=tensor([0.0427, 0.0466, 0.0555, 0.0575, 0.0458, 0.0483, 0.0496, 0.0499], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 16:02:05,328 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.110e+02 1.686e+02 2.112e+02 2.575e+02 6.256e+02, threshold=4.223e+02, percent-clipped=1.0 2023-04-26 16:02:07,807 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0749, 1.2227, 1.3483, 1.5166, 1.4718, 1.6783, 1.4695, 1.4803], device='cuda:1'), covar=tensor([0.6454, 1.0596, 0.9744, 0.8528, 0.9754, 1.5330, 1.1072, 0.9746], device='cuda:1'), in_proj_covar=tensor([0.0312, 0.0399, 0.0320, 0.0327, 0.0350, 0.0414, 0.0384, 0.0338], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-26 16:02:18,849 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=23343.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 16:02:29,820 INFO [finetune.py:976] (1/7) Epoch 5, batch 450, loss[loss=0.183, simple_loss=0.2399, pruned_loss=0.06299, over 4825.00 frames. ], tot_loss[loss=0.2176, simple_loss=0.2778, pruned_loss=0.07869, over 852549.50 frames. ], batch size: 25, lr: 3.95e-03, grad_scale: 32.0 2023-04-26 16:03:14,686 INFO [finetune.py:976] (1/7) Epoch 5, batch 500, loss[loss=0.2581, simple_loss=0.3106, pruned_loss=0.1028, over 4871.00 frames. ], tot_loss[loss=0.2163, simple_loss=0.2756, pruned_loss=0.07853, over 871865.01 frames. ], batch size: 34, lr: 3.95e-03, grad_scale: 32.0 2023-04-26 16:03:29,843 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.283e+02 1.686e+02 2.056e+02 2.772e+02 5.396e+02, threshold=4.112e+02, percent-clipped=3.0 2023-04-26 16:03:36,487 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.4009, 2.9406, 2.5963, 2.6261, 2.2243, 2.4902, 2.6542, 2.2137], device='cuda:1'), covar=tensor([0.2608, 0.1713, 0.1014, 0.2060, 0.3403, 0.1603, 0.2292, 0.2980], device='cuda:1'), in_proj_covar=tensor([0.0311, 0.0331, 0.0240, 0.0304, 0.0323, 0.0287, 0.0273, 0.0296], device='cuda:1'), out_proj_covar=tensor([1.2687e-04, 1.3440e-04, 9.7816e-05, 1.2254e-04, 1.3331e-04, 1.1579e-04, 1.1258e-04, 1.1950e-04], device='cuda:1') 2023-04-26 16:03:47,516 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=23450.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 16:03:50,741 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-04-26 16:03:54,146 INFO [finetune.py:976] (1/7) Epoch 5, batch 550, loss[loss=0.1901, simple_loss=0.251, pruned_loss=0.06457, over 4880.00 frames. ], tot_loss[loss=0.2146, simple_loss=0.2732, pruned_loss=0.07802, over 888397.99 frames. ], batch size: 34, lr: 3.95e-03, grad_scale: 32.0 2023-04-26 16:03:56,065 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=23464.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 16:04:00,007 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.43 vs. limit=5.0 2023-04-26 16:04:07,424 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=23481.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 16:04:12,198 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=23488.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 16:04:16,843 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5092, 1.9828, 1.3899, 1.2183, 1.1954, 1.2293, 1.4679, 1.1093], device='cuda:1'), covar=tensor([0.2034, 0.1623, 0.1878, 0.2266, 0.2868, 0.2203, 0.1391, 0.2443], device='cuda:1'), in_proj_covar=tensor([0.0194, 0.0221, 0.0180, 0.0209, 0.0218, 0.0189, 0.0174, 0.0197], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-26 16:04:19,691 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=23498.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 16:04:27,578 INFO [finetune.py:976] (1/7) Epoch 5, batch 600, loss[loss=0.1792, simple_loss=0.2325, pruned_loss=0.06291, over 4893.00 frames. ], tot_loss[loss=0.2151, simple_loss=0.2733, pruned_loss=0.07843, over 903648.36 frames. ], batch size: 32, lr: 3.94e-03, grad_scale: 32.0 2023-04-26 16:04:36,131 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.343e+02 1.963e+02 2.277e+02 2.707e+02 6.010e+02, threshold=4.553e+02, percent-clipped=1.0 2023-04-26 16:04:39,606 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=23529.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 16:04:53,898 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=23549.0, num_to_drop=1, layers_to_drop={2} 2023-04-26 16:05:01,024 INFO [finetune.py:976] (1/7) Epoch 5, batch 650, loss[loss=0.2911, simple_loss=0.3535, pruned_loss=0.1143, over 4766.00 frames. ], tot_loss[loss=0.2187, simple_loss=0.2777, pruned_loss=0.07982, over 915398.25 frames. ], batch size: 54, lr: 3.94e-03, grad_scale: 32.0 2023-04-26 16:05:08,391 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=23573.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 16:05:16,708 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=23585.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 16:05:22,718 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.1260, 1.5627, 1.4288, 1.8616, 1.7115, 1.7633, 1.4452, 3.3934], device='cuda:1'), covar=tensor([0.0735, 0.0767, 0.0851, 0.1233, 0.0661, 0.0564, 0.0757, 0.0177], device='cuda:1'), in_proj_covar=tensor([0.0040, 0.0040, 0.0042, 0.0046, 0.0041, 0.0041, 0.0040, 0.0063], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0017], device='cuda:1') 2023-04-26 16:05:34,426 INFO [finetune.py:976] (1/7) Epoch 5, batch 700, loss[loss=0.2456, simple_loss=0.3046, pruned_loss=0.09335, over 4821.00 frames. ], tot_loss[loss=0.2212, simple_loss=0.2802, pruned_loss=0.08111, over 923829.87 frames. ], batch size: 40, lr: 3.94e-03, grad_scale: 32.0 2023-04-26 16:05:42,878 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.112e+02 1.928e+02 2.468e+02 2.939e+02 6.493e+02, threshold=4.936e+02, percent-clipped=4.0 2023-04-26 16:06:04,224 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9774, 2.7178, 1.0825, 1.3843, 2.0110, 1.3057, 3.4026, 1.7181], device='cuda:1'), covar=tensor([0.0707, 0.0744, 0.0908, 0.1219, 0.0518, 0.0966, 0.0232, 0.0659], device='cuda:1'), in_proj_covar=tensor([0.0054, 0.0070, 0.0053, 0.0049, 0.0054, 0.0055, 0.0083, 0.0053], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0008, 0.0008, 0.0008, 0.0008, 0.0011, 0.0008], device='cuda:1') 2023-04-26 16:06:19,211 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([0.6975, 1.7111, 1.5908, 1.2881, 1.8067, 1.4109, 2.2384, 1.4048], device='cuda:1'), covar=tensor([0.3557, 0.1468, 0.4232, 0.2835, 0.1579, 0.2073, 0.1250, 0.4251], device='cuda:1'), in_proj_covar=tensor([0.0354, 0.0357, 0.0439, 0.0372, 0.0401, 0.0386, 0.0398, 0.0422], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 16:06:19,709 INFO [finetune.py:976] (1/7) Epoch 5, batch 750, loss[loss=0.2684, simple_loss=0.3129, pruned_loss=0.112, over 4775.00 frames. ], tot_loss[loss=0.2212, simple_loss=0.2803, pruned_loss=0.08105, over 928148.34 frames. ], batch size: 59, lr: 3.94e-03, grad_scale: 32.0 2023-04-26 16:07:26,345 INFO [finetune.py:976] (1/7) Epoch 5, batch 800, loss[loss=0.2298, simple_loss=0.2833, pruned_loss=0.08816, over 4919.00 frames. ], tot_loss[loss=0.2198, simple_loss=0.2796, pruned_loss=0.08001, over 933957.77 frames. ], batch size: 33, lr: 3.94e-03, grad_scale: 32.0 2023-04-26 16:07:34,819 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.218e+02 1.744e+02 2.077e+02 2.568e+02 5.488e+02, threshold=4.154e+02, percent-clipped=2.0 2023-04-26 16:07:54,556 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6125, 1.3174, 1.5971, 1.9241, 1.7700, 1.5306, 1.6089, 1.6247], device='cuda:1'), covar=tensor([1.1639, 1.5034, 1.5179, 1.6886, 1.1990, 1.9275, 1.7875, 1.3581], device='cuda:1'), in_proj_covar=tensor([0.0425, 0.0462, 0.0550, 0.0568, 0.0455, 0.0480, 0.0491, 0.0492], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 16:08:00,116 INFO [finetune.py:976] (1/7) Epoch 5, batch 850, loss[loss=0.2582, simple_loss=0.3095, pruned_loss=0.1035, over 4882.00 frames. ], tot_loss[loss=0.2188, simple_loss=0.2779, pruned_loss=0.07982, over 939431.15 frames. ], batch size: 32, lr: 3.94e-03, grad_scale: 32.0 2023-04-26 16:08:02,025 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=23764.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 16:08:17,360 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=3.92 vs. limit=5.0 2023-04-26 16:08:39,278 INFO [finetune.py:976] (1/7) Epoch 5, batch 900, loss[loss=0.1993, simple_loss=0.2592, pruned_loss=0.06972, over 4816.00 frames. ], tot_loss[loss=0.2163, simple_loss=0.2754, pruned_loss=0.07858, over 945528.52 frames. ], batch size: 39, lr: 3.94e-03, grad_scale: 32.0 2023-04-26 16:08:40,385 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=23812.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 16:08:54,018 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.136e+02 1.680e+02 2.075e+02 2.492e+02 8.869e+02, threshold=4.150e+02, percent-clipped=5.0 2023-04-26 16:09:23,355 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=23844.0, num_to_drop=1, layers_to_drop={2} 2023-04-26 16:09:46,929 INFO [finetune.py:976] (1/7) Epoch 5, batch 950, loss[loss=0.1823, simple_loss=0.2509, pruned_loss=0.05683, over 4825.00 frames. ], tot_loss[loss=0.2156, simple_loss=0.2739, pruned_loss=0.07863, over 948028.31 frames. ], batch size: 51, lr: 3.94e-03, grad_scale: 32.0 2023-04-26 16:09:56,940 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=23873.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 16:10:14,839 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=23885.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 16:10:44,369 INFO [finetune.py:976] (1/7) Epoch 5, batch 1000, loss[loss=0.2353, simple_loss=0.3015, pruned_loss=0.08457, over 4934.00 frames. ], tot_loss[loss=0.2158, simple_loss=0.2744, pruned_loss=0.07865, over 948387.39 frames. ], batch size: 33, lr: 3.94e-03, grad_scale: 32.0 2023-04-26 16:10:45,105 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([0.8012, 1.7424, 1.6027, 1.3390, 1.9140, 1.3512, 2.4594, 1.3422], device='cuda:1'), covar=tensor([0.4308, 0.2008, 0.5670, 0.3957, 0.2105, 0.3164, 0.1517, 0.5240], device='cuda:1'), in_proj_covar=tensor([0.0353, 0.0358, 0.0441, 0.0374, 0.0402, 0.0386, 0.0401, 0.0423], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 16:10:52,030 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=23921.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 16:10:54,330 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.295e+02 1.932e+02 2.275e+02 2.774e+02 5.327e+02, threshold=4.551e+02, percent-clipped=3.0 2023-04-26 16:10:59,245 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=23933.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 16:11:17,666 INFO [finetune.py:976] (1/7) Epoch 5, batch 1050, loss[loss=0.2117, simple_loss=0.2749, pruned_loss=0.0742, over 4865.00 frames. ], tot_loss[loss=0.2171, simple_loss=0.2764, pruned_loss=0.07888, over 951341.42 frames. ], batch size: 31, lr: 3.94e-03, grad_scale: 32.0 2023-04-26 16:11:50,372 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8721, 2.2342, 0.9294, 1.1536, 1.6022, 1.1789, 2.9375, 1.4966], device='cuda:1'), covar=tensor([0.0679, 0.0637, 0.0811, 0.1321, 0.0550, 0.1046, 0.0300, 0.0731], device='cuda:1'), in_proj_covar=tensor([0.0055, 0.0071, 0.0053, 0.0050, 0.0054, 0.0055, 0.0083, 0.0053], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0008, 0.0008, 0.0008, 0.0008, 0.0011, 0.0008], device='cuda:1') 2023-04-26 16:12:31,172 INFO [finetune.py:976] (1/7) Epoch 5, batch 1100, loss[loss=0.2168, simple_loss=0.2745, pruned_loss=0.07959, over 4746.00 frames. ], tot_loss[loss=0.2182, simple_loss=0.2781, pruned_loss=0.07918, over 952217.52 frames. ], batch size: 27, lr: 3.94e-03, grad_scale: 32.0 2023-04-26 16:12:45,514 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.175e+02 1.787e+02 2.210e+02 2.620e+02 5.269e+02, threshold=4.419e+02, percent-clipped=3.0 2023-04-26 16:12:53,373 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=24037.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 16:13:02,936 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.1945, 1.4968, 1.2955, 1.6900, 1.5572, 1.8505, 1.3530, 3.2729], device='cuda:1'), covar=tensor([0.0675, 0.0738, 0.0815, 0.1174, 0.0654, 0.0539, 0.0764, 0.0178], device='cuda:1'), in_proj_covar=tensor([0.0040, 0.0040, 0.0041, 0.0046, 0.0041, 0.0041, 0.0040, 0.0062], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0017], device='cuda:1') 2023-04-26 16:13:06,364 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.5221, 2.4078, 2.5464, 2.9158, 2.7002, 2.3851, 1.9931, 2.5129], device='cuda:1'), covar=tensor([0.0787, 0.0914, 0.0546, 0.0509, 0.0598, 0.0856, 0.0898, 0.0542], device='cuda:1'), in_proj_covar=tensor([0.0207, 0.0210, 0.0186, 0.0182, 0.0183, 0.0199, 0.0170, 0.0194], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 16:13:08,675 INFO [finetune.py:976] (1/7) Epoch 5, batch 1150, loss[loss=0.206, simple_loss=0.2779, pruned_loss=0.06709, over 4887.00 frames. ], tot_loss[loss=0.2192, simple_loss=0.2791, pruned_loss=0.07961, over 953812.64 frames. ], batch size: 43, lr: 3.94e-03, grad_scale: 32.0 2023-04-26 16:13:09,381 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.3526, 3.0341, 0.9499, 1.4442, 2.0305, 1.4495, 4.0658, 2.0169], device='cuda:1'), covar=tensor([0.0627, 0.0691, 0.0899, 0.1304, 0.0565, 0.0971, 0.0199, 0.0620], device='cuda:1'), in_proj_covar=tensor([0.0055, 0.0071, 0.0053, 0.0050, 0.0054, 0.0055, 0.0083, 0.0053], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0008, 0.0008, 0.0008, 0.0008, 0.0011, 0.0008], device='cuda:1') 2023-04-26 16:13:33,484 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=24098.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 16:13:42,166 INFO [finetune.py:976] (1/7) Epoch 5, batch 1200, loss[loss=0.2694, simple_loss=0.318, pruned_loss=0.1104, over 4747.00 frames. ], tot_loss[loss=0.2188, simple_loss=0.2784, pruned_loss=0.07959, over 955027.54 frames. ], batch size: 54, lr: 3.94e-03, grad_scale: 32.0 2023-04-26 16:13:52,168 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.100e+02 1.779e+02 2.090e+02 2.399e+02 4.332e+02, threshold=4.179e+02, percent-clipped=0.0 2023-04-26 16:14:04,698 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=24144.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 16:14:05,406 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-04-26 16:14:15,799 INFO [finetune.py:976] (1/7) Epoch 5, batch 1250, loss[loss=0.2239, simple_loss=0.282, pruned_loss=0.08288, over 4191.00 frames. ], tot_loss[loss=0.2158, simple_loss=0.2749, pruned_loss=0.07837, over 953929.81 frames. ], batch size: 65, lr: 3.94e-03, grad_scale: 32.0 2023-04-26 16:14:42,784 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=24192.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 16:15:00,606 INFO [finetune.py:976] (1/7) Epoch 5, batch 1300, loss[loss=0.2184, simple_loss=0.279, pruned_loss=0.07886, over 4924.00 frames. ], tot_loss[loss=0.2122, simple_loss=0.2712, pruned_loss=0.07659, over 955495.95 frames. ], batch size: 36, lr: 3.94e-03, grad_scale: 32.0 2023-04-26 16:15:10,678 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.317e+02 1.846e+02 2.149e+02 2.713e+02 6.103e+02, threshold=4.299e+02, percent-clipped=1.0 2023-04-26 16:15:28,154 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5947, 1.2447, 1.5993, 1.9068, 1.7958, 1.5213, 1.5711, 1.5985], device='cuda:1'), covar=tensor([1.0418, 1.3634, 1.4904, 1.6132, 1.1615, 1.6805, 1.7679, 1.4333], device='cuda:1'), in_proj_covar=tensor([0.0423, 0.0461, 0.0547, 0.0566, 0.0454, 0.0477, 0.0489, 0.0491], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 16:15:41,359 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-04-26 16:15:49,986 INFO [finetune.py:976] (1/7) Epoch 5, batch 1350, loss[loss=0.2601, simple_loss=0.3308, pruned_loss=0.09464, over 4752.00 frames. ], tot_loss[loss=0.212, simple_loss=0.271, pruned_loss=0.0765, over 955550.52 frames. ], batch size: 54, lr: 3.94e-03, grad_scale: 32.0 2023-04-26 16:16:09,295 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.2084, 2.9373, 2.0962, 1.9680, 1.5932, 1.6326, 2.2886, 1.6391], device='cuda:1'), covar=tensor([0.1994, 0.1774, 0.1901, 0.2373, 0.2930, 0.2270, 0.1422, 0.2361], device='cuda:1'), in_proj_covar=tensor([0.0195, 0.0221, 0.0180, 0.0210, 0.0218, 0.0190, 0.0173, 0.0197], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-26 16:16:29,536 INFO [finetune.py:976] (1/7) Epoch 5, batch 1400, loss[loss=0.1939, simple_loss=0.2507, pruned_loss=0.06855, over 4771.00 frames. ], tot_loss[loss=0.2148, simple_loss=0.2742, pruned_loss=0.07773, over 956974.56 frames. ], batch size: 26, lr: 3.94e-03, grad_scale: 32.0 2023-04-26 16:16:38,503 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.222e+02 1.836e+02 2.129e+02 2.470e+02 6.262e+02, threshold=4.259e+02, percent-clipped=1.0 2023-04-26 16:17:19,153 INFO [finetune.py:976] (1/7) Epoch 5, batch 1450, loss[loss=0.1711, simple_loss=0.2488, pruned_loss=0.04668, over 4774.00 frames. ], tot_loss[loss=0.2168, simple_loss=0.277, pruned_loss=0.07832, over 957163.74 frames. ], batch size: 29, lr: 3.94e-03, grad_scale: 32.0 2023-04-26 16:17:40,630 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.17 vs. limit=5.0 2023-04-26 16:18:00,682 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=24393.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 16:18:10,370 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-04-26 16:18:12,078 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.1735, 1.2774, 1.3080, 1.5009, 1.4647, 1.1652, 0.8307, 1.3016], device='cuda:1'), covar=tensor([0.0950, 0.1216, 0.0846, 0.0573, 0.0719, 0.0972, 0.1117, 0.0674], device='cuda:1'), in_proj_covar=tensor([0.0205, 0.0208, 0.0184, 0.0181, 0.0182, 0.0196, 0.0168, 0.0193], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 16:18:16,237 INFO [finetune.py:976] (1/7) Epoch 5, batch 1500, loss[loss=0.2124, simple_loss=0.2702, pruned_loss=0.07725, over 4803.00 frames. ], tot_loss[loss=0.2187, simple_loss=0.2788, pruned_loss=0.07929, over 956420.93 frames. ], batch size: 25, lr: 3.94e-03, grad_scale: 32.0 2023-04-26 16:18:25,733 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.120e+02 1.847e+02 2.193e+02 2.526e+02 4.286e+02, threshold=4.386e+02, percent-clipped=1.0 2023-04-26 16:18:49,530 INFO [finetune.py:976] (1/7) Epoch 5, batch 1550, loss[loss=0.1731, simple_loss=0.2424, pruned_loss=0.05191, over 4901.00 frames. ], tot_loss[loss=0.2183, simple_loss=0.2785, pruned_loss=0.07903, over 955256.44 frames. ], batch size: 43, lr: 3.94e-03, grad_scale: 32.0 2023-04-26 16:19:22,930 INFO [finetune.py:976] (1/7) Epoch 5, batch 1600, loss[loss=0.228, simple_loss=0.2806, pruned_loss=0.08768, over 4902.00 frames. ], tot_loss[loss=0.2175, simple_loss=0.2771, pruned_loss=0.07895, over 954696.97 frames. ], batch size: 36, lr: 3.94e-03, grad_scale: 32.0 2023-04-26 16:19:32,011 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.098e+02 1.836e+02 2.156e+02 2.625e+02 3.904e+02, threshold=4.311e+02, percent-clipped=0.0 2023-04-26 16:19:37,769 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.66 vs. limit=5.0 2023-04-26 16:19:48,165 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=24547.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 16:19:55,521 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.5407, 1.8720, 2.2747, 2.9075, 2.1477, 1.7527, 1.5066, 2.1683], device='cuda:1'), covar=tensor([0.4306, 0.5127, 0.2207, 0.4268, 0.4730, 0.3944, 0.6494, 0.3946], device='cuda:1'), in_proj_covar=tensor([0.0273, 0.0260, 0.0220, 0.0330, 0.0221, 0.0229, 0.0248, 0.0196], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-26 16:19:56,589 INFO [finetune.py:976] (1/7) Epoch 5, batch 1650, loss[loss=0.2231, simple_loss=0.2823, pruned_loss=0.08194, over 4931.00 frames. ], tot_loss[loss=0.2146, simple_loss=0.2737, pruned_loss=0.07773, over 955268.25 frames. ], batch size: 33, lr: 3.94e-03, grad_scale: 64.0 2023-04-26 16:20:01,612 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=24569.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 16:20:28,442 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=24608.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 16:20:30,137 INFO [finetune.py:976] (1/7) Epoch 5, batch 1700, loss[loss=0.1742, simple_loss=0.23, pruned_loss=0.05923, over 4258.00 frames. ], tot_loss[loss=0.2129, simple_loss=0.2715, pruned_loss=0.07717, over 953786.60 frames. ], batch size: 18, lr: 3.94e-03, grad_scale: 64.0 2023-04-26 16:20:38,596 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.174e+02 1.758e+02 2.083e+02 2.406e+02 5.348e+02, threshold=4.165e+02, percent-clipped=1.0 2023-04-26 16:20:42,274 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=24630.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 16:21:03,218 INFO [finetune.py:976] (1/7) Epoch 5, batch 1750, loss[loss=0.2553, simple_loss=0.2802, pruned_loss=0.1152, over 4053.00 frames. ], tot_loss[loss=0.213, simple_loss=0.2718, pruned_loss=0.07713, over 952605.80 frames. ], batch size: 17, lr: 3.94e-03, grad_scale: 64.0 2023-04-26 16:21:30,489 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.3302, 1.5887, 1.3842, 1.9787, 1.7457, 2.0752, 1.4222, 4.2905], device='cuda:1'), covar=tensor([0.0679, 0.0797, 0.0848, 0.1221, 0.0689, 0.0645, 0.0787, 0.0107], device='cuda:1'), in_proj_covar=tensor([0.0040, 0.0040, 0.0041, 0.0046, 0.0041, 0.0041, 0.0040, 0.0062], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0017], device='cuda:1') 2023-04-26 16:21:35,435 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=24693.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 16:21:48,677 INFO [finetune.py:976] (1/7) Epoch 5, batch 1800, loss[loss=0.1705, simple_loss=0.2466, pruned_loss=0.0472, over 4929.00 frames. ], tot_loss[loss=0.2154, simple_loss=0.2754, pruned_loss=0.07772, over 955435.51 frames. ], batch size: 33, lr: 3.94e-03, grad_scale: 64.0 2023-04-26 16:21:57,279 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.363e+02 2.030e+02 2.431e+02 2.911e+02 5.485e+02, threshold=4.863e+02, percent-clipped=5.0 2023-04-26 16:22:08,100 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=24741.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 16:22:27,731 INFO [finetune.py:976] (1/7) Epoch 5, batch 1850, loss[loss=0.2423, simple_loss=0.308, pruned_loss=0.0883, over 4890.00 frames. ], tot_loss[loss=0.217, simple_loss=0.2772, pruned_loss=0.07834, over 955823.87 frames. ], batch size: 35, lr: 3.94e-03, grad_scale: 64.0 2023-04-26 16:22:27,980 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.32 vs. limit=5.0 2023-04-26 16:22:30,291 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=24765.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 16:23:29,630 INFO [finetune.py:976] (1/7) Epoch 5, batch 1900, loss[loss=0.197, simple_loss=0.2611, pruned_loss=0.06645, over 4816.00 frames. ], tot_loss[loss=0.2173, simple_loss=0.2778, pruned_loss=0.07835, over 954681.95 frames. ], batch size: 38, lr: 3.94e-03, grad_scale: 64.0 2023-04-26 16:23:44,192 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.223e+02 1.736e+02 2.141e+02 2.516e+02 4.890e+02, threshold=4.282e+02, percent-clipped=1.0 2023-04-26 16:23:50,604 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=24826.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 16:24:16,075 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=24848.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 16:24:31,091 INFO [finetune.py:976] (1/7) Epoch 5, batch 1950, loss[loss=0.2196, simple_loss=0.281, pruned_loss=0.07911, over 4816.00 frames. ], tot_loss[loss=0.2167, simple_loss=0.2773, pruned_loss=0.07807, over 957751.49 frames. ], batch size: 39, lr: 3.94e-03, grad_scale: 64.0 2023-04-26 16:24:46,900 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9701, 1.8279, 2.1771, 2.3663, 1.8200, 1.3953, 1.9960, 1.0889], device='cuda:1'), covar=tensor([0.0778, 0.0876, 0.0709, 0.1016, 0.1073, 0.1620, 0.0885, 0.1265], device='cuda:1'), in_proj_covar=tensor([0.0067, 0.0077, 0.0075, 0.0070, 0.0080, 0.0096, 0.0083, 0.0079], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-26 16:24:58,637 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=24903.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 16:25:03,319 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=24909.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 16:25:04,921 INFO [finetune.py:976] (1/7) Epoch 5, batch 2000, loss[loss=0.2136, simple_loss=0.2712, pruned_loss=0.07801, over 4818.00 frames. ], tot_loss[loss=0.214, simple_loss=0.274, pruned_loss=0.07695, over 958028.59 frames. ], batch size: 38, lr: 3.94e-03, grad_scale: 64.0 2023-04-26 16:25:13,836 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.128e+02 1.797e+02 2.078e+02 2.506e+02 3.857e+02, threshold=4.156e+02, percent-clipped=0.0 2023-04-26 16:25:13,920 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=24925.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 16:25:37,718 INFO [finetune.py:976] (1/7) Epoch 5, batch 2050, loss[loss=0.1667, simple_loss=0.2354, pruned_loss=0.04904, over 4765.00 frames. ], tot_loss[loss=0.2105, simple_loss=0.27, pruned_loss=0.07552, over 957054.92 frames. ], batch size: 28, lr: 3.94e-03, grad_scale: 64.0 2023-04-26 16:25:46,067 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6620, 1.6546, 0.9010, 1.3397, 1.9246, 1.5144, 1.4728, 1.4576], device='cuda:1'), covar=tensor([0.0535, 0.0407, 0.0397, 0.0589, 0.0282, 0.0554, 0.0523, 0.0621], device='cuda:1'), in_proj_covar=tensor([0.0031, 0.0027, 0.0025, 0.0031, 0.0022, 0.0030, 0.0030, 0.0031], device='cuda:1'), out_proj_covar=tensor([0.0049, 0.0045, 0.0038, 0.0049, 0.0038, 0.0048, 0.0047, 0.0050], device='cuda:1') 2023-04-26 16:25:55,907 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.83 vs. limit=5.0 2023-04-26 16:25:59,891 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=24995.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 16:26:10,615 INFO [finetune.py:976] (1/7) Epoch 5, batch 2100, loss[loss=0.2051, simple_loss=0.2694, pruned_loss=0.07041, over 4778.00 frames. ], tot_loss[loss=0.2104, simple_loss=0.2695, pruned_loss=0.07561, over 955728.99 frames. ], batch size: 27, lr: 3.94e-03, grad_scale: 64.0 2023-04-26 16:26:21,075 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.078e+02 1.851e+02 2.091e+02 2.608e+02 4.715e+02, threshold=4.183e+02, percent-clipped=2.0 2023-04-26 16:26:22,428 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.0580, 2.0315, 1.6981, 1.7718, 2.1659, 1.7180, 2.7113, 1.4943], device='cuda:1'), covar=tensor([0.4957, 0.2155, 0.6033, 0.3942, 0.2237, 0.3270, 0.1823, 0.5283], device='cuda:1'), in_proj_covar=tensor([0.0347, 0.0353, 0.0433, 0.0364, 0.0394, 0.0381, 0.0393, 0.0414], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 16:26:26,304 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.65 vs. limit=2.0 2023-04-26 16:26:40,915 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=25056.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 16:26:43,815 INFO [finetune.py:976] (1/7) Epoch 5, batch 2150, loss[loss=0.214, simple_loss=0.2813, pruned_loss=0.0733, over 4770.00 frames. ], tot_loss[loss=0.2117, simple_loss=0.2719, pruned_loss=0.07571, over 954800.03 frames. ], batch size: 28, lr: 3.94e-03, grad_scale: 64.0 2023-04-26 16:27:17,134 INFO [finetune.py:976] (1/7) Epoch 5, batch 2200, loss[loss=0.2917, simple_loss=0.342, pruned_loss=0.1207, over 4842.00 frames. ], tot_loss[loss=0.213, simple_loss=0.2742, pruned_loss=0.07592, over 956645.60 frames. ], batch size: 44, lr: 3.94e-03, grad_scale: 64.0 2023-04-26 16:27:19,089 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5976, 1.9541, 1.6683, 1.8324, 1.4644, 1.5327, 1.6178, 1.3178], device='cuda:1'), covar=tensor([0.2126, 0.1517, 0.1036, 0.1414, 0.3724, 0.1548, 0.2026, 0.2683], device='cuda:1'), in_proj_covar=tensor([0.0309, 0.0328, 0.0239, 0.0302, 0.0320, 0.0282, 0.0270, 0.0295], device='cuda:1'), out_proj_covar=tensor([1.2609e-04, 1.3309e-04, 9.7179e-05, 1.2130e-04, 1.3182e-04, 1.1394e-04, 1.1129e-04, 1.1921e-04], device='cuda:1') 2023-04-26 16:27:24,262 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=25121.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 16:27:27,077 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.188e+02 1.776e+02 2.171e+02 2.634e+02 4.750e+02, threshold=4.343e+02, percent-clipped=2.0 2023-04-26 16:27:50,267 INFO [finetune.py:976] (1/7) Epoch 5, batch 2250, loss[loss=0.1794, simple_loss=0.231, pruned_loss=0.06394, over 4436.00 frames. ], tot_loss[loss=0.2144, simple_loss=0.2757, pruned_loss=0.07651, over 957783.64 frames. ], batch size: 19, lr: 3.94e-03, grad_scale: 64.0 2023-04-26 16:28:06,196 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.44 vs. limit=5.0 2023-04-26 16:28:26,332 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=25203.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 16:28:26,892 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=25204.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 16:28:32,137 INFO [finetune.py:976] (1/7) Epoch 5, batch 2300, loss[loss=0.2387, simple_loss=0.2895, pruned_loss=0.09395, over 4784.00 frames. ], tot_loss[loss=0.2151, simple_loss=0.2769, pruned_loss=0.07669, over 958419.37 frames. ], batch size: 26, lr: 3.94e-03, grad_scale: 64.0 2023-04-26 16:28:52,192 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.169e+02 1.823e+02 2.210e+02 2.598e+02 6.851e+02, threshold=4.421e+02, percent-clipped=2.0 2023-04-26 16:28:52,294 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=25225.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 16:29:26,418 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=25251.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 16:29:38,266 INFO [finetune.py:976] (1/7) Epoch 5, batch 2350, loss[loss=0.2239, simple_loss=0.273, pruned_loss=0.08742, over 4800.00 frames. ], tot_loss[loss=0.215, simple_loss=0.2757, pruned_loss=0.07713, over 955397.99 frames. ], batch size: 45, lr: 3.94e-03, grad_scale: 64.0 2023-04-26 16:29:57,349 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=25273.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 16:30:39,277 INFO [finetune.py:976] (1/7) Epoch 5, batch 2400, loss[loss=0.1825, simple_loss=0.2414, pruned_loss=0.06182, over 4801.00 frames. ], tot_loss[loss=0.2117, simple_loss=0.272, pruned_loss=0.07574, over 955693.55 frames. ], batch size: 25, lr: 3.94e-03, grad_scale: 64.0 2023-04-26 16:30:48,373 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.145e+02 1.795e+02 2.098e+02 2.522e+02 5.509e+02, threshold=4.195e+02, percent-clipped=3.0 2023-04-26 16:31:12,374 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=25351.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 16:31:18,436 INFO [finetune.py:976] (1/7) Epoch 5, batch 2450, loss[loss=0.3761, simple_loss=0.3746, pruned_loss=0.1888, over 4207.00 frames. ], tot_loss[loss=0.2095, simple_loss=0.2694, pruned_loss=0.0748, over 956735.77 frames. ], batch size: 65, lr: 3.94e-03, grad_scale: 64.0 2023-04-26 16:32:08,174 INFO [finetune.py:976] (1/7) Epoch 5, batch 2500, loss[loss=0.1861, simple_loss=0.2635, pruned_loss=0.05438, over 4819.00 frames. ], tot_loss[loss=0.211, simple_loss=0.2708, pruned_loss=0.0756, over 955171.32 frames. ], batch size: 39, lr: 3.94e-03, grad_scale: 64.0 2023-04-26 16:32:08,809 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([3.7538, 3.5905, 2.6758, 4.3876, 3.7649, 3.7771, 1.5432, 3.8096], device='cuda:1'), covar=tensor([0.1567, 0.1255, 0.3166, 0.1590, 0.3974, 0.1698, 0.5566, 0.2132], device='cuda:1'), in_proj_covar=tensor([0.0249, 0.0223, 0.0259, 0.0314, 0.0306, 0.0258, 0.0276, 0.0278], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-26 16:32:15,244 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=25421.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 16:32:17,547 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.133e+02 1.862e+02 2.343e+02 2.826e+02 5.817e+02, threshold=4.685e+02, percent-clipped=1.0 2023-04-26 16:32:19,547 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=25428.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 16:32:34,782 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9640, 2.4147, 2.0596, 2.2002, 1.6017, 1.9948, 2.2608, 1.6845], device='cuda:1'), covar=tensor([0.1829, 0.0935, 0.0881, 0.1163, 0.3165, 0.1223, 0.1578, 0.2491], device='cuda:1'), in_proj_covar=tensor([0.0308, 0.0326, 0.0237, 0.0299, 0.0319, 0.0281, 0.0270, 0.0294], device='cuda:1'), out_proj_covar=tensor([1.2576e-04, 1.3249e-04, 9.6716e-05, 1.2040e-04, 1.3126e-04, 1.1368e-04, 1.1112e-04, 1.1875e-04], device='cuda:1') 2023-04-26 16:32:41,938 INFO [finetune.py:976] (1/7) Epoch 5, batch 2550, loss[loss=0.1963, simple_loss=0.2707, pruned_loss=0.06092, over 4834.00 frames. ], tot_loss[loss=0.2147, simple_loss=0.2752, pruned_loss=0.07711, over 955848.49 frames. ], batch size: 49, lr: 3.94e-03, grad_scale: 32.0 2023-04-26 16:32:47,380 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=25469.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 16:33:01,463 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=25489.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 16:33:11,434 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=25504.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 16:33:14,482 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5743, 1.2247, 4.4694, 4.1777, 3.9568, 4.1800, 4.1423, 3.9377], device='cuda:1'), covar=tensor([0.7270, 0.6371, 0.1006, 0.1772, 0.1128, 0.1319, 0.1537, 0.1691], device='cuda:1'), in_proj_covar=tensor([0.0315, 0.0306, 0.0418, 0.0424, 0.0358, 0.0411, 0.0322, 0.0379], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-26 16:33:15,644 INFO [finetune.py:976] (1/7) Epoch 5, batch 2600, loss[loss=0.2292, simple_loss=0.3001, pruned_loss=0.0791, over 4820.00 frames. ], tot_loss[loss=0.2154, simple_loss=0.2764, pruned_loss=0.07723, over 953589.49 frames. ], batch size: 33, lr: 3.94e-03, grad_scale: 32.0 2023-04-26 16:33:25,300 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.303e+02 1.830e+02 2.137e+02 2.517e+02 4.934e+02, threshold=4.274e+02, percent-clipped=1.0 2023-04-26 16:33:29,574 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5046, 2.4336, 2.1268, 2.3131, 2.5788, 2.1228, 3.5384, 2.0099], device='cuda:1'), covar=tensor([0.4608, 0.2436, 0.4764, 0.3889, 0.2138, 0.3135, 0.1824, 0.4284], device='cuda:1'), in_proj_covar=tensor([0.0348, 0.0354, 0.0432, 0.0366, 0.0395, 0.0381, 0.0394, 0.0416], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 16:33:43,927 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=25552.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 16:33:49,355 INFO [finetune.py:976] (1/7) Epoch 5, batch 2650, loss[loss=0.2358, simple_loss=0.2919, pruned_loss=0.08981, over 4838.00 frames. ], tot_loss[loss=0.2157, simple_loss=0.277, pruned_loss=0.07723, over 953644.43 frames. ], batch size: 44, lr: 3.94e-03, grad_scale: 32.0 2023-04-26 16:34:15,326 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([3.6740, 3.5693, 2.7142, 4.2725, 3.7041, 3.6883, 1.6645, 3.6050], device='cuda:1'), covar=tensor([0.1645, 0.1318, 0.2855, 0.1692, 0.1978, 0.1904, 0.5243, 0.2150], device='cuda:1'), in_proj_covar=tensor([0.0247, 0.0223, 0.0257, 0.0310, 0.0304, 0.0256, 0.0273, 0.0276], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-26 16:34:17,876 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5862, 1.1911, 1.5997, 1.9792, 1.7445, 1.5534, 1.5829, 1.6611], device='cuda:1'), covar=tensor([1.1167, 1.5292, 1.5578, 1.6495, 1.2852, 1.7953, 1.7720, 1.3324], device='cuda:1'), in_proj_covar=tensor([0.0423, 0.0460, 0.0547, 0.0567, 0.0454, 0.0478, 0.0488, 0.0490], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 16:34:28,770 INFO [finetune.py:976] (1/7) Epoch 5, batch 2700, loss[loss=0.2023, simple_loss=0.2601, pruned_loss=0.07219, over 4813.00 frames. ], tot_loss[loss=0.2151, simple_loss=0.2766, pruned_loss=0.07683, over 953094.69 frames. ], batch size: 39, lr: 3.94e-03, grad_scale: 32.0 2023-04-26 16:34:48,614 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.304e+02 1.757e+02 2.118e+02 2.619e+02 4.731e+02, threshold=4.237e+02, percent-clipped=2.0 2023-04-26 16:35:20,248 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=25651.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 16:35:37,124 INFO [finetune.py:976] (1/7) Epoch 5, batch 2750, loss[loss=0.2279, simple_loss=0.2787, pruned_loss=0.08854, over 4833.00 frames. ], tot_loss[loss=0.2141, simple_loss=0.2746, pruned_loss=0.07685, over 954493.97 frames. ], batch size: 47, lr: 3.94e-03, grad_scale: 32.0 2023-04-26 16:36:11,160 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9538, 2.3439, 1.9280, 2.2184, 1.6900, 1.7967, 2.1376, 1.7163], device='cuda:1'), covar=tensor([0.2244, 0.1472, 0.1215, 0.1475, 0.3567, 0.1747, 0.1699, 0.3014], device='cuda:1'), in_proj_covar=tensor([0.0310, 0.0329, 0.0239, 0.0301, 0.0320, 0.0283, 0.0271, 0.0294], device='cuda:1'), out_proj_covar=tensor([1.2641e-04, 1.3354e-04, 9.7301e-05, 1.2094e-04, 1.3180e-04, 1.1447e-04, 1.1141e-04, 1.1899e-04], device='cuda:1') 2023-04-26 16:36:24,083 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=25699.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 16:36:32,332 INFO [finetune.py:976] (1/7) Epoch 5, batch 2800, loss[loss=0.1891, simple_loss=0.2498, pruned_loss=0.06415, over 4887.00 frames. ], tot_loss[loss=0.2102, simple_loss=0.2706, pruned_loss=0.07488, over 956505.79 frames. ], batch size: 32, lr: 3.93e-03, grad_scale: 32.0 2023-04-26 16:36:33,259 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.49 vs. limit=2.0 2023-04-26 16:36:47,384 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.154e+02 1.821e+02 2.084e+02 2.532e+02 5.696e+02, threshold=4.167e+02, percent-clipped=4.0 2023-04-26 16:36:50,598 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.1360, 2.1477, 1.7531, 1.8446, 2.2062, 1.7855, 2.7975, 1.4760], device='cuda:1'), covar=tensor([0.4452, 0.1929, 0.4915, 0.3755, 0.2113, 0.2931, 0.1692, 0.4903], device='cuda:1'), in_proj_covar=tensor([0.0350, 0.0354, 0.0432, 0.0367, 0.0396, 0.0383, 0.0395, 0.0418], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 16:36:59,484 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9508, 1.8571, 2.1337, 2.3842, 2.3639, 1.9478, 1.6470, 2.0784], device='cuda:1'), covar=tensor([0.1105, 0.1156, 0.0578, 0.0691, 0.0657, 0.0988, 0.1022, 0.0624], device='cuda:1'), in_proj_covar=tensor([0.0207, 0.0209, 0.0185, 0.0182, 0.0182, 0.0198, 0.0169, 0.0194], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 16:37:13,252 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-04-26 16:37:32,910 INFO [finetune.py:976] (1/7) Epoch 5, batch 2850, loss[loss=0.2095, simple_loss=0.2751, pruned_loss=0.07188, over 4900.00 frames. ], tot_loss[loss=0.2089, simple_loss=0.2687, pruned_loss=0.07452, over 957334.54 frames. ], batch size: 32, lr: 3.93e-03, grad_scale: 32.0 2023-04-26 16:37:46,997 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=25784.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 16:37:59,993 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.3005, 1.4443, 1.5766, 1.7539, 1.6839, 1.8290, 1.6366, 1.7030], device='cuda:1'), covar=tensor([0.6970, 1.1517, 0.8949, 0.8780, 0.9868, 1.5057, 1.1232, 1.0225], device='cuda:1'), in_proj_covar=tensor([0.0314, 0.0396, 0.0316, 0.0325, 0.0348, 0.0412, 0.0379, 0.0335], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-26 16:38:06,695 INFO [finetune.py:976] (1/7) Epoch 5, batch 2900, loss[loss=0.2209, simple_loss=0.2806, pruned_loss=0.08058, over 4247.00 frames. ], tot_loss[loss=0.2133, simple_loss=0.2727, pruned_loss=0.07691, over 955448.41 frames. ], batch size: 65, lr: 3.93e-03, grad_scale: 32.0 2023-04-26 16:38:15,790 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.320e+02 1.753e+02 2.187e+02 2.584e+02 6.163e+02, threshold=4.374e+02, percent-clipped=2.0 2023-04-26 16:38:28,514 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5475, 2.1443, 1.6041, 1.2681, 1.1714, 1.2379, 1.5816, 1.1034], device='cuda:1'), covar=tensor([0.1943, 0.1620, 0.1928, 0.2444, 0.3033, 0.2429, 0.1385, 0.2559], device='cuda:1'), in_proj_covar=tensor([0.0195, 0.0222, 0.0179, 0.0210, 0.0216, 0.0189, 0.0172, 0.0196], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-26 16:38:37,777 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=25858.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 16:38:37,817 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0704, 1.4920, 1.9683, 2.4440, 1.9151, 1.4694, 1.2427, 1.6872], device='cuda:1'), covar=tensor([0.4074, 0.4476, 0.2006, 0.3212, 0.3742, 0.3639, 0.5377, 0.3303], device='cuda:1'), in_proj_covar=tensor([0.0273, 0.0259, 0.0219, 0.0328, 0.0219, 0.0228, 0.0244, 0.0195], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-26 16:38:39,949 INFO [finetune.py:976] (1/7) Epoch 5, batch 2950, loss[loss=0.2554, simple_loss=0.3136, pruned_loss=0.09856, over 4806.00 frames. ], tot_loss[loss=0.216, simple_loss=0.2758, pruned_loss=0.07807, over 956253.37 frames. ], batch size: 41, lr: 3.93e-03, grad_scale: 32.0 2023-04-26 16:38:41,768 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0416, 1.2175, 1.3798, 1.5606, 1.5127, 1.6189, 1.4516, 1.4858], device='cuda:1'), covar=tensor([0.7348, 1.0191, 0.8269, 0.8154, 0.9681, 1.4597, 1.0477, 0.9328], device='cuda:1'), in_proj_covar=tensor([0.0313, 0.0396, 0.0316, 0.0326, 0.0348, 0.0412, 0.0379, 0.0335], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-26 16:39:12,517 INFO [finetune.py:976] (1/7) Epoch 5, batch 3000, loss[loss=0.2591, simple_loss=0.318, pruned_loss=0.1001, over 4815.00 frames. ], tot_loss[loss=0.2171, simple_loss=0.2774, pruned_loss=0.07842, over 956217.62 frames. ], batch size: 38, lr: 3.93e-03, grad_scale: 32.0 2023-04-26 16:39:12,517 INFO [finetune.py:1001] (1/7) Computing validation loss 2023-04-26 16:39:15,418 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.3103, 1.2989, 3.8193, 3.4511, 3.4790, 3.6976, 3.7996, 3.3278], device='cuda:1'), covar=tensor([0.7371, 0.5922, 0.1230, 0.2324, 0.1308, 0.1267, 0.0884, 0.1813], device='cuda:1'), in_proj_covar=tensor([0.0315, 0.0307, 0.0417, 0.0425, 0.0359, 0.0411, 0.0324, 0.0379], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-26 16:39:29,056 INFO [finetune.py:1010] (1/7) Epoch 5, validation: loss=0.1595, simple_loss=0.233, pruned_loss=0.04303, over 2265189.00 frames. 2023-04-26 16:39:29,056 INFO [finetune.py:1011] (1/7) Maximum memory allocated so far is 6391MB 2023-04-26 16:39:41,244 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=25919.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 16:39:51,733 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.259e+02 1.795e+02 2.151e+02 2.692e+02 4.010e+02, threshold=4.303e+02, percent-clipped=0.0 2023-04-26 16:40:35,901 INFO [finetune.py:976] (1/7) Epoch 5, batch 3050, loss[loss=0.2053, simple_loss=0.269, pruned_loss=0.0708, over 4927.00 frames. ], tot_loss[loss=0.2171, simple_loss=0.2783, pruned_loss=0.07793, over 956607.79 frames. ], batch size: 41, lr: 3.93e-03, grad_scale: 32.0 2023-04-26 16:40:45,134 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5172, 1.3164, 1.8901, 1.8491, 1.4203, 1.1309, 1.6374, 1.0303], device='cuda:1'), covar=tensor([0.0896, 0.1112, 0.0613, 0.0957, 0.1186, 0.1650, 0.0928, 0.1238], device='cuda:1'), in_proj_covar=tensor([0.0068, 0.0077, 0.0076, 0.0070, 0.0080, 0.0096, 0.0083, 0.0079], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-26 16:41:32,849 INFO [finetune.py:976] (1/7) Epoch 5, batch 3100, loss[loss=0.2137, simple_loss=0.2703, pruned_loss=0.07853, over 4739.00 frames. ], tot_loss[loss=0.2148, simple_loss=0.2759, pruned_loss=0.07684, over 956086.03 frames. ], batch size: 54, lr: 3.93e-03, grad_scale: 32.0 2023-04-26 16:41:43,947 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.102e+02 1.748e+02 1.976e+02 2.307e+02 6.292e+02, threshold=3.952e+02, percent-clipped=1.0 2023-04-26 16:42:06,252 INFO [finetune.py:976] (1/7) Epoch 5, batch 3150, loss[loss=0.2504, simple_loss=0.2913, pruned_loss=0.1047, over 4890.00 frames. ], tot_loss[loss=0.2122, simple_loss=0.2728, pruned_loss=0.07582, over 957456.71 frames. ], batch size: 35, lr: 3.93e-03, grad_scale: 32.0 2023-04-26 16:42:27,115 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=26084.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 16:42:46,458 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=26102.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 16:42:54,362 INFO [finetune.py:976] (1/7) Epoch 5, batch 3200, loss[loss=0.2097, simple_loss=0.2665, pruned_loss=0.07642, over 4823.00 frames. ], tot_loss[loss=0.2093, simple_loss=0.2694, pruned_loss=0.07463, over 956060.74 frames. ], batch size: 39, lr: 3.93e-03, grad_scale: 32.0 2023-04-26 16:43:09,264 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.183e+02 1.750e+02 2.081e+02 2.590e+02 4.901e+02, threshold=4.161e+02, percent-clipped=2.0 2023-04-26 16:43:14,418 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=26132.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 16:43:14,498 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.3313, 1.4190, 1.5543, 1.6499, 1.6259, 1.7875, 1.6567, 1.6519], device='cuda:1'), covar=tensor([0.6348, 0.9074, 0.8013, 0.7874, 0.9378, 1.2933, 0.9517, 0.8694], device='cuda:1'), in_proj_covar=tensor([0.0314, 0.0395, 0.0316, 0.0326, 0.0347, 0.0411, 0.0379, 0.0335], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-26 16:43:24,893 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6378, 1.4855, 0.8416, 1.3206, 1.4337, 1.4893, 1.4247, 1.4023], device='cuda:1'), covar=tensor([0.0535, 0.0421, 0.0444, 0.0590, 0.0329, 0.0559, 0.0542, 0.0644], device='cuda:1'), in_proj_covar=tensor([0.0031, 0.0027, 0.0024, 0.0031, 0.0022, 0.0030, 0.0030, 0.0031], device='cuda:1'), out_proj_covar=tensor([0.0049, 0.0045, 0.0038, 0.0049, 0.0038, 0.0048, 0.0047, 0.0050], device='cuda:1') 2023-04-26 16:43:31,947 INFO [finetune.py:976] (1/7) Epoch 5, batch 3250, loss[loss=0.2022, simple_loss=0.2713, pruned_loss=0.06652, over 4873.00 frames. ], tot_loss[loss=0.2096, simple_loss=0.2694, pruned_loss=0.07493, over 955050.09 frames. ], batch size: 34, lr: 3.93e-03, grad_scale: 16.0 2023-04-26 16:43:33,283 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=26163.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 16:43:36,827 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=26168.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 16:44:05,489 INFO [finetune.py:976] (1/7) Epoch 5, batch 3300, loss[loss=0.2288, simple_loss=0.2875, pruned_loss=0.08502, over 4865.00 frames. ], tot_loss[loss=0.2129, simple_loss=0.2734, pruned_loss=0.07622, over 953804.01 frames. ], batch size: 44, lr: 3.93e-03, grad_scale: 16.0 2023-04-26 16:44:07,401 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=26214.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 16:44:07,445 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4459, 2.3576, 1.8358, 2.1666, 2.4099, 2.0339, 3.1852, 1.7081], device='cuda:1'), covar=tensor([0.4342, 0.2023, 0.5233, 0.3329, 0.2063, 0.2797, 0.1592, 0.4358], device='cuda:1'), in_proj_covar=tensor([0.0353, 0.0357, 0.0436, 0.0370, 0.0399, 0.0386, 0.0396, 0.0421], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 16:44:16,103 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.343e+02 1.793e+02 2.255e+02 2.677e+02 5.857e+02, threshold=4.510e+02, percent-clipped=2.0 2023-04-26 16:44:17,432 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=26229.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 16:44:32,043 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=26250.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 16:44:38,661 INFO [finetune.py:976] (1/7) Epoch 5, batch 3350, loss[loss=0.1881, simple_loss=0.2654, pruned_loss=0.05541, over 4852.00 frames. ], tot_loss[loss=0.2139, simple_loss=0.2746, pruned_loss=0.07662, over 952677.61 frames. ], batch size: 44, lr: 3.93e-03, grad_scale: 16.0 2023-04-26 16:45:31,123 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.71 vs. limit=2.0 2023-04-26 16:45:39,804 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-04-26 16:45:50,830 INFO [finetune.py:976] (1/7) Epoch 5, batch 3400, loss[loss=0.2455, simple_loss=0.3071, pruned_loss=0.09198, over 4859.00 frames. ], tot_loss[loss=0.2144, simple_loss=0.2755, pruned_loss=0.07667, over 952435.52 frames. ], batch size: 44, lr: 3.93e-03, grad_scale: 16.0 2023-04-26 16:45:50,976 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=26311.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 16:46:13,228 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.088e+02 1.738e+02 2.082e+02 2.437e+02 3.720e+02, threshold=4.164e+02, percent-clipped=0.0 2023-04-26 16:46:56,944 INFO [finetune.py:976] (1/7) Epoch 5, batch 3450, loss[loss=0.2075, simple_loss=0.2702, pruned_loss=0.07242, over 4841.00 frames. ], tot_loss[loss=0.215, simple_loss=0.2763, pruned_loss=0.07681, over 953208.27 frames. ], batch size: 49, lr: 3.93e-03, grad_scale: 16.0 2023-04-26 16:47:07,943 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9956, 1.9616, 2.2368, 2.4656, 2.2654, 1.9725, 1.6196, 1.9818], device='cuda:1'), covar=tensor([0.0931, 0.1013, 0.0567, 0.0502, 0.0610, 0.0845, 0.0914, 0.0597], device='cuda:1'), in_proj_covar=tensor([0.0208, 0.0210, 0.0186, 0.0182, 0.0182, 0.0198, 0.0169, 0.0193], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 16:47:07,956 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([0.9520, 2.0377, 1.8429, 1.7076, 2.0917, 1.7154, 2.7169, 1.5648], device='cuda:1'), covar=tensor([0.4341, 0.1980, 0.5621, 0.3479, 0.1837, 0.2762, 0.1443, 0.4950], device='cuda:1'), in_proj_covar=tensor([0.0353, 0.0358, 0.0439, 0.0371, 0.0401, 0.0387, 0.0397, 0.0423], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 16:47:18,089 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-04-26 16:47:52,438 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.4850, 2.9350, 0.8640, 1.5048, 2.2471, 1.4680, 4.2672, 2.0248], device='cuda:1'), covar=tensor([0.0635, 0.0831, 0.0987, 0.1309, 0.0613, 0.1026, 0.0209, 0.0661], device='cuda:1'), in_proj_covar=tensor([0.0054, 0.0071, 0.0053, 0.0049, 0.0054, 0.0055, 0.0083, 0.0053], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0008, 0.0008, 0.0008, 0.0008, 0.0011, 0.0008], device='cuda:1') 2023-04-26 16:48:00,839 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0924, 1.8064, 2.3611, 2.4017, 1.8792, 1.4699, 2.1430, 1.1753], device='cuda:1'), covar=tensor([0.0767, 0.1190, 0.0659, 0.1317, 0.1211, 0.1631, 0.0986, 0.1317], device='cuda:1'), in_proj_covar=tensor([0.0068, 0.0077, 0.0076, 0.0069, 0.0080, 0.0097, 0.0083, 0.0078], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-26 16:48:03,178 INFO [finetune.py:976] (1/7) Epoch 5, batch 3500, loss[loss=0.1592, simple_loss=0.2227, pruned_loss=0.0478, over 4752.00 frames. ], tot_loss[loss=0.2127, simple_loss=0.2738, pruned_loss=0.07582, over 953739.41 frames. ], batch size: 27, lr: 3.93e-03, grad_scale: 16.0 2023-04-26 16:48:05,091 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=26414.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 16:48:12,728 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4753, 2.4319, 1.9515, 2.2531, 2.4637, 2.0701, 3.3476, 1.7375], device='cuda:1'), covar=tensor([0.4366, 0.2296, 0.5723, 0.3822, 0.2336, 0.2951, 0.1988, 0.4612], device='cuda:1'), in_proj_covar=tensor([0.0354, 0.0358, 0.0440, 0.0371, 0.0401, 0.0387, 0.0398, 0.0424], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 16:48:25,047 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.307e+02 1.713e+02 2.095e+02 2.561e+02 6.592e+02, threshold=4.191e+02, percent-clipped=2.0 2023-04-26 16:49:08,571 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=26458.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 16:49:10,354 INFO [finetune.py:976] (1/7) Epoch 5, batch 3550, loss[loss=0.1917, simple_loss=0.2511, pruned_loss=0.0662, over 4891.00 frames. ], tot_loss[loss=0.2102, simple_loss=0.2708, pruned_loss=0.07478, over 954490.87 frames. ], batch size: 35, lr: 3.93e-03, grad_scale: 16.0 2023-04-26 16:49:24,626 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=26475.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 16:49:49,836 INFO [finetune.py:976] (1/7) Epoch 5, batch 3600, loss[loss=0.2543, simple_loss=0.3041, pruned_loss=0.1023, over 4820.00 frames. ], tot_loss[loss=0.2084, simple_loss=0.2686, pruned_loss=0.07412, over 955663.31 frames. ], batch size: 38, lr: 3.93e-03, grad_scale: 16.0 2023-04-26 16:49:51,741 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=26514.0, num_to_drop=1, layers_to_drop={2} 2023-04-26 16:49:57,836 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=26524.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 16:49:59,591 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.194e+02 1.718e+02 2.043e+02 2.547e+02 4.174e+02, threshold=4.086e+02, percent-clipped=0.0 2023-04-26 16:50:23,342 INFO [finetune.py:976] (1/7) Epoch 5, batch 3650, loss[loss=0.2294, simple_loss=0.2926, pruned_loss=0.08312, over 4899.00 frames. ], tot_loss[loss=0.2108, simple_loss=0.2707, pruned_loss=0.07541, over 954137.88 frames. ], batch size: 35, lr: 3.93e-03, grad_scale: 16.0 2023-04-26 16:50:24,021 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=26562.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 16:50:53,154 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=26606.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 16:50:57,092 INFO [finetune.py:976] (1/7) Epoch 5, batch 3700, loss[loss=0.2286, simple_loss=0.28, pruned_loss=0.08865, over 4823.00 frames. ], tot_loss[loss=0.2155, simple_loss=0.2763, pruned_loss=0.0774, over 955640.52 frames. ], batch size: 33, lr: 3.93e-03, grad_scale: 16.0 2023-04-26 16:51:05,689 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9150, 1.8029, 2.0342, 2.3114, 2.2649, 1.8290, 1.4263, 1.8792], device='cuda:1'), covar=tensor([0.1002, 0.1102, 0.0618, 0.0615, 0.0640, 0.0983, 0.1006, 0.0703], device='cuda:1'), in_proj_covar=tensor([0.0205, 0.0207, 0.0183, 0.0179, 0.0179, 0.0195, 0.0166, 0.0191], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 16:51:06,779 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.079e+02 1.827e+02 2.264e+02 2.659e+02 6.887e+02, threshold=4.529e+02, percent-clipped=2.0 2023-04-26 16:51:08,755 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.2500, 1.6589, 1.5876, 1.9159, 1.7290, 2.1825, 1.5115, 3.4432], device='cuda:1'), covar=tensor([0.0684, 0.0740, 0.0809, 0.1184, 0.0646, 0.0482, 0.0758, 0.0166], device='cuda:1'), in_proj_covar=tensor([0.0040, 0.0040, 0.0041, 0.0046, 0.0041, 0.0041, 0.0040, 0.0062], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0017], device='cuda:1') 2023-04-26 16:51:29,840 INFO [finetune.py:976] (1/7) Epoch 5, batch 3750, loss[loss=0.283, simple_loss=0.3309, pruned_loss=0.1175, over 4814.00 frames. ], tot_loss[loss=0.218, simple_loss=0.2785, pruned_loss=0.0787, over 955286.19 frames. ], batch size: 39, lr: 3.93e-03, grad_scale: 16.0 2023-04-26 16:51:32,791 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-04-26 16:52:20,490 INFO [finetune.py:976] (1/7) Epoch 5, batch 3800, loss[loss=0.2496, simple_loss=0.2941, pruned_loss=0.1025, over 4884.00 frames. ], tot_loss[loss=0.219, simple_loss=0.2796, pruned_loss=0.0792, over 957101.68 frames. ], batch size: 32, lr: 3.93e-03, grad_scale: 16.0 2023-04-26 16:52:31,668 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.252e+02 1.856e+02 2.180e+02 2.581e+02 5.516e+02, threshold=4.361e+02, percent-clipped=1.0 2023-04-26 16:52:52,119 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=26758.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 16:52:54,307 INFO [finetune.py:976] (1/7) Epoch 5, batch 3850, loss[loss=0.2706, simple_loss=0.3185, pruned_loss=0.1113, over 4131.00 frames. ], tot_loss[loss=0.2173, simple_loss=0.2784, pruned_loss=0.07815, over 955302.73 frames. ], batch size: 65, lr: 3.93e-03, grad_scale: 16.0 2023-04-26 16:53:00,827 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=26770.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 16:53:01,457 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4641, 3.9834, 0.8223, 2.0412, 2.1891, 2.7520, 2.2939, 0.9113], device='cuda:1'), covar=tensor([0.1418, 0.0893, 0.2139, 0.1377, 0.1055, 0.1032, 0.1436, 0.2198], device='cuda:1'), in_proj_covar=tensor([0.0120, 0.0258, 0.0145, 0.0127, 0.0136, 0.0158, 0.0123, 0.0126], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-26 16:53:40,968 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=26806.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 16:53:44,435 INFO [finetune.py:976] (1/7) Epoch 5, batch 3900, loss[loss=0.1992, simple_loss=0.2568, pruned_loss=0.07082, over 4822.00 frames. ], tot_loss[loss=0.2148, simple_loss=0.275, pruned_loss=0.07731, over 952704.65 frames. ], batch size: 40, lr: 3.93e-03, grad_scale: 16.0 2023-04-26 16:54:03,745 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=26824.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 16:54:05,468 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.093e+02 1.758e+02 2.108e+02 2.612e+02 6.373e+02, threshold=4.215e+02, percent-clipped=3.0 2023-04-26 16:54:50,051 INFO [finetune.py:976] (1/7) Epoch 5, batch 3950, loss[loss=0.1756, simple_loss=0.2399, pruned_loss=0.05566, over 4775.00 frames. ], tot_loss[loss=0.2098, simple_loss=0.2698, pruned_loss=0.07488, over 954940.49 frames. ], batch size: 28, lr: 3.93e-03, grad_scale: 16.0 2023-04-26 16:55:08,485 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=26872.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 16:55:47,036 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=26904.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 16:55:48,204 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=26906.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 16:55:51,739 INFO [finetune.py:976] (1/7) Epoch 5, batch 4000, loss[loss=0.2371, simple_loss=0.2998, pruned_loss=0.08723, over 4922.00 frames. ], tot_loss[loss=0.211, simple_loss=0.2703, pruned_loss=0.07585, over 954985.32 frames. ], batch size: 43, lr: 3.93e-03, grad_scale: 16.0 2023-04-26 16:56:00,117 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.1904, 1.4882, 1.3299, 1.8355, 1.6183, 2.1020, 1.3404, 3.4006], device='cuda:1'), covar=tensor([0.0745, 0.0819, 0.0938, 0.1222, 0.0715, 0.0511, 0.0816, 0.0163], device='cuda:1'), in_proj_covar=tensor([0.0039, 0.0040, 0.0041, 0.0045, 0.0041, 0.0041, 0.0039, 0.0061], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0017], device='cuda:1') 2023-04-26 16:56:02,925 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.248e+02 1.675e+02 2.014e+02 2.454e+02 6.687e+02, threshold=4.029e+02, percent-clipped=1.0 2023-04-26 16:56:17,490 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=26950.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 16:56:19,903 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=26954.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 16:56:25,064 INFO [finetune.py:976] (1/7) Epoch 5, batch 4050, loss[loss=0.1737, simple_loss=0.24, pruned_loss=0.05367, over 4724.00 frames. ], tot_loss[loss=0.2124, simple_loss=0.2722, pruned_loss=0.07629, over 954137.89 frames. ], batch size: 23, lr: 3.93e-03, grad_scale: 16.0 2023-04-26 16:56:28,124 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=26965.0, num_to_drop=1, layers_to_drop={2} 2023-04-26 16:56:34,137 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-04-26 16:56:58,981 INFO [finetune.py:976] (1/7) Epoch 5, batch 4100, loss[loss=0.1895, simple_loss=0.2575, pruned_loss=0.06071, over 4780.00 frames. ], tot_loss[loss=0.2166, simple_loss=0.2771, pruned_loss=0.07804, over 956715.91 frames. ], batch size: 26, lr: 3.93e-03, grad_scale: 16.0 2023-04-26 16:56:59,090 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=27011.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 16:57:20,558 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.204e+02 1.859e+02 2.211e+02 2.694e+02 5.784e+02, threshold=4.421e+02, percent-clipped=3.0 2023-04-26 16:57:31,889 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5923, 1.2330, 4.3903, 4.0999, 3.8738, 4.1088, 3.9880, 3.8106], device='cuda:1'), covar=tensor([0.7063, 0.6059, 0.1025, 0.1671, 0.1048, 0.1517, 0.1517, 0.1632], device='cuda:1'), in_proj_covar=tensor([0.0313, 0.0306, 0.0415, 0.0419, 0.0353, 0.0405, 0.0320, 0.0376], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-26 16:57:54,743 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.8872, 2.8023, 2.2586, 3.2854, 2.8533, 2.8971, 1.1039, 2.7642], device='cuda:1'), covar=tensor([0.2237, 0.1713, 0.3120, 0.2738, 0.3242, 0.2296, 0.5680, 0.2694], device='cuda:1'), in_proj_covar=tensor([0.0247, 0.0220, 0.0254, 0.0311, 0.0304, 0.0254, 0.0275, 0.0277], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-26 16:58:04,488 INFO [finetune.py:976] (1/7) Epoch 5, batch 4150, loss[loss=0.2406, simple_loss=0.3045, pruned_loss=0.08837, over 4803.00 frames. ], tot_loss[loss=0.2192, simple_loss=0.2796, pruned_loss=0.07938, over 956357.02 frames. ], batch size: 51, lr: 3.93e-03, grad_scale: 16.0 2023-04-26 16:58:13,005 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.1720, 3.2353, 1.0778, 1.6176, 1.5930, 2.1595, 1.8491, 1.0069], device='cuda:1'), covar=tensor([0.1899, 0.1333, 0.2289, 0.1857, 0.1453, 0.1370, 0.1735, 0.2038], device='cuda:1'), in_proj_covar=tensor([0.0120, 0.0258, 0.0144, 0.0127, 0.0137, 0.0158, 0.0123, 0.0126], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-26 16:58:16,103 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=27070.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 16:58:26,751 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8422, 2.3269, 1.9035, 2.1931, 1.8097, 1.9849, 2.0460, 1.6227], device='cuda:1'), covar=tensor([0.2353, 0.1527, 0.1071, 0.1566, 0.3230, 0.1448, 0.2204, 0.3296], device='cuda:1'), in_proj_covar=tensor([0.0309, 0.0328, 0.0237, 0.0302, 0.0320, 0.0282, 0.0270, 0.0293], device='cuda:1'), out_proj_covar=tensor([1.2597e-04, 1.3316e-04, 9.6328e-05, 1.2130e-04, 1.3177e-04, 1.1388e-04, 1.1111e-04, 1.1849e-04], device='cuda:1') 2023-04-26 16:58:33,248 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-04-26 16:58:35,691 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-04-26 16:58:38,680 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.1009, 1.4595, 1.9219, 2.4667, 1.8672, 1.4147, 1.2358, 1.7261], device='cuda:1'), covar=tensor([0.4285, 0.4699, 0.2225, 0.3297, 0.3909, 0.3777, 0.5688, 0.3614], device='cuda:1'), in_proj_covar=tensor([0.0277, 0.0260, 0.0222, 0.0330, 0.0220, 0.0230, 0.0245, 0.0196], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-26 16:58:47,719 INFO [finetune.py:976] (1/7) Epoch 5, batch 4200, loss[loss=0.2113, simple_loss=0.2829, pruned_loss=0.06982, over 4894.00 frames. ], tot_loss[loss=0.2178, simple_loss=0.2789, pruned_loss=0.0783, over 956771.88 frames. ], batch size: 37, lr: 3.93e-03, grad_scale: 16.0 2023-04-26 16:58:47,816 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9590, 2.4477, 0.9498, 1.2780, 1.7508, 1.2114, 2.9352, 1.4757], device='cuda:1'), covar=tensor([0.0701, 0.0612, 0.0769, 0.1209, 0.0485, 0.0957, 0.0284, 0.0671], device='cuda:1'), in_proj_covar=tensor([0.0054, 0.0070, 0.0052, 0.0049, 0.0054, 0.0055, 0.0083, 0.0053], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0011, 0.0008], device='cuda:1') 2023-04-26 16:58:50,178 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=3.76 vs. limit=5.0 2023-04-26 16:58:53,014 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=27118.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 16:58:54,272 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9310, 1.8255, 2.0797, 2.2481, 2.2834, 1.7645, 1.5074, 1.8478], device='cuda:1'), covar=tensor([0.0858, 0.1042, 0.0611, 0.0584, 0.0539, 0.0960, 0.0820, 0.0667], device='cuda:1'), in_proj_covar=tensor([0.0201, 0.0205, 0.0180, 0.0176, 0.0175, 0.0193, 0.0164, 0.0188], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 16:58:58,900 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.181e+01 1.721e+02 2.038e+02 2.503e+02 4.432e+02, threshold=4.077e+02, percent-clipped=1.0 2023-04-26 16:59:04,183 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.3982, 2.2098, 2.6185, 2.7942, 2.7075, 2.1739, 1.9472, 2.2093], device='cuda:1'), covar=tensor([0.0966, 0.1066, 0.0569, 0.0703, 0.0597, 0.0996, 0.0942, 0.0845], device='cuda:1'), in_proj_covar=tensor([0.0201, 0.0205, 0.0180, 0.0176, 0.0175, 0.0193, 0.0164, 0.0188], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 16:59:13,618 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=27148.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 16:59:17,366 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-04-26 16:59:21,417 INFO [finetune.py:976] (1/7) Epoch 5, batch 4250, loss[loss=0.2503, simple_loss=0.2988, pruned_loss=0.1009, over 4816.00 frames. ], tot_loss[loss=0.2149, simple_loss=0.2759, pruned_loss=0.07692, over 956162.38 frames. ], batch size: 40, lr: 3.93e-03, grad_scale: 16.0 2023-04-26 16:59:36,260 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8623, 1.4087, 1.3730, 1.5986, 2.1096, 1.7221, 1.4808, 1.3565], device='cuda:1'), covar=tensor([0.1524, 0.1720, 0.2216, 0.1703, 0.1006, 0.1886, 0.2126, 0.2021], device='cuda:1'), in_proj_covar=tensor([0.0308, 0.0329, 0.0346, 0.0303, 0.0338, 0.0336, 0.0304, 0.0346], device='cuda:1'), out_proj_covar=tensor([6.6137e-05, 7.0405e-05, 7.4915e-05, 6.3343e-05, 7.1643e-05, 7.2900e-05, 6.5929e-05, 7.4481e-05], device='cuda:1') 2023-04-26 16:59:54,270 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=27209.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 16:59:55,345 INFO [finetune.py:976] (1/7) Epoch 5, batch 4300, loss[loss=0.2004, simple_loss=0.2616, pruned_loss=0.06954, over 4824.00 frames. ], tot_loss[loss=0.213, simple_loss=0.273, pruned_loss=0.07649, over 956095.39 frames. ], batch size: 40, lr: 3.93e-03, grad_scale: 16.0 2023-04-26 17:00:17,880 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.057e+02 1.798e+02 2.267e+02 2.763e+02 5.468e+02, threshold=4.535e+02, percent-clipped=5.0 2023-04-26 17:00:25,991 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.2515, 1.1921, 1.3099, 1.5540, 1.6467, 1.2101, 0.8296, 1.3404], device='cuda:1'), covar=tensor([0.0933, 0.1619, 0.0972, 0.0634, 0.0613, 0.1017, 0.1044, 0.0754], device='cuda:1'), in_proj_covar=tensor([0.0201, 0.0205, 0.0179, 0.0176, 0.0175, 0.0193, 0.0164, 0.0188], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 17:00:52,958 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=5.06 vs. limit=5.0 2023-04-26 17:01:04,149 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=27260.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 17:01:04,678 INFO [finetune.py:976] (1/7) Epoch 5, batch 4350, loss[loss=0.2471, simple_loss=0.3051, pruned_loss=0.09456, over 4843.00 frames. ], tot_loss[loss=0.2096, simple_loss=0.2694, pruned_loss=0.07488, over 956006.75 frames. ], batch size: 49, lr: 3.93e-03, grad_scale: 16.0 2023-04-26 17:01:30,287 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.2950, 1.5755, 1.5257, 1.7266, 1.5727, 1.7344, 1.6762, 1.5989], device='cuda:1'), covar=tensor([0.6816, 1.0737, 0.8796, 0.8106, 0.9458, 1.3662, 1.0289, 0.9888], device='cuda:1'), in_proj_covar=tensor([0.0314, 0.0396, 0.0317, 0.0326, 0.0347, 0.0411, 0.0378, 0.0335], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-26 17:01:46,638 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=27306.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 17:01:49,555 INFO [finetune.py:976] (1/7) Epoch 5, batch 4400, loss[loss=0.2044, simple_loss=0.2842, pruned_loss=0.06232, over 4903.00 frames. ], tot_loss[loss=0.2091, simple_loss=0.2692, pruned_loss=0.07445, over 952919.94 frames. ], batch size: 36, lr: 3.93e-03, grad_scale: 16.0 2023-04-26 17:02:00,139 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.042e+02 1.703e+02 2.157e+02 2.468e+02 4.465e+02, threshold=4.314e+02, percent-clipped=0.0 2023-04-26 17:02:23,093 INFO [finetune.py:976] (1/7) Epoch 5, batch 4450, loss[loss=0.2473, simple_loss=0.3159, pruned_loss=0.08929, over 4920.00 frames. ], tot_loss[loss=0.2137, simple_loss=0.2743, pruned_loss=0.07658, over 954988.85 frames. ], batch size: 38, lr: 3.93e-03, grad_scale: 16.0 2023-04-26 17:02:56,797 INFO [finetune.py:976] (1/7) Epoch 5, batch 4500, loss[loss=0.1581, simple_loss=0.2247, pruned_loss=0.04578, over 4777.00 frames. ], tot_loss[loss=0.2143, simple_loss=0.2754, pruned_loss=0.07663, over 954903.20 frames. ], batch size: 26, lr: 3.93e-03, grad_scale: 16.0 2023-04-26 17:03:04,651 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4409, 2.3805, 1.9729, 2.1237, 2.4529, 2.0159, 3.4795, 1.8402], device='cuda:1'), covar=tensor([0.4874, 0.2469, 0.5763, 0.4102, 0.2587, 0.3222, 0.1969, 0.5018], device='cuda:1'), in_proj_covar=tensor([0.0352, 0.0354, 0.0437, 0.0367, 0.0396, 0.0384, 0.0392, 0.0420], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 17:03:17,441 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.259e+02 1.778e+02 2.173e+02 2.607e+02 4.915e+02, threshold=4.346e+02, percent-clipped=1.0 2023-04-26 17:03:57,772 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.2649, 3.3027, 0.9405, 1.7912, 1.8006, 2.4909, 1.8796, 0.9798], device='cuda:1'), covar=tensor([0.1532, 0.0905, 0.2117, 0.1337, 0.1117, 0.0995, 0.1471, 0.2136], device='cuda:1'), in_proj_covar=tensor([0.0121, 0.0260, 0.0146, 0.0128, 0.0138, 0.0160, 0.0124, 0.0127], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-26 17:04:02,851 INFO [finetune.py:976] (1/7) Epoch 5, batch 4550, loss[loss=0.1934, simple_loss=0.2542, pruned_loss=0.06633, over 4921.00 frames. ], tot_loss[loss=0.2143, simple_loss=0.2757, pruned_loss=0.07642, over 955216.90 frames. ], batch size: 33, lr: 3.93e-03, grad_scale: 16.0 2023-04-26 17:04:15,133 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=5.02 vs. limit=5.0 2023-04-26 17:04:31,477 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=27504.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 17:04:36,154 INFO [finetune.py:976] (1/7) Epoch 5, batch 4600, loss[loss=0.1748, simple_loss=0.2457, pruned_loss=0.05198, over 4868.00 frames. ], tot_loss[loss=0.2137, simple_loss=0.2754, pruned_loss=0.07602, over 956618.95 frames. ], batch size: 31, lr: 3.93e-03, grad_scale: 16.0 2023-04-26 17:04:46,279 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.146e+02 1.803e+02 2.105e+02 2.477e+02 5.679e+02, threshold=4.210e+02, percent-clipped=3.0 2023-04-26 17:05:08,949 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=27560.0, num_to_drop=1, layers_to_drop={2} 2023-04-26 17:05:09,447 INFO [finetune.py:976] (1/7) Epoch 5, batch 4650, loss[loss=0.229, simple_loss=0.2813, pruned_loss=0.08831, over 4873.00 frames. ], tot_loss[loss=0.2114, simple_loss=0.2723, pruned_loss=0.07525, over 955572.03 frames. ], batch size: 34, lr: 3.93e-03, grad_scale: 16.0 2023-04-26 17:05:28,059 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8937, 2.4981, 2.0279, 2.3791, 1.8721, 2.0994, 2.1837, 1.6863], device='cuda:1'), covar=tensor([0.2468, 0.1378, 0.1089, 0.1526, 0.3062, 0.1484, 0.2125, 0.3109], device='cuda:1'), in_proj_covar=tensor([0.0308, 0.0326, 0.0236, 0.0301, 0.0320, 0.0281, 0.0269, 0.0293], device='cuda:1'), out_proj_covar=tensor([1.2529e-04, 1.3253e-04, 9.6172e-05, 1.2085e-04, 1.3152e-04, 1.1350e-04, 1.1056e-04, 1.1849e-04], device='cuda:1') 2023-04-26 17:05:40,079 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=27606.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 17:05:41,305 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=27608.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 17:05:43,535 INFO [finetune.py:976] (1/7) Epoch 5, batch 4700, loss[loss=0.1494, simple_loss=0.2211, pruned_loss=0.03886, over 4905.00 frames. ], tot_loss[loss=0.2077, simple_loss=0.2682, pruned_loss=0.07358, over 954755.69 frames. ], batch size: 35, lr: 3.93e-03, grad_scale: 16.0 2023-04-26 17:05:54,155 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.191e+02 1.680e+02 2.058e+02 2.502e+02 5.893e+02, threshold=4.117e+02, percent-clipped=4.0 2023-04-26 17:06:23,333 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=27654.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 17:06:34,207 INFO [finetune.py:976] (1/7) Epoch 5, batch 4750, loss[loss=0.2529, simple_loss=0.2932, pruned_loss=0.1063, over 4781.00 frames. ], tot_loss[loss=0.206, simple_loss=0.2662, pruned_loss=0.07291, over 957263.73 frames. ], batch size: 27, lr: 3.93e-03, grad_scale: 16.0 2023-04-26 17:07:05,753 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9724, 2.2977, 0.9394, 1.3230, 1.7012, 1.1776, 2.9433, 1.5592], device='cuda:1'), covar=tensor([0.0693, 0.0680, 0.0796, 0.1174, 0.0518, 0.0997, 0.0273, 0.0641], device='cuda:1'), in_proj_covar=tensor([0.0054, 0.0070, 0.0052, 0.0049, 0.0054, 0.0055, 0.0082, 0.0053], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0008, 0.0008, 0.0008, 0.0008, 0.0011, 0.0008], device='cuda:1') 2023-04-26 17:07:40,900 INFO [finetune.py:976] (1/7) Epoch 5, batch 4800, loss[loss=0.2574, simple_loss=0.3182, pruned_loss=0.09827, over 4914.00 frames. ], tot_loss[loss=0.2101, simple_loss=0.2704, pruned_loss=0.07493, over 957727.79 frames. ], batch size: 42, lr: 3.93e-03, grad_scale: 16.0 2023-04-26 17:07:48,582 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=27714.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 17:08:02,774 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.360e+02 1.880e+02 2.179e+02 2.503e+02 4.628e+02, threshold=4.358e+02, percent-clipped=2.0 2023-04-26 17:08:08,454 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4723, 1.6006, 0.6029, 1.1766, 1.5306, 1.3499, 1.2217, 1.3068], device='cuda:1'), covar=tensor([0.0574, 0.0415, 0.0474, 0.0632, 0.0335, 0.0577, 0.0570, 0.0656], device='cuda:1'), in_proj_covar=tensor([0.0031, 0.0026, 0.0024, 0.0030, 0.0021, 0.0030, 0.0029, 0.0031], device='cuda:1'), out_proj_covar=tensor([0.0049, 0.0044, 0.0038, 0.0049, 0.0038, 0.0048, 0.0047, 0.0050], device='cuda:1') 2023-04-26 17:08:41,284 INFO [finetune.py:976] (1/7) Epoch 5, batch 4850, loss[loss=0.2303, simple_loss=0.2861, pruned_loss=0.08724, over 4821.00 frames. ], tot_loss[loss=0.2146, simple_loss=0.2752, pruned_loss=0.07699, over 955192.00 frames. ], batch size: 38, lr: 3.92e-03, grad_scale: 16.0 2023-04-26 17:08:50,930 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=27775.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 17:09:12,827 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-04-26 17:09:14,393 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=27804.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 17:09:18,625 INFO [finetune.py:976] (1/7) Epoch 5, batch 4900, loss[loss=0.2101, simple_loss=0.2587, pruned_loss=0.08074, over 4248.00 frames. ], tot_loss[loss=0.215, simple_loss=0.2753, pruned_loss=0.07738, over 951815.08 frames. ], batch size: 18, lr: 3.92e-03, grad_scale: 16.0 2023-04-26 17:09:23,931 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.3814, 3.4972, 0.9504, 1.7978, 1.9144, 2.4060, 1.9817, 0.9834], device='cuda:1'), covar=tensor([0.1566, 0.0865, 0.2035, 0.1403, 0.1180, 0.1137, 0.1520, 0.2161], device='cuda:1'), in_proj_covar=tensor([0.0122, 0.0262, 0.0146, 0.0129, 0.0138, 0.0161, 0.0124, 0.0128], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-26 17:09:30,230 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.253e+02 1.856e+02 2.309e+02 2.750e+02 8.138e+02, threshold=4.618e+02, percent-clipped=6.0 2023-04-26 17:09:42,448 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.8156, 2.7580, 2.3227, 3.2363, 2.8423, 2.8000, 1.2412, 2.7531], device='cuda:1'), covar=tensor([0.2078, 0.1854, 0.3723, 0.3227, 0.3282, 0.2330, 0.5496, 0.2986], device='cuda:1'), in_proj_covar=tensor([0.0248, 0.0223, 0.0256, 0.0315, 0.0307, 0.0258, 0.0278, 0.0278], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-26 17:09:46,104 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=27852.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 17:09:52,541 INFO [finetune.py:976] (1/7) Epoch 5, batch 4950, loss[loss=0.2155, simple_loss=0.2744, pruned_loss=0.07834, over 4787.00 frames. ], tot_loss[loss=0.2155, simple_loss=0.2767, pruned_loss=0.07717, over 953435.64 frames. ], batch size: 29, lr: 3.92e-03, grad_scale: 16.0 2023-04-26 17:10:31,676 INFO [finetune.py:976] (1/7) Epoch 5, batch 5000, loss[loss=0.2104, simple_loss=0.2751, pruned_loss=0.07288, over 4790.00 frames. ], tot_loss[loss=0.2143, simple_loss=0.2751, pruned_loss=0.07676, over 953396.77 frames. ], batch size: 51, lr: 3.92e-03, grad_scale: 16.0 2023-04-26 17:10:41,863 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.1321, 1.8989, 2.1778, 2.4980, 2.3656, 1.9683, 1.6336, 2.1005], device='cuda:1'), covar=tensor([0.0908, 0.1119, 0.0661, 0.0642, 0.0650, 0.1000, 0.0928, 0.0677], device='cuda:1'), in_proj_covar=tensor([0.0204, 0.0207, 0.0182, 0.0179, 0.0179, 0.0195, 0.0166, 0.0190], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 17:10:42,356 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.076e+02 1.656e+02 1.985e+02 2.433e+02 4.076e+02, threshold=3.970e+02, percent-clipped=0.0 2023-04-26 17:10:57,905 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6302, 1.5995, 1.7969, 1.9919, 1.9757, 1.5624, 1.2203, 1.7105], device='cuda:1'), covar=tensor([0.0831, 0.1251, 0.0720, 0.0554, 0.0592, 0.0993, 0.0955, 0.0641], device='cuda:1'), in_proj_covar=tensor([0.0204, 0.0207, 0.0182, 0.0179, 0.0179, 0.0195, 0.0166, 0.0190], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 17:11:04,344 INFO [finetune.py:976] (1/7) Epoch 5, batch 5050, loss[loss=0.2063, simple_loss=0.2604, pruned_loss=0.07608, over 4760.00 frames. ], tot_loss[loss=0.2121, simple_loss=0.2721, pruned_loss=0.07602, over 954021.92 frames. ], batch size: 28, lr: 3.92e-03, grad_scale: 16.0 2023-04-26 17:12:05,263 INFO [finetune.py:976] (1/7) Epoch 5, batch 5100, loss[loss=0.2264, simple_loss=0.2728, pruned_loss=0.09005, over 4856.00 frames. ], tot_loss[loss=0.2092, simple_loss=0.2686, pruned_loss=0.07493, over 953730.78 frames. ], batch size: 44, lr: 3.92e-03, grad_scale: 16.0 2023-04-26 17:12:27,655 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.101e+02 1.777e+02 2.079e+02 2.348e+02 6.496e+02, threshold=4.158e+02, percent-clipped=5.0 2023-04-26 17:13:08,115 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.3868, 1.2748, 3.6726, 3.3850, 3.2726, 3.4470, 3.4857, 3.2821], device='cuda:1'), covar=tensor([0.6882, 0.5881, 0.1125, 0.1825, 0.1262, 0.2102, 0.2386, 0.1474], device='cuda:1'), in_proj_covar=tensor([0.0315, 0.0308, 0.0418, 0.0421, 0.0356, 0.0411, 0.0320, 0.0376], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-26 17:13:10,170 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 2023-04-26 17:13:12,917 INFO [finetune.py:976] (1/7) Epoch 5, batch 5150, loss[loss=0.2207, simple_loss=0.2742, pruned_loss=0.08361, over 4883.00 frames. ], tot_loss[loss=0.2101, simple_loss=0.2694, pruned_loss=0.07546, over 956243.33 frames. ], batch size: 32, lr: 3.92e-03, grad_scale: 16.0 2023-04-26 17:13:31,479 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=28070.0, num_to_drop=1, layers_to_drop={2} 2023-04-26 17:14:09,042 INFO [finetune.py:976] (1/7) Epoch 5, batch 5200, loss[loss=0.2373, simple_loss=0.3063, pruned_loss=0.0841, over 4931.00 frames. ], tot_loss[loss=0.214, simple_loss=0.2738, pruned_loss=0.07715, over 955136.47 frames. ], batch size: 33, lr: 3.92e-03, grad_scale: 16.0 2023-04-26 17:14:15,819 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-04-26 17:14:19,886 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.281e+02 1.900e+02 2.150e+02 2.501e+02 5.102e+02, threshold=4.301e+02, percent-clipped=1.0 2023-04-26 17:14:42,335 INFO [finetune.py:976] (1/7) Epoch 5, batch 5250, loss[loss=0.181, simple_loss=0.2538, pruned_loss=0.05407, over 4861.00 frames. ], tot_loss[loss=0.2157, simple_loss=0.2763, pruned_loss=0.07753, over 954184.41 frames. ], batch size: 34, lr: 3.92e-03, grad_scale: 32.0 2023-04-26 17:14:56,224 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0666, 2.3738, 1.1009, 1.3241, 1.8794, 1.1417, 2.9433, 1.5716], device='cuda:1'), covar=tensor([0.0617, 0.0662, 0.0751, 0.1189, 0.0468, 0.0988, 0.0270, 0.0651], device='cuda:1'), in_proj_covar=tensor([0.0054, 0.0070, 0.0052, 0.0049, 0.0054, 0.0054, 0.0082, 0.0053], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0011, 0.0008], device='cuda:1') 2023-04-26 17:14:59,603 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.2528, 2.1333, 2.4628, 2.7096, 2.6700, 2.1544, 1.8654, 2.2851], device='cuda:1'), covar=tensor([0.1061, 0.1094, 0.0608, 0.0642, 0.0684, 0.1128, 0.1003, 0.0656], device='cuda:1'), in_proj_covar=tensor([0.0204, 0.0207, 0.0183, 0.0179, 0.0179, 0.0195, 0.0166, 0.0190], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 17:15:09,318 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8460, 1.4121, 1.3227, 1.5791, 2.0799, 1.6545, 1.3348, 1.2664], device='cuda:1'), covar=tensor([0.1604, 0.1578, 0.2148, 0.1408, 0.0902, 0.1627, 0.2326, 0.2002], device='cuda:1'), in_proj_covar=tensor([0.0313, 0.0333, 0.0351, 0.0306, 0.0343, 0.0337, 0.0309, 0.0352], device='cuda:1'), out_proj_covar=tensor([6.7151e-05, 7.1373e-05, 7.5979e-05, 6.3902e-05, 7.2567e-05, 7.3232e-05, 6.6956e-05, 7.5957e-05], device='cuda:1') 2023-04-26 17:15:14,087 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.97 vs. limit=2.0 2023-04-26 17:15:16,323 INFO [finetune.py:976] (1/7) Epoch 5, batch 5300, loss[loss=0.1833, simple_loss=0.2488, pruned_loss=0.05889, over 4859.00 frames. ], tot_loss[loss=0.2169, simple_loss=0.2772, pruned_loss=0.07826, over 953905.98 frames. ], batch size: 31, lr: 3.92e-03, grad_scale: 32.0 2023-04-26 17:15:27,004 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.419e+02 1.781e+02 2.085e+02 2.383e+02 5.799e+02, threshold=4.171e+02, percent-clipped=2.0 2023-04-26 17:15:28,323 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=28229.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 17:15:46,212 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=28255.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 17:15:49,729 INFO [finetune.py:976] (1/7) Epoch 5, batch 5350, loss[loss=0.2102, simple_loss=0.2722, pruned_loss=0.07406, over 4868.00 frames. ], tot_loss[loss=0.2162, simple_loss=0.2772, pruned_loss=0.07758, over 954083.75 frames. ], batch size: 34, lr: 3.92e-03, grad_scale: 32.0 2023-04-26 17:16:07,251 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=28287.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 17:16:09,996 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=28290.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 17:16:14,084 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([0.6103, 1.6985, 1.6788, 1.2671, 1.7855, 1.4032, 2.2743, 1.4146], device='cuda:1'), covar=tensor([0.4094, 0.1865, 0.5160, 0.3191, 0.1733, 0.2591, 0.1676, 0.5120], device='cuda:1'), in_proj_covar=tensor([0.0354, 0.0357, 0.0442, 0.0371, 0.0399, 0.0389, 0.0395, 0.0426], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 17:16:23,679 INFO [finetune.py:976] (1/7) Epoch 5, batch 5400, loss[loss=0.2371, simple_loss=0.2932, pruned_loss=0.09046, over 4804.00 frames. ], tot_loss[loss=0.2125, simple_loss=0.2734, pruned_loss=0.0758, over 952563.08 frames. ], batch size: 45, lr: 3.92e-03, grad_scale: 32.0 2023-04-26 17:16:27,323 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=28316.0, num_to_drop=1, layers_to_drop={2} 2023-04-26 17:16:32,209 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4177, 0.6506, 1.2025, 1.8045, 1.5833, 1.3506, 1.2829, 1.3414], device='cuda:1'), covar=tensor([0.9914, 1.2442, 1.3127, 1.4633, 1.1124, 1.4566, 1.5301, 1.3039], device='cuda:1'), in_proj_covar=tensor([0.0421, 0.0454, 0.0538, 0.0559, 0.0450, 0.0472, 0.0486, 0.0485], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 17:16:34,348 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.191e+02 1.689e+02 2.005e+02 2.420e+02 4.760e+02, threshold=4.009e+02, percent-clipped=1.0 2023-04-26 17:16:35,091 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.1541, 1.6433, 1.3670, 1.8631, 1.6127, 2.2612, 1.3664, 3.6370], device='cuda:1'), covar=tensor([0.0678, 0.0736, 0.0792, 0.1121, 0.0626, 0.0462, 0.0721, 0.0141], device='cuda:1'), in_proj_covar=tensor([0.0039, 0.0040, 0.0041, 0.0045, 0.0041, 0.0040, 0.0040, 0.0061], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0017], device='cuda:1') 2023-04-26 17:16:49,208 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=28348.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 17:16:57,452 INFO [finetune.py:976] (1/7) Epoch 5, batch 5450, loss[loss=0.2191, simple_loss=0.2549, pruned_loss=0.09165, over 3965.00 frames. ], tot_loss[loss=0.2116, simple_loss=0.2715, pruned_loss=0.07586, over 953778.81 frames. ], batch size: 17, lr: 3.92e-03, grad_scale: 32.0 2023-04-26 17:17:03,566 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.3086, 3.0579, 0.8834, 1.6748, 1.9578, 2.1340, 1.8711, 0.9754], device='cuda:1'), covar=tensor([0.1470, 0.1232, 0.2113, 0.1435, 0.1050, 0.1195, 0.1483, 0.1875], device='cuda:1'), in_proj_covar=tensor([0.0124, 0.0266, 0.0148, 0.0130, 0.0141, 0.0163, 0.0126, 0.0129], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-26 17:17:03,577 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=28370.0, num_to_drop=1, layers_to_drop={2} 2023-04-26 17:17:33,946 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.8058, 2.7814, 2.2307, 3.2231, 2.8662, 2.7658, 1.1836, 2.7737], device='cuda:1'), covar=tensor([0.2308, 0.1855, 0.3885, 0.3330, 0.4187, 0.2550, 0.5482, 0.3012], device='cuda:1'), in_proj_covar=tensor([0.0248, 0.0222, 0.0259, 0.0316, 0.0307, 0.0258, 0.0279, 0.0279], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-26 17:17:37,439 INFO [finetune.py:976] (1/7) Epoch 5, batch 5500, loss[loss=0.2361, simple_loss=0.2903, pruned_loss=0.09098, over 4816.00 frames. ], tot_loss[loss=0.2078, simple_loss=0.268, pruned_loss=0.07378, over 955902.72 frames. ], batch size: 38, lr: 3.92e-03, grad_scale: 32.0 2023-04-26 17:17:47,035 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=28418.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 17:17:59,135 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.103e+02 1.711e+02 2.116e+02 2.529e+02 5.388e+02, threshold=4.231e+02, percent-clipped=4.0 2023-04-26 17:18:19,172 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.3249, 1.3216, 3.7762, 3.4946, 3.3928, 3.6084, 3.6359, 3.3489], device='cuda:1'), covar=tensor([0.6994, 0.5562, 0.1124, 0.1867, 0.1114, 0.1624, 0.1582, 0.1483], device='cuda:1'), in_proj_covar=tensor([0.0317, 0.0311, 0.0421, 0.0425, 0.0360, 0.0415, 0.0323, 0.0381], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-26 17:18:49,581 INFO [finetune.py:976] (1/7) Epoch 5, batch 5550, loss[loss=0.2299, simple_loss=0.2944, pruned_loss=0.08272, over 4901.00 frames. ], tot_loss[loss=0.2097, simple_loss=0.2701, pruned_loss=0.0747, over 956489.87 frames. ], batch size: 43, lr: 3.92e-03, grad_scale: 32.0 2023-04-26 17:19:09,812 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=28486.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 17:19:11,606 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.60 vs. limit=2.0 2023-04-26 17:19:14,430 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=28492.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 17:19:15,692 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5737, 2.2537, 1.5698, 1.4228, 1.2019, 1.1976, 1.6493, 1.1507], device='cuda:1'), covar=tensor([0.2020, 0.1721, 0.2007, 0.2461, 0.2967, 0.2353, 0.1483, 0.2508], device='cuda:1'), in_proj_covar=tensor([0.0193, 0.0219, 0.0177, 0.0207, 0.0212, 0.0186, 0.0169, 0.0193], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-26 17:19:25,620 INFO [finetune.py:976] (1/7) Epoch 5, batch 5600, loss[loss=0.1968, simple_loss=0.2637, pruned_loss=0.06491, over 4855.00 frames. ], tot_loss[loss=0.2133, simple_loss=0.2744, pruned_loss=0.07611, over 954413.91 frames. ], batch size: 49, lr: 3.92e-03, grad_scale: 32.0 2023-04-26 17:19:34,908 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.156e+02 1.757e+02 2.198e+02 2.682e+02 4.335e+02, threshold=4.395e+02, percent-clipped=1.0 2023-04-26 17:19:47,086 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=28547.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 17:19:51,014 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=28553.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 17:19:55,612 INFO [finetune.py:976] (1/7) Epoch 5, batch 5650, loss[loss=0.2011, simple_loss=0.2734, pruned_loss=0.06437, over 4813.00 frames. ], tot_loss[loss=0.2131, simple_loss=0.275, pruned_loss=0.0756, over 954330.75 frames. ], batch size: 45, lr: 3.92e-03, grad_scale: 32.0 2023-04-26 17:20:09,697 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=28585.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 17:20:21,105 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([0.6784, 1.7031, 1.8231, 1.2599, 1.7882, 1.4860, 2.2641, 1.5515], device='cuda:1'), covar=tensor([0.3497, 0.1513, 0.4077, 0.2421, 0.1500, 0.2172, 0.1289, 0.4116], device='cuda:1'), in_proj_covar=tensor([0.0354, 0.0358, 0.0441, 0.0372, 0.0400, 0.0388, 0.0394, 0.0425], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 17:20:24,601 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([3.2988, 3.2282, 2.4718, 3.7865, 3.3038, 3.3594, 1.4840, 3.2629], device='cuda:1'), covar=tensor([0.1734, 0.1447, 0.2973, 0.2159, 0.3144, 0.1877, 0.5534, 0.2287], device='cuda:1'), in_proj_covar=tensor([0.0245, 0.0219, 0.0255, 0.0311, 0.0302, 0.0254, 0.0274, 0.0275], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-26 17:20:25,136 INFO [finetune.py:976] (1/7) Epoch 5, batch 5700, loss[loss=0.184, simple_loss=0.2369, pruned_loss=0.06559, over 4206.00 frames. ], tot_loss[loss=0.2112, simple_loss=0.2721, pruned_loss=0.07512, over 938429.18 frames. ], batch size: 18, lr: 3.92e-03, grad_scale: 32.0 2023-04-26 17:20:25,191 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=28611.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 17:20:34,600 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.220e+02 1.668e+02 2.021e+02 2.528e+02 3.624e+02, threshold=4.042e+02, percent-clipped=0.0 2023-04-26 17:20:58,935 INFO [finetune.py:976] (1/7) Epoch 6, batch 0, loss[loss=0.2294, simple_loss=0.2843, pruned_loss=0.08728, over 4879.00 frames. ], tot_loss[loss=0.2294, simple_loss=0.2843, pruned_loss=0.08728, over 4879.00 frames. ], batch size: 43, lr: 3.92e-03, grad_scale: 32.0 2023-04-26 17:20:58,935 INFO [finetune.py:1001] (1/7) Computing validation loss 2023-04-26 17:21:08,073 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.3358, 1.2391, 1.5201, 1.4440, 1.2645, 1.0789, 1.3001, 0.9245], device='cuda:1'), covar=tensor([0.0898, 0.0659, 0.0838, 0.0723, 0.0955, 0.1387, 0.0741, 0.0953], device='cuda:1'), in_proj_covar=tensor([0.0067, 0.0075, 0.0074, 0.0068, 0.0079, 0.0096, 0.0082, 0.0077], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-26 17:21:13,391 INFO [finetune.py:1010] (1/7) Epoch 6, validation: loss=0.1605, simple_loss=0.2337, pruned_loss=0.04366, over 2265189.00 frames. 2023-04-26 17:21:13,392 INFO [finetune.py:1011] (1/7) Maximum memory allocated so far is 6391MB 2023-04-26 17:21:19,204 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=28643.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 17:21:56,843 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.2958, 2.2102, 2.5206, 2.7711, 2.7059, 2.0669, 1.8214, 2.3414], device='cuda:1'), covar=tensor([0.0992, 0.1072, 0.0588, 0.0626, 0.0652, 0.1119, 0.0983, 0.0680], device='cuda:1'), in_proj_covar=tensor([0.0202, 0.0205, 0.0181, 0.0178, 0.0177, 0.0193, 0.0165, 0.0188], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 17:21:59,789 INFO [finetune.py:976] (1/7) Epoch 6, batch 50, loss[loss=0.1978, simple_loss=0.2694, pruned_loss=0.06308, over 4886.00 frames. ], tot_loss[loss=0.2107, simple_loss=0.272, pruned_loss=0.07472, over 215628.73 frames. ], batch size: 43, lr: 3.92e-03, grad_scale: 32.0 2023-04-26 17:22:11,068 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.1166, 1.5480, 1.3087, 1.7360, 1.5588, 1.9374, 1.3648, 3.0605], device='cuda:1'), covar=tensor([0.0700, 0.0751, 0.0797, 0.1145, 0.0651, 0.0446, 0.0736, 0.0170], device='cuda:1'), in_proj_covar=tensor([0.0039, 0.0039, 0.0041, 0.0045, 0.0041, 0.0040, 0.0040, 0.0061], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0017], device='cuda:1') 2023-04-26 17:22:24,839 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.026e+02 1.869e+02 2.241e+02 2.636e+02 5.025e+02, threshold=4.483e+02, percent-clipped=7.0 2023-04-26 17:22:33,639 INFO [finetune.py:976] (1/7) Epoch 6, batch 100, loss[loss=0.1753, simple_loss=0.2352, pruned_loss=0.05767, over 4829.00 frames. ], tot_loss[loss=0.207, simple_loss=0.2661, pruned_loss=0.07395, over 379093.04 frames. ], batch size: 33, lr: 3.92e-03, grad_scale: 32.0 2023-04-26 17:22:43,751 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.1656, 1.6089, 1.3944, 1.8067, 1.5597, 2.2492, 1.3528, 3.5020], device='cuda:1'), covar=tensor([0.0645, 0.0712, 0.0762, 0.1079, 0.0611, 0.0484, 0.0709, 0.0140], device='cuda:1'), in_proj_covar=tensor([0.0039, 0.0039, 0.0041, 0.0045, 0.0041, 0.0040, 0.0040, 0.0061], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0017], device='cuda:1') 2023-04-26 17:22:58,430 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.48 vs. limit=2.0 2023-04-26 17:23:06,830 INFO [finetune.py:976] (1/7) Epoch 6, batch 150, loss[loss=0.2146, simple_loss=0.2607, pruned_loss=0.08426, over 4853.00 frames. ], tot_loss[loss=0.2054, simple_loss=0.2635, pruned_loss=0.07364, over 507519.25 frames. ], batch size: 44, lr: 3.92e-03, grad_scale: 32.0 2023-04-26 17:23:07,005 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=3.76 vs. limit=5.0 2023-04-26 17:23:36,887 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.113e+02 1.735e+02 2.162e+02 2.514e+02 4.413e+02, threshold=4.325e+02, percent-clipped=0.0 2023-04-26 17:23:57,269 INFO [finetune.py:976] (1/7) Epoch 6, batch 200, loss[loss=0.1831, simple_loss=0.2492, pruned_loss=0.05855, over 4769.00 frames. ], tot_loss[loss=0.2041, simple_loss=0.2627, pruned_loss=0.07279, over 608534.10 frames. ], batch size: 26, lr: 3.92e-03, grad_scale: 32.0 2023-04-26 17:23:59,202 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=28842.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 17:24:08,153 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=28848.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 17:24:22,303 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=28861.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 17:24:55,334 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=28885.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 17:25:03,860 INFO [finetune.py:976] (1/7) Epoch 6, batch 250, loss[loss=0.2051, simple_loss=0.2837, pruned_loss=0.06324, over 4811.00 frames. ], tot_loss[loss=0.2102, simple_loss=0.2697, pruned_loss=0.07533, over 683482.71 frames. ], batch size: 51, lr: 3.92e-03, grad_scale: 16.0 2023-04-26 17:25:15,805 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7178, 2.2094, 1.8498, 2.1733, 1.6069, 1.8499, 1.8232, 1.5747], device='cuda:1'), covar=tensor([0.2253, 0.1372, 0.0928, 0.1359, 0.3338, 0.1262, 0.1909, 0.2903], device='cuda:1'), in_proj_covar=tensor([0.0305, 0.0323, 0.0235, 0.0298, 0.0317, 0.0278, 0.0266, 0.0290], device='cuda:1'), out_proj_covar=tensor([1.2401e-04, 1.3126e-04, 9.5598e-05, 1.1960e-04, 1.3056e-04, 1.1257e-04, 1.0937e-04, 1.1682e-04], device='cuda:1') 2023-04-26 17:25:18,973 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-04-26 17:25:29,052 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=28911.0, num_to_drop=1, layers_to_drop={2} 2023-04-26 17:25:40,821 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=28922.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 17:25:46,465 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.358e+02 1.791e+02 2.229e+02 2.957e+02 8.157e+02, threshold=4.458e+02, percent-clipped=7.0 2023-04-26 17:25:54,906 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=28933.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 17:25:59,501 INFO [finetune.py:976] (1/7) Epoch 6, batch 300, loss[loss=0.2389, simple_loss=0.2996, pruned_loss=0.08909, over 4725.00 frames. ], tot_loss[loss=0.2129, simple_loss=0.2731, pruned_loss=0.07629, over 743463.33 frames. ], batch size: 54, lr: 3.92e-03, grad_scale: 16.0 2023-04-26 17:26:08,040 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=28943.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 17:26:28,686 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=28959.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 17:27:05,112 INFO [finetune.py:976] (1/7) Epoch 6, batch 350, loss[loss=0.267, simple_loss=0.311, pruned_loss=0.1115, over 4816.00 frames. ], tot_loss[loss=0.2128, simple_loss=0.2742, pruned_loss=0.0757, over 791864.66 frames. ], batch size: 39, lr: 3.92e-03, grad_scale: 16.0 2023-04-26 17:27:11,878 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=28991.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 17:27:24,944 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=29002.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 17:27:53,126 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.162e+02 1.828e+02 2.192e+02 2.525e+02 3.973e+02, threshold=4.385e+02, percent-clipped=0.0 2023-04-26 17:27:59,802 INFO [finetune.py:976] (1/7) Epoch 6, batch 400, loss[loss=0.2067, simple_loss=0.2735, pruned_loss=0.06998, over 4831.00 frames. ], tot_loss[loss=0.2119, simple_loss=0.2742, pruned_loss=0.07478, over 827558.56 frames. ], batch size: 30, lr: 3.92e-03, grad_scale: 16.0 2023-04-26 17:28:01,668 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 2023-04-26 17:28:16,904 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=29063.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 17:28:32,152 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.38 vs. limit=5.0 2023-04-26 17:28:33,169 INFO [finetune.py:976] (1/7) Epoch 6, batch 450, loss[loss=0.1872, simple_loss=0.2508, pruned_loss=0.06177, over 4867.00 frames. ], tot_loss[loss=0.2106, simple_loss=0.2728, pruned_loss=0.07416, over 856369.12 frames. ], batch size: 34, lr: 3.92e-03, grad_scale: 16.0 2023-04-26 17:28:59,674 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.119e+02 1.667e+02 1.998e+02 2.325e+02 3.966e+02, threshold=3.996e+02, percent-clipped=0.0 2023-04-26 17:29:06,386 INFO [finetune.py:976] (1/7) Epoch 6, batch 500, loss[loss=0.2018, simple_loss=0.2681, pruned_loss=0.06771, over 4881.00 frames. ], tot_loss[loss=0.2079, simple_loss=0.2699, pruned_loss=0.07292, over 880019.30 frames. ], batch size: 31, lr: 3.92e-03, grad_scale: 16.0 2023-04-26 17:29:08,314 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=29142.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 17:29:13,414 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=29148.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 17:29:13,650 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2023-04-26 17:29:37,381 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6994, 1.2283, 1.3940, 1.3464, 1.9419, 1.5045, 1.1948, 1.2847], device='cuda:1'), covar=tensor([0.1680, 0.1337, 0.1739, 0.1413, 0.0744, 0.1452, 0.1996, 0.1848], device='cuda:1'), in_proj_covar=tensor([0.0308, 0.0328, 0.0347, 0.0304, 0.0338, 0.0333, 0.0304, 0.0347], device='cuda:1'), out_proj_covar=tensor([6.6162e-05, 6.9969e-05, 7.4965e-05, 6.3463e-05, 7.1518e-05, 7.2190e-05, 6.5961e-05, 7.4830e-05], device='cuda:1') 2023-04-26 17:29:39,681 INFO [finetune.py:976] (1/7) Epoch 6, batch 550, loss[loss=0.2021, simple_loss=0.252, pruned_loss=0.07613, over 4828.00 frames. ], tot_loss[loss=0.2064, simple_loss=0.2676, pruned_loss=0.07265, over 895960.39 frames. ], batch size: 30, lr: 3.92e-03, grad_scale: 16.0 2023-04-26 17:29:40,347 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=29190.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 17:29:44,457 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=29196.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 17:29:48,665 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-04-26 17:29:52,308 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.69 vs. limit=5.0 2023-04-26 17:29:59,118 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=29217.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 17:30:06,230 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.252e+02 1.833e+02 2.160e+02 2.601e+02 5.201e+02, threshold=4.320e+02, percent-clipped=1.0 2023-04-26 17:30:11,162 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=29236.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 17:30:12,893 INFO [finetune.py:976] (1/7) Epoch 6, batch 600, loss[loss=0.174, simple_loss=0.2384, pruned_loss=0.05481, over 4774.00 frames. ], tot_loss[loss=0.2084, simple_loss=0.269, pruned_loss=0.07389, over 908789.95 frames. ], batch size: 26, lr: 3.92e-03, grad_scale: 16.0 2023-04-26 17:30:47,030 INFO [finetune.py:976] (1/7) Epoch 6, batch 650, loss[loss=0.165, simple_loss=0.2367, pruned_loss=0.04661, over 4811.00 frames. ], tot_loss[loss=0.2096, simple_loss=0.271, pruned_loss=0.07411, over 919631.54 frames. ], batch size: 45, lr: 3.92e-03, grad_scale: 16.0 2023-04-26 17:30:57,023 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5827, 1.6375, 0.8550, 1.3059, 1.6731, 1.4760, 1.3687, 1.4291], device='cuda:1'), covar=tensor([0.0576, 0.0422, 0.0417, 0.0617, 0.0318, 0.0583, 0.0584, 0.0657], device='cuda:1'), in_proj_covar=tensor([0.0031, 0.0026, 0.0024, 0.0030, 0.0021, 0.0030, 0.0029, 0.0031], device='cuda:1'), out_proj_covar=tensor([0.0049, 0.0045, 0.0038, 0.0049, 0.0037, 0.0048, 0.0047, 0.0050], device='cuda:1') 2023-04-26 17:30:57,645 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=29297.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 17:31:17,382 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7463, 2.0594, 1.7508, 1.9984, 1.5689, 1.7097, 1.8904, 1.5994], device='cuda:1'), covar=tensor([0.1330, 0.0953, 0.0816, 0.0980, 0.2244, 0.0972, 0.1145, 0.1757], device='cuda:1'), in_proj_covar=tensor([0.0307, 0.0326, 0.0236, 0.0301, 0.0319, 0.0281, 0.0268, 0.0292], device='cuda:1'), out_proj_covar=tensor([1.2491e-04, 1.3243e-04, 9.6161e-05, 1.2093e-04, 1.3140e-04, 1.1365e-04, 1.1034e-04, 1.1792e-04], device='cuda:1') 2023-04-26 17:31:42,431 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.237e+02 1.829e+02 2.148e+02 2.649e+02 3.979e+02, threshold=4.296e+02, percent-clipped=0.0 2023-04-26 17:32:01,093 INFO [finetune.py:976] (1/7) Epoch 6, batch 700, loss[loss=0.2345, simple_loss=0.3015, pruned_loss=0.08376, over 4800.00 frames. ], tot_loss[loss=0.2102, simple_loss=0.2723, pruned_loss=0.07404, over 927830.58 frames. ], batch size: 51, lr: 3.92e-03, grad_scale: 16.0 2023-04-26 17:32:25,022 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=29358.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 17:33:07,833 INFO [finetune.py:976] (1/7) Epoch 6, batch 750, loss[loss=0.2107, simple_loss=0.2756, pruned_loss=0.0729, over 4827.00 frames. ], tot_loss[loss=0.2111, simple_loss=0.2734, pruned_loss=0.07438, over 934038.38 frames. ], batch size: 30, lr: 3.92e-03, grad_scale: 16.0 2023-04-26 17:33:23,039 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.3027, 1.5519, 1.5527, 2.1979, 2.4675, 1.9136, 1.8089, 1.6022], device='cuda:1'), covar=tensor([0.1969, 0.2060, 0.2550, 0.2043, 0.1612, 0.2482, 0.3239, 0.2400], device='cuda:1'), in_proj_covar=tensor([0.0311, 0.0332, 0.0351, 0.0308, 0.0342, 0.0337, 0.0307, 0.0352], device='cuda:1'), out_proj_covar=tensor([6.6792e-05, 7.0915e-05, 7.6051e-05, 6.4234e-05, 7.2273e-05, 7.2948e-05, 6.6471e-05, 7.5794e-05], device='cuda:1') 2023-04-26 17:33:42,162 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.8414, 2.8755, 2.2147, 3.2789, 2.8831, 2.8560, 1.0299, 2.8193], device='cuda:1'), covar=tensor([0.2292, 0.1651, 0.3480, 0.2988, 0.2957, 0.2194, 0.6212, 0.3042], device='cuda:1'), in_proj_covar=tensor([0.0246, 0.0218, 0.0254, 0.0312, 0.0303, 0.0257, 0.0275, 0.0276], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-26 17:34:03,539 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.229e+02 1.738e+02 2.111e+02 2.599e+02 4.246e+02, threshold=4.221e+02, percent-clipped=0.0 2023-04-26 17:34:05,521 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.1451, 2.1833, 2.0664, 1.8816, 2.3806, 1.7756, 2.9835, 1.7990], device='cuda:1'), covar=tensor([0.4256, 0.1965, 0.4973, 0.3457, 0.1927, 0.2886, 0.1359, 0.4373], device='cuda:1'), in_proj_covar=tensor([0.0351, 0.0359, 0.0439, 0.0370, 0.0398, 0.0387, 0.0393, 0.0423], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 17:34:14,519 INFO [finetune.py:976] (1/7) Epoch 6, batch 800, loss[loss=0.2167, simple_loss=0.2612, pruned_loss=0.08609, over 4803.00 frames. ], tot_loss[loss=0.2111, simple_loss=0.2733, pruned_loss=0.07449, over 939368.16 frames. ], batch size: 25, lr: 3.92e-03, grad_scale: 16.0 2023-04-26 17:34:48,534 INFO [finetune.py:976] (1/7) Epoch 6, batch 850, loss[loss=0.1976, simple_loss=0.2648, pruned_loss=0.06517, over 4885.00 frames. ], tot_loss[loss=0.2096, simple_loss=0.2712, pruned_loss=0.07403, over 941497.83 frames. ], batch size: 35, lr: 3.92e-03, grad_scale: 16.0 2023-04-26 17:35:07,986 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=29517.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 17:35:14,512 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.189e+02 1.686e+02 1.967e+02 2.412e+02 4.596e+02, threshold=3.934e+02, percent-clipped=4.0 2023-04-26 17:35:22,137 INFO [finetune.py:976] (1/7) Epoch 6, batch 900, loss[loss=0.2026, simple_loss=0.2706, pruned_loss=0.06726, over 4860.00 frames. ], tot_loss[loss=0.2072, simple_loss=0.268, pruned_loss=0.07325, over 944853.51 frames. ], batch size: 31, lr: 3.91e-03, grad_scale: 16.0 2023-04-26 17:35:30,037 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=29551.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 17:35:38,981 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=29565.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 17:35:56,021 INFO [finetune.py:976] (1/7) Epoch 6, batch 950, loss[loss=0.1985, simple_loss=0.2616, pruned_loss=0.06767, over 4913.00 frames. ], tot_loss[loss=0.2051, simple_loss=0.2661, pruned_loss=0.07208, over 947020.51 frames. ], batch size: 43, lr: 3.91e-03, grad_scale: 16.0 2023-04-26 17:35:57,886 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=29592.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 17:36:10,589 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=29612.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 17:36:27,225 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.239e+02 1.864e+02 2.273e+02 2.659e+02 4.416e+02, threshold=4.547e+02, percent-clipped=3.0 2023-04-26 17:36:39,505 INFO [finetune.py:976] (1/7) Epoch 6, batch 1000, loss[loss=0.2096, simple_loss=0.2794, pruned_loss=0.0699, over 4794.00 frames. ], tot_loss[loss=0.2096, simple_loss=0.2709, pruned_loss=0.07415, over 948848.70 frames. ], batch size: 29, lr: 3.91e-03, grad_scale: 16.0 2023-04-26 17:36:45,979 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6922, 1.7474, 1.9773, 2.2045, 2.1277, 1.8094, 1.3622, 1.8908], device='cuda:1'), covar=tensor([0.1156, 0.1142, 0.0721, 0.0641, 0.0721, 0.0964, 0.1007, 0.0694], device='cuda:1'), in_proj_covar=tensor([0.0203, 0.0206, 0.0181, 0.0179, 0.0179, 0.0193, 0.0166, 0.0190], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 17:37:09,281 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=29658.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 17:37:11,703 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=29662.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 17:37:52,136 INFO [finetune.py:976] (1/7) Epoch 6, batch 1050, loss[loss=0.2459, simple_loss=0.3041, pruned_loss=0.09389, over 4811.00 frames. ], tot_loss[loss=0.2118, simple_loss=0.2738, pruned_loss=0.07485, over 948703.12 frames. ], batch size: 38, lr: 3.91e-03, grad_scale: 16.0 2023-04-26 17:38:06,039 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-04-26 17:38:13,511 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=29706.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 17:38:34,981 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=29723.0, num_to_drop=1, layers_to_drop={2} 2023-04-26 17:38:39,287 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.169e+02 1.744e+02 2.054e+02 2.475e+02 7.840e+02, threshold=4.108e+02, percent-clipped=2.0 2023-04-26 17:38:57,680 INFO [finetune.py:976] (1/7) Epoch 6, batch 1100, loss[loss=0.155, simple_loss=0.2334, pruned_loss=0.0383, over 4771.00 frames. ], tot_loss[loss=0.2131, simple_loss=0.2752, pruned_loss=0.07546, over 951369.91 frames. ], batch size: 28, lr: 3.91e-03, grad_scale: 16.0 2023-04-26 17:39:34,248 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=2.02 vs. limit=2.0 2023-04-26 17:39:58,701 INFO [finetune.py:976] (1/7) Epoch 6, batch 1150, loss[loss=0.1756, simple_loss=0.2306, pruned_loss=0.06034, over 4782.00 frames. ], tot_loss[loss=0.2127, simple_loss=0.275, pruned_loss=0.07519, over 951520.36 frames. ], batch size: 26, lr: 3.91e-03, grad_scale: 16.0 2023-04-26 17:40:07,760 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=29802.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 17:40:24,363 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.056e+02 1.784e+02 2.105e+02 2.513e+02 8.942e+02, threshold=4.210e+02, percent-clipped=4.0 2023-04-26 17:40:31,966 INFO [finetune.py:976] (1/7) Epoch 6, batch 1200, loss[loss=0.2358, simple_loss=0.2866, pruned_loss=0.09247, over 4798.00 frames. ], tot_loss[loss=0.2108, simple_loss=0.273, pruned_loss=0.07434, over 952351.51 frames. ], batch size: 45, lr: 3.91e-03, grad_scale: 16.0 2023-04-26 17:40:33,927 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.2383, 1.6035, 1.5165, 2.0815, 2.3560, 1.9243, 1.8012, 1.5349], device='cuda:1'), covar=tensor([0.1942, 0.1761, 0.1976, 0.1523, 0.1372, 0.1880, 0.1940, 0.1994], device='cuda:1'), in_proj_covar=tensor([0.0312, 0.0333, 0.0352, 0.0308, 0.0343, 0.0337, 0.0307, 0.0353], device='cuda:1'), out_proj_covar=tensor([6.6932e-05, 7.1045e-05, 7.6242e-05, 6.4261e-05, 7.2422e-05, 7.3021e-05, 6.6544e-05, 7.6005e-05], device='cuda:1') 2023-04-26 17:40:35,118 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-04-26 17:40:48,158 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=29863.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 17:40:51,821 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4766, 1.6577, 1.3318, 0.9676, 1.1728, 1.1178, 1.2302, 1.0932], device='cuda:1'), covar=tensor([0.1998, 0.1513, 0.1900, 0.2129, 0.2743, 0.2254, 0.1401, 0.2281], device='cuda:1'), in_proj_covar=tensor([0.0193, 0.0218, 0.0175, 0.0205, 0.0211, 0.0185, 0.0168, 0.0191], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-26 17:41:05,314 INFO [finetune.py:976] (1/7) Epoch 6, batch 1250, loss[loss=0.1979, simple_loss=0.2638, pruned_loss=0.06601, over 4914.00 frames. ], tot_loss[loss=0.2086, simple_loss=0.2706, pruned_loss=0.07331, over 953110.55 frames. ], batch size: 36, lr: 3.91e-03, grad_scale: 16.0 2023-04-26 17:41:07,189 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=29892.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 17:41:09,464 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4485, 3.4513, 0.8240, 1.8363, 2.0589, 2.5962, 1.9315, 0.9779], device='cuda:1'), covar=tensor([0.1408, 0.0966, 0.2237, 0.1333, 0.1011, 0.0942, 0.1517, 0.2077], device='cuda:1'), in_proj_covar=tensor([0.0120, 0.0259, 0.0145, 0.0127, 0.0136, 0.0159, 0.0123, 0.0126], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-26 17:41:17,257 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=29907.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 17:41:28,272 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=3.98 vs. limit=5.0 2023-04-26 17:41:30,521 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.076e+02 1.701e+02 1.937e+02 2.233e+02 5.324e+02, threshold=3.875e+02, percent-clipped=2.0 2023-04-26 17:41:38,730 INFO [finetune.py:976] (1/7) Epoch 6, batch 1300, loss[loss=0.1847, simple_loss=0.2346, pruned_loss=0.06736, over 4866.00 frames. ], tot_loss[loss=0.2052, simple_loss=0.2668, pruned_loss=0.07179, over 953109.96 frames. ], batch size: 31, lr: 3.91e-03, grad_scale: 16.0 2023-04-26 17:41:39,346 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=29940.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 17:41:40,192 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 2023-04-26 17:42:12,107 INFO [finetune.py:976] (1/7) Epoch 6, batch 1350, loss[loss=0.1904, simple_loss=0.2473, pruned_loss=0.06673, over 4730.00 frames. ], tot_loss[loss=0.2068, simple_loss=0.2682, pruned_loss=0.0727, over 952269.95 frames. ], batch size: 54, lr: 3.91e-03, grad_scale: 16.0 2023-04-26 17:42:15,123 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.8511, 2.8043, 2.2693, 3.2751, 2.9092, 2.8542, 1.1554, 2.7433], device='cuda:1'), covar=tensor([0.2309, 0.1791, 0.3465, 0.2925, 0.4675, 0.2482, 0.5803, 0.2988], device='cuda:1'), in_proj_covar=tensor([0.0245, 0.0219, 0.0254, 0.0311, 0.0304, 0.0256, 0.0274, 0.0275], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-26 17:42:32,672 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=30018.0, num_to_drop=1, layers_to_drop={2} 2023-04-26 17:42:39,141 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.331e+02 1.856e+02 2.103e+02 2.497e+02 6.080e+02, threshold=4.207e+02, percent-clipped=3.0 2023-04-26 17:42:41,683 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4453, 1.3057, 4.0401, 3.8019, 3.5902, 3.8760, 3.8097, 3.5419], device='cuda:1'), covar=tensor([0.7623, 0.5685, 0.1017, 0.1681, 0.1192, 0.1746, 0.1889, 0.1410], device='cuda:1'), in_proj_covar=tensor([0.0314, 0.0307, 0.0418, 0.0421, 0.0356, 0.0412, 0.0318, 0.0375], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-26 17:42:52,096 INFO [finetune.py:976] (1/7) Epoch 6, batch 1400, loss[loss=0.2234, simple_loss=0.29, pruned_loss=0.07838, over 4896.00 frames. ], tot_loss[loss=0.2095, simple_loss=0.271, pruned_loss=0.07395, over 952024.91 frames. ], batch size: 35, lr: 3.91e-03, grad_scale: 16.0 2023-04-26 17:43:56,757 INFO [finetune.py:976] (1/7) Epoch 6, batch 1450, loss[loss=0.2233, simple_loss=0.2903, pruned_loss=0.07817, over 4814.00 frames. ], tot_loss[loss=0.2094, simple_loss=0.2714, pruned_loss=0.07369, over 953792.05 frames. ], batch size: 51, lr: 3.91e-03, grad_scale: 16.0 2023-04-26 17:44:42,770 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=30125.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 17:44:44,462 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.421e+02 1.906e+02 2.234e+02 2.642e+02 4.905e+02, threshold=4.468e+02, percent-clipped=1.0 2023-04-26 17:44:54,929 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-04-26 17:44:57,143 INFO [finetune.py:976] (1/7) Epoch 6, batch 1500, loss[loss=0.2299, simple_loss=0.3112, pruned_loss=0.07428, over 4796.00 frames. ], tot_loss[loss=0.2117, simple_loss=0.2737, pruned_loss=0.07481, over 953558.47 frames. ], batch size: 51, lr: 3.91e-03, grad_scale: 16.0 2023-04-26 17:44:57,379 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.99 vs. limit=2.0 2023-04-26 17:45:25,796 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=30158.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 17:45:59,236 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=30186.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 17:46:00,949 INFO [finetune.py:976] (1/7) Epoch 6, batch 1550, loss[loss=0.186, simple_loss=0.2506, pruned_loss=0.06068, over 4805.00 frames. ], tot_loss[loss=0.2111, simple_loss=0.2733, pruned_loss=0.07443, over 952863.00 frames. ], batch size: 39, lr: 3.91e-03, grad_scale: 16.0 2023-04-26 17:46:30,716 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=30207.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 17:46:31,962 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=30209.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 17:46:56,313 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.224e+02 1.842e+02 2.143e+02 2.537e+02 6.404e+02, threshold=4.287e+02, percent-clipped=2.0 2023-04-26 17:47:09,505 INFO [finetune.py:976] (1/7) Epoch 6, batch 1600, loss[loss=0.2365, simple_loss=0.2947, pruned_loss=0.08913, over 4738.00 frames. ], tot_loss[loss=0.2091, simple_loss=0.2709, pruned_loss=0.07369, over 954181.68 frames. ], batch size: 54, lr: 3.91e-03, grad_scale: 16.0 2023-04-26 17:47:31,235 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=30255.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 17:47:37,938 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-04-26 17:47:41,427 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=30270.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 17:47:46,945 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9489, 1.8731, 2.1843, 2.4389, 2.4063, 1.9230, 1.6518, 2.0170], device='cuda:1'), covar=tensor([0.0981, 0.1141, 0.0629, 0.0648, 0.0616, 0.1012, 0.0963, 0.0620], device='cuda:1'), in_proj_covar=tensor([0.0202, 0.0205, 0.0181, 0.0178, 0.0179, 0.0194, 0.0165, 0.0189], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 17:47:53,334 INFO [finetune.py:976] (1/7) Epoch 6, batch 1650, loss[loss=0.202, simple_loss=0.2689, pruned_loss=0.06761, over 4920.00 frames. ], tot_loss[loss=0.2056, simple_loss=0.2675, pruned_loss=0.07191, over 956652.17 frames. ], batch size: 36, lr: 3.91e-03, grad_scale: 16.0 2023-04-26 17:48:13,368 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=30318.0, num_to_drop=1, layers_to_drop={2} 2023-04-26 17:48:16,465 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0843, 1.2997, 1.4490, 1.6321, 1.5264, 1.6853, 1.5293, 1.5846], device='cuda:1'), covar=tensor([0.6114, 0.9289, 0.8114, 0.7670, 0.9394, 1.3454, 0.9328, 0.8351], device='cuda:1'), in_proj_covar=tensor([0.0318, 0.0396, 0.0318, 0.0328, 0.0347, 0.0413, 0.0377, 0.0334], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-26 17:48:19,318 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.157e+02 1.688e+02 2.054e+02 2.477e+02 3.898e+02, threshold=4.108e+02, percent-clipped=0.0 2023-04-26 17:48:26,038 INFO [finetune.py:976] (1/7) Epoch 6, batch 1700, loss[loss=0.2503, simple_loss=0.2935, pruned_loss=0.1036, over 4864.00 frames. ], tot_loss[loss=0.2055, simple_loss=0.2666, pruned_loss=0.07224, over 956535.23 frames. ], batch size: 31, lr: 3.91e-03, grad_scale: 16.0 2023-04-26 17:48:26,383 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.66 vs. limit=2.0 2023-04-26 17:48:51,278 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=30360.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 17:49:00,634 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=30366.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 17:49:26,051 INFO [finetune.py:976] (1/7) Epoch 6, batch 1750, loss[loss=0.1783, simple_loss=0.2491, pruned_loss=0.05373, over 4905.00 frames. ], tot_loss[loss=0.2069, simple_loss=0.2684, pruned_loss=0.07269, over 955771.04 frames. ], batch size: 32, lr: 3.91e-03, grad_scale: 16.0 2023-04-26 17:49:37,083 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([0.9728, 1.2819, 1.1738, 1.5829, 1.3645, 1.3604, 1.2181, 2.1516], device='cuda:1'), covar=tensor([0.0574, 0.0706, 0.0719, 0.1026, 0.0573, 0.0450, 0.0657, 0.0243], device='cuda:1'), in_proj_covar=tensor([0.0039, 0.0039, 0.0041, 0.0045, 0.0041, 0.0040, 0.0040, 0.0061], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0017], device='cuda:1') 2023-04-26 17:50:03,349 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=30421.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 17:50:08,045 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.357e+02 1.827e+02 2.242e+02 2.631e+02 5.410e+02, threshold=4.483e+02, percent-clipped=4.0 2023-04-26 17:50:20,187 INFO [finetune.py:976] (1/7) Epoch 6, batch 1800, loss[loss=0.185, simple_loss=0.2533, pruned_loss=0.05836, over 4792.00 frames. ], tot_loss[loss=0.2094, simple_loss=0.2718, pruned_loss=0.07351, over 955213.85 frames. ], batch size: 29, lr: 3.91e-03, grad_scale: 16.0 2023-04-26 17:50:20,943 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.7665, 1.9178, 1.8245, 1.9933, 1.7413, 2.0202, 1.9361, 1.9335], device='cuda:1'), covar=tensor([0.6252, 1.0575, 0.8602, 0.7986, 0.9552, 1.3068, 1.1441, 0.9375], device='cuda:1'), in_proj_covar=tensor([0.0316, 0.0393, 0.0316, 0.0326, 0.0344, 0.0411, 0.0373, 0.0332], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-26 17:50:43,029 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=30458.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 17:51:09,060 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=30481.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 17:51:13,897 INFO [finetune.py:976] (1/7) Epoch 6, batch 1850, loss[loss=0.1976, simple_loss=0.2757, pruned_loss=0.05969, over 4815.00 frames. ], tot_loss[loss=0.2103, simple_loss=0.2726, pruned_loss=0.07403, over 954112.54 frames. ], batch size: 39, lr: 3.91e-03, grad_scale: 16.0 2023-04-26 17:51:24,866 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=30506.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 17:51:40,049 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.288e+02 1.791e+02 2.106e+02 2.463e+02 4.581e+02, threshold=4.213e+02, percent-clipped=1.0 2023-04-26 17:51:44,968 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 2023-04-26 17:51:47,205 INFO [finetune.py:976] (1/7) Epoch 6, batch 1900, loss[loss=0.2203, simple_loss=0.2942, pruned_loss=0.0732, over 4814.00 frames. ], tot_loss[loss=0.211, simple_loss=0.2735, pruned_loss=0.07422, over 954067.89 frames. ], batch size: 51, lr: 3.91e-03, grad_scale: 16.0 2023-04-26 17:52:17,584 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=30565.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 17:52:34,617 INFO [finetune.py:976] (1/7) Epoch 6, batch 1950, loss[loss=0.1725, simple_loss=0.2304, pruned_loss=0.0573, over 4888.00 frames. ], tot_loss[loss=0.2093, simple_loss=0.2716, pruned_loss=0.07348, over 954986.58 frames. ], batch size: 32, lr: 3.91e-03, grad_scale: 16.0 2023-04-26 17:53:00,332 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.015e+02 1.660e+02 1.977e+02 2.493e+02 4.247e+02, threshold=3.954e+02, percent-clipped=1.0 2023-04-26 17:53:08,042 INFO [finetune.py:976] (1/7) Epoch 6, batch 2000, loss[loss=0.2315, simple_loss=0.2917, pruned_loss=0.08566, over 4924.00 frames. ], tot_loss[loss=0.208, simple_loss=0.2699, pruned_loss=0.07305, over 955372.09 frames. ], batch size: 37, lr: 3.91e-03, grad_scale: 16.0 2023-04-26 17:53:17,900 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=30655.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 17:53:41,351 INFO [finetune.py:976] (1/7) Epoch 6, batch 2050, loss[loss=0.1646, simple_loss=0.234, pruned_loss=0.04758, over 4902.00 frames. ], tot_loss[loss=0.2053, simple_loss=0.2665, pruned_loss=0.07202, over 957137.00 frames. ], batch size: 32, lr: 3.91e-03, grad_scale: 16.0 2023-04-26 17:53:58,927 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=30716.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 17:53:58,994 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=30716.0, num_to_drop=1, layers_to_drop={2} 2023-04-26 17:54:12,376 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.098e+02 1.673e+02 1.948e+02 2.190e+02 5.524e+02, threshold=3.896e+02, percent-clipped=3.0 2023-04-26 17:54:25,815 INFO [finetune.py:976] (1/7) Epoch 6, batch 2100, loss[loss=0.2558, simple_loss=0.3144, pruned_loss=0.09858, over 4842.00 frames. ], tot_loss[loss=0.206, simple_loss=0.2674, pruned_loss=0.07236, over 959358.42 frames. ], batch size: 49, lr: 3.91e-03, grad_scale: 16.0 2023-04-26 17:54:54,245 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=30781.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 17:54:54,987 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.79 vs. limit=5.0 2023-04-26 17:54:59,002 INFO [finetune.py:976] (1/7) Epoch 6, batch 2150, loss[loss=0.2275, simple_loss=0.264, pruned_loss=0.09546, over 3958.00 frames. ], tot_loss[loss=0.2095, simple_loss=0.2713, pruned_loss=0.07382, over 957420.36 frames. ], batch size: 17, lr: 3.91e-03, grad_scale: 16.0 2023-04-26 17:55:40,195 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.101e+02 1.783e+02 2.191e+02 2.615e+02 4.573e+02, threshold=4.381e+02, percent-clipped=3.0 2023-04-26 17:55:40,855 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=30829.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 17:55:54,381 INFO [finetune.py:976] (1/7) Epoch 6, batch 2200, loss[loss=0.2229, simple_loss=0.2792, pruned_loss=0.08329, over 4885.00 frames. ], tot_loss[loss=0.2115, simple_loss=0.273, pruned_loss=0.07498, over 955044.56 frames. ], batch size: 34, lr: 3.91e-03, grad_scale: 16.0 2023-04-26 17:56:23,796 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=30865.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 17:56:50,920 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8392, 1.0246, 1.2946, 1.4515, 1.4285, 1.6310, 1.3920, 1.3549], device='cuda:1'), covar=tensor([0.5921, 0.8027, 0.7093, 0.6472, 0.8023, 1.1273, 0.8336, 0.7640], device='cuda:1'), in_proj_covar=tensor([0.0316, 0.0394, 0.0316, 0.0326, 0.0345, 0.0410, 0.0374, 0.0332], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-26 17:56:51,897 INFO [finetune.py:976] (1/7) Epoch 6, batch 2250, loss[loss=0.216, simple_loss=0.2832, pruned_loss=0.07444, over 4834.00 frames. ], tot_loss[loss=0.2099, simple_loss=0.2722, pruned_loss=0.0738, over 955163.69 frames. ], batch size: 49, lr: 3.91e-03, grad_scale: 32.0 2023-04-26 17:57:11,894 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=30903.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 17:57:22,854 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=30913.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 17:57:35,494 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=30925.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 17:57:42,530 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.062e+02 1.673e+02 1.940e+02 2.404e+02 3.842e+02, threshold=3.880e+02, percent-clipped=0.0 2023-04-26 17:57:46,816 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=30934.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 17:57:50,253 INFO [finetune.py:976] (1/7) Epoch 6, batch 2300, loss[loss=0.1903, simple_loss=0.2519, pruned_loss=0.06437, over 4862.00 frames. ], tot_loss[loss=0.2099, simple_loss=0.2723, pruned_loss=0.07371, over 952635.04 frames. ], batch size: 31, lr: 3.91e-03, grad_scale: 32.0 2023-04-26 17:58:07,554 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=30964.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 17:58:14,180 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8057, 1.2802, 1.3337, 1.4635, 1.9787, 1.5962, 1.2426, 1.2273], device='cuda:1'), covar=tensor([0.1754, 0.1696, 0.2147, 0.1350, 0.0899, 0.1674, 0.2357, 0.2127], device='cuda:1'), in_proj_covar=tensor([0.0310, 0.0331, 0.0349, 0.0305, 0.0340, 0.0335, 0.0308, 0.0351], device='cuda:1'), out_proj_covar=tensor([6.6442e-05, 7.0624e-05, 7.5658e-05, 6.3594e-05, 7.1838e-05, 7.2520e-05, 6.6845e-05, 7.5560e-05], device='cuda:1') 2023-04-26 17:58:21,371 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=30986.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 17:58:23,530 INFO [finetune.py:976] (1/7) Epoch 6, batch 2350, loss[loss=0.2015, simple_loss=0.2629, pruned_loss=0.07009, over 4930.00 frames. ], tot_loss[loss=0.2082, simple_loss=0.2706, pruned_loss=0.0729, over 953981.14 frames. ], batch size: 38, lr: 3.91e-03, grad_scale: 32.0 2023-04-26 17:58:28,690 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=30995.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 17:58:31,880 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.5496, 2.5740, 2.8116, 3.2554, 2.9501, 2.4845, 2.1823, 2.7153], device='cuda:1'), covar=tensor([0.1052, 0.1044, 0.0612, 0.0617, 0.0683, 0.0992, 0.0886, 0.0664], device='cuda:1'), in_proj_covar=tensor([0.0202, 0.0207, 0.0182, 0.0178, 0.0180, 0.0195, 0.0166, 0.0189], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 17:58:32,782 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([3.7384, 3.7235, 2.8226, 4.3942, 3.8078, 3.8047, 1.7243, 3.7239], device='cuda:1'), covar=tensor([0.1814, 0.1129, 0.3446, 0.1774, 0.2409, 0.1949, 0.5929, 0.2349], device='cuda:1'), in_proj_covar=tensor([0.0244, 0.0221, 0.0255, 0.0314, 0.0303, 0.0255, 0.0276, 0.0276], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-26 17:58:39,467 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=31011.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 17:58:39,495 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=31011.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 17:58:42,540 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=31016.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 17:58:49,762 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.071e+02 1.679e+02 2.129e+02 2.680e+02 4.388e+02, threshold=4.258e+02, percent-clipped=1.0 2023-04-26 17:58:51,159 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.2461, 1.4690, 1.5882, 1.6956, 1.6460, 1.7379, 1.6871, 1.6929], device='cuda:1'), covar=tensor([0.5830, 0.8247, 0.6928, 0.6486, 0.7757, 1.1653, 0.8441, 0.7378], device='cuda:1'), in_proj_covar=tensor([0.0317, 0.0395, 0.0317, 0.0328, 0.0346, 0.0412, 0.0376, 0.0333], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-26 17:58:56,918 INFO [finetune.py:976] (1/7) Epoch 6, batch 2400, loss[loss=0.2291, simple_loss=0.2826, pruned_loss=0.08777, over 4831.00 frames. ], tot_loss[loss=0.207, simple_loss=0.2687, pruned_loss=0.07268, over 953286.89 frames. ], batch size: 33, lr: 3.91e-03, grad_scale: 32.0 2023-04-26 17:59:14,967 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=31064.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 17:59:16,206 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6851, 4.0977, 0.8744, 2.3652, 2.1878, 2.6698, 2.4362, 0.8911], device='cuda:1'), covar=tensor([0.1413, 0.0927, 0.2183, 0.1178, 0.1073, 0.1140, 0.1358, 0.2332], device='cuda:1'), in_proj_covar=tensor([0.0123, 0.0263, 0.0148, 0.0128, 0.0138, 0.0161, 0.0123, 0.0127], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-26 17:59:19,937 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=31072.0, num_to_drop=1, layers_to_drop={2} 2023-04-26 17:59:25,515 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6894, 1.6796, 0.7169, 1.3805, 1.8397, 1.5704, 1.4254, 1.4855], device='cuda:1'), covar=tensor([0.0535, 0.0392, 0.0424, 0.0578, 0.0293, 0.0565, 0.0524, 0.0611], device='cuda:1'), in_proj_covar=tensor([0.0031, 0.0027, 0.0024, 0.0030, 0.0021, 0.0030, 0.0029, 0.0031], device='cuda:1'), out_proj_covar=tensor([0.0049, 0.0045, 0.0038, 0.0049, 0.0037, 0.0048, 0.0048, 0.0050], device='cuda:1') 2023-04-26 17:59:28,016 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8262, 2.4365, 1.7541, 1.6606, 1.3485, 1.4342, 1.8703, 1.2481], device='cuda:1'), covar=tensor([0.1968, 0.1655, 0.1846, 0.2273, 0.2699, 0.2168, 0.1326, 0.2406], device='cuda:1'), in_proj_covar=tensor([0.0195, 0.0220, 0.0177, 0.0207, 0.0212, 0.0186, 0.0169, 0.0193], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-26 17:59:30,874 INFO [finetune.py:976] (1/7) Epoch 6, batch 2450, loss[loss=0.2276, simple_loss=0.27, pruned_loss=0.09263, over 3930.00 frames. ], tot_loss[loss=0.2059, simple_loss=0.2666, pruned_loss=0.07259, over 953061.86 frames. ], batch size: 17, lr: 3.91e-03, grad_scale: 32.0 2023-04-26 17:59:50,324 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5098, 1.3487, 1.7685, 1.7023, 1.3796, 1.1418, 1.4199, 0.9605], device='cuda:1'), covar=tensor([0.0793, 0.0858, 0.0617, 0.0892, 0.1008, 0.1341, 0.0875, 0.0977], device='cuda:1'), in_proj_covar=tensor([0.0067, 0.0075, 0.0074, 0.0068, 0.0078, 0.0095, 0.0081, 0.0077], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-26 17:59:57,657 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.091e+02 1.941e+02 2.240e+02 2.729e+02 4.178e+02, threshold=4.480e+02, percent-clipped=0.0 2023-04-26 18:00:04,429 INFO [finetune.py:976] (1/7) Epoch 6, batch 2500, loss[loss=0.2252, simple_loss=0.2861, pruned_loss=0.08216, over 4819.00 frames. ], tot_loss[loss=0.207, simple_loss=0.268, pruned_loss=0.073, over 954056.99 frames. ], batch size: 51, lr: 3.91e-03, grad_scale: 32.0 2023-04-26 18:00:37,627 INFO [finetune.py:976] (1/7) Epoch 6, batch 2550, loss[loss=0.2536, simple_loss=0.3187, pruned_loss=0.09424, over 4871.00 frames. ], tot_loss[loss=0.2116, simple_loss=0.2727, pruned_loss=0.07522, over 953214.20 frames. ], batch size: 34, lr: 3.91e-03, grad_scale: 32.0 2023-04-26 18:01:09,724 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.313e+02 1.772e+02 2.143e+02 2.617e+02 5.206e+02, threshold=4.285e+02, percent-clipped=3.0 2023-04-26 18:01:22,514 INFO [finetune.py:976] (1/7) Epoch 6, batch 2600, loss[loss=0.229, simple_loss=0.2928, pruned_loss=0.08256, over 4824.00 frames. ], tot_loss[loss=0.2105, simple_loss=0.2723, pruned_loss=0.07433, over 954868.28 frames. ], batch size: 51, lr: 3.91e-03, grad_scale: 32.0 2023-04-26 18:01:41,981 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.80 vs. limit=5.0 2023-04-26 18:01:51,873 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=31259.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:02:06,152 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0062, 1.4284, 1.8215, 2.2781, 1.8073, 1.4000, 1.0993, 1.6353], device='cuda:1'), covar=tensor([0.4308, 0.4479, 0.2159, 0.3181, 0.3746, 0.3584, 0.5710, 0.3197], device='cuda:1'), in_proj_covar=tensor([0.0277, 0.0257, 0.0221, 0.0329, 0.0218, 0.0230, 0.0243, 0.0193], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-26 18:02:14,133 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=31281.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:02:24,319 INFO [finetune.py:976] (1/7) Epoch 6, batch 2650, loss[loss=0.2429, simple_loss=0.3113, pruned_loss=0.08729, over 4740.00 frames. ], tot_loss[loss=0.2104, simple_loss=0.2725, pruned_loss=0.07418, over 955044.88 frames. ], batch size: 54, lr: 3.91e-03, grad_scale: 32.0 2023-04-26 18:02:24,977 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=31290.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:02:55,878 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=31311.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:03:05,151 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0600, 1.2232, 5.1604, 4.7379, 4.4905, 4.8959, 4.5883, 4.4525], device='cuda:1'), covar=tensor([0.7172, 0.6562, 0.0964, 0.1855, 0.1063, 0.1436, 0.1322, 0.1617], device='cuda:1'), in_proj_covar=tensor([0.0315, 0.0307, 0.0415, 0.0421, 0.0356, 0.0411, 0.0319, 0.0376], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-26 18:03:18,345 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.052e+02 1.837e+02 2.164e+02 2.605e+02 4.419e+02, threshold=4.327e+02, percent-clipped=1.0 2023-04-26 18:03:30,923 INFO [finetune.py:976] (1/7) Epoch 6, batch 2700, loss[loss=0.1761, simple_loss=0.2491, pruned_loss=0.05154, over 4893.00 frames. ], tot_loss[loss=0.2089, simple_loss=0.2714, pruned_loss=0.07323, over 956102.81 frames. ], batch size: 43, lr: 3.90e-03, grad_scale: 32.0 2023-04-26 18:03:40,157 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.79 vs. limit=2.0 2023-04-26 18:03:52,198 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-04-26 18:04:01,828 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=31359.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:04:14,062 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=31367.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 18:04:44,532 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0244, 1.4520, 1.8248, 1.9867, 1.7344, 1.4018, 1.0419, 1.4826], device='cuda:1'), covar=tensor([0.3669, 0.4227, 0.1849, 0.2952, 0.3456, 0.3337, 0.5372, 0.2998], device='cuda:1'), in_proj_covar=tensor([0.0276, 0.0256, 0.0220, 0.0327, 0.0217, 0.0229, 0.0242, 0.0192], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-26 18:04:44,998 INFO [finetune.py:976] (1/7) Epoch 6, batch 2750, loss[loss=0.2051, simple_loss=0.2669, pruned_loss=0.07162, over 4796.00 frames. ], tot_loss[loss=0.2069, simple_loss=0.2685, pruned_loss=0.07259, over 956783.49 frames. ], batch size: 51, lr: 3.90e-03, grad_scale: 32.0 2023-04-26 18:04:47,578 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8096, 1.6904, 1.8503, 2.1971, 2.1610, 1.7998, 1.3713, 1.7568], device='cuda:1'), covar=tensor([0.0996, 0.1288, 0.0888, 0.0698, 0.0720, 0.0990, 0.1117, 0.0812], device='cuda:1'), in_proj_covar=tensor([0.0202, 0.0207, 0.0182, 0.0179, 0.0180, 0.0195, 0.0166, 0.0189], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 18:05:10,262 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=31409.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:05:28,796 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.050e+02 1.757e+02 2.143e+02 2.485e+02 4.889e+02, threshold=4.286e+02, percent-clipped=1.0 2023-04-26 18:05:35,955 INFO [finetune.py:976] (1/7) Epoch 6, batch 2800, loss[loss=0.1633, simple_loss=0.221, pruned_loss=0.05279, over 4805.00 frames. ], tot_loss[loss=0.2034, simple_loss=0.2649, pruned_loss=0.0709, over 956822.69 frames. ], batch size: 25, lr: 3.90e-03, grad_scale: 32.0 2023-04-26 18:05:57,384 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=31470.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:06:09,971 INFO [finetune.py:976] (1/7) Epoch 6, batch 2850, loss[loss=0.1465, simple_loss=0.2159, pruned_loss=0.03853, over 4761.00 frames. ], tot_loss[loss=0.2026, simple_loss=0.2639, pruned_loss=0.07071, over 956368.37 frames. ], batch size: 26, lr: 3.90e-03, grad_scale: 32.0 2023-04-26 18:06:11,332 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=31491.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 18:06:36,454 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.094e+02 1.825e+02 2.131e+02 2.409e+02 4.261e+02, threshold=4.261e+02, percent-clipped=0.0 2023-04-26 18:06:43,782 INFO [finetune.py:976] (1/7) Epoch 6, batch 2900, loss[loss=0.1634, simple_loss=0.2184, pruned_loss=0.05423, over 4801.00 frames. ], tot_loss[loss=0.2052, simple_loss=0.2668, pruned_loss=0.07178, over 957838.07 frames. ], batch size: 25, lr: 3.90e-03, grad_scale: 32.0 2023-04-26 18:06:51,785 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=31552.0, num_to_drop=1, layers_to_drop={2} 2023-04-26 18:06:55,985 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=31559.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:07:11,685 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=31581.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:07:16,955 INFO [finetune.py:976] (1/7) Epoch 6, batch 2950, loss[loss=0.2834, simple_loss=0.3407, pruned_loss=0.1131, over 4836.00 frames. ], tot_loss[loss=0.2102, simple_loss=0.2721, pruned_loss=0.07409, over 956879.62 frames. ], batch size: 49, lr: 3.90e-03, grad_scale: 32.0 2023-04-26 18:07:17,680 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=31590.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:07:22,118 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-04-26 18:07:28,066 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=31607.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:07:42,550 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.159e+02 1.797e+02 2.161e+02 2.728e+02 5.785e+02, threshold=4.321e+02, percent-clipped=2.0 2023-04-26 18:07:43,219 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=31629.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:07:49,138 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=31638.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:07:49,694 INFO [finetune.py:976] (1/7) Epoch 6, batch 3000, loss[loss=0.2156, simple_loss=0.2813, pruned_loss=0.07498, over 4802.00 frames. ], tot_loss[loss=0.2127, simple_loss=0.2746, pruned_loss=0.07537, over 955115.99 frames. ], batch size: 51, lr: 3.90e-03, grad_scale: 32.0 2023-04-26 18:07:49,694 INFO [finetune.py:1001] (1/7) Computing validation loss 2023-04-26 18:07:53,927 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6123, 1.3521, 1.6576, 1.6711, 1.4222, 1.2560, 1.4093, 1.0107], device='cuda:1'), covar=tensor([0.0500, 0.0913, 0.0728, 0.0624, 0.0789, 0.1158, 0.0632, 0.0814], device='cuda:1'), in_proj_covar=tensor([0.0067, 0.0075, 0.0074, 0.0067, 0.0078, 0.0095, 0.0081, 0.0077], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-26 18:07:59,829 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6554, 1.7015, 1.8426, 2.1488, 2.0446, 1.8200, 1.3114, 1.8206], device='cuda:1'), covar=tensor([0.0864, 0.1140, 0.0794, 0.0577, 0.0599, 0.0838, 0.0930, 0.0635], device='cuda:1'), in_proj_covar=tensor([0.0203, 0.0208, 0.0183, 0.0180, 0.0181, 0.0196, 0.0167, 0.0190], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 18:08:00,214 INFO [finetune.py:1010] (1/7) Epoch 6, validation: loss=0.1565, simple_loss=0.2301, pruned_loss=0.04144, over 2265189.00 frames. 2023-04-26 18:08:00,215 INFO [finetune.py:1011] (1/7) Maximum memory allocated so far is 6391MB 2023-04-26 18:08:17,835 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=31667.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 18:08:48,666 INFO [finetune.py:976] (1/7) Epoch 6, batch 3050, loss[loss=0.2001, simple_loss=0.2634, pruned_loss=0.06834, over 4835.00 frames. ], tot_loss[loss=0.2133, simple_loss=0.2757, pruned_loss=0.07547, over 955382.91 frames. ], batch size: 47, lr: 3.90e-03, grad_scale: 32.0 2023-04-26 18:09:10,797 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=31704.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:09:23,336 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=31715.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 18:09:40,854 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.151e+02 1.692e+02 2.096e+02 2.514e+02 4.783e+02, threshold=4.192e+02, percent-clipped=1.0 2023-04-26 18:09:53,206 INFO [finetune.py:976] (1/7) Epoch 6, batch 3100, loss[loss=0.1454, simple_loss=0.1995, pruned_loss=0.04569, over 4344.00 frames. ], tot_loss[loss=0.2113, simple_loss=0.2733, pruned_loss=0.0747, over 952990.97 frames. ], batch size: 19, lr: 3.90e-03, grad_scale: 32.0 2023-04-26 18:10:25,840 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=31765.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:10:25,887 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=31765.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:10:56,964 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=31788.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:10:57,966 INFO [finetune.py:976] (1/7) Epoch 6, batch 3150, loss[loss=0.2341, simple_loss=0.2875, pruned_loss=0.0903, over 4824.00 frames. ], tot_loss[loss=0.2084, simple_loss=0.27, pruned_loss=0.07343, over 953967.37 frames. ], batch size: 40, lr: 3.90e-03, grad_scale: 32.0 2023-04-26 18:10:59,262 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-04-26 18:11:06,883 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.6909, 1.6794, 1.9265, 2.4673, 2.5429, 2.2803, 2.1339, 1.9626], device='cuda:1'), covar=tensor([0.1781, 0.2369, 0.2445, 0.2453, 0.1603, 0.1986, 0.2721, 0.2262], device='cuda:1'), in_proj_covar=tensor([0.0315, 0.0338, 0.0356, 0.0310, 0.0347, 0.0340, 0.0314, 0.0357], device='cuda:1'), out_proj_covar=tensor([6.7443e-05, 7.2100e-05, 7.7255e-05, 6.4552e-05, 7.3170e-05, 7.3553e-05, 6.8175e-05, 7.6922e-05], device='cuda:1') 2023-04-26 18:11:09,354 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-04-26 18:11:42,899 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=31823.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:11:51,413 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 8.949e+01 1.610e+02 1.947e+02 2.415e+02 6.715e+02, threshold=3.894e+02, percent-clipped=1.0 2023-04-26 18:12:02,483 INFO [finetune.py:976] (1/7) Epoch 6, batch 3200, loss[loss=0.1917, simple_loss=0.2495, pruned_loss=0.06697, over 4829.00 frames. ], tot_loss[loss=0.2039, simple_loss=0.2651, pruned_loss=0.07132, over 953670.69 frames. ], batch size: 40, lr: 3.90e-03, grad_scale: 32.0 2023-04-26 18:12:14,985 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=31847.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 18:12:22,013 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=31849.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:12:36,973 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.2193, 1.5407, 1.7929, 1.9963, 2.2713, 2.0963, 1.7575, 1.6592], device='cuda:1'), covar=tensor([0.1933, 0.2031, 0.1822, 0.1353, 0.1046, 0.1316, 0.2550, 0.1774], device='cuda:1'), in_proj_covar=tensor([0.0315, 0.0338, 0.0356, 0.0310, 0.0346, 0.0340, 0.0314, 0.0357], device='cuda:1'), out_proj_covar=tensor([6.7395e-05, 7.2169e-05, 7.7242e-05, 6.4628e-05, 7.3126e-05, 7.3572e-05, 6.8114e-05, 7.7009e-05], device='cuda:1') 2023-04-26 18:12:57,053 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=31878.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:13:00,727 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=31884.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:13:09,141 INFO [finetune.py:976] (1/7) Epoch 6, batch 3250, loss[loss=0.2661, simple_loss=0.3165, pruned_loss=0.1078, over 4810.00 frames. ], tot_loss[loss=0.2046, simple_loss=0.266, pruned_loss=0.07167, over 953881.76 frames. ], batch size: 39, lr: 3.90e-03, grad_scale: 32.0 2023-04-26 18:13:39,526 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=31915.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:13:57,724 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.120e+02 1.762e+02 2.242e+02 2.827e+02 6.116e+02, threshold=4.483e+02, percent-clipped=8.0 2023-04-26 18:14:10,187 INFO [finetune.py:976] (1/7) Epoch 6, batch 3300, loss[loss=0.1974, simple_loss=0.2699, pruned_loss=0.06245, over 4913.00 frames. ], tot_loss[loss=0.2082, simple_loss=0.2702, pruned_loss=0.07308, over 954795.22 frames. ], batch size: 36, lr: 3.90e-03, grad_scale: 32.0 2023-04-26 18:14:10,325 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=31939.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:14:18,829 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=31950.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:14:35,928 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=31976.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:14:37,784 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.2162, 1.6154, 2.0504, 2.6013, 1.9890, 1.5604, 1.3876, 1.9410], device='cuda:1'), covar=tensor([0.4269, 0.4446, 0.2066, 0.3255, 0.3802, 0.3635, 0.5117, 0.3019], device='cuda:1'), in_proj_covar=tensor([0.0280, 0.0258, 0.0221, 0.0330, 0.0219, 0.0231, 0.0244, 0.0194], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-26 18:14:43,707 INFO [finetune.py:976] (1/7) Epoch 6, batch 3350, loss[loss=0.2302, simple_loss=0.2861, pruned_loss=0.08722, over 4816.00 frames. ], tot_loss[loss=0.2097, simple_loss=0.2718, pruned_loss=0.0738, over 953770.32 frames. ], batch size: 38, lr: 3.90e-03, grad_scale: 32.0 2023-04-26 18:15:00,879 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=32011.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:15:12,029 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.180e+02 1.734e+02 2.033e+02 2.508e+02 4.432e+02, threshold=4.066e+02, percent-clipped=0.0 2023-04-26 18:15:24,408 INFO [finetune.py:976] (1/7) Epoch 6, batch 3400, loss[loss=0.2081, simple_loss=0.2812, pruned_loss=0.06752, over 4812.00 frames. ], tot_loss[loss=0.2099, simple_loss=0.2724, pruned_loss=0.07373, over 955264.79 frames. ], batch size: 33, lr: 3.90e-03, grad_scale: 32.0 2023-04-26 18:15:55,456 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=32060.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:16:04,734 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=32065.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:16:20,371 INFO [finetune.py:976] (1/7) Epoch 6, batch 3450, loss[loss=0.2192, simple_loss=0.2721, pruned_loss=0.08312, over 4885.00 frames. ], tot_loss[loss=0.2085, simple_loss=0.2711, pruned_loss=0.07295, over 954271.82 frames. ], batch size: 35, lr: 3.90e-03, grad_scale: 32.0 2023-04-26 18:16:33,228 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-04-26 18:16:37,411 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=32113.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:16:39,379 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.47 vs. limit=2.0 2023-04-26 18:16:47,433 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.129e+02 1.699e+02 2.063e+02 2.455e+02 6.391e+02, threshold=4.126e+02, percent-clipped=3.0 2023-04-26 18:16:52,370 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0572, 2.9372, 2.2992, 2.5214, 2.0756, 2.3256, 2.5017, 1.9280], device='cuda:1'), covar=tensor([0.2794, 0.1309, 0.1020, 0.1452, 0.3299, 0.1373, 0.2559, 0.3420], device='cuda:1'), in_proj_covar=tensor([0.0306, 0.0327, 0.0237, 0.0300, 0.0322, 0.0282, 0.0268, 0.0291], device='cuda:1'), out_proj_covar=tensor([1.2445e-04, 1.3291e-04, 9.6310e-05, 1.2041e-04, 1.3245e-04, 1.1394e-04, 1.0996e-04, 1.1730e-04], device='cuda:1') 2023-04-26 18:16:54,107 INFO [finetune.py:976] (1/7) Epoch 6, batch 3500, loss[loss=0.2643, simple_loss=0.3055, pruned_loss=0.1115, over 4709.00 frames. ], tot_loss[loss=0.2068, simple_loss=0.2688, pruned_loss=0.07244, over 953472.95 frames. ], batch size: 23, lr: 3.90e-03, grad_scale: 32.0 2023-04-26 18:16:57,312 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=32144.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:16:58,536 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7365, 1.7198, 1.9669, 2.3209, 2.1930, 1.8026, 1.5190, 1.8752], device='cuda:1'), covar=tensor([0.0885, 0.1019, 0.0607, 0.0498, 0.0589, 0.0808, 0.0802, 0.0612], device='cuda:1'), in_proj_covar=tensor([0.0202, 0.0206, 0.0181, 0.0179, 0.0180, 0.0195, 0.0166, 0.0189], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 18:16:59,153 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=32147.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 18:17:21,483 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=32179.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:17:33,407 INFO [finetune.py:976] (1/7) Epoch 6, batch 3550, loss[loss=0.1826, simple_loss=0.2436, pruned_loss=0.06083, over 4754.00 frames. ], tot_loss[loss=0.2045, simple_loss=0.2662, pruned_loss=0.07146, over 955483.97 frames. ], batch size: 28, lr: 3.90e-03, grad_scale: 32.0 2023-04-26 18:17:34,133 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9714, 2.4155, 2.0690, 2.2700, 1.7132, 1.9953, 2.0447, 1.6499], device='cuda:1'), covar=tensor([0.2172, 0.1618, 0.0913, 0.1304, 0.3764, 0.1404, 0.2291, 0.3463], device='cuda:1'), in_proj_covar=tensor([0.0302, 0.0323, 0.0235, 0.0296, 0.0318, 0.0278, 0.0264, 0.0288], device='cuda:1'), out_proj_covar=tensor([1.2289e-04, 1.3126e-04, 9.5317e-05, 1.1900e-04, 1.3089e-04, 1.1244e-04, 1.0866e-04, 1.1592e-04], device='cuda:1') 2023-04-26 18:17:42,298 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.20 vs. limit=5.0 2023-04-26 18:17:43,429 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=32195.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 18:18:29,258 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.147e+02 1.692e+02 2.038e+02 2.540e+02 1.815e+03, threshold=4.076e+02, percent-clipped=3.0 2023-04-26 18:18:39,221 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=32234.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:18:41,144 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6094, 1.1659, 1.6424, 2.0460, 1.7663, 1.5567, 1.5854, 1.6479], device='cuda:1'), covar=tensor([0.8410, 1.1512, 1.1684, 1.1879, 0.9633, 1.3062, 1.4132, 1.0829], device='cuda:1'), in_proj_covar=tensor([0.0416, 0.0442, 0.0529, 0.0550, 0.0445, 0.0465, 0.0478, 0.0476], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 18:18:47,840 INFO [finetune.py:976] (1/7) Epoch 6, batch 3600, loss[loss=0.1671, simple_loss=0.2368, pruned_loss=0.04866, over 4783.00 frames. ], tot_loss[loss=0.2023, simple_loss=0.2635, pruned_loss=0.07053, over 956120.70 frames. ], batch size: 29, lr: 3.90e-03, grad_scale: 32.0 2023-04-26 18:18:48,550 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.1461, 1.5677, 1.3905, 1.7588, 1.6018, 1.9968, 1.4792, 3.3061], device='cuda:1'), covar=tensor([0.0698, 0.0739, 0.0779, 0.1132, 0.0617, 0.0577, 0.0728, 0.0156], device='cuda:1'), in_proj_covar=tensor([0.0039, 0.0040, 0.0041, 0.0045, 0.0041, 0.0040, 0.0040, 0.0061], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0017], device='cuda:1') 2023-04-26 18:19:25,129 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=32271.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:19:55,460 INFO [finetune.py:976] (1/7) Epoch 6, batch 3650, loss[loss=0.1895, simple_loss=0.259, pruned_loss=0.06005, over 4746.00 frames. ], tot_loss[loss=0.2063, simple_loss=0.2677, pruned_loss=0.07247, over 956641.61 frames. ], batch size: 27, lr: 3.90e-03, grad_scale: 32.0 2023-04-26 18:20:16,827 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=32306.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:20:19,313 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([3.6125, 3.5884, 2.5771, 4.2305, 3.7747, 3.5863, 1.8427, 3.6046], device='cuda:1'), covar=tensor([0.1646, 0.1376, 0.3108, 0.1785, 0.3719, 0.1883, 0.5510, 0.2211], device='cuda:1'), in_proj_covar=tensor([0.0242, 0.0219, 0.0254, 0.0310, 0.0300, 0.0253, 0.0275, 0.0273], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-26 18:20:43,391 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.015e+02 1.808e+02 2.117e+02 2.519e+02 3.732e+02, threshold=4.235e+02, percent-clipped=0.0 2023-04-26 18:21:03,308 INFO [finetune.py:976] (1/7) Epoch 6, batch 3700, loss[loss=0.1978, simple_loss=0.2602, pruned_loss=0.06771, over 4866.00 frames. ], tot_loss[loss=0.2095, simple_loss=0.271, pruned_loss=0.07394, over 955465.68 frames. ], batch size: 31, lr: 3.90e-03, grad_scale: 32.0 2023-04-26 18:21:05,887 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.5269, 2.6601, 2.9723, 3.3781, 2.8882, 2.5643, 2.1823, 2.6503], device='cuda:1'), covar=tensor([0.1033, 0.0844, 0.0569, 0.0651, 0.0693, 0.0993, 0.0930, 0.0781], device='cuda:1'), in_proj_covar=tensor([0.0202, 0.0207, 0.0181, 0.0179, 0.0180, 0.0195, 0.0166, 0.0190], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 18:21:13,674 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9924, 2.6469, 2.2016, 2.4576, 1.7849, 2.2148, 2.1017, 1.8592], device='cuda:1'), covar=tensor([0.2093, 0.1061, 0.0860, 0.1120, 0.3263, 0.1117, 0.1980, 0.2612], device='cuda:1'), in_proj_covar=tensor([0.0305, 0.0325, 0.0236, 0.0298, 0.0320, 0.0279, 0.0266, 0.0290], device='cuda:1'), out_proj_covar=tensor([1.2391e-04, 1.3180e-04, 9.5754e-05, 1.1973e-04, 1.3176e-04, 1.1309e-04, 1.0931e-04, 1.1671e-04], device='cuda:1') 2023-04-26 18:21:26,791 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=32360.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:22:04,004 INFO [finetune.py:976] (1/7) Epoch 6, batch 3750, loss[loss=0.1753, simple_loss=0.244, pruned_loss=0.0533, over 4847.00 frames. ], tot_loss[loss=0.2093, simple_loss=0.2715, pruned_loss=0.07352, over 954912.68 frames. ], batch size: 31, lr: 3.90e-03, grad_scale: 32.0 2023-04-26 18:22:21,626 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=32408.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:22:23,512 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([0.3128, 1.3601, 1.3661, 0.9141, 1.4128, 1.1309, 1.6868, 1.2921], device='cuda:1'), covar=tensor([0.4056, 0.1699, 0.5289, 0.3029, 0.1685, 0.2465, 0.1767, 0.4962], device='cuda:1'), in_proj_covar=tensor([0.0349, 0.0357, 0.0437, 0.0366, 0.0395, 0.0385, 0.0388, 0.0418], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 18:22:35,218 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.108e+02 1.756e+02 2.047e+02 2.329e+02 4.527e+02, threshold=4.095e+02, percent-clipped=1.0 2023-04-26 18:22:43,808 INFO [finetune.py:976] (1/7) Epoch 6, batch 3800, loss[loss=0.1861, simple_loss=0.2593, pruned_loss=0.0565, over 4815.00 frames. ], tot_loss[loss=0.2089, simple_loss=0.2717, pruned_loss=0.07303, over 954911.48 frames. ], batch size: 41, lr: 3.90e-03, grad_scale: 32.0 2023-04-26 18:22:53,208 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=32444.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:23:07,875 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-04-26 18:23:21,551 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=32479.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:23:33,783 INFO [finetune.py:976] (1/7) Epoch 6, batch 3850, loss[loss=0.1853, simple_loss=0.2493, pruned_loss=0.06069, over 4891.00 frames. ], tot_loss[loss=0.2075, simple_loss=0.2703, pruned_loss=0.07231, over 955666.75 frames. ], batch size: 32, lr: 3.90e-03, grad_scale: 32.0 2023-04-26 18:23:39,429 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.77 vs. limit=5.0 2023-04-26 18:23:41,605 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=32492.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:23:50,796 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.50 vs. limit=2.0 2023-04-26 18:24:18,831 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=32527.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:24:19,366 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.258e+02 1.697e+02 2.014e+02 2.514e+02 4.116e+02, threshold=4.029e+02, percent-clipped=1.0 2023-04-26 18:24:24,621 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=32534.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:24:28,026 INFO [finetune.py:976] (1/7) Epoch 6, batch 3900, loss[loss=0.173, simple_loss=0.2344, pruned_loss=0.05582, over 4168.00 frames. ], tot_loss[loss=0.2056, simple_loss=0.2675, pruned_loss=0.07186, over 954748.41 frames. ], batch size: 65, lr: 3.90e-03, grad_scale: 32.0 2023-04-26 18:24:31,027 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([4.0118, 3.9030, 2.8262, 4.6664, 4.1832, 3.9790, 1.5990, 3.9185], device='cuda:1'), covar=tensor([0.1529, 0.1129, 0.3109, 0.1311, 0.2414, 0.1727, 0.6325, 0.2272], device='cuda:1'), in_proj_covar=tensor([0.0245, 0.0221, 0.0257, 0.0313, 0.0304, 0.0255, 0.0277, 0.0276], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-26 18:24:53,636 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-04-26 18:25:04,562 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=32571.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:25:14,645 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.5708, 1.7709, 1.7997, 1.9614, 1.7207, 1.9750, 1.8969, 1.8955], device='cuda:1'), covar=tensor([0.6246, 0.9783, 0.8499, 0.7583, 0.9109, 1.2484, 0.9939, 0.8807], device='cuda:1'), in_proj_covar=tensor([0.0319, 0.0398, 0.0320, 0.0329, 0.0347, 0.0413, 0.0377, 0.0334], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-26 18:25:22,258 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=32582.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:25:33,446 INFO [finetune.py:976] (1/7) Epoch 6, batch 3950, loss[loss=0.227, simple_loss=0.2665, pruned_loss=0.09375, over 4336.00 frames. ], tot_loss[loss=0.2027, simple_loss=0.2641, pruned_loss=0.07065, over 954003.56 frames. ], batch size: 19, lr: 3.90e-03, grad_scale: 16.0 2023-04-26 18:25:50,979 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=32606.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:25:58,783 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=32619.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:26:10,336 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.148e+02 1.669e+02 1.908e+02 2.385e+02 4.552e+02, threshold=3.816e+02, percent-clipped=3.0 2023-04-26 18:26:22,979 INFO [finetune.py:976] (1/7) Epoch 6, batch 4000, loss[loss=0.1886, simple_loss=0.2636, pruned_loss=0.05679, over 4911.00 frames. ], tot_loss[loss=0.204, simple_loss=0.2646, pruned_loss=0.07166, over 954692.32 frames. ], batch size: 37, lr: 3.90e-03, grad_scale: 16.0 2023-04-26 18:26:41,964 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([0.7785, 1.9293, 1.8074, 1.4804, 2.0992, 1.5453, 2.5885, 1.4765], device='cuda:1'), covar=tensor([0.4056, 0.1836, 0.4955, 0.3060, 0.1628, 0.2523, 0.1268, 0.4469], device='cuda:1'), in_proj_covar=tensor([0.0351, 0.0358, 0.0438, 0.0368, 0.0397, 0.0387, 0.0388, 0.0421], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 18:26:51,942 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=32654.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:27:19,403 INFO [finetune.py:976] (1/7) Epoch 6, batch 4050, loss[loss=0.2541, simple_loss=0.3051, pruned_loss=0.1015, over 4147.00 frames. ], tot_loss[loss=0.2054, simple_loss=0.2664, pruned_loss=0.07218, over 949359.28 frames. ], batch size: 65, lr: 3.90e-03, grad_scale: 16.0 2023-04-26 18:27:23,551 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.1148, 1.4322, 1.9297, 2.4297, 1.9580, 1.4486, 1.2396, 1.7390], device='cuda:1'), covar=tensor([0.3889, 0.4431, 0.1970, 0.3455, 0.3489, 0.3380, 0.5197, 0.3163], device='cuda:1'), in_proj_covar=tensor([0.0277, 0.0255, 0.0219, 0.0326, 0.0216, 0.0228, 0.0241, 0.0191], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-26 18:27:40,447 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0995, 1.4505, 1.4237, 2.0176, 2.2012, 1.7912, 1.7036, 1.5807], device='cuda:1'), covar=tensor([0.2584, 0.2462, 0.2568, 0.2011, 0.2058, 0.3002, 0.3579, 0.2478], device='cuda:1'), in_proj_covar=tensor([0.0310, 0.0332, 0.0351, 0.0304, 0.0341, 0.0333, 0.0308, 0.0351], device='cuda:1'), out_proj_covar=tensor([6.6185e-05, 7.0781e-05, 7.6104e-05, 6.3222e-05, 7.1994e-05, 7.2087e-05, 6.6746e-05, 7.5606e-05], device='cuda:1') 2023-04-26 18:27:45,419 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=5.11 vs. limit=5.0 2023-04-26 18:27:46,364 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.068e+02 1.872e+02 2.266e+02 2.687e+02 3.966e+02, threshold=4.532e+02, percent-clipped=2.0 2023-04-26 18:27:52,924 INFO [finetune.py:976] (1/7) Epoch 6, batch 4100, loss[loss=0.1908, simple_loss=0.2523, pruned_loss=0.06466, over 4813.00 frames. ], tot_loss[loss=0.2077, simple_loss=0.2697, pruned_loss=0.07283, over 950507.91 frames. ], batch size: 40, lr: 3.90e-03, grad_scale: 16.0 2023-04-26 18:27:55,491 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.52 vs. limit=5.0 2023-04-26 18:28:26,742 INFO [finetune.py:976] (1/7) Epoch 6, batch 4150, loss[loss=0.2377, simple_loss=0.2995, pruned_loss=0.08797, over 4807.00 frames. ], tot_loss[loss=0.2093, simple_loss=0.2712, pruned_loss=0.07365, over 950611.86 frames. ], batch size: 40, lr: 3.90e-03, grad_scale: 16.0 2023-04-26 18:28:30,965 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5260, 3.6627, 0.9204, 1.7997, 1.9787, 2.5843, 2.0733, 1.0418], device='cuda:1'), covar=tensor([0.1518, 0.1177, 0.2161, 0.1534, 0.1183, 0.1159, 0.1495, 0.1997], device='cuda:1'), in_proj_covar=tensor([0.0122, 0.0260, 0.0145, 0.0128, 0.0139, 0.0160, 0.0123, 0.0127], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-26 18:29:05,515 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.362e+02 1.802e+02 2.275e+02 2.602e+02 5.008e+02, threshold=4.549e+02, percent-clipped=2.0 2023-04-26 18:29:12,121 INFO [finetune.py:976] (1/7) Epoch 6, batch 4200, loss[loss=0.1823, simple_loss=0.2396, pruned_loss=0.06255, over 4872.00 frames. ], tot_loss[loss=0.2089, simple_loss=0.2713, pruned_loss=0.0732, over 953364.66 frames. ], batch size: 34, lr: 3.90e-03, grad_scale: 16.0 2023-04-26 18:29:35,762 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7213, 1.9807, 1.6640, 1.9048, 1.4225, 1.5922, 1.6077, 1.3599], device='cuda:1'), covar=tensor([0.1621, 0.1225, 0.0947, 0.1099, 0.3541, 0.1246, 0.1787, 0.2209], device='cuda:1'), in_proj_covar=tensor([0.0307, 0.0328, 0.0238, 0.0300, 0.0323, 0.0280, 0.0269, 0.0292], device='cuda:1'), out_proj_covar=tensor([1.2455e-04, 1.3338e-04, 9.6362e-05, 1.2068e-04, 1.3281e-04, 1.1337e-04, 1.1031e-04, 1.1776e-04], device='cuda:1') 2023-04-26 18:29:45,429 INFO [finetune.py:976] (1/7) Epoch 6, batch 4250, loss[loss=0.1734, simple_loss=0.2295, pruned_loss=0.05869, over 4844.00 frames. ], tot_loss[loss=0.2076, simple_loss=0.2695, pruned_loss=0.07288, over 952665.09 frames. ], batch size: 44, lr: 3.90e-03, grad_scale: 16.0 2023-04-26 18:30:13,035 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.159e+01 1.639e+02 1.929e+02 2.275e+02 6.095e+02, threshold=3.858e+02, percent-clipped=1.0 2023-04-26 18:30:15,088 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.81 vs. limit=2.0 2023-04-26 18:30:16,354 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-04-26 18:30:19,128 INFO [finetune.py:976] (1/7) Epoch 6, batch 4300, loss[loss=0.2046, simple_loss=0.2606, pruned_loss=0.07427, over 4917.00 frames. ], tot_loss[loss=0.2051, simple_loss=0.2667, pruned_loss=0.07176, over 953688.51 frames. ], batch size: 37, lr: 3.90e-03, grad_scale: 16.0 2023-04-26 18:30:39,828 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-04-26 18:30:45,580 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.5199, 2.4380, 1.9247, 2.8540, 2.4970, 2.5498, 1.0573, 2.4169], device='cuda:1'), covar=tensor([0.2069, 0.1463, 0.2663, 0.2126, 0.3226, 0.1738, 0.4747, 0.2609], device='cuda:1'), in_proj_covar=tensor([0.0244, 0.0220, 0.0254, 0.0312, 0.0303, 0.0254, 0.0274, 0.0274], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-26 18:30:47,934 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=32966.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:31:00,247 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-04-26 18:31:19,600 INFO [finetune.py:976] (1/7) Epoch 6, batch 4350, loss[loss=0.193, simple_loss=0.2629, pruned_loss=0.0616, over 4919.00 frames. ], tot_loss[loss=0.2024, simple_loss=0.2636, pruned_loss=0.07062, over 954907.23 frames. ], batch size: 37, lr: 3.90e-03, grad_scale: 16.0 2023-04-26 18:31:29,719 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4937, 2.7135, 2.4341, 2.4289, 2.8003, 2.3565, 3.7568, 2.0861], device='cuda:1'), covar=tensor([0.4525, 0.2288, 0.4618, 0.3812, 0.2325, 0.2953, 0.1649, 0.4723], device='cuda:1'), in_proj_covar=tensor([0.0351, 0.0358, 0.0440, 0.0368, 0.0399, 0.0386, 0.0389, 0.0422], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 18:32:12,347 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=33027.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:32:13,431 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.910e+01 1.721e+02 1.970e+02 2.465e+02 4.001e+02, threshold=3.941e+02, percent-clipped=3.0 2023-04-26 18:32:13,725 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.49 vs. limit=2.0 2023-04-26 18:32:24,806 INFO [finetune.py:976] (1/7) Epoch 6, batch 4400, loss[loss=0.2246, simple_loss=0.2772, pruned_loss=0.08599, over 4912.00 frames. ], tot_loss[loss=0.2065, simple_loss=0.2673, pruned_loss=0.07282, over 955987.25 frames. ], batch size: 37, lr: 3.90e-03, grad_scale: 16.0 2023-04-26 18:33:06,916 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=33069.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:33:29,652 INFO [finetune.py:976] (1/7) Epoch 6, batch 4450, loss[loss=0.1945, simple_loss=0.2621, pruned_loss=0.06342, over 4906.00 frames. ], tot_loss[loss=0.2081, simple_loss=0.2701, pruned_loss=0.07303, over 957072.00 frames. ], batch size: 37, lr: 3.90e-03, grad_scale: 16.0 2023-04-26 18:34:07,538 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0923, 1.3333, 1.4748, 1.6341, 1.5544, 1.7249, 1.5868, 1.5989], device='cuda:1'), covar=tensor([0.5762, 0.8594, 0.7199, 0.6839, 0.8433, 1.1925, 0.8393, 0.7526], device='cuda:1'), in_proj_covar=tensor([0.0313, 0.0389, 0.0315, 0.0324, 0.0340, 0.0405, 0.0370, 0.0327], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-26 18:34:12,568 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.224e+02 1.742e+02 2.062e+02 2.633e+02 5.031e+02, threshold=4.124e+02, percent-clipped=5.0 2023-04-26 18:34:13,316 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=33130.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:34:18,682 INFO [finetune.py:976] (1/7) Epoch 6, batch 4500, loss[loss=0.2028, simple_loss=0.259, pruned_loss=0.07327, over 4233.00 frames. ], tot_loss[loss=0.2092, simple_loss=0.2712, pruned_loss=0.07363, over 955072.25 frames. ], batch size: 65, lr: 3.89e-03, grad_scale: 16.0 2023-04-26 18:34:29,033 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=33155.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:34:48,229 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5016, 1.5637, 0.6965, 1.2554, 1.6283, 1.3852, 1.3026, 1.3883], device='cuda:1'), covar=tensor([0.0569, 0.0404, 0.0438, 0.0596, 0.0323, 0.0572, 0.0543, 0.0631], device='cuda:1'), in_proj_covar=tensor([0.0031, 0.0026, 0.0024, 0.0030, 0.0021, 0.0030, 0.0029, 0.0031], device='cuda:1'), out_proj_covar=tensor([0.0049, 0.0045, 0.0038, 0.0049, 0.0037, 0.0048, 0.0047, 0.0050], device='cuda:1') 2023-04-26 18:34:52,375 INFO [finetune.py:976] (1/7) Epoch 6, batch 4550, loss[loss=0.2073, simple_loss=0.2821, pruned_loss=0.06624, over 4892.00 frames. ], tot_loss[loss=0.2113, simple_loss=0.2735, pruned_loss=0.07453, over 956851.73 frames. ], batch size: 43, lr: 3.89e-03, grad_scale: 16.0 2023-04-26 18:35:09,504 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=33216.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:35:18,794 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.085e+02 1.680e+02 1.969e+02 2.447e+02 4.378e+02, threshold=3.937e+02, percent-clipped=2.0 2023-04-26 18:35:25,829 INFO [finetune.py:976] (1/7) Epoch 6, batch 4600, loss[loss=0.1705, simple_loss=0.2484, pruned_loss=0.0463, over 4849.00 frames. ], tot_loss[loss=0.2091, simple_loss=0.2723, pruned_loss=0.07294, over 958524.09 frames. ], batch size: 44, lr: 3.89e-03, grad_scale: 16.0 2023-04-26 18:35:29,553 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([0.5416, 1.6720, 1.6737, 1.2952, 1.7898, 1.4634, 2.3241, 1.4694], device='cuda:1'), covar=tensor([0.4046, 0.1840, 0.5037, 0.3138, 0.1724, 0.2227, 0.1509, 0.4735], device='cuda:1'), in_proj_covar=tensor([0.0352, 0.0359, 0.0441, 0.0368, 0.0398, 0.0387, 0.0390, 0.0423], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 18:35:59,783 INFO [finetune.py:976] (1/7) Epoch 6, batch 4650, loss[loss=0.2064, simple_loss=0.2615, pruned_loss=0.07569, over 4742.00 frames. ], tot_loss[loss=0.2051, simple_loss=0.2681, pruned_loss=0.07104, over 959361.57 frames. ], batch size: 59, lr: 3.89e-03, grad_scale: 16.0 2023-04-26 18:36:11,415 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.1138, 1.8520, 2.0471, 2.4189, 2.3672, 1.9988, 1.6950, 2.1657], device='cuda:1'), covar=tensor([0.0692, 0.0919, 0.0561, 0.0475, 0.0565, 0.0804, 0.0847, 0.0527], device='cuda:1'), in_proj_covar=tensor([0.0201, 0.0205, 0.0179, 0.0177, 0.0179, 0.0194, 0.0164, 0.0187], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 18:36:20,281 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=33322.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:36:24,421 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6787, 1.3086, 1.8295, 2.0872, 1.8306, 1.6318, 1.7040, 1.7756], device='cuda:1'), covar=tensor([0.8599, 1.2197, 1.1954, 1.2327, 0.9645, 1.4139, 1.4545, 1.2224], device='cuda:1'), in_proj_covar=tensor([0.0414, 0.0440, 0.0524, 0.0545, 0.0443, 0.0462, 0.0475, 0.0473], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 18:36:25,472 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.206e+02 1.668e+02 2.007e+02 2.366e+02 4.187e+02, threshold=4.014e+02, percent-clipped=2.0 2023-04-26 18:36:32,593 INFO [finetune.py:976] (1/7) Epoch 6, batch 4700, loss[loss=0.166, simple_loss=0.2415, pruned_loss=0.0453, over 4897.00 frames. ], tot_loss[loss=0.2015, simple_loss=0.2641, pruned_loss=0.06948, over 957733.22 frames. ], batch size: 32, lr: 3.89e-03, grad_scale: 16.0 2023-04-26 18:37:19,261 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.0586, 2.0499, 1.7219, 1.6568, 2.1149, 1.7535, 2.6446, 1.5168], device='cuda:1'), covar=tensor([0.3982, 0.2024, 0.4923, 0.3490, 0.1977, 0.2637, 0.1438, 0.4482], device='cuda:1'), in_proj_covar=tensor([0.0345, 0.0351, 0.0434, 0.0361, 0.0390, 0.0380, 0.0383, 0.0416], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 18:37:20,897 INFO [finetune.py:976] (1/7) Epoch 6, batch 4750, loss[loss=0.1771, simple_loss=0.2295, pruned_loss=0.06237, over 4824.00 frames. ], tot_loss[loss=0.1993, simple_loss=0.2611, pruned_loss=0.06877, over 956581.14 frames. ], batch size: 25, lr: 3.89e-03, grad_scale: 16.0 2023-04-26 18:37:24,496 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8198, 1.3525, 1.3685, 1.5003, 2.0202, 1.5985, 1.2831, 1.3279], device='cuda:1'), covar=tensor([0.1667, 0.1546, 0.1989, 0.1406, 0.0724, 0.1647, 0.2239, 0.2107], device='cuda:1'), in_proj_covar=tensor([0.0310, 0.0332, 0.0351, 0.0304, 0.0341, 0.0333, 0.0309, 0.0353], device='cuda:1'), out_proj_covar=tensor([6.6272e-05, 7.0761e-05, 7.6099e-05, 6.3203e-05, 7.1943e-05, 7.1894e-05, 6.6962e-05, 7.5948e-05], device='cuda:1') 2023-04-26 18:38:06,051 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=33425.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:38:14,209 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.077e+02 1.777e+02 2.013e+02 2.553e+02 6.681e+02, threshold=4.026e+02, percent-clipped=2.0 2023-04-26 18:38:18,097 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.52 vs. limit=2.0 2023-04-26 18:38:27,279 INFO [finetune.py:976] (1/7) Epoch 6, batch 4800, loss[loss=0.2405, simple_loss=0.2896, pruned_loss=0.09573, over 4748.00 frames. ], tot_loss[loss=0.2036, simple_loss=0.2656, pruned_loss=0.07085, over 955901.86 frames. ], batch size: 27, lr: 3.89e-03, grad_scale: 16.0 2023-04-26 18:39:01,121 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=33472.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:39:05,354 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6559, 1.3968, 1.7329, 2.0571, 1.7852, 1.6078, 1.6502, 1.6475], device='cuda:1'), covar=tensor([0.8147, 1.1019, 1.1766, 1.1408, 0.9523, 1.3289, 1.3525, 1.1694], device='cuda:1'), in_proj_covar=tensor([0.0411, 0.0437, 0.0522, 0.0543, 0.0440, 0.0460, 0.0472, 0.0471], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 18:39:12,307 INFO [finetune.py:976] (1/7) Epoch 6, batch 4850, loss[loss=0.2546, simple_loss=0.3217, pruned_loss=0.09375, over 4900.00 frames. ], tot_loss[loss=0.2064, simple_loss=0.2696, pruned_loss=0.07158, over 956542.54 frames. ], batch size: 43, lr: 3.89e-03, grad_scale: 16.0 2023-04-26 18:39:19,953 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.2244, 2.5244, 1.0941, 1.3332, 1.9458, 1.3548, 3.3776, 1.7619], device='cuda:1'), covar=tensor([0.0657, 0.0628, 0.0858, 0.1474, 0.0565, 0.1078, 0.0351, 0.0699], device='cuda:1'), in_proj_covar=tensor([0.0053, 0.0069, 0.0052, 0.0048, 0.0053, 0.0054, 0.0081, 0.0052], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0011, 0.0008], device='cuda:1') 2023-04-26 18:39:27,142 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=33511.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:39:38,472 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.113e+02 1.796e+02 2.126e+02 2.490e+02 5.374e+02, threshold=4.252e+02, percent-clipped=2.0 2023-04-26 18:39:41,041 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=33533.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:39:44,984 INFO [finetune.py:976] (1/7) Epoch 6, batch 4900, loss[loss=0.2023, simple_loss=0.2595, pruned_loss=0.07259, over 4814.00 frames. ], tot_loss[loss=0.2083, simple_loss=0.2714, pruned_loss=0.07263, over 955468.96 frames. ], batch size: 39, lr: 3.89e-03, grad_scale: 16.0 2023-04-26 18:40:11,221 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.3632, 3.2314, 0.8041, 1.8638, 1.8048, 2.2333, 1.7916, 0.9157], device='cuda:1'), covar=tensor([0.1471, 0.0978, 0.2158, 0.1260, 0.1161, 0.1116, 0.1764, 0.1992], device='cuda:1'), in_proj_covar=tensor([0.0121, 0.0259, 0.0144, 0.0126, 0.0137, 0.0159, 0.0122, 0.0125], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-26 18:40:18,320 INFO [finetune.py:976] (1/7) Epoch 6, batch 4950, loss[loss=0.1833, simple_loss=0.2613, pruned_loss=0.05269, over 4905.00 frames. ], tot_loss[loss=0.2084, simple_loss=0.2719, pruned_loss=0.07251, over 953921.25 frames. ], batch size: 37, lr: 3.89e-03, grad_scale: 16.0 2023-04-26 18:40:27,054 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-04-26 18:40:27,386 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([4.6744, 4.6298, 3.2720, 5.3904, 4.7150, 4.7226, 2.0685, 4.6851], device='cuda:1'), covar=tensor([0.1391, 0.1006, 0.2832, 0.0767, 0.2465, 0.1432, 0.5062, 0.1733], device='cuda:1'), in_proj_covar=tensor([0.0243, 0.0218, 0.0253, 0.0311, 0.0301, 0.0254, 0.0274, 0.0273], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-26 18:40:40,481 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=33622.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:40:44,643 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.256e+02 1.697e+02 1.943e+02 2.276e+02 3.620e+02, threshold=3.887e+02, percent-clipped=0.0 2023-04-26 18:40:51,673 INFO [finetune.py:976] (1/7) Epoch 6, batch 5000, loss[loss=0.1475, simple_loss=0.2089, pruned_loss=0.043, over 4718.00 frames. ], tot_loss[loss=0.2063, simple_loss=0.2699, pruned_loss=0.07142, over 954707.21 frames. ], batch size: 23, lr: 3.89e-03, grad_scale: 16.0 2023-04-26 18:41:24,494 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=33670.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:41:28,785 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=33677.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:41:32,989 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-04-26 18:41:36,417 INFO [finetune.py:976] (1/7) Epoch 6, batch 5050, loss[loss=0.1788, simple_loss=0.2375, pruned_loss=0.06005, over 4752.00 frames. ], tot_loss[loss=0.2058, simple_loss=0.2683, pruned_loss=0.07169, over 954564.75 frames. ], batch size: 27, lr: 3.89e-03, grad_scale: 16.0 2023-04-26 18:42:01,199 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=33725.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:42:03,508 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.184e+02 1.719e+02 2.034e+02 2.333e+02 5.936e+02, threshold=4.068e+02, percent-clipped=3.0 2023-04-26 18:42:14,944 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=33738.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 18:42:15,448 INFO [finetune.py:976] (1/7) Epoch 6, batch 5100, loss[loss=0.1536, simple_loss=0.2285, pruned_loss=0.03933, over 4809.00 frames. ], tot_loss[loss=0.2016, simple_loss=0.2636, pruned_loss=0.06981, over 953578.71 frames. ], batch size: 25, lr: 3.89e-03, grad_scale: 16.0 2023-04-26 18:42:59,574 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-04-26 18:43:00,673 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=33773.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:43:21,823 INFO [finetune.py:976] (1/7) Epoch 6, batch 5150, loss[loss=0.2522, simple_loss=0.3111, pruned_loss=0.0967, over 4827.00 frames. ], tot_loss[loss=0.2031, simple_loss=0.2647, pruned_loss=0.07076, over 953171.40 frames. ], batch size: 47, lr: 3.89e-03, grad_scale: 16.0 2023-04-26 18:43:33,824 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6989, 1.6593, 1.9933, 2.1262, 1.6279, 1.3255, 1.8441, 0.9503], device='cuda:1'), covar=tensor([0.1160, 0.1182, 0.0805, 0.1074, 0.1017, 0.1549, 0.0908, 0.1308], device='cuda:1'), in_proj_covar=tensor([0.0067, 0.0075, 0.0074, 0.0068, 0.0079, 0.0096, 0.0081, 0.0078], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-26 18:43:41,353 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-04-26 18:43:42,912 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=33811.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:43:43,616 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7821, 1.3985, 1.8393, 2.1448, 1.8346, 1.7184, 1.7404, 1.8020], device='cuda:1'), covar=tensor([0.8127, 1.1346, 1.1848, 1.1316, 0.9678, 1.3488, 1.3678, 1.1087], device='cuda:1'), in_proj_covar=tensor([0.0411, 0.0437, 0.0521, 0.0541, 0.0439, 0.0459, 0.0473, 0.0471], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 18:43:54,700 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=33828.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:43:55,206 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.192e+02 1.912e+02 2.137e+02 2.570e+02 6.415e+02, threshold=4.274e+02, percent-clipped=3.0 2023-04-26 18:44:07,682 INFO [finetune.py:976] (1/7) Epoch 6, batch 5200, loss[loss=0.2271, simple_loss=0.2859, pruned_loss=0.08422, over 4903.00 frames. ], tot_loss[loss=0.2048, simple_loss=0.2669, pruned_loss=0.07134, over 953894.39 frames. ], batch size: 36, lr: 3.89e-03, grad_scale: 16.0 2023-04-26 18:44:38,046 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=33859.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:45:14,782 INFO [finetune.py:976] (1/7) Epoch 6, batch 5250, loss[loss=0.2175, simple_loss=0.2909, pruned_loss=0.07207, over 4927.00 frames. ], tot_loss[loss=0.2051, simple_loss=0.2683, pruned_loss=0.071, over 953334.07 frames. ], batch size: 33, lr: 3.89e-03, grad_scale: 16.0 2023-04-26 18:45:58,333 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.252e+02 1.714e+02 2.064e+02 2.470e+02 4.995e+02, threshold=4.129e+02, percent-clipped=2.0 2023-04-26 18:46:00,887 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.1493, 1.5059, 1.1959, 1.7855, 1.5821, 1.9013, 1.3878, 3.6514], device='cuda:1'), covar=tensor([0.0673, 0.0794, 0.0899, 0.1230, 0.0663, 0.0604, 0.0794, 0.0148], device='cuda:1'), in_proj_covar=tensor([0.0039, 0.0039, 0.0041, 0.0045, 0.0040, 0.0040, 0.0039, 0.0061], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0014, 0.0013, 0.0013, 0.0013, 0.0017], device='cuda:1') 2023-04-26 18:46:02,304 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.50 vs. limit=5.0 2023-04-26 18:46:04,410 INFO [finetune.py:976] (1/7) Epoch 6, batch 5300, loss[loss=0.211, simple_loss=0.2792, pruned_loss=0.07141, over 4909.00 frames. ], tot_loss[loss=0.208, simple_loss=0.2709, pruned_loss=0.07253, over 953726.13 frames. ], batch size: 37, lr: 3.89e-03, grad_scale: 16.0 2023-04-26 18:46:04,506 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=33939.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:46:38,099 INFO [finetune.py:976] (1/7) Epoch 6, batch 5350, loss[loss=0.1873, simple_loss=0.2495, pruned_loss=0.0626, over 4772.00 frames. ], tot_loss[loss=0.2075, simple_loss=0.2707, pruned_loss=0.07212, over 954492.81 frames. ], batch size: 26, lr: 3.89e-03, grad_scale: 16.0 2023-04-26 18:46:46,593 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=34000.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:47:06,984 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.964e+01 1.645e+02 1.895e+02 2.308e+02 4.781e+02, threshold=3.789e+02, percent-clipped=2.0 2023-04-26 18:47:09,494 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=34033.0, num_to_drop=1, layers_to_drop={3} 2023-04-26 18:47:13,056 INFO [finetune.py:976] (1/7) Epoch 6, batch 5400, loss[loss=0.2321, simple_loss=0.2777, pruned_loss=0.09324, over 4845.00 frames. ], tot_loss[loss=0.2044, simple_loss=0.2672, pruned_loss=0.07084, over 953242.49 frames. ], batch size: 47, lr: 3.89e-03, grad_scale: 16.0 2023-04-26 18:47:14,995 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6496, 1.9824, 1.5712, 1.3656, 1.3074, 1.2885, 1.6231, 1.1891], device='cuda:1'), covar=tensor([0.1719, 0.1510, 0.1706, 0.1982, 0.2689, 0.2102, 0.1201, 0.2291], device='cuda:1'), in_proj_covar=tensor([0.0195, 0.0219, 0.0175, 0.0206, 0.0211, 0.0185, 0.0166, 0.0191], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-26 18:47:16,177 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=34044.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:47:21,442 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0042, 2.7132, 2.1634, 2.5724, 1.9199, 2.2658, 2.1663, 1.8287], device='cuda:1'), covar=tensor([0.2243, 0.1249, 0.1027, 0.1145, 0.3251, 0.1307, 0.1900, 0.3070], device='cuda:1'), in_proj_covar=tensor([0.0308, 0.0325, 0.0238, 0.0298, 0.0320, 0.0279, 0.0267, 0.0291], device='cuda:1'), out_proj_covar=tensor([1.2495e-04, 1.3155e-04, 9.6451e-05, 1.1963e-04, 1.3170e-04, 1.1278e-04, 1.0935e-04, 1.1709e-04], device='cuda:1') 2023-04-26 18:47:46,753 INFO [finetune.py:976] (1/7) Epoch 6, batch 5450, loss[loss=0.212, simple_loss=0.2404, pruned_loss=0.09182, over 3726.00 frames. ], tot_loss[loss=0.2021, simple_loss=0.2639, pruned_loss=0.07014, over 952067.11 frames. ], batch size: 15, lr: 3.89e-03, grad_scale: 16.0 2023-04-26 18:47:57,053 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=34105.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:47:58,228 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0222, 2.4264, 0.9249, 1.2465, 1.7457, 1.1276, 3.0775, 1.5482], device='cuda:1'), covar=tensor([0.0724, 0.0589, 0.0820, 0.1437, 0.0576, 0.1187, 0.0342, 0.0705], device='cuda:1'), in_proj_covar=tensor([0.0054, 0.0070, 0.0052, 0.0049, 0.0053, 0.0054, 0.0082, 0.0052], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0011, 0.0008], device='cuda:1') 2023-04-26 18:48:01,939 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.6328, 1.9178, 1.7937, 1.9747, 1.7310, 2.0267, 1.9077, 1.8540], device='cuda:1'), covar=tensor([0.6716, 1.1067, 0.8420, 0.7254, 0.9165, 1.2103, 1.1012, 0.9881], device='cuda:1'), in_proj_covar=tensor([0.0320, 0.0395, 0.0319, 0.0329, 0.0346, 0.0412, 0.0374, 0.0331], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-26 18:48:12,465 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=34128.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:48:12,948 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.072e+02 1.622e+02 2.010e+02 2.382e+02 4.007e+02, threshold=4.020e+02, percent-clipped=1.0 2023-04-26 18:48:25,396 INFO [finetune.py:976] (1/7) Epoch 6, batch 5500, loss[loss=0.2989, simple_loss=0.3271, pruned_loss=0.1354, over 4749.00 frames. ], tot_loss[loss=0.2007, simple_loss=0.2617, pruned_loss=0.0699, over 949972.24 frames. ], batch size: 54, lr: 3.89e-03, grad_scale: 16.0 2023-04-26 18:48:28,593 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.1749, 2.0136, 2.5247, 2.5684, 1.9167, 1.6695, 2.1823, 1.0925], device='cuda:1'), covar=tensor([0.0651, 0.1068, 0.0547, 0.0894, 0.1097, 0.1392, 0.0922, 0.1223], device='cuda:1'), in_proj_covar=tensor([0.0068, 0.0075, 0.0074, 0.0068, 0.0079, 0.0096, 0.0081, 0.0078], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-26 18:48:29,206 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=34145.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:49:10,833 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=34176.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:49:22,960 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0031, 3.8652, 1.3109, 2.2527, 2.3694, 2.7210, 2.2988, 1.4022], device='cuda:1'), covar=tensor([0.1351, 0.1196, 0.1861, 0.1305, 0.1079, 0.1096, 0.1462, 0.1794], device='cuda:1'), in_proj_covar=tensor([0.0120, 0.0257, 0.0143, 0.0126, 0.0137, 0.0158, 0.0122, 0.0125], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-26 18:49:30,902 INFO [finetune.py:976] (1/7) Epoch 6, batch 5550, loss[loss=0.2338, simple_loss=0.3084, pruned_loss=0.07962, over 4819.00 frames. ], tot_loss[loss=0.2027, simple_loss=0.2633, pruned_loss=0.07101, over 947932.96 frames. ], batch size: 38, lr: 3.89e-03, grad_scale: 16.0 2023-04-26 18:49:32,114 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5536, 1.6468, 1.0244, 1.3040, 2.0227, 1.4722, 1.4351, 1.3552], device='cuda:1'), covar=tensor([0.0540, 0.0392, 0.0365, 0.0564, 0.0278, 0.0532, 0.0495, 0.0605], device='cuda:1'), in_proj_covar=tensor([0.0031, 0.0026, 0.0024, 0.0030, 0.0021, 0.0030, 0.0029, 0.0031], device='cuda:1'), out_proj_covar=tensor([0.0049, 0.0045, 0.0038, 0.0050, 0.0038, 0.0048, 0.0048, 0.0050], device='cuda:1') 2023-04-26 18:49:41,391 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7906, 2.6685, 2.1930, 2.5141, 1.9384, 2.1854, 2.3468, 1.7903], device='cuda:1'), covar=tensor([0.2744, 0.1447, 0.0902, 0.1355, 0.3003, 0.1403, 0.1935, 0.2664], device='cuda:1'), in_proj_covar=tensor([0.0304, 0.0322, 0.0235, 0.0295, 0.0316, 0.0277, 0.0264, 0.0287], device='cuda:1'), out_proj_covar=tensor([1.2338e-04, 1.3054e-04, 9.5168e-05, 1.1846e-04, 1.3000e-04, 1.1184e-04, 1.0814e-04, 1.1539e-04], device='cuda:1') 2023-04-26 18:49:52,184 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=34206.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:50:03,991 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4758, 0.9898, 1.2418, 1.1000, 1.6426, 1.2923, 1.0012, 1.1497], device='cuda:1'), covar=tensor([0.1756, 0.1557, 0.1853, 0.1649, 0.0774, 0.1469, 0.2287, 0.1858], device='cuda:1'), in_proj_covar=tensor([0.0309, 0.0331, 0.0350, 0.0302, 0.0340, 0.0330, 0.0309, 0.0353], device='cuda:1'), out_proj_covar=tensor([6.5979e-05, 7.0634e-05, 7.5857e-05, 6.2880e-05, 7.1584e-05, 7.1252e-05, 6.6777e-05, 7.6020e-05], device='cuda:1') 2023-04-26 18:50:24,791 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.272e+02 1.890e+02 2.190e+02 2.677e+02 6.143e+02, threshold=4.380e+02, percent-clipped=3.0 2023-04-26 18:50:36,987 INFO [finetune.py:976] (1/7) Epoch 6, batch 5600, loss[loss=0.2042, simple_loss=0.2798, pruned_loss=0.06428, over 4812.00 frames. ], tot_loss[loss=0.2055, simple_loss=0.2671, pruned_loss=0.07194, over 948120.07 frames. ], batch size: 40, lr: 3.89e-03, grad_scale: 16.0 2023-04-26 18:51:28,269 INFO [finetune.py:976] (1/7) Epoch 6, batch 5650, loss[loss=0.1642, simple_loss=0.2364, pruned_loss=0.046, over 4761.00 frames. ], tot_loss[loss=0.2079, simple_loss=0.2701, pruned_loss=0.07288, over 947425.13 frames. ], batch size: 28, lr: 3.89e-03, grad_scale: 16.0 2023-04-26 18:51:31,878 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=34295.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:51:32,567 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4533, 1.2263, 1.7095, 1.5827, 1.3243, 1.2259, 1.3015, 0.7164], device='cuda:1'), covar=tensor([0.0602, 0.0828, 0.0448, 0.0697, 0.0866, 0.1364, 0.0635, 0.1010], device='cuda:1'), in_proj_covar=tensor([0.0068, 0.0076, 0.0074, 0.0068, 0.0079, 0.0096, 0.0082, 0.0078], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-26 18:51:33,749 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.3038, 1.5175, 1.5605, 1.7228, 1.5486, 1.7150, 1.6647, 1.6158], device='cuda:1'), covar=tensor([0.6886, 1.0293, 0.8709, 0.7436, 0.9143, 1.3188, 0.9977, 0.9243], device='cuda:1'), in_proj_covar=tensor([0.0321, 0.0395, 0.0320, 0.0329, 0.0346, 0.0412, 0.0374, 0.0331], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-26 18:51:35,966 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.2649, 1.3550, 3.8255, 3.5627, 3.4021, 3.7236, 3.6864, 3.4024], device='cuda:1'), covar=tensor([0.7397, 0.5800, 0.1269, 0.2029, 0.1257, 0.1965, 0.1485, 0.1583], device='cuda:1'), in_proj_covar=tensor([0.0314, 0.0307, 0.0411, 0.0415, 0.0353, 0.0407, 0.0314, 0.0370], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-26 18:52:03,000 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.004e+02 1.702e+02 2.093e+02 2.545e+02 5.557e+02, threshold=4.186e+02, percent-clipped=1.0 2023-04-26 18:52:05,453 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=34333.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:52:09,023 INFO [finetune.py:976] (1/7) Epoch 6, batch 5700, loss[loss=0.2139, simple_loss=0.2485, pruned_loss=0.08968, over 4368.00 frames. ], tot_loss[loss=0.2035, simple_loss=0.2643, pruned_loss=0.07137, over 928416.42 frames. ], batch size: 19, lr: 3.89e-03, grad_scale: 16.0 2023-04-26 18:52:16,195 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.3364, 1.6622, 1.5234, 2.0097, 1.7146, 2.0330, 1.5481, 3.4484], device='cuda:1'), covar=tensor([0.0613, 0.0665, 0.0738, 0.1090, 0.0574, 0.0474, 0.0694, 0.0222], device='cuda:1'), in_proj_covar=tensor([0.0039, 0.0039, 0.0041, 0.0045, 0.0041, 0.0040, 0.0039, 0.0060], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0017], device='cuda:1') 2023-04-26 18:52:39,656 INFO [finetune.py:976] (1/7) Epoch 7, batch 0, loss[loss=0.2227, simple_loss=0.2808, pruned_loss=0.08235, over 4813.00 frames. ], tot_loss[loss=0.2227, simple_loss=0.2808, pruned_loss=0.08235, over 4813.00 frames. ], batch size: 33, lr: 3.89e-03, grad_scale: 16.0 2023-04-26 18:52:39,656 INFO [finetune.py:1001] (1/7) Computing validation loss 2023-04-26 18:52:49,755 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5850, 2.1145, 1.7392, 2.0268, 1.6858, 1.7377, 1.6979, 1.4515], device='cuda:1'), covar=tensor([0.2435, 0.1423, 0.1165, 0.1385, 0.3710, 0.1708, 0.2062, 0.2913], device='cuda:1'), in_proj_covar=tensor([0.0304, 0.0321, 0.0235, 0.0295, 0.0315, 0.0277, 0.0263, 0.0286], device='cuda:1'), out_proj_covar=tensor([1.2337e-04, 1.3009e-04, 9.5201e-05, 1.1844e-04, 1.2973e-04, 1.1183e-04, 1.0785e-04, 1.1517e-04], device='cuda:1') 2023-04-26 18:52:50,221 INFO [finetune.py:1010] (1/7) Epoch 7, validation: loss=0.1579, simple_loss=0.2317, pruned_loss=0.04207, over 2265189.00 frames. 2023-04-26 18:52:50,221 INFO [finetune.py:1011] (1/7) Maximum memory allocated so far is 6391MB 2023-04-26 18:52:59,337 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=34381.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:53:11,519 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=34400.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:53:23,180 INFO [finetune.py:976] (1/7) Epoch 7, batch 50, loss[loss=0.2064, simple_loss=0.2658, pruned_loss=0.07352, over 4827.00 frames. ], tot_loss[loss=0.2109, simple_loss=0.2722, pruned_loss=0.07483, over 215012.79 frames. ], batch size: 30, lr: 3.89e-03, grad_scale: 16.0 2023-04-26 18:53:23,281 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([0.9722, 1.3984, 1.2737, 1.5905, 1.4026, 1.4428, 1.2979, 2.4719], device='cuda:1'), covar=tensor([0.0674, 0.0767, 0.0788, 0.1217, 0.0654, 0.0546, 0.0747, 0.0249], device='cuda:1'), in_proj_covar=tensor([0.0039, 0.0039, 0.0041, 0.0045, 0.0041, 0.0040, 0.0039, 0.0061], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0017], device='cuda:1') 2023-04-26 18:53:32,046 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.224e+02 1.800e+02 2.094e+02 2.566e+02 4.468e+02, threshold=4.189e+02, percent-clipped=1.0 2023-04-26 18:53:34,641 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-04-26 18:53:56,438 INFO [finetune.py:976] (1/7) Epoch 7, batch 100, loss[loss=0.19, simple_loss=0.2573, pruned_loss=0.06137, over 4746.00 frames. ], tot_loss[loss=0.2038, simple_loss=0.2657, pruned_loss=0.07101, over 379965.88 frames. ], batch size: 26, lr: 3.89e-03, grad_scale: 16.0 2023-04-26 18:54:12,157 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.3621, 1.7120, 1.4762, 1.7745, 1.6612, 1.9123, 1.4014, 3.6687], device='cuda:1'), covar=tensor([0.0612, 0.0749, 0.0782, 0.1191, 0.0648, 0.0550, 0.0750, 0.0161], device='cuda:1'), in_proj_covar=tensor([0.0039, 0.0039, 0.0041, 0.0044, 0.0040, 0.0040, 0.0039, 0.0060], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0014, 0.0013, 0.0013, 0.0013, 0.0017], device='cuda:1') 2023-04-26 18:54:19,339 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=34501.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:54:29,756 INFO [finetune.py:976] (1/7) Epoch 7, batch 150, loss[loss=0.1803, simple_loss=0.2409, pruned_loss=0.05988, over 4821.00 frames. ], tot_loss[loss=0.1991, simple_loss=0.2605, pruned_loss=0.06885, over 507210.26 frames. ], batch size: 38, lr: 3.89e-03, grad_scale: 16.0 2023-04-26 18:54:30,491 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7954, 1.7389, 2.0249, 2.1822, 1.7393, 1.4123, 1.8220, 1.0868], device='cuda:1'), covar=tensor([0.1001, 0.1023, 0.0720, 0.1122, 0.0975, 0.1268, 0.1015, 0.1067], device='cuda:1'), in_proj_covar=tensor([0.0067, 0.0075, 0.0073, 0.0067, 0.0078, 0.0095, 0.0080, 0.0077], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-26 18:54:39,165 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.220e+02 1.717e+02 2.069e+02 2.417e+02 7.374e+02, threshold=4.138e+02, percent-clipped=4.0 2023-04-26 18:54:40,566 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.1989, 1.4546, 1.5137, 1.6649, 1.6132, 1.7843, 1.7329, 1.6357], device='cuda:1'), covar=tensor([0.6019, 0.8796, 0.7757, 0.7452, 0.8051, 1.2225, 0.8310, 0.7721], device='cuda:1'), in_proj_covar=tensor([0.0320, 0.0393, 0.0318, 0.0329, 0.0344, 0.0410, 0.0374, 0.0330], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-26 18:55:03,718 INFO [finetune.py:976] (1/7) Epoch 7, batch 200, loss[loss=0.1994, simple_loss=0.2674, pruned_loss=0.06572, over 4873.00 frames. ], tot_loss[loss=0.1995, simple_loss=0.2605, pruned_loss=0.06929, over 607413.44 frames. ], batch size: 31, lr: 3.89e-03, grad_scale: 32.0 2023-04-26 18:55:27,235 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6012, 3.6518, 1.1686, 1.9271, 1.9964, 2.5041, 2.1923, 1.0520], device='cuda:1'), covar=tensor([0.1459, 0.0947, 0.1867, 0.1414, 0.1155, 0.1134, 0.1402, 0.2005], device='cuda:1'), in_proj_covar=tensor([0.0120, 0.0257, 0.0143, 0.0126, 0.0137, 0.0158, 0.0121, 0.0125], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-26 18:55:27,271 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=34593.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:55:33,660 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=34595.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:56:04,707 INFO [finetune.py:976] (1/7) Epoch 7, batch 250, loss[loss=0.1595, simple_loss=0.2309, pruned_loss=0.04403, over 4762.00 frames. ], tot_loss[loss=0.2041, simple_loss=0.2659, pruned_loss=0.07119, over 684456.70 frames. ], batch size: 27, lr: 3.88e-03, grad_scale: 32.0 2023-04-26 18:56:09,046 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=34622.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:56:17,341 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=34626.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:56:19,023 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.076e+02 1.883e+02 2.183e+02 2.769e+02 4.729e+02, threshold=4.366e+02, percent-clipped=3.0 2023-04-26 18:56:39,248 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=34643.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:56:39,424 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.54 vs. limit=2.0 2023-04-26 18:56:51,389 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=34654.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:57:11,709 INFO [finetune.py:976] (1/7) Epoch 7, batch 300, loss[loss=0.1648, simple_loss=0.2342, pruned_loss=0.04772, over 4774.00 frames. ], tot_loss[loss=0.2057, simple_loss=0.2677, pruned_loss=0.07184, over 742065.63 frames. ], batch size: 28, lr: 3.88e-03, grad_scale: 32.0 2023-04-26 18:57:35,511 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=34683.0, num_to_drop=1, layers_to_drop={2} 2023-04-26 18:57:37,980 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=34687.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:57:51,404 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=34700.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:58:03,702 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=34710.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 18:58:03,850 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 2023-04-26 18:58:04,505 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-04-26 18:58:12,740 INFO [finetune.py:976] (1/7) Epoch 7, batch 350, loss[loss=0.239, simple_loss=0.2924, pruned_loss=0.09284, over 4762.00 frames. ], tot_loss[loss=0.2073, simple_loss=0.27, pruned_loss=0.07233, over 788994.68 frames. ], batch size: 28, lr: 3.88e-03, grad_scale: 32.0 2023-04-26 18:58:25,950 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.13 vs. limit=2.0 2023-04-26 18:58:27,620 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-04-26 18:58:28,031 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.287e+02 1.776e+02 2.211e+02 2.674e+02 4.769e+02, threshold=4.422e+02, percent-clipped=1.0 2023-04-26 18:58:49,832 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-04-26 18:58:50,329 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=34748.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:59:02,246 INFO [finetune.py:976] (1/7) Epoch 7, batch 400, loss[loss=0.2662, simple_loss=0.3196, pruned_loss=0.1064, over 4931.00 frames. ], tot_loss[loss=0.2072, simple_loss=0.2705, pruned_loss=0.07194, over 825901.35 frames. ], batch size: 42, lr: 3.88e-03, grad_scale: 32.0 2023-04-26 18:59:05,944 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=34771.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 18:59:14,092 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.4727, 1.8754, 2.2628, 3.0266, 2.2871, 1.7368, 1.6545, 2.2540], device='cuda:1'), covar=tensor([0.4383, 0.4376, 0.2119, 0.3379, 0.3785, 0.3659, 0.5110, 0.3112], device='cuda:1'), in_proj_covar=tensor([0.0277, 0.0254, 0.0218, 0.0323, 0.0213, 0.0228, 0.0238, 0.0190], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-26 18:59:20,063 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6941, 2.2726, 1.7285, 1.4835, 1.2520, 1.2820, 1.7733, 1.2230], device='cuda:1'), covar=tensor([0.1921, 0.1525, 0.1724, 0.2097, 0.2815, 0.2287, 0.1269, 0.2456], device='cuda:1'), in_proj_covar=tensor([0.0194, 0.0218, 0.0173, 0.0204, 0.0209, 0.0185, 0.0166, 0.0190], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-26 18:59:26,207 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=34801.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:59:31,776 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8461, 1.6740, 2.0421, 2.2628, 2.0001, 1.7375, 1.8746, 1.9346], device='cuda:1'), covar=tensor([0.9228, 1.1842, 1.4866, 1.2910, 1.0550, 1.7607, 1.8242, 1.4277], device='cuda:1'), in_proj_covar=tensor([0.0413, 0.0436, 0.0522, 0.0542, 0.0440, 0.0461, 0.0475, 0.0473], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 18:59:36,185 INFO [finetune.py:976] (1/7) Epoch 7, batch 450, loss[loss=0.1691, simple_loss=0.24, pruned_loss=0.0491, over 4818.00 frames. ], tot_loss[loss=0.2059, simple_loss=0.269, pruned_loss=0.07139, over 854244.86 frames. ], batch size: 38, lr: 3.88e-03, grad_scale: 32.0 2023-04-26 18:59:45,537 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.159e+02 1.735e+02 2.065e+02 2.383e+02 5.194e+02, threshold=4.130e+02, percent-clipped=2.0 2023-04-26 18:59:52,165 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4447, 1.7397, 1.2871, 0.9814, 1.1766, 1.1004, 1.2741, 1.0685], device='cuda:1'), covar=tensor([0.1851, 0.1461, 0.1747, 0.2094, 0.2635, 0.2274, 0.1177, 0.2159], device='cuda:1'), in_proj_covar=tensor([0.0194, 0.0218, 0.0173, 0.0205, 0.0210, 0.0185, 0.0166, 0.0190], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-26 18:59:58,665 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=34849.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:00:09,434 INFO [finetune.py:976] (1/7) Epoch 7, batch 500, loss[loss=0.2192, simple_loss=0.2755, pruned_loss=0.08138, over 4858.00 frames. ], tot_loss[loss=0.2039, simple_loss=0.2665, pruned_loss=0.07068, over 877742.53 frames. ], batch size: 31, lr: 3.88e-03, grad_scale: 32.0 2023-04-26 19:00:42,229 INFO [finetune.py:976] (1/7) Epoch 7, batch 550, loss[loss=0.215, simple_loss=0.2934, pruned_loss=0.06836, over 4801.00 frames. ], tot_loss[loss=0.2013, simple_loss=0.2633, pruned_loss=0.06966, over 897256.10 frames. ], batch size: 45, lr: 3.88e-03, grad_scale: 32.0 2023-04-26 19:00:50,617 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8507, 2.5063, 1.7930, 1.7506, 1.3290, 1.4014, 1.9257, 1.2313], device='cuda:1'), covar=tensor([0.1843, 0.1499, 0.1749, 0.2004, 0.2713, 0.2304, 0.1200, 0.2355], device='cuda:1'), in_proj_covar=tensor([0.0194, 0.0219, 0.0174, 0.0205, 0.0210, 0.0186, 0.0166, 0.0190], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-26 19:00:51,085 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.195e+02 1.669e+02 2.030e+02 2.418e+02 4.759e+02, threshold=4.059e+02, percent-clipped=2.0 2023-04-26 19:00:51,211 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7526, 1.6441, 1.9615, 2.1577, 1.5777, 1.3117, 1.7363, 1.1509], device='cuda:1'), covar=tensor([0.0899, 0.1015, 0.0849, 0.0956, 0.0956, 0.1404, 0.1064, 0.1077], device='cuda:1'), in_proj_covar=tensor([0.0067, 0.0075, 0.0073, 0.0068, 0.0078, 0.0095, 0.0081, 0.0077], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-26 19:00:52,267 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0904, 2.6465, 1.0217, 1.2883, 2.0154, 1.2564, 3.2634, 1.6147], device='cuda:1'), covar=tensor([0.0646, 0.0846, 0.0968, 0.1174, 0.0496, 0.0961, 0.0225, 0.0640], device='cuda:1'), in_proj_covar=tensor([0.0053, 0.0069, 0.0051, 0.0048, 0.0053, 0.0054, 0.0081, 0.0052], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0011, 0.0008], device='cuda:1') 2023-04-26 19:01:03,580 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-04-26 19:01:10,922 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=34949.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:01:12,887 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.73 vs. limit=2.0 2023-04-26 19:01:21,560 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5551, 0.9719, 1.2726, 1.2036, 1.7174, 1.3622, 1.0945, 1.2356], device='cuda:1'), covar=tensor([0.1565, 0.1438, 0.1844, 0.1291, 0.0775, 0.1218, 0.1865, 0.1814], device='cuda:1'), in_proj_covar=tensor([0.0312, 0.0334, 0.0354, 0.0305, 0.0341, 0.0332, 0.0310, 0.0356], device='cuda:1'), out_proj_covar=tensor([6.6580e-05, 7.1199e-05, 7.6750e-05, 6.3551e-05, 7.1887e-05, 7.1636e-05, 6.7168e-05, 7.6708e-05], device='cuda:1') 2023-04-26 19:01:25,740 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([3.3261, 3.2847, 2.3761, 3.8696, 3.3536, 3.3529, 1.4232, 3.2758], device='cuda:1'), covar=tensor([0.2157, 0.1302, 0.3449, 0.2356, 0.3385, 0.2112, 0.6160, 0.2823], device='cuda:1'), in_proj_covar=tensor([0.0242, 0.0218, 0.0253, 0.0310, 0.0300, 0.0253, 0.0272, 0.0274], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-26 19:01:32,183 INFO [finetune.py:976] (1/7) Epoch 7, batch 600, loss[loss=0.1703, simple_loss=0.2449, pruned_loss=0.04784, over 4898.00 frames. ], tot_loss[loss=0.2022, simple_loss=0.2641, pruned_loss=0.07012, over 910202.87 frames. ], batch size: 32, lr: 3.88e-03, grad_scale: 32.0 2023-04-26 19:01:45,332 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=34978.0, num_to_drop=1, layers_to_drop={3} 2023-04-26 19:01:54,262 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=34982.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:02:27,943 INFO [finetune.py:976] (1/7) Epoch 7, batch 650, loss[loss=0.2503, simple_loss=0.3157, pruned_loss=0.09244, over 4819.00 frames. ], tot_loss[loss=0.2041, simple_loss=0.267, pruned_loss=0.07058, over 921417.07 frames. ], batch size: 40, lr: 3.88e-03, grad_scale: 32.0 2023-04-26 19:02:42,342 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.097e+02 1.760e+02 2.002e+02 2.405e+02 4.301e+02, threshold=4.004e+02, percent-clipped=1.0 2023-04-26 19:02:46,716 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.60 vs. limit=2.0 2023-04-26 19:03:07,891 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.2136, 2.1112, 1.7437, 1.7858, 2.1288, 1.7271, 2.4366, 1.5227], device='cuda:1'), covar=tensor([0.3479, 0.1493, 0.4007, 0.2874, 0.1624, 0.2266, 0.1726, 0.4241], device='cuda:1'), in_proj_covar=tensor([0.0350, 0.0356, 0.0440, 0.0367, 0.0394, 0.0385, 0.0390, 0.0421], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 19:03:19,049 INFO [finetune.py:976] (1/7) Epoch 7, batch 700, loss[loss=0.2241, simple_loss=0.2906, pruned_loss=0.07878, over 4924.00 frames. ], tot_loss[loss=0.2043, simple_loss=0.268, pruned_loss=0.07027, over 928805.20 frames. ], batch size: 42, lr: 3.88e-03, grad_scale: 32.0 2023-04-26 19:03:19,113 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=35066.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 19:03:47,934 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=35092.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:04:19,896 INFO [finetune.py:976] (1/7) Epoch 7, batch 750, loss[loss=0.2182, simple_loss=0.2879, pruned_loss=0.07432, over 4831.00 frames. ], tot_loss[loss=0.2044, simple_loss=0.2684, pruned_loss=0.07014, over 934533.67 frames. ], batch size: 49, lr: 3.88e-03, grad_scale: 32.0 2023-04-26 19:04:20,636 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9310, 2.3184, 0.9521, 1.1077, 1.6490, 1.1913, 3.0046, 1.4393], device='cuda:1'), covar=tensor([0.0700, 0.0674, 0.0828, 0.1375, 0.0522, 0.1073, 0.0293, 0.0735], device='cuda:1'), in_proj_covar=tensor([0.0053, 0.0069, 0.0051, 0.0048, 0.0052, 0.0054, 0.0081, 0.0052], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0011, 0.0008], device='cuda:1') 2023-04-26 19:04:33,722 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.243e+02 1.705e+02 2.087e+02 2.408e+02 7.580e+02, threshold=4.175e+02, percent-clipped=5.0 2023-04-26 19:05:05,758 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=35153.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:05:25,181 INFO [finetune.py:976] (1/7) Epoch 7, batch 800, loss[loss=0.2271, simple_loss=0.2807, pruned_loss=0.08677, over 4866.00 frames. ], tot_loss[loss=0.2043, simple_loss=0.2683, pruned_loss=0.07012, over 938641.26 frames. ], batch size: 31, lr: 3.88e-03, grad_scale: 32.0 2023-04-26 19:06:20,320 INFO [finetune.py:976] (1/7) Epoch 7, batch 850, loss[loss=0.1911, simple_loss=0.2559, pruned_loss=0.06315, over 4702.00 frames. ], tot_loss[loss=0.2033, simple_loss=0.2666, pruned_loss=0.06996, over 943134.09 frames. ], batch size: 23, lr: 3.88e-03, grad_scale: 32.0 2023-04-26 19:06:32,704 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.117e+02 1.732e+02 2.031e+02 2.393e+02 4.995e+02, threshold=4.061e+02, percent-clipped=2.0 2023-04-26 19:06:58,280 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=35249.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:07:20,078 INFO [finetune.py:976] (1/7) Epoch 7, batch 900, loss[loss=0.1866, simple_loss=0.256, pruned_loss=0.05858, over 4841.00 frames. ], tot_loss[loss=0.2, simple_loss=0.2634, pruned_loss=0.06833, over 945620.55 frames. ], batch size: 49, lr: 3.88e-03, grad_scale: 32.0 2023-04-26 19:07:38,490 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=35278.0, num_to_drop=1, layers_to_drop={2} 2023-04-26 19:07:40,990 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=35282.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:08:02,072 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=35297.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:08:14,479 INFO [finetune.py:976] (1/7) Epoch 7, batch 950, loss[loss=0.1756, simple_loss=0.2411, pruned_loss=0.055, over 4775.00 frames. ], tot_loss[loss=0.1988, simple_loss=0.2617, pruned_loss=0.06792, over 947618.63 frames. ], batch size: 28, lr: 3.88e-03, grad_scale: 32.0 2023-04-26 19:08:27,229 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=35326.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:08:29,019 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.023e+02 1.623e+02 1.972e+02 2.345e+02 3.927e+02, threshold=3.944e+02, percent-clipped=0.0 2023-04-26 19:08:29,702 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=35330.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:08:50,413 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=35345.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 19:09:20,045 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.37 vs. limit=5.0 2023-04-26 19:09:20,442 INFO [finetune.py:976] (1/7) Epoch 7, batch 1000, loss[loss=0.2098, simple_loss=0.2754, pruned_loss=0.07206, over 4894.00 frames. ], tot_loss[loss=0.2007, simple_loss=0.2638, pruned_loss=0.06876, over 948756.24 frames. ], batch size: 35, lr: 3.88e-03, grad_scale: 32.0 2023-04-26 19:09:20,563 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=35366.0, num_to_drop=1, layers_to_drop={2} 2023-04-26 19:09:29,980 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=35374.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:10:06,708 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=35406.0, num_to_drop=1, layers_to_drop={2} 2023-04-26 19:10:18,209 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=35414.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 19:10:25,351 INFO [finetune.py:976] (1/7) Epoch 7, batch 1050, loss[loss=0.2495, simple_loss=0.3056, pruned_loss=0.09665, over 4860.00 frames. ], tot_loss[loss=0.2037, simple_loss=0.2675, pruned_loss=0.06993, over 948687.80 frames. ], batch size: 31, lr: 3.88e-03, grad_scale: 32.0 2023-04-26 19:10:39,105 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.246e+02 1.810e+02 2.204e+02 2.753e+02 5.526e+02, threshold=4.408e+02, percent-clipped=1.0 2023-04-26 19:10:48,435 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=35435.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:10:58,440 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.1904, 1.1448, 1.2460, 1.5408, 1.5922, 1.2375, 0.8511, 1.3189], device='cuda:1'), covar=tensor([0.0984, 0.1572, 0.1035, 0.0651, 0.0687, 0.1080, 0.1104, 0.0712], device='cuda:1'), in_proj_covar=tensor([0.0204, 0.0209, 0.0183, 0.0179, 0.0182, 0.0197, 0.0166, 0.0191], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 19:11:02,531 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=35448.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:11:09,518 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-04-26 19:11:31,123 INFO [finetune.py:976] (1/7) Epoch 7, batch 1100, loss[loss=0.2226, simple_loss=0.2851, pruned_loss=0.08012, over 4884.00 frames. ], tot_loss[loss=0.2037, simple_loss=0.2677, pruned_loss=0.06986, over 949023.60 frames. ], batch size: 35, lr: 3.88e-03, grad_scale: 32.0 2023-04-26 19:11:33,093 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8156, 1.1351, 1.2510, 1.4469, 1.4372, 1.6056, 1.3942, 1.3445], device='cuda:1'), covar=tensor([0.5652, 0.6890, 0.6372, 0.5617, 0.7177, 0.9739, 0.6517, 0.6226], device='cuda:1'), in_proj_covar=tensor([0.0320, 0.0391, 0.0316, 0.0327, 0.0343, 0.0408, 0.0371, 0.0328], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-26 19:12:09,915 INFO [finetune.py:976] (1/7) Epoch 7, batch 1150, loss[loss=0.1423, simple_loss=0.2228, pruned_loss=0.03091, over 4738.00 frames. ], tot_loss[loss=0.2039, simple_loss=0.268, pruned_loss=0.06993, over 950284.85 frames. ], batch size: 27, lr: 3.88e-03, grad_scale: 32.0 2023-04-26 19:12:18,334 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.291e+02 1.835e+02 2.101e+02 2.448e+02 6.183e+02, threshold=4.202e+02, percent-clipped=2.0 2023-04-26 19:12:19,093 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=35530.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:12:41,558 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.1835, 2.7307, 1.0807, 1.4234, 2.0913, 1.1936, 3.8251, 1.9641], device='cuda:1'), covar=tensor([0.0727, 0.0700, 0.0911, 0.1377, 0.0529, 0.1139, 0.0262, 0.0641], device='cuda:1'), in_proj_covar=tensor([0.0053, 0.0069, 0.0051, 0.0048, 0.0053, 0.0054, 0.0080, 0.0052], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0011, 0.0007], device='cuda:1') 2023-04-26 19:12:43,336 INFO [finetune.py:976] (1/7) Epoch 7, batch 1200, loss[loss=0.1669, simple_loss=0.2494, pruned_loss=0.04218, over 4791.00 frames. ], tot_loss[loss=0.2029, simple_loss=0.2664, pruned_loss=0.06966, over 949878.65 frames. ], batch size: 51, lr: 3.88e-03, grad_scale: 32.0 2023-04-26 19:13:00,096 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=35591.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:13:06,216 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([3.4786, 3.3299, 2.6300, 4.0375, 3.4083, 3.4839, 1.4763, 3.4108], device='cuda:1'), covar=tensor([0.1731, 0.1378, 0.3489, 0.1967, 0.3043, 0.2001, 0.5673, 0.2533], device='cuda:1'), in_proj_covar=tensor([0.0243, 0.0218, 0.0253, 0.0309, 0.0302, 0.0253, 0.0273, 0.0273], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-26 19:13:17,330 INFO [finetune.py:976] (1/7) Epoch 7, batch 1250, loss[loss=0.2609, simple_loss=0.3037, pruned_loss=0.109, over 4193.00 frames. ], tot_loss[loss=0.202, simple_loss=0.2643, pruned_loss=0.06987, over 950257.04 frames. ], batch size: 65, lr: 3.88e-03, grad_scale: 32.0 2023-04-26 19:13:26,240 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.030e+02 1.763e+02 2.030e+02 2.428e+02 4.549e+02, threshold=4.060e+02, percent-clipped=1.0 2023-04-26 19:13:27,683 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=3.62 vs. limit=5.0 2023-04-26 19:13:51,317 INFO [finetune.py:976] (1/7) Epoch 7, batch 1300, loss[loss=0.1982, simple_loss=0.2619, pruned_loss=0.06728, over 4894.00 frames. ], tot_loss[loss=0.1984, simple_loss=0.2609, pruned_loss=0.06799, over 952355.95 frames. ], batch size: 32, lr: 3.88e-03, grad_scale: 32.0 2023-04-26 19:14:04,568 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7309, 2.4887, 1.3154, 1.5206, 2.5532, 1.7064, 1.5423, 1.7023], device='cuda:1'), covar=tensor([0.0546, 0.0327, 0.0328, 0.0565, 0.0220, 0.0550, 0.0533, 0.0601], device='cuda:1'), in_proj_covar=tensor([0.0031, 0.0026, 0.0024, 0.0030, 0.0021, 0.0030, 0.0029, 0.0031], device='cuda:1'), out_proj_covar=tensor([0.0049, 0.0045, 0.0038, 0.0050, 0.0038, 0.0048, 0.0048, 0.0050], device='cuda:1') 2023-04-26 19:14:14,078 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=35701.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 19:14:14,744 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.4874, 3.1056, 0.8377, 1.5253, 2.1286, 1.4620, 4.2425, 2.1793], device='cuda:1'), covar=tensor([0.0630, 0.0976, 0.1069, 0.1314, 0.0620, 0.1068, 0.0211, 0.0633], device='cuda:1'), in_proj_covar=tensor([0.0052, 0.0068, 0.0051, 0.0048, 0.0052, 0.0053, 0.0080, 0.0051], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0011, 0.0007], device='cuda:1') 2023-04-26 19:14:35,192 INFO [finetune.py:976] (1/7) Epoch 7, batch 1350, loss[loss=0.1776, simple_loss=0.2502, pruned_loss=0.05254, over 4926.00 frames. ], tot_loss[loss=0.1997, simple_loss=0.2621, pruned_loss=0.06871, over 954679.02 frames. ], batch size: 38, lr: 3.88e-03, grad_scale: 32.0 2023-04-26 19:14:45,779 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.2953, 1.4622, 1.5763, 2.1273, 2.3455, 1.9750, 1.8986, 1.6864], device='cuda:1'), covar=tensor([0.1686, 0.1825, 0.2090, 0.1762, 0.1326, 0.1949, 0.2128, 0.1927], device='cuda:1'), in_proj_covar=tensor([0.0309, 0.0330, 0.0352, 0.0302, 0.0338, 0.0329, 0.0308, 0.0354], device='cuda:1'), out_proj_covar=tensor([6.6065e-05, 7.0308e-05, 7.6306e-05, 6.2799e-05, 7.1215e-05, 7.0893e-05, 6.6579e-05, 7.6214e-05], device='cuda:1') 2023-04-26 19:14:54,290 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.038e+02 1.639e+02 1.977e+02 2.381e+02 4.004e+02, threshold=3.953e+02, percent-clipped=0.0 2023-04-26 19:14:54,986 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=35730.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:15:09,272 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5107, 1.4018, 1.8754, 1.7975, 1.4114, 1.1471, 1.6196, 1.0041], device='cuda:1'), covar=tensor([0.0728, 0.0974, 0.0480, 0.0836, 0.1010, 0.1383, 0.0763, 0.0948], device='cuda:1'), in_proj_covar=tensor([0.0066, 0.0075, 0.0072, 0.0067, 0.0077, 0.0095, 0.0080, 0.0076], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-26 19:15:17,757 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=35748.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:15:17,840 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.13 vs. limit=5.0 2023-04-26 19:15:40,835 INFO [finetune.py:976] (1/7) Epoch 7, batch 1400, loss[loss=0.2512, simple_loss=0.3112, pruned_loss=0.09567, over 4762.00 frames. ], tot_loss[loss=0.2019, simple_loss=0.2648, pruned_loss=0.06955, over 952580.96 frames. ], batch size: 54, lr: 3.88e-03, grad_scale: 32.0 2023-04-26 19:15:44,429 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=35771.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:16:06,190 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=35796.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:16:06,404 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.29 vs. limit=5.0 2023-04-26 19:16:08,652 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-04-26 19:16:26,223 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=35811.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:16:30,140 INFO [finetune.py:976] (1/7) Epoch 7, batch 1450, loss[loss=0.18, simple_loss=0.2353, pruned_loss=0.06239, over 4414.00 frames. ], tot_loss[loss=0.2024, simple_loss=0.2656, pruned_loss=0.06959, over 951561.66 frames. ], batch size: 19, lr: 3.88e-03, grad_scale: 32.0 2023-04-26 19:16:50,309 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.278e+02 1.770e+02 2.190e+02 2.562e+02 4.335e+02, threshold=4.381e+02, percent-clipped=3.0 2023-04-26 19:16:58,467 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=35832.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:17:10,457 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-04-26 19:17:13,347 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.1792, 2.6503, 1.0092, 1.3752, 1.9633, 1.2334, 3.3758, 1.7334], device='cuda:1'), covar=tensor([0.0638, 0.0571, 0.0798, 0.1305, 0.0494, 0.1068, 0.0315, 0.0652], device='cuda:1'), in_proj_covar=tensor([0.0052, 0.0068, 0.0050, 0.0047, 0.0052, 0.0053, 0.0079, 0.0051], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0007, 0.0008, 0.0008, 0.0011, 0.0007], device='cuda:1') 2023-04-26 19:17:41,157 INFO [finetune.py:976] (1/7) Epoch 7, batch 1500, loss[loss=0.2237, simple_loss=0.2944, pruned_loss=0.07655, over 4813.00 frames. ], tot_loss[loss=0.2038, simple_loss=0.2673, pruned_loss=0.07008, over 953031.95 frames. ], batch size: 47, lr: 3.88e-03, grad_scale: 32.0 2023-04-26 19:17:46,030 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=35872.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:17:55,858 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=35886.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:18:01,447 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-04-26 19:18:31,248 INFO [finetune.py:976] (1/7) Epoch 7, batch 1550, loss[loss=0.1892, simple_loss=0.254, pruned_loss=0.06227, over 4742.00 frames. ], tot_loss[loss=0.2032, simple_loss=0.2671, pruned_loss=0.0697, over 951161.86 frames. ], batch size: 26, lr: 3.88e-03, grad_scale: 32.0 2023-04-26 19:18:51,453 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.224e+02 1.799e+02 2.113e+02 2.516e+02 6.961e+02, threshold=4.225e+02, percent-clipped=2.0 2023-04-26 19:19:06,499 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-04-26 19:19:34,641 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.12 vs. limit=2.0 2023-04-26 19:19:38,741 INFO [finetune.py:976] (1/7) Epoch 7, batch 1600, loss[loss=0.2196, simple_loss=0.2846, pruned_loss=0.07724, over 4898.00 frames. ], tot_loss[loss=0.2021, simple_loss=0.2651, pruned_loss=0.06957, over 951055.47 frames. ], batch size: 32, lr: 3.88e-03, grad_scale: 32.0 2023-04-26 19:19:50,062 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.0645, 1.6262, 1.4292, 1.8860, 1.6028, 1.9888, 1.3822, 3.5769], device='cuda:1'), covar=tensor([0.0711, 0.0757, 0.0831, 0.1172, 0.0672, 0.0533, 0.0787, 0.0170], device='cuda:1'), in_proj_covar=tensor([0.0039, 0.0039, 0.0041, 0.0044, 0.0040, 0.0040, 0.0039, 0.0060], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0014, 0.0013, 0.0013, 0.0013, 0.0017], device='cuda:1') 2023-04-26 19:20:30,768 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=36001.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 19:20:45,229 INFO [finetune.py:976] (1/7) Epoch 7, batch 1650, loss[loss=0.2051, simple_loss=0.2608, pruned_loss=0.07466, over 4931.00 frames. ], tot_loss[loss=0.1992, simple_loss=0.2619, pruned_loss=0.06823, over 952842.90 frames. ], batch size: 38, lr: 3.88e-03, grad_scale: 32.0 2023-04-26 19:20:59,791 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.070e+02 1.778e+02 2.122e+02 2.492e+02 4.114e+02, threshold=4.243e+02, percent-clipped=0.0 2023-04-26 19:21:00,972 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=36030.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:21:26,272 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=36049.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 19:21:28,203 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.3799, 1.6052, 1.6774, 2.2139, 2.3891, 1.9597, 1.9399, 1.8090], device='cuda:1'), covar=tensor([0.1851, 0.2088, 0.2204, 0.1890, 0.1268, 0.2225, 0.2841, 0.2003], device='cuda:1'), in_proj_covar=tensor([0.0310, 0.0331, 0.0354, 0.0303, 0.0340, 0.0330, 0.0309, 0.0356], device='cuda:1'), out_proj_covar=tensor([6.6114e-05, 7.0521e-05, 7.6686e-05, 6.2935e-05, 7.1550e-05, 7.1280e-05, 6.6810e-05, 7.6555e-05], device='cuda:1') 2023-04-26 19:21:48,196 INFO [finetune.py:976] (1/7) Epoch 7, batch 1700, loss[loss=0.22, simple_loss=0.2672, pruned_loss=0.08636, over 4820.00 frames. ], tot_loss[loss=0.1992, simple_loss=0.2611, pruned_loss=0.06866, over 951990.47 frames. ], batch size: 25, lr: 3.88e-03, grad_scale: 32.0 2023-04-26 19:21:56,084 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=36078.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:22:21,090 INFO [finetune.py:976] (1/7) Epoch 7, batch 1750, loss[loss=0.221, simple_loss=0.2943, pruned_loss=0.07379, over 4752.00 frames. ], tot_loss[loss=0.2025, simple_loss=0.2645, pruned_loss=0.07027, over 952829.32 frames. ], batch size: 59, lr: 3.88e-03, grad_scale: 32.0 2023-04-26 19:22:28,801 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=36127.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:22:29,945 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.331e+02 1.866e+02 2.133e+02 2.694e+02 4.644e+02, threshold=4.265e+02, percent-clipped=2.0 2023-04-26 19:22:54,678 INFO [finetune.py:976] (1/7) Epoch 7, batch 1800, loss[loss=0.2305, simple_loss=0.2946, pruned_loss=0.08323, over 4758.00 frames. ], tot_loss[loss=0.2039, simple_loss=0.2664, pruned_loss=0.07068, over 953525.37 frames. ], batch size: 59, lr: 3.88e-03, grad_scale: 32.0 2023-04-26 19:22:55,364 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=36167.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:22:56,618 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4884, 3.2892, 0.9701, 1.8747, 1.8656, 2.3517, 1.9137, 1.0270], device='cuda:1'), covar=tensor([0.1387, 0.1104, 0.1951, 0.1306, 0.1114, 0.1026, 0.1586, 0.1917], device='cuda:1'), in_proj_covar=tensor([0.0121, 0.0258, 0.0145, 0.0126, 0.0137, 0.0158, 0.0122, 0.0125], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-26 19:22:58,538 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-04-26 19:23:05,454 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=36181.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:23:08,499 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=36186.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:23:41,869 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.1216, 2.1264, 4.5675, 4.2725, 4.0581, 4.3897, 4.2636, 4.0665], device='cuda:1'), covar=tensor([0.6113, 0.4703, 0.0986, 0.1615, 0.1059, 0.1517, 0.1151, 0.1238], device='cuda:1'), in_proj_covar=tensor([0.0310, 0.0304, 0.0409, 0.0414, 0.0350, 0.0407, 0.0317, 0.0365], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-26 19:23:44,253 INFO [finetune.py:976] (1/7) Epoch 7, batch 1850, loss[loss=0.2009, simple_loss=0.2693, pruned_loss=0.06624, over 4840.00 frames. ], tot_loss[loss=0.2055, simple_loss=0.2682, pruned_loss=0.07138, over 952148.56 frames. ], batch size: 49, lr: 3.88e-03, grad_scale: 32.0 2023-04-26 19:24:03,616 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.013e+02 1.720e+02 2.075e+02 2.567e+02 3.756e+02, threshold=4.150e+02, percent-clipped=0.0 2023-04-26 19:24:07,670 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=36234.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:24:19,060 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=36242.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:24:44,996 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-04-26 19:24:45,930 INFO [finetune.py:976] (1/7) Epoch 7, batch 1900, loss[loss=0.2577, simple_loss=0.3085, pruned_loss=0.1035, over 4262.00 frames. ], tot_loss[loss=0.2054, simple_loss=0.2689, pruned_loss=0.071, over 952222.42 frames. ], batch size: 65, lr: 3.87e-03, grad_scale: 32.0 2023-04-26 19:24:52,267 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-04-26 19:24:52,868 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-04-26 19:25:19,583 INFO [finetune.py:976] (1/7) Epoch 7, batch 1950, loss[loss=0.1575, simple_loss=0.2159, pruned_loss=0.0495, over 4724.00 frames. ], tot_loss[loss=0.203, simple_loss=0.2664, pruned_loss=0.06986, over 950970.80 frames. ], batch size: 23, lr: 3.87e-03, grad_scale: 32.0 2023-04-26 19:25:27,367 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.259e+02 1.677e+02 1.937e+02 2.378e+02 4.833e+02, threshold=3.873e+02, percent-clipped=2.0 2023-04-26 19:25:31,077 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-04-26 19:26:09,641 INFO [finetune.py:976] (1/7) Epoch 7, batch 2000, loss[loss=0.1919, simple_loss=0.2524, pruned_loss=0.06568, over 4850.00 frames. ], tot_loss[loss=0.2014, simple_loss=0.2642, pruned_loss=0.06932, over 952105.64 frames. ], batch size: 44, lr: 3.87e-03, grad_scale: 32.0 2023-04-26 19:27:00,245 INFO [finetune.py:976] (1/7) Epoch 7, batch 2050, loss[loss=0.1844, simple_loss=0.2462, pruned_loss=0.06125, over 4938.00 frames. ], tot_loss[loss=0.1997, simple_loss=0.262, pruned_loss=0.06867, over 953790.80 frames. ], batch size: 38, lr: 3.87e-03, grad_scale: 32.0 2023-04-26 19:27:03,565 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-04-26 19:27:06,699 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.74 vs. limit=5.0 2023-04-26 19:27:07,120 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=36427.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:27:08,260 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.074e+02 1.550e+02 1.908e+02 2.415e+02 5.206e+02, threshold=3.816e+02, percent-clipped=3.0 2023-04-26 19:27:37,474 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.20 vs. limit=5.0 2023-04-26 19:27:43,883 INFO [finetune.py:976] (1/7) Epoch 7, batch 2100, loss[loss=0.1616, simple_loss=0.2388, pruned_loss=0.04219, over 4907.00 frames. ], tot_loss[loss=0.1985, simple_loss=0.2611, pruned_loss=0.06798, over 954473.84 frames. ], batch size: 35, lr: 3.87e-03, grad_scale: 32.0 2023-04-26 19:27:45,093 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=36467.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:27:49,906 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=36475.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:28:17,168 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=36515.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:28:17,708 INFO [finetune.py:976] (1/7) Epoch 7, batch 2150, loss[loss=0.2669, simple_loss=0.3303, pruned_loss=0.1017, over 4817.00 frames. ], tot_loss[loss=0.2023, simple_loss=0.2654, pruned_loss=0.06958, over 953193.39 frames. ], batch size: 40, lr: 3.87e-03, grad_scale: 32.0 2023-04-26 19:28:25,936 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.260e+02 1.822e+02 2.196e+02 2.668e+02 7.231e+02, threshold=4.392e+02, percent-clipped=1.0 2023-04-26 19:28:30,877 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=36537.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:28:35,162 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([4.7680, 4.7272, 3.3705, 5.5271, 4.8614, 4.7990, 2.2771, 4.7238], device='cuda:1'), covar=tensor([0.1558, 0.0909, 0.2604, 0.0766, 0.2193, 0.1626, 0.5362, 0.1690], device='cuda:1'), in_proj_covar=tensor([0.0245, 0.0219, 0.0253, 0.0313, 0.0305, 0.0254, 0.0276, 0.0275], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-26 19:28:51,287 INFO [finetune.py:976] (1/7) Epoch 7, batch 2200, loss[loss=0.2102, simple_loss=0.2792, pruned_loss=0.07065, over 4757.00 frames. ], tot_loss[loss=0.2033, simple_loss=0.2666, pruned_loss=0.07001, over 952935.68 frames. ], batch size: 28, lr: 3.87e-03, grad_scale: 64.0 2023-04-26 19:28:58,645 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.55 vs. limit=5.0 2023-04-26 19:29:27,978 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=36605.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:29:30,333 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=36608.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:29:30,353 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9109, 2.6182, 1.8933, 1.8180, 1.3922, 1.3735, 2.0262, 1.3853], device='cuda:1'), covar=tensor([0.1852, 0.1774, 0.1699, 0.2200, 0.2768, 0.2171, 0.1276, 0.2252], device='cuda:1'), in_proj_covar=tensor([0.0194, 0.0218, 0.0173, 0.0204, 0.0209, 0.0186, 0.0165, 0.0190], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-26 19:29:38,205 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=36612.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:29:41,092 INFO [finetune.py:976] (1/7) Epoch 7, batch 2250, loss[loss=0.2129, simple_loss=0.2738, pruned_loss=0.07598, over 4814.00 frames. ], tot_loss[loss=0.2049, simple_loss=0.2682, pruned_loss=0.07087, over 952896.19 frames. ], batch size: 39, lr: 3.87e-03, grad_scale: 64.0 2023-04-26 19:30:00,404 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.228e+02 1.830e+02 2.114e+02 2.598e+02 4.357e+02, threshold=4.229e+02, percent-clipped=0.0 2023-04-26 19:30:10,773 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([4.8633, 4.8499, 3.3549, 5.5777, 4.8801, 4.8592, 2.4810, 4.7802], device='cuda:1'), covar=tensor([0.1522, 0.0810, 0.2653, 0.0793, 0.2958, 0.1445, 0.4747, 0.1738], device='cuda:1'), in_proj_covar=tensor([0.0244, 0.0218, 0.0252, 0.0310, 0.0303, 0.0252, 0.0273, 0.0272], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-26 19:30:12,677 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([0.3289, 1.3973, 1.4596, 1.0620, 1.4507, 1.1348, 1.7503, 1.4420], device='cuda:1'), covar=tensor([0.4005, 0.1897, 0.6327, 0.3001, 0.1722, 0.2627, 0.1908, 0.5025], device='cuda:1'), in_proj_covar=tensor([0.0351, 0.0356, 0.0441, 0.0368, 0.0394, 0.0389, 0.0391, 0.0425], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 19:30:25,376 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0730, 2.5440, 1.1412, 1.3988, 2.0517, 1.2453, 3.4215, 1.7052], device='cuda:1'), covar=tensor([0.0693, 0.0722, 0.0948, 0.1399, 0.0504, 0.1055, 0.0341, 0.0674], device='cuda:1'), in_proj_covar=tensor([0.0053, 0.0069, 0.0051, 0.0048, 0.0052, 0.0053, 0.0080, 0.0051], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0011, 0.0007], device='cuda:1') 2023-04-26 19:30:40,442 INFO [finetune.py:976] (1/7) Epoch 7, batch 2300, loss[loss=0.1854, simple_loss=0.2616, pruned_loss=0.05459, over 4784.00 frames. ], tot_loss[loss=0.2055, simple_loss=0.2693, pruned_loss=0.07088, over 950884.02 frames. ], batch size: 51, lr: 3.87e-03, grad_scale: 64.0 2023-04-26 19:30:40,574 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=36666.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:30:42,392 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=36669.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:30:42,411 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0589, 1.2525, 1.4947, 1.6571, 1.5784, 1.6877, 1.6201, 1.6089], device='cuda:1'), covar=tensor([0.5931, 0.7128, 0.6425, 0.5881, 0.7238, 1.0464, 0.7124, 0.6500], device='cuda:1'), in_proj_covar=tensor([0.0323, 0.0394, 0.0319, 0.0330, 0.0347, 0.0410, 0.0374, 0.0331], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-26 19:30:45,355 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=36673.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:31:00,434 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5763, 2.0592, 1.7683, 1.8946, 1.5277, 1.7358, 1.7862, 1.4577], device='cuda:1'), covar=tensor([0.1951, 0.1216, 0.0839, 0.1178, 0.3316, 0.0981, 0.1602, 0.2229], device='cuda:1'), in_proj_covar=tensor([0.0304, 0.0322, 0.0234, 0.0295, 0.0316, 0.0274, 0.0263, 0.0285], device='cuda:1'), out_proj_covar=tensor([1.2335e-04, 1.3026e-04, 9.4845e-05, 1.1845e-04, 1.3010e-04, 1.1063e-04, 1.0774e-04, 1.1474e-04], device='cuda:1') 2023-04-26 19:31:02,264 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=36700.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 19:31:23,676 INFO [finetune.py:976] (1/7) Epoch 7, batch 2350, loss[loss=0.2083, simple_loss=0.267, pruned_loss=0.07483, over 4751.00 frames. ], tot_loss[loss=0.2043, simple_loss=0.2675, pruned_loss=0.07051, over 951274.88 frames. ], batch size: 26, lr: 3.87e-03, grad_scale: 64.0 2023-04-26 19:31:38,225 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.607e+01 1.814e+02 2.081e+02 2.499e+02 5.543e+02, threshold=4.163e+02, percent-clipped=2.0 2023-04-26 19:32:21,172 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=36761.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 19:32:30,354 INFO [finetune.py:976] (1/7) Epoch 7, batch 2400, loss[loss=0.1732, simple_loss=0.2384, pruned_loss=0.05399, over 4815.00 frames. ], tot_loss[loss=0.2015, simple_loss=0.2644, pruned_loss=0.06931, over 949748.24 frames. ], batch size: 41, lr: 3.87e-03, grad_scale: 64.0 2023-04-26 19:32:47,037 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.2060, 2.5823, 0.8338, 1.4968, 1.5006, 1.9373, 1.5881, 0.8182], device='cuda:1'), covar=tensor([0.1421, 0.1219, 0.1694, 0.1394, 0.1180, 0.0973, 0.1608, 0.1711], device='cuda:1'), in_proj_covar=tensor([0.0119, 0.0254, 0.0143, 0.0125, 0.0135, 0.0156, 0.0121, 0.0124], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-26 19:32:56,378 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=36805.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 19:33:03,827 INFO [finetune.py:976] (1/7) Epoch 7, batch 2450, loss[loss=0.1647, simple_loss=0.2251, pruned_loss=0.05211, over 4805.00 frames. ], tot_loss[loss=0.2001, simple_loss=0.2626, pruned_loss=0.06883, over 950858.92 frames. ], batch size: 25, lr: 3.87e-03, grad_scale: 64.0 2023-04-26 19:33:12,663 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.003e+02 1.628e+02 1.856e+02 2.205e+02 3.828e+02, threshold=3.712e+02, percent-clipped=0.0 2023-04-26 19:33:18,554 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=36837.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:33:25,262 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.73 vs. limit=2.0 2023-04-26 19:33:37,053 INFO [finetune.py:976] (1/7) Epoch 7, batch 2500, loss[loss=0.2064, simple_loss=0.2817, pruned_loss=0.06562, over 4796.00 frames. ], tot_loss[loss=0.2005, simple_loss=0.2633, pruned_loss=0.06884, over 950604.64 frames. ], batch size: 45, lr: 3.87e-03, grad_scale: 64.0 2023-04-26 19:33:37,185 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=36866.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 19:33:51,058 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=36885.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:34:04,990 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8036, 0.9995, 1.3035, 1.5001, 1.4049, 1.5562, 1.4214, 1.3794], device='cuda:1'), covar=tensor([0.4778, 0.7062, 0.5980, 0.5549, 0.7288, 1.0369, 0.6864, 0.6344], device='cuda:1'), in_proj_covar=tensor([0.0321, 0.0393, 0.0318, 0.0328, 0.0346, 0.0409, 0.0372, 0.0329], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-26 19:34:07,396 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=36911.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:34:10,362 INFO [finetune.py:976] (1/7) Epoch 7, batch 2550, loss[loss=0.2322, simple_loss=0.2919, pruned_loss=0.08623, over 4747.00 frames. ], tot_loss[loss=0.2022, simple_loss=0.2657, pruned_loss=0.06931, over 952462.77 frames. ], batch size: 27, lr: 3.87e-03, grad_scale: 64.0 2023-04-26 19:34:20,132 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.245e+02 1.780e+02 2.108e+02 2.520e+02 3.900e+02, threshold=4.217e+02, percent-clipped=2.0 2023-04-26 19:34:45,088 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=36958.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:34:52,519 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=36961.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:34:54,325 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=36964.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:34:55,485 INFO [finetune.py:976] (1/7) Epoch 7, batch 2600, loss[loss=0.2169, simple_loss=0.271, pruned_loss=0.08144, over 4774.00 frames. ], tot_loss[loss=0.2038, simple_loss=0.2676, pruned_loss=0.07005, over 953399.81 frames. ], batch size: 27, lr: 3.87e-03, grad_scale: 64.0 2023-04-26 19:35:00,962 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=36968.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:35:04,488 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=36972.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:35:48,170 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=37004.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:36:01,123 INFO [finetune.py:976] (1/7) Epoch 7, batch 2650, loss[loss=0.2123, simple_loss=0.2844, pruned_loss=0.0701, over 4788.00 frames. ], tot_loss[loss=0.2045, simple_loss=0.2686, pruned_loss=0.07022, over 949766.86 frames. ], batch size: 29, lr: 3.87e-03, grad_scale: 64.0 2023-04-26 19:36:03,058 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=37019.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:36:04,915 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.3503, 1.6788, 1.6128, 1.8503, 1.6588, 1.8043, 1.7750, 1.7053], device='cuda:1'), covar=tensor([0.5671, 0.8299, 0.7767, 0.7132, 0.8432, 1.1936, 0.8367, 0.8044], device='cuda:1'), in_proj_covar=tensor([0.0320, 0.0390, 0.0316, 0.0326, 0.0343, 0.0406, 0.0370, 0.0328], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-26 19:36:09,955 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.140e+02 1.731e+02 2.035e+02 2.586e+02 4.493e+02, threshold=4.069e+02, percent-clipped=1.0 2023-04-26 19:36:28,799 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=37056.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 19:36:34,305 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=37065.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:36:34,793 INFO [finetune.py:976] (1/7) Epoch 7, batch 2700, loss[loss=0.199, simple_loss=0.2561, pruned_loss=0.0709, over 4826.00 frames. ], tot_loss[loss=0.2029, simple_loss=0.2668, pruned_loss=0.0695, over 949902.88 frames. ], batch size: 30, lr: 3.87e-03, grad_scale: 64.0 2023-04-26 19:36:38,624 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4954, 1.8273, 0.8967, 1.2706, 1.6718, 1.3878, 1.3265, 1.3549], device='cuda:1'), covar=tensor([0.0643, 0.0315, 0.0425, 0.0611, 0.0328, 0.0676, 0.0643, 0.0655], device='cuda:1'), in_proj_covar=tensor([0.0031, 0.0026, 0.0024, 0.0030, 0.0021, 0.0030, 0.0029, 0.0030], device='cuda:1'), out_proj_covar=tensor([0.0049, 0.0044, 0.0038, 0.0050, 0.0038, 0.0048, 0.0048, 0.0050], device='cuda:1') 2023-04-26 19:36:46,703 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5268, 1.4318, 0.5335, 1.2258, 1.4609, 1.4084, 1.2821, 1.3159], device='cuda:1'), covar=tensor([0.0581, 0.0411, 0.0499, 0.0619, 0.0330, 0.0635, 0.0579, 0.0664], device='cuda:1'), in_proj_covar=tensor([0.0031, 0.0026, 0.0024, 0.0030, 0.0021, 0.0030, 0.0030, 0.0030], device='cuda:1'), out_proj_covar=tensor([0.0049, 0.0044, 0.0038, 0.0050, 0.0038, 0.0048, 0.0048, 0.0050], device='cuda:1') 2023-04-26 19:37:10,773 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.1519, 1.6930, 1.7778, 2.1657, 1.8108, 2.0710, 1.5135, 4.5340], device='cuda:1'), covar=tensor([0.0653, 0.0803, 0.0780, 0.1196, 0.0661, 0.0539, 0.0736, 0.0125], device='cuda:1'), in_proj_covar=tensor([0.0039, 0.0039, 0.0041, 0.0044, 0.0041, 0.0040, 0.0039, 0.0060], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0017], device='cuda:1') 2023-04-26 19:37:19,207 INFO [finetune.py:976] (1/7) Epoch 7, batch 2750, loss[loss=0.1784, simple_loss=0.2453, pruned_loss=0.05569, over 4929.00 frames. ], tot_loss[loss=0.1998, simple_loss=0.2634, pruned_loss=0.06815, over 950444.67 frames. ], batch size: 33, lr: 3.87e-03, grad_scale: 64.0 2023-04-26 19:37:29,354 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8494, 1.6797, 1.9815, 2.2810, 2.2503, 1.9040, 1.4847, 1.8942], device='cuda:1'), covar=tensor([0.0924, 0.1213, 0.0593, 0.0595, 0.0684, 0.0905, 0.0988, 0.0711], device='cuda:1'), in_proj_covar=tensor([0.0201, 0.0207, 0.0182, 0.0179, 0.0181, 0.0194, 0.0164, 0.0190], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 19:37:32,274 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.030e+02 1.627e+02 1.918e+02 2.292e+02 3.296e+02, threshold=3.835e+02, percent-clipped=0.0 2023-04-26 19:37:45,642 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.9202, 2.0857, 1.8701, 2.1155, 1.8624, 2.1340, 2.0311, 1.9548], device='cuda:1'), covar=tensor([0.5852, 0.9892, 0.9160, 0.7107, 0.9128, 1.1370, 1.0825, 0.9612], device='cuda:1'), in_proj_covar=tensor([0.0322, 0.0392, 0.0318, 0.0329, 0.0345, 0.0409, 0.0372, 0.0330], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-26 19:37:59,226 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4246, 3.5110, 0.9287, 2.0640, 1.9528, 2.4427, 1.9755, 0.9564], device='cuda:1'), covar=tensor([0.1425, 0.0886, 0.1988, 0.1282, 0.1075, 0.1131, 0.1599, 0.2179], device='cuda:1'), in_proj_covar=tensor([0.0119, 0.0256, 0.0144, 0.0126, 0.0136, 0.0157, 0.0122, 0.0125], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-26 19:38:11,744 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=37161.0, num_to_drop=1, layers_to_drop={3} 2023-04-26 19:38:20,449 INFO [finetune.py:976] (1/7) Epoch 7, batch 2800, loss[loss=0.1573, simple_loss=0.228, pruned_loss=0.04328, over 4772.00 frames. ], tot_loss[loss=0.1965, simple_loss=0.26, pruned_loss=0.06652, over 953685.19 frames. ], batch size: 28, lr: 3.87e-03, grad_scale: 64.0 2023-04-26 19:39:00,020 INFO [finetune.py:976] (1/7) Epoch 7, batch 2850, loss[loss=0.1703, simple_loss=0.2399, pruned_loss=0.05036, over 4933.00 frames. ], tot_loss[loss=0.196, simple_loss=0.2589, pruned_loss=0.06651, over 955175.20 frames. ], batch size: 38, lr: 3.87e-03, grad_scale: 32.0 2023-04-26 19:39:08,542 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.213e+02 1.711e+02 1.920e+02 2.384e+02 5.364e+02, threshold=3.840e+02, percent-clipped=2.0 2023-04-26 19:39:10,439 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.8567, 2.8686, 2.2302, 3.2941, 2.9328, 2.8722, 1.2108, 2.7913], device='cuda:1'), covar=tensor([0.2054, 0.1629, 0.3500, 0.2672, 0.4105, 0.2051, 0.5400, 0.2657], device='cuda:1'), in_proj_covar=tensor([0.0243, 0.0217, 0.0250, 0.0307, 0.0301, 0.0251, 0.0273, 0.0270], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-26 19:39:28,467 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7651, 1.7572, 2.0211, 2.3300, 2.2955, 1.9247, 1.4796, 1.9001], device='cuda:1'), covar=tensor([0.0935, 0.1183, 0.0714, 0.0613, 0.0611, 0.0948, 0.0958, 0.0681], device='cuda:1'), in_proj_covar=tensor([0.0202, 0.0207, 0.0183, 0.0179, 0.0181, 0.0195, 0.0164, 0.0190], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 19:39:30,814 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=37261.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:39:32,633 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=37264.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:39:33,740 INFO [finetune.py:976] (1/7) Epoch 7, batch 2900, loss[loss=0.2162, simple_loss=0.2952, pruned_loss=0.06854, over 4921.00 frames. ], tot_loss[loss=0.199, simple_loss=0.2619, pruned_loss=0.06805, over 955003.48 frames. ], batch size: 42, lr: 3.87e-03, grad_scale: 32.0 2023-04-26 19:39:34,411 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=37267.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:39:35,023 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=37268.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:39:35,037 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=37268.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:40:02,584 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=37309.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:40:10,641 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=37312.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:40:11,891 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=37314.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:40:13,061 INFO [finetune.py:976] (1/7) Epoch 7, batch 2950, loss[loss=0.2098, simple_loss=0.2824, pruned_loss=0.06859, over 4788.00 frames. ], tot_loss[loss=0.2007, simple_loss=0.2649, pruned_loss=0.06825, over 955968.78 frames. ], batch size: 51, lr: 3.87e-03, grad_scale: 32.0 2023-04-26 19:40:13,122 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=37316.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:40:22,848 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.1877, 1.4090, 1.4980, 1.6582, 1.5073, 1.6269, 1.5929, 1.5606], device='cuda:1'), covar=tensor([0.5727, 0.8245, 0.7023, 0.6374, 0.7683, 1.1248, 0.8271, 0.7527], device='cuda:1'), in_proj_covar=tensor([0.0322, 0.0392, 0.0318, 0.0328, 0.0344, 0.0410, 0.0372, 0.0329], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-26 19:40:32,651 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=37329.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:40:33,130 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.057e+02 1.889e+02 2.237e+02 2.673e+02 8.962e+02, threshold=4.474e+02, percent-clipped=7.0 2023-04-26 19:41:05,697 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=37356.0, num_to_drop=1, layers_to_drop={2} 2023-04-26 19:41:08,605 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=37360.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:41:17,426 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=37364.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:41:18,550 INFO [finetune.py:976] (1/7) Epoch 7, batch 3000, loss[loss=0.1568, simple_loss=0.2037, pruned_loss=0.055, over 4286.00 frames. ], tot_loss[loss=0.2027, simple_loss=0.2664, pruned_loss=0.06947, over 954939.85 frames. ], batch size: 18, lr: 3.87e-03, grad_scale: 32.0 2023-04-26 19:41:18,550 INFO [finetune.py:1001] (1/7) Computing validation loss 2023-04-26 19:41:30,336 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.1902, 1.7007, 2.0755, 2.3134, 1.9935, 1.6678, 1.0896, 1.7722], device='cuda:1'), covar=tensor([0.4342, 0.4286, 0.1884, 0.2844, 0.3627, 0.3361, 0.5090, 0.3091], device='cuda:1'), in_proj_covar=tensor([0.0283, 0.0256, 0.0219, 0.0325, 0.0217, 0.0231, 0.0239, 0.0191], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-26 19:41:31,748 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.3900, 1.2482, 1.6291, 1.5705, 1.2975, 1.1621, 1.3421, 0.8854], device='cuda:1'), covar=tensor([0.0700, 0.0795, 0.0560, 0.0662, 0.0927, 0.1330, 0.0653, 0.0840], device='cuda:1'), in_proj_covar=tensor([0.0066, 0.0075, 0.0073, 0.0067, 0.0078, 0.0096, 0.0081, 0.0077], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-26 19:41:40,583 INFO [finetune.py:1010] (1/7) Epoch 7, validation: loss=0.1559, simple_loss=0.2289, pruned_loss=0.04148, over 2265189.00 frames. 2023-04-26 19:41:40,583 INFO [finetune.py:1011] (1/7) Maximum memory allocated so far is 6391MB 2023-04-26 19:42:00,280 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6882, 2.2107, 1.8232, 2.1002, 1.6367, 1.8676, 1.8323, 1.4236], device='cuda:1'), covar=tensor([0.2318, 0.1636, 0.1045, 0.1380, 0.3409, 0.1390, 0.2122, 0.2829], device='cuda:1'), in_proj_covar=tensor([0.0308, 0.0326, 0.0236, 0.0298, 0.0320, 0.0277, 0.0266, 0.0289], device='cuda:1'), out_proj_covar=tensor([1.2498e-04, 1.3198e-04, 9.5639e-05, 1.1971e-04, 1.3171e-04, 1.1192e-04, 1.0917e-04, 1.1630e-04], device='cuda:1') 2023-04-26 19:42:11,839 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([3.1675, 2.3056, 2.2212, 2.3518, 2.1312, 2.3761, 2.4402, 2.3128], device='cuda:1'), covar=tensor([0.5484, 0.8637, 0.7070, 0.6201, 0.7638, 0.9896, 0.8586, 0.7132], device='cuda:1'), in_proj_covar=tensor([0.0322, 0.0391, 0.0317, 0.0327, 0.0343, 0.0408, 0.0370, 0.0328], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-26 19:42:12,570 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([0.6449, 1.5834, 1.5997, 1.2781, 1.7683, 1.3311, 2.3018, 1.4038], device='cuda:1'), covar=tensor([0.4318, 0.2017, 0.5672, 0.3782, 0.2113, 0.3000, 0.1651, 0.5436], device='cuda:1'), in_proj_covar=tensor([0.0349, 0.0356, 0.0439, 0.0368, 0.0393, 0.0388, 0.0388, 0.0424], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 19:42:15,398 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=37404.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 19:42:16,076 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.0779, 0.7730, 0.8965, 0.7707, 1.2346, 0.9795, 0.8522, 0.9122], device='cuda:1'), covar=tensor([0.1648, 0.1474, 0.2096, 0.1505, 0.0893, 0.1418, 0.1929, 0.2131], device='cuda:1'), in_proj_covar=tensor([0.0308, 0.0329, 0.0354, 0.0303, 0.0341, 0.0328, 0.0308, 0.0353], device='cuda:1'), out_proj_covar=tensor([6.5722e-05, 7.0091e-05, 7.6668e-05, 6.2969e-05, 7.1792e-05, 7.0875e-05, 6.6442e-05, 7.5878e-05], device='cuda:1') 2023-04-26 19:42:23,694 INFO [finetune.py:976] (1/7) Epoch 7, batch 3050, loss[loss=0.1778, simple_loss=0.2527, pruned_loss=0.05142, over 4731.00 frames. ], tot_loss[loss=0.2042, simple_loss=0.2681, pruned_loss=0.07009, over 956118.04 frames. ], batch size: 59, lr: 3.87e-03, grad_scale: 32.0 2023-04-26 19:42:30,260 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=37425.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:42:32,444 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7208, 2.3507, 1.9521, 2.1826, 1.6721, 1.8996, 1.9420, 1.4365], device='cuda:1'), covar=tensor([0.2523, 0.1304, 0.0949, 0.1434, 0.3480, 0.1313, 0.2176, 0.3130], device='cuda:1'), in_proj_covar=tensor([0.0306, 0.0324, 0.0235, 0.0297, 0.0318, 0.0275, 0.0265, 0.0288], device='cuda:1'), out_proj_covar=tensor([1.2427e-04, 1.3118e-04, 9.5201e-05, 1.1894e-04, 1.3095e-04, 1.1115e-04, 1.0846e-04, 1.1573e-04], device='cuda:1') 2023-04-26 19:42:34,151 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 8.257e+01 1.766e+02 2.163e+02 2.758e+02 4.245e+02, threshold=4.325e+02, percent-clipped=0.0 2023-04-26 19:42:54,524 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=37461.0, num_to_drop=1, layers_to_drop={2} 2023-04-26 19:42:57,505 INFO [finetune.py:976] (1/7) Epoch 7, batch 3100, loss[loss=0.1914, simple_loss=0.2526, pruned_loss=0.06508, over 4832.00 frames. ], tot_loss[loss=0.2029, simple_loss=0.267, pruned_loss=0.06941, over 956481.90 frames. ], batch size: 25, lr: 3.87e-03, grad_scale: 32.0 2023-04-26 19:43:05,721 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([3.7707, 3.7745, 2.7602, 4.4114, 3.8635, 3.8544, 1.6129, 3.7245], device='cuda:1'), covar=tensor([0.1428, 0.1125, 0.2646, 0.1494, 0.2528, 0.1652, 0.5523, 0.2179], device='cuda:1'), in_proj_covar=tensor([0.0244, 0.0219, 0.0251, 0.0310, 0.0304, 0.0253, 0.0274, 0.0272], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-26 19:43:14,106 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0641, 2.3774, 1.0259, 1.3141, 1.7972, 1.1705, 2.9950, 1.4717], device='cuda:1'), covar=tensor([0.0654, 0.0590, 0.0735, 0.1235, 0.0483, 0.1025, 0.0238, 0.0697], device='cuda:1'), in_proj_covar=tensor([0.0052, 0.0068, 0.0051, 0.0048, 0.0052, 0.0053, 0.0079, 0.0051], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0007], device='cuda:1') 2023-04-26 19:43:16,628 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.58 vs. limit=2.0 2023-04-26 19:43:26,077 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=37509.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 19:43:31,233 INFO [finetune.py:976] (1/7) Epoch 7, batch 3150, loss[loss=0.2231, simple_loss=0.2694, pruned_loss=0.08839, over 4810.00 frames. ], tot_loss[loss=0.2011, simple_loss=0.2649, pruned_loss=0.06862, over 956330.96 frames. ], batch size: 41, lr: 3.87e-03, grad_scale: 32.0 2023-04-26 19:43:52,227 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.195e+02 1.638e+02 2.016e+02 2.443e+02 5.326e+02, threshold=4.032e+02, percent-clipped=1.0 2023-04-26 19:44:25,127 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([0.8420, 1.9078, 1.8302, 1.5836, 2.0962, 1.7296, 2.6482, 1.5063], device='cuda:1'), covar=tensor([0.3468, 0.1557, 0.4256, 0.2618, 0.1518, 0.2182, 0.1144, 0.4816], device='cuda:1'), in_proj_covar=tensor([0.0343, 0.0350, 0.0431, 0.0362, 0.0387, 0.0381, 0.0383, 0.0417], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 19:44:36,861 INFO [finetune.py:976] (1/7) Epoch 7, batch 3200, loss[loss=0.225, simple_loss=0.2744, pruned_loss=0.0878, over 4875.00 frames. ], tot_loss[loss=0.1976, simple_loss=0.2612, pruned_loss=0.06703, over 957751.97 frames. ], batch size: 31, lr: 3.87e-03, grad_scale: 32.0 2023-04-26 19:44:37,556 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=37567.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:45:07,998 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.37 vs. limit=5.0 2023-04-26 19:45:33,718 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5323, 0.6921, 1.3168, 1.8413, 1.5614, 1.3789, 1.3730, 1.4316], device='cuda:1'), covar=tensor([0.8657, 1.1289, 1.2625, 1.3147, 1.0709, 1.4524, 1.4214, 1.1769], device='cuda:1'), in_proj_covar=tensor([0.0413, 0.0433, 0.0518, 0.0538, 0.0441, 0.0461, 0.0472, 0.0470], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 19:45:42,214 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=37614.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:45:43,284 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=37615.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:45:43,827 INFO [finetune.py:976] (1/7) Epoch 7, batch 3250, loss[loss=0.1814, simple_loss=0.2447, pruned_loss=0.059, over 4825.00 frames. ], tot_loss[loss=0.1973, simple_loss=0.2606, pruned_loss=0.06696, over 956793.30 frames. ], batch size: 33, lr: 3.87e-03, grad_scale: 32.0 2023-04-26 19:45:55,384 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=37624.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:46:03,765 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.639e+01 1.714e+02 1.975e+02 2.453e+02 9.656e+02, threshold=3.950e+02, percent-clipped=3.0 2023-04-26 19:46:41,392 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=37660.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:46:42,585 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=37662.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:46:45,458 INFO [finetune.py:976] (1/7) Epoch 7, batch 3300, loss[loss=0.2277, simple_loss=0.289, pruned_loss=0.08325, over 4910.00 frames. ], tot_loss[loss=0.2005, simple_loss=0.2643, pruned_loss=0.06832, over 956755.97 frames. ], batch size: 43, lr: 3.87e-03, grad_scale: 32.0 2023-04-26 19:47:48,146 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=37708.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:47:57,368 INFO [finetune.py:976] (1/7) Epoch 7, batch 3350, loss[loss=0.2322, simple_loss=0.285, pruned_loss=0.08969, over 4838.00 frames. ], tot_loss[loss=0.2009, simple_loss=0.2652, pruned_loss=0.06829, over 955612.51 frames. ], batch size: 49, lr: 3.87e-03, grad_scale: 32.0 2023-04-26 19:47:59,883 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=37720.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:48:01,188 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.45 vs. limit=5.0 2023-04-26 19:48:12,882 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.265e+02 1.789e+02 2.091e+02 2.525e+02 4.436e+02, threshold=4.182e+02, percent-clipped=1.0 2023-04-26 19:48:22,436 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6282, 2.3032, 1.0035, 1.3424, 2.3439, 1.5101, 1.4067, 1.5850], device='cuda:1'), covar=tensor([0.0557, 0.0367, 0.0379, 0.0616, 0.0249, 0.0567, 0.0575, 0.0651], device='cuda:1'), in_proj_covar=tensor([0.0030, 0.0026, 0.0024, 0.0030, 0.0021, 0.0030, 0.0029, 0.0030], device='cuda:1'), out_proj_covar=tensor([0.0049, 0.0044, 0.0038, 0.0050, 0.0037, 0.0048, 0.0048, 0.0050], device='cuda:1') 2023-04-26 19:48:36,820 INFO [finetune.py:976] (1/7) Epoch 7, batch 3400, loss[loss=0.1955, simple_loss=0.27, pruned_loss=0.06046, over 4810.00 frames. ], tot_loss[loss=0.2016, simple_loss=0.2661, pruned_loss=0.0685, over 955410.37 frames. ], batch size: 25, lr: 3.87e-03, grad_scale: 32.0 2023-04-26 19:49:36,097 INFO [finetune.py:976] (1/7) Epoch 7, batch 3450, loss[loss=0.1913, simple_loss=0.2554, pruned_loss=0.06366, over 4867.00 frames. ], tot_loss[loss=0.2017, simple_loss=0.2666, pruned_loss=0.06836, over 955247.87 frames. ], batch size: 31, lr: 3.87e-03, grad_scale: 32.0 2023-04-26 19:49:55,704 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.199e+01 1.617e+02 2.034e+02 2.503e+02 3.981e+02, threshold=4.069e+02, percent-clipped=0.0 2023-04-26 19:50:37,107 INFO [finetune.py:976] (1/7) Epoch 7, batch 3500, loss[loss=0.1539, simple_loss=0.2185, pruned_loss=0.04465, over 4720.00 frames. ], tot_loss[loss=0.2005, simple_loss=0.2644, pruned_loss=0.06833, over 953685.72 frames. ], batch size: 54, lr: 3.86e-03, grad_scale: 32.0 2023-04-26 19:50:53,292 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8470, 1.3865, 1.6606, 1.6760, 1.6224, 1.3589, 0.7287, 1.3009], device='cuda:1'), covar=tensor([0.4373, 0.4244, 0.1982, 0.2975, 0.3350, 0.3270, 0.5230, 0.3014], device='cuda:1'), in_proj_covar=tensor([0.0282, 0.0255, 0.0219, 0.0325, 0.0218, 0.0231, 0.0239, 0.0190], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-26 19:51:17,010 INFO [finetune.py:976] (1/7) Epoch 7, batch 3550, loss[loss=0.2044, simple_loss=0.2643, pruned_loss=0.07223, over 4695.00 frames. ], tot_loss[loss=0.1968, simple_loss=0.2606, pruned_loss=0.06649, over 955218.71 frames. ], batch size: 23, lr: 3.86e-03, grad_scale: 32.0 2023-04-26 19:51:27,503 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=37924.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:51:36,395 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.005e+02 1.609e+02 1.946e+02 2.387e+02 4.801e+02, threshold=3.892e+02, percent-clipped=2.0 2023-04-26 19:51:53,750 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-04-26 19:52:01,352 INFO [finetune.py:976] (1/7) Epoch 7, batch 3600, loss[loss=0.2038, simple_loss=0.2718, pruned_loss=0.06793, over 4932.00 frames. ], tot_loss[loss=0.1953, simple_loss=0.2586, pruned_loss=0.06599, over 956576.14 frames. ], batch size: 33, lr: 3.86e-03, grad_scale: 32.0 2023-04-26 19:52:05,026 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=37972.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:52:09,404 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=37979.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:52:35,872 INFO [finetune.py:976] (1/7) Epoch 7, batch 3650, loss[loss=0.2548, simple_loss=0.3276, pruned_loss=0.09099, over 4841.00 frames. ], tot_loss[loss=0.1983, simple_loss=0.2615, pruned_loss=0.06756, over 956892.58 frames. ], batch size: 49, lr: 3.86e-03, grad_scale: 32.0 2023-04-26 19:52:38,384 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=38020.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:52:44,352 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.213e+02 1.731e+02 1.989e+02 2.580e+02 1.075e+03, threshold=3.977e+02, percent-clipped=3.0 2023-04-26 19:52:49,796 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.3279, 2.9605, 0.8972, 1.6276, 1.6383, 2.0840, 1.7155, 0.8778], device='cuda:1'), covar=tensor([0.1444, 0.0913, 0.1969, 0.1360, 0.1256, 0.1069, 0.1626, 0.2147], device='cuda:1'), in_proj_covar=tensor([0.0120, 0.0256, 0.0144, 0.0126, 0.0137, 0.0157, 0.0121, 0.0124], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-26 19:52:51,026 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=38040.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:53:08,987 INFO [finetune.py:976] (1/7) Epoch 7, batch 3700, loss[loss=0.2338, simple_loss=0.3039, pruned_loss=0.08185, over 4133.00 frames. ], tot_loss[loss=0.2025, simple_loss=0.2657, pruned_loss=0.06964, over 954382.67 frames. ], batch size: 65, lr: 3.86e-03, grad_scale: 32.0 2023-04-26 19:53:10,241 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=38068.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:53:41,674 INFO [finetune.py:976] (1/7) Epoch 7, batch 3750, loss[loss=0.1598, simple_loss=0.2258, pruned_loss=0.04685, over 4782.00 frames. ], tot_loss[loss=0.2042, simple_loss=0.2678, pruned_loss=0.07028, over 956970.44 frames. ], batch size: 26, lr: 3.86e-03, grad_scale: 32.0 2023-04-26 19:53:50,668 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.250e+02 1.891e+02 2.141e+02 2.716e+02 4.079e+02, threshold=4.281e+02, percent-clipped=1.0 2023-04-26 19:54:31,497 INFO [finetune.py:976] (1/7) Epoch 7, batch 3800, loss[loss=0.1335, simple_loss=0.1994, pruned_loss=0.0338, over 4734.00 frames. ], tot_loss[loss=0.2049, simple_loss=0.2686, pruned_loss=0.07062, over 956272.29 frames. ], batch size: 23, lr: 3.86e-03, grad_scale: 32.0 2023-04-26 19:54:31,622 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([0.9295, 2.0038, 1.7102, 1.6533, 2.0916, 1.6418, 2.5398, 1.5044], device='cuda:1'), covar=tensor([0.3636, 0.1513, 0.5178, 0.2857, 0.1670, 0.2414, 0.1233, 0.4484], device='cuda:1'), in_proj_covar=tensor([0.0347, 0.0352, 0.0433, 0.0362, 0.0389, 0.0383, 0.0383, 0.0419], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 19:54:50,071 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([3.6020, 3.5910, 2.6445, 4.2089, 3.4995, 3.6524, 1.6278, 3.5491], device='cuda:1'), covar=tensor([0.1813, 0.1245, 0.3791, 0.1610, 0.2931, 0.1882, 0.5830, 0.2506], device='cuda:1'), in_proj_covar=tensor([0.0246, 0.0220, 0.0254, 0.0311, 0.0305, 0.0255, 0.0276, 0.0274], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-26 19:54:55,061 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6015, 1.6185, 1.9094, 1.9232, 1.6531, 1.3067, 1.7116, 1.0931], device='cuda:1'), covar=tensor([0.0945, 0.0781, 0.0599, 0.0931, 0.0798, 0.1190, 0.0738, 0.1075], device='cuda:1'), in_proj_covar=tensor([0.0065, 0.0074, 0.0072, 0.0066, 0.0076, 0.0095, 0.0080, 0.0075], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-26 19:55:05,503 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0201, 2.9928, 2.4005, 2.4967, 2.0514, 2.3325, 2.4690, 1.9305], device='cuda:1'), covar=tensor([0.2510, 0.1374, 0.1005, 0.1651, 0.3410, 0.1424, 0.2259, 0.3226], device='cuda:1'), in_proj_covar=tensor([0.0304, 0.0323, 0.0235, 0.0295, 0.0316, 0.0275, 0.0263, 0.0288], device='cuda:1'), out_proj_covar=tensor([1.2325e-04, 1.3053e-04, 9.4949e-05, 1.1824e-04, 1.3001e-04, 1.1103e-04, 1.0799e-04, 1.1574e-04], device='cuda:1') 2023-04-26 19:55:36,932 INFO [finetune.py:976] (1/7) Epoch 7, batch 3850, loss[loss=0.2359, simple_loss=0.2855, pruned_loss=0.09318, over 4915.00 frames. ], tot_loss[loss=0.2033, simple_loss=0.2669, pruned_loss=0.06984, over 955120.49 frames. ], batch size: 46, lr: 3.86e-03, grad_scale: 32.0 2023-04-26 19:55:56,086 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.094e+02 1.647e+02 1.925e+02 2.333e+02 3.999e+02, threshold=3.850e+02, percent-clipped=0.0 2023-04-26 19:56:41,861 INFO [finetune.py:976] (1/7) Epoch 7, batch 3900, loss[loss=0.194, simple_loss=0.2603, pruned_loss=0.06386, over 4749.00 frames. ], tot_loss[loss=0.2009, simple_loss=0.264, pruned_loss=0.06891, over 954756.94 frames. ], batch size: 27, lr: 3.86e-03, grad_scale: 32.0 2023-04-26 19:57:02,032 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([3.8629, 3.7990, 2.7823, 4.4584, 3.9283, 3.9100, 1.7429, 3.7345], device='cuda:1'), covar=tensor([0.1769, 0.1156, 0.3002, 0.1592, 0.3297, 0.1898, 0.5782, 0.2610], device='cuda:1'), in_proj_covar=tensor([0.0247, 0.0220, 0.0254, 0.0312, 0.0306, 0.0256, 0.0276, 0.0275], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-26 19:57:25,207 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.8535, 2.7968, 2.3349, 3.2544, 2.8534, 2.8566, 1.3074, 2.7815], device='cuda:1'), covar=tensor([0.2378, 0.1878, 0.3956, 0.3480, 0.3130, 0.2599, 0.5454, 0.3227], device='cuda:1'), in_proj_covar=tensor([0.0246, 0.0220, 0.0254, 0.0311, 0.0306, 0.0256, 0.0276, 0.0275], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-26 19:57:47,849 INFO [finetune.py:976] (1/7) Epoch 7, batch 3950, loss[loss=0.2267, simple_loss=0.2792, pruned_loss=0.0871, over 4927.00 frames. ], tot_loss[loss=0.1975, simple_loss=0.2606, pruned_loss=0.06725, over 956344.65 frames. ], batch size: 38, lr: 3.86e-03, grad_scale: 32.0 2023-04-26 19:58:09,819 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.007e+02 1.650e+02 1.940e+02 2.337e+02 4.288e+02, threshold=3.879e+02, percent-clipped=1.0 2023-04-26 19:58:18,646 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=38335.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:58:43,687 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=38360.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:58:53,666 INFO [finetune.py:976] (1/7) Epoch 7, batch 4000, loss[loss=0.1923, simple_loss=0.2596, pruned_loss=0.06243, over 4747.00 frames. ], tot_loss[loss=0.1972, simple_loss=0.2598, pruned_loss=0.06732, over 957103.29 frames. ], batch size: 54, lr: 3.86e-03, grad_scale: 32.0 2023-04-26 20:00:00,681 INFO [finetune.py:976] (1/7) Epoch 7, batch 4050, loss[loss=0.1516, simple_loss=0.2265, pruned_loss=0.0383, over 4779.00 frames. ], tot_loss[loss=0.2012, simple_loss=0.2644, pruned_loss=0.06896, over 957568.07 frames. ], batch size: 26, lr: 3.86e-03, grad_scale: 32.0 2023-04-26 20:00:08,971 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-04-26 20:00:10,086 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=38421.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:00:21,444 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.926e+01 1.887e+02 2.250e+02 2.854e+02 6.950e+02, threshold=4.500e+02, percent-clipped=7.0 2023-04-26 20:01:01,406 INFO [finetune.py:976] (1/7) Epoch 7, batch 4100, loss[loss=0.2219, simple_loss=0.2838, pruned_loss=0.07996, over 4888.00 frames. ], tot_loss[loss=0.2039, simple_loss=0.268, pruned_loss=0.06991, over 958731.04 frames. ], batch size: 43, lr: 3.86e-03, grad_scale: 32.0 2023-04-26 20:01:07,931 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.76 vs. limit=2.0 2023-04-26 20:01:09,484 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.2017, 1.1854, 1.2711, 1.5777, 1.5870, 1.2708, 0.9945, 1.3624], device='cuda:1'), covar=tensor([0.0993, 0.1434, 0.0916, 0.0636, 0.0657, 0.0979, 0.0971, 0.0726], device='cuda:1'), in_proj_covar=tensor([0.0202, 0.0207, 0.0184, 0.0180, 0.0181, 0.0195, 0.0164, 0.0189], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 20:02:13,352 INFO [finetune.py:976] (1/7) Epoch 7, batch 4150, loss[loss=0.2227, simple_loss=0.2894, pruned_loss=0.07799, over 4822.00 frames. ], tot_loss[loss=0.2051, simple_loss=0.2695, pruned_loss=0.07032, over 958601.09 frames. ], batch size: 47, lr: 3.86e-03, grad_scale: 32.0 2023-04-26 20:02:28,364 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.215e+02 1.719e+02 1.998e+02 2.310e+02 4.665e+02, threshold=3.995e+02, percent-clipped=1.0 2023-04-26 20:02:52,781 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.50 vs. limit=2.0 2023-04-26 20:02:57,365 INFO [finetune.py:976] (1/7) Epoch 7, batch 4200, loss[loss=0.1577, simple_loss=0.2105, pruned_loss=0.05246, over 3979.00 frames. ], tot_loss[loss=0.2057, simple_loss=0.2699, pruned_loss=0.07075, over 956048.56 frames. ], batch size: 17, lr: 3.86e-03, grad_scale: 32.0 2023-04-26 20:02:58,734 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.4073, 1.6112, 1.6982, 1.8635, 1.6868, 1.7482, 1.8279, 1.7453], device='cuda:1'), covar=tensor([0.5989, 0.9030, 0.7365, 0.6606, 0.8018, 1.1595, 0.8301, 0.8332], device='cuda:1'), in_proj_covar=tensor([0.0323, 0.0392, 0.0316, 0.0327, 0.0342, 0.0407, 0.0370, 0.0329], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-26 20:03:29,904 INFO [finetune.py:976] (1/7) Epoch 7, batch 4250, loss[loss=0.2042, simple_loss=0.2713, pruned_loss=0.06858, over 4791.00 frames. ], tot_loss[loss=0.2022, simple_loss=0.266, pruned_loss=0.06924, over 954845.45 frames. ], batch size: 51, lr: 3.86e-03, grad_scale: 32.0 2023-04-26 20:03:35,991 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.67 vs. limit=2.0 2023-04-26 20:03:40,412 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.148e+02 1.629e+02 2.024e+02 2.405e+02 4.304e+02, threshold=4.048e+02, percent-clipped=1.0 2023-04-26 20:03:42,856 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([0.7774, 1.8013, 1.7283, 1.3262, 2.0142, 1.5513, 2.4278, 1.5654], device='cuda:1'), covar=tensor([0.3798, 0.1785, 0.4341, 0.3347, 0.1664, 0.2195, 0.1398, 0.4254], device='cuda:1'), in_proj_covar=tensor([0.0348, 0.0353, 0.0432, 0.0363, 0.0389, 0.0382, 0.0383, 0.0420], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 20:03:44,043 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=38635.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:04:03,667 INFO [finetune.py:976] (1/7) Epoch 7, batch 4300, loss[loss=0.1918, simple_loss=0.2555, pruned_loss=0.06401, over 4872.00 frames. ], tot_loss[loss=0.2011, simple_loss=0.264, pruned_loss=0.06912, over 956735.76 frames. ], batch size: 34, lr: 3.86e-03, grad_scale: 32.0 2023-04-26 20:04:26,425 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=38683.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:05:03,900 INFO [finetune.py:976] (1/7) Epoch 7, batch 4350, loss[loss=0.2545, simple_loss=0.3006, pruned_loss=0.1042, over 4918.00 frames. ], tot_loss[loss=0.1978, simple_loss=0.2605, pruned_loss=0.06754, over 957693.96 frames. ], batch size: 43, lr: 3.86e-03, grad_scale: 32.0 2023-04-26 20:05:08,719 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=38716.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:05:18,152 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.164e+02 1.707e+02 2.023e+02 2.548e+02 4.531e+02, threshold=4.045e+02, percent-clipped=2.0 2023-04-26 20:05:48,653 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-04-26 20:05:52,730 INFO [finetune.py:976] (1/7) Epoch 7, batch 4400, loss[loss=0.1905, simple_loss=0.2662, pruned_loss=0.0574, over 4910.00 frames. ], tot_loss[loss=0.1993, simple_loss=0.2619, pruned_loss=0.06833, over 957646.80 frames. ], batch size: 37, lr: 3.86e-03, grad_scale: 32.0 2023-04-26 20:06:12,853 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([3.3538, 3.2842, 2.5440, 3.8734, 3.4341, 3.3761, 1.5181, 3.2166], device='cuda:1'), covar=tensor([0.2064, 0.1490, 0.3322, 0.2368, 0.4966, 0.2226, 0.6018, 0.3015], device='cuda:1'), in_proj_covar=tensor([0.0245, 0.0219, 0.0254, 0.0311, 0.0304, 0.0255, 0.0276, 0.0274], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-26 20:06:15,159 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.3668, 3.6048, 0.9285, 1.6646, 1.7584, 2.4852, 1.9087, 0.9813], device='cuda:1'), covar=tensor([0.1473, 0.0707, 0.2030, 0.1406, 0.1191, 0.0984, 0.1538, 0.2174], device='cuda:1'), in_proj_covar=tensor([0.0119, 0.0255, 0.0143, 0.0125, 0.0136, 0.0156, 0.0121, 0.0124], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-26 20:06:23,125 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0188, 1.4240, 1.8325, 2.3086, 1.7652, 1.4497, 1.1505, 1.6371], device='cuda:1'), covar=tensor([0.3841, 0.4221, 0.1970, 0.2873, 0.3446, 0.3354, 0.5177, 0.2966], device='cuda:1'), in_proj_covar=tensor([0.0279, 0.0253, 0.0218, 0.0322, 0.0214, 0.0229, 0.0236, 0.0188], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-26 20:06:26,655 INFO [finetune.py:976] (1/7) Epoch 7, batch 4450, loss[loss=0.1941, simple_loss=0.2613, pruned_loss=0.06344, over 4759.00 frames. ], tot_loss[loss=0.2016, simple_loss=0.265, pruned_loss=0.06915, over 956697.61 frames. ], batch size: 27, lr: 3.86e-03, grad_scale: 32.0 2023-04-26 20:06:36,510 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=38823.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:06:45,522 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.120e+02 1.756e+02 2.089e+02 2.730e+02 5.109e+02, threshold=4.178e+02, percent-clipped=5.0 2023-04-26 20:07:37,018 INFO [finetune.py:976] (1/7) Epoch 7, batch 4500, loss[loss=0.2038, simple_loss=0.2789, pruned_loss=0.0644, over 4899.00 frames. ], tot_loss[loss=0.204, simple_loss=0.2674, pruned_loss=0.07025, over 957215.94 frames. ], batch size: 37, lr: 3.86e-03, grad_scale: 32.0 2023-04-26 20:07:59,562 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=38884.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:08:12,772 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([3.7079, 3.6244, 2.7197, 4.2737, 3.6962, 3.6772, 1.7696, 3.6754], device='cuda:1'), covar=tensor([0.1462, 0.1241, 0.3402, 0.1357, 0.2414, 0.1729, 0.5317, 0.2122], device='cuda:1'), in_proj_covar=tensor([0.0244, 0.0217, 0.0252, 0.0310, 0.0302, 0.0253, 0.0274, 0.0272], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-26 20:08:44,168 INFO [finetune.py:976] (1/7) Epoch 7, batch 4550, loss[loss=0.2248, simple_loss=0.2991, pruned_loss=0.07528, over 4770.00 frames. ], tot_loss[loss=0.2034, simple_loss=0.2671, pruned_loss=0.06985, over 953867.26 frames. ], batch size: 29, lr: 3.86e-03, grad_scale: 32.0 2023-04-26 20:08:58,071 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.107e+02 1.848e+02 2.064e+02 2.542e+02 5.076e+02, threshold=4.127e+02, percent-clipped=2.0 2023-04-26 20:09:04,825 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.2241, 1.4753, 1.2445, 1.6746, 1.5472, 1.7236, 1.3328, 3.2038], device='cuda:1'), covar=tensor([0.0641, 0.0779, 0.0820, 0.1171, 0.0631, 0.0594, 0.0750, 0.0164], device='cuda:1'), in_proj_covar=tensor([0.0039, 0.0039, 0.0041, 0.0044, 0.0040, 0.0040, 0.0039, 0.0060], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0014, 0.0013, 0.0013, 0.0013, 0.0017], device='cuda:1') 2023-04-26 20:09:07,757 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4818, 1.1841, 0.4630, 1.2115, 1.4398, 1.3556, 1.2265, 1.2365], device='cuda:1'), covar=tensor([0.0582, 0.0474, 0.0471, 0.0645, 0.0323, 0.0594, 0.0612, 0.0697], device='cuda:1'), in_proj_covar=tensor([0.0031, 0.0026, 0.0024, 0.0031, 0.0021, 0.0030, 0.0030, 0.0031], device='cuda:1'), out_proj_covar=tensor([0.0049, 0.0045, 0.0038, 0.0050, 0.0038, 0.0048, 0.0048, 0.0050], device='cuda:1') 2023-04-26 20:09:50,305 INFO [finetune.py:976] (1/7) Epoch 7, batch 4600, loss[loss=0.174, simple_loss=0.2455, pruned_loss=0.05128, over 4892.00 frames. ], tot_loss[loss=0.2016, simple_loss=0.2661, pruned_loss=0.06858, over 955504.21 frames. ], batch size: 32, lr: 3.86e-03, grad_scale: 32.0 2023-04-26 20:10:14,029 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=3.61 vs. limit=5.0 2023-04-26 20:10:56,081 INFO [finetune.py:976] (1/7) Epoch 7, batch 4650, loss[loss=0.1887, simple_loss=0.2415, pruned_loss=0.06791, over 4249.00 frames. ], tot_loss[loss=0.2005, simple_loss=0.2641, pruned_loss=0.06841, over 956209.49 frames. ], batch size: 65, lr: 3.86e-03, grad_scale: 32.0 2023-04-26 20:10:56,160 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=39016.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:11:16,118 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.165e+02 1.581e+02 1.907e+02 2.292e+02 5.308e+02, threshold=3.814e+02, percent-clipped=1.0 2023-04-26 20:11:51,919 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.1499, 1.5571, 1.5318, 2.0461, 2.2805, 1.9271, 1.8334, 1.6143], device='cuda:1'), covar=tensor([0.1886, 0.1840, 0.1922, 0.1572, 0.1243, 0.1917, 0.2000, 0.1915], device='cuda:1'), in_proj_covar=tensor([0.0307, 0.0326, 0.0354, 0.0302, 0.0339, 0.0326, 0.0307, 0.0353], device='cuda:1'), out_proj_covar=tensor([6.5245e-05, 6.9355e-05, 7.6717e-05, 6.2667e-05, 7.1485e-05, 7.0252e-05, 6.6288e-05, 7.5909e-05], device='cuda:1') 2023-04-26 20:12:02,779 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=39064.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:12:03,960 INFO [finetune.py:976] (1/7) Epoch 7, batch 4700, loss[loss=0.1831, simple_loss=0.2459, pruned_loss=0.06014, over 4830.00 frames. ], tot_loss[loss=0.1976, simple_loss=0.2605, pruned_loss=0.0673, over 954427.11 frames. ], batch size: 30, lr: 3.86e-03, grad_scale: 32.0 2023-04-26 20:13:04,273 INFO [finetune.py:976] (1/7) Epoch 7, batch 4750, loss[loss=0.2107, simple_loss=0.2696, pruned_loss=0.07589, over 4905.00 frames. ], tot_loss[loss=0.196, simple_loss=0.2586, pruned_loss=0.06666, over 955269.78 frames. ], batch size: 37, lr: 3.86e-03, grad_scale: 32.0 2023-04-26 20:13:24,484 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.169e+02 1.720e+02 2.202e+02 2.598e+02 5.909e+02, threshold=4.403e+02, percent-clipped=7.0 2023-04-26 20:13:31,788 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7111, 1.8106, 0.9152, 1.3541, 1.9208, 1.5765, 1.4376, 1.5031], device='cuda:1'), covar=tensor([0.0542, 0.0394, 0.0396, 0.0580, 0.0281, 0.0569, 0.0561, 0.0620], device='cuda:1'), in_proj_covar=tensor([0.0030, 0.0026, 0.0024, 0.0030, 0.0021, 0.0030, 0.0029, 0.0030], device='cuda:1'), out_proj_covar=tensor([0.0049, 0.0044, 0.0038, 0.0050, 0.0037, 0.0048, 0.0048, 0.0050], device='cuda:1') 2023-04-26 20:13:55,066 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7908, 1.8791, 1.2685, 1.4666, 2.1819, 1.6471, 1.5078, 1.5405], device='cuda:1'), covar=tensor([0.0541, 0.0391, 0.0327, 0.0571, 0.0246, 0.0533, 0.0528, 0.0602], device='cuda:1'), in_proj_covar=tensor([0.0030, 0.0026, 0.0024, 0.0030, 0.0021, 0.0030, 0.0029, 0.0030], device='cuda:1'), out_proj_covar=tensor([0.0049, 0.0044, 0.0038, 0.0050, 0.0037, 0.0048, 0.0047, 0.0049], device='cuda:1') 2023-04-26 20:14:16,831 INFO [finetune.py:976] (1/7) Epoch 7, batch 4800, loss[loss=0.2415, simple_loss=0.3076, pruned_loss=0.08773, over 4928.00 frames. ], tot_loss[loss=0.1976, simple_loss=0.2605, pruned_loss=0.06735, over 955370.35 frames. ], batch size: 38, lr: 3.86e-03, grad_scale: 32.0 2023-04-26 20:14:31,376 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=39179.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:14:40,491 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=39185.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:15:12,278 INFO [finetune.py:976] (1/7) Epoch 7, batch 4850, loss[loss=0.1863, simple_loss=0.2427, pruned_loss=0.065, over 4874.00 frames. ], tot_loss[loss=0.2008, simple_loss=0.2644, pruned_loss=0.06864, over 955963.19 frames. ], batch size: 31, lr: 3.86e-03, grad_scale: 64.0 2023-04-26 20:15:17,192 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4808, 1.3458, 1.7700, 1.7030, 1.3756, 1.1717, 1.4547, 0.8780], device='cuda:1'), covar=tensor([0.0716, 0.0843, 0.0540, 0.0797, 0.0881, 0.1307, 0.0756, 0.0962], device='cuda:1'), in_proj_covar=tensor([0.0066, 0.0074, 0.0073, 0.0067, 0.0077, 0.0095, 0.0080, 0.0076], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-26 20:15:21,785 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.088e+02 1.823e+02 2.117e+02 2.667e+02 5.725e+02, threshold=4.234e+02, percent-clipped=1.0 2023-04-26 20:15:31,634 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=39246.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:15:45,481 INFO [finetune.py:976] (1/7) Epoch 7, batch 4900, loss[loss=0.2244, simple_loss=0.2938, pruned_loss=0.07749, over 4816.00 frames. ], tot_loss[loss=0.2038, simple_loss=0.2673, pruned_loss=0.07021, over 956071.10 frames. ], batch size: 40, lr: 3.86e-03, grad_scale: 64.0 2023-04-26 20:16:09,173 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=39284.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:16:30,279 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8037, 1.2935, 1.6273, 1.6088, 1.5456, 1.2887, 0.7050, 1.2800], device='cuda:1'), covar=tensor([0.4147, 0.4189, 0.2096, 0.2814, 0.3338, 0.3298, 0.5145, 0.2771], device='cuda:1'), in_proj_covar=tensor([0.0280, 0.0254, 0.0219, 0.0323, 0.0214, 0.0229, 0.0238, 0.0188], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-26 20:16:40,462 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=3.72 vs. limit=5.0 2023-04-26 20:16:44,225 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-04-26 20:16:50,964 INFO [finetune.py:976] (1/7) Epoch 7, batch 4950, loss[loss=0.2033, simple_loss=0.2663, pruned_loss=0.07014, over 4858.00 frames. ], tot_loss[loss=0.2046, simple_loss=0.2687, pruned_loss=0.07023, over 957087.17 frames. ], batch size: 31, lr: 3.86e-03, grad_scale: 64.0 2023-04-26 20:16:56,301 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4563, 1.3346, 4.2998, 4.0252, 3.8276, 4.1277, 4.0114, 3.7958], device='cuda:1'), covar=tensor([0.6928, 0.5964, 0.1013, 0.1649, 0.0982, 0.2007, 0.1510, 0.1337], device='cuda:1'), in_proj_covar=tensor([0.0311, 0.0308, 0.0410, 0.0415, 0.0350, 0.0408, 0.0319, 0.0369], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-26 20:17:06,088 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.056e+02 1.715e+02 2.035e+02 2.513e+02 5.677e+02, threshold=4.070e+02, percent-clipped=1.0 2023-04-26 20:17:25,339 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=39345.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:17:46,319 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.01 vs. limit=5.0 2023-04-26 20:17:49,760 INFO [finetune.py:976] (1/7) Epoch 7, batch 5000, loss[loss=0.2304, simple_loss=0.2829, pruned_loss=0.089, over 4713.00 frames. ], tot_loss[loss=0.2032, simple_loss=0.2674, pruned_loss=0.06953, over 957472.66 frames. ], batch size: 23, lr: 3.86e-03, grad_scale: 64.0 2023-04-26 20:18:23,436 INFO [finetune.py:976] (1/7) Epoch 7, batch 5050, loss[loss=0.1991, simple_loss=0.2663, pruned_loss=0.06591, over 4898.00 frames. ], tot_loss[loss=0.2011, simple_loss=0.2646, pruned_loss=0.06881, over 958687.38 frames. ], batch size: 32, lr: 3.85e-03, grad_scale: 64.0 2023-04-26 20:18:33,400 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.120e+02 1.684e+02 2.008e+02 2.415e+02 4.173e+02, threshold=4.016e+02, percent-clipped=2.0 2023-04-26 20:18:54,472 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=39463.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:18:56,691 INFO [finetune.py:976] (1/7) Epoch 7, batch 5100, loss[loss=0.1763, simple_loss=0.2416, pruned_loss=0.05555, over 4828.00 frames. ], tot_loss[loss=0.1982, simple_loss=0.2615, pruned_loss=0.06749, over 958702.40 frames. ], batch size: 33, lr: 3.85e-03, grad_scale: 64.0 2023-04-26 20:18:57,362 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6036, 1.3794, 4.4558, 4.1333, 3.9648, 4.1750, 4.1276, 3.9708], device='cuda:1'), covar=tensor([0.7021, 0.5810, 0.1049, 0.1852, 0.1130, 0.1892, 0.1648, 0.1453], device='cuda:1'), in_proj_covar=tensor([0.0310, 0.0308, 0.0409, 0.0415, 0.0349, 0.0406, 0.0317, 0.0368], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-26 20:19:05,635 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=39479.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:19:29,889 INFO [finetune.py:976] (1/7) Epoch 7, batch 5150, loss[loss=0.1866, simple_loss=0.2455, pruned_loss=0.06381, over 4882.00 frames. ], tot_loss[loss=0.1991, simple_loss=0.2619, pruned_loss=0.0682, over 957122.23 frames. ], batch size: 32, lr: 3.85e-03, grad_scale: 32.0 2023-04-26 20:19:35,793 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.2661, 1.6792, 1.6473, 2.1283, 1.8680, 2.0687, 1.4796, 4.3563], device='cuda:1'), covar=tensor([0.0752, 0.0961, 0.0941, 0.1350, 0.0719, 0.0606, 0.0945, 0.0221], device='cuda:1'), in_proj_covar=tensor([0.0039, 0.0039, 0.0041, 0.0044, 0.0040, 0.0040, 0.0039, 0.0060], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0014, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:1') 2023-04-26 20:19:35,825 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=39524.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:19:38,070 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=39527.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:19:40,451 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.309e+02 1.798e+02 2.123e+02 2.599e+02 5.486e+02, threshold=4.247e+02, percent-clipped=4.0 2023-04-26 20:19:47,585 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=39541.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:19:59,235 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.51 vs. limit=2.0 2023-04-26 20:20:03,313 INFO [finetune.py:976] (1/7) Epoch 7, batch 5200, loss[loss=0.211, simple_loss=0.2858, pruned_loss=0.0681, over 4745.00 frames. ], tot_loss[loss=0.2005, simple_loss=0.2639, pruned_loss=0.0686, over 955839.64 frames. ], batch size: 54, lr: 3.85e-03, grad_scale: 32.0 2023-04-26 20:20:12,850 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.27 vs. limit=5.0 2023-04-26 20:20:37,216 INFO [finetune.py:976] (1/7) Epoch 7, batch 5250, loss[loss=0.2046, simple_loss=0.2711, pruned_loss=0.06906, over 4812.00 frames. ], tot_loss[loss=0.202, simple_loss=0.2659, pruned_loss=0.06903, over 955865.26 frames. ], batch size: 33, lr: 3.85e-03, grad_scale: 32.0 2023-04-26 20:20:43,313 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.8097, 1.9618, 1.9174, 2.0951, 1.9486, 2.1027, 2.0130, 1.9339], device='cuda:1'), covar=tensor([0.5728, 0.9889, 0.8714, 0.7770, 0.8370, 1.1604, 1.1227, 0.9527], device='cuda:1'), in_proj_covar=tensor([0.0322, 0.0391, 0.0316, 0.0325, 0.0342, 0.0407, 0.0368, 0.0329], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-26 20:20:47,796 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.203e+02 1.776e+02 2.030e+02 2.603e+02 8.469e+02, threshold=4.060e+02, percent-clipped=1.0 2023-04-26 20:20:53,203 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=39639.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 20:20:53,786 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=39640.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:20:53,874 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5819, 1.2961, 1.6655, 2.0246, 1.7278, 1.5493, 1.5743, 1.6076], device='cuda:1'), covar=tensor([0.6570, 0.9106, 0.8875, 0.9374, 0.7752, 1.0525, 1.1029, 0.9204], device='cuda:1'), in_proj_covar=tensor([0.0409, 0.0428, 0.0511, 0.0534, 0.0439, 0.0457, 0.0471, 0.0466], device='cuda:1'), out_proj_covar=tensor([9.9457e-05, 1.0631e-04, 1.1542e-04, 1.2652e-04, 1.0684e-04, 1.1074e-04, 1.1316e-04, 1.1348e-04], device='cuda:1') 2023-04-26 20:21:10,369 INFO [finetune.py:976] (1/7) Epoch 7, batch 5300, loss[loss=0.2159, simple_loss=0.2901, pruned_loss=0.07082, over 4825.00 frames. ], tot_loss[loss=0.2027, simple_loss=0.267, pruned_loss=0.06919, over 954538.66 frames. ], batch size: 40, lr: 3.85e-03, grad_scale: 32.0 2023-04-26 20:21:50,865 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=39700.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 20:22:11,712 INFO [finetune.py:976] (1/7) Epoch 7, batch 5350, loss[loss=0.1764, simple_loss=0.2534, pruned_loss=0.04973, over 4794.00 frames. ], tot_loss[loss=0.2023, simple_loss=0.2665, pruned_loss=0.06912, over 950466.09 frames. ], batch size: 45, lr: 3.85e-03, grad_scale: 32.0 2023-04-26 20:22:31,528 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.006e+02 1.675e+02 2.001e+02 2.311e+02 3.893e+02, threshold=4.002e+02, percent-clipped=0.0 2023-04-26 20:23:17,037 INFO [finetune.py:976] (1/7) Epoch 7, batch 5400, loss[loss=0.1473, simple_loss=0.2168, pruned_loss=0.03885, over 4765.00 frames. ], tot_loss[loss=0.1985, simple_loss=0.2626, pruned_loss=0.06716, over 949486.31 frames. ], batch size: 27, lr: 3.85e-03, grad_scale: 32.0 2023-04-26 20:23:17,766 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=39767.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:24:18,120 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=39815.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:24:18,613 INFO [finetune.py:976] (1/7) Epoch 7, batch 5450, loss[loss=0.2083, simple_loss=0.2694, pruned_loss=0.07361, over 4819.00 frames. ], tot_loss[loss=0.1969, simple_loss=0.2605, pruned_loss=0.06664, over 951480.00 frames. ], batch size: 38, lr: 3.85e-03, grad_scale: 32.0 2023-04-26 20:24:20,495 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=39819.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:24:31,607 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=39828.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:24:33,271 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.150e+02 1.624e+02 1.910e+02 2.215e+02 3.764e+02, threshold=3.819e+02, percent-clipped=0.0 2023-04-26 20:24:40,441 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=39841.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:24:43,538 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.58 vs. limit=5.0 2023-04-26 20:24:57,926 INFO [finetune.py:976] (1/7) Epoch 7, batch 5500, loss[loss=0.1604, simple_loss=0.2361, pruned_loss=0.04238, over 4791.00 frames. ], tot_loss[loss=0.1951, simple_loss=0.258, pruned_loss=0.06606, over 952096.93 frames. ], batch size: 29, lr: 3.85e-03, grad_scale: 32.0 2023-04-26 20:25:04,136 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=39876.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:25:12,053 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=39889.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:25:23,291 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8952, 2.4349, 2.0562, 2.2810, 1.6645, 2.0068, 2.0607, 1.6421], device='cuda:1'), covar=tensor([0.2465, 0.1457, 0.0898, 0.1394, 0.3768, 0.1324, 0.2215, 0.3022], device='cuda:1'), in_proj_covar=tensor([0.0302, 0.0321, 0.0234, 0.0293, 0.0317, 0.0274, 0.0262, 0.0286], device='cuda:1'), out_proj_covar=tensor([1.2267e-04, 1.2994e-04, 9.4562e-05, 1.1755e-04, 1.3033e-04, 1.1071e-04, 1.0731e-04, 1.1501e-04], device='cuda:1') 2023-04-26 20:25:31,741 INFO [finetune.py:976] (1/7) Epoch 7, batch 5550, loss[loss=0.1997, simple_loss=0.2821, pruned_loss=0.05866, over 4801.00 frames. ], tot_loss[loss=0.1961, simple_loss=0.2593, pruned_loss=0.06643, over 953048.97 frames. ], batch size: 45, lr: 3.85e-03, grad_scale: 32.0 2023-04-26 20:25:39,277 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7994, 1.4819, 1.2654, 1.5923, 2.1031, 1.6667, 1.4167, 1.2835], device='cuda:1'), covar=tensor([0.1771, 0.1520, 0.2127, 0.1346, 0.0792, 0.1664, 0.2339, 0.2223], device='cuda:1'), in_proj_covar=tensor([0.0309, 0.0329, 0.0359, 0.0303, 0.0342, 0.0328, 0.0311, 0.0355], device='cuda:1'), out_proj_covar=tensor([6.5709e-05, 6.9938e-05, 7.7797e-05, 6.2885e-05, 7.2096e-05, 7.0719e-05, 6.7125e-05, 7.6234e-05], device='cuda:1') 2023-04-26 20:25:40,907 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.156e+02 1.695e+02 2.043e+02 2.631e+02 5.173e+02, threshold=4.085e+02, percent-clipped=3.0 2023-04-26 20:25:46,474 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=39940.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:25:48,316 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 2023-04-26 20:26:02,970 INFO [finetune.py:976] (1/7) Epoch 7, batch 5600, loss[loss=0.2125, simple_loss=0.2767, pruned_loss=0.07416, over 4909.00 frames. ], tot_loss[loss=0.1994, simple_loss=0.2635, pruned_loss=0.06764, over 954229.07 frames. ], batch size: 37, lr: 3.85e-03, grad_scale: 32.0 2023-04-26 20:26:14,918 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.48 vs. limit=5.0 2023-04-26 20:26:15,865 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=39988.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:26:19,990 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=39995.0, num_to_drop=1, layers_to_drop={2} 2023-04-26 20:26:38,931 INFO [finetune.py:976] (1/7) Epoch 7, batch 5650, loss[loss=0.1841, simple_loss=0.2531, pruned_loss=0.05754, over 4934.00 frames. ], tot_loss[loss=0.202, simple_loss=0.267, pruned_loss=0.06847, over 954722.44 frames. ], batch size: 33, lr: 3.85e-03, grad_scale: 32.0 2023-04-26 20:26:53,974 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.232e+02 1.674e+02 2.052e+02 2.439e+02 4.352e+02, threshold=4.105e+02, percent-clipped=2.0 2023-04-26 20:27:20,304 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6338, 1.5079, 0.9523, 1.3768, 1.4687, 1.4864, 1.4251, 1.4251], device='cuda:1'), covar=tensor([0.0484, 0.0356, 0.0398, 0.0519, 0.0329, 0.0510, 0.0506, 0.0576], device='cuda:1'), in_proj_covar=tensor([0.0030, 0.0026, 0.0024, 0.0030, 0.0021, 0.0030, 0.0029, 0.0030], device='cuda:1'), out_proj_covar=tensor([0.0049, 0.0044, 0.0038, 0.0050, 0.0037, 0.0048, 0.0048, 0.0050], device='cuda:1') 2023-04-26 20:27:27,070 INFO [finetune.py:976] (1/7) Epoch 7, batch 5700, loss[loss=0.1393, simple_loss=0.1934, pruned_loss=0.04259, over 4032.00 frames. ], tot_loss[loss=0.2006, simple_loss=0.2638, pruned_loss=0.06866, over 938256.91 frames. ], batch size: 17, lr: 3.85e-03, grad_scale: 32.0 2023-04-26 20:27:47,427 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9682, 1.9179, 2.1170, 2.3534, 2.3567, 1.9140, 1.6576, 1.9969], device='cuda:1'), covar=tensor([0.0731, 0.1021, 0.0542, 0.0479, 0.0531, 0.0901, 0.0773, 0.0659], device='cuda:1'), in_proj_covar=tensor([0.0198, 0.0205, 0.0181, 0.0178, 0.0179, 0.0194, 0.0162, 0.0188], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 20:28:14,049 INFO [finetune.py:976] (1/7) Epoch 8, batch 0, loss[loss=0.2181, simple_loss=0.2888, pruned_loss=0.07373, over 4839.00 frames. ], tot_loss[loss=0.2181, simple_loss=0.2888, pruned_loss=0.07373, over 4839.00 frames. ], batch size: 49, lr: 3.85e-03, grad_scale: 32.0 2023-04-26 20:28:14,049 INFO [finetune.py:1001] (1/7) Computing validation loss 2023-04-26 20:28:25,212 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.3959, 1.2235, 1.6496, 1.5376, 1.2761, 1.1396, 1.2991, 0.8179], device='cuda:1'), covar=tensor([0.0692, 0.1067, 0.0548, 0.0804, 0.0908, 0.1421, 0.0755, 0.0906], device='cuda:1'), in_proj_covar=tensor([0.0066, 0.0074, 0.0073, 0.0067, 0.0077, 0.0096, 0.0080, 0.0075], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-26 20:28:30,560 INFO [finetune.py:1010] (1/7) Epoch 8, validation: loss=0.1574, simple_loss=0.2299, pruned_loss=0.04247, over 2265189.00 frames. 2023-04-26 20:28:30,561 INFO [finetune.py:1011] (1/7) Maximum memory allocated so far is 6391MB 2023-04-26 20:28:44,868 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7107, 1.9482, 0.9788, 1.4802, 2.1309, 1.6014, 1.5466, 1.5846], device='cuda:1'), covar=tensor([0.0506, 0.0336, 0.0340, 0.0539, 0.0268, 0.0530, 0.0474, 0.0576], device='cuda:1'), in_proj_covar=tensor([0.0030, 0.0026, 0.0024, 0.0030, 0.0021, 0.0030, 0.0029, 0.0030], device='cuda:1'), out_proj_covar=tensor([0.0049, 0.0044, 0.0038, 0.0050, 0.0037, 0.0048, 0.0048, 0.0049], device='cuda:1') 2023-04-26 20:29:03,042 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=40119.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:29:05,425 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=40123.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:29:10,788 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.231e+02 1.752e+02 2.089e+02 2.536e+02 4.447e+02, threshold=4.178e+02, percent-clipped=1.0 2023-04-26 20:29:20,979 INFO [finetune.py:976] (1/7) Epoch 8, batch 50, loss[loss=0.1989, simple_loss=0.2618, pruned_loss=0.06804, over 4855.00 frames. ], tot_loss[loss=0.2057, simple_loss=0.2705, pruned_loss=0.07045, over 217687.51 frames. ], batch size: 44, lr: 3.85e-03, grad_scale: 32.0 2023-04-26 20:29:36,060 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=40167.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:29:38,551 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=40171.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:29:54,377 INFO [finetune.py:976] (1/7) Epoch 8, batch 100, loss[loss=0.1905, simple_loss=0.255, pruned_loss=0.06298, over 4927.00 frames. ], tot_loss[loss=0.1976, simple_loss=0.2606, pruned_loss=0.06728, over 380928.06 frames. ], batch size: 37, lr: 3.85e-03, grad_scale: 32.0 2023-04-26 20:29:55,614 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=40195.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:30:09,361 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.67 vs. limit=5.0 2023-04-26 20:30:10,504 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=40218.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:30:18,313 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.006e+02 1.606e+02 1.925e+02 2.421e+02 4.111e+02, threshold=3.850e+02, percent-clipped=0.0 2023-04-26 20:30:20,803 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.51 vs. limit=2.0 2023-04-26 20:30:24,692 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.3648, 1.1513, 1.6108, 1.4875, 1.2717, 1.0914, 1.2857, 0.8491], device='cuda:1'), covar=tensor([0.0634, 0.0955, 0.0520, 0.0837, 0.0813, 0.1386, 0.0646, 0.0852], device='cuda:1'), in_proj_covar=tensor([0.0065, 0.0073, 0.0072, 0.0066, 0.0076, 0.0095, 0.0080, 0.0075], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-26 20:30:28,097 INFO [finetune.py:976] (1/7) Epoch 8, batch 150, loss[loss=0.1757, simple_loss=0.2382, pruned_loss=0.0566, over 4931.00 frames. ], tot_loss[loss=0.1935, simple_loss=0.2555, pruned_loss=0.0657, over 510110.71 frames. ], batch size: 38, lr: 3.85e-03, grad_scale: 32.0 2023-04-26 20:30:29,553 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 2023-04-26 20:30:36,428 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=40256.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:30:41,946 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.1718, 0.7022, 0.9294, 0.6770, 1.3036, 0.9357, 0.7857, 1.0482], device='cuda:1'), covar=tensor([0.1707, 0.1717, 0.2361, 0.1790, 0.1172, 0.1499, 0.1909, 0.2211], device='cuda:1'), in_proj_covar=tensor([0.0307, 0.0327, 0.0355, 0.0302, 0.0341, 0.0326, 0.0308, 0.0353], device='cuda:1'), out_proj_covar=tensor([6.5316e-05, 6.9444e-05, 7.7040e-05, 6.2704e-05, 7.1784e-05, 7.0352e-05, 6.6608e-05, 7.5773e-05], device='cuda:1') 2023-04-26 20:30:46,191 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.3777, 2.4015, 1.8899, 2.1533, 2.4800, 1.8503, 3.2125, 1.8715], device='cuda:1'), covar=tensor([0.4650, 0.2346, 0.5223, 0.3572, 0.2048, 0.3189, 0.1845, 0.4415], device='cuda:1'), in_proj_covar=tensor([0.0345, 0.0350, 0.0432, 0.0361, 0.0386, 0.0380, 0.0382, 0.0417], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 20:30:46,193 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=40272.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:30:50,428 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=40279.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:31:01,726 INFO [finetune.py:976] (1/7) Epoch 8, batch 200, loss[loss=0.1984, simple_loss=0.2667, pruned_loss=0.06505, over 4908.00 frames. ], tot_loss[loss=0.1936, simple_loss=0.2557, pruned_loss=0.06578, over 610396.36 frames. ], batch size: 36, lr: 3.85e-03, grad_scale: 32.0 2023-04-26 20:31:02,440 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=40295.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 20:31:24,759 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=40330.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:31:25,245 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.151e+02 1.700e+02 1.993e+02 2.528e+02 5.564e+02, threshold=3.985e+02, percent-clipped=2.0 2023-04-26 20:31:26,594 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=40333.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:31:34,055 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=40343.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 20:31:35,035 INFO [finetune.py:976] (1/7) Epoch 8, batch 250, loss[loss=0.225, simple_loss=0.2944, pruned_loss=0.07782, over 4801.00 frames. ], tot_loss[loss=0.1986, simple_loss=0.2613, pruned_loss=0.06791, over 686592.47 frames. ], batch size: 41, lr: 3.85e-03, grad_scale: 32.0 2023-04-26 20:32:05,768 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=40391.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:32:07,909 INFO [finetune.py:976] (1/7) Epoch 8, batch 300, loss[loss=0.192, simple_loss=0.2339, pruned_loss=0.07505, over 4127.00 frames. ], tot_loss[loss=0.2001, simple_loss=0.2643, pruned_loss=0.06796, over 745877.38 frames. ], batch size: 17, lr: 3.85e-03, grad_scale: 32.0 2023-04-26 20:32:27,221 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=40423.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:32:31,997 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.115e+02 1.648e+02 2.026e+02 2.422e+02 3.959e+02, threshold=4.051e+02, percent-clipped=0.0 2023-04-26 20:32:32,253 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.49 vs. limit=2.0 2023-04-26 20:32:45,857 INFO [finetune.py:976] (1/7) Epoch 8, batch 350, loss[loss=0.1857, simple_loss=0.2604, pruned_loss=0.05546, over 4842.00 frames. ], tot_loss[loss=0.2006, simple_loss=0.2652, pruned_loss=0.06806, over 790793.84 frames. ], batch size: 49, lr: 3.85e-03, grad_scale: 32.0 2023-04-26 20:33:27,709 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=40471.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:33:27,757 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=40471.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:33:52,004 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=40492.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:33:53,120 INFO [finetune.py:976] (1/7) Epoch 8, batch 400, loss[loss=0.1731, simple_loss=0.2471, pruned_loss=0.04955, over 4928.00 frames. ], tot_loss[loss=0.203, simple_loss=0.2676, pruned_loss=0.06916, over 828157.06 frames. ], batch size: 42, lr: 3.85e-03, grad_scale: 32.0 2023-04-26 20:34:32,780 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=40519.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:34:45,208 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.151e+02 1.623e+02 1.962e+02 2.378e+02 5.703e+02, threshold=3.923e+02, percent-clipped=1.0 2023-04-26 20:35:05,026 INFO [finetune.py:976] (1/7) Epoch 8, batch 450, loss[loss=0.2041, simple_loss=0.2647, pruned_loss=0.07168, over 4820.00 frames. ], tot_loss[loss=0.2015, simple_loss=0.2656, pruned_loss=0.06869, over 854764.47 frames. ], batch size: 38, lr: 3.85e-03, grad_scale: 32.0 2023-04-26 20:35:16,806 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=40551.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:35:18,063 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-04-26 20:35:18,618 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=40553.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:35:35,911 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.1057, 1.5294, 1.9884, 2.2005, 1.9107, 1.4935, 1.1897, 1.6769], device='cuda:1'), covar=tensor([0.3616, 0.3841, 0.1740, 0.2977, 0.3437, 0.3136, 0.5170, 0.2567], device='cuda:1'), in_proj_covar=tensor([0.0278, 0.0252, 0.0218, 0.0321, 0.0213, 0.0228, 0.0236, 0.0188], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-26 20:35:38,784 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=40574.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:35:56,422 INFO [finetune.py:976] (1/7) Epoch 8, batch 500, loss[loss=0.1662, simple_loss=0.2226, pruned_loss=0.05487, over 4751.00 frames. ], tot_loss[loss=0.2001, simple_loss=0.2639, pruned_loss=0.06816, over 878398.40 frames. ], batch size: 26, lr: 3.85e-03, grad_scale: 32.0 2023-04-26 20:36:37,616 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=40628.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:36:39,378 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.145e+02 1.684e+02 2.034e+02 2.425e+02 5.158e+02, threshold=4.068e+02, percent-clipped=3.0 2023-04-26 20:36:52,501 INFO [finetune.py:976] (1/7) Epoch 8, batch 550, loss[loss=0.1849, simple_loss=0.2574, pruned_loss=0.05625, over 4903.00 frames. ], tot_loss[loss=0.1977, simple_loss=0.261, pruned_loss=0.06721, over 896362.86 frames. ], batch size: 43, lr: 3.85e-03, grad_scale: 32.0 2023-04-26 20:36:56,967 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=40651.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:37:47,168 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=40686.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:37:52,034 INFO [finetune.py:976] (1/7) Epoch 8, batch 600, loss[loss=0.1918, simple_loss=0.2611, pruned_loss=0.06124, over 4784.00 frames. ], tot_loss[loss=0.1989, simple_loss=0.2621, pruned_loss=0.06787, over 910996.53 frames. ], batch size: 25, lr: 3.84e-03, grad_scale: 32.0 2023-04-26 20:38:21,439 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=40712.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:38:22,125 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.61 vs. limit=2.0 2023-04-26 20:38:44,482 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.285e+02 1.754e+02 2.110e+02 2.532e+02 4.405e+02, threshold=4.220e+02, percent-clipped=2.0 2023-04-26 20:38:54,536 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5526, 1.4078, 1.8645, 1.8387, 1.4086, 1.1773, 1.5482, 1.0050], device='cuda:1'), covar=tensor([0.0738, 0.0966, 0.0473, 0.0784, 0.1036, 0.1489, 0.0851, 0.0999], device='cuda:1'), in_proj_covar=tensor([0.0066, 0.0074, 0.0072, 0.0067, 0.0076, 0.0096, 0.0080, 0.0075], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-26 20:39:02,720 INFO [finetune.py:976] (1/7) Epoch 8, batch 650, loss[loss=0.195, simple_loss=0.2724, pruned_loss=0.05875, over 4807.00 frames. ], tot_loss[loss=0.2007, simple_loss=0.2649, pruned_loss=0.06823, over 921087.25 frames. ], batch size: 51, lr: 3.84e-03, grad_scale: 32.0 2023-04-26 20:39:14,004 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-04-26 20:39:36,202 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.3755, 1.6605, 1.3197, 0.9868, 1.0752, 1.0665, 1.2746, 1.0174], device='cuda:1'), covar=tensor([0.1860, 0.1523, 0.1797, 0.2188, 0.2764, 0.2231, 0.1216, 0.2274], device='cuda:1'), in_proj_covar=tensor([0.0195, 0.0217, 0.0173, 0.0205, 0.0206, 0.0186, 0.0163, 0.0189], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-26 20:40:08,310 INFO [finetune.py:976] (1/7) Epoch 8, batch 700, loss[loss=0.1736, simple_loss=0.215, pruned_loss=0.06609, over 4168.00 frames. ], tot_loss[loss=0.2003, simple_loss=0.2648, pruned_loss=0.06786, over 926176.30 frames. ], batch size: 18, lr: 3.84e-03, grad_scale: 32.0 2023-04-26 20:40:55,215 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.166e+02 1.764e+02 2.066e+02 2.587e+02 3.741e+02, threshold=4.132e+02, percent-clipped=0.0 2023-04-26 20:41:14,342 INFO [finetune.py:976] (1/7) Epoch 8, batch 750, loss[loss=0.2308, simple_loss=0.3026, pruned_loss=0.07954, over 4847.00 frames. ], tot_loss[loss=0.2021, simple_loss=0.2667, pruned_loss=0.06875, over 931873.91 frames. ], batch size: 44, lr: 3.84e-03, grad_scale: 32.0 2023-04-26 20:41:16,846 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=40848.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:41:17,643 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.81 vs. limit=2.0 2023-04-26 20:41:24,321 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=40851.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:41:38,681 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=40865.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:41:44,666 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=40874.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:41:57,692 INFO [finetune.py:976] (1/7) Epoch 8, batch 800, loss[loss=0.1632, simple_loss=0.2362, pruned_loss=0.04514, over 4827.00 frames. ], tot_loss[loss=0.2027, simple_loss=0.2677, pruned_loss=0.06891, over 937357.42 frames. ], batch size: 47, lr: 3.84e-03, grad_scale: 32.0 2023-04-26 20:42:00,808 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=40899.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:42:01,459 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=40900.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:42:16,688 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=40922.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:42:19,179 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=40926.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:42:20,382 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=40928.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:42:22,146 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.118e+02 1.615e+02 1.856e+02 2.266e+02 4.587e+02, threshold=3.712e+02, percent-clipped=2.0 2023-04-26 20:42:31,004 INFO [finetune.py:976] (1/7) Epoch 8, batch 850, loss[loss=0.1942, simple_loss=0.2575, pruned_loss=0.06546, over 4928.00 frames. ], tot_loss[loss=0.2015, simple_loss=0.266, pruned_loss=0.06854, over 941802.22 frames. ], batch size: 38, lr: 3.84e-03, grad_scale: 32.0 2023-04-26 20:42:41,364 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=40961.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 20:42:48,396 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-04-26 20:42:52,268 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=40976.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:42:58,797 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=40986.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:43:03,444 INFO [finetune.py:976] (1/7) Epoch 8, batch 900, loss[loss=0.1892, simple_loss=0.2448, pruned_loss=0.06682, over 4802.00 frames. ], tot_loss[loss=0.199, simple_loss=0.2631, pruned_loss=0.06745, over 945402.28 frames. ], batch size: 29, lr: 3.84e-03, grad_scale: 32.0 2023-04-26 20:43:05,770 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=40997.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:43:12,067 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=41007.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:43:28,962 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.199e+02 1.671e+02 1.992e+02 2.459e+02 6.072e+02, threshold=3.985e+02, percent-clipped=2.0 2023-04-26 20:43:30,875 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=41034.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:43:37,319 INFO [finetune.py:976] (1/7) Epoch 8, batch 950, loss[loss=0.1696, simple_loss=0.2394, pruned_loss=0.04987, over 4908.00 frames. ], tot_loss[loss=0.1985, simple_loss=0.2621, pruned_loss=0.06745, over 945976.73 frames. ], batch size: 37, lr: 3.84e-03, grad_scale: 32.0 2023-04-26 20:43:38,039 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=41045.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:43:46,386 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=41058.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:44:10,578 INFO [finetune.py:976] (1/7) Epoch 8, batch 1000, loss[loss=0.1793, simple_loss=0.2389, pruned_loss=0.05987, over 4903.00 frames. ], tot_loss[loss=0.2007, simple_loss=0.2647, pruned_loss=0.06836, over 950009.94 frames. ], batch size: 32, lr: 3.84e-03, grad_scale: 32.0 2023-04-26 20:44:16,008 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=41102.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:44:18,373 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=41106.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:44:23,864 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.1692, 1.8674, 2.1360, 2.5235, 2.4221, 2.0994, 1.7201, 2.1610], device='cuda:1'), covar=tensor([0.0866, 0.1107, 0.0637, 0.0613, 0.0694, 0.0879, 0.0895, 0.0646], device='cuda:1'), in_proj_covar=tensor([0.0200, 0.0206, 0.0183, 0.0180, 0.0179, 0.0193, 0.0163, 0.0188], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 20:44:35,824 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.105e+02 1.783e+02 2.181e+02 2.591e+02 5.810e+02, threshold=4.362e+02, percent-clipped=3.0 2023-04-26 20:44:39,575 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.1668, 2.6494, 1.9670, 2.1722, 1.7031, 2.0153, 2.4190, 1.7598], device='cuda:1'), covar=tensor([0.2202, 0.1760, 0.1411, 0.1954, 0.3229, 0.1749, 0.1783, 0.2616], device='cuda:1'), in_proj_covar=tensor([0.0302, 0.0323, 0.0234, 0.0296, 0.0320, 0.0275, 0.0263, 0.0287], device='cuda:1'), out_proj_covar=tensor([1.2271e-04, 1.3066e-04, 9.4635e-05, 1.1864e-04, 1.3138e-04, 1.1117e-04, 1.0805e-04, 1.1523e-04], device='cuda:1') 2023-04-26 20:44:44,111 INFO [finetune.py:976] (1/7) Epoch 8, batch 1050, loss[loss=0.2633, simple_loss=0.3265, pruned_loss=0.1, over 4823.00 frames. ], tot_loss[loss=0.2032, simple_loss=0.2682, pruned_loss=0.0691, over 951920.85 frames. ], batch size: 47, lr: 3.84e-03, grad_scale: 32.0 2023-04-26 20:44:46,604 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=41148.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:44:56,176 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=41163.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:45:27,977 INFO [finetune.py:976] (1/7) Epoch 8, batch 1100, loss[loss=0.2137, simple_loss=0.2856, pruned_loss=0.07096, over 4810.00 frames. ], tot_loss[loss=0.2041, simple_loss=0.2688, pruned_loss=0.06968, over 953191.31 frames. ], batch size: 39, lr: 3.84e-03, grad_scale: 32.0 2023-04-26 20:45:29,272 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=41196.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:45:50,742 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=41221.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:45:58,234 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.055e+02 1.713e+02 2.125e+02 2.514e+02 5.728e+02, threshold=4.249e+02, percent-clipped=2.0 2023-04-26 20:46:17,813 INFO [finetune.py:976] (1/7) Epoch 8, batch 1150, loss[loss=0.2054, simple_loss=0.2795, pruned_loss=0.06564, over 4853.00 frames. ], tot_loss[loss=0.2033, simple_loss=0.2685, pruned_loss=0.0691, over 953746.10 frames. ], batch size: 49, lr: 3.84e-03, grad_scale: 32.0 2023-04-26 20:46:36,333 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=41256.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 20:47:11,487 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.60 vs. limit=2.0 2023-04-26 20:47:24,548 INFO [finetune.py:976] (1/7) Epoch 8, batch 1200, loss[loss=0.2216, simple_loss=0.2781, pruned_loss=0.08255, over 4814.00 frames. ], tot_loss[loss=0.2013, simple_loss=0.266, pruned_loss=0.0683, over 954559.56 frames. ], batch size: 33, lr: 3.84e-03, grad_scale: 32.0 2023-04-26 20:47:44,102 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=41307.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:47:59,148 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.011e+02 1.672e+02 1.949e+02 2.366e+02 5.490e+02, threshold=3.899e+02, percent-clipped=2.0 2023-04-26 20:48:08,614 INFO [finetune.py:976] (1/7) Epoch 8, batch 1250, loss[loss=0.181, simple_loss=0.2406, pruned_loss=0.06071, over 4744.00 frames. ], tot_loss[loss=0.1981, simple_loss=0.2622, pruned_loss=0.06704, over 954039.69 frames. ], batch size: 23, lr: 3.84e-03, grad_scale: 32.0 2023-04-26 20:48:14,608 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=41353.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:48:15,817 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=41355.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:48:36,652 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.3039, 3.2079, 2.5624, 2.6470, 2.1975, 2.5274, 2.5988, 2.1066], device='cuda:1'), covar=tensor([0.2607, 0.1341, 0.0944, 0.1519, 0.3192, 0.1403, 0.2351, 0.3136], device='cuda:1'), in_proj_covar=tensor([0.0301, 0.0321, 0.0233, 0.0294, 0.0319, 0.0273, 0.0262, 0.0285], device='cuda:1'), out_proj_covar=tensor([1.2208e-04, 1.2957e-04, 9.4091e-05, 1.1761e-04, 1.3090e-04, 1.1030e-04, 1.0727e-04, 1.1450e-04], device='cuda:1') 2023-04-26 20:48:42,247 INFO [finetune.py:976] (1/7) Epoch 8, batch 1300, loss[loss=0.1779, simple_loss=0.2361, pruned_loss=0.05982, over 4809.00 frames. ], tot_loss[loss=0.1952, simple_loss=0.2589, pruned_loss=0.06579, over 953986.22 frames. ], batch size: 45, lr: 3.84e-03, grad_scale: 32.0 2023-04-26 20:48:46,730 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=41401.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:49:05,847 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.125e+02 1.753e+02 2.019e+02 2.733e+02 6.111e+02, threshold=4.038e+02, percent-clipped=8.0 2023-04-26 20:49:15,122 INFO [finetune.py:976] (1/7) Epoch 8, batch 1350, loss[loss=0.2045, simple_loss=0.2728, pruned_loss=0.06814, over 4863.00 frames. ], tot_loss[loss=0.1963, simple_loss=0.2597, pruned_loss=0.06646, over 953498.85 frames. ], batch size: 34, lr: 3.84e-03, grad_scale: 32.0 2023-04-26 20:49:20,557 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.80 vs. limit=5.0 2023-04-26 20:49:23,864 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.0973, 1.4332, 1.3033, 1.6682, 1.4370, 1.5831, 1.3252, 2.9880], device='cuda:1'), covar=tensor([0.0714, 0.0755, 0.0789, 0.1151, 0.0670, 0.0596, 0.0747, 0.0200], device='cuda:1'), in_proj_covar=tensor([0.0039, 0.0039, 0.0041, 0.0044, 0.0041, 0.0040, 0.0039, 0.0060], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0014, 0.0013, 0.0013, 0.0013, 0.0017], device='cuda:1') 2023-04-26 20:49:25,036 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=41458.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:49:27,478 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.6753, 1.9239, 1.8656, 2.0803, 1.7883, 2.0277, 1.9927, 1.8837], device='cuda:1'), covar=tensor([0.6426, 0.9232, 0.7942, 0.6963, 0.8974, 1.1967, 1.0599, 0.9009], device='cuda:1'), in_proj_covar=tensor([0.0324, 0.0389, 0.0319, 0.0329, 0.0343, 0.0408, 0.0369, 0.0332], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-26 20:49:48,137 INFO [finetune.py:976] (1/7) Epoch 8, batch 1400, loss[loss=0.2111, simple_loss=0.2526, pruned_loss=0.08479, over 4062.00 frames. ], tot_loss[loss=0.199, simple_loss=0.2629, pruned_loss=0.06751, over 952215.13 frames. ], batch size: 17, lr: 3.84e-03, grad_scale: 64.0 2023-04-26 20:50:06,404 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=41521.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:50:10,643 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6533, 1.3579, 1.8717, 1.9076, 1.4154, 1.2404, 1.5732, 0.9725], device='cuda:1'), covar=tensor([0.0646, 0.1145, 0.0464, 0.0760, 0.0976, 0.1421, 0.0870, 0.0951], device='cuda:1'), in_proj_covar=tensor([0.0065, 0.0074, 0.0072, 0.0067, 0.0076, 0.0096, 0.0080, 0.0075], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-26 20:50:12,930 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.248e+02 1.632e+02 2.112e+02 2.588e+02 6.346e+02, threshold=4.225e+02, percent-clipped=4.0 2023-04-26 20:50:21,305 INFO [finetune.py:976] (1/7) Epoch 8, batch 1450, loss[loss=0.263, simple_loss=0.3154, pruned_loss=0.1053, over 4799.00 frames. ], tot_loss[loss=0.1987, simple_loss=0.2634, pruned_loss=0.06693, over 953124.27 frames. ], batch size: 51, lr: 3.84e-03, grad_scale: 64.0 2023-04-26 20:50:30,574 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=41556.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 20:50:30,725 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-04-26 20:50:38,925 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=41569.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:50:54,401 INFO [finetune.py:976] (1/7) Epoch 8, batch 1500, loss[loss=0.2251, simple_loss=0.2826, pruned_loss=0.08383, over 4924.00 frames. ], tot_loss[loss=0.1998, simple_loss=0.265, pruned_loss=0.06731, over 951784.09 frames. ], batch size: 33, lr: 3.84e-03, grad_scale: 64.0 2023-04-26 20:50:59,298 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6526, 1.2964, 1.7119, 2.0839, 1.7640, 1.6050, 1.6933, 1.7097], device='cuda:1'), covar=tensor([0.7460, 1.0193, 1.0586, 1.0418, 0.9182, 1.2609, 1.2372, 1.1278], device='cuda:1'), in_proj_covar=tensor([0.0408, 0.0426, 0.0507, 0.0529, 0.0437, 0.0456, 0.0467, 0.0466], device='cuda:1'), out_proj_covar=tensor([9.9199e-05, 1.0551e-04, 1.1469e-04, 1.2538e-04, 1.0615e-04, 1.1051e-04, 1.1254e-04, 1.1310e-04], device='cuda:1') 2023-04-26 20:51:02,366 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=41604.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:51:11,869 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7296, 2.0923, 1.0464, 1.4184, 2.1287, 1.6279, 1.5488, 1.6289], device='cuda:1'), covar=tensor([0.0547, 0.0357, 0.0385, 0.0598, 0.0278, 0.0591, 0.0541, 0.0606], device='cuda:1'), in_proj_covar=tensor([0.0030, 0.0026, 0.0024, 0.0030, 0.0021, 0.0030, 0.0030, 0.0030], device='cuda:1'), out_proj_covar=tensor([0.0050, 0.0045, 0.0038, 0.0050, 0.0038, 0.0049, 0.0048, 0.0050], device='cuda:1') 2023-04-26 20:51:25,118 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.252e+02 1.649e+02 2.032e+02 2.388e+02 6.025e+02, threshold=4.063e+02, percent-clipped=2.0 2023-04-26 20:51:37,136 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.1602, 2.4963, 1.1593, 1.3539, 2.0730, 1.2029, 3.5801, 1.8284], device='cuda:1'), covar=tensor([0.0702, 0.0853, 0.0844, 0.1448, 0.0531, 0.1102, 0.0377, 0.0684], device='cuda:1'), in_proj_covar=tensor([0.0053, 0.0069, 0.0051, 0.0049, 0.0053, 0.0053, 0.0080, 0.0052], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0011, 0.0008], device='cuda:1') 2023-04-26 20:51:38,880 INFO [finetune.py:976] (1/7) Epoch 8, batch 1550, loss[loss=0.1829, simple_loss=0.2503, pruned_loss=0.05777, over 4747.00 frames. ], tot_loss[loss=0.1989, simple_loss=0.2642, pruned_loss=0.06677, over 951345.68 frames. ], batch size: 27, lr: 3.84e-03, grad_scale: 64.0 2023-04-26 20:51:48,297 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7339, 1.4146, 1.8260, 2.1634, 1.8761, 1.7249, 1.8068, 1.8354], device='cuda:1'), covar=tensor([0.6796, 0.9104, 0.9134, 0.8931, 0.7795, 1.1045, 1.0793, 0.9667], device='cuda:1'), in_proj_covar=tensor([0.0406, 0.0424, 0.0505, 0.0526, 0.0435, 0.0454, 0.0465, 0.0464], device='cuda:1'), out_proj_covar=tensor([9.8680e-05, 1.0509e-04, 1.1414e-04, 1.2488e-04, 1.0580e-04, 1.1004e-04, 1.1201e-04, 1.1269e-04], device='cuda:1') 2023-04-26 20:51:51,052 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=41653.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:51:58,451 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=41654.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:52:47,105 INFO [finetune.py:976] (1/7) Epoch 8, batch 1600, loss[loss=0.2126, simple_loss=0.2625, pruned_loss=0.08137, over 4824.00 frames. ], tot_loss[loss=0.1974, simple_loss=0.2622, pruned_loss=0.06635, over 952781.50 frames. ], batch size: 33, lr: 3.84e-03, grad_scale: 64.0 2023-04-26 20:52:56,994 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=41701.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:52:57,021 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=41701.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:52:57,635 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.1792, 2.5198, 1.1309, 1.4445, 1.9809, 1.2223, 3.5481, 1.8912], device='cuda:1'), covar=tensor([0.0628, 0.0749, 0.0788, 0.1256, 0.0511, 0.1012, 0.0331, 0.0633], device='cuda:1'), in_proj_covar=tensor([0.0053, 0.0069, 0.0051, 0.0048, 0.0053, 0.0053, 0.0080, 0.0052], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0011, 0.0008], device='cuda:1') 2023-04-26 20:53:10,575 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.63 vs. limit=5.0 2023-04-26 20:53:12,230 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=41715.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 20:53:15,166 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7207, 1.3910, 1.7907, 2.0986, 1.8999, 1.6400, 1.7006, 1.7211], device='cuda:1'), covar=tensor([0.6899, 0.9953, 0.9592, 0.9578, 0.7632, 1.2185, 1.2076, 1.0709], device='cuda:1'), in_proj_covar=tensor([0.0407, 0.0425, 0.0505, 0.0528, 0.0436, 0.0455, 0.0467, 0.0465], device='cuda:1'), out_proj_covar=tensor([9.9046e-05, 1.0530e-04, 1.1431e-04, 1.2518e-04, 1.0611e-04, 1.1019e-04, 1.1235e-04, 1.1291e-04], device='cuda:1') 2023-04-26 20:53:20,611 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.7402, 1.9070, 1.8633, 2.1277, 1.8784, 2.0832, 1.9163, 1.8618], device='cuda:1'), covar=tensor([0.5866, 0.9840, 0.7749, 0.6239, 0.7743, 1.0894, 1.0011, 0.8806], device='cuda:1'), in_proj_covar=tensor([0.0321, 0.0386, 0.0316, 0.0327, 0.0341, 0.0405, 0.0367, 0.0329], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-26 20:53:21,822 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.2576, 1.6211, 2.0775, 2.5396, 2.0294, 1.6023, 1.2308, 1.8489], device='cuda:1'), covar=tensor([0.4147, 0.3952, 0.1939, 0.2974, 0.3437, 0.3336, 0.5045, 0.2593], device='cuda:1'), in_proj_covar=tensor([0.0282, 0.0253, 0.0219, 0.0323, 0.0215, 0.0230, 0.0237, 0.0188], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-26 20:53:22,287 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.077e+02 1.620e+02 1.919e+02 2.261e+02 4.784e+02, threshold=3.838e+02, percent-clipped=1.0 2023-04-26 20:53:30,671 INFO [finetune.py:976] (1/7) Epoch 8, batch 1650, loss[loss=0.1619, simple_loss=0.2243, pruned_loss=0.0498, over 4902.00 frames. ], tot_loss[loss=0.1947, simple_loss=0.259, pruned_loss=0.06518, over 951867.01 frames. ], batch size: 32, lr: 3.84e-03, grad_scale: 64.0 2023-04-26 20:53:33,794 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=41749.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:53:40,311 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=41758.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:53:58,741 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5815, 1.9934, 1.5024, 1.2235, 1.1594, 1.2149, 1.5576, 1.1242], device='cuda:1'), covar=tensor([0.1836, 0.1605, 0.1721, 0.2181, 0.2804, 0.2278, 0.1225, 0.2356], device='cuda:1'), in_proj_covar=tensor([0.0195, 0.0219, 0.0173, 0.0205, 0.0207, 0.0186, 0.0164, 0.0190], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-26 20:54:03,919 INFO [finetune.py:976] (1/7) Epoch 8, batch 1700, loss[loss=0.1351, simple_loss=0.2093, pruned_loss=0.03042, over 4799.00 frames. ], tot_loss[loss=0.1951, simple_loss=0.259, pruned_loss=0.06562, over 954676.28 frames. ], batch size: 26, lr: 3.84e-03, grad_scale: 64.0 2023-04-26 20:54:11,112 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-04-26 20:54:11,550 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=41806.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:54:26,447 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.50 vs. limit=2.0 2023-04-26 20:54:29,248 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.116e+02 1.826e+02 2.142e+02 2.567e+02 5.646e+02, threshold=4.284e+02, percent-clipped=4.0 2023-04-26 20:54:33,667 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.1620, 1.5915, 2.1181, 2.6791, 2.0568, 1.6751, 1.5217, 1.9424], device='cuda:1'), covar=tensor([0.3828, 0.4112, 0.1897, 0.3136, 0.3505, 0.3338, 0.4839, 0.2791], device='cuda:1'), in_proj_covar=tensor([0.0282, 0.0254, 0.0220, 0.0324, 0.0216, 0.0230, 0.0237, 0.0189], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-26 20:54:37,045 INFO [finetune.py:976] (1/7) Epoch 8, batch 1750, loss[loss=0.2413, simple_loss=0.3028, pruned_loss=0.08992, over 4876.00 frames. ], tot_loss[loss=0.1955, simple_loss=0.2596, pruned_loss=0.06573, over 954871.06 frames. ], batch size: 34, lr: 3.84e-03, grad_scale: 64.0 2023-04-26 20:55:10,268 INFO [finetune.py:976] (1/7) Epoch 8, batch 1800, loss[loss=0.2006, simple_loss=0.2707, pruned_loss=0.06525, over 4896.00 frames. ], tot_loss[loss=0.1996, simple_loss=0.2641, pruned_loss=0.06756, over 955738.62 frames. ], batch size: 35, lr: 3.84e-03, grad_scale: 64.0 2023-04-26 20:55:18,141 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=41906.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:55:31,466 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([0.4231, 1.3981, 1.4235, 1.1053, 1.4157, 1.1727, 1.7898, 1.3442], device='cuda:1'), covar=tensor([0.3689, 0.1876, 0.5278, 0.2754, 0.1692, 0.2385, 0.1883, 0.5030], device='cuda:1'), in_proj_covar=tensor([0.0350, 0.0353, 0.0436, 0.0365, 0.0392, 0.0386, 0.0385, 0.0421], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 20:55:35,587 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.171e+02 1.841e+02 2.087e+02 2.578e+02 5.698e+02, threshold=4.173e+02, percent-clipped=3.0 2023-04-26 20:55:43,854 INFO [finetune.py:976] (1/7) Epoch 8, batch 1850, loss[loss=0.1647, simple_loss=0.2427, pruned_loss=0.04341, over 4776.00 frames. ], tot_loss[loss=0.2013, simple_loss=0.266, pruned_loss=0.06835, over 955823.66 frames. ], batch size: 28, lr: 3.84e-03, grad_scale: 64.0 2023-04-26 20:55:46,384 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=41948.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 20:55:50,624 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.49 vs. limit=2.0 2023-04-26 20:55:58,818 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=41967.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:56:17,000 INFO [finetune.py:976] (1/7) Epoch 8, batch 1900, loss[loss=0.1921, simple_loss=0.2605, pruned_loss=0.06186, over 4730.00 frames. ], tot_loss[loss=0.2028, simple_loss=0.2678, pruned_loss=0.0689, over 955465.23 frames. ], batch size: 59, lr: 3.84e-03, grad_scale: 32.0 2023-04-26 20:56:21,980 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9994, 2.2893, 0.9695, 1.2227, 1.6701, 1.1628, 2.9682, 1.5138], device='cuda:1'), covar=tensor([0.0733, 0.0714, 0.0839, 0.1530, 0.0569, 0.1145, 0.0449, 0.0781], device='cuda:1'), in_proj_covar=tensor([0.0052, 0.0068, 0.0050, 0.0048, 0.0052, 0.0053, 0.0080, 0.0052], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0011, 0.0008], device='cuda:1') 2023-04-26 20:56:27,978 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=42009.0, num_to_drop=1, layers_to_drop={3} 2023-04-26 20:56:28,519 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=42010.0, num_to_drop=1, layers_to_drop={2} 2023-04-26 20:56:49,879 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.129e+02 1.688e+02 2.018e+02 2.426e+02 3.814e+02, threshold=4.036e+02, percent-clipped=0.0 2023-04-26 20:57:08,688 INFO [finetune.py:976] (1/7) Epoch 8, batch 1950, loss[loss=0.1477, simple_loss=0.2116, pruned_loss=0.04184, over 4217.00 frames. ], tot_loss[loss=0.2002, simple_loss=0.2653, pruned_loss=0.06749, over 955648.31 frames. ], batch size: 65, lr: 3.84e-03, grad_scale: 32.0 2023-04-26 20:57:10,604 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6733, 1.4724, 1.6627, 2.0792, 2.0088, 1.6886, 1.4116, 1.8177], device='cuda:1'), covar=tensor([0.0813, 0.1189, 0.0723, 0.0531, 0.0580, 0.0764, 0.0756, 0.0558], device='cuda:1'), in_proj_covar=tensor([0.0200, 0.0207, 0.0183, 0.0180, 0.0180, 0.0193, 0.0162, 0.0189], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 20:58:13,786 INFO [finetune.py:976] (1/7) Epoch 8, batch 2000, loss[loss=0.2167, simple_loss=0.2684, pruned_loss=0.08251, over 4249.00 frames. ], tot_loss[loss=0.1993, simple_loss=0.2638, pruned_loss=0.06737, over 956533.12 frames. ], batch size: 66, lr: 3.84e-03, grad_scale: 32.0 2023-04-26 20:59:06,781 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.093e+02 1.619e+02 1.918e+02 2.281e+02 4.896e+02, threshold=3.837e+02, percent-clipped=1.0 2023-04-26 20:59:17,303 INFO [finetune.py:976] (1/7) Epoch 8, batch 2050, loss[loss=0.2435, simple_loss=0.3134, pruned_loss=0.08679, over 4853.00 frames. ], tot_loss[loss=0.196, simple_loss=0.2603, pruned_loss=0.06586, over 956794.34 frames. ], batch size: 47, lr: 3.84e-03, grad_scale: 32.0 2023-04-26 20:59:28,157 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4068, 1.3904, 4.0789, 3.8042, 3.5542, 3.7839, 3.7576, 3.5109], device='cuda:1'), covar=tensor([0.7460, 0.5848, 0.1142, 0.1813, 0.1274, 0.2382, 0.1825, 0.1516], device='cuda:1'), in_proj_covar=tensor([0.0312, 0.0308, 0.0409, 0.0415, 0.0351, 0.0408, 0.0318, 0.0370], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-26 21:00:12,508 INFO [finetune.py:976] (1/7) Epoch 8, batch 2100, loss[loss=0.1651, simple_loss=0.2408, pruned_loss=0.04466, over 4828.00 frames. ], tot_loss[loss=0.197, simple_loss=0.2608, pruned_loss=0.06662, over 957407.35 frames. ], batch size: 33, lr: 3.83e-03, grad_scale: 32.0 2023-04-26 21:00:38,854 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.172e+02 1.610e+02 1.994e+02 2.459e+02 7.557e+02, threshold=3.988e+02, percent-clipped=2.0 2023-04-26 21:00:46,719 INFO [finetune.py:976] (1/7) Epoch 8, batch 2150, loss[loss=0.2064, simple_loss=0.2695, pruned_loss=0.07163, over 4896.00 frames. ], tot_loss[loss=0.1993, simple_loss=0.2636, pruned_loss=0.06751, over 957945.28 frames. ], batch size: 32, lr: 3.83e-03, grad_scale: 32.0 2023-04-26 21:00:56,461 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-04-26 21:00:58,115 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=42262.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 21:01:20,322 INFO [finetune.py:976] (1/7) Epoch 8, batch 2200, loss[loss=0.1869, simple_loss=0.2574, pruned_loss=0.05822, over 4818.00 frames. ], tot_loss[loss=0.1999, simple_loss=0.2647, pruned_loss=0.06756, over 956761.46 frames. ], batch size: 38, lr: 3.83e-03, grad_scale: 32.0 2023-04-26 21:01:26,906 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=42304.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 21:01:28,926 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2023-04-26 21:01:30,513 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=42310.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 21:01:44,806 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.211e+02 1.704e+02 2.094e+02 2.427e+02 3.760e+02, threshold=4.188e+02, percent-clipped=0.0 2023-04-26 21:01:52,460 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6539, 1.6207, 0.8869, 1.3196, 1.8646, 1.5501, 1.4535, 1.4867], device='cuda:1'), covar=tensor([0.0520, 0.0391, 0.0375, 0.0586, 0.0274, 0.0538, 0.0505, 0.0580], device='cuda:1'), in_proj_covar=tensor([0.0030, 0.0026, 0.0024, 0.0030, 0.0021, 0.0030, 0.0029, 0.0030], device='cuda:1'), out_proj_covar=tensor([0.0049, 0.0044, 0.0038, 0.0050, 0.0038, 0.0048, 0.0048, 0.0050], device='cuda:1') 2023-04-26 21:01:52,972 INFO [finetune.py:976] (1/7) Epoch 8, batch 2250, loss[loss=0.1588, simple_loss=0.2422, pruned_loss=0.0377, over 4882.00 frames. ], tot_loss[loss=0.2019, simple_loss=0.2665, pruned_loss=0.06862, over 956106.06 frames. ], batch size: 43, lr: 3.83e-03, grad_scale: 32.0 2023-04-26 21:01:55,424 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.3921, 1.3514, 1.7251, 1.6671, 1.3589, 1.0887, 1.4202, 1.0485], device='cuda:1'), covar=tensor([0.0748, 0.0681, 0.0476, 0.0664, 0.0822, 0.1309, 0.0721, 0.0803], device='cuda:1'), in_proj_covar=tensor([0.0066, 0.0073, 0.0072, 0.0067, 0.0076, 0.0096, 0.0080, 0.0075], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-26 21:02:02,511 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=42358.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 21:02:11,120 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.0830, 2.0127, 1.7235, 1.7802, 2.1459, 1.7480, 2.4808, 1.5194], device='cuda:1'), covar=tensor([0.3881, 0.1673, 0.5077, 0.2754, 0.1584, 0.2240, 0.1641, 0.4705], device='cuda:1'), in_proj_covar=tensor([0.0347, 0.0352, 0.0434, 0.0363, 0.0389, 0.0384, 0.0384, 0.0422], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 21:02:15,803 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=42379.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 21:02:24,540 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-04-26 21:02:26,180 INFO [finetune.py:976] (1/7) Epoch 8, batch 2300, loss[loss=0.1941, simple_loss=0.2638, pruned_loss=0.06216, over 4852.00 frames. ], tot_loss[loss=0.2018, simple_loss=0.267, pruned_loss=0.06833, over 955479.76 frames. ], batch size: 49, lr: 3.83e-03, grad_scale: 32.0 2023-04-26 21:02:50,915 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.077e+02 1.665e+02 1.908e+02 2.261e+02 6.563e+02, threshold=3.817e+02, percent-clipped=1.0 2023-04-26 21:02:51,726 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.81 vs. limit=2.0 2023-04-26 21:02:56,327 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=42440.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 21:02:57,386 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.3904, 1.6545, 1.6012, 1.8973, 1.7077, 1.9617, 1.4401, 3.6486], device='cuda:1'), covar=tensor([0.0635, 0.0805, 0.0794, 0.1182, 0.0665, 0.0501, 0.0785, 0.0146], device='cuda:1'), in_proj_covar=tensor([0.0039, 0.0039, 0.0040, 0.0044, 0.0041, 0.0040, 0.0039, 0.0059], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0014, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:1') 2023-04-26 21:02:58,701 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-04-26 21:02:59,687 INFO [finetune.py:976] (1/7) Epoch 8, batch 2350, loss[loss=0.2071, simple_loss=0.2717, pruned_loss=0.07126, over 4771.00 frames. ], tot_loss[loss=0.1983, simple_loss=0.2634, pruned_loss=0.06664, over 955550.88 frames. ], batch size: 26, lr: 3.83e-03, grad_scale: 32.0 2023-04-26 21:03:02,745 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.2657, 1.4721, 1.6491, 1.8246, 1.7039, 1.8154, 1.7597, 1.7162], device='cuda:1'), covar=tensor([0.5536, 0.7384, 0.6440, 0.5897, 0.7216, 1.0356, 0.7410, 0.7175], device='cuda:1'), in_proj_covar=tensor([0.0321, 0.0384, 0.0315, 0.0325, 0.0341, 0.0403, 0.0367, 0.0327], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-26 21:03:20,362 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-04-26 21:03:48,469 INFO [finetune.py:976] (1/7) Epoch 8, batch 2400, loss[loss=0.2225, simple_loss=0.2726, pruned_loss=0.08615, over 4295.00 frames. ], tot_loss[loss=0.1969, simple_loss=0.2611, pruned_loss=0.0664, over 955936.44 frames. ], batch size: 65, lr: 3.83e-03, grad_scale: 32.0 2023-04-26 21:04:08,551 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=42506.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 21:04:40,672 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.220e+02 1.697e+02 2.030e+02 2.440e+02 3.399e+02, threshold=4.060e+02, percent-clipped=0.0 2023-04-26 21:04:44,378 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6252, 1.5122, 1.9704, 1.9491, 1.5449, 1.2737, 1.6503, 1.0830], device='cuda:1'), covar=tensor([0.0618, 0.0805, 0.0484, 0.0719, 0.0864, 0.1210, 0.0757, 0.0966], device='cuda:1'), in_proj_covar=tensor([0.0066, 0.0073, 0.0072, 0.0067, 0.0076, 0.0096, 0.0079, 0.0075], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-26 21:04:54,582 INFO [finetune.py:976] (1/7) Epoch 8, batch 2450, loss[loss=0.1851, simple_loss=0.2655, pruned_loss=0.05236, over 4898.00 frames. ], tot_loss[loss=0.1938, simple_loss=0.2578, pruned_loss=0.0649, over 956230.35 frames. ], batch size: 32, lr: 3.83e-03, grad_scale: 32.0 2023-04-26 21:04:55,309 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.1243, 1.3665, 1.2387, 1.7133, 1.4805, 1.6436, 1.2950, 2.3932], device='cuda:1'), covar=tensor([0.0663, 0.0811, 0.0829, 0.1180, 0.0680, 0.0473, 0.0762, 0.0244], device='cuda:1'), in_proj_covar=tensor([0.0039, 0.0039, 0.0040, 0.0043, 0.0040, 0.0039, 0.0039, 0.0059], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0014, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:1') 2023-04-26 21:05:24,351 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=42562.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 21:05:26,768 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.1952, 1.3878, 1.7233, 1.8628, 1.7519, 1.9497, 1.7586, 1.7631], device='cuda:1'), covar=tensor([0.5187, 0.7106, 0.6080, 0.5933, 0.7283, 0.9436, 0.6764, 0.6311], device='cuda:1'), in_proj_covar=tensor([0.0323, 0.0387, 0.0317, 0.0328, 0.0343, 0.0407, 0.0369, 0.0329], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-26 21:05:27,980 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=42567.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 21:05:37,438 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([0.6119, 1.6147, 1.5992, 1.2966, 1.7589, 1.4298, 2.3065, 1.3825], device='cuda:1'), covar=tensor([0.4158, 0.1752, 0.4454, 0.3076, 0.1702, 0.2395, 0.1322, 0.4777], device='cuda:1'), in_proj_covar=tensor([0.0351, 0.0354, 0.0437, 0.0368, 0.0392, 0.0387, 0.0387, 0.0425], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 21:05:58,926 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-04-26 21:06:00,482 INFO [finetune.py:976] (1/7) Epoch 8, batch 2500, loss[loss=0.2111, simple_loss=0.2775, pruned_loss=0.07235, over 4911.00 frames. ], tot_loss[loss=0.1981, simple_loss=0.2616, pruned_loss=0.06728, over 956076.80 frames. ], batch size: 37, lr: 3.83e-03, grad_scale: 32.0 2023-04-26 21:06:20,335 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=42604.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 21:06:29,255 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=42610.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 21:06:54,510 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.169e+02 1.868e+02 2.156e+02 2.482e+02 4.292e+02, threshold=4.312e+02, percent-clipped=2.0 2023-04-26 21:07:08,191 INFO [finetune.py:976] (1/7) Epoch 8, batch 2550, loss[loss=0.182, simple_loss=0.2556, pruned_loss=0.05418, over 4901.00 frames. ], tot_loss[loss=0.1998, simple_loss=0.264, pruned_loss=0.06774, over 956888.44 frames. ], batch size: 32, lr: 3.83e-03, grad_scale: 32.0 2023-04-26 21:07:11,910 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=42649.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 21:07:14,665 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=42652.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 21:07:17,858 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.2351, 1.4984, 1.5731, 1.7357, 1.5653, 1.7265, 1.6809, 1.6206], device='cuda:1'), covar=tensor([0.6118, 0.8255, 0.6717, 0.5875, 0.8123, 1.0945, 0.8114, 0.7327], device='cuda:1'), in_proj_covar=tensor([0.0322, 0.0385, 0.0316, 0.0327, 0.0342, 0.0405, 0.0368, 0.0328], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-26 21:07:41,702 INFO [finetune.py:976] (1/7) Epoch 8, batch 2600, loss[loss=0.2049, simple_loss=0.2708, pruned_loss=0.0695, over 4831.00 frames. ], tot_loss[loss=0.2003, simple_loss=0.2649, pruned_loss=0.06787, over 954103.83 frames. ], batch size: 49, lr: 3.83e-03, grad_scale: 32.0 2023-04-26 21:07:52,439 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9074, 1.8570, 2.3356, 2.3518, 1.8417, 1.5478, 1.8999, 1.1168], device='cuda:1'), covar=tensor([0.0871, 0.0895, 0.0546, 0.1127, 0.0880, 0.1280, 0.0908, 0.1099], device='cuda:1'), in_proj_covar=tensor([0.0066, 0.0073, 0.0072, 0.0067, 0.0076, 0.0095, 0.0079, 0.0075], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-26 21:07:53,064 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=42710.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 21:08:07,498 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.967e+01 1.665e+02 1.922e+02 2.451e+02 5.774e+02, threshold=3.843e+02, percent-clipped=4.0 2023-04-26 21:08:09,403 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=42735.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 21:08:14,790 INFO [finetune.py:976] (1/7) Epoch 8, batch 2650, loss[loss=0.1979, simple_loss=0.2668, pruned_loss=0.0645, over 4772.00 frames. ], tot_loss[loss=0.2007, simple_loss=0.2659, pruned_loss=0.06774, over 956790.53 frames. ], batch size: 26, lr: 3.83e-03, grad_scale: 32.0 2023-04-26 21:08:41,214 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.67 vs. limit=2.0 2023-04-26 21:08:45,398 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4706, 1.4050, 4.0870, 3.8084, 3.5935, 3.8678, 3.8223, 3.6246], device='cuda:1'), covar=tensor([0.7001, 0.5772, 0.1149, 0.1824, 0.1238, 0.1793, 0.1625, 0.1549], device='cuda:1'), in_proj_covar=tensor([0.0311, 0.0309, 0.0412, 0.0416, 0.0353, 0.0409, 0.0319, 0.0372], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-26 21:08:48,362 INFO [finetune.py:976] (1/7) Epoch 8, batch 2700, loss[loss=0.1869, simple_loss=0.2536, pruned_loss=0.06009, over 4828.00 frames. ], tot_loss[loss=0.198, simple_loss=0.2638, pruned_loss=0.06613, over 957067.64 frames. ], batch size: 47, lr: 3.83e-03, grad_scale: 32.0 2023-04-26 21:09:14,556 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.600e+01 1.577e+02 1.842e+02 2.218e+02 2.993e+02, threshold=3.685e+02, percent-clipped=0.0 2023-04-26 21:09:21,935 INFO [finetune.py:976] (1/7) Epoch 8, batch 2750, loss[loss=0.1634, simple_loss=0.2323, pruned_loss=0.04726, over 4909.00 frames. ], tot_loss[loss=0.1943, simple_loss=0.26, pruned_loss=0.06431, over 956658.78 frames. ], batch size: 35, lr: 3.83e-03, grad_scale: 32.0 2023-04-26 21:09:24,498 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4242, 1.2077, 1.5549, 1.5185, 1.2857, 1.1280, 1.2803, 0.8507], device='cuda:1'), covar=tensor([0.0522, 0.0840, 0.0469, 0.0737, 0.0807, 0.1116, 0.0663, 0.0769], device='cuda:1'), in_proj_covar=tensor([0.0066, 0.0073, 0.0072, 0.0067, 0.0076, 0.0096, 0.0079, 0.0075], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-26 21:09:31,555 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=42859.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 21:09:34,270 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=42862.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 21:09:57,313 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=42879.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 21:09:58,009 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.68 vs. limit=5.0 2023-04-26 21:10:17,377 INFO [finetune.py:976] (1/7) Epoch 8, batch 2800, loss[loss=0.1729, simple_loss=0.2289, pruned_loss=0.05846, over 4827.00 frames. ], tot_loss[loss=0.1922, simple_loss=0.2571, pruned_loss=0.06362, over 958157.63 frames. ], batch size: 30, lr: 3.83e-03, grad_scale: 32.0 2023-04-26 21:10:51,610 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=42920.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 21:11:10,643 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.078e+02 1.665e+02 1.974e+02 2.383e+02 4.846e+02, threshold=3.948e+02, percent-clipped=3.0 2023-04-26 21:11:16,191 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=42940.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 21:11:17,596 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-04-26 21:11:18,534 INFO [finetune.py:976] (1/7) Epoch 8, batch 2850, loss[loss=0.1966, simple_loss=0.2729, pruned_loss=0.0602, over 4887.00 frames. ], tot_loss[loss=0.1926, simple_loss=0.2571, pruned_loss=0.0641, over 957980.58 frames. ], batch size: 32, lr: 3.83e-03, grad_scale: 32.0 2023-04-26 21:11:56,701 INFO [finetune.py:976] (1/7) Epoch 8, batch 2900, loss[loss=0.1982, simple_loss=0.2727, pruned_loss=0.06188, over 4151.00 frames. ], tot_loss[loss=0.1952, simple_loss=0.2601, pruned_loss=0.06518, over 958465.53 frames. ], batch size: 65, lr: 3.83e-03, grad_scale: 32.0 2023-04-26 21:12:03,835 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6042, 1.4828, 1.6686, 1.9880, 2.0229, 1.5616, 1.1617, 1.6928], device='cuda:1'), covar=tensor([0.0968, 0.1312, 0.0728, 0.0633, 0.0684, 0.0985, 0.1008, 0.0698], device='cuda:1'), in_proj_covar=tensor([0.0196, 0.0203, 0.0179, 0.0176, 0.0177, 0.0189, 0.0159, 0.0185], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 21:12:15,324 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=43005.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 21:12:49,754 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.3340, 1.2391, 4.0827, 3.7967, 3.6180, 3.8562, 3.7894, 3.6287], device='cuda:1'), covar=tensor([0.7235, 0.5872, 0.1091, 0.1728, 0.1094, 0.1832, 0.1705, 0.1379], device='cuda:1'), in_proj_covar=tensor([0.0313, 0.0311, 0.0413, 0.0418, 0.0354, 0.0409, 0.0319, 0.0374], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-26 21:12:51,485 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.213e+02 1.808e+02 2.091e+02 2.545e+02 3.466e+02, threshold=4.182e+02, percent-clipped=0.0 2023-04-26 21:12:59,360 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=43035.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 21:12:59,409 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9993, 1.0732, 1.4585, 1.6062, 1.5756, 1.7100, 1.5193, 1.5017], device='cuda:1'), covar=tensor([0.3882, 0.6406, 0.6062, 0.5390, 0.6604, 0.9421, 0.6212, 0.6262], device='cuda:1'), in_proj_covar=tensor([0.0323, 0.0384, 0.0316, 0.0326, 0.0342, 0.0405, 0.0366, 0.0327], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-26 21:13:01,768 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-04-26 21:13:10,703 INFO [finetune.py:976] (1/7) Epoch 8, batch 2950, loss[loss=0.2366, simple_loss=0.2946, pruned_loss=0.0893, over 4806.00 frames. ], tot_loss[loss=0.1999, simple_loss=0.2652, pruned_loss=0.06734, over 958317.48 frames. ], batch size: 45, lr: 3.83e-03, grad_scale: 32.0 2023-04-26 21:13:26,504 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7702, 4.0074, 0.8359, 2.2466, 2.2195, 2.8729, 2.4741, 1.0336], device='cuda:1'), covar=tensor([0.1321, 0.0875, 0.2186, 0.1284, 0.1067, 0.0942, 0.1342, 0.2209], device='cuda:1'), in_proj_covar=tensor([0.0119, 0.0255, 0.0144, 0.0125, 0.0137, 0.0157, 0.0122, 0.0123], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-26 21:13:28,843 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5092, 3.0495, 0.8889, 1.7198, 1.7630, 2.2811, 1.8735, 0.9974], device='cuda:1'), covar=tensor([0.1376, 0.1038, 0.2066, 0.1353, 0.1135, 0.0983, 0.1484, 0.1873], device='cuda:1'), in_proj_covar=tensor([0.0119, 0.0255, 0.0144, 0.0125, 0.0137, 0.0157, 0.0122, 0.0123], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-26 21:13:37,378 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=43083.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 21:13:44,490 INFO [finetune.py:976] (1/7) Epoch 8, batch 3000, loss[loss=0.2375, simple_loss=0.2915, pruned_loss=0.09169, over 4879.00 frames. ], tot_loss[loss=0.2014, simple_loss=0.2667, pruned_loss=0.06802, over 957742.29 frames. ], batch size: 35, lr: 3.83e-03, grad_scale: 32.0 2023-04-26 21:13:44,490 INFO [finetune.py:1001] (1/7) Computing validation loss 2023-04-26 21:13:47,354 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4095, 3.0235, 0.9215, 1.6261, 1.7643, 2.3069, 1.8211, 0.9471], device='cuda:1'), covar=tensor([0.1357, 0.0853, 0.1954, 0.1346, 0.1074, 0.0862, 0.1473, 0.1754], device='cuda:1'), in_proj_covar=tensor([0.0118, 0.0255, 0.0144, 0.0125, 0.0137, 0.0157, 0.0122, 0.0123], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-26 21:13:47,979 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4562, 1.4959, 3.8697, 3.5737, 3.5117, 3.7613, 3.7734, 3.4231], device='cuda:1'), covar=tensor([0.6847, 0.4822, 0.1194, 0.2033, 0.1292, 0.1399, 0.0833, 0.1600], device='cuda:1'), in_proj_covar=tensor([0.0311, 0.0309, 0.0411, 0.0415, 0.0351, 0.0407, 0.0318, 0.0372], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-26 21:13:54,963 INFO [finetune.py:1010] (1/7) Epoch 8, validation: loss=0.1551, simple_loss=0.2273, pruned_loss=0.04149, over 2265189.00 frames. 2023-04-26 21:13:54,964 INFO [finetune.py:1011] (1/7) Maximum memory allocated so far is 6391MB 2023-04-26 21:14:18,552 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.136e+02 1.720e+02 2.087e+02 2.508e+02 4.995e+02, threshold=4.174e+02, percent-clipped=1.0 2023-04-26 21:14:27,800 INFO [finetune.py:976] (1/7) Epoch 8, batch 3050, loss[loss=0.1928, simple_loss=0.2573, pruned_loss=0.0641, over 4819.00 frames. ], tot_loss[loss=0.2025, simple_loss=0.2683, pruned_loss=0.06832, over 958659.15 frames. ], batch size: 33, lr: 3.83e-03, grad_scale: 32.0 2023-04-26 21:14:28,981 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.3244, 2.9142, 2.3832, 2.6318, 2.0873, 2.4687, 2.6423, 1.9299], device='cuda:1'), covar=tensor([0.2592, 0.1626, 0.1040, 0.1551, 0.3777, 0.1627, 0.2136, 0.3159], device='cuda:1'), in_proj_covar=tensor([0.0305, 0.0325, 0.0235, 0.0298, 0.0322, 0.0278, 0.0264, 0.0290], device='cuda:1'), out_proj_covar=tensor([1.2374e-04, 1.3136e-04, 9.4960e-05, 1.1949e-04, 1.3250e-04, 1.1230e-04, 1.0830e-04, 1.1637e-04], device='cuda:1') 2023-04-26 21:14:37,229 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9971, 1.4303, 5.3051, 4.9781, 4.5755, 5.1553, 4.6778, 4.6363], device='cuda:1'), covar=tensor([0.6957, 0.6257, 0.1123, 0.1718, 0.1037, 0.1058, 0.1048, 0.1389], device='cuda:1'), in_proj_covar=tensor([0.0310, 0.0308, 0.0410, 0.0414, 0.0350, 0.0406, 0.0317, 0.0371], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-26 21:14:39,653 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=43162.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 21:15:00,984 INFO [finetune.py:976] (1/7) Epoch 8, batch 3100, loss[loss=0.1943, simple_loss=0.2553, pruned_loss=0.06662, over 4833.00 frames. ], tot_loss[loss=0.1986, simple_loss=0.2642, pruned_loss=0.06653, over 956163.68 frames. ], batch size: 30, lr: 3.83e-03, grad_scale: 32.0 2023-04-26 21:15:11,983 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=43210.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 21:15:15,120 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=43215.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 21:15:17,008 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6779, 1.2911, 1.3335, 1.3511, 1.8429, 1.4579, 1.1900, 1.2207], device='cuda:1'), covar=tensor([0.1735, 0.1464, 0.1994, 0.1336, 0.0847, 0.1480, 0.1900, 0.2100], device='cuda:1'), in_proj_covar=tensor([0.0309, 0.0324, 0.0354, 0.0301, 0.0340, 0.0325, 0.0310, 0.0354], device='cuda:1'), out_proj_covar=tensor([6.5580e-05, 6.8784e-05, 7.6547e-05, 6.2505e-05, 7.1669e-05, 6.9902e-05, 6.6784e-05, 7.6156e-05], device='cuda:1') 2023-04-26 21:15:25,431 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.200e+02 1.610e+02 1.868e+02 2.269e+02 4.698e+02, threshold=3.736e+02, percent-clipped=1.0 2023-04-26 21:15:27,879 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=43235.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 21:15:34,183 INFO [finetune.py:976] (1/7) Epoch 8, batch 3150, loss[loss=0.1938, simple_loss=0.2554, pruned_loss=0.06614, over 4891.00 frames. ], tot_loss[loss=0.1967, simple_loss=0.2614, pruned_loss=0.06602, over 957226.91 frames. ], batch size: 35, lr: 3.83e-03, grad_scale: 32.0 2023-04-26 21:15:53,957 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.13 vs. limit=5.0 2023-04-26 21:16:18,114 INFO [finetune.py:976] (1/7) Epoch 8, batch 3200, loss[loss=0.1913, simple_loss=0.2604, pruned_loss=0.06112, over 4905.00 frames. ], tot_loss[loss=0.1928, simple_loss=0.2568, pruned_loss=0.06441, over 954343.20 frames. ], batch size: 32, lr: 3.83e-03, grad_scale: 32.0 2023-04-26 21:16:30,947 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=43305.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 21:16:49,129 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.2257, 2.9883, 0.8536, 1.6480, 1.6799, 2.1393, 1.8048, 0.9114], device='cuda:1'), covar=tensor([0.1547, 0.0948, 0.2048, 0.1423, 0.1149, 0.1006, 0.1422, 0.1967], device='cuda:1'), in_proj_covar=tensor([0.0117, 0.0251, 0.0141, 0.0123, 0.0135, 0.0154, 0.0120, 0.0121], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-26 21:17:03,159 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.48 vs. limit=2.0 2023-04-26 21:17:03,559 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.128e+02 1.615e+02 2.007e+02 2.471e+02 4.400e+02, threshold=4.014e+02, percent-clipped=2.0 2023-04-26 21:17:14,619 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([3.3654, 3.2756, 2.3715, 3.8748, 3.4016, 3.3638, 1.3420, 3.2380], device='cuda:1'), covar=tensor([0.1896, 0.1667, 0.3474, 0.2373, 0.4253, 0.2293, 0.6153, 0.2962], device='cuda:1'), in_proj_covar=tensor([0.0241, 0.0215, 0.0247, 0.0306, 0.0296, 0.0248, 0.0268, 0.0269], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-26 21:17:20,615 INFO [finetune.py:976] (1/7) Epoch 8, batch 3250, loss[loss=0.184, simple_loss=0.2468, pruned_loss=0.06059, over 4797.00 frames. ], tot_loss[loss=0.1945, simple_loss=0.2584, pruned_loss=0.06536, over 954098.73 frames. ], batch size: 25, lr: 3.83e-03, grad_scale: 32.0 2023-04-26 21:17:33,417 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=43353.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 21:18:26,606 INFO [finetune.py:976] (1/7) Epoch 8, batch 3300, loss[loss=0.2361, simple_loss=0.3083, pruned_loss=0.08198, over 4851.00 frames. ], tot_loss[loss=0.1981, simple_loss=0.2627, pruned_loss=0.06672, over 955112.20 frames. ], batch size: 44, lr: 3.83e-03, grad_scale: 32.0 2023-04-26 21:18:59,198 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.201e+02 1.779e+02 2.083e+02 2.593e+02 4.743e+02, threshold=4.166e+02, percent-clipped=4.0 2023-04-26 21:19:06,534 INFO [finetune.py:976] (1/7) Epoch 8, batch 3350, loss[loss=0.1775, simple_loss=0.2396, pruned_loss=0.05767, over 4808.00 frames. ], tot_loss[loss=0.1986, simple_loss=0.2639, pruned_loss=0.06665, over 955017.46 frames. ], batch size: 25, lr: 3.83e-03, grad_scale: 32.0 2023-04-26 21:19:27,860 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.53 vs. limit=5.0 2023-04-26 21:19:40,384 INFO [finetune.py:976] (1/7) Epoch 8, batch 3400, loss[loss=0.2114, simple_loss=0.279, pruned_loss=0.07192, over 4863.00 frames. ], tot_loss[loss=0.1992, simple_loss=0.2648, pruned_loss=0.06682, over 954705.11 frames. ], batch size: 34, lr: 3.83e-03, grad_scale: 32.0 2023-04-26 21:19:54,922 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=43514.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 21:19:55,558 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=43515.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 21:20:06,649 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.125e+02 1.715e+02 2.177e+02 2.420e+02 5.380e+02, threshold=4.353e+02, percent-clipped=2.0 2023-04-26 21:20:08,562 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=43535.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 21:20:11,000 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([0.9475, 2.0005, 1.7735, 1.7597, 2.1490, 1.7290, 2.6616, 1.5197], device='cuda:1'), covar=tensor([0.4071, 0.1756, 0.4945, 0.3116, 0.1736, 0.2594, 0.1435, 0.4673], device='cuda:1'), in_proj_covar=tensor([0.0348, 0.0353, 0.0433, 0.0365, 0.0388, 0.0386, 0.0386, 0.0421], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 21:20:13,907 INFO [finetune.py:976] (1/7) Epoch 8, batch 3450, loss[loss=0.2146, simple_loss=0.2696, pruned_loss=0.07982, over 4815.00 frames. ], tot_loss[loss=0.1991, simple_loss=0.2649, pruned_loss=0.06665, over 954823.67 frames. ], batch size: 38, lr: 3.83e-03, grad_scale: 32.0 2023-04-26 21:20:26,910 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=43563.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 21:20:35,619 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=43575.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 21:20:40,952 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=43583.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 21:20:47,607 INFO [finetune.py:976] (1/7) Epoch 8, batch 3500, loss[loss=0.2179, simple_loss=0.2863, pruned_loss=0.07477, over 4796.00 frames. ], tot_loss[loss=0.197, simple_loss=0.2623, pruned_loss=0.06581, over 954977.94 frames. ], batch size: 29, lr: 3.83e-03, grad_scale: 32.0 2023-04-26 21:21:13,732 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.083e+02 1.646e+02 1.959e+02 2.415e+02 4.063e+02, threshold=3.918e+02, percent-clipped=0.0 2023-04-26 21:21:21,526 INFO [finetune.py:976] (1/7) Epoch 8, batch 3550, loss[loss=0.1899, simple_loss=0.2482, pruned_loss=0.06578, over 4898.00 frames. ], tot_loss[loss=0.1943, simple_loss=0.2591, pruned_loss=0.06475, over 953934.98 frames. ], batch size: 36, lr: 3.82e-03, grad_scale: 32.0 2023-04-26 21:21:33,859 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7326, 1.5727, 3.9351, 3.7071, 3.4755, 3.5657, 3.5761, 3.5533], device='cuda:1'), covar=tensor([0.6407, 0.4934, 0.0972, 0.1376, 0.1145, 0.1936, 0.2980, 0.1208], device='cuda:1'), in_proj_covar=tensor([0.0308, 0.0306, 0.0406, 0.0411, 0.0349, 0.0403, 0.0317, 0.0369], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-26 21:22:11,564 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=3.95 vs. limit=5.0 2023-04-26 21:22:22,680 INFO [finetune.py:976] (1/7) Epoch 8, batch 3600, loss[loss=0.1894, simple_loss=0.2667, pruned_loss=0.05599, over 4918.00 frames. ], tot_loss[loss=0.1929, simple_loss=0.2573, pruned_loss=0.0643, over 954853.65 frames. ], batch size: 43, lr: 3.82e-03, grad_scale: 32.0 2023-04-26 21:22:42,522 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-04-26 21:22:51,282 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.75 vs. limit=5.0 2023-04-26 21:23:15,350 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.186e+02 1.664e+02 2.003e+02 2.703e+02 5.988e+02, threshold=4.006e+02, percent-clipped=3.0 2023-04-26 21:23:34,071 INFO [finetune.py:976] (1/7) Epoch 8, batch 3650, loss[loss=0.1894, simple_loss=0.2641, pruned_loss=0.05733, over 4919.00 frames. ], tot_loss[loss=0.1953, simple_loss=0.2594, pruned_loss=0.06558, over 954393.03 frames. ], batch size: 36, lr: 3.82e-03, grad_scale: 32.0 2023-04-26 21:24:29,559 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=43787.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 21:24:30,830 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-04-26 21:24:31,981 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.0380, 1.5085, 1.4428, 1.7481, 1.5705, 1.8813, 1.2711, 3.7540], device='cuda:1'), covar=tensor([0.0857, 0.1087, 0.1038, 0.1428, 0.0912, 0.0701, 0.1084, 0.0189], device='cuda:1'), in_proj_covar=tensor([0.0039, 0.0039, 0.0040, 0.0044, 0.0040, 0.0039, 0.0039, 0.0059], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0014, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:1') 2023-04-26 21:24:33,740 INFO [finetune.py:976] (1/7) Epoch 8, batch 3700, loss[loss=0.1718, simple_loss=0.2461, pruned_loss=0.04873, over 4891.00 frames. ], tot_loss[loss=0.1991, simple_loss=0.264, pruned_loss=0.06707, over 955877.75 frames. ], batch size: 35, lr: 3.82e-03, grad_scale: 32.0 2023-04-26 21:24:52,318 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6430, 2.0286, 1.0094, 1.3374, 1.9393, 1.4913, 1.4027, 1.4518], device='cuda:1'), covar=tensor([0.0612, 0.0303, 0.0366, 0.0636, 0.0289, 0.0697, 0.0701, 0.0645], device='cuda:1'), in_proj_covar=tensor([0.0030, 0.0026, 0.0024, 0.0030, 0.0021, 0.0030, 0.0029, 0.0030], device='cuda:1'), out_proj_covar=tensor([0.0049, 0.0044, 0.0038, 0.0050, 0.0038, 0.0048, 0.0048, 0.0050], device='cuda:1') 2023-04-26 21:25:03,743 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.187e+02 1.778e+02 2.044e+02 2.450e+02 4.108e+02, threshold=4.087e+02, percent-clipped=2.0 2023-04-26 21:25:11,087 INFO [finetune.py:976] (1/7) Epoch 8, batch 3750, loss[loss=0.2497, simple_loss=0.3168, pruned_loss=0.0913, over 4889.00 frames. ], tot_loss[loss=0.2001, simple_loss=0.2657, pruned_loss=0.06726, over 956716.40 frames. ], batch size: 43, lr: 3.82e-03, grad_scale: 32.0 2023-04-26 21:25:14,039 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=43848.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 21:25:27,354 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=43870.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 21:25:39,449 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.2421, 2.9388, 1.0686, 1.4047, 1.9733, 1.2856, 3.9799, 1.6033], device='cuda:1'), covar=tensor([0.0849, 0.0904, 0.1131, 0.1670, 0.0705, 0.1364, 0.0390, 0.0990], device='cuda:1'), in_proj_covar=tensor([0.0052, 0.0068, 0.0050, 0.0048, 0.0052, 0.0053, 0.0079, 0.0052], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0011, 0.0007], device='cuda:1') 2023-04-26 21:25:44,749 INFO [finetune.py:976] (1/7) Epoch 8, batch 3800, loss[loss=0.2205, simple_loss=0.2781, pruned_loss=0.08141, over 4896.00 frames. ], tot_loss[loss=0.2007, simple_loss=0.2664, pruned_loss=0.06755, over 953829.89 frames. ], batch size: 36, lr: 3.82e-03, grad_scale: 32.0 2023-04-26 21:25:47,320 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.1836, 1.4462, 1.2796, 1.6819, 1.5247, 1.7970, 1.2552, 3.6457], device='cuda:1'), covar=tensor([0.0804, 0.1081, 0.1091, 0.1444, 0.0871, 0.0728, 0.1061, 0.0174], device='cuda:1'), in_proj_covar=tensor([0.0039, 0.0039, 0.0040, 0.0044, 0.0040, 0.0039, 0.0039, 0.0059], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0014, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:1') 2023-04-26 21:25:54,496 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=43909.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 21:26:09,807 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.001e+02 1.636e+02 2.019e+02 2.329e+02 3.676e+02, threshold=4.038e+02, percent-clipped=0.0 2023-04-26 21:26:18,038 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([3.3105, 3.1996, 2.3777, 3.8402, 3.3767, 3.2997, 1.3977, 3.3030], device='cuda:1'), covar=tensor([0.1922, 0.1392, 0.3242, 0.2326, 0.3061, 0.1932, 0.5731, 0.2365], device='cuda:1'), in_proj_covar=tensor([0.0242, 0.0215, 0.0248, 0.0306, 0.0298, 0.0249, 0.0269, 0.0267], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-26 21:26:18,580 INFO [finetune.py:976] (1/7) Epoch 8, batch 3850, loss[loss=0.1985, simple_loss=0.27, pruned_loss=0.06354, over 4825.00 frames. ], tot_loss[loss=0.1987, simple_loss=0.2648, pruned_loss=0.06629, over 955040.07 frames. ], batch size: 39, lr: 3.82e-03, grad_scale: 32.0 2023-04-26 21:26:25,312 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6836, 1.4098, 1.8355, 2.1262, 1.8378, 1.6108, 1.6684, 1.7389], device='cuda:1'), covar=tensor([0.7502, 0.9770, 1.0130, 0.9948, 0.8528, 1.2111, 1.2308, 1.0877], device='cuda:1'), in_proj_covar=tensor([0.0409, 0.0424, 0.0508, 0.0530, 0.0438, 0.0457, 0.0470, 0.0467], device='cuda:1'), out_proj_covar=tensor([9.9473e-05, 1.0504e-04, 1.1485e-04, 1.2563e-04, 1.0648e-04, 1.1039e-04, 1.1298e-04, 1.1329e-04], device='cuda:1') 2023-04-26 21:26:28,335 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=43959.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 21:26:35,134 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=43970.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 21:27:02,339 INFO [finetune.py:976] (1/7) Epoch 8, batch 3900, loss[loss=0.1821, simple_loss=0.2389, pruned_loss=0.06268, over 4790.00 frames. ], tot_loss[loss=0.1968, simple_loss=0.2622, pruned_loss=0.06571, over 955062.31 frames. ], batch size: 45, lr: 3.82e-03, grad_scale: 64.0 2023-04-26 21:27:14,292 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5785, 1.2058, 4.1345, 3.5180, 3.7063, 3.9250, 3.7965, 3.3890], device='cuda:1'), covar=tensor([0.9393, 0.8742, 0.1621, 0.3093, 0.2231, 0.2670, 0.2642, 0.2897], device='cuda:1'), in_proj_covar=tensor([0.0307, 0.0305, 0.0406, 0.0409, 0.0348, 0.0401, 0.0315, 0.0368], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-26 21:27:35,501 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=44020.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 21:27:48,404 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.103e+02 1.747e+02 1.960e+02 2.337e+02 4.630e+02, threshold=3.919e+02, percent-clipped=3.0 2023-04-26 21:28:08,675 INFO [finetune.py:976] (1/7) Epoch 8, batch 3950, loss[loss=0.1713, simple_loss=0.2396, pruned_loss=0.05153, over 4759.00 frames. ], tot_loss[loss=0.1923, simple_loss=0.2571, pruned_loss=0.06376, over 955463.22 frames. ], batch size: 26, lr: 3.82e-03, grad_scale: 64.0 2023-04-26 21:28:20,338 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([0.7266, 1.8557, 1.6513, 1.5178, 1.9224, 1.6045, 2.5260, 1.5451], device='cuda:1'), covar=tensor([0.4543, 0.1841, 0.5168, 0.3645, 0.1966, 0.2675, 0.1498, 0.4679], device='cuda:1'), in_proj_covar=tensor([0.0346, 0.0349, 0.0429, 0.0363, 0.0385, 0.0381, 0.0381, 0.0418], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 21:28:40,428 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.1215, 1.2699, 4.9865, 4.5802, 4.2601, 4.7203, 4.3812, 4.3648], device='cuda:1'), covar=tensor([0.6978, 0.6668, 0.0968, 0.1860, 0.1247, 0.1582, 0.1891, 0.1564], device='cuda:1'), in_proj_covar=tensor([0.0307, 0.0304, 0.0406, 0.0409, 0.0347, 0.0401, 0.0314, 0.0368], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-26 21:28:48,367 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=6.15 vs. limit=5.0 2023-04-26 21:28:58,844 INFO [finetune.py:976] (1/7) Epoch 8, batch 4000, loss[loss=0.2035, simple_loss=0.2757, pruned_loss=0.06568, over 4934.00 frames. ], tot_loss[loss=0.1938, simple_loss=0.2575, pruned_loss=0.06501, over 954892.91 frames. ], batch size: 38, lr: 3.82e-03, grad_scale: 64.0 2023-04-26 21:29:49,240 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.897e+01 1.638e+02 2.029e+02 2.445e+02 5.584e+02, threshold=4.057e+02, percent-clipped=1.0 2023-04-26 21:30:02,112 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=44143.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 21:30:02,663 INFO [finetune.py:976] (1/7) Epoch 8, batch 4050, loss[loss=0.2847, simple_loss=0.3467, pruned_loss=0.1114, over 4858.00 frames. ], tot_loss[loss=0.1966, simple_loss=0.2609, pruned_loss=0.06614, over 954673.79 frames. ], batch size: 44, lr: 3.82e-03, grad_scale: 64.0 2023-04-26 21:30:35,439 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=44170.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 21:31:07,890 INFO [finetune.py:976] (1/7) Epoch 8, batch 4100, loss[loss=0.2367, simple_loss=0.2956, pruned_loss=0.08887, over 4866.00 frames. ], tot_loss[loss=0.1993, simple_loss=0.2642, pruned_loss=0.06721, over 954049.49 frames. ], batch size: 34, lr: 3.82e-03, grad_scale: 64.0 2023-04-26 21:31:19,881 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.2261, 1.6442, 1.3020, 1.9438, 1.7096, 1.9397, 1.4472, 4.1219], device='cuda:1'), covar=tensor([0.0592, 0.0752, 0.0833, 0.1165, 0.0655, 0.0630, 0.0762, 0.0115], device='cuda:1'), in_proj_covar=tensor([0.0039, 0.0039, 0.0041, 0.0044, 0.0040, 0.0040, 0.0039, 0.0059], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:1') 2023-04-26 21:31:41,622 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=44218.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 21:31:42,258 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.2138, 2.9837, 0.8637, 1.6299, 1.6394, 2.1742, 1.7100, 0.9743], device='cuda:1'), covar=tensor([0.1518, 0.0926, 0.2034, 0.1408, 0.1213, 0.0986, 0.1621, 0.1914], device='cuda:1'), in_proj_covar=tensor([0.0118, 0.0253, 0.0142, 0.0124, 0.0136, 0.0155, 0.0120, 0.0122], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-26 21:32:00,836 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.207e+02 1.648e+02 2.017e+02 2.380e+02 4.436e+02, threshold=4.034e+02, percent-clipped=1.0 2023-04-26 21:32:13,834 INFO [finetune.py:976] (1/7) Epoch 8, batch 4150, loss[loss=0.2316, simple_loss=0.3082, pruned_loss=0.07746, over 4927.00 frames. ], tot_loss[loss=0.2028, simple_loss=0.2679, pruned_loss=0.06883, over 952942.47 frames. ], batch size: 42, lr: 3.82e-03, grad_scale: 64.0 2023-04-26 21:32:23,552 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.2128, 1.1256, 3.7701, 3.4629, 3.3332, 3.6197, 3.5716, 3.3028], device='cuda:1'), covar=tensor([0.7459, 0.6333, 0.1305, 0.2044, 0.1351, 0.1527, 0.1831, 0.1712], device='cuda:1'), in_proj_covar=tensor([0.0312, 0.0308, 0.0413, 0.0415, 0.0352, 0.0408, 0.0319, 0.0373], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-26 21:32:34,619 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=44265.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 21:32:51,525 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9856, 2.3871, 1.2100, 1.4252, 1.8623, 1.3076, 2.8248, 1.5477], device='cuda:1'), covar=tensor([0.0645, 0.0732, 0.0774, 0.1094, 0.0459, 0.0912, 0.0277, 0.0645], device='cuda:1'), in_proj_covar=tensor([0.0052, 0.0068, 0.0050, 0.0048, 0.0052, 0.0053, 0.0079, 0.0052], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0011, 0.0008], device='cuda:1') 2023-04-26 21:32:52,669 INFO [finetune.py:976] (1/7) Epoch 8, batch 4200, loss[loss=0.1962, simple_loss=0.2523, pruned_loss=0.07005, over 4696.00 frames. ], tot_loss[loss=0.2021, simple_loss=0.2675, pruned_loss=0.06836, over 952299.45 frames. ], batch size: 59, lr: 3.82e-03, grad_scale: 64.0 2023-04-26 21:33:08,262 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=44315.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 21:33:18,929 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.191e+02 1.745e+02 1.959e+02 2.417e+02 4.699e+02, threshold=3.917e+02, percent-clipped=3.0 2023-04-26 21:33:26,340 INFO [finetune.py:976] (1/7) Epoch 8, batch 4250, loss[loss=0.1793, simple_loss=0.2435, pruned_loss=0.05757, over 4771.00 frames. ], tot_loss[loss=0.199, simple_loss=0.2638, pruned_loss=0.06706, over 952031.17 frames. ], batch size: 26, lr: 3.82e-03, grad_scale: 64.0 2023-04-26 21:33:48,048 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([0.7336, 1.8993, 1.7573, 1.4429, 2.0369, 1.5247, 2.5979, 1.5551], device='cuda:1'), covar=tensor([0.4080, 0.1731, 0.5125, 0.3288, 0.1735, 0.2602, 0.1569, 0.4848], device='cuda:1'), in_proj_covar=tensor([0.0347, 0.0351, 0.0432, 0.0367, 0.0388, 0.0383, 0.0384, 0.0421], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 21:33:54,172 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0716, 1.9378, 2.4791, 2.6735, 1.9481, 1.7048, 2.0783, 1.1760], device='cuda:1'), covar=tensor([0.0632, 0.0957, 0.0458, 0.0681, 0.0954, 0.1364, 0.0923, 0.1082], device='cuda:1'), in_proj_covar=tensor([0.0065, 0.0073, 0.0072, 0.0067, 0.0076, 0.0096, 0.0079, 0.0074], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-26 21:34:00,167 INFO [finetune.py:976] (1/7) Epoch 8, batch 4300, loss[loss=0.1537, simple_loss=0.2187, pruned_loss=0.04438, over 4814.00 frames. ], tot_loss[loss=0.1969, simple_loss=0.2612, pruned_loss=0.06628, over 953736.12 frames. ], batch size: 25, lr: 3.82e-03, grad_scale: 64.0 2023-04-26 21:34:21,537 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=3.75 vs. limit=5.0 2023-04-26 21:34:26,193 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.110e+02 1.667e+02 1.928e+02 2.256e+02 4.744e+02, threshold=3.855e+02, percent-clipped=2.0 2023-04-26 21:34:37,499 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4597, 1.7131, 1.3237, 0.9704, 1.1633, 1.1735, 1.3320, 1.0941], device='cuda:1'), covar=tensor([0.1742, 0.1385, 0.1678, 0.1942, 0.2555, 0.1973, 0.1189, 0.2111], device='cuda:1'), in_proj_covar=tensor([0.0195, 0.0216, 0.0171, 0.0204, 0.0206, 0.0185, 0.0163, 0.0189], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-26 21:34:44,423 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=44443.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 21:34:44,910 INFO [finetune.py:976] (1/7) Epoch 8, batch 4350, loss[loss=0.189, simple_loss=0.2512, pruned_loss=0.06339, over 4765.00 frames. ], tot_loss[loss=0.1942, simple_loss=0.2578, pruned_loss=0.06532, over 955064.80 frames. ], batch size: 27, lr: 3.82e-03, grad_scale: 64.0 2023-04-26 21:35:46,505 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=44491.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 21:35:48,244 INFO [finetune.py:976] (1/7) Epoch 8, batch 4400, loss[loss=0.2342, simple_loss=0.298, pruned_loss=0.08521, over 4814.00 frames. ], tot_loss[loss=0.1944, simple_loss=0.2581, pruned_loss=0.06537, over 954717.22 frames. ], batch size: 40, lr: 3.82e-03, grad_scale: 64.0 2023-04-26 21:36:11,291 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=44527.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 21:36:14,647 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.189e+02 1.807e+02 2.142e+02 2.505e+02 5.650e+02, threshold=4.284e+02, percent-clipped=2.0 2023-04-26 21:36:33,284 INFO [finetune.py:976] (1/7) Epoch 8, batch 4450, loss[loss=0.1857, simple_loss=0.2462, pruned_loss=0.06256, over 4825.00 frames. ], tot_loss[loss=0.1978, simple_loss=0.2623, pruned_loss=0.0666, over 955765.08 frames. ], batch size: 30, lr: 3.82e-03, grad_scale: 64.0 2023-04-26 21:36:42,760 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=44550.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 21:36:58,010 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=44565.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 21:37:30,487 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=44588.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 21:37:39,769 INFO [finetune.py:976] (1/7) Epoch 8, batch 4500, loss[loss=0.2185, simple_loss=0.2833, pruned_loss=0.07687, over 4793.00 frames. ], tot_loss[loss=0.1989, simple_loss=0.2638, pruned_loss=0.06704, over 955714.93 frames. ], batch size: 51, lr: 3.82e-03, grad_scale: 64.0 2023-04-26 21:37:49,426 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4207, 1.1661, 0.3825, 1.1688, 1.2305, 1.3482, 1.2456, 1.2258], device='cuda:1'), covar=tensor([0.0515, 0.0409, 0.0460, 0.0580, 0.0304, 0.0524, 0.0503, 0.0579], device='cuda:1'), in_proj_covar=tensor([0.0030, 0.0026, 0.0024, 0.0030, 0.0021, 0.0030, 0.0029, 0.0030], device='cuda:1'), out_proj_covar=tensor([0.0049, 0.0044, 0.0038, 0.0049, 0.0037, 0.0048, 0.0048, 0.0049], device='cuda:1') 2023-04-26 21:37:57,808 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=44611.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 21:37:58,975 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=44613.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 21:38:00,233 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=44615.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 21:38:23,239 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.4775, 2.8187, 1.2391, 1.6539, 2.2980, 1.4565, 3.6640, 1.9829], device='cuda:1'), covar=tensor([0.0610, 0.0586, 0.0786, 0.1306, 0.0462, 0.1019, 0.0229, 0.0630], device='cuda:1'), in_proj_covar=tensor([0.0052, 0.0068, 0.0050, 0.0048, 0.0052, 0.0053, 0.0079, 0.0051], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0007], device='cuda:1') 2023-04-26 21:38:30,727 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.141e+02 1.764e+02 2.099e+02 2.413e+02 4.489e+02, threshold=4.197e+02, percent-clipped=1.0 2023-04-26 21:38:43,666 INFO [finetune.py:976] (1/7) Epoch 8, batch 4550, loss[loss=0.195, simple_loss=0.2723, pruned_loss=0.05883, over 4919.00 frames. ], tot_loss[loss=0.1985, simple_loss=0.264, pruned_loss=0.0665, over 956195.92 frames. ], batch size: 41, lr: 3.82e-03, grad_scale: 32.0 2023-04-26 21:38:55,365 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=44663.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 21:38:56,086 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-04-26 21:39:07,221 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-04-26 21:39:17,128 INFO [finetune.py:976] (1/7) Epoch 8, batch 4600, loss[loss=0.1744, simple_loss=0.2377, pruned_loss=0.05556, over 4659.00 frames. ], tot_loss[loss=0.1972, simple_loss=0.2629, pruned_loss=0.06573, over 955671.68 frames. ], batch size: 23, lr: 3.82e-03, grad_scale: 32.0 2023-04-26 21:39:38,851 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.91 vs. limit=5.0 2023-04-26 21:39:43,235 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.165e+02 1.725e+02 1.987e+02 2.350e+02 3.510e+02, threshold=3.973e+02, percent-clipped=0.0 2023-04-26 21:39:55,901 INFO [finetune.py:976] (1/7) Epoch 8, batch 4650, loss[loss=0.1948, simple_loss=0.2564, pruned_loss=0.06659, over 4889.00 frames. ], tot_loss[loss=0.1956, simple_loss=0.2605, pruned_loss=0.06537, over 956593.15 frames. ], batch size: 35, lr: 3.82e-03, grad_scale: 32.0 2023-04-26 21:41:01,771 INFO [finetune.py:976] (1/7) Epoch 8, batch 4700, loss[loss=0.1976, simple_loss=0.257, pruned_loss=0.0691, over 4819.00 frames. ], tot_loss[loss=0.1917, simple_loss=0.2564, pruned_loss=0.06348, over 956408.19 frames. ], batch size: 38, lr: 3.82e-03, grad_scale: 32.0 2023-04-26 21:41:25,425 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0837, 2.4541, 0.9906, 1.3595, 1.8074, 1.3588, 3.2553, 1.6752], device='cuda:1'), covar=tensor([0.0667, 0.0704, 0.0768, 0.1274, 0.0532, 0.0939, 0.0218, 0.0651], device='cuda:1'), in_proj_covar=tensor([0.0052, 0.0068, 0.0050, 0.0048, 0.0052, 0.0053, 0.0079, 0.0052], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0011, 0.0007], device='cuda:1') 2023-04-26 21:41:26,395 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.155e+02 1.673e+02 1.965e+02 2.397e+02 5.509e+02, threshold=3.929e+02, percent-clipped=2.0 2023-04-26 21:41:35,060 INFO [finetune.py:976] (1/7) Epoch 8, batch 4750, loss[loss=0.1853, simple_loss=0.2463, pruned_loss=0.06213, over 4809.00 frames. ], tot_loss[loss=0.1904, simple_loss=0.2548, pruned_loss=0.06304, over 956525.08 frames. ], batch size: 25, lr: 3.82e-03, grad_scale: 32.0 2023-04-26 21:41:39,333 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-04-26 21:41:59,218 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=44883.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 21:42:08,192 INFO [finetune.py:976] (1/7) Epoch 8, batch 4800, loss[loss=0.1281, simple_loss=0.1895, pruned_loss=0.03334, over 4257.00 frames. ], tot_loss[loss=0.1942, simple_loss=0.2581, pruned_loss=0.06517, over 954395.99 frames. ], batch size: 18, lr: 3.82e-03, grad_scale: 32.0 2023-04-26 21:42:13,701 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.2798, 3.0349, 0.8953, 1.5376, 1.7394, 1.4954, 3.8981, 1.7622], device='cuda:1'), covar=tensor([0.0687, 0.1072, 0.1045, 0.1341, 0.0710, 0.1017, 0.0209, 0.0738], device='cuda:1'), in_proj_covar=tensor([0.0052, 0.0068, 0.0050, 0.0048, 0.0052, 0.0053, 0.0079, 0.0052], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0011, 0.0007], device='cuda:1') 2023-04-26 21:42:16,075 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=44906.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 21:42:32,274 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.212e+02 1.751e+02 2.130e+02 2.730e+02 4.658e+02, threshold=4.261e+02, percent-clipped=2.0 2023-04-26 21:42:40,903 INFO [finetune.py:976] (1/7) Epoch 8, batch 4850, loss[loss=0.2311, simple_loss=0.2884, pruned_loss=0.08689, over 4773.00 frames. ], tot_loss[loss=0.1969, simple_loss=0.2616, pruned_loss=0.06608, over 952301.00 frames. ], batch size: 26, lr: 3.82e-03, grad_scale: 32.0 2023-04-26 21:43:27,975 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.2247, 2.9321, 0.8408, 1.5908, 1.6729, 2.0860, 1.6799, 0.9185], device='cuda:1'), covar=tensor([0.1450, 0.0898, 0.1971, 0.1331, 0.1121, 0.0958, 0.1507, 0.2022], device='cuda:1'), in_proj_covar=tensor([0.0116, 0.0251, 0.0141, 0.0123, 0.0135, 0.0154, 0.0119, 0.0121], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-26 21:43:40,824 INFO [finetune.py:976] (1/7) Epoch 8, batch 4900, loss[loss=0.2072, simple_loss=0.2794, pruned_loss=0.06754, over 4895.00 frames. ], tot_loss[loss=0.199, simple_loss=0.2644, pruned_loss=0.06681, over 950580.07 frames. ], batch size: 43, lr: 3.82e-03, grad_scale: 32.0 2023-04-26 21:44:04,798 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.0520, 1.3214, 1.1335, 1.6706, 1.3679, 1.4105, 1.1785, 2.4976], device='cuda:1'), covar=tensor([0.0728, 0.0994, 0.1012, 0.1332, 0.0828, 0.0607, 0.0957, 0.0306], device='cuda:1'), in_proj_covar=tensor([0.0038, 0.0039, 0.0040, 0.0044, 0.0040, 0.0039, 0.0039, 0.0059], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0014, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:1') 2023-04-26 21:44:16,758 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.0639, 2.1716, 1.8470, 1.8162, 2.2289, 1.7291, 2.8314, 1.7191], device='cuda:1'), covar=tensor([0.4226, 0.1857, 0.4579, 0.3709, 0.2093, 0.2829, 0.1501, 0.4555], device='cuda:1'), in_proj_covar=tensor([0.0348, 0.0352, 0.0433, 0.0366, 0.0389, 0.0385, 0.0384, 0.0421], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 21:44:34,742 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.241e+02 1.669e+02 1.942e+02 2.230e+02 4.872e+02, threshold=3.883e+02, percent-clipped=3.0 2023-04-26 21:44:46,773 INFO [finetune.py:976] (1/7) Epoch 8, batch 4950, loss[loss=0.1889, simple_loss=0.261, pruned_loss=0.05841, over 4748.00 frames. ], tot_loss[loss=0.1997, simple_loss=0.2656, pruned_loss=0.06692, over 953133.28 frames. ], batch size: 54, lr: 3.81e-03, grad_scale: 32.0 2023-04-26 21:45:08,563 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=45056.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 21:45:35,345 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=45081.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 21:45:47,620 INFO [finetune.py:976] (1/7) Epoch 8, batch 5000, loss[loss=0.2098, simple_loss=0.2631, pruned_loss=0.07826, over 4883.00 frames. ], tot_loss[loss=0.1989, simple_loss=0.2645, pruned_loss=0.06663, over 953854.43 frames. ], batch size: 32, lr: 3.81e-03, grad_scale: 32.0 2023-04-26 21:46:01,733 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8048, 2.4944, 1.8242, 1.6797, 1.3229, 1.4050, 1.9284, 1.3504], device='cuda:1'), covar=tensor([0.1801, 0.1436, 0.1596, 0.2070, 0.2515, 0.2111, 0.1074, 0.2056], device='cuda:1'), in_proj_covar=tensor([0.0196, 0.0218, 0.0173, 0.0206, 0.0206, 0.0186, 0.0163, 0.0189], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-26 21:46:04,180 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=45117.0, num_to_drop=1, layers_to_drop={3} 2023-04-26 21:46:13,679 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.059e+02 1.645e+02 2.005e+02 2.467e+02 4.350e+02, threshold=4.011e+02, percent-clipped=3.0 2023-04-26 21:46:19,260 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=45142.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 21:46:20,393 INFO [finetune.py:976] (1/7) Epoch 8, batch 5050, loss[loss=0.1833, simple_loss=0.2489, pruned_loss=0.05886, over 4911.00 frames. ], tot_loss[loss=0.1961, simple_loss=0.2611, pruned_loss=0.06559, over 956671.91 frames. ], batch size: 37, lr: 3.81e-03, grad_scale: 32.0 2023-04-26 21:46:47,023 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=45183.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 21:46:53,634 INFO [finetune.py:976] (1/7) Epoch 8, batch 5100, loss[loss=0.1675, simple_loss=0.2337, pruned_loss=0.05071, over 4753.00 frames. ], tot_loss[loss=0.1939, simple_loss=0.258, pruned_loss=0.06486, over 957247.80 frames. ], batch size: 27, lr: 3.81e-03, grad_scale: 32.0 2023-04-26 21:47:02,087 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=45206.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 21:47:13,912 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9473, 1.3523, 1.7901, 2.3514, 1.8849, 1.4050, 1.1036, 1.6908], device='cuda:1'), covar=tensor([0.4155, 0.4443, 0.2281, 0.3035, 0.3539, 0.3321, 0.5479, 0.2605], device='cuda:1'), in_proj_covar=tensor([0.0280, 0.0251, 0.0218, 0.0319, 0.0213, 0.0227, 0.0235, 0.0187], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-26 21:47:17,600 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6040, 1.3093, 1.7586, 2.1283, 1.7939, 1.6382, 1.7142, 1.7182], device='cuda:1'), covar=tensor([0.6624, 0.8908, 0.8456, 0.8516, 0.7836, 1.0172, 1.0984, 0.9588], device='cuda:1'), in_proj_covar=tensor([0.0409, 0.0424, 0.0507, 0.0528, 0.0439, 0.0457, 0.0469, 0.0466], device='cuda:1'), out_proj_covar=tensor([9.9634e-05, 1.0508e-04, 1.1449e-04, 1.2526e-04, 1.0672e-04, 1.1055e-04, 1.1289e-04, 1.1296e-04], device='cuda:1') 2023-04-26 21:47:18,715 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=45231.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 21:47:19,883 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.132e+02 1.617e+02 1.880e+02 2.299e+02 5.532e+02, threshold=3.760e+02, percent-clipped=2.0 2023-04-26 21:47:26,586 INFO [finetune.py:976] (1/7) Epoch 8, batch 5150, loss[loss=0.2621, simple_loss=0.318, pruned_loss=0.1031, over 4810.00 frames. ], tot_loss[loss=0.1967, simple_loss=0.2601, pruned_loss=0.06662, over 956715.29 frames. ], batch size: 41, lr: 3.81e-03, grad_scale: 32.0 2023-04-26 21:47:32,144 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7538, 4.0799, 0.7145, 2.0623, 2.2806, 2.7031, 2.4545, 0.9806], device='cuda:1'), covar=tensor([0.1366, 0.0988, 0.2383, 0.1368, 0.1045, 0.1071, 0.1345, 0.2168], device='cuda:1'), in_proj_covar=tensor([0.0117, 0.0251, 0.0141, 0.0123, 0.0135, 0.0155, 0.0119, 0.0122], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-26 21:47:32,723 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=45254.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 21:48:00,050 INFO [finetune.py:976] (1/7) Epoch 8, batch 5200, loss[loss=0.1794, simple_loss=0.2542, pruned_loss=0.05227, over 4831.00 frames. ], tot_loss[loss=0.1994, simple_loss=0.2635, pruned_loss=0.06768, over 956983.35 frames. ], batch size: 30, lr: 3.81e-03, grad_scale: 32.0 2023-04-26 21:48:35,329 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.3840, 1.6902, 1.7511, 1.8981, 1.7371, 1.8529, 1.8903, 1.7897], device='cuda:1'), covar=tensor([0.4939, 0.7581, 0.6097, 0.5792, 0.7107, 0.9858, 0.7058, 0.6992], device='cuda:1'), in_proj_covar=tensor([0.0319, 0.0381, 0.0314, 0.0323, 0.0339, 0.0400, 0.0361, 0.0324], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-26 21:48:44,628 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.118e+02 1.724e+02 2.073e+02 2.442e+02 4.626e+02, threshold=4.147e+02, percent-clipped=2.0 2023-04-26 21:48:56,369 INFO [finetune.py:976] (1/7) Epoch 8, batch 5250, loss[loss=0.2007, simple_loss=0.2432, pruned_loss=0.07906, over 4239.00 frames. ], tot_loss[loss=0.1998, simple_loss=0.2648, pruned_loss=0.06742, over 956973.47 frames. ], batch size: 18, lr: 3.81e-03, grad_scale: 32.0 2023-04-26 21:48:58,921 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=45348.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 21:49:00,219 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-04-26 21:49:19,336 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9801, 2.3678, 2.0373, 2.2152, 1.8816, 1.9412, 2.0918, 1.6955], device='cuda:1'), covar=tensor([0.1936, 0.1307, 0.0932, 0.1244, 0.3429, 0.1309, 0.1770, 0.2303], device='cuda:1'), in_proj_covar=tensor([0.0302, 0.0323, 0.0232, 0.0295, 0.0320, 0.0277, 0.0261, 0.0284], device='cuda:1'), out_proj_covar=tensor([1.2220e-04, 1.3019e-04, 9.3652e-05, 1.1817e-04, 1.3161e-04, 1.1178e-04, 1.0701e-04, 1.1393e-04], device='cuda:1') 2023-04-26 21:50:01,146 INFO [finetune.py:976] (1/7) Epoch 8, batch 5300, loss[loss=0.1987, simple_loss=0.2676, pruned_loss=0.06495, over 4816.00 frames. ], tot_loss[loss=0.1996, simple_loss=0.2646, pruned_loss=0.06724, over 952066.28 frames. ], batch size: 25, lr: 3.81e-03, grad_scale: 32.0 2023-04-26 21:50:21,699 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=45409.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 21:50:23,488 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=45412.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 21:50:55,375 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.314e+02 1.711e+02 1.928e+02 2.396e+02 5.196e+02, threshold=3.857e+02, percent-clipped=2.0 2023-04-26 21:50:57,928 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=45437.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 21:51:07,771 INFO [finetune.py:976] (1/7) Epoch 8, batch 5350, loss[loss=0.1849, simple_loss=0.2463, pruned_loss=0.06176, over 4714.00 frames. ], tot_loss[loss=0.1987, simple_loss=0.2643, pruned_loss=0.06656, over 953289.79 frames. ], batch size: 23, lr: 3.81e-03, grad_scale: 32.0 2023-04-26 21:51:07,903 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5601, 1.4418, 1.8918, 1.8793, 1.4221, 1.2039, 1.5252, 0.9955], device='cuda:1'), covar=tensor([0.0652, 0.0983, 0.0565, 0.0929, 0.0944, 0.1248, 0.0881, 0.0995], device='cuda:1'), in_proj_covar=tensor([0.0066, 0.0073, 0.0072, 0.0067, 0.0076, 0.0096, 0.0078, 0.0074], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-26 21:51:08,599 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.13 vs. limit=5.0 2023-04-26 21:51:43,694 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.55 vs. limit=5.0 2023-04-26 21:51:44,903 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=3.76 vs. limit=5.0 2023-04-26 21:51:45,336 INFO [finetune.py:976] (1/7) Epoch 8, batch 5400, loss[loss=0.1433, simple_loss=0.2176, pruned_loss=0.03453, over 4767.00 frames. ], tot_loss[loss=0.1973, simple_loss=0.2625, pruned_loss=0.06608, over 954319.36 frames. ], batch size: 28, lr: 3.81e-03, grad_scale: 32.0 2023-04-26 21:51:58,131 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.53 vs. limit=5.0 2023-04-26 21:52:11,223 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.182e+02 1.671e+02 1.868e+02 2.359e+02 4.891e+02, threshold=3.736e+02, percent-clipped=4.0 2023-04-26 21:52:18,380 INFO [finetune.py:976] (1/7) Epoch 8, batch 5450, loss[loss=0.1872, simple_loss=0.2487, pruned_loss=0.06287, over 4816.00 frames. ], tot_loss[loss=0.1945, simple_loss=0.2592, pruned_loss=0.0649, over 953677.07 frames. ], batch size: 41, lr: 3.81e-03, grad_scale: 32.0 2023-04-26 21:52:42,435 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-04-26 21:52:51,571 INFO [finetune.py:976] (1/7) Epoch 8, batch 5500, loss[loss=0.1789, simple_loss=0.2446, pruned_loss=0.05665, over 4764.00 frames. ], tot_loss[loss=0.1912, simple_loss=0.2556, pruned_loss=0.06343, over 952578.07 frames. ], batch size: 26, lr: 3.81e-03, grad_scale: 32.0 2023-04-26 21:53:11,180 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9868, 2.6881, 1.9976, 2.0805, 1.4071, 1.4616, 2.2405, 1.4089], device='cuda:1'), covar=tensor([0.1782, 0.1661, 0.1488, 0.1947, 0.2534, 0.2147, 0.1056, 0.2144], device='cuda:1'), in_proj_covar=tensor([0.0194, 0.0216, 0.0171, 0.0204, 0.0205, 0.0184, 0.0161, 0.0188], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-26 21:53:16,426 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.060e+02 1.725e+02 1.934e+02 2.387e+02 4.697e+02, threshold=3.868e+02, percent-clipped=2.0 2023-04-26 21:53:24,570 INFO [finetune.py:976] (1/7) Epoch 8, batch 5550, loss[loss=0.1708, simple_loss=0.2394, pruned_loss=0.05107, over 4716.00 frames. ], tot_loss[loss=0.1927, simple_loss=0.2572, pruned_loss=0.06404, over 954948.81 frames. ], batch size: 23, lr: 3.81e-03, grad_scale: 32.0 2023-04-26 21:53:35,097 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.63 vs. limit=2.0 2023-04-26 21:53:42,457 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5829, 1.0005, 1.3451, 1.2036, 1.7091, 1.4320, 1.1827, 1.3090], device='cuda:1'), covar=tensor([0.1730, 0.1552, 0.2233, 0.1524, 0.0930, 0.1294, 0.1875, 0.2078], device='cuda:1'), in_proj_covar=tensor([0.0313, 0.0326, 0.0358, 0.0303, 0.0342, 0.0328, 0.0313, 0.0359], device='cuda:1'), out_proj_covar=tensor([6.6278e-05, 6.9228e-05, 7.7363e-05, 6.2824e-05, 7.1917e-05, 7.0542e-05, 6.7392e-05, 7.7128e-05], device='cuda:1') 2023-04-26 21:54:06,991 INFO [finetune.py:976] (1/7) Epoch 8, batch 5600, loss[loss=0.1726, simple_loss=0.2451, pruned_loss=0.05008, over 4765.00 frames. ], tot_loss[loss=0.1969, simple_loss=0.2619, pruned_loss=0.0659, over 956348.09 frames. ], batch size: 26, lr: 3.81e-03, grad_scale: 32.0 2023-04-26 21:54:12,171 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=45702.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 21:54:13,282 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=45704.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 21:54:17,969 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=45712.0, num_to_drop=1, layers_to_drop={2} 2023-04-26 21:54:29,792 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-04-26 21:54:30,171 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.094e+02 1.723e+02 2.023e+02 2.514e+02 5.159e+02, threshold=4.046e+02, percent-clipped=6.0 2023-04-26 21:54:32,580 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=45737.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 21:54:36,957 INFO [finetune.py:976] (1/7) Epoch 8, batch 5650, loss[loss=0.2499, simple_loss=0.325, pruned_loss=0.08742, over 4845.00 frames. ], tot_loss[loss=0.1979, simple_loss=0.2644, pruned_loss=0.06574, over 956335.55 frames. ], batch size: 47, lr: 3.81e-03, grad_scale: 32.0 2023-04-26 21:54:46,839 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=45760.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 21:54:47,494 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.2430, 1.7827, 2.1860, 2.7282, 2.2477, 1.8280, 1.7971, 2.0017], device='cuda:1'), covar=tensor([0.3040, 0.3439, 0.1701, 0.2621, 0.2868, 0.2867, 0.4156, 0.2476], device='cuda:1'), in_proj_covar=tensor([0.0282, 0.0253, 0.0220, 0.0320, 0.0214, 0.0229, 0.0236, 0.0189], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-26 21:54:48,641 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=45763.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 21:54:54,999 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=45774.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 21:55:07,628 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=45785.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 21:55:18,509 INFO [finetune.py:976] (1/7) Epoch 8, batch 5700, loss[loss=0.1412, simple_loss=0.1965, pruned_loss=0.04298, over 4409.00 frames. ], tot_loss[loss=0.1969, simple_loss=0.2621, pruned_loss=0.06589, over 939619.64 frames. ], batch size: 19, lr: 3.81e-03, grad_scale: 32.0 2023-04-26 21:55:59,024 INFO [finetune.py:976] (1/7) Epoch 9, batch 0, loss[loss=0.2132, simple_loss=0.2729, pruned_loss=0.0768, over 4820.00 frames. ], tot_loss[loss=0.2132, simple_loss=0.2729, pruned_loss=0.0768, over 4820.00 frames. ], batch size: 39, lr: 3.81e-03, grad_scale: 32.0 2023-04-26 21:55:59,024 INFO [finetune.py:1001] (1/7) Computing validation loss 2023-04-26 21:56:04,722 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5221, 2.8881, 1.0837, 1.8388, 1.9730, 2.2959, 1.9007, 1.2014], device='cuda:1'), covar=tensor([0.1084, 0.0889, 0.1675, 0.1114, 0.0798, 0.0788, 0.1379, 0.1592], device='cuda:1'), in_proj_covar=tensor([0.0118, 0.0253, 0.0142, 0.0124, 0.0135, 0.0156, 0.0120, 0.0122], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-26 21:56:14,947 INFO [finetune.py:1010] (1/7) Epoch 9, validation: loss=0.1554, simple_loss=0.2289, pruned_loss=0.04093, over 2265189.00 frames. 2023-04-26 21:56:14,948 INFO [finetune.py:1011] (1/7) Maximum memory allocated so far is 6391MB 2023-04-26 21:56:33,614 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.064e+02 1.748e+02 2.064e+02 2.356e+02 2.984e+02, threshold=4.129e+02, percent-clipped=0.0 2023-04-26 21:56:34,959 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=45835.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 21:56:46,615 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.2291, 2.8991, 0.9941, 1.4425, 1.9905, 1.3429, 3.7514, 1.6620], device='cuda:1'), covar=tensor([0.0637, 0.0941, 0.0874, 0.1216, 0.0534, 0.0908, 0.0176, 0.0639], device='cuda:1'), in_proj_covar=tensor([0.0052, 0.0068, 0.0050, 0.0048, 0.0052, 0.0053, 0.0079, 0.0052], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0011, 0.0008], device='cuda:1') 2023-04-26 21:57:05,568 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=45859.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 21:57:19,424 INFO [finetune.py:976] (1/7) Epoch 9, batch 50, loss[loss=0.1868, simple_loss=0.2568, pruned_loss=0.05841, over 4865.00 frames. ], tot_loss[loss=0.1963, simple_loss=0.2624, pruned_loss=0.06513, over 215657.34 frames. ], batch size: 31, lr: 3.81e-03, grad_scale: 32.0 2023-04-26 21:57:38,672 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0069, 1.6012, 1.5270, 1.7520, 2.2449, 1.8263, 1.6187, 1.4512], device='cuda:1'), covar=tensor([0.1886, 0.1560, 0.1920, 0.1410, 0.0770, 0.1506, 0.2041, 0.2221], device='cuda:1'), in_proj_covar=tensor([0.0312, 0.0326, 0.0354, 0.0303, 0.0341, 0.0327, 0.0311, 0.0357], device='cuda:1'), out_proj_covar=tensor([6.6122e-05, 6.9124e-05, 7.6568e-05, 6.2735e-05, 7.1757e-05, 7.0199e-05, 6.6910e-05, 7.6727e-05], device='cuda:1') 2023-04-26 21:57:51,854 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=45920.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 21:57:52,935 INFO [finetune.py:976] (1/7) Epoch 9, batch 100, loss[loss=0.1773, simple_loss=0.2397, pruned_loss=0.05743, over 4860.00 frames. ], tot_loss[loss=0.1914, simple_loss=0.2561, pruned_loss=0.06332, over 380536.87 frames. ], batch size: 34, lr: 3.81e-03, grad_scale: 32.0 2023-04-26 21:58:01,519 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.187e+02 1.655e+02 1.945e+02 2.383e+02 5.251e+02, threshold=3.889e+02, percent-clipped=1.0 2023-04-26 21:58:08,981 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=3.93 vs. limit=5.0 2023-04-26 21:58:26,128 INFO [finetune.py:976] (1/7) Epoch 9, batch 150, loss[loss=0.1668, simple_loss=0.2412, pruned_loss=0.04615, over 4781.00 frames. ], tot_loss[loss=0.1874, simple_loss=0.2519, pruned_loss=0.06147, over 508804.06 frames. ], batch size: 29, lr: 3.81e-03, grad_scale: 32.0 2023-04-26 21:58:34,006 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=3.66 vs. limit=5.0 2023-04-26 21:58:47,788 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=46004.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 21:58:47,833 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7495, 1.5448, 1.8092, 2.1294, 1.9406, 1.6641, 1.7351, 1.7194], device='cuda:1'), covar=tensor([0.5140, 0.6837, 0.6393, 0.6968, 0.5675, 0.7854, 0.7977, 0.8112], device='cuda:1'), in_proj_covar=tensor([0.0409, 0.0422, 0.0505, 0.0524, 0.0439, 0.0456, 0.0467, 0.0467], device='cuda:1'), out_proj_covar=tensor([9.9385e-05, 1.0465e-04, 1.1398e-04, 1.2450e-04, 1.0671e-04, 1.1029e-04, 1.1252e-04, 1.1287e-04], device='cuda:1') 2023-04-26 21:58:59,027 INFO [finetune.py:976] (1/7) Epoch 9, batch 200, loss[loss=0.1866, simple_loss=0.2515, pruned_loss=0.0609, over 4814.00 frames. ], tot_loss[loss=0.1874, simple_loss=0.2512, pruned_loss=0.06184, over 607426.81 frames. ], batch size: 54, lr: 3.81e-03, grad_scale: 32.0 2023-04-26 21:59:07,994 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.665e+01 1.625e+02 2.034e+02 2.353e+02 6.037e+02, threshold=4.068e+02, percent-clipped=2.0 2023-04-26 21:59:19,776 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=46052.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 21:59:19,837 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5743, 1.7396, 1.4211, 1.0716, 1.1808, 1.1797, 1.4269, 1.1199], device='cuda:1'), covar=tensor([0.1956, 0.1475, 0.1839, 0.2076, 0.2735, 0.2250, 0.1265, 0.2278], device='cuda:1'), in_proj_covar=tensor([0.0195, 0.0217, 0.0172, 0.0205, 0.0206, 0.0185, 0.0163, 0.0189], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-26 21:59:23,446 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=46058.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 21:59:32,393 INFO [finetune.py:976] (1/7) Epoch 9, batch 250, loss[loss=0.2181, simple_loss=0.2857, pruned_loss=0.0753, over 4902.00 frames. ], tot_loss[loss=0.1902, simple_loss=0.2542, pruned_loss=0.06314, over 683699.03 frames. ], batch size: 36, lr: 3.81e-03, grad_scale: 32.0 2023-04-26 22:00:05,474 INFO [finetune.py:976] (1/7) Epoch 9, batch 300, loss[loss=0.2092, simple_loss=0.2751, pruned_loss=0.07168, over 4933.00 frames. ], tot_loss[loss=0.1931, simple_loss=0.258, pruned_loss=0.06415, over 743514.32 frames. ], batch size: 33, lr: 3.81e-03, grad_scale: 32.0 2023-04-26 22:00:08,186 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2023-04-26 22:00:10,893 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=46130.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:00:12,630 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.245e+02 1.689e+02 1.934e+02 2.285e+02 6.162e+02, threshold=3.867e+02, percent-clipped=1.0 2023-04-26 22:00:19,441 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9615, 1.6496, 1.9607, 2.2830, 2.3692, 1.8075, 1.4636, 1.9259], device='cuda:1'), covar=tensor([0.0861, 0.1202, 0.0619, 0.0604, 0.0564, 0.0844, 0.0969, 0.0658], device='cuda:1'), in_proj_covar=tensor([0.0197, 0.0204, 0.0180, 0.0177, 0.0178, 0.0190, 0.0160, 0.0185], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 22:00:19,504 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-04-26 22:00:42,792 INFO [finetune.py:976] (1/7) Epoch 9, batch 350, loss[loss=0.26, simple_loss=0.3104, pruned_loss=0.1048, over 4736.00 frames. ], tot_loss[loss=0.1989, simple_loss=0.2636, pruned_loss=0.06715, over 788367.82 frames. ], batch size: 54, lr: 3.81e-03, grad_scale: 32.0 2023-04-26 22:01:38,755 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=46215.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:01:38,902 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-04-26 22:01:48,010 INFO [finetune.py:976] (1/7) Epoch 9, batch 400, loss[loss=0.2356, simple_loss=0.3057, pruned_loss=0.08278, over 4906.00 frames. ], tot_loss[loss=0.2001, simple_loss=0.2655, pruned_loss=0.06735, over 827529.87 frames. ], batch size: 43, lr: 3.80e-03, grad_scale: 32.0 2023-04-26 22:02:01,095 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.329e+02 1.830e+02 2.182e+02 2.472e+02 8.213e+02, threshold=4.364e+02, percent-clipped=3.0 2023-04-26 22:02:14,220 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4535, 0.6610, 1.3522, 1.8554, 1.6164, 1.4139, 1.4112, 1.4618], device='cuda:1'), covar=tensor([0.5087, 0.7134, 0.6395, 0.7688, 0.6401, 0.8291, 0.8191, 0.7479], device='cuda:1'), in_proj_covar=tensor([0.0408, 0.0421, 0.0504, 0.0523, 0.0437, 0.0455, 0.0467, 0.0465], device='cuda:1'), out_proj_covar=tensor([9.9194e-05, 1.0446e-04, 1.1379e-04, 1.2434e-04, 1.0621e-04, 1.1007e-04, 1.1240e-04, 1.1258e-04], device='cuda:1') 2023-04-26 22:02:26,905 INFO [finetune.py:976] (1/7) Epoch 9, batch 450, loss[loss=0.1786, simple_loss=0.2515, pruned_loss=0.05288, over 4806.00 frames. ], tot_loss[loss=0.1988, simple_loss=0.2644, pruned_loss=0.06663, over 857883.17 frames. ], batch size: 39, lr: 3.80e-03, grad_scale: 32.0 2023-04-26 22:02:29,459 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=46276.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:02:41,295 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.78 vs. limit=5.0 2023-04-26 22:02:55,932 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0462, 2.4206, 0.9300, 1.3091, 1.6432, 1.1834, 3.0445, 1.6975], device='cuda:1'), covar=tensor([0.0678, 0.0607, 0.0773, 0.1252, 0.0550, 0.0986, 0.0286, 0.0634], device='cuda:1'), in_proj_covar=tensor([0.0052, 0.0068, 0.0050, 0.0048, 0.0052, 0.0053, 0.0079, 0.0051], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0011, 0.0007], device='cuda:1') 2023-04-26 22:03:00,646 INFO [finetune.py:976] (1/7) Epoch 9, batch 500, loss[loss=0.248, simple_loss=0.2977, pruned_loss=0.09918, over 4757.00 frames. ], tot_loss[loss=0.1974, simple_loss=0.262, pruned_loss=0.06638, over 881574.11 frames. ], batch size: 28, lr: 3.80e-03, grad_scale: 32.0 2023-04-26 22:03:07,760 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.224e+02 1.712e+02 1.942e+02 2.347e+02 4.298e+02, threshold=3.884e+02, percent-clipped=0.0 2023-04-26 22:03:10,765 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=46337.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:03:25,994 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=46358.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:03:34,347 INFO [finetune.py:976] (1/7) Epoch 9, batch 550, loss[loss=0.2082, simple_loss=0.2582, pruned_loss=0.07916, over 4907.00 frames. ], tot_loss[loss=0.1944, simple_loss=0.2586, pruned_loss=0.06513, over 896845.46 frames. ], batch size: 36, lr: 3.80e-03, grad_scale: 32.0 2023-04-26 22:03:43,974 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.1620, 2.5943, 1.0538, 1.4171, 1.9453, 1.2546, 3.3384, 1.8755], device='cuda:1'), covar=tensor([0.0630, 0.0634, 0.0738, 0.1259, 0.0513, 0.0995, 0.0294, 0.0601], device='cuda:1'), in_proj_covar=tensor([0.0052, 0.0068, 0.0050, 0.0048, 0.0052, 0.0052, 0.0079, 0.0051], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0007], device='cuda:1') 2023-04-26 22:03:48,763 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.14 vs. limit=5.0 2023-04-26 22:03:56,736 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0046, 1.5426, 5.1319, 4.7754, 4.4968, 4.9670, 4.5045, 4.5511], device='cuda:1'), covar=tensor([0.7346, 0.6360, 0.1050, 0.1938, 0.1114, 0.1838, 0.1270, 0.1514], device='cuda:1'), in_proj_covar=tensor([0.0312, 0.0309, 0.0409, 0.0414, 0.0350, 0.0408, 0.0319, 0.0372], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-26 22:03:57,935 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=46406.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:04:07,673 INFO [finetune.py:976] (1/7) Epoch 9, batch 600, loss[loss=0.2141, simple_loss=0.2675, pruned_loss=0.08038, over 4902.00 frames. ], tot_loss[loss=0.1941, simple_loss=0.2583, pruned_loss=0.0649, over 909752.93 frames. ], batch size: 43, lr: 3.80e-03, grad_scale: 32.0 2023-04-26 22:04:12,568 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=46430.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:04:14,306 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.294e+02 1.692e+02 1.978e+02 2.446e+02 4.653e+02, threshold=3.955e+02, percent-clipped=2.0 2023-04-26 22:04:15,120 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.05 vs. limit=5.0 2023-04-26 22:04:21,555 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5361, 2.0093, 1.7241, 1.9821, 1.5958, 1.5687, 1.7095, 1.4151], device='cuda:1'), covar=tensor([0.2033, 0.1386, 0.0919, 0.1177, 0.3557, 0.1349, 0.2025, 0.2642], device='cuda:1'), in_proj_covar=tensor([0.0299, 0.0320, 0.0229, 0.0291, 0.0317, 0.0274, 0.0259, 0.0283], device='cuda:1'), out_proj_covar=tensor([1.2111e-04, 1.2903e-04, 9.2322e-05, 1.1643e-04, 1.3037e-04, 1.1041e-04, 1.0614e-04, 1.1365e-04], device='cuda:1') 2023-04-26 22:04:30,721 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0478, 2.5842, 1.2195, 1.4025, 2.0369, 1.3114, 3.1368, 1.6095], device='cuda:1'), covar=tensor([0.0699, 0.0960, 0.0806, 0.1247, 0.0463, 0.0927, 0.0226, 0.0647], device='cuda:1'), in_proj_covar=tensor([0.0052, 0.0068, 0.0050, 0.0048, 0.0052, 0.0053, 0.0080, 0.0052], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0011, 0.0008], device='cuda:1') 2023-04-26 22:04:40,995 INFO [finetune.py:976] (1/7) Epoch 9, batch 650, loss[loss=0.2491, simple_loss=0.2954, pruned_loss=0.1013, over 4749.00 frames. ], tot_loss[loss=0.1965, simple_loss=0.2612, pruned_loss=0.06587, over 921045.57 frames. ], batch size: 59, lr: 3.80e-03, grad_scale: 32.0 2023-04-26 22:04:44,729 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=46478.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:05:10,686 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=46515.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:05:14,901 INFO [finetune.py:976] (1/7) Epoch 9, batch 700, loss[loss=0.1882, simple_loss=0.2509, pruned_loss=0.0628, over 4882.00 frames. ], tot_loss[loss=0.1971, simple_loss=0.2624, pruned_loss=0.06585, over 927770.84 frames. ], batch size: 32, lr: 3.80e-03, grad_scale: 32.0 2023-04-26 22:05:21,596 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.297e+02 1.714e+02 2.090e+02 2.739e+02 4.841e+02, threshold=4.179e+02, percent-clipped=4.0 2023-04-26 22:05:39,607 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.0474, 1.0108, 1.2137, 1.1473, 1.0195, 0.8971, 1.0063, 0.5923], device='cuda:1'), covar=tensor([0.0549, 0.0751, 0.0623, 0.0632, 0.0771, 0.1328, 0.0544, 0.0923], device='cuda:1'), in_proj_covar=tensor([0.0067, 0.0074, 0.0072, 0.0067, 0.0076, 0.0097, 0.0079, 0.0074], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-26 22:05:43,197 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=46563.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:05:53,547 INFO [finetune.py:976] (1/7) Epoch 9, batch 750, loss[loss=0.22, simple_loss=0.2941, pruned_loss=0.07298, over 4910.00 frames. ], tot_loss[loss=0.1983, simple_loss=0.2644, pruned_loss=0.06611, over 933718.72 frames. ], batch size: 37, lr: 3.80e-03, grad_scale: 32.0 2023-04-26 22:06:18,452 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=46595.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:06:20,171 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0559, 2.5642, 1.0313, 1.3680, 1.9147, 1.2084, 3.3703, 1.7579], device='cuda:1'), covar=tensor([0.0730, 0.0723, 0.0913, 0.1371, 0.0540, 0.1045, 0.0266, 0.0658], device='cuda:1'), in_proj_covar=tensor([0.0052, 0.0068, 0.0050, 0.0048, 0.0052, 0.0053, 0.0079, 0.0051], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0011, 0.0007], device='cuda:1') 2023-04-26 22:06:28,359 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=46602.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:06:51,721 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7321, 1.2734, 1.4192, 1.4102, 1.9464, 1.6188, 1.3042, 1.3813], device='cuda:1'), covar=tensor([0.1923, 0.1406, 0.1872, 0.1483, 0.0778, 0.1325, 0.1869, 0.1775], device='cuda:1'), in_proj_covar=tensor([0.0312, 0.0324, 0.0354, 0.0301, 0.0341, 0.0325, 0.0308, 0.0356], device='cuda:1'), out_proj_covar=tensor([6.6088e-05, 6.8751e-05, 7.6413e-05, 6.2314e-05, 7.1761e-05, 6.9741e-05, 6.6381e-05, 7.6504e-05], device='cuda:1') 2023-04-26 22:06:57,598 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.16 vs. limit=5.0 2023-04-26 22:06:58,615 INFO [finetune.py:976] (1/7) Epoch 9, batch 800, loss[loss=0.1985, simple_loss=0.2706, pruned_loss=0.06317, over 4754.00 frames. ], tot_loss[loss=0.197, simple_loss=0.263, pruned_loss=0.06554, over 938513.99 frames. ], batch size: 28, lr: 3.80e-03, grad_scale: 64.0 2023-04-26 22:07:09,958 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=46632.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:07:10,488 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.096e+02 1.620e+02 1.991e+02 2.380e+02 4.805e+02, threshold=3.982e+02, percent-clipped=1.0 2023-04-26 22:07:25,511 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([3.5165, 3.4371, 2.5442, 4.1046, 3.4867, 3.6104, 1.5360, 3.4953], device='cuda:1'), covar=tensor([0.1807, 0.1403, 0.3504, 0.1833, 0.2953, 0.1874, 0.5281, 0.2328], device='cuda:1'), in_proj_covar=tensor([0.0244, 0.0217, 0.0252, 0.0306, 0.0301, 0.0252, 0.0271, 0.0273], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-26 22:07:26,182 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=46656.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 22:07:28,570 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.36 vs. limit=5.0 2023-04-26 22:07:31,864 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=46663.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:07:37,206 INFO [finetune.py:976] (1/7) Epoch 9, batch 850, loss[loss=0.1903, simple_loss=0.2454, pruned_loss=0.06763, over 4896.00 frames. ], tot_loss[loss=0.1943, simple_loss=0.2603, pruned_loss=0.06414, over 944137.15 frames. ], batch size: 35, lr: 3.80e-03, grad_scale: 64.0 2023-04-26 22:07:44,571 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.1644, 1.5721, 1.3951, 1.7577, 1.5960, 1.8443, 1.3341, 3.2881], device='cuda:1'), covar=tensor([0.0647, 0.0747, 0.0768, 0.1175, 0.0623, 0.0563, 0.0754, 0.0144], device='cuda:1'), in_proj_covar=tensor([0.0039, 0.0039, 0.0040, 0.0044, 0.0040, 0.0039, 0.0039, 0.0059], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:1') 2023-04-26 22:08:10,356 INFO [finetune.py:976] (1/7) Epoch 9, batch 900, loss[loss=0.1895, simple_loss=0.253, pruned_loss=0.06305, over 4899.00 frames. ], tot_loss[loss=0.1914, simple_loss=0.2567, pruned_loss=0.06307, over 947900.76 frames. ], batch size: 32, lr: 3.80e-03, grad_scale: 64.0 2023-04-26 22:08:17,019 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.197e+02 1.643e+02 2.086e+02 2.450e+02 5.487e+02, threshold=4.172e+02, percent-clipped=3.0 2023-04-26 22:08:43,449 INFO [finetune.py:976] (1/7) Epoch 9, batch 950, loss[loss=0.2196, simple_loss=0.2728, pruned_loss=0.08316, over 4912.00 frames. ], tot_loss[loss=0.1914, simple_loss=0.2563, pruned_loss=0.06328, over 950124.39 frames. ], batch size: 36, lr: 3.80e-03, grad_scale: 64.0 2023-04-26 22:08:56,372 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5369, 1.4875, 1.9441, 1.8454, 1.4726, 1.2618, 1.5863, 0.9989], device='cuda:1'), covar=tensor([0.0767, 0.0873, 0.0461, 0.0857, 0.0914, 0.1291, 0.0779, 0.0935], device='cuda:1'), in_proj_covar=tensor([0.0067, 0.0073, 0.0072, 0.0067, 0.0076, 0.0096, 0.0079, 0.0074], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-26 22:09:07,654 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=46810.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:09:16,375 INFO [finetune.py:976] (1/7) Epoch 9, batch 1000, loss[loss=0.198, simple_loss=0.2705, pruned_loss=0.06279, over 4817.00 frames. ], tot_loss[loss=0.1926, simple_loss=0.258, pruned_loss=0.06355, over 952625.42 frames. ], batch size: 51, lr: 3.80e-03, grad_scale: 64.0 2023-04-26 22:09:22,994 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.171e+02 1.599e+02 1.946e+02 2.521e+02 4.625e+02, threshold=3.893e+02, percent-clipped=2.0 2023-04-26 22:09:26,870 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-04-26 22:09:49,147 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=46871.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:09:49,626 INFO [finetune.py:976] (1/7) Epoch 9, batch 1050, loss[loss=0.2279, simple_loss=0.2952, pruned_loss=0.08031, over 4904.00 frames. ], tot_loss[loss=0.1938, simple_loss=0.26, pruned_loss=0.06378, over 951816.30 frames. ], batch size: 43, lr: 3.80e-03, grad_scale: 64.0 2023-04-26 22:09:52,894 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.71 vs. limit=2.0 2023-04-26 22:10:12,225 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.22 vs. limit=5.0 2023-04-26 22:10:22,798 INFO [finetune.py:976] (1/7) Epoch 9, batch 1100, loss[loss=0.2139, simple_loss=0.2851, pruned_loss=0.07137, over 4928.00 frames. ], tot_loss[loss=0.1976, simple_loss=0.2635, pruned_loss=0.06584, over 953174.45 frames. ], batch size: 38, lr: 3.80e-03, grad_scale: 32.0 2023-04-26 22:10:27,187 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-04-26 22:10:29,446 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=46932.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:10:30,585 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.171e+02 1.585e+02 1.927e+02 2.625e+02 4.611e+02, threshold=3.853e+02, percent-clipped=4.0 2023-04-26 22:10:39,158 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.1369, 0.7074, 0.9234, 0.7743, 1.2704, 0.9477, 0.8594, 0.9562], device='cuda:1'), covar=tensor([0.1468, 0.1515, 0.1844, 0.1482, 0.0847, 0.1241, 0.1590, 0.1841], device='cuda:1'), in_proj_covar=tensor([0.0310, 0.0324, 0.0353, 0.0301, 0.0339, 0.0324, 0.0309, 0.0354], device='cuda:1'), out_proj_covar=tensor([6.5763e-05, 6.8811e-05, 7.6274e-05, 6.2285e-05, 7.1210e-05, 6.9480e-05, 6.6426e-05, 7.6008e-05], device='cuda:1') 2023-04-26 22:10:39,174 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7963, 1.2943, 1.5895, 1.6076, 1.5525, 1.2704, 0.6812, 1.2545], device='cuda:1'), covar=tensor([0.3455, 0.3802, 0.1889, 0.2487, 0.2856, 0.2933, 0.4978, 0.2499], device='cuda:1'), in_proj_covar=tensor([0.0284, 0.0252, 0.0220, 0.0321, 0.0214, 0.0230, 0.0236, 0.0188], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-26 22:10:40,916 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=46951.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 22:10:45,716 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=46958.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:10:56,081 INFO [finetune.py:976] (1/7) Epoch 9, batch 1150, loss[loss=0.1747, simple_loss=0.2396, pruned_loss=0.05493, over 4291.00 frames. ], tot_loss[loss=0.1975, simple_loss=0.2637, pruned_loss=0.06568, over 951968.95 frames. ], batch size: 66, lr: 3.80e-03, grad_scale: 32.0 2023-04-26 22:11:01,460 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=46980.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:11:20,711 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7331, 1.7141, 0.8783, 1.3952, 1.9172, 1.5909, 1.5317, 1.5259], device='cuda:1'), covar=tensor([0.0496, 0.0369, 0.0371, 0.0555, 0.0278, 0.0535, 0.0496, 0.0581], device='cuda:1'), in_proj_covar=tensor([0.0030, 0.0026, 0.0024, 0.0030, 0.0021, 0.0029, 0.0029, 0.0030], device='cuda:1'), out_proj_covar=tensor([0.0049, 0.0044, 0.0038, 0.0050, 0.0038, 0.0048, 0.0048, 0.0050], device='cuda:1') 2023-04-26 22:11:43,958 INFO [finetune.py:976] (1/7) Epoch 9, batch 1200, loss[loss=0.1874, simple_loss=0.2591, pruned_loss=0.05783, over 4907.00 frames. ], tot_loss[loss=0.1958, simple_loss=0.2619, pruned_loss=0.06486, over 952507.78 frames. ], batch size: 37, lr: 3.80e-03, grad_scale: 32.0 2023-04-26 22:12:03,718 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.043e+02 1.560e+02 1.899e+02 2.282e+02 5.310e+02, threshold=3.798e+02, percent-clipped=1.0 2023-04-26 22:12:04,533 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-04-26 22:12:55,840 INFO [finetune.py:976] (1/7) Epoch 9, batch 1250, loss[loss=0.1863, simple_loss=0.2489, pruned_loss=0.06183, over 4890.00 frames. ], tot_loss[loss=0.193, simple_loss=0.2583, pruned_loss=0.06387, over 952467.92 frames. ], batch size: 32, lr: 3.80e-03, grad_scale: 32.0 2023-04-26 22:13:10,240 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8467, 2.4042, 2.0018, 2.1825, 1.7640, 2.1158, 2.0360, 1.5401], device='cuda:1'), covar=tensor([0.2404, 0.1252, 0.0996, 0.1366, 0.3092, 0.1028, 0.2040, 0.3111], device='cuda:1'), in_proj_covar=tensor([0.0303, 0.0323, 0.0232, 0.0296, 0.0321, 0.0276, 0.0263, 0.0288], device='cuda:1'), out_proj_covar=tensor([1.2295e-04, 1.3025e-04, 9.3588e-05, 1.1835e-04, 1.3209e-04, 1.1146e-04, 1.0773e-04, 1.1572e-04], device='cuda:1') 2023-04-26 22:13:21,127 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.1077, 2.2003, 1.9553, 2.0150, 2.4420, 1.8345, 2.9901, 1.7298], device='cuda:1'), covar=tensor([0.4095, 0.2125, 0.5503, 0.2926, 0.1859, 0.2904, 0.1546, 0.4735], device='cuda:1'), in_proj_covar=tensor([0.0347, 0.0352, 0.0432, 0.0364, 0.0389, 0.0387, 0.0384, 0.0423], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 22:14:00,073 INFO [finetune.py:976] (1/7) Epoch 9, batch 1300, loss[loss=0.1969, simple_loss=0.2571, pruned_loss=0.06837, over 4815.00 frames. ], tot_loss[loss=0.1916, simple_loss=0.2561, pruned_loss=0.06355, over 953289.44 frames. ], batch size: 51, lr: 3.80e-03, grad_scale: 32.0 2023-04-26 22:14:15,264 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.174e+02 1.673e+02 1.914e+02 2.315e+02 4.014e+02, threshold=3.828e+02, percent-clipped=2.0 2023-04-26 22:14:35,995 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9808, 2.3216, 0.9723, 1.2927, 1.6965, 1.1500, 2.9726, 1.5403], device='cuda:1'), covar=tensor([0.0685, 0.0644, 0.0786, 0.1232, 0.0536, 0.1030, 0.0255, 0.0632], device='cuda:1'), in_proj_covar=tensor([0.0052, 0.0067, 0.0050, 0.0047, 0.0052, 0.0052, 0.0078, 0.0051], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0007], device='cuda:1') 2023-04-26 22:14:58,580 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=47166.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:15:07,950 INFO [finetune.py:976] (1/7) Epoch 9, batch 1350, loss[loss=0.2476, simple_loss=0.3216, pruned_loss=0.08675, over 4764.00 frames. ], tot_loss[loss=0.1935, simple_loss=0.2575, pruned_loss=0.06469, over 951657.47 frames. ], batch size: 54, lr: 3.80e-03, grad_scale: 32.0 2023-04-26 22:15:58,006 INFO [finetune.py:976] (1/7) Epoch 9, batch 1400, loss[loss=0.2016, simple_loss=0.293, pruned_loss=0.05509, over 4803.00 frames. ], tot_loss[loss=0.1963, simple_loss=0.261, pruned_loss=0.06581, over 954020.64 frames. ], batch size: 45, lr: 3.80e-03, grad_scale: 32.0 2023-04-26 22:16:04,635 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.3599, 1.6199, 1.6734, 1.7928, 1.6940, 1.9041, 1.7947, 1.7364], device='cuda:1'), covar=tensor([0.5287, 0.7322, 0.6740, 0.6405, 0.7393, 1.0256, 0.7214, 0.6720], device='cuda:1'), in_proj_covar=tensor([0.0323, 0.0383, 0.0317, 0.0326, 0.0342, 0.0403, 0.0364, 0.0327], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-26 22:16:06,780 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.248e+02 1.819e+02 2.122e+02 2.454e+02 4.582e+02, threshold=4.244e+02, percent-clipped=5.0 2023-04-26 22:16:08,285 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-04-26 22:16:08,816 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.1064, 0.7098, 0.9071, 0.7110, 1.2027, 0.9843, 0.8670, 0.9317], device='cuda:1'), covar=tensor([0.1414, 0.1378, 0.1742, 0.1492, 0.0800, 0.1255, 0.1665, 0.1889], device='cuda:1'), in_proj_covar=tensor([0.0312, 0.0326, 0.0357, 0.0303, 0.0341, 0.0327, 0.0311, 0.0358], device='cuda:1'), out_proj_covar=tensor([6.6169e-05, 6.9295e-05, 7.7158e-05, 6.2828e-05, 7.1482e-05, 7.0335e-05, 6.6951e-05, 7.6858e-05], device='cuda:1') 2023-04-26 22:16:09,451 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4937, 1.3251, 1.8188, 1.7129, 1.4116, 1.2219, 1.4611, 1.0723], device='cuda:1'), covar=tensor([0.0683, 0.0938, 0.0504, 0.0706, 0.0834, 0.1197, 0.0777, 0.0787], device='cuda:1'), in_proj_covar=tensor([0.0067, 0.0074, 0.0073, 0.0068, 0.0077, 0.0096, 0.0079, 0.0075], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-26 22:16:11,735 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=47241.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:16:17,771 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=47251.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:16:22,000 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=47258.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:16:30,889 INFO [finetune.py:976] (1/7) Epoch 9, batch 1450, loss[loss=0.2016, simple_loss=0.2729, pruned_loss=0.06522, over 4821.00 frames. ], tot_loss[loss=0.1977, simple_loss=0.2631, pruned_loss=0.06614, over 954058.89 frames. ], batch size: 39, lr: 3.80e-03, grad_scale: 32.0 2023-04-26 22:16:50,059 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=47299.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:16:51,908 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=47302.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:16:54,266 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=47306.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:17:03,913 INFO [finetune.py:976] (1/7) Epoch 9, batch 1500, loss[loss=0.213, simple_loss=0.2775, pruned_loss=0.07424, over 4923.00 frames. ], tot_loss[loss=0.1978, simple_loss=0.263, pruned_loss=0.06629, over 953031.41 frames. ], batch size: 38, lr: 3.80e-03, grad_scale: 32.0 2023-04-26 22:17:12,630 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.146e+02 1.650e+02 1.825e+02 2.290e+02 4.290e+02, threshold=3.651e+02, percent-clipped=1.0 2023-04-26 22:18:02,736 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.3753, 1.7759, 2.2310, 2.7866, 2.2042, 1.7761, 1.6612, 1.9804], device='cuda:1'), covar=tensor([0.4002, 0.4029, 0.1938, 0.3454, 0.3486, 0.3161, 0.4746, 0.3173], device='cuda:1'), in_proj_covar=tensor([0.0281, 0.0251, 0.0219, 0.0319, 0.0213, 0.0228, 0.0234, 0.0187], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-26 22:18:03,817 INFO [finetune.py:976] (1/7) Epoch 9, batch 1550, loss[loss=0.1733, simple_loss=0.2407, pruned_loss=0.05292, over 4804.00 frames. ], tot_loss[loss=0.1964, simple_loss=0.2624, pruned_loss=0.06521, over 955897.38 frames. ], batch size: 45, lr: 3.80e-03, grad_scale: 32.0 2023-04-26 22:18:13,627 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=47379.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:18:14,290 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5846, 1.0971, 1.6615, 2.1226, 1.7391, 1.5589, 1.6463, 1.6530], device='cuda:1'), covar=tensor([0.6156, 0.8817, 0.8330, 0.8091, 0.7532, 0.9996, 0.9819, 0.9563], device='cuda:1'), in_proj_covar=tensor([0.0409, 0.0421, 0.0505, 0.0524, 0.0438, 0.0458, 0.0468, 0.0466], device='cuda:1'), out_proj_covar=tensor([9.9631e-05, 1.0440e-04, 1.1404e-04, 1.2451e-04, 1.0651e-04, 1.1073e-04, 1.1255e-04, 1.1267e-04], device='cuda:1') 2023-04-26 22:19:10,271 INFO [finetune.py:976] (1/7) Epoch 9, batch 1600, loss[loss=0.1495, simple_loss=0.209, pruned_loss=0.04502, over 4805.00 frames. ], tot_loss[loss=0.194, simple_loss=0.26, pruned_loss=0.06406, over 956538.91 frames. ], batch size: 25, lr: 3.80e-03, grad_scale: 32.0 2023-04-26 22:19:23,854 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.166e+02 1.649e+02 1.912e+02 2.310e+02 4.649e+02, threshold=3.824e+02, percent-clipped=3.0 2023-04-26 22:19:34,502 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=47440.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:19:57,011 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.78 vs. limit=2.0 2023-04-26 22:20:08,741 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=47466.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:20:17,937 INFO [finetune.py:976] (1/7) Epoch 9, batch 1650, loss[loss=0.1457, simple_loss=0.208, pruned_loss=0.0417, over 4721.00 frames. ], tot_loss[loss=0.1899, simple_loss=0.2557, pruned_loss=0.06208, over 958056.87 frames. ], batch size: 23, lr: 3.80e-03, grad_scale: 32.0 2023-04-26 22:20:39,812 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.47 vs. limit=5.0 2023-04-26 22:21:00,462 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=47514.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:21:00,698 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-04-26 22:21:05,270 INFO [finetune.py:976] (1/7) Epoch 9, batch 1700, loss[loss=0.1321, simple_loss=0.208, pruned_loss=0.0281, over 4797.00 frames. ], tot_loss[loss=0.1899, simple_loss=0.2553, pruned_loss=0.06229, over 957914.15 frames. ], batch size: 25, lr: 3.80e-03, grad_scale: 32.0 2023-04-26 22:21:12,555 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.160e+02 1.638e+02 2.054e+02 2.588e+02 3.948e+02, threshold=4.108e+02, percent-clipped=3.0 2023-04-26 22:21:38,482 INFO [finetune.py:976] (1/7) Epoch 9, batch 1750, loss[loss=0.1952, simple_loss=0.2721, pruned_loss=0.05913, over 4762.00 frames. ], tot_loss[loss=0.1931, simple_loss=0.2585, pruned_loss=0.06387, over 955755.53 frames. ], batch size: 54, lr: 3.80e-03, grad_scale: 32.0 2023-04-26 22:21:48,883 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-04-26 22:21:51,424 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.84 vs. limit=5.0 2023-04-26 22:21:54,721 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=47597.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:22:11,819 INFO [finetune.py:976] (1/7) Epoch 9, batch 1800, loss[loss=0.1895, simple_loss=0.2591, pruned_loss=0.05998, over 4907.00 frames. ], tot_loss[loss=0.196, simple_loss=0.2622, pruned_loss=0.06493, over 957281.80 frames. ], batch size: 36, lr: 3.79e-03, grad_scale: 32.0 2023-04-26 22:22:16,846 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8735, 2.4496, 1.9610, 2.2619, 1.7106, 2.0923, 2.1208, 1.5869], device='cuda:1'), covar=tensor([0.2119, 0.1309, 0.0974, 0.1154, 0.3299, 0.1242, 0.1778, 0.2678], device='cuda:1'), in_proj_covar=tensor([0.0299, 0.0319, 0.0229, 0.0292, 0.0317, 0.0274, 0.0260, 0.0284], device='cuda:1'), out_proj_covar=tensor([1.2102e-04, 1.2828e-04, 9.2224e-05, 1.1705e-04, 1.3037e-04, 1.1026e-04, 1.0636e-04, 1.1396e-04], device='cuda:1') 2023-04-26 22:22:19,171 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.171e+02 1.815e+02 2.187e+02 2.572e+02 4.070e+02, threshold=4.375e+02, percent-clipped=0.0 2023-04-26 22:22:45,311 INFO [finetune.py:976] (1/7) Epoch 9, batch 1850, loss[loss=0.1795, simple_loss=0.2385, pruned_loss=0.0603, over 4704.00 frames. ], tot_loss[loss=0.1981, simple_loss=0.264, pruned_loss=0.06606, over 955753.43 frames. ], batch size: 23, lr: 3.79e-03, grad_scale: 32.0 2023-04-26 22:23:44,011 INFO [finetune.py:976] (1/7) Epoch 9, batch 1900, loss[loss=0.1759, simple_loss=0.254, pruned_loss=0.04892, over 4746.00 frames. ], tot_loss[loss=0.1986, simple_loss=0.265, pruned_loss=0.06612, over 954977.59 frames. ], batch size: 27, lr: 3.79e-03, grad_scale: 16.0 2023-04-26 22:24:02,108 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.223e+02 1.717e+02 1.936e+02 2.333e+02 4.969e+02, threshold=3.871e+02, percent-clipped=1.0 2023-04-26 22:24:02,194 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=47735.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:24:04,112 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=47738.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:24:33,494 INFO [finetune.py:976] (1/7) Epoch 9, batch 1950, loss[loss=0.2139, simple_loss=0.2629, pruned_loss=0.08249, over 4855.00 frames. ], tot_loss[loss=0.1968, simple_loss=0.2631, pruned_loss=0.06522, over 956833.72 frames. ], batch size: 31, lr: 3.79e-03, grad_scale: 16.0 2023-04-26 22:24:43,262 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=47788.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:24:44,615 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-04-26 22:24:50,428 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=47799.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:24:50,496 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=3.60 vs. limit=5.0 2023-04-26 22:24:57,040 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=47809.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 22:25:11,249 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.84 vs. limit=5.0 2023-04-26 22:25:11,660 INFO [finetune.py:976] (1/7) Epoch 9, batch 2000, loss[loss=0.1807, simple_loss=0.2491, pruned_loss=0.05615, over 4828.00 frames. ], tot_loss[loss=0.1953, simple_loss=0.2605, pruned_loss=0.06507, over 956973.20 frames. ], batch size: 40, lr: 3.79e-03, grad_scale: 16.0 2023-04-26 22:25:30,746 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.076e+02 1.686e+02 2.019e+02 2.338e+02 4.477e+02, threshold=4.037e+02, percent-clipped=4.0 2023-04-26 22:25:44,891 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=47849.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:26:17,390 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=47870.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 22:26:24,626 INFO [finetune.py:976] (1/7) Epoch 9, batch 2050, loss[loss=0.2278, simple_loss=0.2835, pruned_loss=0.086, over 4822.00 frames. ], tot_loss[loss=0.193, simple_loss=0.2576, pruned_loss=0.06426, over 956117.82 frames. ], batch size: 30, lr: 3.79e-03, grad_scale: 16.0 2023-04-26 22:26:50,445 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-04-26 22:26:52,053 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=47897.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:27:30,785 INFO [finetune.py:976] (1/7) Epoch 9, batch 2100, loss[loss=0.1721, simple_loss=0.25, pruned_loss=0.04707, over 4853.00 frames. ], tot_loss[loss=0.193, simple_loss=0.2577, pruned_loss=0.06421, over 955020.13 frames. ], batch size: 49, lr: 3.79e-03, grad_scale: 16.0 2023-04-26 22:27:39,268 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.172e+02 1.719e+02 2.114e+02 2.573e+02 5.213e+02, threshold=4.228e+02, percent-clipped=2.0 2023-04-26 22:27:45,498 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=47945.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:27:48,595 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6521, 2.0113, 1.9514, 2.3810, 2.0917, 2.1828, 1.7710, 4.6852], device='cuda:1'), covar=tensor([0.0571, 0.0681, 0.0733, 0.1049, 0.0571, 0.0527, 0.0716, 0.0103], device='cuda:1'), in_proj_covar=tensor([0.0038, 0.0039, 0.0040, 0.0044, 0.0040, 0.0039, 0.0039, 0.0059], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0014, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:1') 2023-04-26 22:27:49,208 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([0.8240, 1.8346, 1.7526, 1.5116, 1.8698, 1.5231, 2.5110, 1.4170], device='cuda:1'), covar=tensor([0.3257, 0.1482, 0.3913, 0.2514, 0.1594, 0.2514, 0.1258, 0.4290], device='cuda:1'), in_proj_covar=tensor([0.0348, 0.0353, 0.0432, 0.0362, 0.0391, 0.0387, 0.0383, 0.0422], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 22:28:20,274 INFO [finetune.py:976] (1/7) Epoch 9, batch 2150, loss[loss=0.1918, simple_loss=0.2603, pruned_loss=0.06168, over 4738.00 frames. ], tot_loss[loss=0.1953, simple_loss=0.2599, pruned_loss=0.0653, over 955904.32 frames. ], batch size: 54, lr: 3.79e-03, grad_scale: 16.0 2023-04-26 22:28:26,464 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5682, 1.0104, 1.5200, 2.0375, 1.6667, 1.5377, 1.5500, 1.6404], device='cuda:1'), covar=tensor([0.6175, 0.8653, 0.8952, 0.8798, 0.7850, 1.0458, 1.1011, 0.9438], device='cuda:1'), in_proj_covar=tensor([0.0409, 0.0422, 0.0506, 0.0524, 0.0440, 0.0459, 0.0469, 0.0467], device='cuda:1'), out_proj_covar=tensor([9.9683e-05, 1.0458e-04, 1.1434e-04, 1.2455e-04, 1.0699e-04, 1.1099e-04, 1.1286e-04, 1.1315e-04], device='cuda:1') 2023-04-26 22:29:17,268 INFO [finetune.py:976] (1/7) Epoch 9, batch 2200, loss[loss=0.184, simple_loss=0.2525, pruned_loss=0.05776, over 4871.00 frames. ], tot_loss[loss=0.1978, simple_loss=0.2633, pruned_loss=0.06617, over 956178.65 frames. ], batch size: 34, lr: 3.79e-03, grad_scale: 16.0 2023-04-26 22:29:32,259 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.341e+02 1.685e+02 2.065e+02 2.367e+02 5.445e+02, threshold=4.130e+02, percent-clipped=2.0 2023-04-26 22:29:32,359 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=48035.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:29:44,114 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.1590, 2.6190, 0.9568, 1.3723, 1.9404, 1.2666, 3.6406, 1.9307], device='cuda:1'), covar=tensor([0.0654, 0.0810, 0.0854, 0.1303, 0.0547, 0.1018, 0.0302, 0.0629], device='cuda:1'), in_proj_covar=tensor([0.0051, 0.0068, 0.0050, 0.0047, 0.0052, 0.0053, 0.0078, 0.0051], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0007], device='cuda:1') 2023-04-26 22:30:13,121 INFO [finetune.py:976] (1/7) Epoch 9, batch 2250, loss[loss=0.2186, simple_loss=0.282, pruned_loss=0.07761, over 4817.00 frames. ], tot_loss[loss=0.1973, simple_loss=0.2635, pruned_loss=0.06555, over 957500.94 frames. ], batch size: 25, lr: 3.79e-03, grad_scale: 16.0 2023-04-26 22:30:18,370 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=48072.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:30:32,861 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=48083.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:30:45,744 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=48094.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:31:21,272 INFO [finetune.py:976] (1/7) Epoch 9, batch 2300, loss[loss=0.1969, simple_loss=0.2671, pruned_loss=0.06334, over 4882.00 frames. ], tot_loss[loss=0.1968, simple_loss=0.2632, pruned_loss=0.06517, over 957443.10 frames. ], batch size: 35, lr: 3.79e-03, grad_scale: 16.0 2023-04-26 22:31:22,676 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-04-26 22:31:32,176 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9928, 2.6472, 1.9671, 1.7917, 1.3796, 1.4253, 1.9886, 1.3363], device='cuda:1'), covar=tensor([0.1786, 0.1597, 0.1642, 0.2066, 0.2770, 0.2159, 0.1214, 0.2218], device='cuda:1'), in_proj_covar=tensor([0.0194, 0.0215, 0.0170, 0.0204, 0.0205, 0.0183, 0.0161, 0.0187], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-26 22:31:35,701 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=48133.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:31:36,814 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.203e+02 1.671e+02 1.981e+02 2.243e+02 3.920e+02, threshold=3.963e+02, percent-clipped=0.0 2023-04-26 22:31:50,252 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=48144.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:32:14,432 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=48165.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 22:32:25,262 INFO [finetune.py:976] (1/7) Epoch 9, batch 2350, loss[loss=0.2088, simple_loss=0.2696, pruned_loss=0.07399, over 4738.00 frames. ], tot_loss[loss=0.1945, simple_loss=0.2607, pruned_loss=0.06418, over 957435.32 frames. ], batch size: 59, lr: 3.79e-03, grad_scale: 16.0 2023-04-26 22:32:36,985 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.4433, 1.9090, 2.3003, 2.8508, 2.4003, 1.8158, 1.6553, 2.2117], device='cuda:1'), covar=tensor([0.3503, 0.3478, 0.1733, 0.3173, 0.3151, 0.2954, 0.4839, 0.2803], device='cuda:1'), in_proj_covar=tensor([0.0280, 0.0250, 0.0219, 0.0318, 0.0212, 0.0227, 0.0233, 0.0186], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-26 22:32:47,980 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8775, 2.2149, 1.2980, 1.6435, 2.1079, 1.7482, 1.7303, 1.8059], device='cuda:1'), covar=tensor([0.0589, 0.0310, 0.0332, 0.0612, 0.0251, 0.0642, 0.0601, 0.0596], device='cuda:1'), in_proj_covar=tensor([0.0030, 0.0026, 0.0023, 0.0030, 0.0021, 0.0029, 0.0029, 0.0030], device='cuda:1'), out_proj_covar=tensor([0.0049, 0.0044, 0.0038, 0.0050, 0.0037, 0.0048, 0.0048, 0.0050], device='cuda:1') 2023-04-26 22:33:19,226 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=48212.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 22:33:30,439 INFO [finetune.py:976] (1/7) Epoch 9, batch 2400, loss[loss=0.1946, simple_loss=0.2564, pruned_loss=0.06635, over 4939.00 frames. ], tot_loss[loss=0.1925, simple_loss=0.258, pruned_loss=0.06352, over 957297.88 frames. ], batch size: 33, lr: 3.79e-03, grad_scale: 16.0 2023-04-26 22:33:45,094 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.152e+02 1.712e+02 1.962e+02 2.337e+02 6.803e+02, threshold=3.925e+02, percent-clipped=4.0 2023-04-26 22:34:34,223 INFO [finetune.py:976] (1/7) Epoch 9, batch 2450, loss[loss=0.2202, simple_loss=0.2835, pruned_loss=0.07842, over 4918.00 frames. ], tot_loss[loss=0.1898, simple_loss=0.2547, pruned_loss=0.06242, over 956710.61 frames. ], batch size: 37, lr: 3.79e-03, grad_scale: 16.0 2023-04-26 22:34:34,997 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=48273.0, num_to_drop=1, layers_to_drop={3} 2023-04-26 22:34:39,137 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.56 vs. limit=5.0 2023-04-26 22:35:25,353 INFO [finetune.py:976] (1/7) Epoch 9, batch 2500, loss[loss=0.2014, simple_loss=0.2779, pruned_loss=0.06248, over 4815.00 frames. ], tot_loss[loss=0.1934, simple_loss=0.258, pruned_loss=0.06442, over 955281.07 frames. ], batch size: 38, lr: 3.79e-03, grad_scale: 16.0 2023-04-26 22:35:26,649 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=48324.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:35:32,711 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9876, 1.4836, 1.8198, 2.1152, 1.8283, 1.4577, 0.9542, 1.5209], device='cuda:1'), covar=tensor([0.3973, 0.3819, 0.1972, 0.2617, 0.2935, 0.3092, 0.5181, 0.2798], device='cuda:1'), in_proj_covar=tensor([0.0282, 0.0251, 0.0220, 0.0320, 0.0213, 0.0228, 0.0235, 0.0187], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-26 22:35:34,244 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.038e+02 1.734e+02 2.061e+02 2.480e+02 6.002e+02, threshold=4.122e+02, percent-clipped=6.0 2023-04-26 22:35:53,080 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4478, 3.7203, 0.9253, 1.8354, 1.8375, 2.4590, 2.1647, 0.8465], device='cuda:1'), covar=tensor([0.1822, 0.1528, 0.2527, 0.1809, 0.1386, 0.1351, 0.1740, 0.2540], device='cuda:1'), in_proj_covar=tensor([0.0117, 0.0251, 0.0142, 0.0123, 0.0136, 0.0155, 0.0118, 0.0121], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-26 22:36:14,628 INFO [finetune.py:976] (1/7) Epoch 9, batch 2550, loss[loss=0.1973, simple_loss=0.2671, pruned_loss=0.06378, over 4820.00 frames. ], tot_loss[loss=0.1959, simple_loss=0.2613, pruned_loss=0.06526, over 954913.93 frames. ], batch size: 38, lr: 3.79e-03, grad_scale: 16.0 2023-04-26 22:36:34,694 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=48385.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:36:47,904 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=48394.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:36:55,534 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=48397.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:37:28,005 INFO [finetune.py:976] (1/7) Epoch 9, batch 2600, loss[loss=0.197, simple_loss=0.2676, pruned_loss=0.06321, over 4843.00 frames. ], tot_loss[loss=0.1973, simple_loss=0.2631, pruned_loss=0.06574, over 955225.06 frames. ], batch size: 44, lr: 3.79e-03, grad_scale: 16.0 2023-04-26 22:37:28,730 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=48423.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:37:31,742 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=48428.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:37:41,881 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.066e+02 1.714e+02 2.123e+02 2.498e+02 4.905e+02, threshold=4.246e+02, percent-clipped=1.0 2023-04-26 22:37:52,435 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=48442.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:37:54,158 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=48444.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:38:13,984 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=48458.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:38:24,203 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=48465.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 22:38:34,661 INFO [finetune.py:976] (1/7) Epoch 9, batch 2650, loss[loss=0.1963, simple_loss=0.2569, pruned_loss=0.06782, over 4118.00 frames. ], tot_loss[loss=0.1985, simple_loss=0.2644, pruned_loss=0.06632, over 953474.05 frames. ], batch size: 65, lr: 3.79e-03, grad_scale: 16.0 2023-04-26 22:38:46,944 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=48484.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:38:57,118 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=48492.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:39:21,187 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=48513.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 22:39:32,428 INFO [finetune.py:976] (1/7) Epoch 9, batch 2700, loss[loss=0.1806, simple_loss=0.2587, pruned_loss=0.05128, over 4810.00 frames. ], tot_loss[loss=0.1968, simple_loss=0.2629, pruned_loss=0.06531, over 954301.17 frames. ], batch size: 45, lr: 3.79e-03, grad_scale: 16.0 2023-04-26 22:39:40,875 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.141e+02 1.732e+02 1.990e+02 2.443e+02 3.754e+02, threshold=3.980e+02, percent-clipped=0.0 2023-04-26 22:40:20,119 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=48568.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 22:40:28,389 INFO [finetune.py:976] (1/7) Epoch 9, batch 2750, loss[loss=0.1792, simple_loss=0.2361, pruned_loss=0.0611, over 4936.00 frames. ], tot_loss[loss=0.1943, simple_loss=0.2602, pruned_loss=0.06421, over 956091.30 frames. ], batch size: 33, lr: 3.79e-03, grad_scale: 16.0 2023-04-26 22:41:33,773 INFO [finetune.py:976] (1/7) Epoch 9, batch 2800, loss[loss=0.1677, simple_loss=0.2414, pruned_loss=0.04704, over 4921.00 frames. ], tot_loss[loss=0.1917, simple_loss=0.2566, pruned_loss=0.06343, over 955546.15 frames. ], batch size: 43, lr: 3.79e-03, grad_scale: 16.0 2023-04-26 22:41:34,462 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9228, 4.3132, 0.8278, 2.1765, 2.4333, 2.8574, 2.5248, 1.0506], device='cuda:1'), covar=tensor([0.1414, 0.0911, 0.2379, 0.1426, 0.1109, 0.1178, 0.1511, 0.2180], device='cuda:1'), in_proj_covar=tensor([0.0117, 0.0251, 0.0142, 0.0123, 0.0135, 0.0155, 0.0118, 0.0121], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-26 22:41:43,989 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5854, 1.0135, 1.5239, 2.0830, 1.7267, 1.5024, 1.5274, 1.5836], device='cuda:1'), covar=tensor([0.5734, 0.8061, 0.7818, 0.7715, 0.7040, 0.9889, 0.9419, 0.9778], device='cuda:1'), in_proj_covar=tensor([0.0409, 0.0422, 0.0504, 0.0523, 0.0439, 0.0459, 0.0470, 0.0467], device='cuda:1'), out_proj_covar=tensor([9.9607e-05, 1.0455e-04, 1.1384e-04, 1.2437e-04, 1.0656e-04, 1.1087e-04, 1.1302e-04, 1.1301e-04], device='cuda:1') 2023-04-26 22:41:46,303 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.102e+02 1.615e+02 1.895e+02 2.284e+02 5.384e+02, threshold=3.791e+02, percent-clipped=3.0 2023-04-26 22:42:38,717 INFO [finetune.py:976] (1/7) Epoch 9, batch 2850, loss[loss=0.1398, simple_loss=0.2089, pruned_loss=0.03535, over 4832.00 frames. ], tot_loss[loss=0.1898, simple_loss=0.2549, pruned_loss=0.06239, over 958672.58 frames. ], batch size: 25, lr: 3.79e-03, grad_scale: 16.0 2023-04-26 22:42:49,433 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=48680.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:43:11,160 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.29 vs. limit=5.0 2023-04-26 22:43:44,377 INFO [finetune.py:976] (1/7) Epoch 9, batch 2900, loss[loss=0.239, simple_loss=0.2986, pruned_loss=0.08967, over 4922.00 frames. ], tot_loss[loss=0.1939, simple_loss=0.259, pruned_loss=0.06442, over 956869.53 frames. ], batch size: 38, lr: 3.79e-03, grad_scale: 16.0 2023-04-26 22:43:48,139 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=48728.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:43:52,843 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.096e+02 1.707e+02 2.011e+02 2.462e+02 4.094e+02, threshold=4.022e+02, percent-clipped=3.0 2023-04-26 22:43:57,252 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.5584, 2.0785, 2.5661, 3.1145, 2.3642, 1.9394, 1.8827, 2.3991], device='cuda:1'), covar=tensor([0.4011, 0.3869, 0.1806, 0.2959, 0.3367, 0.3168, 0.4510, 0.2623], device='cuda:1'), in_proj_covar=tensor([0.0281, 0.0250, 0.0220, 0.0318, 0.0213, 0.0228, 0.0233, 0.0187], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-26 22:44:04,267 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=48753.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:44:10,217 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.80 vs. limit=2.0 2023-04-26 22:44:17,828 INFO [finetune.py:976] (1/7) Epoch 9, batch 2950, loss[loss=0.1756, simple_loss=0.2351, pruned_loss=0.05803, over 4760.00 frames. ], tot_loss[loss=0.1965, simple_loss=0.2624, pruned_loss=0.06526, over 955515.65 frames. ], batch size: 23, lr: 3.79e-03, grad_scale: 16.0 2023-04-26 22:44:20,345 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=48776.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:44:22,221 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=48779.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:44:25,462 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 2023-04-26 22:44:51,972 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-04-26 22:45:04,156 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7242, 2.2348, 1.8685, 2.0819, 1.5925, 1.8391, 1.9939, 1.5478], device='cuda:1'), covar=tensor([0.2599, 0.2005, 0.1348, 0.1797, 0.3602, 0.1879, 0.1959, 0.2795], device='cuda:1'), in_proj_covar=tensor([0.0294, 0.0315, 0.0227, 0.0288, 0.0312, 0.0269, 0.0255, 0.0279], device='cuda:1'), out_proj_covar=tensor([1.1904e-04, 1.2703e-04, 9.1230e-05, 1.1543e-04, 1.2796e-04, 1.0860e-04, 1.0459e-04, 1.1191e-04], device='cuda:1') 2023-04-26 22:45:10,549 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=48821.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:45:11,035 INFO [finetune.py:976] (1/7) Epoch 9, batch 3000, loss[loss=0.2347, simple_loss=0.2974, pruned_loss=0.08599, over 4847.00 frames. ], tot_loss[loss=0.1984, simple_loss=0.2643, pruned_loss=0.06625, over 955348.42 frames. ], batch size: 44, lr: 3.79e-03, grad_scale: 16.0 2023-04-26 22:45:11,035 INFO [finetune.py:1001] (1/7) Computing validation loss 2023-04-26 22:45:27,358 INFO [finetune.py:1010] (1/7) Epoch 9, validation: loss=0.1543, simple_loss=0.2267, pruned_loss=0.04097, over 2265189.00 frames. 2023-04-26 22:45:27,359 INFO [finetune.py:1011] (1/7) Maximum memory allocated so far is 6391MB 2023-04-26 22:45:46,033 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.141e+02 1.720e+02 1.968e+02 2.331e+02 3.766e+02, threshold=3.936e+02, percent-clipped=0.0 2023-04-26 22:45:47,982 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=48838.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:46:30,395 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=48868.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 22:46:32,696 INFO [finetune.py:976] (1/7) Epoch 9, batch 3050, loss[loss=0.1786, simple_loss=0.2566, pruned_loss=0.05035, over 4722.00 frames. ], tot_loss[loss=0.1978, simple_loss=0.2642, pruned_loss=0.06571, over 955232.14 frames. ], batch size: 59, lr: 3.79e-03, grad_scale: 16.0 2023-04-26 22:46:32,837 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.3712, 1.5908, 1.6937, 1.8315, 1.6415, 1.7419, 1.7881, 1.7512], device='cuda:1'), covar=tensor([0.5349, 0.7082, 0.6167, 0.5604, 0.7130, 0.9985, 0.7022, 0.6833], device='cuda:1'), in_proj_covar=tensor([0.0324, 0.0385, 0.0318, 0.0327, 0.0343, 0.0406, 0.0365, 0.0327], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-26 22:46:51,111 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=48882.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:46:53,575 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.5280, 2.1871, 2.6455, 3.0068, 2.5011, 2.0046, 1.7765, 2.4839], device='cuda:1'), covar=tensor([0.3804, 0.3324, 0.1599, 0.3032, 0.3029, 0.2761, 0.4371, 0.2342], device='cuda:1'), in_proj_covar=tensor([0.0280, 0.0249, 0.0219, 0.0318, 0.0212, 0.0227, 0.0232, 0.0186], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-26 22:47:12,136 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=48899.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:47:28,836 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=48916.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 22:47:32,444 INFO [finetune.py:976] (1/7) Epoch 9, batch 3100, loss[loss=0.1728, simple_loss=0.2457, pruned_loss=0.04991, over 4726.00 frames. ], tot_loss[loss=0.1953, simple_loss=0.2615, pruned_loss=0.0646, over 955641.31 frames. ], batch size: 23, lr: 3.79e-03, grad_scale: 16.0 2023-04-26 22:47:42,246 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.118e+02 1.703e+02 2.023e+02 2.523e+02 5.306e+02, threshold=4.046e+02, percent-clipped=4.0 2023-04-26 22:48:02,823 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-04-26 22:48:05,673 INFO [finetune.py:976] (1/7) Epoch 9, batch 3150, loss[loss=0.1403, simple_loss=0.2111, pruned_loss=0.03478, over 4710.00 frames. ], tot_loss[loss=0.1929, simple_loss=0.2583, pruned_loss=0.06372, over 955400.74 frames. ], batch size: 59, lr: 3.78e-03, grad_scale: 16.0 2023-04-26 22:48:11,545 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=48980.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:48:21,548 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([3.3607, 3.3391, 2.4782, 3.9266, 3.3513, 3.4415, 1.5681, 3.3746], device='cuda:1'), covar=tensor([0.1965, 0.1371, 0.3249, 0.2450, 0.2978, 0.2015, 0.5856, 0.2638], device='cuda:1'), in_proj_covar=tensor([0.0241, 0.0216, 0.0249, 0.0303, 0.0298, 0.0250, 0.0268, 0.0269], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-26 22:48:38,108 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4459, 1.3350, 1.7267, 1.7264, 1.3306, 1.1122, 1.4829, 1.0053], device='cuda:1'), covar=tensor([0.0662, 0.0770, 0.0474, 0.0635, 0.0922, 0.1186, 0.0810, 0.0819], device='cuda:1'), in_proj_covar=tensor([0.0066, 0.0073, 0.0072, 0.0066, 0.0076, 0.0095, 0.0078, 0.0074], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-26 22:48:38,602 INFO [finetune.py:976] (1/7) Epoch 9, batch 3200, loss[loss=0.2055, simple_loss=0.2628, pruned_loss=0.07411, over 4805.00 frames. ], tot_loss[loss=0.1903, simple_loss=0.2552, pruned_loss=0.06266, over 957123.82 frames. ], batch size: 51, lr: 3.78e-03, grad_scale: 16.0 2023-04-26 22:48:42,797 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=49028.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:48:47,938 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.162e+02 1.546e+02 1.866e+02 2.430e+02 3.518e+02, threshold=3.733e+02, percent-clipped=0.0 2023-04-26 22:48:50,414 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5245, 1.3167, 1.8130, 1.7361, 1.3780, 1.2525, 1.4769, 1.0260], device='cuda:1'), covar=tensor([0.0681, 0.1015, 0.0537, 0.0764, 0.0845, 0.1436, 0.0752, 0.0843], device='cuda:1'), in_proj_covar=tensor([0.0066, 0.0073, 0.0072, 0.0066, 0.0076, 0.0095, 0.0078, 0.0074], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-26 22:48:55,697 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0393, 2.5801, 1.1446, 1.3208, 1.8753, 1.2434, 3.2639, 1.6750], device='cuda:1'), covar=tensor([0.0696, 0.0711, 0.0818, 0.1249, 0.0513, 0.1018, 0.0221, 0.0659], device='cuda:1'), in_proj_covar=tensor([0.0051, 0.0067, 0.0049, 0.0047, 0.0051, 0.0052, 0.0078, 0.0051], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0007], device='cuda:1') 2023-04-26 22:48:59,905 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=49053.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:49:12,011 INFO [finetune.py:976] (1/7) Epoch 9, batch 3250, loss[loss=0.2033, simple_loss=0.2824, pruned_loss=0.06212, over 4799.00 frames. ], tot_loss[loss=0.1902, simple_loss=0.2553, pruned_loss=0.06262, over 955401.72 frames. ], batch size: 51, lr: 3.78e-03, grad_scale: 16.0 2023-04-26 22:49:16,900 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=49079.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:49:32,073 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=49101.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:49:40,136 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-04-26 22:49:45,283 INFO [finetune.py:976] (1/7) Epoch 9, batch 3300, loss[loss=0.1661, simple_loss=0.2407, pruned_loss=0.04572, over 4754.00 frames. ], tot_loss[loss=0.1947, simple_loss=0.2603, pruned_loss=0.06453, over 954777.02 frames. ], batch size: 27, lr: 3.78e-03, grad_scale: 16.0 2023-04-26 22:49:48,898 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=49127.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:49:51,982 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([3.3536, 2.7669, 2.6108, 2.7897, 2.5533, 2.7521, 2.6857, 2.6449], device='cuda:1'), covar=tensor([0.4405, 0.6911, 0.6291, 0.5215, 0.6629, 0.7814, 0.7414, 0.6629], device='cuda:1'), in_proj_covar=tensor([0.0322, 0.0381, 0.0315, 0.0325, 0.0340, 0.0402, 0.0362, 0.0324], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-26 22:49:54,142 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.094e+02 1.776e+02 1.982e+02 2.505e+02 4.368e+02, threshold=3.963e+02, percent-clipped=2.0 2023-04-26 22:50:18,718 INFO [finetune.py:976] (1/7) Epoch 9, batch 3350, loss[loss=0.1921, simple_loss=0.2564, pruned_loss=0.06387, over 4884.00 frames. ], tot_loss[loss=0.1966, simple_loss=0.2626, pruned_loss=0.06529, over 954331.60 frames. ], batch size: 32, lr: 3.78e-03, grad_scale: 16.0 2023-04-26 22:50:21,870 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=49177.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:50:46,360 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=49194.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:50:58,847 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6962, 1.3159, 0.6089, 1.3441, 1.5459, 1.5590, 1.4304, 1.4202], device='cuda:1'), covar=tensor([0.0518, 0.0428, 0.0405, 0.0593, 0.0279, 0.0533, 0.0523, 0.0592], device='cuda:1'), in_proj_covar=tensor([0.0030, 0.0026, 0.0024, 0.0030, 0.0021, 0.0030, 0.0029, 0.0030], device='cuda:1'), out_proj_covar=tensor([0.0049, 0.0045, 0.0038, 0.0050, 0.0038, 0.0048, 0.0048, 0.0050], device='cuda:1') 2023-04-26 22:51:20,199 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.67 vs. limit=2.0 2023-04-26 22:51:20,614 INFO [finetune.py:976] (1/7) Epoch 9, batch 3400, loss[loss=0.211, simple_loss=0.2884, pruned_loss=0.06682, over 4915.00 frames. ], tot_loss[loss=0.1979, simple_loss=0.2644, pruned_loss=0.06573, over 956083.06 frames. ], batch size: 36, lr: 3.78e-03, grad_scale: 16.0 2023-04-26 22:51:38,491 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.071e+02 1.560e+02 1.880e+02 2.304e+02 5.617e+02, threshold=3.759e+02, percent-clipped=2.0 2023-04-26 22:51:39,195 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=49236.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:52:08,716 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.79 vs. limit=2.0 2023-04-26 22:52:13,324 INFO [finetune.py:976] (1/7) Epoch 9, batch 3450, loss[loss=0.1894, simple_loss=0.271, pruned_loss=0.05389, over 4815.00 frames. ], tot_loss[loss=0.197, simple_loss=0.2642, pruned_loss=0.0649, over 956640.53 frames. ], batch size: 33, lr: 3.78e-03, grad_scale: 16.0 2023-04-26 22:52:30,780 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=49297.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:52:40,886 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.57 vs. limit=2.0 2023-04-26 22:53:02,579 INFO [finetune.py:976] (1/7) Epoch 9, batch 3500, loss[loss=0.1945, simple_loss=0.2549, pruned_loss=0.06708, over 4873.00 frames. ], tot_loss[loss=0.1954, simple_loss=0.2617, pruned_loss=0.06457, over 954094.56 frames. ], batch size: 34, lr: 3.78e-03, grad_scale: 16.0 2023-04-26 22:53:15,830 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.247e+02 1.587e+02 1.893e+02 2.359e+02 3.521e+02, threshold=3.785e+02, percent-clipped=0.0 2023-04-26 22:54:01,001 INFO [finetune.py:976] (1/7) Epoch 9, batch 3550, loss[loss=0.1995, simple_loss=0.2698, pruned_loss=0.06457, over 4873.00 frames. ], tot_loss[loss=0.1942, simple_loss=0.2594, pruned_loss=0.06444, over 956194.29 frames. ], batch size: 34, lr: 3.78e-03, grad_scale: 16.0 2023-04-26 22:55:07,552 INFO [finetune.py:976] (1/7) Epoch 9, batch 3600, loss[loss=0.2096, simple_loss=0.2656, pruned_loss=0.07684, over 4824.00 frames. ], tot_loss[loss=0.1919, simple_loss=0.2566, pruned_loss=0.06362, over 954520.67 frames. ], batch size: 40, lr: 3.78e-03, grad_scale: 16.0 2023-04-26 22:55:07,655 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=49422.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:55:24,157 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=49432.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 22:55:25,887 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.037e+02 1.649e+02 1.885e+02 2.480e+02 5.265e+02, threshold=3.771e+02, percent-clipped=3.0 2023-04-26 22:56:18,973 INFO [finetune.py:976] (1/7) Epoch 9, batch 3650, loss[loss=0.1877, simple_loss=0.2599, pruned_loss=0.05779, over 4908.00 frames. ], tot_loss[loss=0.1937, simple_loss=0.2587, pruned_loss=0.06439, over 954346.14 frames. ], batch size: 36, lr: 3.78e-03, grad_scale: 16.0 2023-04-26 22:56:22,167 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=49477.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:56:30,050 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=49483.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:56:36,685 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=49493.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 22:56:37,252 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=49494.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:56:56,124 INFO [finetune.py:976] (1/7) Epoch 9, batch 3700, loss[loss=0.2021, simple_loss=0.2791, pruned_loss=0.06258, over 4826.00 frames. ], tot_loss[loss=0.1956, simple_loss=0.2615, pruned_loss=0.06482, over 955514.90 frames. ], batch size: 47, lr: 3.78e-03, grad_scale: 16.0 2023-04-26 22:56:58,021 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=49525.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:57:09,404 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.074e+02 1.745e+02 1.976e+02 2.490e+02 4.358e+02, threshold=3.952e+02, percent-clipped=2.0 2023-04-26 22:57:13,689 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=49542.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:57:56,799 INFO [finetune.py:976] (1/7) Epoch 9, batch 3750, loss[loss=0.1662, simple_loss=0.2431, pruned_loss=0.04459, over 4897.00 frames. ], tot_loss[loss=0.1966, simple_loss=0.2626, pruned_loss=0.06531, over 953882.64 frames. ], batch size: 43, lr: 3.78e-03, grad_scale: 16.0 2023-04-26 22:58:16,674 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0132, 2.3707, 1.1094, 1.4181, 1.8874, 1.2505, 3.2862, 1.7119], device='cuda:1'), covar=tensor([0.0653, 0.0695, 0.0800, 0.1208, 0.0518, 0.0977, 0.0193, 0.0637], device='cuda:1'), in_proj_covar=tensor([0.0051, 0.0066, 0.0049, 0.0047, 0.0051, 0.0052, 0.0078, 0.0051], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0007, 0.0008, 0.0008, 0.0010, 0.0007], device='cuda:1') 2023-04-26 22:58:23,846 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=49592.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:58:44,872 INFO [finetune.py:976] (1/7) Epoch 9, batch 3800, loss[loss=0.1799, simple_loss=0.2461, pruned_loss=0.05685, over 4922.00 frames. ], tot_loss[loss=0.1983, simple_loss=0.2648, pruned_loss=0.06591, over 953962.91 frames. ], batch size: 41, lr: 3.78e-03, grad_scale: 16.0 2023-04-26 22:58:52,836 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.166e+02 1.711e+02 2.053e+02 2.525e+02 6.309e+02, threshold=4.105e+02, percent-clipped=4.0 2023-04-26 22:59:17,906 INFO [finetune.py:976] (1/7) Epoch 9, batch 3850, loss[loss=0.2124, simple_loss=0.2766, pruned_loss=0.07413, over 4794.00 frames. ], tot_loss[loss=0.1965, simple_loss=0.2633, pruned_loss=0.06484, over 954937.69 frames. ], batch size: 29, lr: 3.78e-03, grad_scale: 16.0 2023-04-26 22:59:21,626 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.2525, 2.9845, 1.2045, 1.6183, 2.1867, 1.4617, 3.6491, 1.8372], device='cuda:1'), covar=tensor([0.0614, 0.0919, 0.0849, 0.1008, 0.0452, 0.0830, 0.0184, 0.0559], device='cuda:1'), in_proj_covar=tensor([0.0051, 0.0066, 0.0049, 0.0047, 0.0051, 0.0052, 0.0078, 0.0051], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0007], device='cuda:1') 2023-04-26 22:59:22,883 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7718, 1.2480, 1.4620, 1.4006, 1.9340, 1.5677, 1.2639, 1.4050], device='cuda:1'), covar=tensor([0.1680, 0.1452, 0.1628, 0.1329, 0.0834, 0.1394, 0.1960, 0.2007], device='cuda:1'), in_proj_covar=tensor([0.0308, 0.0325, 0.0353, 0.0297, 0.0335, 0.0322, 0.0308, 0.0356], device='cuda:1'), out_proj_covar=tensor([6.5139e-05, 6.8904e-05, 7.6235e-05, 6.1447e-05, 7.0237e-05, 6.9112e-05, 6.6192e-05, 7.6469e-05], device='cuda:1') 2023-04-26 22:59:49,730 INFO [finetune.py:976] (1/7) Epoch 9, batch 3900, loss[loss=0.2477, simple_loss=0.2942, pruned_loss=0.1006, over 4325.00 frames. ], tot_loss[loss=0.1947, simple_loss=0.2604, pruned_loss=0.06447, over 954902.94 frames. ], batch size: 65, lr: 3.78e-03, grad_scale: 32.0 2023-04-26 22:59:58,108 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.293e+02 1.654e+02 1.955e+02 2.414e+02 4.265e+02, threshold=3.910e+02, percent-clipped=1.0 2023-04-26 23:00:00,093 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8857, 1.4878, 1.9813, 2.2814, 1.9636, 1.8315, 1.9618, 1.9184], device='cuda:1'), covar=tensor([0.5527, 0.8054, 0.8537, 0.7502, 0.7527, 0.9538, 0.9774, 0.9611], device='cuda:1'), in_proj_covar=tensor([0.0409, 0.0420, 0.0505, 0.0524, 0.0439, 0.0459, 0.0470, 0.0468], device='cuda:1'), out_proj_covar=tensor([9.9373e-05, 1.0419e-04, 1.1392e-04, 1.2438e-04, 1.0661e-04, 1.1102e-04, 1.1292e-04, 1.1312e-04], device='cuda:1') 2023-04-26 23:00:16,873 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.3320, 1.5303, 1.5815, 1.8023, 1.6597, 1.7683, 1.7715, 1.6816], device='cuda:1'), covar=tensor([0.5415, 0.6585, 0.5703, 0.5064, 0.6464, 0.8927, 0.6490, 0.6164], device='cuda:1'), in_proj_covar=tensor([0.0319, 0.0378, 0.0312, 0.0323, 0.0337, 0.0398, 0.0358, 0.0320], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-26 23:00:18,068 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=49766.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:00:21,569 INFO [finetune.py:976] (1/7) Epoch 9, batch 3950, loss[loss=0.251, simple_loss=0.2882, pruned_loss=0.107, over 4893.00 frames. ], tot_loss[loss=0.1917, simple_loss=0.2568, pruned_loss=0.06336, over 954852.08 frames. ], batch size: 35, lr: 3.78e-03, grad_scale: 32.0 2023-04-26 23:00:26,172 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8136, 1.8331, 1.0564, 1.5658, 1.9390, 1.6905, 1.6172, 1.6325], device='cuda:1'), covar=tensor([0.0542, 0.0368, 0.0397, 0.0577, 0.0284, 0.0538, 0.0536, 0.0596], device='cuda:1'), in_proj_covar=tensor([0.0030, 0.0026, 0.0023, 0.0030, 0.0021, 0.0029, 0.0029, 0.0030], device='cuda:1'), out_proj_covar=tensor([0.0049, 0.0045, 0.0038, 0.0050, 0.0038, 0.0048, 0.0048, 0.0050], device='cuda:1') 2023-04-26 23:00:27,218 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=49778.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:00:33,309 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=49788.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 23:00:55,289 INFO [finetune.py:976] (1/7) Epoch 9, batch 4000, loss[loss=0.2218, simple_loss=0.2817, pruned_loss=0.08097, over 4935.00 frames. ], tot_loss[loss=0.1922, simple_loss=0.2565, pruned_loss=0.06399, over 953787.55 frames. ], batch size: 38, lr: 3.78e-03, grad_scale: 32.0 2023-04-26 23:00:55,361 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([4.6331, 4.4674, 3.1480, 5.3374, 4.6684, 4.5592, 2.1159, 4.5546], device='cuda:1'), covar=tensor([0.1399, 0.0933, 0.3139, 0.0783, 0.3949, 0.1555, 0.5552, 0.1957], device='cuda:1'), in_proj_covar=tensor([0.0242, 0.0216, 0.0249, 0.0303, 0.0299, 0.0250, 0.0269, 0.0270], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-26 23:00:58,934 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=49827.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:01:03,405 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9830, 1.4783, 5.3888, 5.0911, 4.7312, 5.0354, 4.6815, 4.8132], device='cuda:1'), covar=tensor([0.6566, 0.6229, 0.0899, 0.1644, 0.1017, 0.2154, 0.1198, 0.1556], device='cuda:1'), in_proj_covar=tensor([0.0309, 0.0307, 0.0404, 0.0410, 0.0348, 0.0402, 0.0313, 0.0368], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-26 23:01:05,129 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.158e+02 1.602e+02 1.886e+02 2.314e+02 3.383e+02, threshold=3.771e+02, percent-clipped=0.0 2023-04-26 23:01:44,300 INFO [finetune.py:976] (1/7) Epoch 9, batch 4050, loss[loss=0.1712, simple_loss=0.2572, pruned_loss=0.04259, over 4847.00 frames. ], tot_loss[loss=0.1963, simple_loss=0.2608, pruned_loss=0.06594, over 953918.75 frames. ], batch size: 49, lr: 3.78e-03, grad_scale: 32.0 2023-04-26 23:01:52,859 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-04-26 23:01:55,048 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([0.8804, 1.9374, 1.7211, 1.4667, 2.0134, 1.6674, 2.5927, 1.5746], device='cuda:1'), covar=tensor([0.3900, 0.1773, 0.5408, 0.3309, 0.1880, 0.2468, 0.1466, 0.4709], device='cuda:1'), in_proj_covar=tensor([0.0348, 0.0351, 0.0435, 0.0363, 0.0391, 0.0386, 0.0383, 0.0424], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 23:02:03,180 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([0.9124, 1.8761, 1.5715, 1.3988, 1.9431, 1.6236, 2.4543, 1.4197], device='cuda:1'), covar=tensor([0.3865, 0.1769, 0.5323, 0.3543, 0.1893, 0.2613, 0.1520, 0.4815], device='cuda:1'), in_proj_covar=tensor([0.0348, 0.0351, 0.0436, 0.0363, 0.0392, 0.0387, 0.0383, 0.0424], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 23:02:14,665 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=49892.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:02:49,741 INFO [finetune.py:976] (1/7) Epoch 9, batch 4100, loss[loss=0.1532, simple_loss=0.2328, pruned_loss=0.03682, over 4854.00 frames. ], tot_loss[loss=0.1961, simple_loss=0.2621, pruned_loss=0.06509, over 956408.56 frames. ], batch size: 31, lr: 3.78e-03, grad_scale: 32.0 2023-04-26 23:03:10,347 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.345e+02 1.694e+02 2.037e+02 2.558e+02 4.844e+02, threshold=4.074e+02, percent-clipped=3.0 2023-04-26 23:03:19,228 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=49940.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:03:33,528 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.3988, 1.6343, 1.6849, 1.7863, 1.6535, 1.7793, 1.9013, 1.7487], device='cuda:1'), covar=tensor([0.4983, 0.7186, 0.6062, 0.5549, 0.6847, 1.0002, 0.6999, 0.6649], device='cuda:1'), in_proj_covar=tensor([0.0321, 0.0380, 0.0313, 0.0324, 0.0339, 0.0400, 0.0360, 0.0322], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-26 23:03:53,584 INFO [finetune.py:976] (1/7) Epoch 9, batch 4150, loss[loss=0.2163, simple_loss=0.2809, pruned_loss=0.07584, over 4832.00 frames. ], tot_loss[loss=0.1968, simple_loss=0.2629, pruned_loss=0.06534, over 954453.18 frames. ], batch size: 47, lr: 3.78e-03, grad_scale: 32.0 2023-04-26 23:04:03,520 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-04-26 23:04:33,308 INFO [finetune.py:976] (1/7) Epoch 9, batch 4200, loss[loss=0.2245, simple_loss=0.2866, pruned_loss=0.08116, over 4815.00 frames. ], tot_loss[loss=0.1952, simple_loss=0.2618, pruned_loss=0.06433, over 952804.02 frames. ], batch size: 38, lr: 3.78e-03, grad_scale: 32.0 2023-04-26 23:04:41,640 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.155e+02 1.800e+02 2.150e+02 2.461e+02 7.133e+02, threshold=4.301e+02, percent-clipped=1.0 2023-04-26 23:05:05,674 INFO [finetune.py:976] (1/7) Epoch 9, batch 4250, loss[loss=0.1964, simple_loss=0.2472, pruned_loss=0.07278, over 4176.00 frames. ], tot_loss[loss=0.1946, simple_loss=0.2606, pruned_loss=0.06427, over 953975.41 frames. ], batch size: 18, lr: 3.78e-03, grad_scale: 32.0 2023-04-26 23:05:09,945 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=50078.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:05:16,048 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=50088.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 23:05:20,690 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6412, 1.7660, 0.8785, 1.3458, 1.9550, 1.5299, 1.4452, 1.4896], device='cuda:1'), covar=tensor([0.0533, 0.0376, 0.0386, 0.0584, 0.0290, 0.0568, 0.0518, 0.0597], device='cuda:1'), in_proj_covar=tensor([0.0030, 0.0026, 0.0024, 0.0030, 0.0021, 0.0029, 0.0029, 0.0030], device='cuda:1'), out_proj_covar=tensor([0.0049, 0.0045, 0.0038, 0.0050, 0.0038, 0.0048, 0.0048, 0.0050], device='cuda:1') 2023-04-26 23:05:37,584 INFO [finetune.py:976] (1/7) Epoch 9, batch 4300, loss[loss=0.2034, simple_loss=0.2683, pruned_loss=0.06926, over 4828.00 frames. ], tot_loss[loss=0.1927, simple_loss=0.2581, pruned_loss=0.06364, over 955267.25 frames. ], batch size: 30, lr: 3.78e-03, grad_scale: 32.0 2023-04-26 23:05:37,666 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=50122.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:05:40,094 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=50126.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:05:41,356 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=50128.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:05:46,442 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.165e+02 1.729e+02 1.986e+02 2.496e+02 5.058e+02, threshold=3.971e+02, percent-clipped=3.0 2023-04-26 23:05:47,120 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=50136.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 23:06:10,816 INFO [finetune.py:976] (1/7) Epoch 9, batch 4350, loss[loss=0.1831, simple_loss=0.2431, pruned_loss=0.0615, over 4755.00 frames. ], tot_loss[loss=0.1894, simple_loss=0.2544, pruned_loss=0.06224, over 957213.35 frames. ], batch size: 27, lr: 3.78e-03, grad_scale: 32.0 2023-04-26 23:06:22,151 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=50189.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:06:27,952 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=50197.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 23:06:31,564 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.56 vs. limit=5.0 2023-04-26 23:06:44,562 INFO [finetune.py:976] (1/7) Epoch 9, batch 4400, loss[loss=0.2152, simple_loss=0.292, pruned_loss=0.06919, over 4749.00 frames. ], tot_loss[loss=0.1916, simple_loss=0.2561, pruned_loss=0.06354, over 954687.34 frames. ], batch size: 54, lr: 3.78e-03, grad_scale: 32.0 2023-04-26 23:06:52,520 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.016e+02 1.620e+02 1.881e+02 2.281e+02 3.883e+02, threshold=3.762e+02, percent-clipped=0.0 2023-04-26 23:06:56,028 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.3095, 1.5459, 1.6967, 1.8243, 1.7219, 1.8094, 1.8685, 1.7905], device='cuda:1'), covar=tensor([0.5404, 0.7392, 0.6337, 0.6347, 0.7011, 1.0156, 0.7123, 0.6277], device='cuda:1'), in_proj_covar=tensor([0.0322, 0.0380, 0.0313, 0.0324, 0.0338, 0.0400, 0.0359, 0.0322], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-26 23:07:13,977 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=50258.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 23:07:27,674 INFO [finetune.py:976] (1/7) Epoch 9, batch 4450, loss[loss=0.1985, simple_loss=0.2706, pruned_loss=0.06317, over 4810.00 frames. ], tot_loss[loss=0.1947, simple_loss=0.2604, pruned_loss=0.06452, over 954479.23 frames. ], batch size: 38, lr: 3.77e-03, grad_scale: 32.0 2023-04-26 23:07:59,850 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8706, 1.7019, 2.1652, 2.3750, 1.9828, 1.8178, 1.9709, 2.0210], device='cuda:1'), covar=tensor([0.6970, 0.9228, 1.0187, 0.9194, 0.8427, 1.1738, 1.2006, 1.0018], device='cuda:1'), in_proj_covar=tensor([0.0409, 0.0419, 0.0505, 0.0523, 0.0438, 0.0457, 0.0469, 0.0466], device='cuda:1'), out_proj_covar=tensor([9.9641e-05, 1.0396e-04, 1.1393e-04, 1.2421e-04, 1.0642e-04, 1.1055e-04, 1.1263e-04, 1.1267e-04], device='cuda:1') 2023-04-26 23:08:07,930 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([3.8943, 3.7571, 2.6240, 4.4842, 3.9049, 3.8585, 1.8971, 3.8947], device='cuda:1'), covar=tensor([0.1686, 0.1324, 0.3293, 0.1455, 0.3125, 0.1846, 0.5612, 0.2374], device='cuda:1'), in_proj_covar=tensor([0.0241, 0.0215, 0.0248, 0.0301, 0.0299, 0.0248, 0.0267, 0.0269], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-26 23:08:19,315 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7210, 1.6204, 1.7657, 2.0763, 2.2058, 1.6760, 1.3041, 1.9012], device='cuda:1'), covar=tensor([0.0850, 0.1091, 0.0703, 0.0590, 0.0508, 0.0775, 0.0828, 0.0502], device='cuda:1'), in_proj_covar=tensor([0.0196, 0.0204, 0.0181, 0.0175, 0.0178, 0.0188, 0.0159, 0.0184], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 23:08:33,131 INFO [finetune.py:976] (1/7) Epoch 9, batch 4500, loss[loss=0.1476, simple_loss=0.2178, pruned_loss=0.03869, over 4756.00 frames. ], tot_loss[loss=0.1947, simple_loss=0.2608, pruned_loss=0.0643, over 951908.23 frames. ], batch size: 27, lr: 3.77e-03, grad_scale: 32.0 2023-04-26 23:08:40,148 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8036, 1.5751, 4.4100, 3.8356, 3.9791, 4.1658, 3.9370, 3.8217], device='cuda:1'), covar=tensor([0.7842, 0.7678, 0.1398, 0.2914, 0.1773, 0.2768, 0.2201, 0.2898], device='cuda:1'), in_proj_covar=tensor([0.0309, 0.0309, 0.0404, 0.0411, 0.0350, 0.0405, 0.0315, 0.0370], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-26 23:08:40,306 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.51 vs. limit=2.0 2023-04-26 23:08:46,790 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.096e+02 1.697e+02 2.110e+02 2.509e+02 5.255e+02, threshold=4.219e+02, percent-clipped=3.0 2023-04-26 23:09:12,328 INFO [finetune.py:976] (1/7) Epoch 9, batch 4550, loss[loss=0.2332, simple_loss=0.2852, pruned_loss=0.09059, over 4838.00 frames. ], tot_loss[loss=0.1967, simple_loss=0.2629, pruned_loss=0.06527, over 952435.37 frames. ], batch size: 47, lr: 3.77e-03, grad_scale: 32.0 2023-04-26 23:09:22,179 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7249, 2.0919, 1.1207, 1.4923, 2.2779, 1.6588, 1.6389, 1.5997], device='cuda:1'), covar=tensor([0.0533, 0.0372, 0.0333, 0.0596, 0.0242, 0.0570, 0.0544, 0.0617], device='cuda:1'), in_proj_covar=tensor([0.0030, 0.0026, 0.0023, 0.0030, 0.0021, 0.0029, 0.0029, 0.0030], device='cuda:1'), out_proj_covar=tensor([0.0049, 0.0044, 0.0038, 0.0050, 0.0038, 0.0048, 0.0048, 0.0050], device='cuda:1') 2023-04-26 23:09:44,390 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.1609, 1.8954, 2.0253, 2.5393, 2.4852, 2.0205, 1.5540, 2.1574], device='cuda:1'), covar=tensor([0.0904, 0.1092, 0.0733, 0.0574, 0.0573, 0.0904, 0.0906, 0.0574], device='cuda:1'), in_proj_covar=tensor([0.0196, 0.0205, 0.0182, 0.0176, 0.0178, 0.0189, 0.0160, 0.0184], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 23:10:20,888 INFO [finetune.py:976] (1/7) Epoch 9, batch 4600, loss[loss=0.1535, simple_loss=0.223, pruned_loss=0.04197, over 4871.00 frames. ], tot_loss[loss=0.1959, simple_loss=0.262, pruned_loss=0.0649, over 955149.25 frames. ], batch size: 31, lr: 3.77e-03, grad_scale: 32.0 2023-04-26 23:10:21,001 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=50422.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:10:30,604 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0970, 2.6110, 1.0336, 1.4196, 2.0611, 1.2445, 3.5845, 1.7507], device='cuda:1'), covar=tensor([0.0631, 0.0711, 0.0861, 0.1272, 0.0500, 0.0981, 0.0222, 0.0661], device='cuda:1'), in_proj_covar=tensor([0.0051, 0.0066, 0.0049, 0.0047, 0.0051, 0.0052, 0.0078, 0.0050], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0007, 0.0008, 0.0008, 0.0010, 0.0007], device='cuda:1') 2023-04-26 23:10:39,535 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.116e+02 1.674e+02 1.954e+02 2.272e+02 3.604e+02, threshold=3.908e+02, percent-clipped=0.0 2023-04-26 23:11:14,981 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=50470.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:11:16,115 INFO [finetune.py:976] (1/7) Epoch 9, batch 4650, loss[loss=0.1934, simple_loss=0.2574, pruned_loss=0.06471, over 4820.00 frames. ], tot_loss[loss=0.1931, simple_loss=0.2586, pruned_loss=0.06378, over 955220.71 frames. ], batch size: 40, lr: 3.77e-03, grad_scale: 32.0 2023-04-26 23:11:23,418 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=50484.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:11:49,307 INFO [finetune.py:976] (1/7) Epoch 9, batch 4700, loss[loss=0.1622, simple_loss=0.234, pruned_loss=0.0452, over 4754.00 frames. ], tot_loss[loss=0.1893, simple_loss=0.2548, pruned_loss=0.06187, over 956526.10 frames. ], batch size: 28, lr: 3.77e-03, grad_scale: 32.0 2023-04-26 23:11:57,184 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.176e+02 1.640e+02 1.914e+02 2.343e+02 4.102e+02, threshold=3.828e+02, percent-clipped=1.0 2023-04-26 23:12:04,102 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=50546.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:12:08,318 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=50553.0, num_to_drop=1, layers_to_drop={3} 2023-04-26 23:12:22,117 INFO [finetune.py:976] (1/7) Epoch 9, batch 4750, loss[loss=0.1918, simple_loss=0.2571, pruned_loss=0.06319, over 4800.00 frames. ], tot_loss[loss=0.1884, simple_loss=0.2536, pruned_loss=0.06157, over 954412.48 frames. ], batch size: 29, lr: 3.77e-03, grad_scale: 32.0 2023-04-26 23:12:44,843 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8622, 2.4491, 0.7867, 1.1989, 1.6119, 1.0606, 3.2702, 1.5192], device='cuda:1'), covar=tensor([0.0912, 0.0879, 0.1075, 0.1785, 0.0813, 0.1506, 0.0387, 0.0980], device='cuda:1'), in_proj_covar=tensor([0.0052, 0.0066, 0.0049, 0.0047, 0.0051, 0.0052, 0.0078, 0.0051], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0007], device='cuda:1') 2023-04-26 23:12:53,740 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0602, 1.9090, 2.5307, 2.6717, 1.9001, 1.6184, 1.9938, 1.1642], device='cuda:1'), covar=tensor([0.0723, 0.0982, 0.0492, 0.0566, 0.0921, 0.1344, 0.0861, 0.1233], device='cuda:1'), in_proj_covar=tensor([0.0066, 0.0073, 0.0071, 0.0066, 0.0075, 0.0096, 0.0078, 0.0073], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-26 23:12:54,843 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=50607.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:13:16,244 INFO [finetune.py:976] (1/7) Epoch 9, batch 4800, loss[loss=0.1564, simple_loss=0.2335, pruned_loss=0.03967, over 4888.00 frames. ], tot_loss[loss=0.1898, simple_loss=0.2554, pruned_loss=0.06213, over 955085.22 frames. ], batch size: 35, lr: 3.77e-03, grad_scale: 32.0 2023-04-26 23:13:26,094 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.63 vs. limit=5.0 2023-04-26 23:13:30,153 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.165e+02 1.707e+02 2.037e+02 2.400e+02 5.618e+02, threshold=4.074e+02, percent-clipped=3.0 2023-04-26 23:13:35,096 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0464, 3.8261, 1.3525, 2.2190, 2.3817, 2.8440, 2.1916, 1.4494], device='cuda:1'), covar=tensor([0.1191, 0.0897, 0.1847, 0.1219, 0.0978, 0.0838, 0.1554, 0.1629], device='cuda:1'), in_proj_covar=tensor([0.0117, 0.0250, 0.0140, 0.0122, 0.0135, 0.0153, 0.0118, 0.0121], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-26 23:13:46,535 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4878, 1.3397, 4.3195, 4.0055, 3.7943, 4.1089, 3.9959, 3.8294], device='cuda:1'), covar=tensor([0.7513, 0.6066, 0.1080, 0.1912, 0.1179, 0.2035, 0.1409, 0.1718], device='cuda:1'), in_proj_covar=tensor([0.0309, 0.0307, 0.0404, 0.0411, 0.0349, 0.0404, 0.0314, 0.0371], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-26 23:13:55,578 INFO [finetune.py:976] (1/7) Epoch 9, batch 4850, loss[loss=0.2106, simple_loss=0.2877, pruned_loss=0.06671, over 4798.00 frames. ], tot_loss[loss=0.1921, simple_loss=0.2585, pruned_loss=0.06289, over 954594.17 frames. ], batch size: 45, lr: 3.77e-03, grad_scale: 32.0 2023-04-26 23:14:02,501 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=5.11 vs. limit=5.0 2023-04-26 23:14:28,053 INFO [finetune.py:976] (1/7) Epoch 9, batch 4900, loss[loss=0.2739, simple_loss=0.316, pruned_loss=0.1159, over 4891.00 frames. ], tot_loss[loss=0.1936, simple_loss=0.2599, pruned_loss=0.06365, over 953575.83 frames. ], batch size: 35, lr: 3.77e-03, grad_scale: 32.0 2023-04-26 23:14:36,898 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.087e+02 1.673e+02 1.942e+02 2.428e+02 3.700e+02, threshold=3.884e+02, percent-clipped=0.0 2023-04-26 23:15:14,094 INFO [finetune.py:976] (1/7) Epoch 9, batch 4950, loss[loss=0.2487, simple_loss=0.3141, pruned_loss=0.09162, over 4816.00 frames. ], tot_loss[loss=0.197, simple_loss=0.2634, pruned_loss=0.06531, over 955945.51 frames. ], batch size: 38, lr: 3.77e-03, grad_scale: 32.0 2023-04-26 23:15:32,756 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.36 vs. limit=5.0 2023-04-26 23:15:33,260 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=50782.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:15:34,462 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=50784.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:16:02,474 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=50808.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 23:16:11,333 INFO [finetune.py:976] (1/7) Epoch 9, batch 5000, loss[loss=0.2166, simple_loss=0.2639, pruned_loss=0.08465, over 4738.00 frames. ], tot_loss[loss=0.1958, simple_loss=0.2617, pruned_loss=0.06497, over 955289.92 frames. ], batch size: 23, lr: 3.77e-03, grad_scale: 32.0 2023-04-26 23:16:19,878 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=50832.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:16:21,621 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.072e+02 1.711e+02 2.099e+02 2.479e+02 5.783e+02, threshold=4.198e+02, percent-clipped=3.0 2023-04-26 23:16:26,691 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=50843.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:16:31,726 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 2023-04-26 23:16:32,730 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=50853.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 23:16:36,948 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6163, 1.4939, 4.3792, 4.1029, 3.8361, 4.0825, 4.0324, 3.8645], device='cuda:1'), covar=tensor([0.6899, 0.5666, 0.1069, 0.1700, 0.0988, 0.1711, 0.1548, 0.1496], device='cuda:1'), in_proj_covar=tensor([0.0307, 0.0306, 0.0402, 0.0409, 0.0347, 0.0403, 0.0312, 0.0369], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-26 23:16:42,419 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=50869.0, num_to_drop=1, layers_to_drop={3} 2023-04-26 23:16:44,512 INFO [finetune.py:976] (1/7) Epoch 9, batch 5050, loss[loss=0.1661, simple_loss=0.2343, pruned_loss=0.04897, over 4739.00 frames. ], tot_loss[loss=0.1926, simple_loss=0.2581, pruned_loss=0.0636, over 954978.81 frames. ], batch size: 54, lr: 3.77e-03, grad_scale: 32.0 2023-04-26 23:16:53,901 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.1518, 2.5601, 1.0283, 1.4599, 2.0631, 1.2996, 3.6024, 1.8678], device='cuda:1'), covar=tensor([0.0645, 0.0693, 0.0777, 0.1313, 0.0527, 0.1014, 0.0237, 0.0643], device='cuda:1'), in_proj_covar=tensor([0.0051, 0.0066, 0.0049, 0.0047, 0.0051, 0.0052, 0.0078, 0.0050], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0007, 0.0008, 0.0008, 0.0010, 0.0007], device='cuda:1') 2023-04-26 23:17:04,640 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=50901.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 23:17:05,266 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=50902.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:17:13,165 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4803, 2.0246, 1.9110, 2.2866, 2.2501, 2.1956, 1.6886, 4.6716], device='cuda:1'), covar=tensor([0.0579, 0.0719, 0.0748, 0.1088, 0.0580, 0.0549, 0.0733, 0.0094], device='cuda:1'), in_proj_covar=tensor([0.0038, 0.0039, 0.0040, 0.0044, 0.0040, 0.0039, 0.0039, 0.0058], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0014, 0.0013, 0.0012, 0.0013, 0.0016], device='cuda:1') 2023-04-26 23:17:17,313 INFO [finetune.py:976] (1/7) Epoch 9, batch 5100, loss[loss=0.1704, simple_loss=0.2252, pruned_loss=0.05773, over 4712.00 frames. ], tot_loss[loss=0.1899, simple_loss=0.2548, pruned_loss=0.0625, over 956748.73 frames. ], batch size: 23, lr: 3.77e-03, grad_scale: 32.0 2023-04-26 23:17:26,162 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.072e+02 1.643e+02 1.891e+02 2.439e+02 4.473e+02, threshold=3.781e+02, percent-clipped=2.0 2023-04-26 23:17:34,470 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=50946.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:17:50,268 INFO [finetune.py:976] (1/7) Epoch 9, batch 5150, loss[loss=0.1549, simple_loss=0.2281, pruned_loss=0.04089, over 4797.00 frames. ], tot_loss[loss=0.1906, simple_loss=0.2552, pruned_loss=0.06298, over 953994.69 frames. ], batch size: 29, lr: 3.77e-03, grad_scale: 32.0 2023-04-26 23:18:05,889 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2023-04-26 23:18:06,951 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.64 vs. limit=2.0 2023-04-26 23:18:25,491 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=51007.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:18:40,597 INFO [finetune.py:976] (1/7) Epoch 9, batch 5200, loss[loss=0.2292, simple_loss=0.2928, pruned_loss=0.0828, over 4757.00 frames. ], tot_loss[loss=0.1928, simple_loss=0.2585, pruned_loss=0.06355, over 955840.32 frames. ], batch size: 59, lr: 3.77e-03, grad_scale: 32.0 2023-04-26 23:18:49,051 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.109e+02 1.700e+02 2.036e+02 2.415e+02 4.035e+02, threshold=4.072e+02, percent-clipped=2.0 2023-04-26 23:19:01,912 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([0.9269, 2.0580, 1.7008, 1.4496, 1.9454, 1.6161, 2.5884, 1.3852], device='cuda:1'), covar=tensor([0.3473, 0.1766, 0.4329, 0.3311, 0.1698, 0.2587, 0.1342, 0.4678], device='cuda:1'), in_proj_covar=tensor([0.0344, 0.0347, 0.0430, 0.0361, 0.0386, 0.0382, 0.0380, 0.0419], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 23:19:08,566 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=51062.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:19:14,489 INFO [finetune.py:976] (1/7) Epoch 9, batch 5250, loss[loss=0.1721, simple_loss=0.2521, pruned_loss=0.04601, over 4808.00 frames. ], tot_loss[loss=0.1937, simple_loss=0.2594, pruned_loss=0.06395, over 951679.03 frames. ], batch size: 51, lr: 3.77e-03, grad_scale: 32.0 2023-04-26 23:19:47,750 INFO [finetune.py:976] (1/7) Epoch 9, batch 5300, loss[loss=0.1843, simple_loss=0.2577, pruned_loss=0.05544, over 4799.00 frames. ], tot_loss[loss=0.1963, simple_loss=0.2623, pruned_loss=0.06518, over 954416.68 frames. ], batch size: 45, lr: 3.77e-03, grad_scale: 32.0 2023-04-26 23:19:48,488 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=51123.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:19:56,069 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.143e+02 1.811e+02 2.045e+02 2.583e+02 4.950e+02, threshold=4.090e+02, percent-clipped=1.0 2023-04-26 23:19:57,962 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=51138.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:20:15,867 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=51164.0, num_to_drop=1, layers_to_drop={2} 2023-04-26 23:20:20,663 INFO [finetune.py:976] (1/7) Epoch 9, batch 5350, loss[loss=0.1915, simple_loss=0.2448, pruned_loss=0.06908, over 4697.00 frames. ], tot_loss[loss=0.1969, simple_loss=0.2631, pruned_loss=0.06539, over 953090.69 frames. ], batch size: 23, lr: 3.77e-03, grad_scale: 32.0 2023-04-26 23:20:20,776 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=51172.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 23:20:21,392 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0009, 2.7096, 2.1593, 2.0268, 1.4086, 1.4730, 2.2566, 1.4414], device='cuda:1'), covar=tensor([0.1667, 0.1632, 0.1485, 0.1934, 0.2621, 0.2125, 0.1018, 0.2123], device='cuda:1'), in_proj_covar=tensor([0.0194, 0.0216, 0.0171, 0.0204, 0.0204, 0.0183, 0.0160, 0.0187], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-26 23:20:46,714 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=51202.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:20:50,320 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.96 vs. limit=2.0 2023-04-26 23:21:10,254 INFO [finetune.py:976] (1/7) Epoch 9, batch 5400, loss[loss=0.2238, simple_loss=0.2887, pruned_loss=0.07949, over 4755.00 frames. ], tot_loss[loss=0.1948, simple_loss=0.2603, pruned_loss=0.0647, over 952711.76 frames. ], batch size: 54, lr: 3.77e-03, grad_scale: 32.0 2023-04-26 23:21:21,593 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=51233.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 23:21:22,681 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.244e+02 1.662e+02 1.922e+02 2.270e+02 4.708e+02, threshold=3.844e+02, percent-clipped=3.0 2023-04-26 23:21:43,668 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=51250.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:22:14,586 INFO [finetune.py:976] (1/7) Epoch 9, batch 5450, loss[loss=0.1764, simple_loss=0.2362, pruned_loss=0.05831, over 4750.00 frames. ], tot_loss[loss=0.1912, simple_loss=0.2566, pruned_loss=0.06291, over 952242.65 frames. ], batch size: 28, lr: 3.77e-03, grad_scale: 32.0 2023-04-26 23:22:47,594 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=51302.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:22:54,444 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5065, 1.1538, 0.4509, 1.1640, 1.0968, 1.3842, 1.2560, 1.2485], device='cuda:1'), covar=tensor([0.0529, 0.0411, 0.0443, 0.0583, 0.0328, 0.0540, 0.0513, 0.0581], device='cuda:1'), in_proj_covar=tensor([0.0030, 0.0026, 0.0023, 0.0030, 0.0021, 0.0029, 0.0029, 0.0030], device='cuda:1'), out_proj_covar=tensor([0.0049, 0.0045, 0.0038, 0.0050, 0.0038, 0.0048, 0.0048, 0.0050], device='cuda:1') 2023-04-26 23:23:10,597 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([0.5979, 1.6786, 1.6366, 1.1869, 1.8308, 1.3867, 2.2899, 1.4085], device='cuda:1'), covar=tensor([0.4083, 0.1849, 0.4930, 0.2997, 0.1569, 0.2647, 0.1563, 0.5192], device='cuda:1'), in_proj_covar=tensor([0.0346, 0.0349, 0.0431, 0.0363, 0.0388, 0.0384, 0.0380, 0.0421], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 23:23:19,200 INFO [finetune.py:976] (1/7) Epoch 9, batch 5500, loss[loss=0.1661, simple_loss=0.2122, pruned_loss=0.06002, over 4214.00 frames. ], tot_loss[loss=0.1895, simple_loss=0.2544, pruned_loss=0.06233, over 952315.92 frames. ], batch size: 17, lr: 3.77e-03, grad_scale: 32.0 2023-04-26 23:23:32,910 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.202e+01 1.588e+02 1.902e+02 2.243e+02 3.887e+02, threshold=3.804e+02, percent-clipped=1.0 2023-04-26 23:23:40,037 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-04-26 23:24:14,746 INFO [finetune.py:976] (1/7) Epoch 9, batch 5550, loss[loss=0.1774, simple_loss=0.2507, pruned_loss=0.0521, over 4790.00 frames. ], tot_loss[loss=0.1927, simple_loss=0.2572, pruned_loss=0.06404, over 950235.26 frames. ], batch size: 29, lr: 3.77e-03, grad_scale: 32.0 2023-04-26 23:24:16,708 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6025, 1.4273, 4.2260, 3.8079, 3.7916, 3.8385, 3.8507, 3.5886], device='cuda:1'), covar=tensor([0.8914, 0.8146, 0.1599, 0.3401, 0.2186, 0.4588, 0.2972, 0.3504], device='cuda:1'), in_proj_covar=tensor([0.0313, 0.0311, 0.0413, 0.0416, 0.0355, 0.0411, 0.0319, 0.0375], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-26 23:24:20,973 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=51382.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:24:26,981 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=51392.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:24:42,925 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=51418.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:24:45,249 INFO [finetune.py:976] (1/7) Epoch 9, batch 5600, loss[loss=0.168, simple_loss=0.2063, pruned_loss=0.06486, over 4033.00 frames. ], tot_loss[loss=0.1943, simple_loss=0.2596, pruned_loss=0.06447, over 948249.79 frames. ], batch size: 17, lr: 3.77e-03, grad_scale: 32.0 2023-04-26 23:24:47,068 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.1738, 1.4421, 1.5357, 1.6952, 1.5402, 1.6858, 1.6165, 1.5620], device='cuda:1'), covar=tensor([0.6097, 0.7306, 0.6082, 0.5591, 0.7196, 1.0238, 0.7051, 0.6515], device='cuda:1'), in_proj_covar=tensor([0.0323, 0.0383, 0.0316, 0.0326, 0.0338, 0.0400, 0.0359, 0.0323], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-26 23:24:47,785 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2023-04-26 23:24:52,682 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.274e+02 1.747e+02 2.120e+02 2.551e+02 6.497e+02, threshold=4.239e+02, percent-clipped=3.0 2023-04-26 23:24:54,539 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=51438.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:24:57,469 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=51443.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:25:03,251 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=51453.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:25:10,020 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=51464.0, num_to_drop=1, layers_to_drop={2} 2023-04-26 23:25:15,161 INFO [finetune.py:976] (1/7) Epoch 9, batch 5650, loss[loss=0.2552, simple_loss=0.3084, pruned_loss=0.101, over 4888.00 frames. ], tot_loss[loss=0.1968, simple_loss=0.2627, pruned_loss=0.06542, over 950262.60 frames. ], batch size: 32, lr: 3.77e-03, grad_scale: 32.0 2023-04-26 23:25:16,427 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9789, 2.4478, 2.0577, 2.3124, 1.8444, 1.9568, 2.0543, 1.6021], device='cuda:1'), covar=tensor([0.1639, 0.1202, 0.0875, 0.1116, 0.2675, 0.1281, 0.1778, 0.2411], device='cuda:1'), in_proj_covar=tensor([0.0297, 0.0317, 0.0227, 0.0289, 0.0317, 0.0273, 0.0259, 0.0281], device='cuda:1'), out_proj_covar=tensor([1.2026e-04, 1.2788e-04, 9.1286e-05, 1.1600e-04, 1.2995e-04, 1.1016e-04, 1.0583e-04, 1.1291e-04], device='cuda:1') 2023-04-26 23:25:23,729 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=51486.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:25:29,028 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=51495.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:25:34,350 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9832, 2.0994, 1.1430, 1.7150, 2.1921, 1.8535, 1.7626, 1.8650], device='cuda:1'), covar=tensor([0.0459, 0.0364, 0.0369, 0.0553, 0.0243, 0.0503, 0.0507, 0.0533], device='cuda:1'), in_proj_covar=tensor([0.0030, 0.0025, 0.0023, 0.0030, 0.0021, 0.0029, 0.0029, 0.0030], device='cuda:1'), out_proj_covar=tensor([0.0049, 0.0044, 0.0038, 0.0050, 0.0037, 0.0048, 0.0048, 0.0049], device='cuda:1') 2023-04-26 23:25:36,724 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([3.3805, 3.3323, 2.5032, 3.8413, 3.3625, 3.3250, 1.6686, 3.2433], device='cuda:1'), covar=tensor([0.1917, 0.1229, 0.3097, 0.2082, 0.3214, 0.2135, 0.5227, 0.2850], device='cuda:1'), in_proj_covar=tensor([0.0241, 0.0214, 0.0247, 0.0299, 0.0297, 0.0248, 0.0267, 0.0270], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-26 23:25:39,094 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=51512.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 23:25:44,895 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=3.92 vs. limit=5.0 2023-04-26 23:25:45,276 INFO [finetune.py:976] (1/7) Epoch 9, batch 5700, loss[loss=0.1567, simple_loss=0.2087, pruned_loss=0.05232, over 4639.00 frames. ], tot_loss[loss=0.1946, simple_loss=0.2587, pruned_loss=0.06527, over 929377.12 frames. ], batch size: 20, lr: 3.77e-03, grad_scale: 32.0 2023-04-26 23:25:48,918 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=51528.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 23:25:53,110 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.028e+02 1.631e+02 1.982e+02 2.330e+02 4.156e+02, threshold=3.963e+02, percent-clipped=0.0 2023-04-26 23:25:56,299 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 2023-04-26 23:26:16,105 INFO [finetune.py:976] (1/7) Epoch 10, batch 0, loss[loss=0.1696, simple_loss=0.2375, pruned_loss=0.05082, over 4904.00 frames. ], tot_loss[loss=0.1696, simple_loss=0.2375, pruned_loss=0.05082, over 4904.00 frames. ], batch size: 33, lr: 3.76e-03, grad_scale: 32.0 2023-04-26 23:26:16,105 INFO [finetune.py:1001] (1/7) Computing validation loss 2023-04-26 23:26:31,820 INFO [finetune.py:1010] (1/7) Epoch 10, validation: loss=0.1558, simple_loss=0.2282, pruned_loss=0.04164, over 2265189.00 frames. 2023-04-26 23:26:31,820 INFO [finetune.py:1011] (1/7) Maximum memory allocated so far is 6391MB 2023-04-26 23:26:37,654 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=51556.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:26:41,446 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.50 vs. limit=2.0 2023-04-26 23:26:49,312 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-04-26 23:27:05,702 INFO [finetune.py:976] (1/7) Epoch 10, batch 50, loss[loss=0.1836, simple_loss=0.2395, pruned_loss=0.06383, over 4745.00 frames. ], tot_loss[loss=0.1979, simple_loss=0.2644, pruned_loss=0.06569, over 215743.58 frames. ], batch size: 23, lr: 3.76e-03, grad_scale: 32.0 2023-04-26 23:27:08,563 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=51602.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:27:24,340 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.1132, 1.4609, 1.9354, 2.3271, 1.8251, 1.4543, 1.0952, 1.6360], device='cuda:1'), covar=tensor([0.3604, 0.4001, 0.1910, 0.2647, 0.3073, 0.3114, 0.5008, 0.2666], device='cuda:1'), in_proj_covar=tensor([0.0283, 0.0251, 0.0220, 0.0318, 0.0214, 0.0227, 0.0233, 0.0186], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-26 23:27:31,934 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.265e+02 1.709e+02 2.017e+02 2.493e+02 1.011e+03, threshold=4.035e+02, percent-clipped=6.0 2023-04-26 23:27:43,979 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6897, 1.2582, 1.7864, 2.1898, 1.8550, 1.6539, 1.7154, 1.7114], device='cuda:1'), covar=tensor([0.6325, 0.9033, 0.8850, 0.8162, 0.7162, 1.0446, 1.1088, 1.0330], device='cuda:1'), in_proj_covar=tensor([0.0407, 0.0415, 0.0501, 0.0518, 0.0436, 0.0454, 0.0466, 0.0463], device='cuda:1'), out_proj_covar=tensor([9.9039e-05, 1.0286e-04, 1.1289e-04, 1.2325e-04, 1.0586e-04, 1.0995e-04, 1.1191e-04, 1.1190e-04], device='cuda:1') 2023-04-26 23:27:45,650 INFO [finetune.py:976] (1/7) Epoch 10, batch 100, loss[loss=0.2081, simple_loss=0.277, pruned_loss=0.06966, over 4902.00 frames. ], tot_loss[loss=0.1939, simple_loss=0.2585, pruned_loss=0.06466, over 378554.81 frames. ], batch size: 43, lr: 3.76e-03, grad_scale: 16.0 2023-04-26 23:27:46,796 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=51650.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:28:21,315 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6378, 3.5497, 0.9950, 2.0814, 2.0975, 2.4934, 2.1128, 0.9916], device='cuda:1'), covar=tensor([0.1244, 0.0741, 0.1818, 0.1077, 0.0952, 0.0998, 0.1314, 0.2146], device='cuda:1'), in_proj_covar=tensor([0.0117, 0.0249, 0.0141, 0.0122, 0.0135, 0.0154, 0.0118, 0.0122], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-26 23:28:38,560 INFO [finetune.py:976] (1/7) Epoch 10, batch 150, loss[loss=0.1775, simple_loss=0.241, pruned_loss=0.05702, over 4844.00 frames. ], tot_loss[loss=0.1887, simple_loss=0.2529, pruned_loss=0.0623, over 507434.43 frames. ], batch size: 47, lr: 3.76e-03, grad_scale: 16.0 2023-04-26 23:29:02,025 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([4.6701, 4.4640, 3.0699, 5.3301, 4.6027, 4.5827, 2.1645, 4.5333], device='cuda:1'), covar=tensor([0.1508, 0.1057, 0.3181, 0.0888, 0.3243, 0.1719, 0.5814, 0.2119], device='cuda:1'), in_proj_covar=tensor([0.0242, 0.0215, 0.0248, 0.0301, 0.0298, 0.0248, 0.0268, 0.0269], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-26 23:29:04,452 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=51718.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:29:12,104 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.54 vs. limit=5.0 2023-04-26 23:29:16,234 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6991, 2.3814, 1.7934, 1.6378, 1.2492, 1.2997, 1.8793, 1.2325], device='cuda:1'), covar=tensor([0.1875, 0.1549, 0.1668, 0.2171, 0.2872, 0.2284, 0.1151, 0.2304], device='cuda:1'), in_proj_covar=tensor([0.0194, 0.0216, 0.0171, 0.0205, 0.0205, 0.0184, 0.0160, 0.0187], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-26 23:29:20,968 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.275e+02 1.676e+02 2.023e+02 2.458e+02 4.768e+02, threshold=4.046e+02, percent-clipped=1.0 2023-04-26 23:29:22,215 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=51738.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:29:29,229 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=51747.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:29:29,796 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=51748.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:29:30,341 INFO [finetune.py:976] (1/7) Epoch 10, batch 200, loss[loss=0.1587, simple_loss=0.2262, pruned_loss=0.04557, over 4790.00 frames. ], tot_loss[loss=0.1898, simple_loss=0.254, pruned_loss=0.06276, over 609179.61 frames. ], batch size: 29, lr: 3.76e-03, grad_scale: 16.0 2023-04-26 23:29:42,677 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=51766.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:30:04,062 INFO [finetune.py:976] (1/7) Epoch 10, batch 250, loss[loss=0.2126, simple_loss=0.3033, pruned_loss=0.06091, over 4845.00 frames. ], tot_loss[loss=0.1932, simple_loss=0.2583, pruned_loss=0.06398, over 687194.16 frames. ], batch size: 49, lr: 3.76e-03, grad_scale: 16.0 2023-04-26 23:30:11,165 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=51808.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:30:23,869 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=51828.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 23:30:28,669 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.252e+02 1.711e+02 1.990e+02 2.532e+02 5.377e+02, threshold=3.981e+02, percent-clipped=2.0 2023-04-26 23:30:37,590 INFO [finetune.py:976] (1/7) Epoch 10, batch 300, loss[loss=0.1435, simple_loss=0.2184, pruned_loss=0.03428, over 4806.00 frames. ], tot_loss[loss=0.1951, simple_loss=0.2612, pruned_loss=0.06449, over 747072.09 frames. ], batch size: 25, lr: 3.76e-03, grad_scale: 16.0 2023-04-26 23:30:38,885 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=51851.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:30:44,179 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5518, 0.6289, 1.3403, 1.9707, 1.6699, 1.4741, 1.4453, 1.5180], device='cuda:1'), covar=tensor([0.5420, 0.7511, 0.7684, 0.7718, 0.6496, 0.8285, 0.8387, 0.8026], device='cuda:1'), in_proj_covar=tensor([0.0409, 0.0416, 0.0502, 0.0520, 0.0437, 0.0456, 0.0467, 0.0465], device='cuda:1'), out_proj_covar=tensor([9.9536e-05, 1.0319e-04, 1.1320e-04, 1.2373e-04, 1.0616e-04, 1.1038e-04, 1.1219e-04, 1.1223e-04], device='cuda:1') 2023-04-26 23:30:56,475 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=51876.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 23:31:10,834 INFO [finetune.py:976] (1/7) Epoch 10, batch 350, loss[loss=0.248, simple_loss=0.293, pruned_loss=0.1015, over 4899.00 frames. ], tot_loss[loss=0.195, simple_loss=0.2616, pruned_loss=0.06424, over 791307.65 frames. ], batch size: 32, lr: 3.76e-03, grad_scale: 16.0 2023-04-26 23:31:41,424 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.954e+01 1.688e+02 1.992e+02 2.453e+02 5.822e+02, threshold=3.984e+02, percent-clipped=3.0 2023-04-26 23:32:00,714 INFO [finetune.py:976] (1/7) Epoch 10, batch 400, loss[loss=0.2042, simple_loss=0.268, pruned_loss=0.0702, over 4847.00 frames. ], tot_loss[loss=0.1953, simple_loss=0.2624, pruned_loss=0.06407, over 827656.67 frames. ], batch size: 49, lr: 3.76e-03, grad_scale: 16.0 2023-04-26 23:32:51,604 INFO [finetune.py:976] (1/7) Epoch 10, batch 450, loss[loss=0.2067, simple_loss=0.2658, pruned_loss=0.07379, over 4384.00 frames. ], tot_loss[loss=0.1928, simple_loss=0.2599, pruned_loss=0.06287, over 855851.09 frames. ], batch size: 66, lr: 3.76e-03, grad_scale: 16.0 2023-04-26 23:33:45,355 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.213e+02 1.582e+02 1.952e+02 2.304e+02 4.077e+02, threshold=3.905e+02, percent-clipped=1.0 2023-04-26 23:33:46,697 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=52038.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:33:53,300 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=52040.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:33:58,079 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=52048.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:34:04,124 INFO [finetune.py:976] (1/7) Epoch 10, batch 500, loss[loss=0.1945, simple_loss=0.2626, pruned_loss=0.06319, over 4859.00 frames. ], tot_loss[loss=0.1916, simple_loss=0.2578, pruned_loss=0.06271, over 879147.89 frames. ], batch size: 47, lr: 3.76e-03, grad_scale: 16.0 2023-04-26 23:34:25,461 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.74 vs. limit=5.0 2023-04-26 23:34:51,209 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.2188, 1.6347, 2.0993, 2.3692, 2.0075, 1.6027, 1.2612, 1.8230], device='cuda:1'), covar=tensor([0.3809, 0.4062, 0.1902, 0.2943, 0.3174, 0.3160, 0.5111, 0.2612], device='cuda:1'), in_proj_covar=tensor([0.0283, 0.0251, 0.0220, 0.0319, 0.0213, 0.0227, 0.0234, 0.0186], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-26 23:34:51,735 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=52086.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:35:03,595 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=52096.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:35:11,182 INFO [finetune.py:976] (1/7) Epoch 10, batch 550, loss[loss=0.1619, simple_loss=0.2243, pruned_loss=0.04976, over 4774.00 frames. ], tot_loss[loss=0.1888, simple_loss=0.2548, pruned_loss=0.06142, over 897274.18 frames. ], batch size: 26, lr: 3.76e-03, grad_scale: 16.0 2023-04-26 23:35:12,520 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=52101.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:35:13,631 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=52103.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:36:05,085 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.236e+02 1.854e+02 2.167e+02 2.503e+02 5.657e+02, threshold=4.334e+02, percent-clipped=4.0 2023-04-26 23:36:07,114 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6891, 2.3611, 1.8705, 1.7357, 1.2542, 1.2761, 1.9350, 1.2171], device='cuda:1'), covar=tensor([0.1771, 0.1616, 0.1504, 0.1923, 0.2547, 0.2198, 0.1048, 0.2238], device='cuda:1'), in_proj_covar=tensor([0.0195, 0.0216, 0.0171, 0.0205, 0.0205, 0.0184, 0.0161, 0.0188], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-26 23:36:18,684 INFO [finetune.py:976] (1/7) Epoch 10, batch 600, loss[loss=0.214, simple_loss=0.2838, pruned_loss=0.0721, over 4817.00 frames. ], tot_loss[loss=0.1912, simple_loss=0.2565, pruned_loss=0.06295, over 909267.31 frames. ], batch size: 40, lr: 3.76e-03, grad_scale: 16.0 2023-04-26 23:36:19,995 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=52151.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:37:10,905 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.2584, 1.6557, 2.1293, 2.5117, 2.0551, 1.6342, 1.2854, 1.7312], device='cuda:1'), covar=tensor([0.3928, 0.4324, 0.1864, 0.2855, 0.3296, 0.3293, 0.5265, 0.2983], device='cuda:1'), in_proj_covar=tensor([0.0284, 0.0251, 0.0220, 0.0319, 0.0214, 0.0227, 0.0234, 0.0186], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-26 23:37:12,638 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6436, 1.9425, 0.8993, 1.3081, 2.1445, 1.5519, 1.5078, 1.4359], device='cuda:1'), covar=tensor([0.0540, 0.0376, 0.0359, 0.0604, 0.0267, 0.0550, 0.0532, 0.0612], device='cuda:1'), in_proj_covar=tensor([0.0030, 0.0025, 0.0023, 0.0030, 0.0021, 0.0029, 0.0029, 0.0030], device='cuda:1'), out_proj_covar=tensor([0.0049, 0.0044, 0.0038, 0.0050, 0.0038, 0.0048, 0.0048, 0.0050], device='cuda:1') 2023-04-26 23:37:16,290 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=52193.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:37:25,227 INFO [finetune.py:976] (1/7) Epoch 10, batch 650, loss[loss=0.1894, simple_loss=0.2569, pruned_loss=0.06097, over 4915.00 frames. ], tot_loss[loss=0.193, simple_loss=0.259, pruned_loss=0.06345, over 920689.08 frames. ], batch size: 32, lr: 3.76e-03, grad_scale: 16.0 2023-04-26 23:37:25,287 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=52199.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:38:17,627 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.057e+02 1.623e+02 1.914e+02 2.334e+02 8.066e+02, threshold=3.828e+02, percent-clipped=2.0 2023-04-26 23:38:18,951 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.1652, 1.5676, 1.3998, 1.8226, 1.6082, 1.7804, 1.4244, 3.1425], device='cuda:1'), covar=tensor([0.0659, 0.0748, 0.0790, 0.1137, 0.0627, 0.0476, 0.0721, 0.0201], device='cuda:1'), in_proj_covar=tensor([0.0038, 0.0039, 0.0040, 0.0044, 0.0040, 0.0039, 0.0039, 0.0059], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0014, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:1') 2023-04-26 23:38:30,845 INFO [finetune.py:976] (1/7) Epoch 10, batch 700, loss[loss=0.2491, simple_loss=0.2845, pruned_loss=0.1068, over 4748.00 frames. ], tot_loss[loss=0.1937, simple_loss=0.2598, pruned_loss=0.06381, over 927828.90 frames. ], batch size: 27, lr: 3.76e-03, grad_scale: 16.0 2023-04-26 23:38:38,114 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-04-26 23:38:39,787 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=52254.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:39:44,471 INFO [finetune.py:976] (1/7) Epoch 10, batch 750, loss[loss=0.1677, simple_loss=0.2343, pruned_loss=0.0506, over 4766.00 frames. ], tot_loss[loss=0.1945, simple_loss=0.2606, pruned_loss=0.06417, over 933003.52 frames. ], batch size: 26, lr: 3.76e-03, grad_scale: 16.0 2023-04-26 23:39:56,788 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=52312.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:40:13,564 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.225e+02 1.731e+02 1.996e+02 2.532e+02 4.524e+02, threshold=3.992e+02, percent-clipped=2.0 2023-04-26 23:40:13,716 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.7541, 1.9614, 1.8414, 2.0332, 1.8493, 1.9714, 1.9802, 1.9002], device='cuda:1'), covar=tensor([0.5364, 0.8026, 0.6785, 0.5932, 0.7113, 0.9537, 0.7663, 0.7513], device='cuda:1'), in_proj_covar=tensor([0.0324, 0.0382, 0.0316, 0.0325, 0.0339, 0.0402, 0.0360, 0.0323], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-26 23:40:22,484 INFO [finetune.py:976] (1/7) Epoch 10, batch 800, loss[loss=0.1705, simple_loss=0.2321, pruned_loss=0.05441, over 4823.00 frames. ], tot_loss[loss=0.1958, simple_loss=0.2621, pruned_loss=0.06477, over 937864.12 frames. ], batch size: 30, lr: 3.76e-03, grad_scale: 16.0 2023-04-26 23:40:27,733 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-04-26 23:40:37,965 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=52373.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:40:54,331 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=52396.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:40:56,135 INFO [finetune.py:976] (1/7) Epoch 10, batch 850, loss[loss=0.179, simple_loss=0.2453, pruned_loss=0.05635, over 4804.00 frames. ], tot_loss[loss=0.1935, simple_loss=0.2596, pruned_loss=0.06369, over 941804.38 frames. ], batch size: 51, lr: 3.76e-03, grad_scale: 16.0 2023-04-26 23:40:58,772 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=52403.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:41:30,356 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8594, 1.1714, 4.9727, 4.7064, 4.3128, 4.6693, 4.3878, 4.3812], device='cuda:1'), covar=tensor([0.6725, 0.6620, 0.0978, 0.1549, 0.0941, 0.1350, 0.1465, 0.1285], device='cuda:1'), in_proj_covar=tensor([0.0309, 0.0308, 0.0407, 0.0410, 0.0349, 0.0405, 0.0314, 0.0371], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-26 23:41:30,883 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.154e+02 1.587e+02 1.831e+02 2.127e+02 6.660e+02, threshold=3.662e+02, percent-clipped=3.0 2023-04-26 23:41:33,556 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-04-26 23:41:33,605 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.74 vs. limit=2.0 2023-04-26 23:41:41,225 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-04-26 23:41:44,733 INFO [finetune.py:976] (1/7) Epoch 10, batch 900, loss[loss=0.1642, simple_loss=0.2353, pruned_loss=0.04657, over 4756.00 frames. ], tot_loss[loss=0.19, simple_loss=0.2556, pruned_loss=0.0622, over 945482.64 frames. ], batch size: 54, lr: 3.76e-03, grad_scale: 16.0 2023-04-26 23:41:51,262 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=52451.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:42:39,326 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.32 vs. limit=5.0 2023-04-26 23:42:48,723 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.2480, 1.5485, 1.3391, 1.4600, 1.2866, 1.2229, 1.3948, 1.0828], device='cuda:1'), covar=tensor([0.1758, 0.1212, 0.0968, 0.1334, 0.3763, 0.1333, 0.1637, 0.2350], device='cuda:1'), in_proj_covar=tensor([0.0294, 0.0314, 0.0226, 0.0287, 0.0316, 0.0271, 0.0256, 0.0278], device='cuda:1'), out_proj_covar=tensor([1.1901e-04, 1.2666e-04, 9.0749e-05, 1.1522e-04, 1.2934e-04, 1.0904e-04, 1.0460e-04, 1.1162e-04], device='cuda:1') 2023-04-26 23:42:49,241 INFO [finetune.py:976] (1/7) Epoch 10, batch 950, loss[loss=0.2064, simple_loss=0.2745, pruned_loss=0.06909, over 4804.00 frames. ], tot_loss[loss=0.1894, simple_loss=0.2546, pruned_loss=0.06208, over 949349.03 frames. ], batch size: 45, lr: 3.76e-03, grad_scale: 16.0 2023-04-26 23:43:00,205 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-04-26 23:43:29,396 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 2023-04-26 23:43:41,296 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.114e+02 1.638e+02 1.935e+02 2.436e+02 4.596e+02, threshold=3.871e+02, percent-clipped=3.0 2023-04-26 23:43:55,594 INFO [finetune.py:976] (1/7) Epoch 10, batch 1000, loss[loss=0.1784, simple_loss=0.2588, pruned_loss=0.04903, over 4802.00 frames. ], tot_loss[loss=0.194, simple_loss=0.2591, pruned_loss=0.06442, over 952742.40 frames. ], batch size: 51, lr: 3.76e-03, grad_scale: 16.0 2023-04-26 23:44:01,403 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=52549.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:44:04,507 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-04-26 23:44:05,066 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=52555.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 23:44:28,280 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.74 vs. limit=2.0 2023-04-26 23:45:07,088 INFO [finetune.py:976] (1/7) Epoch 10, batch 1050, loss[loss=0.2029, simple_loss=0.272, pruned_loss=0.06694, over 4898.00 frames. ], tot_loss[loss=0.1948, simple_loss=0.261, pruned_loss=0.0643, over 951471.89 frames. ], batch size: 32, lr: 3.76e-03, grad_scale: 16.0 2023-04-26 23:45:29,129 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=52616.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 23:45:43,665 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-04-26 23:45:52,966 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.267e+02 1.739e+02 2.075e+02 2.346e+02 3.891e+02, threshold=4.149e+02, percent-clipped=1.0 2023-04-26 23:46:06,290 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.1848, 1.9263, 2.2120, 2.6838, 2.5702, 2.1884, 1.8041, 2.2219], device='cuda:1'), covar=tensor([0.0892, 0.1127, 0.0615, 0.0543, 0.0610, 0.0795, 0.0897, 0.0625], device='cuda:1'), in_proj_covar=tensor([0.0196, 0.0205, 0.0183, 0.0176, 0.0178, 0.0189, 0.0160, 0.0185], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 23:46:13,537 INFO [finetune.py:976] (1/7) Epoch 10, batch 1100, loss[loss=0.1441, simple_loss=0.2092, pruned_loss=0.03948, over 4748.00 frames. ], tot_loss[loss=0.1951, simple_loss=0.2615, pruned_loss=0.06438, over 951602.91 frames. ], batch size: 23, lr: 3.76e-03, grad_scale: 16.0 2023-04-26 23:46:36,537 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=52668.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:46:47,485 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.3984, 2.4342, 2.1320, 2.2300, 2.5949, 2.1199, 3.4883, 1.8473], device='cuda:1'), covar=tensor([0.4344, 0.2274, 0.5073, 0.3846, 0.2047, 0.2935, 0.1620, 0.4875], device='cuda:1'), in_proj_covar=tensor([0.0346, 0.0350, 0.0432, 0.0364, 0.0390, 0.0386, 0.0382, 0.0422], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 23:46:52,290 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.67 vs. limit=2.0 2023-04-26 23:46:55,072 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9375, 2.5380, 2.3154, 1.9548, 1.3303, 1.4414, 2.4755, 1.4279], device='cuda:1'), covar=tensor([0.1730, 0.1743, 0.1257, 0.1804, 0.2366, 0.1933, 0.0820, 0.2104], device='cuda:1'), in_proj_covar=tensor([0.0195, 0.0216, 0.0170, 0.0204, 0.0204, 0.0184, 0.0160, 0.0188], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-26 23:46:56,134 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=52696.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:46:57,843 INFO [finetune.py:976] (1/7) Epoch 10, batch 1150, loss[loss=0.1603, simple_loss=0.2362, pruned_loss=0.04227, over 4805.00 frames. ], tot_loss[loss=0.1959, simple_loss=0.2628, pruned_loss=0.06451, over 954902.18 frames. ], batch size: 25, lr: 3.75e-03, grad_scale: 16.0 2023-04-26 23:46:58,112 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-04-26 23:47:19,005 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-04-26 23:47:21,723 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.059e+02 1.582e+02 1.906e+02 2.303e+02 4.186e+02, threshold=3.812e+02, percent-clipped=1.0 2023-04-26 23:47:33,673 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=52744.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:47:43,257 INFO [finetune.py:976] (1/7) Epoch 10, batch 1200, loss[loss=0.2424, simple_loss=0.3027, pruned_loss=0.0911, over 4741.00 frames. ], tot_loss[loss=0.1942, simple_loss=0.2611, pruned_loss=0.06372, over 955527.84 frames. ], batch size: 23, lr: 3.75e-03, grad_scale: 16.0 2023-04-26 23:47:45,181 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=52752.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:47:54,433 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.5452, 1.5970, 1.7640, 1.8828, 1.6173, 1.8428, 1.9545, 1.8518], device='cuda:1'), covar=tensor([0.4483, 0.7096, 0.6475, 0.5624, 0.6866, 0.9084, 0.6419, 0.6318], device='cuda:1'), in_proj_covar=tensor([0.0325, 0.0381, 0.0316, 0.0325, 0.0340, 0.0401, 0.0360, 0.0324], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-26 23:48:19,659 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.1919, 1.2877, 1.6133, 1.7268, 1.5445, 1.8204, 1.7369, 1.6853], device='cuda:1'), covar=tensor([0.4346, 0.6372, 0.5549, 0.5103, 0.6385, 0.7949, 0.5969, 0.5680], device='cuda:1'), in_proj_covar=tensor([0.0324, 0.0381, 0.0316, 0.0325, 0.0339, 0.0401, 0.0360, 0.0323], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-26 23:48:48,804 INFO [finetune.py:976] (1/7) Epoch 10, batch 1250, loss[loss=0.2265, simple_loss=0.2881, pruned_loss=0.08242, over 4871.00 frames. ], tot_loss[loss=0.1917, simple_loss=0.2583, pruned_loss=0.0626, over 957020.88 frames. ], batch size: 31, lr: 3.75e-03, grad_scale: 16.0 2023-04-26 23:49:09,934 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=52813.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:49:12,337 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6095, 3.9531, 0.8444, 2.2223, 2.1899, 2.6298, 2.3822, 1.0500], device='cuda:1'), covar=tensor([0.1287, 0.0950, 0.2046, 0.1193, 0.0977, 0.1094, 0.1359, 0.2028], device='cuda:1'), in_proj_covar=tensor([0.0117, 0.0248, 0.0141, 0.0122, 0.0135, 0.0154, 0.0119, 0.0121], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-26 23:49:34,634 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.104e+02 1.609e+02 1.926e+02 2.217e+02 3.154e+02, threshold=3.851e+02, percent-clipped=0.0 2023-04-26 23:49:48,761 INFO [finetune.py:976] (1/7) Epoch 10, batch 1300, loss[loss=0.2009, simple_loss=0.2501, pruned_loss=0.07581, over 4313.00 frames. ], tot_loss[loss=0.1879, simple_loss=0.2542, pruned_loss=0.06086, over 957261.55 frames. ], batch size: 65, lr: 3.75e-03, grad_scale: 16.0 2023-04-26 23:49:48,870 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=52849.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:49:49,577 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.70 vs. limit=2.0 2023-04-26 23:50:20,538 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=52897.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:50:21,745 INFO [finetune.py:976] (1/7) Epoch 10, batch 1350, loss[loss=0.1598, simple_loss=0.2288, pruned_loss=0.04538, over 4727.00 frames. ], tot_loss[loss=0.1875, simple_loss=0.2534, pruned_loss=0.0608, over 956344.99 frames. ], batch size: 23, lr: 3.75e-03, grad_scale: 16.0 2023-04-26 23:50:31,129 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=52911.0, num_to_drop=1, layers_to_drop={3} 2023-04-26 23:50:44,368 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([0.3764, 1.3671, 1.3708, 1.0087, 1.4474, 1.1799, 1.7536, 1.2102], device='cuda:1'), covar=tensor([0.3215, 0.1438, 0.4294, 0.2276, 0.1303, 0.1828, 0.1428, 0.4684], device='cuda:1'), in_proj_covar=tensor([0.0343, 0.0348, 0.0428, 0.0362, 0.0387, 0.0383, 0.0379, 0.0419], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-26 23:50:46,658 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.089e+02 1.765e+02 2.062e+02 2.587e+02 4.153e+02, threshold=4.125e+02, percent-clipped=1.0 2023-04-26 23:50:55,066 INFO [finetune.py:976] (1/7) Epoch 10, batch 1400, loss[loss=0.216, simple_loss=0.2886, pruned_loss=0.0717, over 4757.00 frames. ], tot_loss[loss=0.1921, simple_loss=0.2582, pruned_loss=0.06299, over 957804.41 frames. ], batch size: 27, lr: 3.75e-03, grad_scale: 16.0 2023-04-26 23:50:58,848 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.72 vs. limit=2.0 2023-04-26 23:51:09,507 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=52968.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:51:10,285 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-04-26 23:51:25,780 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([3.3480, 3.2195, 2.6431, 3.8913, 3.2593, 3.2783, 1.8586, 3.3132], device='cuda:1'), covar=tensor([0.1868, 0.1485, 0.4087, 0.1626, 0.2542, 0.1971, 0.4485, 0.2401], device='cuda:1'), in_proj_covar=tensor([0.0243, 0.0217, 0.0248, 0.0302, 0.0299, 0.0249, 0.0268, 0.0270], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-26 23:51:28,729 INFO [finetune.py:976] (1/7) Epoch 10, batch 1450, loss[loss=0.2627, simple_loss=0.3108, pruned_loss=0.1073, over 4086.00 frames. ], tot_loss[loss=0.194, simple_loss=0.2601, pruned_loss=0.06393, over 954170.06 frames. ], batch size: 65, lr: 3.75e-03, grad_scale: 16.0 2023-04-26 23:51:31,455 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.4155, 1.8277, 2.3006, 2.8220, 2.1836, 1.7844, 1.6448, 2.1410], device='cuda:1'), covar=tensor([0.3863, 0.3932, 0.1835, 0.3223, 0.3405, 0.3104, 0.4808, 0.2717], device='cuda:1'), in_proj_covar=tensor([0.0284, 0.0252, 0.0220, 0.0319, 0.0215, 0.0229, 0.0234, 0.0187], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-26 23:51:41,185 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=53016.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:52:04,980 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.250e+02 1.692e+02 2.005e+02 2.440e+02 4.945e+02, threshold=4.009e+02, percent-clipped=1.0 2023-04-26 23:52:18,927 INFO [finetune.py:976] (1/7) Epoch 10, batch 1500, loss[loss=0.2036, simple_loss=0.2617, pruned_loss=0.07273, over 4804.00 frames. ], tot_loss[loss=0.195, simple_loss=0.261, pruned_loss=0.06456, over 955009.57 frames. ], batch size: 41, lr: 3.75e-03, grad_scale: 16.0 2023-04-26 23:52:28,803 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=53056.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:53:19,945 INFO [finetune.py:976] (1/7) Epoch 10, batch 1550, loss[loss=0.2317, simple_loss=0.2814, pruned_loss=0.09101, over 4804.00 frames. ], tot_loss[loss=0.1939, simple_loss=0.2602, pruned_loss=0.06386, over 955215.77 frames. ], batch size: 41, lr: 3.75e-03, grad_scale: 16.0 2023-04-26 23:53:25,473 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=53108.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:53:32,443 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=53117.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:53:45,207 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.067e+02 1.713e+02 2.006e+02 2.452e+02 5.414e+02, threshold=4.013e+02, percent-clipped=2.0 2023-04-26 23:53:53,127 INFO [finetune.py:976] (1/7) Epoch 10, batch 1600, loss[loss=0.1942, simple_loss=0.2618, pruned_loss=0.06325, over 4902.00 frames. ], tot_loss[loss=0.1921, simple_loss=0.258, pruned_loss=0.06312, over 955596.20 frames. ], batch size: 36, lr: 3.75e-03, grad_scale: 16.0 2023-04-26 23:54:44,152 INFO [finetune.py:976] (1/7) Epoch 10, batch 1650, loss[loss=0.1892, simple_loss=0.2544, pruned_loss=0.06197, over 4882.00 frames. ], tot_loss[loss=0.1901, simple_loss=0.2558, pruned_loss=0.06217, over 956290.40 frames. ], batch size: 35, lr: 3.75e-03, grad_scale: 16.0 2023-04-26 23:54:47,348 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.4722, 1.8018, 1.8673, 2.0117, 1.7966, 1.9266, 1.9754, 1.9343], device='cuda:1'), covar=tensor([0.5395, 0.7648, 0.6066, 0.5516, 0.7009, 0.9890, 0.6980, 0.6346], device='cuda:1'), in_proj_covar=tensor([0.0328, 0.0384, 0.0318, 0.0327, 0.0342, 0.0404, 0.0363, 0.0327], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-26 23:54:49,096 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9744, 1.3361, 5.1433, 4.8230, 4.4674, 4.8699, 4.5118, 4.5852], device='cuda:1'), covar=tensor([0.6343, 0.6163, 0.0944, 0.1643, 0.0936, 0.1302, 0.1181, 0.1249], device='cuda:1'), in_proj_covar=tensor([0.0306, 0.0305, 0.0404, 0.0407, 0.0348, 0.0404, 0.0313, 0.0370], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-26 23:54:52,155 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=53211.0, num_to_drop=1, layers_to_drop={2} 2023-04-26 23:55:09,622 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.078e+02 1.636e+02 1.864e+02 2.338e+02 5.656e+02, threshold=3.728e+02, percent-clipped=1.0 2023-04-26 23:55:12,824 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=53241.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:55:17,567 INFO [finetune.py:976] (1/7) Epoch 10, batch 1700, loss[loss=0.2226, simple_loss=0.2804, pruned_loss=0.08241, over 4862.00 frames. ], tot_loss[loss=0.1879, simple_loss=0.2535, pruned_loss=0.06116, over 957026.72 frames. ], batch size: 34, lr: 3.75e-03, grad_scale: 16.0 2023-04-26 23:55:24,103 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=53259.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 23:55:31,326 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5478, 2.0072, 1.9185, 2.3198, 2.1756, 2.2971, 1.9218, 4.7753], device='cuda:1'), covar=tensor([0.0598, 0.0777, 0.0789, 0.1175, 0.0634, 0.0495, 0.0750, 0.0155], device='cuda:1'), in_proj_covar=tensor([0.0038, 0.0039, 0.0040, 0.0044, 0.0040, 0.0039, 0.0039, 0.0059], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:1') 2023-04-26 23:55:31,351 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.3381, 2.0136, 2.4566, 2.8921, 1.9726, 1.4711, 1.9962, 1.1205], device='cuda:1'), covar=tensor([0.0569, 0.0919, 0.0495, 0.0728, 0.0801, 0.1680, 0.0947, 0.1191], device='cuda:1'), in_proj_covar=tensor([0.0066, 0.0072, 0.0070, 0.0067, 0.0075, 0.0094, 0.0077, 0.0073], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-26 23:55:39,503 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5884, 1.5397, 0.7477, 1.2625, 1.6906, 1.4497, 1.3660, 1.4011], device='cuda:1'), covar=tensor([0.0514, 0.0399, 0.0388, 0.0577, 0.0295, 0.0551, 0.0512, 0.0562], device='cuda:1'), in_proj_covar=tensor([0.0030, 0.0025, 0.0023, 0.0030, 0.0021, 0.0029, 0.0029, 0.0030], device='cuda:1'), out_proj_covar=tensor([0.0049, 0.0044, 0.0038, 0.0050, 0.0038, 0.0048, 0.0048, 0.0050], device='cuda:1') 2023-04-26 23:55:51,480 INFO [finetune.py:976] (1/7) Epoch 10, batch 1750, loss[loss=0.2619, simple_loss=0.3157, pruned_loss=0.1041, over 4087.00 frames. ], tot_loss[loss=0.1894, simple_loss=0.2551, pruned_loss=0.06187, over 955703.94 frames. ], batch size: 65, lr: 3.75e-03, grad_scale: 16.0 2023-04-26 23:55:53,407 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=53302.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:55:54,613 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=53304.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:56:01,775 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=53315.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:56:16,877 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.153e+02 1.803e+02 2.178e+02 2.575e+02 5.782e+02, threshold=4.356e+02, percent-clipped=6.0 2023-04-26 23:56:25,435 INFO [finetune.py:976] (1/7) Epoch 10, batch 1800, loss[loss=0.2093, simple_loss=0.2717, pruned_loss=0.07346, over 4184.00 frames. ], tot_loss[loss=0.1922, simple_loss=0.2588, pruned_loss=0.06285, over 955660.09 frames. ], batch size: 65, lr: 3.75e-03, grad_scale: 16.0 2023-04-26 23:56:35,681 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=53365.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:56:36,902 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=53367.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:56:42,899 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=53376.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:56:59,051 INFO [finetune.py:976] (1/7) Epoch 10, batch 1850, loss[loss=0.2144, simple_loss=0.2716, pruned_loss=0.07866, over 4823.00 frames. ], tot_loss[loss=0.1941, simple_loss=0.2608, pruned_loss=0.06365, over 956121.55 frames. ], batch size: 33, lr: 3.75e-03, grad_scale: 16.0 2023-04-26 23:57:10,974 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=53408.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:57:13,874 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=53412.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:57:34,992 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=53428.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 23:57:45,405 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.032e+02 1.670e+02 2.068e+02 2.536e+02 6.143e+02, threshold=4.136e+02, percent-clipped=4.0 2023-04-26 23:58:05,924 INFO [finetune.py:976] (1/7) Epoch 10, batch 1900, loss[loss=0.1776, simple_loss=0.2547, pruned_loss=0.05028, over 4923.00 frames. ], tot_loss[loss=0.1957, simple_loss=0.2626, pruned_loss=0.06445, over 957052.35 frames. ], batch size: 42, lr: 3.75e-03, grad_scale: 16.0 2023-04-26 23:58:15,433 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=53456.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:58:48,691 INFO [finetune.py:976] (1/7) Epoch 10, batch 1950, loss[loss=0.1654, simple_loss=0.2342, pruned_loss=0.04829, over 4857.00 frames. ], tot_loss[loss=0.1931, simple_loss=0.2601, pruned_loss=0.06304, over 957237.62 frames. ], batch size: 44, lr: 3.75e-03, grad_scale: 16.0 2023-04-26 23:59:03,223 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.8777, 2.8037, 2.1921, 3.2813, 2.8761, 2.8177, 1.1822, 2.7615], device='cuda:1'), covar=tensor([0.2206, 0.1687, 0.3525, 0.2893, 0.3486, 0.2392, 0.6061, 0.2989], device='cuda:1'), in_proj_covar=tensor([0.0243, 0.0218, 0.0250, 0.0303, 0.0301, 0.0251, 0.0270, 0.0272], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-26 23:59:12,238 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.213e+02 1.722e+02 1.968e+02 2.257e+02 3.746e+02, threshold=3.936e+02, percent-clipped=0.0 2023-04-26 23:59:16,907 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.9263, 2.0524, 2.0241, 2.2242, 1.9258, 2.1158, 2.1735, 2.0801], device='cuda:1'), covar=tensor([0.5323, 0.7972, 0.6054, 0.5496, 0.7150, 0.9127, 0.7702, 0.6867], device='cuda:1'), in_proj_covar=tensor([0.0325, 0.0382, 0.0316, 0.0326, 0.0339, 0.0403, 0.0361, 0.0325], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-26 23:59:17,495 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5106, 1.3418, 1.8690, 1.8786, 1.3964, 1.2205, 1.4987, 1.0265], device='cuda:1'), covar=tensor([0.0791, 0.1014, 0.0485, 0.0627, 0.0866, 0.1114, 0.0751, 0.0850], device='cuda:1'), in_proj_covar=tensor([0.0066, 0.0073, 0.0070, 0.0066, 0.0076, 0.0095, 0.0077, 0.0073], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-26 23:59:18,883 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-04-26 23:59:22,127 INFO [finetune.py:976] (1/7) Epoch 10, batch 2000, loss[loss=0.1682, simple_loss=0.2364, pruned_loss=0.05002, over 4765.00 frames. ], tot_loss[loss=0.191, simple_loss=0.2573, pruned_loss=0.06239, over 956387.83 frames. ], batch size: 27, lr: 3.75e-03, grad_scale: 16.0 2023-04-26 23:59:38,584 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4595, 1.8822, 1.6702, 2.2506, 2.0447, 2.0895, 1.8622, 4.4234], device='cuda:1'), covar=tensor([0.0569, 0.0777, 0.0792, 0.1110, 0.0627, 0.0539, 0.0704, 0.0112], device='cuda:1'), in_proj_covar=tensor([0.0038, 0.0039, 0.0040, 0.0044, 0.0040, 0.0039, 0.0039, 0.0059], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:1') 2023-04-26 23:59:53,817 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.78 vs. limit=5.0 2023-04-26 23:59:55,729 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.78 vs. limit=5.0 2023-04-27 00:00:04,619 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=53597.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 00:00:05,761 INFO [finetune.py:976] (1/7) Epoch 10, batch 2050, loss[loss=0.1664, simple_loss=0.2336, pruned_loss=0.04963, over 4745.00 frames. ], tot_loss[loss=0.1893, simple_loss=0.2547, pruned_loss=0.06193, over 957667.18 frames. ], batch size: 54, lr: 3.75e-03, grad_scale: 16.0 2023-04-27 00:00:13,454 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.1412, 0.7434, 0.9078, 0.7684, 1.2515, 0.9546, 0.8478, 1.0004], device='cuda:1'), covar=tensor([0.1797, 0.1646, 0.2324, 0.1783, 0.1069, 0.1601, 0.1757, 0.2319], device='cuda:1'), in_proj_covar=tensor([0.0306, 0.0321, 0.0355, 0.0298, 0.0335, 0.0322, 0.0308, 0.0359], device='cuda:1'), out_proj_covar=tensor([6.4474e-05, 6.8075e-05, 7.6699e-05, 6.1534e-05, 7.0010e-05, 6.9002e-05, 6.5992e-05, 7.7012e-05], device='cuda:1') 2023-04-27 00:00:34,503 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.104e+02 1.698e+02 1.922e+02 2.404e+02 4.326e+02, threshold=3.844e+02, percent-clipped=2.0 2023-04-27 00:00:44,012 INFO [finetune.py:976] (1/7) Epoch 10, batch 2100, loss[loss=0.2538, simple_loss=0.3015, pruned_loss=0.103, over 4925.00 frames. ], tot_loss[loss=0.1887, simple_loss=0.2539, pruned_loss=0.06175, over 957438.57 frames. ], batch size: 38, lr: 3.75e-03, grad_scale: 32.0 2023-04-27 00:00:51,821 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=53660.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 00:00:55,123 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.80 vs. limit=2.0 2023-04-27 00:00:58,523 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=53671.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 00:01:15,970 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5478, 2.1320, 1.7839, 1.5519, 1.1325, 1.1489, 1.8843, 1.1978], device='cuda:1'), covar=tensor([0.1906, 0.1505, 0.1533, 0.1853, 0.2635, 0.2074, 0.1035, 0.2218], device='cuda:1'), in_proj_covar=tensor([0.0195, 0.0216, 0.0170, 0.0204, 0.0204, 0.0184, 0.0161, 0.0188], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-27 00:01:16,441 INFO [finetune.py:976] (1/7) Epoch 10, batch 2150, loss[loss=0.1967, simple_loss=0.2794, pruned_loss=0.05704, over 4932.00 frames. ], tot_loss[loss=0.1915, simple_loss=0.2574, pruned_loss=0.06277, over 957841.96 frames. ], batch size: 38, lr: 3.75e-03, grad_scale: 32.0 2023-04-27 00:01:22,404 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.3078, 1.6099, 1.6231, 1.8397, 1.6718, 1.8253, 1.7574, 1.6811], device='cuda:1'), covar=tensor([0.4837, 0.7025, 0.5785, 0.5355, 0.6727, 0.9034, 0.6732, 0.6505], device='cuda:1'), in_proj_covar=tensor([0.0326, 0.0383, 0.0316, 0.0327, 0.0341, 0.0402, 0.0362, 0.0325], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 00:01:26,024 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=53712.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 00:01:32,721 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=53723.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 00:01:38,801 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.0369, 1.8912, 1.6479, 1.5621, 2.0908, 1.6806, 2.4396, 1.4332], device='cuda:1'), covar=tensor([0.2952, 0.1511, 0.4115, 0.2639, 0.1330, 0.2004, 0.1258, 0.4392], device='cuda:1'), in_proj_covar=tensor([0.0342, 0.0345, 0.0429, 0.0360, 0.0386, 0.0381, 0.0377, 0.0419], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 00:01:41,150 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.123e+02 1.745e+02 2.178e+02 2.516e+02 3.729e+02, threshold=4.356e+02, percent-clipped=0.0 2023-04-27 00:01:49,681 INFO [finetune.py:976] (1/7) Epoch 10, batch 2200, loss[loss=0.1767, simple_loss=0.2511, pruned_loss=0.05115, over 4855.00 frames. ], tot_loss[loss=0.1939, simple_loss=0.2601, pruned_loss=0.0638, over 956956.00 frames. ], batch size: 31, lr: 3.75e-03, grad_scale: 32.0 2023-04-27 00:01:56,942 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-04-27 00:01:57,880 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=53760.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 00:02:14,205 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.1156, 1.3018, 1.1977, 1.5586, 1.3977, 1.5264, 1.2771, 2.4601], device='cuda:1'), covar=tensor([0.0613, 0.0842, 0.0871, 0.1175, 0.0667, 0.0523, 0.0766, 0.0247], device='cuda:1'), in_proj_covar=tensor([0.0038, 0.0039, 0.0040, 0.0044, 0.0040, 0.0039, 0.0039, 0.0059], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:1') 2023-04-27 00:02:19,686 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7749, 3.5046, 1.3060, 2.0475, 2.1527, 2.5440, 2.2262, 1.3302], device='cuda:1'), covar=tensor([0.1195, 0.1050, 0.1729, 0.1170, 0.0876, 0.0959, 0.1350, 0.1591], device='cuda:1'), in_proj_covar=tensor([0.0118, 0.0251, 0.0141, 0.0123, 0.0135, 0.0155, 0.0120, 0.0122], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-27 00:02:22,061 INFO [finetune.py:976] (1/7) Epoch 10, batch 2250, loss[loss=0.1573, simple_loss=0.2187, pruned_loss=0.04793, over 4750.00 frames. ], tot_loss[loss=0.1938, simple_loss=0.2607, pruned_loss=0.06347, over 956959.55 frames. ], batch size: 23, lr: 3.75e-03, grad_scale: 32.0 2023-04-27 00:02:46,629 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.157e+02 1.602e+02 1.955e+02 2.365e+02 4.712e+02, threshold=3.910e+02, percent-clipped=2.0 2023-04-27 00:02:56,344 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.82 vs. limit=5.0 2023-04-27 00:03:06,374 INFO [finetune.py:976] (1/7) Epoch 10, batch 2300, loss[loss=0.1857, simple_loss=0.2632, pruned_loss=0.05415, over 4766.00 frames. ], tot_loss[loss=0.193, simple_loss=0.2604, pruned_loss=0.06282, over 957328.80 frames. ], batch size: 28, lr: 3.75e-03, grad_scale: 32.0 2023-04-27 00:03:07,670 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0726, 2.7155, 1.0956, 1.3764, 2.1703, 1.1478, 3.6690, 1.7835], device='cuda:1'), covar=tensor([0.0690, 0.0653, 0.0778, 0.1349, 0.0510, 0.1110, 0.0269, 0.0673], device='cuda:1'), in_proj_covar=tensor([0.0052, 0.0067, 0.0049, 0.0047, 0.0051, 0.0052, 0.0078, 0.0051], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0007], device='cuda:1') 2023-04-27 00:03:18,367 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.1139, 1.7019, 1.9358, 2.4670, 2.0179, 1.5984, 1.6084, 1.8517], device='cuda:1'), covar=tensor([0.3221, 0.3328, 0.1729, 0.2267, 0.2587, 0.2695, 0.4621, 0.2473], device='cuda:1'), in_proj_covar=tensor([0.0282, 0.0250, 0.0220, 0.0315, 0.0212, 0.0227, 0.0234, 0.0185], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-27 00:04:00,270 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=3.56 vs. limit=5.0 2023-04-27 00:04:10,839 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=53897.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 00:04:11,997 INFO [finetune.py:976] (1/7) Epoch 10, batch 2350, loss[loss=0.1753, simple_loss=0.2505, pruned_loss=0.05003, over 4781.00 frames. ], tot_loss[loss=0.1913, simple_loss=0.2587, pruned_loss=0.062, over 958062.11 frames. ], batch size: 29, lr: 3.75e-03, grad_scale: 32.0 2023-04-27 00:04:57,666 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.162e+02 1.640e+02 1.984e+02 2.431e+02 6.591e+02, threshold=3.969e+02, percent-clipped=4.0 2023-04-27 00:05:08,688 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=53945.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 00:05:16,067 INFO [finetune.py:976] (1/7) Epoch 10, batch 2400, loss[loss=0.2075, simple_loss=0.2546, pruned_loss=0.08018, over 4818.00 frames. ], tot_loss[loss=0.1896, simple_loss=0.2558, pruned_loss=0.06167, over 955625.69 frames. ], batch size: 41, lr: 3.74e-03, grad_scale: 32.0 2023-04-27 00:05:23,342 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=53960.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 00:05:32,000 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=53971.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 00:05:47,717 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6935, 1.7723, 0.7087, 1.3492, 1.8742, 1.5550, 1.4382, 1.4350], device='cuda:1'), covar=tensor([0.0517, 0.0388, 0.0399, 0.0588, 0.0271, 0.0541, 0.0550, 0.0605], device='cuda:1'), in_proj_covar=tensor([0.0030, 0.0025, 0.0023, 0.0030, 0.0021, 0.0029, 0.0029, 0.0030], device='cuda:1'), out_proj_covar=tensor([0.0049, 0.0045, 0.0038, 0.0050, 0.0038, 0.0048, 0.0048, 0.0050], device='cuda:1') 2023-04-27 00:05:49,429 INFO [finetune.py:976] (1/7) Epoch 10, batch 2450, loss[loss=0.2099, simple_loss=0.2609, pruned_loss=0.07949, over 4094.00 frames. ], tot_loss[loss=0.1882, simple_loss=0.2535, pruned_loss=0.06141, over 954527.21 frames. ], batch size: 65, lr: 3.74e-03, grad_scale: 32.0 2023-04-27 00:05:56,236 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=54008.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 00:06:04,941 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=54019.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 00:06:07,941 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=54023.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 00:06:15,670 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.227e+02 1.726e+02 2.083e+02 2.546e+02 5.806e+02, threshold=4.166e+02, percent-clipped=1.0 2023-04-27 00:06:24,052 INFO [finetune.py:976] (1/7) Epoch 10, batch 2500, loss[loss=0.1959, simple_loss=0.2614, pruned_loss=0.06518, over 4127.00 frames. ], tot_loss[loss=0.188, simple_loss=0.2535, pruned_loss=0.06122, over 954837.22 frames. ], batch size: 65, lr: 3.74e-03, grad_scale: 32.0 2023-04-27 00:06:39,518 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=54071.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 00:06:57,700 INFO [finetune.py:976] (1/7) Epoch 10, batch 2550, loss[loss=0.2099, simple_loss=0.2643, pruned_loss=0.07771, over 4811.00 frames. ], tot_loss[loss=0.1913, simple_loss=0.2575, pruned_loss=0.06254, over 953328.40 frames. ], batch size: 40, lr: 3.74e-03, grad_scale: 32.0 2023-04-27 00:07:19,819 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.35 vs. limit=5.0 2023-04-27 00:07:22,641 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.109e+02 1.598e+02 1.963e+02 2.419e+02 4.837e+02, threshold=3.926e+02, percent-clipped=1.0 2023-04-27 00:07:30,643 INFO [finetune.py:976] (1/7) Epoch 10, batch 2600, loss[loss=0.1932, simple_loss=0.2674, pruned_loss=0.05951, over 4892.00 frames. ], tot_loss[loss=0.1937, simple_loss=0.2605, pruned_loss=0.06342, over 954741.94 frames. ], batch size: 43, lr: 3.74e-03, grad_scale: 32.0 2023-04-27 00:08:04,393 INFO [finetune.py:976] (1/7) Epoch 10, batch 2650, loss[loss=0.2142, simple_loss=0.2792, pruned_loss=0.07463, over 4795.00 frames. ], tot_loss[loss=0.1946, simple_loss=0.2616, pruned_loss=0.0638, over 954566.18 frames. ], batch size: 51, lr: 3.74e-03, grad_scale: 32.0 2023-04-27 00:08:39,721 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.224e+02 1.638e+02 1.968e+02 2.302e+02 4.272e+02, threshold=3.936e+02, percent-clipped=1.0 2023-04-27 00:08:52,971 INFO [finetune.py:976] (1/7) Epoch 10, batch 2700, loss[loss=0.1914, simple_loss=0.2542, pruned_loss=0.06427, over 4759.00 frames. ], tot_loss[loss=0.1933, simple_loss=0.2607, pruned_loss=0.06297, over 954553.37 frames. ], batch size: 26, lr: 3.74e-03, grad_scale: 32.0 2023-04-27 00:10:03,551 INFO [finetune.py:976] (1/7) Epoch 10, batch 2750, loss[loss=0.196, simple_loss=0.2571, pruned_loss=0.06751, over 4921.00 frames. ], tot_loss[loss=0.1901, simple_loss=0.257, pruned_loss=0.06158, over 956547.98 frames. ], batch size: 33, lr: 3.74e-03, grad_scale: 32.0 2023-04-27 00:10:49,705 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.176e+02 1.579e+02 1.855e+02 2.381e+02 3.900e+02, threshold=3.711e+02, percent-clipped=0.0 2023-04-27 00:10:58,945 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.1474, 1.8954, 2.1501, 2.4257, 2.5152, 1.8506, 1.6312, 2.2924], device='cuda:1'), covar=tensor([0.0804, 0.1063, 0.0627, 0.0539, 0.0538, 0.0954, 0.0801, 0.0503], device='cuda:1'), in_proj_covar=tensor([0.0196, 0.0203, 0.0182, 0.0175, 0.0178, 0.0189, 0.0160, 0.0184], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 00:10:58,972 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8938, 2.4816, 1.9013, 1.6689, 1.2812, 1.3881, 1.9782, 1.2517], device='cuda:1'), covar=tensor([0.1726, 0.1380, 0.1541, 0.1918, 0.2558, 0.2067, 0.1098, 0.2218], device='cuda:1'), in_proj_covar=tensor([0.0194, 0.0215, 0.0170, 0.0204, 0.0204, 0.0184, 0.0160, 0.0187], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-27 00:11:09,555 INFO [finetune.py:976] (1/7) Epoch 10, batch 2800, loss[loss=0.1709, simple_loss=0.2283, pruned_loss=0.05677, over 4815.00 frames. ], tot_loss[loss=0.1871, simple_loss=0.2533, pruned_loss=0.06045, over 955605.78 frames. ], batch size: 25, lr: 3.74e-03, grad_scale: 32.0 2023-04-27 00:11:48,128 INFO [finetune.py:976] (1/7) Epoch 10, batch 2850, loss[loss=0.1636, simple_loss=0.2369, pruned_loss=0.04518, over 4818.00 frames. ], tot_loss[loss=0.1852, simple_loss=0.2513, pruned_loss=0.05955, over 956194.57 frames. ], batch size: 38, lr: 3.74e-03, grad_scale: 32.0 2023-04-27 00:12:11,872 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.049e+02 1.761e+02 2.016e+02 2.397e+02 4.207e+02, threshold=4.033e+02, percent-clipped=3.0 2023-04-27 00:12:21,810 INFO [finetune.py:976] (1/7) Epoch 10, batch 2900, loss[loss=0.2011, simple_loss=0.2706, pruned_loss=0.06575, over 4832.00 frames. ], tot_loss[loss=0.1872, simple_loss=0.2536, pruned_loss=0.06043, over 955556.86 frames. ], batch size: 30, lr: 3.74e-03, grad_scale: 32.0 2023-04-27 00:12:55,779 INFO [finetune.py:976] (1/7) Epoch 10, batch 2950, loss[loss=0.1636, simple_loss=0.2395, pruned_loss=0.04384, over 4760.00 frames. ], tot_loss[loss=0.1903, simple_loss=0.2572, pruned_loss=0.06172, over 954675.58 frames. ], batch size: 28, lr: 3.74e-03, grad_scale: 32.0 2023-04-27 00:13:19,227 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.151e+02 1.620e+02 2.110e+02 2.463e+02 5.874e+02, threshold=4.221e+02, percent-clipped=2.0 2023-04-27 00:13:24,626 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.6805, 2.1141, 1.8830, 2.1103, 1.9429, 2.0755, 1.9754, 1.8975], device='cuda:1'), covar=tensor([0.5028, 0.7283, 0.6534, 0.5504, 0.6587, 0.8372, 0.7987, 0.7220], device='cuda:1'), in_proj_covar=tensor([0.0322, 0.0377, 0.0312, 0.0323, 0.0335, 0.0397, 0.0356, 0.0320], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 00:13:29,123 INFO [finetune.py:976] (1/7) Epoch 10, batch 3000, loss[loss=0.173, simple_loss=0.2466, pruned_loss=0.0497, over 4885.00 frames. ], tot_loss[loss=0.1916, simple_loss=0.2587, pruned_loss=0.06224, over 953523.70 frames. ], batch size: 43, lr: 3.74e-03, grad_scale: 32.0 2023-04-27 00:13:29,123 INFO [finetune.py:1001] (1/7) Computing validation loss 2023-04-27 00:13:45,198 INFO [finetune.py:1010] (1/7) Epoch 10, validation: loss=0.1531, simple_loss=0.2257, pruned_loss=0.04026, over 2265189.00 frames. 2023-04-27 00:13:45,198 INFO [finetune.py:1011] (1/7) Maximum memory allocated so far is 6391MB 2023-04-27 00:14:10,048 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4883, 1.5230, 4.1359, 3.8232, 3.6410, 3.8987, 3.8972, 3.6125], device='cuda:1'), covar=tensor([0.6791, 0.5368, 0.1044, 0.1867, 0.1099, 0.1405, 0.1170, 0.1456], device='cuda:1'), in_proj_covar=tensor([0.0309, 0.0308, 0.0406, 0.0409, 0.0349, 0.0407, 0.0315, 0.0371], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-27 00:14:32,350 INFO [finetune.py:976] (1/7) Epoch 10, batch 3050, loss[loss=0.1915, simple_loss=0.239, pruned_loss=0.07206, over 4089.00 frames. ], tot_loss[loss=0.1915, simple_loss=0.2589, pruned_loss=0.06208, over 953760.78 frames. ], batch size: 17, lr: 3.74e-03, grad_scale: 32.0 2023-04-27 00:14:50,521 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.1750, 2.5054, 1.1536, 1.3107, 1.9793, 1.2522, 3.5800, 1.7703], device='cuda:1'), covar=tensor([0.0635, 0.0606, 0.0744, 0.1361, 0.0527, 0.1036, 0.0250, 0.0665], device='cuda:1'), in_proj_covar=tensor([0.0052, 0.0067, 0.0050, 0.0047, 0.0051, 0.0053, 0.0079, 0.0051], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0011, 0.0007], device='cuda:1') 2023-04-27 00:14:57,005 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.192e+02 1.553e+02 1.895e+02 2.251e+02 3.622e+02, threshold=3.789e+02, percent-clipped=0.0 2023-04-27 00:15:05,021 INFO [finetune.py:976] (1/7) Epoch 10, batch 3100, loss[loss=0.1779, simple_loss=0.2429, pruned_loss=0.05645, over 4818.00 frames. ], tot_loss[loss=0.1901, simple_loss=0.2571, pruned_loss=0.06159, over 954590.77 frames. ], batch size: 39, lr: 3.74e-03, grad_scale: 32.0 2023-04-27 00:15:27,262 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=54664.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 00:15:34,906 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-04-27 00:16:10,660 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9569, 1.4911, 1.8299, 2.1016, 1.7858, 1.4310, 1.0376, 1.4909], device='cuda:1'), covar=tensor([0.3761, 0.4063, 0.1946, 0.2604, 0.3143, 0.3219, 0.4939, 0.2571], device='cuda:1'), in_proj_covar=tensor([0.0284, 0.0251, 0.0220, 0.0317, 0.0214, 0.0228, 0.0234, 0.0186], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-27 00:16:11,743 INFO [finetune.py:976] (1/7) Epoch 10, batch 3150, loss[loss=0.167, simple_loss=0.2215, pruned_loss=0.05626, over 4244.00 frames. ], tot_loss[loss=0.1876, simple_loss=0.254, pruned_loss=0.06059, over 954415.69 frames. ], batch size: 18, lr: 3.74e-03, grad_scale: 32.0 2023-04-27 00:16:53,623 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=54725.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 00:17:04,903 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.175e+02 1.617e+02 1.915e+02 2.289e+02 4.542e+02, threshold=3.830e+02, percent-clipped=1.0 2023-04-27 00:17:25,057 INFO [finetune.py:976] (1/7) Epoch 10, batch 3200, loss[loss=0.1393, simple_loss=0.2108, pruned_loss=0.03391, over 4819.00 frames. ], tot_loss[loss=0.184, simple_loss=0.2501, pruned_loss=0.05897, over 955592.70 frames. ], batch size: 38, lr: 3.74e-03, grad_scale: 32.0 2023-04-27 00:18:08,050 INFO [finetune.py:976] (1/7) Epoch 10, batch 3250, loss[loss=0.2111, simple_loss=0.2698, pruned_loss=0.07617, over 4933.00 frames. ], tot_loss[loss=0.1859, simple_loss=0.2513, pruned_loss=0.06019, over 953428.87 frames. ], batch size: 33, lr: 3.74e-03, grad_scale: 32.0 2023-04-27 00:18:33,642 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.070e+02 1.703e+02 2.055e+02 2.409e+02 4.051e+02, threshold=4.111e+02, percent-clipped=1.0 2023-04-27 00:18:42,051 INFO [finetune.py:976] (1/7) Epoch 10, batch 3300, loss[loss=0.2136, simple_loss=0.2728, pruned_loss=0.07715, over 4928.00 frames. ], tot_loss[loss=0.1897, simple_loss=0.2556, pruned_loss=0.0619, over 952307.92 frames. ], batch size: 33, lr: 3.74e-03, grad_scale: 32.0 2023-04-27 00:18:42,797 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4099, 2.3532, 1.9305, 2.0436, 2.5039, 2.1065, 3.3114, 1.8193], device='cuda:1'), covar=tensor([0.4373, 0.2659, 0.5498, 0.3993, 0.2168, 0.2944, 0.1810, 0.4806], device='cuda:1'), in_proj_covar=tensor([0.0343, 0.0347, 0.0431, 0.0360, 0.0386, 0.0382, 0.0378, 0.0418], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 00:18:42,834 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7663, 1.6471, 2.0483, 2.2605, 1.9020, 1.7096, 1.8433, 1.8581], device='cuda:1'), covar=tensor([0.6232, 0.8742, 0.9383, 0.8310, 0.7449, 1.1207, 1.1689, 1.0700], device='cuda:1'), in_proj_covar=tensor([0.0407, 0.0412, 0.0494, 0.0514, 0.0435, 0.0453, 0.0464, 0.0462], device='cuda:1'), out_proj_covar=tensor([9.9175e-05, 1.0212e-04, 1.1154e-04, 1.2221e-04, 1.0548e-04, 1.0956e-04, 1.1130e-04, 1.1125e-04], device='cuda:1') 2023-04-27 00:18:46,469 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=3.58 vs. limit=5.0 2023-04-27 00:18:50,594 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=54862.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 00:19:12,763 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.71 vs. limit=5.0 2023-04-27 00:19:15,377 INFO [finetune.py:976] (1/7) Epoch 10, batch 3350, loss[loss=0.2223, simple_loss=0.2804, pruned_loss=0.08209, over 4218.00 frames. ], tot_loss[loss=0.1918, simple_loss=0.2573, pruned_loss=0.06321, over 949525.00 frames. ], batch size: 66, lr: 3.74e-03, grad_scale: 32.0 2023-04-27 00:19:30,560 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=54923.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 00:19:39,395 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=3.61 vs. limit=5.0 2023-04-27 00:19:39,813 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.272e+02 1.802e+02 2.053e+02 2.510e+02 4.625e+02, threshold=4.107e+02, percent-clipped=1.0 2023-04-27 00:19:47,702 INFO [finetune.py:976] (1/7) Epoch 10, batch 3400, loss[loss=0.2015, simple_loss=0.2755, pruned_loss=0.06372, over 4763.00 frames. ], tot_loss[loss=0.193, simple_loss=0.259, pruned_loss=0.06351, over 951429.71 frames. ], batch size: 26, lr: 3.74e-03, grad_scale: 32.0 2023-04-27 00:19:56,214 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0611, 1.8374, 2.0477, 2.3170, 2.4501, 1.8600, 1.6382, 2.1126], device='cuda:1'), covar=tensor([0.0945, 0.1040, 0.0697, 0.0711, 0.0655, 0.0959, 0.0865, 0.0634], device='cuda:1'), in_proj_covar=tensor([0.0198, 0.0206, 0.0185, 0.0178, 0.0181, 0.0191, 0.0162, 0.0186], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 00:20:20,501 INFO [finetune.py:976] (1/7) Epoch 10, batch 3450, loss[loss=0.1526, simple_loss=0.2193, pruned_loss=0.04289, over 4773.00 frames. ], tot_loss[loss=0.1942, simple_loss=0.2604, pruned_loss=0.06397, over 952706.04 frames. ], batch size: 26, lr: 3.74e-03, grad_scale: 32.0 2023-04-27 00:20:33,869 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=55020.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 00:20:45,488 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.144e+02 1.637e+02 2.017e+02 2.373e+02 5.913e+02, threshold=4.034e+02, percent-clipped=2.0 2023-04-27 00:20:53,437 INFO [finetune.py:976] (1/7) Epoch 10, batch 3500, loss[loss=0.1775, simple_loss=0.2454, pruned_loss=0.05477, over 4768.00 frames. ], tot_loss[loss=0.191, simple_loss=0.2572, pruned_loss=0.06247, over 954171.90 frames. ], batch size: 27, lr: 3.74e-03, grad_scale: 32.0 2023-04-27 00:21:37,025 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-04-27 00:21:37,465 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7780, 3.6422, 0.8627, 1.9074, 2.2542, 2.6114, 2.0690, 0.9946], device='cuda:1'), covar=tensor([0.1284, 0.0870, 0.2048, 0.1329, 0.0907, 0.0993, 0.1426, 0.1979], device='cuda:1'), in_proj_covar=tensor([0.0119, 0.0254, 0.0142, 0.0124, 0.0137, 0.0156, 0.0120, 0.0123], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-27 00:21:46,489 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-04-27 00:21:46,851 INFO [finetune.py:976] (1/7) Epoch 10, batch 3550, loss[loss=0.2127, simple_loss=0.2644, pruned_loss=0.08047, over 4750.00 frames. ], tot_loss[loss=0.1879, simple_loss=0.2537, pruned_loss=0.06107, over 953948.72 frames. ], batch size: 23, lr: 3.74e-03, grad_scale: 32.0 2023-04-27 00:22:27,074 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 8.344e+01 1.599e+02 1.898e+02 2.251e+02 4.553e+02, threshold=3.796e+02, percent-clipped=2.0 2023-04-27 00:22:30,128 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.96 vs. limit=2.0 2023-04-27 00:22:35,878 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=55148.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 00:22:36,396 INFO [finetune.py:976] (1/7) Epoch 10, batch 3600, loss[loss=0.1376, simple_loss=0.2177, pruned_loss=0.02875, over 4750.00 frames. ], tot_loss[loss=0.1862, simple_loss=0.2516, pruned_loss=0.06044, over 955531.95 frames. ], batch size: 27, lr: 3.74e-03, grad_scale: 32.0 2023-04-27 00:23:26,775 INFO [finetune.py:976] (1/7) Epoch 10, batch 3650, loss[loss=0.2977, simple_loss=0.3399, pruned_loss=0.1278, over 4814.00 frames. ], tot_loss[loss=0.1886, simple_loss=0.254, pruned_loss=0.06161, over 954625.59 frames. ], batch size: 38, lr: 3.74e-03, grad_scale: 32.0 2023-04-27 00:23:27,826 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-04-27 00:23:33,114 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=55209.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 00:23:34,368 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=55211.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 00:23:39,095 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=55218.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 00:23:41,013 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-04-27 00:23:50,726 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.180e+02 1.786e+02 2.133e+02 2.599e+02 5.918e+02, threshold=4.267e+02, percent-clipped=5.0 2023-04-27 00:23:56,640 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.88 vs. limit=2.0 2023-04-27 00:23:57,689 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6148, 1.1581, 1.6954, 2.1148, 1.7430, 1.6261, 1.7109, 1.6595], device='cuda:1'), covar=tensor([0.5552, 0.7871, 0.8165, 0.7313, 0.6837, 0.8768, 0.9252, 0.8793], device='cuda:1'), in_proj_covar=tensor([0.0410, 0.0416, 0.0501, 0.0518, 0.0439, 0.0458, 0.0469, 0.0466], device='cuda:1'), out_proj_covar=tensor([9.9820e-05, 1.0307e-04, 1.1312e-04, 1.2318e-04, 1.0653e-04, 1.1056e-04, 1.1238e-04, 1.1214e-04], device='cuda:1') 2023-04-27 00:24:00,573 INFO [finetune.py:976] (1/7) Epoch 10, batch 3700, loss[loss=0.1446, simple_loss=0.2087, pruned_loss=0.04021, over 4746.00 frames. ], tot_loss[loss=0.1917, simple_loss=0.258, pruned_loss=0.06268, over 954176.07 frames. ], batch size: 23, lr: 3.73e-03, grad_scale: 32.0 2023-04-27 00:24:05,600 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2023-04-27 00:24:15,032 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=55272.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 00:24:18,793 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.56 vs. limit=2.0 2023-04-27 00:24:25,921 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.2647, 1.3228, 1.4265, 1.5640, 1.6789, 1.2354, 0.9140, 1.4446], device='cuda:1'), covar=tensor([0.0960, 0.1326, 0.0872, 0.0691, 0.0684, 0.0973, 0.0994, 0.0666], device='cuda:1'), in_proj_covar=tensor([0.0196, 0.0204, 0.0182, 0.0175, 0.0178, 0.0189, 0.0160, 0.0184], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 00:24:33,930 INFO [finetune.py:976] (1/7) Epoch 10, batch 3750, loss[loss=0.2209, simple_loss=0.2788, pruned_loss=0.08154, over 4888.00 frames. ], tot_loss[loss=0.1931, simple_loss=0.26, pruned_loss=0.06309, over 955386.74 frames. ], batch size: 43, lr: 3.73e-03, grad_scale: 32.0 2023-04-27 00:24:47,368 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=55320.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 00:24:57,144 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.148e+02 1.679e+02 1.930e+02 2.224e+02 3.496e+02, threshold=3.860e+02, percent-clipped=0.0 2023-04-27 00:25:07,102 INFO [finetune.py:976] (1/7) Epoch 10, batch 3800, loss[loss=0.2079, simple_loss=0.2722, pruned_loss=0.07177, over 4910.00 frames. ], tot_loss[loss=0.1938, simple_loss=0.2609, pruned_loss=0.06335, over 956002.37 frames. ], batch size: 43, lr: 3.73e-03, grad_scale: 32.0 2023-04-27 00:25:19,728 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=55368.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 00:25:28,906 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=3.66 vs. limit=5.0 2023-04-27 00:25:40,048 INFO [finetune.py:976] (1/7) Epoch 10, batch 3850, loss[loss=0.1689, simple_loss=0.2481, pruned_loss=0.0449, over 4808.00 frames. ], tot_loss[loss=0.1916, simple_loss=0.2589, pruned_loss=0.06221, over 955483.90 frames. ], batch size: 45, lr: 3.73e-03, grad_scale: 32.0 2023-04-27 00:25:48,395 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7137, 1.7798, 1.0324, 1.3482, 1.7513, 1.5762, 1.5119, 1.4910], device='cuda:1'), covar=tensor([0.0459, 0.0340, 0.0350, 0.0549, 0.0277, 0.0517, 0.0480, 0.0532], device='cuda:1'), in_proj_covar=tensor([0.0030, 0.0026, 0.0023, 0.0030, 0.0021, 0.0029, 0.0029, 0.0030], device='cuda:1'), out_proj_covar=tensor([0.0049, 0.0045, 0.0038, 0.0050, 0.0038, 0.0048, 0.0048, 0.0050], device='cuda:1') 2023-04-27 00:25:51,535 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-04-27 00:26:04,612 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.147e+02 1.602e+02 1.915e+02 2.260e+02 3.579e+02, threshold=3.830e+02, percent-clipped=0.0 2023-04-27 00:26:12,958 INFO [finetune.py:976] (1/7) Epoch 10, batch 3900, loss[loss=0.1825, simple_loss=0.2538, pruned_loss=0.05558, over 4872.00 frames. ], tot_loss[loss=0.1889, simple_loss=0.256, pruned_loss=0.06093, over 957481.26 frames. ], batch size: 34, lr: 3.73e-03, grad_scale: 32.0 2023-04-27 00:26:14,117 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([3.3390, 3.2243, 2.5484, 3.8734, 3.3830, 3.3475, 1.3572, 3.3156], device='cuda:1'), covar=tensor([0.1740, 0.1401, 0.2898, 0.2303, 0.2552, 0.1888, 0.5697, 0.2321], device='cuda:1'), in_proj_covar=tensor([0.0244, 0.0216, 0.0249, 0.0306, 0.0301, 0.0250, 0.0270, 0.0270], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 00:27:03,046 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.6328, 2.1722, 2.6167, 3.1733, 2.5043, 1.9615, 2.0114, 2.4334], device='cuda:1'), covar=tensor([0.3626, 0.3534, 0.1724, 0.2875, 0.3197, 0.2871, 0.4302, 0.2362], device='cuda:1'), in_proj_covar=tensor([0.0285, 0.0250, 0.0220, 0.0318, 0.0215, 0.0228, 0.0235, 0.0186], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-27 00:27:07,759 INFO [finetune.py:976] (1/7) Epoch 10, batch 3950, loss[loss=0.1556, simple_loss=0.2123, pruned_loss=0.04945, over 4260.00 frames. ], tot_loss[loss=0.1855, simple_loss=0.2526, pruned_loss=0.05926, over 957670.03 frames. ], batch size: 18, lr: 3.73e-03, grad_scale: 32.0 2023-04-27 00:27:15,666 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=55504.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 00:27:36,355 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=55518.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 00:27:47,672 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.461e+01 1.659e+02 2.039e+02 2.496e+02 7.335e+02, threshold=4.077e+02, percent-clipped=3.0 2023-04-27 00:27:56,582 INFO [finetune.py:976] (1/7) Epoch 10, batch 4000, loss[loss=0.1735, simple_loss=0.2636, pruned_loss=0.04173, over 4815.00 frames. ], tot_loss[loss=0.1866, simple_loss=0.2529, pruned_loss=0.06011, over 958090.95 frames. ], batch size: 41, lr: 3.73e-03, grad_scale: 32.0 2023-04-27 00:28:09,058 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=55566.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 00:28:12,129 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=55567.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 00:28:43,369 INFO [finetune.py:976] (1/7) Epoch 10, batch 4050, loss[loss=0.1331, simple_loss=0.2045, pruned_loss=0.03089, over 4712.00 frames. ], tot_loss[loss=0.1905, simple_loss=0.257, pruned_loss=0.06202, over 955441.74 frames. ], batch size: 23, lr: 3.73e-03, grad_scale: 32.0 2023-04-27 00:29:09,491 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.009e+02 1.669e+02 1.991e+02 2.505e+02 4.319e+02, threshold=3.981e+02, percent-clipped=1.0 2023-04-27 00:29:16,837 INFO [finetune.py:976] (1/7) Epoch 10, batch 4100, loss[loss=0.2059, simple_loss=0.2804, pruned_loss=0.06566, over 4901.00 frames. ], tot_loss[loss=0.1919, simple_loss=0.2592, pruned_loss=0.06227, over 956452.16 frames. ], batch size: 35, lr: 3.73e-03, grad_scale: 32.0 2023-04-27 00:29:49,694 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-04-27 00:29:50,669 INFO [finetune.py:976] (1/7) Epoch 10, batch 4150, loss[loss=0.1945, simple_loss=0.2657, pruned_loss=0.06166, over 4921.00 frames. ], tot_loss[loss=0.1938, simple_loss=0.2611, pruned_loss=0.06323, over 957074.45 frames. ], batch size: 33, lr: 3.73e-03, grad_scale: 32.0 2023-04-27 00:30:16,063 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.104e+02 1.755e+02 2.031e+02 2.318e+02 3.746e+02, threshold=4.063e+02, percent-clipped=0.0 2023-04-27 00:30:23,238 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=55748.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 00:30:23,766 INFO [finetune.py:976] (1/7) Epoch 10, batch 4200, loss[loss=0.2488, simple_loss=0.2988, pruned_loss=0.09944, over 4223.00 frames. ], tot_loss[loss=0.1946, simple_loss=0.2618, pruned_loss=0.06369, over 953428.23 frames. ], batch size: 65, lr: 3.73e-03, grad_scale: 32.0 2023-04-27 00:30:57,481 INFO [finetune.py:976] (1/7) Epoch 10, batch 4250, loss[loss=0.1592, simple_loss=0.2343, pruned_loss=0.04208, over 4786.00 frames. ], tot_loss[loss=0.1929, simple_loss=0.2597, pruned_loss=0.06303, over 954080.26 frames. ], batch size: 51, lr: 3.73e-03, grad_scale: 32.0 2023-04-27 00:30:58,300 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6161, 1.2737, 4.4733, 4.1758, 3.8901, 4.1879, 4.1093, 3.9386], device='cuda:1'), covar=tensor([0.6971, 0.6262, 0.1013, 0.1819, 0.1096, 0.1397, 0.1507, 0.1579], device='cuda:1'), in_proj_covar=tensor([0.0311, 0.0310, 0.0407, 0.0411, 0.0352, 0.0408, 0.0317, 0.0373], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-27 00:31:00,762 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=55804.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 00:31:04,329 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=55809.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 00:31:23,548 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.518e+01 1.564e+02 1.887e+02 2.474e+02 5.983e+02, threshold=3.773e+02, percent-clipped=1.0 2023-04-27 00:31:30,793 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=55848.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 00:31:31,265 INFO [finetune.py:976] (1/7) Epoch 10, batch 4300, loss[loss=0.1786, simple_loss=0.2402, pruned_loss=0.05848, over 4724.00 frames. ], tot_loss[loss=0.1905, simple_loss=0.2564, pruned_loss=0.06227, over 954307.55 frames. ], batch size: 23, lr: 3.73e-03, grad_scale: 32.0 2023-04-27 00:31:33,046 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=55852.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 00:31:48,275 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=55867.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 00:32:31,697 INFO [finetune.py:976] (1/7) Epoch 10, batch 4350, loss[loss=0.1656, simple_loss=0.2215, pruned_loss=0.05488, over 4092.00 frames. ], tot_loss[loss=0.1876, simple_loss=0.2535, pruned_loss=0.06089, over 952989.77 frames. ], batch size: 17, lr: 3.73e-03, grad_scale: 32.0 2023-04-27 00:32:43,366 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=55909.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 00:32:46,922 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=55915.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 00:33:02,710 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.042e+02 1.769e+02 2.030e+02 2.376e+02 3.891e+02, threshold=4.060e+02, percent-clipped=2.0 2023-04-27 00:33:10,568 INFO [finetune.py:976] (1/7) Epoch 10, batch 4400, loss[loss=0.1735, simple_loss=0.248, pruned_loss=0.0495, over 4862.00 frames. ], tot_loss[loss=0.1883, simple_loss=0.2547, pruned_loss=0.06101, over 953741.35 frames. ], batch size: 44, lr: 3.73e-03, grad_scale: 32.0 2023-04-27 00:34:18,723 INFO [finetune.py:976] (1/7) Epoch 10, batch 4450, loss[loss=0.1963, simple_loss=0.2691, pruned_loss=0.06175, over 4792.00 frames. ], tot_loss[loss=0.1902, simple_loss=0.2574, pruned_loss=0.06145, over 955549.95 frames. ], batch size: 45, lr: 3.73e-03, grad_scale: 32.0 2023-04-27 00:34:28,840 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.3563, 3.2781, 1.1323, 1.7186, 1.6605, 2.4078, 1.8418, 1.0885], device='cuda:1'), covar=tensor([0.1379, 0.0829, 0.1672, 0.1312, 0.1132, 0.0899, 0.1394, 0.1998], device='cuda:1'), in_proj_covar=tensor([0.0118, 0.0249, 0.0141, 0.0123, 0.0135, 0.0154, 0.0118, 0.0122], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-27 00:34:31,968 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-04-27 00:34:57,415 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.139e+02 1.646e+02 1.926e+02 2.335e+02 3.268e+02, threshold=3.851e+02, percent-clipped=0.0 2023-04-27 00:35:05,273 INFO [finetune.py:976] (1/7) Epoch 10, batch 4500, loss[loss=0.2045, simple_loss=0.2737, pruned_loss=0.06766, over 4719.00 frames. ], tot_loss[loss=0.1923, simple_loss=0.2599, pruned_loss=0.06234, over 956853.35 frames. ], batch size: 59, lr: 3.73e-03, grad_scale: 32.0 2023-04-27 00:35:38,736 INFO [finetune.py:976] (1/7) Epoch 10, batch 4550, loss[loss=0.2035, simple_loss=0.2659, pruned_loss=0.07052, over 4255.00 frames. ], tot_loss[loss=0.1953, simple_loss=0.2623, pruned_loss=0.06413, over 956813.59 frames. ], batch size: 65, lr: 3.73e-03, grad_scale: 32.0 2023-04-27 00:35:41,832 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=56104.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 00:35:48,467 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.2175, 3.3169, 2.5488, 2.7179, 2.2274, 2.7066, 2.6804, 1.9214], device='cuda:1'), covar=tensor([0.2483, 0.1052, 0.0850, 0.1467, 0.3183, 0.1171, 0.2473, 0.3256], device='cuda:1'), in_proj_covar=tensor([0.0300, 0.0321, 0.0232, 0.0293, 0.0320, 0.0274, 0.0260, 0.0283], device='cuda:1'), out_proj_covar=tensor([1.2117e-04, 1.2935e-04, 9.3162e-05, 1.1750e-04, 1.3082e-04, 1.1009e-04, 1.0583e-04, 1.1376e-04], device='cuda:1') 2023-04-27 00:35:58,682 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7322, 1.2535, 1.7835, 2.1716, 1.8102, 1.7024, 1.7455, 1.7793], device='cuda:1'), covar=tensor([0.5748, 0.8128, 0.7882, 0.8102, 0.7570, 0.9690, 1.0034, 0.8714], device='cuda:1'), in_proj_covar=tensor([0.0411, 0.0414, 0.0499, 0.0518, 0.0439, 0.0459, 0.0470, 0.0467], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 00:36:03,340 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.622e+01 1.743e+02 2.073e+02 2.397e+02 5.279e+02, threshold=4.145e+02, percent-clipped=1.0 2023-04-27 00:36:09,777 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=56145.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 00:36:12,138 INFO [finetune.py:976] (1/7) Epoch 10, batch 4600, loss[loss=0.1612, simple_loss=0.2187, pruned_loss=0.05185, over 4914.00 frames. ], tot_loss[loss=0.1942, simple_loss=0.2608, pruned_loss=0.06376, over 956498.97 frames. ], batch size: 38, lr: 3.73e-03, grad_scale: 32.0 2023-04-27 00:36:45,855 INFO [finetune.py:976] (1/7) Epoch 10, batch 4650, loss[loss=0.1681, simple_loss=0.2292, pruned_loss=0.05355, over 4754.00 frames. ], tot_loss[loss=0.1916, simple_loss=0.2575, pruned_loss=0.06286, over 953590.56 frames. ], batch size: 27, lr: 3.73e-03, grad_scale: 32.0 2023-04-27 00:36:49,606 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=56204.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 00:36:50,842 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=56206.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 00:36:55,067 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([4.0790, 3.9183, 3.0113, 4.6379, 4.1057, 4.0285, 2.0292, 3.8477], device='cuda:1'), covar=tensor([0.1778, 0.1023, 0.2861, 0.1411, 0.2454, 0.1930, 0.5309, 0.2310], device='cuda:1'), in_proj_covar=tensor([0.0241, 0.0214, 0.0246, 0.0302, 0.0297, 0.0249, 0.0266, 0.0267], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 00:36:55,160 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9491, 1.6822, 2.1647, 2.4403, 2.1097, 1.9242, 2.0686, 1.9927], device='cuda:1'), covar=tensor([0.5643, 0.7201, 0.7424, 0.7213, 0.6689, 0.9085, 0.8792, 0.8223], device='cuda:1'), in_proj_covar=tensor([0.0409, 0.0412, 0.0496, 0.0516, 0.0437, 0.0456, 0.0467, 0.0464], device='cuda:1'), out_proj_covar=tensor([9.9742e-05, 1.0207e-04, 1.1191e-04, 1.2258e-04, 1.0588e-04, 1.1021e-04, 1.1196e-04, 1.1165e-04], device='cuda:1') 2023-04-27 00:37:05,721 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 2023-04-27 00:37:16,889 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 8.791e+01 1.513e+02 1.802e+02 2.222e+02 6.871e+02, threshold=3.603e+02, percent-clipped=2.0 2023-04-27 00:37:18,268 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8607, 1.4075, 1.6973, 1.6027, 1.6036, 1.3345, 0.7287, 1.3249], device='cuda:1'), covar=tensor([0.3473, 0.3581, 0.1708, 0.2368, 0.2701, 0.2736, 0.4521, 0.2358], device='cuda:1'), in_proj_covar=tensor([0.0283, 0.0249, 0.0220, 0.0315, 0.0213, 0.0227, 0.0233, 0.0184], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-27 00:37:24,882 INFO [finetune.py:976] (1/7) Epoch 10, batch 4700, loss[loss=0.1686, simple_loss=0.2362, pruned_loss=0.05054, over 4901.00 frames. ], tot_loss[loss=0.1889, simple_loss=0.2544, pruned_loss=0.06165, over 954495.63 frames. ], batch size: 32, lr: 3.73e-03, grad_scale: 32.0 2023-04-27 00:37:44,918 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-04-27 00:37:49,606 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.4720, 1.6793, 1.8410, 2.0050, 1.8548, 1.9316, 1.9114, 1.8936], device='cuda:1'), covar=tensor([0.4722, 0.6897, 0.5846, 0.5163, 0.6508, 0.8820, 0.6715, 0.6128], device='cuda:1'), in_proj_covar=tensor([0.0326, 0.0378, 0.0313, 0.0325, 0.0337, 0.0399, 0.0358, 0.0323], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 00:37:57,893 INFO [finetune.py:976] (1/7) Epoch 10, batch 4750, loss[loss=0.1702, simple_loss=0.2334, pruned_loss=0.05351, over 4823.00 frames. ], tot_loss[loss=0.1868, simple_loss=0.252, pruned_loss=0.06081, over 955205.27 frames. ], batch size: 30, lr: 3.73e-03, grad_scale: 32.0 2023-04-27 00:38:23,668 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.024e+02 1.579e+02 1.869e+02 2.281e+02 6.054e+02, threshold=3.738e+02, percent-clipped=4.0 2023-04-27 00:38:31,957 INFO [finetune.py:976] (1/7) Epoch 10, batch 4800, loss[loss=0.2366, simple_loss=0.3006, pruned_loss=0.08629, over 4907.00 frames. ], tot_loss[loss=0.1892, simple_loss=0.2551, pruned_loss=0.06168, over 955812.74 frames. ], batch size: 36, lr: 3.73e-03, grad_scale: 32.0 2023-04-27 00:38:32,666 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4778, 3.2439, 0.8241, 1.6677, 1.7338, 2.3075, 1.8571, 0.9176], device='cuda:1'), covar=tensor([0.1375, 0.1060, 0.2070, 0.1414, 0.1140, 0.1082, 0.1433, 0.2072], device='cuda:1'), in_proj_covar=tensor([0.0118, 0.0251, 0.0142, 0.0123, 0.0135, 0.0154, 0.0119, 0.0122], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-27 00:39:33,342 INFO [finetune.py:976] (1/7) Epoch 10, batch 4850, loss[loss=0.1962, simple_loss=0.2705, pruned_loss=0.06095, over 4887.00 frames. ], tot_loss[loss=0.1921, simple_loss=0.2587, pruned_loss=0.06277, over 956008.20 frames. ], batch size: 32, lr: 3.73e-03, grad_scale: 32.0 2023-04-27 00:39:43,323 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=56404.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 00:40:14,075 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.161e+02 1.686e+02 2.043e+02 2.423e+02 5.639e+02, threshold=4.086e+02, percent-clipped=3.0 2023-04-27 00:40:22,379 INFO [finetune.py:976] (1/7) Epoch 10, batch 4900, loss[loss=0.184, simple_loss=0.2525, pruned_loss=0.05772, over 4828.00 frames. ], tot_loss[loss=0.1945, simple_loss=0.2612, pruned_loss=0.06393, over 953790.73 frames. ], batch size: 49, lr: 3.73e-03, grad_scale: 32.0 2023-04-27 00:40:24,262 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=56452.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 00:40:56,301 INFO [finetune.py:976] (1/7) Epoch 10, batch 4950, loss[loss=0.2008, simple_loss=0.2426, pruned_loss=0.07946, over 4391.00 frames. ], tot_loss[loss=0.1951, simple_loss=0.2621, pruned_loss=0.06411, over 954821.19 frames. ], batch size: 19, lr: 3.72e-03, grad_scale: 32.0 2023-04-27 00:40:57,602 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=56501.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 00:40:59,499 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=56504.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 00:41:12,940 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=56523.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 00:41:21,844 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.129e+02 1.631e+02 2.001e+02 2.395e+02 7.277e+02, threshold=4.002e+02, percent-clipped=1.0 2023-04-27 00:41:29,726 INFO [finetune.py:976] (1/7) Epoch 10, batch 5000, loss[loss=0.1799, simple_loss=0.2431, pruned_loss=0.05831, over 4090.00 frames. ], tot_loss[loss=0.1926, simple_loss=0.2598, pruned_loss=0.06274, over 954754.54 frames. ], batch size: 17, lr: 3.72e-03, grad_scale: 32.0 2023-04-27 00:41:32,067 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=56552.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 00:41:53,721 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=56584.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 00:42:03,263 INFO [finetune.py:976] (1/7) Epoch 10, batch 5050, loss[loss=0.1572, simple_loss=0.2383, pruned_loss=0.03807, over 4917.00 frames. ], tot_loss[loss=0.1898, simple_loss=0.2561, pruned_loss=0.0617, over 953866.94 frames. ], batch size: 37, lr: 3.72e-03, grad_scale: 32.0 2023-04-27 00:42:56,296 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.103e+02 1.627e+02 1.925e+02 2.295e+02 4.200e+02, threshold=3.850e+02, percent-clipped=1.0 2023-04-27 00:43:09,522 INFO [finetune.py:976] (1/7) Epoch 10, batch 5100, loss[loss=0.1472, simple_loss=0.2096, pruned_loss=0.04239, over 4174.00 frames. ], tot_loss[loss=0.1861, simple_loss=0.2526, pruned_loss=0.05981, over 954334.72 frames. ], batch size: 65, lr: 3.72e-03, grad_scale: 32.0 2023-04-27 00:43:59,161 INFO [finetune.py:976] (1/7) Epoch 10, batch 5150, loss[loss=0.2069, simple_loss=0.2835, pruned_loss=0.06516, over 4898.00 frames. ], tot_loss[loss=0.1865, simple_loss=0.2531, pruned_loss=0.05995, over 955501.87 frames. ], batch size: 35, lr: 3.72e-03, grad_scale: 32.0 2023-04-27 00:44:25,424 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.088e+02 1.696e+02 2.008e+02 2.313e+02 3.981e+02, threshold=4.015e+02, percent-clipped=1.0 2023-04-27 00:44:33,209 INFO [finetune.py:976] (1/7) Epoch 10, batch 5200, loss[loss=0.1477, simple_loss=0.2365, pruned_loss=0.02944, over 4937.00 frames. ], tot_loss[loss=0.19, simple_loss=0.2572, pruned_loss=0.06141, over 953711.87 frames. ], batch size: 33, lr: 3.72e-03, grad_scale: 32.0 2023-04-27 00:45:00,951 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.1353, 0.8192, 0.9673, 0.7503, 1.2601, 0.9797, 0.8579, 0.9875], device='cuda:1'), covar=tensor([0.1444, 0.1254, 0.1814, 0.1475, 0.0931, 0.1203, 0.1499, 0.1867], device='cuda:1'), in_proj_covar=tensor([0.0304, 0.0320, 0.0353, 0.0297, 0.0335, 0.0319, 0.0305, 0.0360], device='cuda:1'), out_proj_covar=tensor([6.3977e-05, 6.7645e-05, 7.6100e-05, 6.1213e-05, 7.0042e-05, 6.8228e-05, 6.5375e-05, 7.7369e-05], device='cuda:1') 2023-04-27 00:45:18,108 INFO [finetune.py:976] (1/7) Epoch 10, batch 5250, loss[loss=0.2358, simple_loss=0.3023, pruned_loss=0.08462, over 4905.00 frames. ], tot_loss[loss=0.1922, simple_loss=0.26, pruned_loss=0.06225, over 953429.88 frames. ], batch size: 36, lr: 3.72e-03, grad_scale: 32.0 2023-04-27 00:45:20,019 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=56801.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 00:45:21,229 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=56803.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 00:45:37,893 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7309, 2.0222, 1.3849, 1.5114, 2.0227, 1.6594, 1.5474, 1.5575], device='cuda:1'), covar=tensor([0.0555, 0.0270, 0.0321, 0.0505, 0.0301, 0.0552, 0.0498, 0.0559], device='cuda:1'), in_proj_covar=tensor([0.0030, 0.0025, 0.0023, 0.0030, 0.0020, 0.0029, 0.0029, 0.0030], device='cuda:1'), out_proj_covar=tensor([0.0049, 0.0045, 0.0038, 0.0050, 0.0037, 0.0048, 0.0048, 0.0050], device='cuda:1') 2023-04-27 00:45:37,906 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9380, 2.3385, 1.9908, 2.1835, 1.6637, 1.9741, 2.1397, 1.6822], device='cuda:1'), covar=tensor([0.1600, 0.0940, 0.0758, 0.0977, 0.2752, 0.0957, 0.1403, 0.2085], device='cuda:1'), in_proj_covar=tensor([0.0296, 0.0315, 0.0228, 0.0287, 0.0315, 0.0271, 0.0255, 0.0278], device='cuda:1'), out_proj_covar=tensor([1.1965e-04, 1.2686e-04, 9.1354e-05, 1.1495e-04, 1.2881e-04, 1.0887e-04, 1.0380e-04, 1.1158e-04], device='cuda:1') 2023-04-27 00:45:43,652 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.105e+02 1.715e+02 1.995e+02 2.714e+02 4.040e+02, threshold=3.989e+02, percent-clipped=1.0 2023-04-27 00:45:51,393 INFO [finetune.py:976] (1/7) Epoch 10, batch 5300, loss[loss=0.1975, simple_loss=0.2751, pruned_loss=0.05989, over 4832.00 frames. ], tot_loss[loss=0.1941, simple_loss=0.2623, pruned_loss=0.06298, over 953113.96 frames. ], batch size: 47, lr: 3.72e-03, grad_scale: 32.0 2023-04-27 00:45:51,452 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=56849.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 00:45:51,537 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.7970, 1.9130, 1.9630, 2.0760, 1.7942, 1.9885, 2.1198, 2.0350], device='cuda:1'), covar=tensor([0.4830, 0.7785, 0.5823, 0.5628, 0.6940, 0.9161, 0.7010, 0.6474], device='cuda:1'), in_proj_covar=tensor([0.0328, 0.0381, 0.0315, 0.0326, 0.0338, 0.0401, 0.0359, 0.0322], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 00:46:01,124 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=56864.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 00:46:10,803 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-04-27 00:46:11,155 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=56879.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 00:46:12,494 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.80 vs. limit=2.0 2023-04-27 00:46:25,223 INFO [finetune.py:976] (1/7) Epoch 10, batch 5350, loss[loss=0.1621, simple_loss=0.2415, pruned_loss=0.04138, over 4806.00 frames. ], tot_loss[loss=0.1924, simple_loss=0.2609, pruned_loss=0.06198, over 952053.31 frames. ], batch size: 39, lr: 3.72e-03, grad_scale: 32.0 2023-04-27 00:46:31,345 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=56908.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 00:46:50,603 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.133e+02 1.631e+02 1.870e+02 2.324e+02 4.447e+02, threshold=3.741e+02, percent-clipped=2.0 2023-04-27 00:46:58,344 INFO [finetune.py:976] (1/7) Epoch 10, batch 5400, loss[loss=0.1791, simple_loss=0.2437, pruned_loss=0.05727, over 4843.00 frames. ], tot_loss[loss=0.1896, simple_loss=0.2571, pruned_loss=0.061, over 951551.46 frames. ], batch size: 49, lr: 3.72e-03, grad_scale: 32.0 2023-04-27 00:46:59,080 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-04-27 00:47:11,574 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=56969.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 00:47:20,891 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([3.7733, 3.5681, 2.6592, 4.3114, 3.7433, 3.7407, 1.7229, 3.7122], device='cuda:1'), covar=tensor([0.1589, 0.1284, 0.3453, 0.1616, 0.3467, 0.1721, 0.5496, 0.2291], device='cuda:1'), in_proj_covar=tensor([0.0241, 0.0214, 0.0248, 0.0302, 0.0298, 0.0248, 0.0266, 0.0270], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 00:47:28,573 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=56993.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 00:47:29,784 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=56995.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 00:47:32,098 INFO [finetune.py:976] (1/7) Epoch 10, batch 5450, loss[loss=0.2082, simple_loss=0.2615, pruned_loss=0.07748, over 4934.00 frames. ], tot_loss[loss=0.1867, simple_loss=0.2535, pruned_loss=0.05992, over 952710.46 frames. ], batch size: 38, lr: 3.72e-03, grad_scale: 32.0 2023-04-27 00:48:18,180 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.140e+02 1.695e+02 1.974e+02 2.503e+02 6.150e+02, threshold=3.947e+02, percent-clipped=1.0 2023-04-27 00:48:38,382 INFO [finetune.py:976] (1/7) Epoch 10, batch 5500, loss[loss=0.1915, simple_loss=0.2484, pruned_loss=0.0673, over 4773.00 frames. ], tot_loss[loss=0.1851, simple_loss=0.2514, pruned_loss=0.05943, over 952478.39 frames. ], batch size: 28, lr: 3.72e-03, grad_scale: 32.0 2023-04-27 00:48:47,142 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=57054.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 00:48:48,377 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=57056.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 00:49:19,124 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.6060, 3.3571, 2.4168, 2.6340, 1.9691, 1.8990, 2.6458, 1.9257], device='cuda:1'), covar=tensor([0.1524, 0.1271, 0.1458, 0.1589, 0.2268, 0.1880, 0.1001, 0.1987], device='cuda:1'), in_proj_covar=tensor([0.0194, 0.0213, 0.0169, 0.0202, 0.0203, 0.0185, 0.0158, 0.0186], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-27 00:49:35,989 INFO [finetune.py:976] (1/7) Epoch 10, batch 5550, loss[loss=0.1624, simple_loss=0.2307, pruned_loss=0.04712, over 4885.00 frames. ], tot_loss[loss=0.1874, simple_loss=0.2532, pruned_loss=0.06077, over 951429.96 frames. ], batch size: 32, lr: 3.72e-03, grad_scale: 32.0 2023-04-27 00:49:43,955 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4785, 3.1943, 0.8873, 1.6719, 1.7171, 2.3765, 1.7908, 0.9994], device='cuda:1'), covar=tensor([0.1357, 0.0934, 0.2049, 0.1370, 0.1155, 0.0935, 0.1519, 0.2068], device='cuda:1'), in_proj_covar=tensor([0.0116, 0.0247, 0.0141, 0.0121, 0.0133, 0.0152, 0.0118, 0.0122], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-27 00:49:59,961 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.204e+02 1.628e+02 1.818e+02 2.115e+02 5.040e+02, threshold=3.636e+02, percent-clipped=1.0 2023-04-27 00:50:06,923 INFO [finetune.py:976] (1/7) Epoch 10, batch 5600, loss[loss=0.1886, simple_loss=0.2671, pruned_loss=0.05501, over 4904.00 frames. ], tot_loss[loss=0.1923, simple_loss=0.2592, pruned_loss=0.06272, over 953598.06 frames. ], batch size: 35, lr: 3.72e-03, grad_scale: 32.0 2023-04-27 00:50:12,764 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=57159.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 00:50:24,869 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=57179.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 00:50:29,403 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=57186.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 00:50:35,768 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=57197.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 00:50:36,870 INFO [finetune.py:976] (1/7) Epoch 10, batch 5650, loss[loss=0.1232, simple_loss=0.1962, pruned_loss=0.02509, over 3964.00 frames. ], tot_loss[loss=0.1935, simple_loss=0.261, pruned_loss=0.06299, over 951785.61 frames. ], batch size: 17, lr: 3.72e-03, grad_scale: 32.0 2023-04-27 00:50:53,686 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=57227.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 00:50:54,868 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.0127, 1.1324, 3.1142, 2.7633, 2.7702, 2.8204, 2.9403, 2.5826], device='cuda:1'), covar=tensor([0.9850, 0.7713, 0.2240, 0.3545, 0.3162, 0.4402, 0.3007, 0.3534], device='cuda:1'), in_proj_covar=tensor([0.0309, 0.0306, 0.0405, 0.0406, 0.0350, 0.0405, 0.0315, 0.0367], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-27 00:50:59,542 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.003e+02 1.673e+02 1.923e+02 2.288e+02 6.250e+02, threshold=3.847e+02, percent-clipped=4.0 2023-04-27 00:51:03,823 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=57244.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 00:51:04,401 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.4888, 3.0116, 1.2670, 1.6468, 2.4138, 1.3907, 3.9098, 2.3089], device='cuda:1'), covar=tensor([0.0606, 0.0554, 0.0766, 0.1267, 0.0489, 0.1013, 0.0267, 0.0561], device='cuda:1'), in_proj_covar=tensor([0.0053, 0.0067, 0.0050, 0.0047, 0.0052, 0.0053, 0.0079, 0.0052], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0011, 0.0008], device='cuda:1') 2023-04-27 00:51:05,615 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=57247.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 00:51:06,699 INFO [finetune.py:976] (1/7) Epoch 10, batch 5700, loss[loss=0.2091, simple_loss=0.2481, pruned_loss=0.08507, over 3888.00 frames. ], tot_loss[loss=0.1904, simple_loss=0.256, pruned_loss=0.06239, over 930063.13 frames. ], batch size: 17, lr: 3.72e-03, grad_scale: 32.0 2023-04-27 00:51:10,402 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4438, 1.4770, 3.7716, 3.5403, 3.3103, 3.6013, 3.5811, 3.3263], device='cuda:1'), covar=tensor([0.6368, 0.5140, 0.1163, 0.1754, 0.1260, 0.1644, 0.1356, 0.1757], device='cuda:1'), in_proj_covar=tensor([0.0308, 0.0305, 0.0403, 0.0405, 0.0349, 0.0404, 0.0314, 0.0366], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-27 00:51:12,250 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=57258.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 00:51:15,769 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=57264.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 00:51:38,655 INFO [finetune.py:976] (1/7) Epoch 11, batch 0, loss[loss=0.1879, simple_loss=0.2559, pruned_loss=0.06001, over 4727.00 frames. ], tot_loss[loss=0.1879, simple_loss=0.2559, pruned_loss=0.06001, over 4727.00 frames. ], batch size: 59, lr: 3.72e-03, grad_scale: 32.0 2023-04-27 00:51:38,655 INFO [finetune.py:1001] (1/7) Computing validation loss 2023-04-27 00:51:59,885 INFO [finetune.py:1010] (1/7) Epoch 11, validation: loss=0.1558, simple_loss=0.2272, pruned_loss=0.04225, over 2265189.00 frames. 2023-04-27 00:51:59,886 INFO [finetune.py:1011] (1/7) Maximum memory allocated so far is 6391MB 2023-04-27 00:52:28,821 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=57305.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 00:52:32,464 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.56 vs. limit=5.0 2023-04-27 00:52:42,962 INFO [finetune.py:976] (1/7) Epoch 11, batch 50, loss[loss=0.1942, simple_loss=0.2531, pruned_loss=0.06763, over 4814.00 frames. ], tot_loss[loss=0.1997, simple_loss=0.2655, pruned_loss=0.06692, over 217997.60 frames. ], batch size: 33, lr: 3.72e-03, grad_scale: 32.0 2023-04-27 00:52:45,387 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7568, 2.3235, 1.6460, 1.5950, 1.3048, 1.3161, 1.7420, 1.2361], device='cuda:1'), covar=tensor([0.1713, 0.1377, 0.1628, 0.1887, 0.2464, 0.2015, 0.1120, 0.2095], device='cuda:1'), in_proj_covar=tensor([0.0195, 0.0214, 0.0169, 0.0204, 0.0204, 0.0185, 0.0159, 0.0186], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-27 00:52:49,891 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.045e+02 1.652e+02 2.063e+02 2.429e+02 4.586e+02, threshold=4.127e+02, percent-clipped=3.0 2023-04-27 00:52:57,809 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=57349.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 00:52:59,007 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=57351.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 00:53:20,600 INFO [finetune.py:976] (1/7) Epoch 11, batch 100, loss[loss=0.1693, simple_loss=0.2452, pruned_loss=0.04671, over 4890.00 frames. ], tot_loss[loss=0.1917, simple_loss=0.2571, pruned_loss=0.06315, over 380946.11 frames. ], batch size: 32, lr: 3.72e-03, grad_scale: 32.0 2023-04-27 00:53:34,833 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.5116, 2.3804, 2.6160, 2.8757, 2.9796, 2.3663, 2.0142, 2.6530], device='cuda:1'), covar=tensor([0.0979, 0.0928, 0.0542, 0.0710, 0.0569, 0.0982, 0.0928, 0.0617], device='cuda:1'), in_proj_covar=tensor([0.0194, 0.0201, 0.0180, 0.0173, 0.0176, 0.0186, 0.0158, 0.0182], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 00:53:40,244 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=57398.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 00:54:10,097 INFO [finetune.py:976] (1/7) Epoch 11, batch 150, loss[loss=0.1761, simple_loss=0.2299, pruned_loss=0.06118, over 4895.00 frames. ], tot_loss[loss=0.1864, simple_loss=0.2511, pruned_loss=0.06088, over 507989.65 frames. ], batch size: 32, lr: 3.72e-03, grad_scale: 32.0 2023-04-27 00:54:27,172 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.157e+02 1.639e+02 1.968e+02 2.285e+02 5.137e+02, threshold=3.937e+02, percent-clipped=1.0 2023-04-27 00:54:29,608 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([4.4682, 4.2631, 3.0082, 5.0796, 4.2319, 4.4259, 1.9046, 4.3500], device='cuda:1'), covar=tensor([0.1432, 0.1141, 0.3278, 0.0978, 0.2909, 0.1539, 0.5587, 0.1952], device='cuda:1'), in_proj_covar=tensor([0.0241, 0.0214, 0.0247, 0.0302, 0.0296, 0.0248, 0.0266, 0.0269], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 00:54:37,561 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-04-27 00:54:52,699 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=57459.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 00:54:52,723 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=57459.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 00:55:00,023 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=57462.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 00:55:02,966 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6393, 1.4441, 4.5564, 4.2109, 3.9583, 4.2729, 4.2035, 3.9646], device='cuda:1'), covar=tensor([0.7196, 0.6042, 0.1037, 0.1833, 0.1219, 0.1518, 0.1356, 0.1767], device='cuda:1'), in_proj_covar=tensor([0.0310, 0.0307, 0.0406, 0.0408, 0.0350, 0.0406, 0.0315, 0.0367], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-27 00:55:02,996 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8025, 1.7972, 0.9349, 1.4381, 1.9132, 1.6472, 1.5085, 1.5432], device='cuda:1'), covar=tensor([0.0475, 0.0400, 0.0382, 0.0574, 0.0286, 0.0511, 0.0548, 0.0587], device='cuda:1'), in_proj_covar=tensor([0.0030, 0.0025, 0.0023, 0.0030, 0.0020, 0.0029, 0.0029, 0.0029], device='cuda:1'), out_proj_covar=tensor([0.0049, 0.0044, 0.0037, 0.0050, 0.0037, 0.0048, 0.0048, 0.0050], device='cuda:1') 2023-04-27 00:55:10,526 INFO [finetune.py:976] (1/7) Epoch 11, batch 200, loss[loss=0.2003, simple_loss=0.2685, pruned_loss=0.06603, over 4914.00 frames. ], tot_loss[loss=0.1854, simple_loss=0.2501, pruned_loss=0.06039, over 608765.24 frames. ], batch size: 46, lr: 3.72e-03, grad_scale: 32.0 2023-04-27 00:55:18,957 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8412, 1.1316, 1.4590, 1.5642, 1.5122, 1.6014, 1.4906, 1.4509], device='cuda:1'), covar=tensor([0.4567, 0.5553, 0.4861, 0.4838, 0.5647, 0.8493, 0.5282, 0.5347], device='cuda:1'), in_proj_covar=tensor([0.0329, 0.0379, 0.0315, 0.0326, 0.0338, 0.0402, 0.0360, 0.0324], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 00:55:20,201 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.45 vs. limit=5.0 2023-04-27 00:55:22,544 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=57495.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 00:55:30,370 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=57507.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 00:55:40,652 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=57523.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 00:55:42,940 INFO [finetune.py:976] (1/7) Epoch 11, batch 250, loss[loss=0.1481, simple_loss=0.2176, pruned_loss=0.03931, over 4773.00 frames. ], tot_loss[loss=0.1874, simple_loss=0.2528, pruned_loss=0.06097, over 684827.68 frames. ], batch size: 28, lr: 3.71e-03, grad_scale: 32.0 2023-04-27 00:55:50,073 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.267e+02 1.758e+02 2.095e+02 2.670e+02 5.121e+02, threshold=4.190e+02, percent-clipped=8.0 2023-04-27 00:55:52,329 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([4.3442, 4.2116, 3.0449, 4.9981, 4.2742, 4.2990, 1.8385, 4.3379], device='cuda:1'), covar=tensor([0.1495, 0.1064, 0.3413, 0.0852, 0.2834, 0.1580, 0.5395, 0.1941], device='cuda:1'), in_proj_covar=tensor([0.0241, 0.0214, 0.0248, 0.0301, 0.0296, 0.0248, 0.0267, 0.0269], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 00:55:54,617 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=57542.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 00:56:01,285 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=57553.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 00:56:03,138 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=57556.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 00:56:04,905 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8167, 1.4843, 1.8986, 2.2797, 1.9321, 1.8014, 1.8778, 1.9184], device='cuda:1'), covar=tensor([0.5653, 0.7534, 0.8425, 0.7492, 0.7279, 0.9730, 0.9699, 0.8026], device='cuda:1'), in_proj_covar=tensor([0.0407, 0.0412, 0.0497, 0.0516, 0.0437, 0.0457, 0.0467, 0.0466], device='cuda:1'), out_proj_covar=tensor([9.9207e-05, 1.0217e-04, 1.1214e-04, 1.2278e-04, 1.0577e-04, 1.1042e-04, 1.1193e-04, 1.1199e-04], device='cuda:1') 2023-04-27 00:56:07,254 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7543, 1.4105, 4.6990, 4.3949, 4.1116, 4.4891, 4.1935, 4.1817], device='cuda:1'), covar=tensor([0.7005, 0.6112, 0.1003, 0.1800, 0.1118, 0.1149, 0.1974, 0.1443], device='cuda:1'), in_proj_covar=tensor([0.0309, 0.0305, 0.0405, 0.0408, 0.0349, 0.0405, 0.0314, 0.0365], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-27 00:56:08,501 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=57564.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 00:56:16,735 INFO [finetune.py:976] (1/7) Epoch 11, batch 300, loss[loss=0.2207, simple_loss=0.2913, pruned_loss=0.07511, over 4745.00 frames. ], tot_loss[loss=0.1917, simple_loss=0.2579, pruned_loss=0.06272, over 745960.21 frames. ], batch size: 59, lr: 3.71e-03, grad_scale: 32.0 2023-04-27 00:56:17,559 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=3.91 vs. limit=5.0 2023-04-27 00:56:27,334 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([3.3202, 3.2425, 2.7139, 3.8814, 3.0961, 3.2809, 1.8558, 3.3285], device='cuda:1'), covar=tensor([0.1730, 0.1405, 0.3897, 0.1743, 0.3347, 0.1898, 0.4524, 0.2397], device='cuda:1'), in_proj_covar=tensor([0.0242, 0.0215, 0.0249, 0.0303, 0.0297, 0.0249, 0.0268, 0.0270], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 00:56:32,857 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=57600.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 00:56:37,147 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5527, 2.6643, 2.3044, 2.3775, 2.7252, 2.3770, 3.6995, 2.1841], device='cuda:1'), covar=tensor([0.4327, 0.2200, 0.4713, 0.3827, 0.2145, 0.2908, 0.1516, 0.4162], device='cuda:1'), in_proj_covar=tensor([0.0338, 0.0341, 0.0424, 0.0356, 0.0381, 0.0377, 0.0377, 0.0413], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 00:56:37,187 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0028, 1.7774, 2.1784, 2.5691, 2.0777, 1.9419, 2.0878, 2.0877], device='cuda:1'), covar=tensor([0.6271, 0.8688, 0.9980, 0.7792, 0.7979, 1.1344, 1.1747, 1.0115], device='cuda:1'), in_proj_covar=tensor([0.0408, 0.0413, 0.0497, 0.0517, 0.0437, 0.0458, 0.0467, 0.0466], device='cuda:1'), out_proj_covar=tensor([9.9286e-05, 1.0225e-04, 1.1216e-04, 1.2286e-04, 1.0591e-04, 1.1066e-04, 1.1198e-04, 1.1207e-04], device='cuda:1') 2023-04-27 00:56:40,095 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=57612.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 00:56:41,881 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-04-27 00:56:50,090 INFO [finetune.py:976] (1/7) Epoch 11, batch 350, loss[loss=0.1746, simple_loss=0.2563, pruned_loss=0.04642, over 4895.00 frames. ], tot_loss[loss=0.1913, simple_loss=0.258, pruned_loss=0.06223, over 792199.64 frames. ], batch size: 37, lr: 3.71e-03, grad_scale: 32.0 2023-04-27 00:56:50,160 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([3.8485, 3.7685, 2.7443, 4.4336, 3.8583, 3.8682, 1.7188, 3.7425], device='cuda:1'), covar=tensor([0.1422, 0.1157, 0.2758, 0.1401, 0.2980, 0.1535, 0.5363, 0.2375], device='cuda:1'), in_proj_covar=tensor([0.0241, 0.0215, 0.0248, 0.0302, 0.0296, 0.0248, 0.0267, 0.0269], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 00:56:52,175 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.91 vs. limit=5.0 2023-04-27 00:56:56,644 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.085e+02 1.613e+02 1.965e+02 2.419e+02 4.062e+02, threshold=3.929e+02, percent-clipped=0.0 2023-04-27 00:57:11,402 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=57649.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 00:57:13,093 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=57651.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 00:57:46,786 INFO [finetune.py:976] (1/7) Epoch 11, batch 400, loss[loss=0.1895, simple_loss=0.2597, pruned_loss=0.05967, over 4827.00 frames. ], tot_loss[loss=0.1913, simple_loss=0.2588, pruned_loss=0.06192, over 828585.35 frames. ], batch size: 47, lr: 3.71e-03, grad_scale: 64.0 2023-04-27 00:58:06,587 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8363, 2.8154, 2.1740, 2.3340, 1.9388, 2.3494, 2.4081, 1.7617], device='cuda:1'), covar=tensor([0.2545, 0.1077, 0.0898, 0.1518, 0.3403, 0.1268, 0.2162, 0.2659], device='cuda:1'), in_proj_covar=tensor([0.0294, 0.0315, 0.0228, 0.0285, 0.0314, 0.0267, 0.0253, 0.0276], device='cuda:1'), out_proj_covar=tensor([1.1856e-04, 1.2659e-04, 9.1447e-05, 1.1417e-04, 1.2835e-04, 1.0747e-04, 1.0324e-04, 1.1079e-04], device='cuda:1') 2023-04-27 00:58:10,658 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=57697.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 00:58:12,862 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=57699.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 00:58:27,233 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.2459, 1.6979, 2.0794, 2.4760, 2.0689, 1.6075, 1.2639, 1.8881], device='cuda:1'), covar=tensor([0.3163, 0.3349, 0.1506, 0.2418, 0.2612, 0.2760, 0.4547, 0.2223], device='cuda:1'), in_proj_covar=tensor([0.0284, 0.0248, 0.0220, 0.0317, 0.0213, 0.0227, 0.0232, 0.0185], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-27 00:58:30,752 INFO [finetune.py:976] (1/7) Epoch 11, batch 450, loss[loss=0.1442, simple_loss=0.2246, pruned_loss=0.03188, over 4870.00 frames. ], tot_loss[loss=0.1909, simple_loss=0.2584, pruned_loss=0.06174, over 857785.34 frames. ], batch size: 34, lr: 3.71e-03, grad_scale: 64.0 2023-04-27 00:58:33,256 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-04-27 00:58:34,980 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=57733.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 00:58:37,317 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.069e+02 1.604e+02 1.948e+02 2.335e+02 4.408e+02, threshold=3.896e+02, percent-clipped=1.0 2023-04-27 00:59:00,418 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=57754.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 00:59:32,061 INFO [finetune.py:976] (1/7) Epoch 11, batch 500, loss[loss=0.1837, simple_loss=0.2409, pruned_loss=0.06328, over 4771.00 frames. ], tot_loss[loss=0.1892, simple_loss=0.256, pruned_loss=0.06121, over 880602.59 frames. ], batch size: 28, lr: 3.71e-03, grad_scale: 64.0 2023-04-27 00:59:36,346 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.3856, 1.8633, 2.2753, 2.7740, 2.2009, 1.7641, 1.6161, 2.0954], device='cuda:1'), covar=tensor([0.3986, 0.3931, 0.1973, 0.3291, 0.3291, 0.3222, 0.4888, 0.2783], device='cuda:1'), in_proj_covar=tensor([0.0286, 0.0250, 0.0222, 0.0319, 0.0214, 0.0229, 0.0234, 0.0186], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-27 00:59:48,114 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7229, 1.5912, 1.7732, 2.0960, 2.1595, 1.6815, 1.4083, 1.8929], device='cuda:1'), covar=tensor([0.0816, 0.1084, 0.0694, 0.0516, 0.0476, 0.0811, 0.0763, 0.0489], device='cuda:1'), in_proj_covar=tensor([0.0193, 0.0200, 0.0180, 0.0172, 0.0175, 0.0185, 0.0157, 0.0181], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 00:59:54,255 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=57794.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 00:59:59,051 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6639, 1.4244, 1.9225, 1.8924, 1.4607, 1.3069, 1.6037, 1.1117], device='cuda:1'), covar=tensor([0.0686, 0.0898, 0.0529, 0.0777, 0.0934, 0.1422, 0.0843, 0.0837], device='cuda:1'), in_proj_covar=tensor([0.0065, 0.0071, 0.0069, 0.0066, 0.0074, 0.0094, 0.0075, 0.0071], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-27 01:00:25,877 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=57818.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:00:31,730 INFO [finetune.py:976] (1/7) Epoch 11, batch 550, loss[loss=0.1843, simple_loss=0.2461, pruned_loss=0.0612, over 4705.00 frames. ], tot_loss[loss=0.1866, simple_loss=0.2529, pruned_loss=0.06014, over 895754.97 frames. ], batch size: 23, lr: 3.71e-03, grad_scale: 64.0 2023-04-27 01:00:38,242 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.096e+02 1.623e+02 1.943e+02 2.373e+02 5.716e+02, threshold=3.887e+02, percent-clipped=4.0 2023-04-27 01:00:41,335 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=57842.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:00:47,293 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=57851.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:00:49,051 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=57853.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 01:00:52,585 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.7905, 2.7546, 2.0742, 3.2493, 2.8449, 2.7955, 1.1968, 2.7507], device='cuda:1'), covar=tensor([0.2250, 0.1947, 0.3951, 0.3247, 0.4579, 0.2414, 0.6212, 0.3473], device='cuda:1'), in_proj_covar=tensor([0.0241, 0.0215, 0.0248, 0.0301, 0.0297, 0.0249, 0.0268, 0.0269], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 01:01:04,583 INFO [finetune.py:976] (1/7) Epoch 11, batch 600, loss[loss=0.2371, simple_loss=0.289, pruned_loss=0.09261, over 4910.00 frames. ], tot_loss[loss=0.1868, simple_loss=0.2531, pruned_loss=0.06029, over 911826.54 frames. ], batch size: 43, lr: 3.71e-03, grad_scale: 64.0 2023-04-27 01:01:13,413 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=57890.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:01:20,161 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=57900.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:01:20,737 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=57901.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:01:38,048 INFO [finetune.py:976] (1/7) Epoch 11, batch 650, loss[loss=0.2236, simple_loss=0.2977, pruned_loss=0.07468, over 4821.00 frames. ], tot_loss[loss=0.1895, simple_loss=0.2563, pruned_loss=0.06133, over 923110.49 frames. ], batch size: 51, lr: 3.71e-03, grad_scale: 64.0 2023-04-27 01:01:45,068 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.147e+02 1.753e+02 2.161e+02 2.818e+02 6.431e+02, threshold=4.321e+02, percent-clipped=6.0 2023-04-27 01:01:51,794 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=57948.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:01:53,686 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7141, 2.1159, 1.6639, 1.5825, 1.4054, 1.4095, 1.6673, 1.3504], device='cuda:1'), covar=tensor([0.1184, 0.1241, 0.1142, 0.1431, 0.1788, 0.1512, 0.0854, 0.1511], device='cuda:1'), in_proj_covar=tensor([0.0195, 0.0215, 0.0170, 0.0204, 0.0204, 0.0185, 0.0159, 0.0187], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-27 01:02:11,760 INFO [finetune.py:976] (1/7) Epoch 11, batch 700, loss[loss=0.1684, simple_loss=0.2471, pruned_loss=0.0448, over 4764.00 frames. ], tot_loss[loss=0.1906, simple_loss=0.2581, pruned_loss=0.06149, over 929071.79 frames. ], batch size: 54, lr: 3.71e-03, grad_scale: 64.0 2023-04-27 01:02:25,953 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.2873, 1.7609, 2.1018, 2.5544, 2.0815, 1.6328, 1.3242, 1.8961], device='cuda:1'), covar=tensor([0.3353, 0.3341, 0.1650, 0.2513, 0.2867, 0.2784, 0.4799, 0.2362], device='cuda:1'), in_proj_covar=tensor([0.0285, 0.0249, 0.0221, 0.0318, 0.0213, 0.0228, 0.0233, 0.0185], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-27 01:02:57,541 INFO [finetune.py:976] (1/7) Epoch 11, batch 750, loss[loss=0.203, simple_loss=0.2804, pruned_loss=0.06278, over 4828.00 frames. ], tot_loss[loss=0.1928, simple_loss=0.2604, pruned_loss=0.06257, over 936406.79 frames. ], batch size: 47, lr: 3.71e-03, grad_scale: 32.0 2023-04-27 01:03:04,227 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.211e+02 1.766e+02 1.999e+02 2.458e+02 5.140e+02, threshold=3.998e+02, percent-clipped=2.0 2023-04-27 01:03:14,988 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=58054.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:03:31,278 INFO [finetune.py:976] (1/7) Epoch 11, batch 800, loss[loss=0.1608, simple_loss=0.2182, pruned_loss=0.05167, over 4068.00 frames. ], tot_loss[loss=0.1924, simple_loss=0.2602, pruned_loss=0.06234, over 940663.82 frames. ], batch size: 17, lr: 3.71e-03, grad_scale: 32.0 2023-04-27 01:03:38,599 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=58089.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:03:47,429 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=58102.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:03:58,170 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=58118.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:04:04,034 INFO [finetune.py:976] (1/7) Epoch 11, batch 850, loss[loss=0.1958, simple_loss=0.2585, pruned_loss=0.06655, over 4758.00 frames. ], tot_loss[loss=0.191, simple_loss=0.2583, pruned_loss=0.06188, over 942248.10 frames. ], batch size: 54, lr: 3.71e-03, grad_scale: 32.0 2023-04-27 01:04:09,580 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([0.8403, 1.8531, 1.7085, 1.5205, 1.9547, 1.6110, 2.5027, 1.5263], device='cuda:1'), covar=tensor([0.4108, 0.2106, 0.4909, 0.3252, 0.1857, 0.2672, 0.1454, 0.4904], device='cuda:1'), in_proj_covar=tensor([0.0342, 0.0346, 0.0426, 0.0358, 0.0385, 0.0382, 0.0379, 0.0417], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 01:04:10,688 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.079e+02 1.590e+02 1.973e+02 2.582e+02 4.894e+02, threshold=3.945e+02, percent-clipped=3.0 2023-04-27 01:04:12,612 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=58141.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:04:14,899 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=58144.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:04:19,566 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=58151.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:04:22,034 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8109, 2.9982, 2.5810, 2.6360, 2.9871, 2.6246, 3.9362, 2.4754], device='cuda:1'), covar=tensor([0.3917, 0.1921, 0.3527, 0.3169, 0.1916, 0.2599, 0.1268, 0.3605], device='cuda:1'), in_proj_covar=tensor([0.0342, 0.0346, 0.0427, 0.0358, 0.0385, 0.0382, 0.0379, 0.0417], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 01:04:39,248 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=58166.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:04:50,556 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7457, 3.5455, 1.1344, 2.2143, 2.1122, 2.6880, 2.2618, 1.3556], device='cuda:1'), covar=tensor([0.1118, 0.0850, 0.1702, 0.1071, 0.0922, 0.0835, 0.1227, 0.1931], device='cuda:1'), in_proj_covar=tensor([0.0117, 0.0250, 0.0141, 0.0123, 0.0134, 0.0154, 0.0119, 0.0123], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-27 01:04:52,829 INFO [finetune.py:976] (1/7) Epoch 11, batch 900, loss[loss=0.1569, simple_loss=0.2277, pruned_loss=0.04311, over 4749.00 frames. ], tot_loss[loss=0.1878, simple_loss=0.2548, pruned_loss=0.06042, over 944520.88 frames. ], batch size: 27, lr: 3.71e-03, grad_scale: 32.0 2023-04-27 01:05:19,781 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=58199.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:05:22,311 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=58202.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:05:28,588 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=58205.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:05:42,972 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7209, 1.8101, 1.2581, 1.5154, 2.0671, 1.6531, 1.5819, 1.6009], device='cuda:1'), covar=tensor([0.0458, 0.0318, 0.0339, 0.0458, 0.0278, 0.0454, 0.0436, 0.0503], device='cuda:1'), in_proj_covar=tensor([0.0030, 0.0025, 0.0023, 0.0030, 0.0020, 0.0029, 0.0029, 0.0029], device='cuda:1'), out_proj_covar=tensor([0.0049, 0.0044, 0.0037, 0.0050, 0.0037, 0.0048, 0.0048, 0.0050], device='cuda:1') 2023-04-27 01:05:54,289 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.4368, 1.5800, 1.7239, 1.7745, 1.6850, 1.8557, 1.8740, 1.7876], device='cuda:1'), covar=tensor([0.4207, 0.6172, 0.5500, 0.5387, 0.6309, 0.8367, 0.6262, 0.6063], device='cuda:1'), in_proj_covar=tensor([0.0328, 0.0379, 0.0315, 0.0326, 0.0338, 0.0401, 0.0360, 0.0324], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 01:05:54,759 INFO [finetune.py:976] (1/7) Epoch 11, batch 950, loss[loss=0.2171, simple_loss=0.2766, pruned_loss=0.07879, over 4916.00 frames. ], tot_loss[loss=0.1861, simple_loss=0.2526, pruned_loss=0.05981, over 947334.90 frames. ], batch size: 37, lr: 3.71e-03, grad_scale: 32.0 2023-04-27 01:06:10,243 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.115e+02 1.752e+02 2.086e+02 2.428e+02 7.149e+02, threshold=4.172e+02, percent-clipped=3.0 2023-04-27 01:06:47,389 INFO [finetune.py:976] (1/7) Epoch 11, batch 1000, loss[loss=0.2103, simple_loss=0.2732, pruned_loss=0.07371, over 4888.00 frames. ], tot_loss[loss=0.1888, simple_loss=0.255, pruned_loss=0.06125, over 947123.53 frames. ], batch size: 32, lr: 3.71e-03, grad_scale: 32.0 2023-04-27 01:07:20,925 INFO [finetune.py:976] (1/7) Epoch 11, batch 1050, loss[loss=0.1304, simple_loss=0.2011, pruned_loss=0.02984, over 4757.00 frames. ], tot_loss[loss=0.1899, simple_loss=0.2574, pruned_loss=0.06127, over 947950.16 frames. ], batch size: 26, lr: 3.71e-03, grad_scale: 32.0 2023-04-27 01:07:28,201 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.162e+02 1.762e+02 2.120e+02 2.435e+02 5.017e+02, threshold=4.240e+02, percent-clipped=2.0 2023-04-27 01:07:37,029 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.62 vs. limit=5.0 2023-04-27 01:07:41,064 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5733, 1.0347, 1.3187, 1.1613, 1.6145, 1.3338, 1.0857, 1.2985], device='cuda:1'), covar=tensor([0.1680, 0.1559, 0.2164, 0.1422, 0.0909, 0.1586, 0.1848, 0.1966], device='cuda:1'), in_proj_covar=tensor([0.0306, 0.0321, 0.0352, 0.0298, 0.0335, 0.0320, 0.0306, 0.0362], device='cuda:1'), out_proj_covar=tensor([6.4240e-05, 6.7929e-05, 7.5606e-05, 6.1568e-05, 7.0034e-05, 6.8307e-05, 6.5496e-05, 7.7729e-05], device='cuda:1') 2023-04-27 01:07:52,945 INFO [finetune.py:976] (1/7) Epoch 11, batch 1100, loss[loss=0.2251, simple_loss=0.2929, pruned_loss=0.07864, over 4817.00 frames. ], tot_loss[loss=0.1912, simple_loss=0.2585, pruned_loss=0.06192, over 950206.54 frames. ], batch size: 38, lr: 3.71e-03, grad_scale: 32.0 2023-04-27 01:07:55,444 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.08 vs. limit=5.0 2023-04-27 01:08:01,698 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=58389.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:08:26,953 INFO [finetune.py:976] (1/7) Epoch 11, batch 1150, loss[loss=0.2045, simple_loss=0.2669, pruned_loss=0.07103, over 4879.00 frames. ], tot_loss[loss=0.1931, simple_loss=0.2606, pruned_loss=0.06278, over 953406.48 frames. ], batch size: 32, lr: 3.71e-03, grad_scale: 32.0 2023-04-27 01:08:34,049 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=58437.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:08:35,086 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.202e+02 1.681e+02 2.024e+02 2.460e+02 8.000e+02, threshold=4.047e+02, percent-clipped=3.0 2023-04-27 01:08:43,197 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8901, 1.4000, 1.6396, 1.6846, 1.6061, 1.3449, 0.6891, 1.3824], device='cuda:1'), covar=tensor([0.3345, 0.3500, 0.1706, 0.2313, 0.2796, 0.2697, 0.4598, 0.2306], device='cuda:1'), in_proj_covar=tensor([0.0285, 0.0249, 0.0221, 0.0317, 0.0214, 0.0228, 0.0233, 0.0185], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-27 01:08:48,063 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.0931, 1.0418, 1.2436, 1.1693, 1.0503, 0.9083, 1.0482, 0.5671], device='cuda:1'), covar=tensor([0.0601, 0.0668, 0.0648, 0.0512, 0.0792, 0.1312, 0.0515, 0.0972], device='cuda:1'), in_proj_covar=tensor([0.0066, 0.0072, 0.0070, 0.0067, 0.0075, 0.0096, 0.0076, 0.0072], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-27 01:08:53,264 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6416, 1.5739, 1.7800, 2.0256, 2.0581, 1.6343, 1.3276, 1.8714], device='cuda:1'), covar=tensor([0.0802, 0.1182, 0.0726, 0.0571, 0.0571, 0.0886, 0.0802, 0.0524], device='cuda:1'), in_proj_covar=tensor([0.0192, 0.0199, 0.0179, 0.0172, 0.0176, 0.0185, 0.0158, 0.0180], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 01:08:59,896 INFO [finetune.py:976] (1/7) Epoch 11, batch 1200, loss[loss=0.1789, simple_loss=0.2587, pruned_loss=0.04959, over 4814.00 frames. ], tot_loss[loss=0.1922, simple_loss=0.2594, pruned_loss=0.0625, over 955052.72 frames. ], batch size: 40, lr: 3.71e-03, grad_scale: 32.0 2023-04-27 01:09:07,990 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.3459, 1.2270, 1.6052, 1.5007, 1.2498, 1.1420, 1.2405, 0.7392], device='cuda:1'), covar=tensor([0.0556, 0.0706, 0.0457, 0.0646, 0.0829, 0.1121, 0.0601, 0.0724], device='cuda:1'), in_proj_covar=tensor([0.0066, 0.0071, 0.0069, 0.0066, 0.0075, 0.0095, 0.0075, 0.0071], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-27 01:09:13,882 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=58497.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:09:15,684 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=58500.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:09:32,952 INFO [finetune.py:976] (1/7) Epoch 11, batch 1250, loss[loss=0.1606, simple_loss=0.2261, pruned_loss=0.0475, over 4696.00 frames. ], tot_loss[loss=0.1905, simple_loss=0.2572, pruned_loss=0.0619, over 954932.09 frames. ], batch size: 23, lr: 3.71e-03, grad_scale: 32.0 2023-04-27 01:09:36,128 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-04-27 01:09:41,127 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.214e+02 1.641e+02 1.959e+02 2.303e+02 4.487e+02, threshold=3.918e+02, percent-clipped=1.0 2023-04-27 01:10:22,918 INFO [finetune.py:976] (1/7) Epoch 11, batch 1300, loss[loss=0.1322, simple_loss=0.2021, pruned_loss=0.03112, over 4786.00 frames. ], tot_loss[loss=0.1874, simple_loss=0.2539, pruned_loss=0.06046, over 953571.40 frames. ], batch size: 29, lr: 3.71e-03, grad_scale: 32.0 2023-04-27 01:11:07,803 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.1248, 2.8346, 0.7940, 1.3754, 1.5242, 2.0864, 1.6742, 0.9488], device='cuda:1'), covar=tensor([0.1819, 0.1779, 0.2346, 0.2014, 0.1328, 0.1347, 0.1605, 0.2122], device='cuda:1'), in_proj_covar=tensor([0.0116, 0.0246, 0.0139, 0.0121, 0.0132, 0.0152, 0.0117, 0.0121], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-27 01:11:29,576 INFO [finetune.py:976] (1/7) Epoch 11, batch 1350, loss[loss=0.188, simple_loss=0.2497, pruned_loss=0.06319, over 4158.00 frames. ], tot_loss[loss=0.1872, simple_loss=0.2537, pruned_loss=0.0604, over 953855.60 frames. ], batch size: 65, lr: 3.71e-03, grad_scale: 32.0 2023-04-27 01:11:46,674 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.165e+02 1.586e+02 1.956e+02 2.328e+02 4.668e+02, threshold=3.913e+02, percent-clipped=1.0 2023-04-27 01:11:59,972 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.0593, 1.5513, 1.3467, 1.7616, 1.6414, 1.8362, 1.4175, 3.3102], device='cuda:1'), covar=tensor([0.0701, 0.0784, 0.0834, 0.1164, 0.0629, 0.0480, 0.0730, 0.0163], device='cuda:1'), in_proj_covar=tensor([0.0038, 0.0039, 0.0040, 0.0044, 0.0040, 0.0039, 0.0039, 0.0058], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0014, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:1') 2023-04-27 01:12:32,653 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.2082, 2.5070, 0.7735, 1.4097, 1.5558, 1.7800, 1.5952, 0.8144], device='cuda:1'), covar=tensor([0.1405, 0.1299, 0.1942, 0.1516, 0.1148, 0.1095, 0.1671, 0.1921], device='cuda:1'), in_proj_covar=tensor([0.0115, 0.0246, 0.0139, 0.0121, 0.0132, 0.0152, 0.0117, 0.0121], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-27 01:12:34,437 INFO [finetune.py:976] (1/7) Epoch 11, batch 1400, loss[loss=0.1747, simple_loss=0.2591, pruned_loss=0.04521, over 4806.00 frames. ], tot_loss[loss=0.1903, simple_loss=0.2573, pruned_loss=0.06158, over 954972.04 frames. ], batch size: 41, lr: 3.71e-03, grad_scale: 32.0 2023-04-27 01:12:42,439 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=58681.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:13:42,639 INFO [finetune.py:976] (1/7) Epoch 11, batch 1450, loss[loss=0.1655, simple_loss=0.2208, pruned_loss=0.05507, over 4327.00 frames. ], tot_loss[loss=0.1938, simple_loss=0.2611, pruned_loss=0.06326, over 952980.36 frames. ], batch size: 19, lr: 3.71e-03, grad_scale: 32.0 2023-04-27 01:14:01,639 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.919e+01 1.615e+02 1.986e+02 2.296e+02 3.609e+02, threshold=3.972e+02, percent-clipped=0.0 2023-04-27 01:14:04,681 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=58742.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:14:05,933 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.25 vs. limit=5.0 2023-04-27 01:14:37,114 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=58775.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:14:38,276 INFO [finetune.py:976] (1/7) Epoch 11, batch 1500, loss[loss=0.1629, simple_loss=0.2384, pruned_loss=0.04372, over 4831.00 frames. ], tot_loss[loss=0.1948, simple_loss=0.2622, pruned_loss=0.06366, over 954117.68 frames. ], batch size: 30, lr: 3.70e-03, grad_scale: 32.0 2023-04-27 01:14:46,674 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.2088, 2.4955, 1.0538, 1.3264, 1.9051, 1.1847, 3.1189, 1.6372], device='cuda:1'), covar=tensor([0.0651, 0.0555, 0.0750, 0.1444, 0.0508, 0.1114, 0.0317, 0.0695], device='cuda:1'), in_proj_covar=tensor([0.0052, 0.0067, 0.0050, 0.0047, 0.0051, 0.0053, 0.0078, 0.0052], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:1') 2023-04-27 01:14:52,523 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=58797.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:14:54,839 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.0464, 2.5233, 0.8735, 1.4950, 1.5063, 1.8298, 1.5787, 0.8306], device='cuda:1'), covar=tensor([0.1587, 0.1203, 0.1797, 0.1356, 0.1161, 0.1006, 0.1506, 0.1799], device='cuda:1'), in_proj_covar=tensor([0.0116, 0.0247, 0.0139, 0.0122, 0.0133, 0.0152, 0.0118, 0.0122], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-27 01:14:54,854 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=58800.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:14:56,668 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.1972, 2.5272, 1.2048, 1.4678, 2.1033, 1.2577, 3.6988, 1.7972], device='cuda:1'), covar=tensor([0.0655, 0.0732, 0.0803, 0.1367, 0.0523, 0.1099, 0.0310, 0.0685], device='cuda:1'), in_proj_covar=tensor([0.0052, 0.0067, 0.0050, 0.0047, 0.0051, 0.0053, 0.0078, 0.0052], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0011, 0.0008], device='cuda:1') 2023-04-27 01:15:08,108 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5465, 1.5715, 0.7683, 1.3021, 1.6588, 1.4344, 1.3652, 1.4493], device='cuda:1'), covar=tensor([0.0524, 0.0388, 0.0384, 0.0561, 0.0284, 0.0542, 0.0509, 0.0608], device='cuda:1'), in_proj_covar=tensor([0.0029, 0.0025, 0.0023, 0.0029, 0.0020, 0.0029, 0.0029, 0.0029], device='cuda:1'), out_proj_covar=tensor([0.0049, 0.0044, 0.0038, 0.0050, 0.0037, 0.0048, 0.0048, 0.0049], device='cuda:1') 2023-04-27 01:15:11,710 INFO [finetune.py:976] (1/7) Epoch 11, batch 1550, loss[loss=0.1467, simple_loss=0.2271, pruned_loss=0.03315, over 4922.00 frames. ], tot_loss[loss=0.1937, simple_loss=0.2613, pruned_loss=0.06303, over 953474.41 frames. ], batch size: 38, lr: 3.70e-03, grad_scale: 32.0 2023-04-27 01:15:18,204 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=58836.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:15:19,336 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.373e+01 1.631e+02 1.941e+02 2.305e+02 6.407e+02, threshold=3.882e+02, percent-clipped=1.0 2023-04-27 01:15:19,434 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4549, 1.3133, 4.1450, 3.8474, 3.6763, 4.0013, 3.9448, 3.6428], device='cuda:1'), covar=tensor([0.6730, 0.5990, 0.1085, 0.1838, 0.1073, 0.1769, 0.1404, 0.1433], device='cuda:1'), in_proj_covar=tensor([0.0309, 0.0306, 0.0402, 0.0406, 0.0348, 0.0404, 0.0312, 0.0367], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-27 01:15:24,175 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=58845.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:15:26,504 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=58848.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:15:30,007 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([3.3253, 3.3715, 2.6225, 3.8438, 3.2927, 3.3525, 1.5226, 3.3157], device='cuda:1'), covar=tensor([0.2030, 0.1263, 0.2966, 0.1996, 0.2664, 0.1905, 0.5574, 0.2448], device='cuda:1'), in_proj_covar=tensor([0.0241, 0.0215, 0.0247, 0.0300, 0.0296, 0.0249, 0.0267, 0.0267], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 01:15:44,971 INFO [finetune.py:976] (1/7) Epoch 11, batch 1600, loss[loss=0.1881, simple_loss=0.2558, pruned_loss=0.06017, over 4916.00 frames. ], tot_loss[loss=0.1911, simple_loss=0.2584, pruned_loss=0.06191, over 954550.26 frames. ], batch size: 37, lr: 3.70e-03, grad_scale: 32.0 2023-04-27 01:16:05,011 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=58906.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:16:18,090 INFO [finetune.py:976] (1/7) Epoch 11, batch 1650, loss[loss=0.2091, simple_loss=0.2542, pruned_loss=0.08202, over 4938.00 frames. ], tot_loss[loss=0.1896, simple_loss=0.2563, pruned_loss=0.06146, over 955327.76 frames. ], batch size: 33, lr: 3.70e-03, grad_scale: 32.0 2023-04-27 01:16:18,894 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.72 vs. limit=2.0 2023-04-27 01:16:25,722 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.044e+02 1.662e+02 1.988e+02 2.360e+02 4.922e+02, threshold=3.975e+02, percent-clipped=3.0 2023-04-27 01:16:44,577 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0735, 2.6360, 2.1320, 2.4654, 1.7617, 2.2170, 2.2483, 1.7070], device='cuda:1'), covar=tensor([0.1936, 0.1019, 0.0782, 0.1197, 0.3022, 0.1050, 0.1877, 0.2613], device='cuda:1'), in_proj_covar=tensor([0.0293, 0.0313, 0.0227, 0.0282, 0.0311, 0.0267, 0.0251, 0.0273], device='cuda:1'), out_proj_covar=tensor([1.1812e-04, 1.2568e-04, 9.1132e-05, 1.1304e-04, 1.2722e-04, 1.0753e-04, 1.0248e-04, 1.0953e-04], device='cuda:1') 2023-04-27 01:17:00,459 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5935, 1.1715, 1.7239, 2.1577, 1.7963, 1.5685, 1.6131, 1.6808], device='cuda:1'), covar=tensor([0.5293, 0.7176, 0.6817, 0.6697, 0.6271, 0.8295, 0.8628, 0.8052], device='cuda:1'), in_proj_covar=tensor([0.0408, 0.0412, 0.0498, 0.0517, 0.0439, 0.0457, 0.0468, 0.0467], device='cuda:1'), out_proj_covar=tensor([9.9374e-05, 1.0213e-04, 1.1239e-04, 1.2277e-04, 1.0623e-04, 1.1054e-04, 1.1222e-04, 1.1217e-04], device='cuda:1') 2023-04-27 01:17:07,486 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=58967.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:17:13,432 INFO [finetune.py:976] (1/7) Epoch 11, batch 1700, loss[loss=0.1625, simple_loss=0.233, pruned_loss=0.04598, over 4866.00 frames. ], tot_loss[loss=0.1872, simple_loss=0.2534, pruned_loss=0.06046, over 956538.96 frames. ], batch size: 31, lr: 3.70e-03, grad_scale: 32.0 2023-04-27 01:17:40,502 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.3057, 1.3803, 4.0674, 3.8122, 3.5424, 3.8396, 3.8568, 3.5771], device='cuda:1'), covar=tensor([0.7396, 0.5926, 0.1095, 0.1794, 0.1301, 0.2206, 0.1598, 0.1664], device='cuda:1'), in_proj_covar=tensor([0.0309, 0.0305, 0.0402, 0.0406, 0.0349, 0.0405, 0.0312, 0.0366], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-27 01:17:40,513 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.2234, 3.2543, 0.8185, 1.7158, 1.6808, 2.1702, 1.7566, 0.9817], device='cuda:1'), covar=tensor([0.1524, 0.0851, 0.2118, 0.1271, 0.1136, 0.1097, 0.1670, 0.2050], device='cuda:1'), in_proj_covar=tensor([0.0117, 0.0249, 0.0141, 0.0122, 0.0134, 0.0154, 0.0119, 0.0123], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-27 01:17:47,002 INFO [finetune.py:976] (1/7) Epoch 11, batch 1750, loss[loss=0.2081, simple_loss=0.2656, pruned_loss=0.07534, over 4833.00 frames. ], tot_loss[loss=0.1894, simple_loss=0.2556, pruned_loss=0.06163, over 957834.24 frames. ], batch size: 30, lr: 3.70e-03, grad_scale: 32.0 2023-04-27 01:17:53,224 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=59037.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:17:53,741 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.098e+02 1.792e+02 2.172e+02 2.715e+02 5.550e+02, threshold=4.344e+02, percent-clipped=4.0 2023-04-27 01:18:18,310 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.2257, 2.0128, 2.2039, 2.5993, 2.7090, 2.1438, 1.7706, 2.3085], device='cuda:1'), covar=tensor([0.0743, 0.0910, 0.0610, 0.0472, 0.0469, 0.0743, 0.0922, 0.0478], device='cuda:1'), in_proj_covar=tensor([0.0196, 0.0204, 0.0183, 0.0176, 0.0179, 0.0189, 0.0160, 0.0185], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 01:18:30,832 INFO [finetune.py:976] (1/7) Epoch 11, batch 1800, loss[loss=0.2702, simple_loss=0.3281, pruned_loss=0.1061, over 4768.00 frames. ], tot_loss[loss=0.1915, simple_loss=0.2583, pruned_loss=0.06232, over 955437.11 frames. ], batch size: 54, lr: 3.70e-03, grad_scale: 32.0 2023-04-27 01:19:09,377 INFO [finetune.py:976] (1/7) Epoch 11, batch 1850, loss[loss=0.2067, simple_loss=0.2753, pruned_loss=0.06905, over 4846.00 frames. ], tot_loss[loss=0.192, simple_loss=0.2594, pruned_loss=0.06229, over 955286.25 frames. ], batch size: 49, lr: 3.70e-03, grad_scale: 32.0 2023-04-27 01:19:11,898 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=59131.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:19:21,288 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.048e+02 1.699e+02 2.019e+02 2.381e+02 7.878e+02, threshold=4.039e+02, percent-clipped=1.0 2023-04-27 01:19:30,571 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6460, 1.3346, 1.9124, 2.1464, 1.8034, 1.7145, 1.7936, 1.7592], device='cuda:1'), covar=tensor([0.5802, 0.7708, 0.7703, 0.8028, 0.7543, 1.0067, 0.9157, 0.9005], device='cuda:1'), in_proj_covar=tensor([0.0409, 0.0411, 0.0499, 0.0517, 0.0438, 0.0458, 0.0468, 0.0467], device='cuda:1'), out_proj_covar=tensor([9.9528e-05, 1.0196e-04, 1.1254e-04, 1.2269e-04, 1.0610e-04, 1.1069e-04, 1.1221e-04, 1.1228e-04], device='cuda:1') 2023-04-27 01:19:30,693 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 2023-04-27 01:19:52,232 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=59164.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:20:10,995 INFO [finetune.py:976] (1/7) Epoch 11, batch 1900, loss[loss=0.1655, simple_loss=0.2347, pruned_loss=0.04811, over 4771.00 frames. ], tot_loss[loss=0.1928, simple_loss=0.2601, pruned_loss=0.06275, over 955119.78 frames. ], batch size: 26, lr: 3.70e-03, grad_scale: 32.0 2023-04-27 01:21:00,133 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=59225.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:21:01,213 INFO [finetune.py:976] (1/7) Epoch 11, batch 1950, loss[loss=0.1966, simple_loss=0.2649, pruned_loss=0.06409, over 4760.00 frames. ], tot_loss[loss=0.191, simple_loss=0.2585, pruned_loss=0.06177, over 955811.93 frames. ], batch size: 27, lr: 3.70e-03, grad_scale: 32.0 2023-04-27 01:21:06,102 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.5280, 1.6956, 1.8827, 2.0042, 1.8337, 1.9295, 1.9811, 1.9202], device='cuda:1'), covar=tensor([0.5184, 0.7961, 0.6494, 0.6093, 0.7058, 0.9326, 0.6943, 0.6703], device='cuda:1'), in_proj_covar=tensor([0.0327, 0.0380, 0.0315, 0.0326, 0.0339, 0.0400, 0.0359, 0.0323], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 01:21:08,358 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.114e+02 1.600e+02 1.889e+02 2.365e+02 4.581e+02, threshold=3.778e+02, percent-clipped=1.0 2023-04-27 01:21:20,106 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-04-27 01:21:24,551 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=59262.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:21:35,025 INFO [finetune.py:976] (1/7) Epoch 11, batch 2000, loss[loss=0.1506, simple_loss=0.2188, pruned_loss=0.04124, over 4877.00 frames. ], tot_loss[loss=0.1888, simple_loss=0.2558, pruned_loss=0.06092, over 953602.65 frames. ], batch size: 31, lr: 3.70e-03, grad_scale: 32.0 2023-04-27 01:22:03,863 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.55 vs. limit=2.0 2023-04-27 01:22:08,861 INFO [finetune.py:976] (1/7) Epoch 11, batch 2050, loss[loss=0.1514, simple_loss=0.2154, pruned_loss=0.0437, over 4765.00 frames. ], tot_loss[loss=0.1851, simple_loss=0.2514, pruned_loss=0.05942, over 951426.67 frames. ], batch size: 27, lr: 3.70e-03, grad_scale: 32.0 2023-04-27 01:22:15,375 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=59337.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:22:15,893 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.113e+02 1.589e+02 2.008e+02 2.324e+02 3.763e+02, threshold=4.015e+02, percent-clipped=0.0 2023-04-27 01:22:48,351 INFO [finetune.py:976] (1/7) Epoch 11, batch 2100, loss[loss=0.2028, simple_loss=0.2672, pruned_loss=0.06922, over 4814.00 frames. ], tot_loss[loss=0.1841, simple_loss=0.2502, pruned_loss=0.05906, over 951258.51 frames. ], batch size: 51, lr: 3.70e-03, grad_scale: 32.0 2023-04-27 01:22:53,365 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=59385.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:22:58,199 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.3886, 1.6156, 1.7941, 1.9174, 1.8090, 1.9260, 1.8825, 1.8183], device='cuda:1'), covar=tensor([0.4767, 0.6660, 0.5779, 0.5584, 0.6648, 0.8307, 0.6515, 0.6005], device='cuda:1'), in_proj_covar=tensor([0.0326, 0.0379, 0.0314, 0.0325, 0.0338, 0.0400, 0.0357, 0.0322], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 01:23:32,963 INFO [finetune.py:976] (1/7) Epoch 11, batch 2150, loss[loss=0.218, simple_loss=0.2845, pruned_loss=0.0757, over 4038.00 frames. ], tot_loss[loss=0.1862, simple_loss=0.2528, pruned_loss=0.05978, over 952442.25 frames. ], batch size: 65, lr: 3.70e-03, grad_scale: 32.0 2023-04-27 01:23:35,533 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=59431.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:23:44,712 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.256e+02 1.698e+02 2.142e+02 2.512e+02 4.265e+02, threshold=4.284e+02, percent-clipped=3.0 2023-04-27 01:23:58,241 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.3603, 1.7670, 1.5106, 2.0699, 2.0442, 2.0821, 1.6557, 4.3500], device='cuda:1'), covar=tensor([0.0595, 0.0822, 0.0835, 0.1181, 0.0626, 0.0513, 0.0732, 0.0114], device='cuda:1'), in_proj_covar=tensor([0.0038, 0.0039, 0.0040, 0.0044, 0.0040, 0.0038, 0.0038, 0.0058], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0014, 0.0013, 0.0012, 0.0013, 0.0016], device='cuda:1') 2023-04-27 01:24:09,599 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0243, 1.3374, 1.8069, 2.4046, 1.8658, 1.4456, 1.1849, 1.6322], device='cuda:1'), covar=tensor([0.3724, 0.4130, 0.1948, 0.2740, 0.3174, 0.2914, 0.5014, 0.2615], device='cuda:1'), in_proj_covar=tensor([0.0286, 0.0248, 0.0222, 0.0316, 0.0215, 0.0228, 0.0232, 0.0185], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-27 01:24:15,852 INFO [finetune.py:976] (1/7) Epoch 11, batch 2200, loss[loss=0.167, simple_loss=0.2326, pruned_loss=0.05069, over 4788.00 frames. ], tot_loss[loss=0.1882, simple_loss=0.2557, pruned_loss=0.06039, over 953250.83 frames. ], batch size: 25, lr: 3.70e-03, grad_scale: 32.0 2023-04-27 01:24:18,112 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=59479.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:24:37,674 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=59502.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:24:54,703 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=59520.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:24:59,358 INFO [finetune.py:976] (1/7) Epoch 11, batch 2250, loss[loss=0.1787, simple_loss=0.2544, pruned_loss=0.05153, over 4756.00 frames. ], tot_loss[loss=0.1896, simple_loss=0.2572, pruned_loss=0.06101, over 952409.92 frames. ], batch size: 28, lr: 3.70e-03, grad_scale: 32.0 2023-04-27 01:25:13,213 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.038e+02 1.641e+02 2.091e+02 2.357e+02 4.911e+02, threshold=4.183e+02, percent-clipped=2.0 2023-04-27 01:25:44,079 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=3.71 vs. limit=5.0 2023-04-27 01:25:45,853 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=59562.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:25:46,493 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=59563.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:25:55,720 INFO [finetune.py:976] (1/7) Epoch 11, batch 2300, loss[loss=0.1945, simple_loss=0.272, pruned_loss=0.05855, over 4820.00 frames. ], tot_loss[loss=0.1905, simple_loss=0.2586, pruned_loss=0.06121, over 954651.35 frames. ], batch size: 38, lr: 3.70e-03, grad_scale: 32.0 2023-04-27 01:26:02,302 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-04-27 01:26:07,014 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.3038, 1.5375, 1.3431, 1.5953, 1.2979, 1.2668, 1.4479, 1.0477], device='cuda:1'), covar=tensor([0.1639, 0.1324, 0.1026, 0.1192, 0.3516, 0.1325, 0.1489, 0.2177], device='cuda:1'), in_proj_covar=tensor([0.0293, 0.0312, 0.0228, 0.0283, 0.0313, 0.0269, 0.0254, 0.0274], device='cuda:1'), out_proj_covar=tensor([1.1858e-04, 1.2538e-04, 9.1223e-05, 1.1338e-04, 1.2776e-04, 1.0822e-04, 1.0327e-04, 1.0952e-04], device='cuda:1') 2023-04-27 01:26:17,900 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=59610.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:26:29,150 INFO [finetune.py:976] (1/7) Epoch 11, batch 2350, loss[loss=0.2377, simple_loss=0.2816, pruned_loss=0.09691, over 4043.00 frames. ], tot_loss[loss=0.1901, simple_loss=0.2576, pruned_loss=0.06124, over 955741.15 frames. ], batch size: 65, lr: 3.70e-03, grad_scale: 32.0 2023-04-27 01:26:37,789 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.079e+02 1.642e+02 1.943e+02 2.312e+02 3.854e+02, threshold=3.885e+02, percent-clipped=0.0 2023-04-27 01:27:02,575 INFO [finetune.py:976] (1/7) Epoch 11, batch 2400, loss[loss=0.1782, simple_loss=0.2483, pruned_loss=0.05405, over 4750.00 frames. ], tot_loss[loss=0.1869, simple_loss=0.2541, pruned_loss=0.05984, over 955784.94 frames. ], batch size: 27, lr: 3.70e-03, grad_scale: 32.0 2023-04-27 01:27:36,065 INFO [finetune.py:976] (1/7) Epoch 11, batch 2450, loss[loss=0.1943, simple_loss=0.2578, pruned_loss=0.06538, over 4103.00 frames. ], tot_loss[loss=0.1838, simple_loss=0.2503, pruned_loss=0.05864, over 956554.14 frames. ], batch size: 65, lr: 3.70e-03, grad_scale: 32.0 2023-04-27 01:27:39,827 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.96 vs. limit=2.0 2023-04-27 01:27:43,829 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.154e+02 1.580e+02 1.900e+02 2.443e+02 4.290e+02, threshold=3.800e+02, percent-clipped=1.0 2023-04-27 01:28:09,958 INFO [finetune.py:976] (1/7) Epoch 11, batch 2500, loss[loss=0.17, simple_loss=0.2347, pruned_loss=0.05267, over 4740.00 frames. ], tot_loss[loss=0.1864, simple_loss=0.2525, pruned_loss=0.06014, over 953558.01 frames. ], batch size: 23, lr: 3.70e-03, grad_scale: 32.0 2023-04-27 01:28:44,949 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=59820.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:28:54,838 INFO [finetune.py:976] (1/7) Epoch 11, batch 2550, loss[loss=0.2059, simple_loss=0.2762, pruned_loss=0.0678, over 4851.00 frames. ], tot_loss[loss=0.189, simple_loss=0.2561, pruned_loss=0.06089, over 953983.18 frames. ], batch size: 44, lr: 3.70e-03, grad_scale: 32.0 2023-04-27 01:29:12,510 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.167e+02 1.688e+02 2.032e+02 2.494e+02 8.629e+02, threshold=4.065e+02, percent-clipped=5.0 2023-04-27 01:29:23,405 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=59848.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:29:35,314 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=59858.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:29:46,952 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=59868.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:29:57,640 INFO [finetune.py:976] (1/7) Epoch 11, batch 2600, loss[loss=0.1354, simple_loss=0.2012, pruned_loss=0.03475, over 3927.00 frames. ], tot_loss[loss=0.1908, simple_loss=0.2582, pruned_loss=0.06173, over 954795.21 frames. ], batch size: 17, lr: 3.70e-03, grad_scale: 32.0 2023-04-27 01:30:10,823 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=59889.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:30:22,124 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9128, 1.6183, 2.0225, 2.3474, 2.0218, 1.7817, 1.8990, 1.9390], device='cuda:1'), covar=tensor([0.5766, 0.8278, 0.9044, 0.7289, 0.7251, 1.0876, 1.0954, 1.0638], device='cuda:1'), in_proj_covar=tensor([0.0408, 0.0410, 0.0497, 0.0517, 0.0439, 0.0457, 0.0466, 0.0467], device='cuda:1'), out_proj_covar=tensor([9.9118e-05, 1.0185e-04, 1.1218e-04, 1.2272e-04, 1.0617e-04, 1.1056e-04, 1.1183e-04, 1.1213e-04], device='cuda:1') 2023-04-27 01:30:29,465 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=59909.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:30:32,470 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=3.48 vs. limit=5.0 2023-04-27 01:30:51,533 INFO [finetune.py:976] (1/7) Epoch 11, batch 2650, loss[loss=0.2221, simple_loss=0.2925, pruned_loss=0.07589, over 4744.00 frames. ], tot_loss[loss=0.1922, simple_loss=0.2598, pruned_loss=0.06233, over 955455.65 frames. ], batch size: 54, lr: 3.70e-03, grad_scale: 32.0 2023-04-27 01:31:04,460 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.138e+02 1.693e+02 1.978e+02 2.480e+02 4.060e+02, threshold=3.956e+02, percent-clipped=0.0 2023-04-27 01:31:21,159 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=59950.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:31:41,179 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-04-27 01:31:41,228 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-04-27 01:31:43,484 INFO [finetune.py:976] (1/7) Epoch 11, batch 2700, loss[loss=0.1922, simple_loss=0.2426, pruned_loss=0.07088, over 3914.00 frames. ], tot_loss[loss=0.1906, simple_loss=0.2585, pruned_loss=0.0613, over 954595.28 frames. ], batch size: 17, lr: 3.70e-03, grad_scale: 32.0 2023-04-27 01:32:17,130 INFO [finetune.py:976] (1/7) Epoch 11, batch 2750, loss[loss=0.2027, simple_loss=0.2681, pruned_loss=0.0686, over 4893.00 frames. ], tot_loss[loss=0.1881, simple_loss=0.2555, pruned_loss=0.06034, over 954756.89 frames. ], batch size: 32, lr: 3.69e-03, grad_scale: 64.0 2023-04-27 01:32:24,354 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.196e+02 1.624e+02 1.944e+02 2.490e+02 4.678e+02, threshold=3.889e+02, percent-clipped=1.0 2023-04-27 01:32:25,707 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=60040.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:32:33,975 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=60052.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:32:42,795 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=60065.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:32:50,411 INFO [finetune.py:976] (1/7) Epoch 11, batch 2800, loss[loss=0.2011, simple_loss=0.2555, pruned_loss=0.07342, over 4916.00 frames. ], tot_loss[loss=0.1851, simple_loss=0.2518, pruned_loss=0.05922, over 954887.97 frames. ], batch size: 36, lr: 3.69e-03, grad_scale: 64.0 2023-04-27 01:32:51,103 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=60078.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:32:58,433 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.1165, 1.4854, 1.4256, 1.7872, 1.6090, 1.6842, 1.3877, 3.0466], device='cuda:1'), covar=tensor([0.0716, 0.0795, 0.0827, 0.1193, 0.0659, 0.0525, 0.0737, 0.0182], device='cuda:1'), in_proj_covar=tensor([0.0038, 0.0039, 0.0040, 0.0044, 0.0040, 0.0039, 0.0039, 0.0058], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:1') 2023-04-27 01:33:05,639 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=60101.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 01:33:14,372 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=60113.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 01:33:22,792 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=60126.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:33:23,286 INFO [finetune.py:976] (1/7) Epoch 11, batch 2850, loss[loss=0.2028, simple_loss=0.2696, pruned_loss=0.06801, over 4812.00 frames. ], tot_loss[loss=0.1864, simple_loss=0.2525, pruned_loss=0.06016, over 953709.99 frames. ], batch size: 45, lr: 3.69e-03, grad_scale: 64.0 2023-04-27 01:33:29,989 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.009e+02 1.671e+02 1.998e+02 2.360e+02 5.267e+02, threshold=3.997e+02, percent-clipped=5.0 2023-04-27 01:33:30,725 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=60139.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:33:43,077 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=60158.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:33:55,931 INFO [finetune.py:976] (1/7) Epoch 11, batch 2900, loss[loss=0.3374, simple_loss=0.3661, pruned_loss=0.1544, over 4775.00 frames. ], tot_loss[loss=0.1897, simple_loss=0.256, pruned_loss=0.06167, over 954173.05 frames. ], batch size: 59, lr: 3.69e-03, grad_scale: 64.0 2023-04-27 01:34:06,328 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.3783, 1.5816, 1.6886, 1.8753, 1.6251, 1.7546, 1.8069, 1.7425], device='cuda:1'), covar=tensor([0.5019, 0.6822, 0.5611, 0.5047, 0.7012, 0.9021, 0.6737, 0.6158], device='cuda:1'), in_proj_covar=tensor([0.0327, 0.0379, 0.0314, 0.0325, 0.0340, 0.0400, 0.0358, 0.0323], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 01:34:15,781 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=60198.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:34:16,348 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([3.9568, 3.9438, 2.7410, 4.6083, 4.0005, 3.9483, 2.0333, 3.9141], device='cuda:1'), covar=tensor([0.1547, 0.1349, 0.3400, 0.1459, 0.3280, 0.1952, 0.5459, 0.2682], device='cuda:1'), in_proj_covar=tensor([0.0242, 0.0217, 0.0250, 0.0304, 0.0298, 0.0250, 0.0268, 0.0271], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 01:34:25,134 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=60204.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:34:26,335 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=60206.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:34:38,984 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.0780, 2.1204, 1.8479, 1.6286, 2.2931, 1.8492, 2.7840, 1.6256], device='cuda:1'), covar=tensor([0.4539, 0.2297, 0.4869, 0.3672, 0.1962, 0.2759, 0.1646, 0.4744], device='cuda:1'), in_proj_covar=tensor([0.0341, 0.0347, 0.0424, 0.0357, 0.0382, 0.0380, 0.0378, 0.0419], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 01:34:58,743 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5267, 3.3710, 0.8837, 1.6750, 1.8993, 2.4290, 1.8881, 1.0288], device='cuda:1'), covar=tensor([0.1283, 0.0872, 0.2078, 0.1311, 0.0973, 0.0986, 0.1476, 0.2046], device='cuda:1'), in_proj_covar=tensor([0.0116, 0.0247, 0.0140, 0.0121, 0.0133, 0.0152, 0.0117, 0.0121], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-27 01:34:59,278 INFO [finetune.py:976] (1/7) Epoch 11, batch 2950, loss[loss=0.21, simple_loss=0.2792, pruned_loss=0.07044, over 4812.00 frames. ], tot_loss[loss=0.1906, simple_loss=0.2579, pruned_loss=0.06163, over 954074.14 frames. ], batch size: 45, lr: 3.69e-03, grad_scale: 32.0 2023-04-27 01:35:09,913 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-04-27 01:35:12,690 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 8.759e+01 1.634e+02 1.896e+02 2.531e+02 4.222e+02, threshold=3.792e+02, percent-clipped=1.0 2023-04-27 01:35:13,461 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.38 vs. limit=5.0 2023-04-27 01:35:21,688 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=60245.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:35:36,405 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=60259.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:35:54,405 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.12 vs. limit=2.0 2023-04-27 01:36:05,411 INFO [finetune.py:976] (1/7) Epoch 11, batch 3000, loss[loss=0.1978, simple_loss=0.273, pruned_loss=0.0613, over 4835.00 frames. ], tot_loss[loss=0.1923, simple_loss=0.2598, pruned_loss=0.0624, over 953758.38 frames. ], batch size: 49, lr: 3.69e-03, grad_scale: 32.0 2023-04-27 01:36:05,411 INFO [finetune.py:1001] (1/7) Computing validation loss 2023-04-27 01:36:11,621 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.2361, 2.5549, 0.9908, 1.3327, 1.9302, 1.3328, 3.0530, 1.7500], device='cuda:1'), covar=tensor([0.0602, 0.0574, 0.0790, 0.1347, 0.0439, 0.0928, 0.0278, 0.0578], device='cuda:1'), in_proj_covar=tensor([0.0052, 0.0067, 0.0049, 0.0047, 0.0051, 0.0053, 0.0078, 0.0052], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:1') 2023-04-27 01:36:27,793 INFO [finetune.py:1010] (1/7) Epoch 11, validation: loss=0.1531, simple_loss=0.2255, pruned_loss=0.04032, over 2265189.00 frames. 2023-04-27 01:36:27,793 INFO [finetune.py:1011] (1/7) Maximum memory allocated so far is 6391MB 2023-04-27 01:36:39,643 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.5052, 3.4060, 2.7094, 2.9449, 2.5935, 2.8715, 2.8145, 2.1084], device='cuda:1'), covar=tensor([0.2052, 0.1254, 0.0819, 0.1357, 0.2659, 0.1024, 0.2062, 0.2787], device='cuda:1'), in_proj_covar=tensor([0.0294, 0.0316, 0.0229, 0.0285, 0.0314, 0.0271, 0.0255, 0.0277], device='cuda:1'), out_proj_covar=tensor([1.1883e-04, 1.2696e-04, 9.1587e-05, 1.1421e-04, 1.2853e-04, 1.0869e-04, 1.0397e-04, 1.1093e-04], device='cuda:1') 2023-04-27 01:37:31,377 INFO [finetune.py:976] (1/7) Epoch 11, batch 3050, loss[loss=0.1494, simple_loss=0.2263, pruned_loss=0.0363, over 4788.00 frames. ], tot_loss[loss=0.1925, simple_loss=0.2602, pruned_loss=0.06245, over 953605.04 frames. ], batch size: 29, lr: 3.69e-03, grad_scale: 32.0 2023-04-27 01:37:42,869 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-04-27 01:37:45,692 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.005e+02 1.734e+02 2.169e+02 2.489e+02 4.339e+02, threshold=4.338e+02, percent-clipped=2.0 2023-04-27 01:38:26,535 INFO [finetune.py:976] (1/7) Epoch 11, batch 3100, loss[loss=0.2135, simple_loss=0.2709, pruned_loss=0.07799, over 4923.00 frames. ], tot_loss[loss=0.1905, simple_loss=0.2577, pruned_loss=0.06165, over 954370.14 frames. ], batch size: 38, lr: 3.69e-03, grad_scale: 32.0 2023-04-27 01:38:27,804 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.1178, 0.7789, 0.9482, 0.7989, 1.2540, 0.9682, 0.9193, 0.9889], device='cuda:1'), covar=tensor([0.1517, 0.1635, 0.1996, 0.1624, 0.1107, 0.1640, 0.1915, 0.2317], device='cuda:1'), in_proj_covar=tensor([0.0304, 0.0319, 0.0350, 0.0297, 0.0334, 0.0318, 0.0304, 0.0360], device='cuda:1'), out_proj_covar=tensor([6.3760e-05, 6.7513e-05, 7.5091e-05, 6.1023e-05, 6.9832e-05, 6.7815e-05, 6.5012e-05, 7.7292e-05], device='cuda:1') 2023-04-27 01:38:40,717 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=60396.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 01:38:48,251 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=60408.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 01:38:50,805 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-04-27 01:38:56,657 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=60421.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:39:00,267 INFO [finetune.py:976] (1/7) Epoch 11, batch 3150, loss[loss=0.1821, simple_loss=0.2567, pruned_loss=0.05379, over 4765.00 frames. ], tot_loss[loss=0.1892, simple_loss=0.2557, pruned_loss=0.06138, over 955463.04 frames. ], batch size: 26, lr: 3.69e-03, grad_scale: 32.0 2023-04-27 01:39:06,450 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=60434.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:39:09,789 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.219e+02 1.640e+02 1.939e+02 2.354e+02 4.857e+02, threshold=3.879e+02, percent-clipped=2.0 2023-04-27 01:39:21,383 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.4377, 1.9412, 2.4600, 3.0612, 2.3595, 1.8449, 1.6839, 2.2984], device='cuda:1'), covar=tensor([0.3585, 0.3481, 0.1627, 0.2523, 0.2999, 0.2744, 0.4191, 0.2418], device='cuda:1'), in_proj_covar=tensor([0.0282, 0.0246, 0.0219, 0.0312, 0.0212, 0.0226, 0.0229, 0.0183], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-27 01:39:26,233 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0997, 2.5723, 0.9258, 1.3986, 1.7788, 1.2892, 3.1041, 1.7486], device='cuda:1'), covar=tensor([0.0664, 0.0538, 0.0774, 0.1280, 0.0531, 0.1007, 0.0300, 0.0599], device='cuda:1'), in_proj_covar=tensor([0.0052, 0.0066, 0.0049, 0.0047, 0.0051, 0.0053, 0.0078, 0.0051], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:1') 2023-04-27 01:39:33,959 INFO [finetune.py:976] (1/7) Epoch 11, batch 3200, loss[loss=0.1815, simple_loss=0.2508, pruned_loss=0.05614, over 4709.00 frames. ], tot_loss[loss=0.1865, simple_loss=0.2526, pruned_loss=0.06016, over 956596.29 frames. ], batch size: 23, lr: 3.69e-03, grad_scale: 32.0 2023-04-27 01:39:53,427 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=60504.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:40:07,940 INFO [finetune.py:976] (1/7) Epoch 11, batch 3250, loss[loss=0.1715, simple_loss=0.248, pruned_loss=0.04748, over 4807.00 frames. ], tot_loss[loss=0.1878, simple_loss=0.2538, pruned_loss=0.06088, over 954876.46 frames. ], batch size: 45, lr: 3.69e-03, grad_scale: 32.0 2023-04-27 01:40:15,718 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.064e+02 1.675e+02 1.982e+02 2.288e+02 4.621e+02, threshold=3.964e+02, percent-clipped=1.0 2023-04-27 01:40:17,619 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-04-27 01:40:20,475 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=60545.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:40:25,156 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=60552.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:40:26,841 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=60554.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:40:42,227 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.2925, 1.3600, 3.7835, 3.5078, 3.3568, 3.6618, 3.6217, 3.3657], device='cuda:1'), covar=tensor([0.6700, 0.5278, 0.1019, 0.1820, 0.1176, 0.1584, 0.1753, 0.1279], device='cuda:1'), in_proj_covar=tensor([0.0313, 0.0309, 0.0404, 0.0410, 0.0351, 0.0409, 0.0315, 0.0369], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-27 01:40:52,025 INFO [finetune.py:976] (1/7) Epoch 11, batch 3300, loss[loss=0.2085, simple_loss=0.282, pruned_loss=0.06747, over 4906.00 frames. ], tot_loss[loss=0.1908, simple_loss=0.2573, pruned_loss=0.06217, over 953989.99 frames. ], batch size: 37, lr: 3.69e-03, grad_scale: 32.0 2023-04-27 01:41:12,294 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9988, 2.7218, 0.9654, 1.3920, 1.8506, 1.2221, 3.4415, 1.7690], device='cuda:1'), covar=tensor([0.0763, 0.0733, 0.0968, 0.1342, 0.0575, 0.1055, 0.0339, 0.0653], device='cuda:1'), in_proj_covar=tensor([0.0052, 0.0066, 0.0049, 0.0047, 0.0051, 0.0052, 0.0078, 0.0051], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:1') 2023-04-27 01:41:14,046 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=60593.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:41:58,102 INFO [finetune.py:976] (1/7) Epoch 11, batch 3350, loss[loss=0.2112, simple_loss=0.2931, pruned_loss=0.06466, over 4801.00 frames. ], tot_loss[loss=0.1934, simple_loss=0.2606, pruned_loss=0.06314, over 954317.70 frames. ], batch size: 51, lr: 3.69e-03, grad_scale: 32.0 2023-04-27 01:42:11,849 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.079e+02 1.646e+02 2.104e+02 2.646e+02 5.419e+02, threshold=4.209e+02, percent-clipped=4.0 2023-04-27 01:42:52,680 INFO [finetune.py:976] (1/7) Epoch 11, batch 3400, loss[loss=0.1887, simple_loss=0.2616, pruned_loss=0.05785, over 4930.00 frames. ], tot_loss[loss=0.1922, simple_loss=0.2593, pruned_loss=0.06253, over 951145.25 frames. ], batch size: 42, lr: 3.69e-03, grad_scale: 32.0 2023-04-27 01:42:52,857 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9150, 1.4528, 1.9508, 2.2821, 1.9766, 1.8391, 1.9022, 1.8897], device='cuda:1'), covar=tensor([0.5658, 0.7094, 0.7305, 0.8105, 0.6759, 0.9645, 0.9392, 0.7795], device='cuda:1'), in_proj_covar=tensor([0.0407, 0.0409, 0.0496, 0.0514, 0.0437, 0.0456, 0.0465, 0.0465], device='cuda:1'), out_proj_covar=tensor([9.8934e-05, 1.0132e-04, 1.1178e-04, 1.2193e-04, 1.0589e-04, 1.1026e-04, 1.1133e-04, 1.1164e-04], device='cuda:1') 2023-04-27 01:42:58,359 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-04-27 01:43:05,403 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=60696.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:43:14,141 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=60708.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:43:20,704 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=60718.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:43:22,463 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=60721.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:43:26,003 INFO [finetune.py:976] (1/7) Epoch 11, batch 3450, loss[loss=0.1934, simple_loss=0.247, pruned_loss=0.06991, over 4864.00 frames. ], tot_loss[loss=0.1918, simple_loss=0.2592, pruned_loss=0.06218, over 952623.40 frames. ], batch size: 31, lr: 3.69e-03, grad_scale: 32.0 2023-04-27 01:43:30,800 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=60734.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:43:31,417 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.4513, 2.2236, 2.4568, 2.8415, 2.8560, 2.2126, 1.7855, 2.5707], device='cuda:1'), covar=tensor([0.0734, 0.0952, 0.0581, 0.0545, 0.0519, 0.0861, 0.1016, 0.0503], device='cuda:1'), in_proj_covar=tensor([0.0199, 0.0205, 0.0185, 0.0177, 0.0181, 0.0191, 0.0161, 0.0186], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 01:43:33,701 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.084e+02 1.685e+02 2.029e+02 2.444e+02 3.873e+02, threshold=4.057e+02, percent-clipped=0.0 2023-04-27 01:43:36,747 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=60744.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:43:45,471 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=60756.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:43:54,319 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=60769.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:43:54,923 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5922, 1.6472, 4.4161, 4.1076, 3.8834, 4.1352, 4.0441, 3.8743], device='cuda:1'), covar=tensor([0.7237, 0.5192, 0.0939, 0.1643, 0.1111, 0.1472, 0.1713, 0.1446], device='cuda:1'), in_proj_covar=tensor([0.0315, 0.0312, 0.0408, 0.0412, 0.0355, 0.0412, 0.0318, 0.0372], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-27 01:43:59,111 INFO [finetune.py:976] (1/7) Epoch 11, batch 3500, loss[loss=0.1514, simple_loss=0.2261, pruned_loss=0.03829, over 4824.00 frames. ], tot_loss[loss=0.1892, simple_loss=0.2561, pruned_loss=0.06109, over 954964.08 frames. ], batch size: 30, lr: 3.69e-03, grad_scale: 32.0 2023-04-27 01:44:00,483 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=60779.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:44:01,727 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-04-27 01:44:02,193 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=60782.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:44:27,685 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.41 vs. limit=5.0 2023-04-27 01:44:28,238 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.77 vs. limit=2.0 2023-04-27 01:44:32,809 INFO [finetune.py:976] (1/7) Epoch 11, batch 3550, loss[loss=0.1593, simple_loss=0.2289, pruned_loss=0.04482, over 4850.00 frames. ], tot_loss[loss=0.1862, simple_loss=0.2528, pruned_loss=0.05979, over 954976.51 frames. ], batch size: 47, lr: 3.69e-03, grad_scale: 32.0 2023-04-27 01:44:40,088 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.124e+02 1.561e+02 1.887e+02 2.297e+02 4.767e+02, threshold=3.774e+02, percent-clipped=1.0 2023-04-27 01:44:44,359 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4984, 3.4271, 0.8050, 1.8864, 1.8977, 2.3734, 1.9688, 1.0450], device='cuda:1'), covar=tensor([0.1325, 0.0930, 0.2170, 0.1183, 0.1047, 0.1048, 0.1426, 0.2114], device='cuda:1'), in_proj_covar=tensor([0.0117, 0.0248, 0.0141, 0.0121, 0.0134, 0.0153, 0.0118, 0.0121], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-27 01:44:50,271 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=60854.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:45:06,039 INFO [finetune.py:976] (1/7) Epoch 11, batch 3600, loss[loss=0.1845, simple_loss=0.2512, pruned_loss=0.05895, over 4793.00 frames. ], tot_loss[loss=0.1844, simple_loss=0.2506, pruned_loss=0.05911, over 955669.87 frames. ], batch size: 51, lr: 3.69e-03, grad_scale: 32.0 2023-04-27 01:45:18,278 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-04-27 01:45:21,747 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=60902.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:45:39,840 INFO [finetune.py:976] (1/7) Epoch 11, batch 3650, loss[loss=0.2307, simple_loss=0.2991, pruned_loss=0.08113, over 4815.00 frames. ], tot_loss[loss=0.1868, simple_loss=0.2532, pruned_loss=0.06015, over 952994.68 frames. ], batch size: 40, lr: 3.69e-03, grad_scale: 32.0 2023-04-27 01:45:47,175 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.103e+02 1.725e+02 2.086e+02 2.393e+02 4.773e+02, threshold=4.172e+02, percent-clipped=3.0 2023-04-27 01:46:04,732 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.2996, 2.8043, 0.9535, 1.4304, 2.0586, 1.2891, 3.8006, 1.6596], device='cuda:1'), covar=tensor([0.0672, 0.1050, 0.0907, 0.1183, 0.0498, 0.0978, 0.0238, 0.0631], device='cuda:1'), in_proj_covar=tensor([0.0052, 0.0066, 0.0049, 0.0047, 0.0051, 0.0052, 0.0077, 0.0051], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:1') 2023-04-27 01:46:25,826 INFO [finetune.py:976] (1/7) Epoch 11, batch 3700, loss[loss=0.2113, simple_loss=0.2864, pruned_loss=0.06805, over 4806.00 frames. ], tot_loss[loss=0.189, simple_loss=0.2564, pruned_loss=0.06078, over 953754.11 frames. ], batch size: 51, lr: 3.69e-03, grad_scale: 32.0 2023-04-27 01:47:05,592 INFO [finetune.py:976] (1/7) Epoch 11, batch 3750, loss[loss=0.1793, simple_loss=0.2371, pruned_loss=0.06071, over 4290.00 frames. ], tot_loss[loss=0.1893, simple_loss=0.2567, pruned_loss=0.06097, over 952181.48 frames. ], batch size: 18, lr: 3.69e-03, grad_scale: 32.0 2023-04-27 01:47:18,727 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.172e+02 1.566e+02 1.879e+02 2.369e+02 3.505e+02, threshold=3.758e+02, percent-clipped=0.0 2023-04-27 01:47:20,755 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0230, 1.5782, 1.5735, 1.8723, 2.2301, 1.8064, 1.5225, 1.4716], device='cuda:1'), covar=tensor([0.1452, 0.1353, 0.1873, 0.1119, 0.0792, 0.1626, 0.1986, 0.1878], device='cuda:1'), in_proj_covar=tensor([0.0306, 0.0318, 0.0350, 0.0295, 0.0331, 0.0317, 0.0305, 0.0359], device='cuda:1'), out_proj_covar=tensor([6.4072e-05, 6.7233e-05, 7.5328e-05, 6.0703e-05, 6.9103e-05, 6.7556e-05, 6.5118e-05, 7.7168e-05], device='cuda:1') 2023-04-27 01:47:29,910 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7821, 1.4879, 1.3895, 1.7137, 2.0234, 1.6659, 1.4207, 1.3061], device='cuda:1'), covar=tensor([0.1345, 0.1453, 0.1757, 0.1183, 0.0791, 0.1729, 0.2087, 0.2196], device='cuda:1'), in_proj_covar=tensor([0.0306, 0.0318, 0.0350, 0.0295, 0.0331, 0.0317, 0.0305, 0.0359], device='cuda:1'), out_proj_covar=tensor([6.4105e-05, 6.7217e-05, 7.5355e-05, 6.0722e-05, 6.9122e-05, 6.7552e-05, 6.5112e-05, 7.7167e-05], device='cuda:1') 2023-04-27 01:47:40,541 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=61061.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:48:01,153 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=61074.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:48:05,286 INFO [finetune.py:976] (1/7) Epoch 11, batch 3800, loss[loss=0.2164, simple_loss=0.2848, pruned_loss=0.07404, over 4891.00 frames. ], tot_loss[loss=0.1901, simple_loss=0.258, pruned_loss=0.0611, over 953475.93 frames. ], batch size: 43, lr: 3.69e-03, grad_scale: 32.0 2023-04-27 01:48:56,479 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=61122.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:48:59,869 INFO [finetune.py:976] (1/7) Epoch 11, batch 3850, loss[loss=0.1767, simple_loss=0.2425, pruned_loss=0.05548, over 4905.00 frames. ], tot_loss[loss=0.1887, simple_loss=0.2567, pruned_loss=0.06033, over 956838.91 frames. ], batch size: 36, lr: 3.69e-03, grad_scale: 32.0 2023-04-27 01:49:08,085 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.118e+02 1.671e+02 1.909e+02 2.247e+02 3.528e+02, threshold=3.817e+02, percent-clipped=0.0 2023-04-27 01:49:33,129 INFO [finetune.py:976] (1/7) Epoch 11, batch 3900, loss[loss=0.2351, simple_loss=0.2954, pruned_loss=0.0874, over 4844.00 frames. ], tot_loss[loss=0.1871, simple_loss=0.2544, pruned_loss=0.05984, over 954900.36 frames. ], batch size: 49, lr: 3.69e-03, grad_scale: 32.0 2023-04-27 01:49:50,541 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0801, 2.5405, 1.0124, 1.3692, 1.8953, 1.2367, 3.2190, 1.7411], device='cuda:1'), covar=tensor([0.0656, 0.0649, 0.0823, 0.1217, 0.0497, 0.1002, 0.0279, 0.0614], device='cuda:1'), in_proj_covar=tensor([0.0051, 0.0066, 0.0049, 0.0047, 0.0051, 0.0052, 0.0077, 0.0051], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0007, 0.0008, 0.0008, 0.0010, 0.0007], device='cuda:1') 2023-04-27 01:49:57,116 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.1460, 1.3178, 1.5599, 1.7632, 1.6697, 1.8273, 1.6645, 1.6666], device='cuda:1'), covar=tensor([0.4266, 0.5971, 0.5144, 0.4839, 0.5950, 0.8271, 0.5508, 0.5082], device='cuda:1'), in_proj_covar=tensor([0.0324, 0.0376, 0.0314, 0.0325, 0.0338, 0.0398, 0.0356, 0.0320], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 01:50:05,886 INFO [finetune.py:976] (1/7) Epoch 11, batch 3950, loss[loss=0.1665, simple_loss=0.2302, pruned_loss=0.05142, over 4860.00 frames. ], tot_loss[loss=0.1843, simple_loss=0.2511, pruned_loss=0.05878, over 956027.24 frames. ], batch size: 44, lr: 3.68e-03, grad_scale: 32.0 2023-04-27 01:50:15,547 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.075e+02 1.581e+02 1.829e+02 2.221e+02 4.088e+02, threshold=3.658e+02, percent-clipped=2.0 2023-04-27 01:50:33,798 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.71 vs. limit=2.0 2023-04-27 01:50:39,618 INFO [finetune.py:976] (1/7) Epoch 11, batch 4000, loss[loss=0.1881, simple_loss=0.2601, pruned_loss=0.05799, over 4803.00 frames. ], tot_loss[loss=0.1837, simple_loss=0.2504, pruned_loss=0.05846, over 956812.38 frames. ], batch size: 51, lr: 3.68e-03, grad_scale: 32.0 2023-04-27 01:50:58,623 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8408, 2.3594, 1.8872, 2.1936, 1.7016, 1.9113, 1.8795, 1.4420], device='cuda:1'), covar=tensor([0.1831, 0.1212, 0.0867, 0.1046, 0.3142, 0.1234, 0.2031, 0.2689], device='cuda:1'), in_proj_covar=tensor([0.0291, 0.0313, 0.0225, 0.0282, 0.0312, 0.0270, 0.0253, 0.0274], device='cuda:1'), out_proj_covar=tensor([1.1756e-04, 1.2569e-04, 9.0175e-05, 1.1285e-04, 1.2734e-04, 1.0825e-04, 1.0309e-04, 1.0972e-04], device='cuda:1') 2023-04-27 01:51:00,496 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.1948, 1.6177, 1.4063, 1.7746, 1.7358, 1.8715, 1.3852, 3.3858], device='cuda:1'), covar=tensor([0.0647, 0.0762, 0.0790, 0.1215, 0.0587, 0.0514, 0.0778, 0.0154], device='cuda:1'), in_proj_covar=tensor([0.0038, 0.0039, 0.0040, 0.0043, 0.0040, 0.0038, 0.0039, 0.0057], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0014, 0.0013, 0.0012, 0.0013, 0.0016], device='cuda:1') 2023-04-27 01:51:18,202 INFO [finetune.py:976] (1/7) Epoch 11, batch 4050, loss[loss=0.1931, simple_loss=0.284, pruned_loss=0.05116, over 4814.00 frames. ], tot_loss[loss=0.1867, simple_loss=0.254, pruned_loss=0.05967, over 957044.28 frames. ], batch size: 45, lr: 3.68e-03, grad_scale: 32.0 2023-04-27 01:51:37,048 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.086e+02 1.723e+02 1.975e+02 2.601e+02 5.201e+02, threshold=3.950e+02, percent-clipped=3.0 2023-04-27 01:52:21,677 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=61374.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:52:23,440 INFO [finetune.py:976] (1/7) Epoch 11, batch 4100, loss[loss=0.1534, simple_loss=0.2313, pruned_loss=0.03772, over 4761.00 frames. ], tot_loss[loss=0.189, simple_loss=0.2567, pruned_loss=0.06066, over 955041.97 frames. ], batch size: 28, lr: 3.68e-03, grad_scale: 32.0 2023-04-27 01:53:16,504 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=61417.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:53:24,318 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4514, 0.6123, 1.3615, 1.8436, 1.5642, 1.3596, 1.3560, 1.4349], device='cuda:1'), covar=tensor([0.5077, 0.6934, 0.6534, 0.7341, 0.6164, 0.8168, 0.8288, 0.7757], device='cuda:1'), in_proj_covar=tensor([0.0412, 0.0411, 0.0500, 0.0517, 0.0441, 0.0461, 0.0472, 0.0470], device='cuda:1'), out_proj_covar=tensor([9.9957e-05, 1.0197e-04, 1.1278e-04, 1.2287e-04, 1.0686e-04, 1.1146e-04, 1.1298e-04, 1.1290e-04], device='cuda:1') 2023-04-27 01:53:25,485 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=61422.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:53:28,494 INFO [finetune.py:976] (1/7) Epoch 11, batch 4150, loss[loss=0.2802, simple_loss=0.3235, pruned_loss=0.1184, over 4169.00 frames. ], tot_loss[loss=0.1914, simple_loss=0.2591, pruned_loss=0.06189, over 954507.70 frames. ], batch size: 65, lr: 3.68e-03, grad_scale: 32.0 2023-04-27 01:53:28,711 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.12 vs. limit=2.0 2023-04-27 01:53:48,240 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.138e+02 1.660e+02 1.903e+02 2.358e+02 4.640e+02, threshold=3.807e+02, percent-clipped=2.0 2023-04-27 01:54:37,128 INFO [finetune.py:976] (1/7) Epoch 11, batch 4200, loss[loss=0.2245, simple_loss=0.2861, pruned_loss=0.0814, over 4810.00 frames. ], tot_loss[loss=0.1914, simple_loss=0.2593, pruned_loss=0.06182, over 954958.96 frames. ], batch size: 41, lr: 3.68e-03, grad_scale: 32.0 2023-04-27 01:55:45,102 INFO [finetune.py:976] (1/7) Epoch 11, batch 4250, loss[loss=0.1454, simple_loss=0.2108, pruned_loss=0.04005, over 4789.00 frames. ], tot_loss[loss=0.1896, simple_loss=0.2572, pruned_loss=0.06107, over 956331.91 frames. ], batch size: 29, lr: 3.68e-03, grad_scale: 32.0 2023-04-27 01:55:48,862 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5287, 1.2481, 4.4890, 4.1504, 3.9327, 4.2914, 4.1416, 3.9657], device='cuda:1'), covar=tensor([0.7353, 0.6418, 0.0994, 0.1828, 0.1060, 0.1496, 0.1437, 0.1497], device='cuda:1'), in_proj_covar=tensor([0.0313, 0.0310, 0.0405, 0.0412, 0.0354, 0.0411, 0.0317, 0.0373], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-27 01:55:54,439 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.1347, 2.6984, 2.0881, 2.4751, 1.8362, 2.1253, 2.2934, 1.5074], device='cuda:1'), covar=tensor([0.1979, 0.1129, 0.0923, 0.1130, 0.2974, 0.1292, 0.1795, 0.2923], device='cuda:1'), in_proj_covar=tensor([0.0293, 0.0315, 0.0227, 0.0284, 0.0313, 0.0271, 0.0254, 0.0276], device='cuda:1'), out_proj_covar=tensor([1.1848e-04, 1.2663e-04, 9.0938e-05, 1.1359e-04, 1.2808e-04, 1.0888e-04, 1.0350e-04, 1.1050e-04], device='cuda:1') 2023-04-27 01:55:55,050 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.2349, 1.4538, 1.6582, 1.7986, 1.6641, 1.7743, 1.7524, 1.6673], device='cuda:1'), covar=tensor([0.4520, 0.6098, 0.5315, 0.4992, 0.6153, 0.8540, 0.6275, 0.5899], device='cuda:1'), in_proj_covar=tensor([0.0324, 0.0375, 0.0314, 0.0324, 0.0338, 0.0397, 0.0356, 0.0321], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 01:55:57,237 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.015e+02 1.633e+02 1.870e+02 2.370e+02 4.371e+02, threshold=3.739e+02, percent-clipped=2.0 2023-04-27 01:56:49,826 INFO [finetune.py:976] (1/7) Epoch 11, batch 4300, loss[loss=0.1786, simple_loss=0.2416, pruned_loss=0.0578, over 4932.00 frames. ], tot_loss[loss=0.1867, simple_loss=0.2537, pruned_loss=0.05984, over 957103.12 frames. ], batch size: 38, lr: 3.68e-03, grad_scale: 32.0 2023-04-27 01:57:04,116 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.68 vs. limit=5.0 2023-04-27 01:57:47,102 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-04-27 01:57:57,464 INFO [finetune.py:976] (1/7) Epoch 11, batch 4350, loss[loss=0.1516, simple_loss=0.2242, pruned_loss=0.03951, over 4767.00 frames. ], tot_loss[loss=0.1839, simple_loss=0.2503, pruned_loss=0.05879, over 956162.03 frames. ], batch size: 26, lr: 3.68e-03, grad_scale: 32.0 2023-04-27 01:58:10,463 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.036e+02 1.619e+02 1.912e+02 2.174e+02 4.082e+02, threshold=3.823e+02, percent-clipped=2.0 2023-04-27 01:58:16,930 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 2023-04-27 01:58:49,420 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4020, 1.3350, 1.7962, 1.6615, 1.2961, 1.1146, 1.4392, 0.9249], device='cuda:1'), covar=tensor([0.0807, 0.0829, 0.0483, 0.0906, 0.0865, 0.1215, 0.0749, 0.0913], device='cuda:1'), in_proj_covar=tensor([0.0068, 0.0072, 0.0071, 0.0067, 0.0075, 0.0097, 0.0076, 0.0072], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-27 01:59:02,120 INFO [finetune.py:976] (1/7) Epoch 11, batch 4400, loss[loss=0.2293, simple_loss=0.2827, pruned_loss=0.08796, over 4801.00 frames. ], tot_loss[loss=0.1862, simple_loss=0.2521, pruned_loss=0.06012, over 954984.43 frames. ], batch size: 29, lr: 3.68e-03, grad_scale: 32.0 2023-04-27 01:59:23,300 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.2060, 1.6994, 2.0601, 2.4726, 2.0611, 1.5959, 1.2981, 1.8911], device='cuda:1'), covar=tensor([0.3641, 0.3669, 0.1855, 0.2815, 0.3047, 0.3091, 0.4580, 0.2286], device='cuda:1'), in_proj_covar=tensor([0.0285, 0.0249, 0.0221, 0.0316, 0.0214, 0.0227, 0.0231, 0.0186], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-27 01:59:57,221 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=61717.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:00:08,835 INFO [finetune.py:976] (1/7) Epoch 11, batch 4450, loss[loss=0.1716, simple_loss=0.2418, pruned_loss=0.05072, over 4736.00 frames. ], tot_loss[loss=0.189, simple_loss=0.2558, pruned_loss=0.06106, over 954673.98 frames. ], batch size: 27, lr: 3.68e-03, grad_scale: 32.0 2023-04-27 02:00:17,325 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=61732.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:00:21,987 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.107e+02 1.778e+02 2.036e+02 2.485e+02 5.642e+02, threshold=4.071e+02, percent-clipped=3.0 2023-04-27 02:00:24,570 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=61743.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:00:45,386 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=61765.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:01:03,250 INFO [finetune.py:976] (1/7) Epoch 11, batch 4500, loss[loss=0.1685, simple_loss=0.2494, pruned_loss=0.04374, over 4786.00 frames. ], tot_loss[loss=0.1915, simple_loss=0.2585, pruned_loss=0.0622, over 956521.00 frames. ], batch size: 51, lr: 3.68e-03, grad_scale: 32.0 2023-04-27 02:01:19,132 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=61793.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:01:26,330 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=61804.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:01:42,632 INFO [finetune.py:976] (1/7) Epoch 11, batch 4550, loss[loss=0.2151, simple_loss=0.2831, pruned_loss=0.07352, over 4755.00 frames. ], tot_loss[loss=0.1932, simple_loss=0.2602, pruned_loss=0.06306, over 956160.01 frames. ], batch size: 54, lr: 3.68e-03, grad_scale: 32.0 2023-04-27 02:01:49,951 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.176e+02 1.694e+02 2.002e+02 2.452e+02 5.471e+02, threshold=4.003e+02, percent-clipped=1.0 2023-04-27 02:02:16,332 INFO [finetune.py:976] (1/7) Epoch 11, batch 4600, loss[loss=0.1739, simple_loss=0.2367, pruned_loss=0.0556, over 4827.00 frames. ], tot_loss[loss=0.1914, simple_loss=0.2587, pruned_loss=0.06205, over 955664.96 frames. ], batch size: 30, lr: 3.68e-03, grad_scale: 32.0 2023-04-27 02:02:24,251 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9376, 2.2628, 0.9310, 1.2143, 1.5419, 1.1476, 2.5240, 1.4032], device='cuda:1'), covar=tensor([0.0709, 0.0648, 0.0663, 0.1228, 0.0442, 0.0985, 0.0313, 0.0667], device='cuda:1'), in_proj_covar=tensor([0.0052, 0.0066, 0.0049, 0.0047, 0.0051, 0.0053, 0.0077, 0.0051], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:1') 2023-04-27 02:02:40,213 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=61914.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:02:49,392 INFO [finetune.py:976] (1/7) Epoch 11, batch 4650, loss[loss=0.1658, simple_loss=0.2375, pruned_loss=0.04698, over 4888.00 frames. ], tot_loss[loss=0.1894, simple_loss=0.2559, pruned_loss=0.06146, over 955476.01 frames. ], batch size: 35, lr: 3.68e-03, grad_scale: 32.0 2023-04-27 02:02:50,761 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8945, 2.3734, 1.8693, 1.7973, 1.4793, 1.4770, 1.9203, 1.4048], device='cuda:1'), covar=tensor([0.1379, 0.1294, 0.1380, 0.1648, 0.2179, 0.1764, 0.0936, 0.1847], device='cuda:1'), in_proj_covar=tensor([0.0195, 0.0214, 0.0169, 0.0203, 0.0202, 0.0183, 0.0158, 0.0186], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-27 02:02:55,580 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.1375, 1.3478, 1.5726, 1.7479, 1.6664, 1.8467, 1.6493, 1.6017], device='cuda:1'), covar=tensor([0.4060, 0.5523, 0.4880, 0.4364, 0.5374, 0.7452, 0.5302, 0.5018], device='cuda:1'), in_proj_covar=tensor([0.0325, 0.0375, 0.0313, 0.0325, 0.0337, 0.0398, 0.0356, 0.0320], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 02:02:56,646 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.191e+02 1.561e+02 1.984e+02 2.276e+02 4.966e+02, threshold=3.968e+02, percent-clipped=2.0 2023-04-27 02:03:20,499 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=61975.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:03:22,011 INFO [finetune.py:976] (1/7) Epoch 11, batch 4700, loss[loss=0.1722, simple_loss=0.2282, pruned_loss=0.05812, over 4917.00 frames. ], tot_loss[loss=0.186, simple_loss=0.2518, pruned_loss=0.06008, over 954475.19 frames. ], batch size: 43, lr: 3.68e-03, grad_scale: 32.0 2023-04-27 02:03:23,376 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6922, 1.1734, 1.7944, 2.1896, 1.8001, 1.6752, 1.7394, 1.7328], device='cuda:1'), covar=tensor([0.5096, 0.7329, 0.7486, 0.6722, 0.6686, 0.8644, 0.8816, 0.7960], device='cuda:1'), in_proj_covar=tensor([0.0411, 0.0411, 0.0499, 0.0515, 0.0442, 0.0460, 0.0470, 0.0470], device='cuda:1'), out_proj_covar=tensor([9.9877e-05, 1.0195e-04, 1.1259e-04, 1.2248e-04, 1.0699e-04, 1.1128e-04, 1.1258e-04, 1.1275e-04], device='cuda:1') 2023-04-27 02:04:06,619 INFO [finetune.py:976] (1/7) Epoch 11, batch 4750, loss[loss=0.224, simple_loss=0.298, pruned_loss=0.07498, over 4833.00 frames. ], tot_loss[loss=0.1864, simple_loss=0.2518, pruned_loss=0.06053, over 954382.23 frames. ], batch size: 51, lr: 3.68e-03, grad_scale: 32.0 2023-04-27 02:04:20,378 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.139e+02 1.587e+02 1.879e+02 2.301e+02 4.862e+02, threshold=3.757e+02, percent-clipped=3.0 2023-04-27 02:04:28,550 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7617, 1.2537, 1.5784, 1.6820, 1.5589, 1.2596, 0.7263, 1.3103], device='cuda:1'), covar=tensor([0.3438, 0.3560, 0.1834, 0.2285, 0.2695, 0.2701, 0.4727, 0.2344], device='cuda:1'), in_proj_covar=tensor([0.0286, 0.0250, 0.0222, 0.0317, 0.0215, 0.0228, 0.0232, 0.0187], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-27 02:04:30,363 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.0267, 0.5839, 0.8905, 0.6811, 1.1870, 0.9396, 0.7962, 0.9263], device='cuda:1'), covar=tensor([0.1788, 0.1820, 0.2268, 0.1845, 0.1289, 0.1409, 0.2017, 0.2213], device='cuda:1'), in_proj_covar=tensor([0.0306, 0.0320, 0.0351, 0.0295, 0.0332, 0.0317, 0.0306, 0.0359], device='cuda:1'), out_proj_covar=tensor([6.4104e-05, 6.7432e-05, 7.5574e-05, 6.0671e-05, 6.9433e-05, 6.7577e-05, 6.5401e-05, 7.7081e-05], device='cuda:1') 2023-04-27 02:04:47,589 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5430, 2.5523, 2.1168, 2.2529, 2.6944, 2.2767, 3.5316, 2.0694], device='cuda:1'), covar=tensor([0.4100, 0.2293, 0.4665, 0.3359, 0.1768, 0.2820, 0.1454, 0.4197], device='cuda:1'), in_proj_covar=tensor([0.0344, 0.0351, 0.0429, 0.0362, 0.0386, 0.0384, 0.0382, 0.0424], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 02:04:50,599 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.60 vs. limit=2.0 2023-04-27 02:04:56,946 INFO [finetune.py:976] (1/7) Epoch 11, batch 4800, loss[loss=0.1885, simple_loss=0.2486, pruned_loss=0.06423, over 4889.00 frames. ], tot_loss[loss=0.1882, simple_loss=0.2548, pruned_loss=0.06078, over 955367.10 frames. ], batch size: 32, lr: 3.68e-03, grad_scale: 32.0 2023-04-27 02:05:05,179 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=62088.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:05:11,913 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=62099.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:05:20,883 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9190, 1.8759, 2.1170, 2.4189, 2.4146, 1.8508, 1.6289, 2.1574], device='cuda:1'), covar=tensor([0.1039, 0.1034, 0.0651, 0.0553, 0.0622, 0.1023, 0.0948, 0.0614], device='cuda:1'), in_proj_covar=tensor([0.0197, 0.0204, 0.0184, 0.0175, 0.0180, 0.0189, 0.0160, 0.0185], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 02:05:25,063 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-04-27 02:05:27,339 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=62122.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:05:30,801 INFO [finetune.py:976] (1/7) Epoch 11, batch 4850, loss[loss=0.2059, simple_loss=0.2762, pruned_loss=0.06777, over 4823.00 frames. ], tot_loss[loss=0.19, simple_loss=0.2577, pruned_loss=0.06117, over 956066.68 frames. ], batch size: 39, lr: 3.68e-03, grad_scale: 32.0 2023-04-27 02:05:39,086 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.265e+02 1.826e+02 2.171e+02 2.650e+02 4.437e+02, threshold=4.341e+02, percent-clipped=4.0 2023-04-27 02:05:59,040 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.52 vs. limit=2.0 2023-04-27 02:06:08,686 INFO [finetune.py:976] (1/7) Epoch 11, batch 4900, loss[loss=0.2243, simple_loss=0.2783, pruned_loss=0.08515, over 4762.00 frames. ], tot_loss[loss=0.1909, simple_loss=0.2585, pruned_loss=0.06168, over 954412.02 frames. ], batch size: 26, lr: 3.68e-03, grad_scale: 32.0 2023-04-27 02:06:18,150 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=62183.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:06:20,023 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.55 vs. limit=2.0 2023-04-27 02:06:52,810 INFO [finetune.py:976] (1/7) Epoch 11, batch 4950, loss[loss=0.2514, simple_loss=0.3029, pruned_loss=0.09992, over 4849.00 frames. ], tot_loss[loss=0.1912, simple_loss=0.2587, pruned_loss=0.06186, over 952926.87 frames. ], batch size: 31, lr: 3.68e-03, grad_scale: 64.0 2023-04-27 02:07:00,209 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.77 vs. limit=2.0 2023-04-27 02:07:01,570 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.127e+02 1.624e+02 1.966e+02 2.483e+02 3.537e+02, threshold=3.932e+02, percent-clipped=0.0 2023-04-27 02:07:21,451 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=62270.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:07:23,322 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=62273.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:07:26,147 INFO [finetune.py:976] (1/7) Epoch 11, batch 5000, loss[loss=0.1655, simple_loss=0.2355, pruned_loss=0.04773, over 4816.00 frames. ], tot_loss[loss=0.1899, simple_loss=0.2573, pruned_loss=0.06131, over 952570.23 frames. ], batch size: 39, lr: 3.68e-03, grad_scale: 32.0 2023-04-27 02:07:34,580 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.1049, 1.4692, 1.6179, 1.7669, 2.2141, 1.8964, 1.4998, 1.5005], device='cuda:1'), covar=tensor([0.1576, 0.1875, 0.2027, 0.1416, 0.0973, 0.1448, 0.2224, 0.2185], device='cuda:1'), in_proj_covar=tensor([0.0304, 0.0318, 0.0350, 0.0293, 0.0330, 0.0315, 0.0305, 0.0357], device='cuda:1'), out_proj_covar=tensor([6.3586e-05, 6.7110e-05, 7.5235e-05, 6.0145e-05, 6.8833e-05, 6.7190e-05, 6.5173e-05, 7.6637e-05], device='cuda:1') 2023-04-27 02:07:53,044 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.1032, 2.1115, 1.7458, 1.7037, 2.1539, 1.6999, 2.7360, 1.5725], device='cuda:1'), covar=tensor([0.3772, 0.1964, 0.4833, 0.2964, 0.1871, 0.2759, 0.1402, 0.4713], device='cuda:1'), in_proj_covar=tensor([0.0341, 0.0348, 0.0425, 0.0358, 0.0384, 0.0380, 0.0377, 0.0419], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 02:07:58,882 INFO [finetune.py:976] (1/7) Epoch 11, batch 5050, loss[loss=0.1557, simple_loss=0.2269, pruned_loss=0.04224, over 4841.00 frames. ], tot_loss[loss=0.1883, simple_loss=0.255, pruned_loss=0.06078, over 952989.08 frames. ], batch size: 25, lr: 3.68e-03, grad_scale: 32.0 2023-04-27 02:08:01,402 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2023-04-27 02:08:03,182 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.65 vs. limit=2.0 2023-04-27 02:08:04,231 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=62334.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:08:08,227 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.081e+02 1.670e+02 2.040e+02 2.398e+02 4.126e+02, threshold=4.080e+02, percent-clipped=2.0 2023-04-27 02:08:24,425 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-04-27 02:08:27,326 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.3467, 1.5292, 1.4693, 1.8428, 1.6772, 2.0507, 1.3718, 3.4176], device='cuda:1'), covar=tensor([0.0631, 0.0848, 0.0820, 0.1145, 0.0635, 0.0446, 0.0770, 0.0158], device='cuda:1'), in_proj_covar=tensor([0.0038, 0.0038, 0.0040, 0.0043, 0.0039, 0.0038, 0.0038, 0.0057], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0014, 0.0013, 0.0012, 0.0013, 0.0016], device='cuda:1') 2023-04-27 02:08:32,082 INFO [finetune.py:976] (1/7) Epoch 11, batch 5100, loss[loss=0.1691, simple_loss=0.2391, pruned_loss=0.04954, over 4814.00 frames. ], tot_loss[loss=0.1865, simple_loss=0.2528, pruned_loss=0.06009, over 954962.22 frames. ], batch size: 41, lr: 3.68e-03, grad_scale: 32.0 2023-04-27 02:08:40,374 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=62388.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:08:42,198 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=62391.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:08:47,532 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=62399.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:09:06,023 INFO [finetune.py:976] (1/7) Epoch 11, batch 5150, loss[loss=0.1877, simple_loss=0.2411, pruned_loss=0.06718, over 4133.00 frames. ], tot_loss[loss=0.1878, simple_loss=0.2533, pruned_loss=0.06118, over 953538.78 frames. ], batch size: 18, lr: 3.67e-03, grad_scale: 32.0 2023-04-27 02:09:12,052 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=62436.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:09:14,302 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([3.8662, 3.7349, 2.8417, 4.4802, 3.8345, 3.8391, 1.7489, 3.8817], device='cuda:1'), covar=tensor([0.1887, 0.1445, 0.4202, 0.1454, 0.4728, 0.2226, 0.6327, 0.2457], device='cuda:1'), in_proj_covar=tensor([0.0244, 0.0218, 0.0252, 0.0306, 0.0299, 0.0249, 0.0270, 0.0270], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 02:09:14,848 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.955e+01 1.646e+02 2.010e+02 2.557e+02 5.535e+02, threshold=4.020e+02, percent-clipped=1.0 2023-04-27 02:09:25,275 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=62447.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:09:33,233 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=62452.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:10:01,359 INFO [finetune.py:976] (1/7) Epoch 11, batch 5200, loss[loss=0.2147, simple_loss=0.2853, pruned_loss=0.07204, over 4095.00 frames. ], tot_loss[loss=0.1913, simple_loss=0.2576, pruned_loss=0.06249, over 952984.72 frames. ], batch size: 65, lr: 3.67e-03, grad_scale: 32.0 2023-04-27 02:10:02,044 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=62478.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:10:26,480 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.3742, 1.0866, 4.1345, 3.8545, 3.6845, 3.9250, 3.8637, 3.6785], device='cuda:1'), covar=tensor([0.7175, 0.6156, 0.1048, 0.1790, 0.1130, 0.1706, 0.1332, 0.1474], device='cuda:1'), in_proj_covar=tensor([0.0308, 0.0305, 0.0400, 0.0406, 0.0348, 0.0405, 0.0311, 0.0368], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-27 02:10:48,050 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.79 vs. limit=2.0 2023-04-27 02:10:54,519 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7730, 2.1405, 1.2318, 1.4258, 2.2576, 1.6481, 1.5806, 1.5887], device='cuda:1'), covar=tensor([0.0506, 0.0331, 0.0301, 0.0556, 0.0240, 0.0521, 0.0496, 0.0580], device='cuda:1'), in_proj_covar=tensor([0.0030, 0.0025, 0.0023, 0.0030, 0.0020, 0.0029, 0.0029, 0.0029], device='cuda:1'), out_proj_covar=tensor([0.0050, 0.0045, 0.0038, 0.0050, 0.0037, 0.0048, 0.0048, 0.0050], device='cuda:1') 2023-04-27 02:10:58,656 INFO [finetune.py:976] (1/7) Epoch 11, batch 5250, loss[loss=0.1708, simple_loss=0.2539, pruned_loss=0.04389, over 4308.00 frames. ], tot_loss[loss=0.1934, simple_loss=0.2601, pruned_loss=0.06336, over 953267.61 frames. ], batch size: 65, lr: 3.67e-03, grad_scale: 32.0 2023-04-27 02:11:07,063 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.221e+02 1.615e+02 2.039e+02 2.344e+02 5.619e+02, threshold=4.078e+02, percent-clipped=3.0 2023-04-27 02:11:10,690 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=62545.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:11:28,230 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=62570.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:11:32,447 INFO [finetune.py:976] (1/7) Epoch 11, batch 5300, loss[loss=0.2149, simple_loss=0.2755, pruned_loss=0.07718, over 4803.00 frames. ], tot_loss[loss=0.1936, simple_loss=0.2602, pruned_loss=0.06348, over 950765.11 frames. ], batch size: 51, lr: 3.67e-03, grad_scale: 32.0 2023-04-27 02:11:37,332 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=62585.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:11:52,273 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=62606.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:12:00,033 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=62618.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:12:03,572 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.2053, 2.4925, 1.3612, 1.5696, 2.1915, 1.4767, 3.2244, 1.9252], device='cuda:1'), covar=tensor([0.0562, 0.0734, 0.0718, 0.1041, 0.0404, 0.0794, 0.0222, 0.0502], device='cuda:1'), in_proj_covar=tensor([0.0052, 0.0067, 0.0049, 0.0047, 0.0051, 0.0053, 0.0077, 0.0051], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:1') 2023-04-27 02:12:05,914 INFO [finetune.py:976] (1/7) Epoch 11, batch 5350, loss[loss=0.1877, simple_loss=0.2538, pruned_loss=0.06082, over 4723.00 frames. ], tot_loss[loss=0.1914, simple_loss=0.2586, pruned_loss=0.06211, over 951565.27 frames. ], batch size: 54, lr: 3.67e-03, grad_scale: 32.0 2023-04-27 02:12:07,195 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=62629.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:12:13,875 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.227e+02 1.690e+02 2.017e+02 2.486e+02 5.260e+02, threshold=4.033e+02, percent-clipped=4.0 2023-04-27 02:12:18,147 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=62646.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:12:32,052 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=62666.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:12:39,515 INFO [finetune.py:976] (1/7) Epoch 11, batch 5400, loss[loss=0.1786, simple_loss=0.2448, pruned_loss=0.05616, over 4934.00 frames. ], tot_loss[loss=0.1871, simple_loss=0.2546, pruned_loss=0.05979, over 954573.28 frames. ], batch size: 33, lr: 3.67e-03, grad_scale: 32.0 2023-04-27 02:13:12,193 INFO [finetune.py:976] (1/7) Epoch 11, batch 5450, loss[loss=0.197, simple_loss=0.2541, pruned_loss=0.06994, over 4929.00 frames. ], tot_loss[loss=0.1855, simple_loss=0.2522, pruned_loss=0.05935, over 954363.87 frames. ], batch size: 46, lr: 3.67e-03, grad_scale: 32.0 2023-04-27 02:13:12,306 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=62727.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:13:20,548 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.017e+02 1.672e+02 1.919e+02 2.198e+02 3.769e+02, threshold=3.837e+02, percent-clipped=0.0 2023-04-27 02:13:24,917 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=62747.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:13:32,203 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.6695, 1.6929, 1.8001, 2.4694, 2.6083, 2.1902, 2.1301, 1.8826], device='cuda:1'), covar=tensor([0.1463, 0.1825, 0.2033, 0.1456, 0.1133, 0.2083, 0.2417, 0.2149], device='cuda:1'), in_proj_covar=tensor([0.0305, 0.0319, 0.0350, 0.0295, 0.0331, 0.0316, 0.0305, 0.0358], device='cuda:1'), out_proj_covar=tensor([6.3838e-05, 6.7097e-05, 7.5322e-05, 6.0558e-05, 6.9040e-05, 6.7467e-05, 6.5136e-05, 7.6817e-05], device='cuda:1') 2023-04-27 02:13:42,646 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=62773.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:13:45,485 INFO [finetune.py:976] (1/7) Epoch 11, batch 5500, loss[loss=0.2139, simple_loss=0.2759, pruned_loss=0.07593, over 4804.00 frames. ], tot_loss[loss=0.1836, simple_loss=0.2496, pruned_loss=0.05875, over 953917.45 frames. ], batch size: 51, lr: 3.67e-03, grad_scale: 32.0 2023-04-27 02:13:46,157 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=62778.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:14:17,883 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=62826.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:14:18,931 INFO [finetune.py:976] (1/7) Epoch 11, batch 5550, loss[loss=0.1814, simple_loss=0.2291, pruned_loss=0.0669, over 4407.00 frames. ], tot_loss[loss=0.1851, simple_loss=0.2515, pruned_loss=0.05937, over 953986.47 frames. ], batch size: 19, lr: 3.67e-03, grad_scale: 32.0 2023-04-27 02:14:20,268 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.3636, 2.3242, 2.6661, 2.9990, 2.8645, 2.3314, 2.0622, 2.5473], device='cuda:1'), covar=tensor([0.0994, 0.0928, 0.0575, 0.0614, 0.0598, 0.0971, 0.0808, 0.0624], device='cuda:1'), in_proj_covar=tensor([0.0198, 0.0204, 0.0184, 0.0175, 0.0179, 0.0189, 0.0159, 0.0185], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 02:14:23,728 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=62834.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:14:27,256 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.295e+02 1.658e+02 2.068e+02 2.664e+02 6.166e+02, threshold=4.137e+02, percent-clipped=3.0 2023-04-27 02:14:53,937 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8126, 1.2963, 1.8502, 2.2426, 1.8758, 1.7705, 1.8246, 1.8074], device='cuda:1'), covar=tensor([0.5419, 0.7115, 0.7709, 0.7139, 0.7072, 0.8850, 0.8811, 0.8321], device='cuda:1'), in_proj_covar=tensor([0.0407, 0.0407, 0.0493, 0.0512, 0.0437, 0.0456, 0.0464, 0.0465], device='cuda:1'), out_proj_covar=tensor([9.8881e-05, 1.0064e-04, 1.1131e-04, 1.2155e-04, 1.0603e-04, 1.1012e-04, 1.1133e-04, 1.1156e-04], device='cuda:1') 2023-04-27 02:15:05,418 INFO [finetune.py:976] (1/7) Epoch 11, batch 5600, loss[loss=0.1574, simple_loss=0.2358, pruned_loss=0.03948, over 4791.00 frames. ], tot_loss[loss=0.1885, simple_loss=0.256, pruned_loss=0.06045, over 952098.12 frames. ], batch size: 26, lr: 3.67e-03, grad_scale: 32.0 2023-04-27 02:15:37,389 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=62901.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:16:10,654 INFO [finetune.py:976] (1/7) Epoch 11, batch 5650, loss[loss=0.1461, simple_loss=0.2108, pruned_loss=0.04071, over 4700.00 frames. ], tot_loss[loss=0.1898, simple_loss=0.2583, pruned_loss=0.06061, over 952010.92 frames. ], batch size: 23, lr: 3.67e-03, grad_scale: 32.0 2023-04-27 02:16:11,896 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.1095, 1.5548, 1.4258, 1.7072, 1.6538, 2.1761, 1.3866, 3.6458], device='cuda:1'), covar=tensor([0.0632, 0.0777, 0.0797, 0.1200, 0.0636, 0.0462, 0.0787, 0.0131], device='cuda:1'), in_proj_covar=tensor([0.0038, 0.0038, 0.0040, 0.0043, 0.0040, 0.0038, 0.0038, 0.0057], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0014, 0.0013, 0.0012, 0.0013, 0.0016], device='cuda:1') 2023-04-27 02:16:11,903 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=62929.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:16:20,564 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.6191, 2.9718, 1.0995, 1.8680, 2.4031, 1.9476, 4.5904, 2.3329], device='cuda:1'), covar=tensor([0.0634, 0.0757, 0.0848, 0.1247, 0.0538, 0.0878, 0.0382, 0.0599], device='cuda:1'), in_proj_covar=tensor([0.0052, 0.0066, 0.0049, 0.0047, 0.0051, 0.0052, 0.0077, 0.0051], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:1') 2023-04-27 02:16:23,344 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.053e+02 1.567e+02 1.821e+02 2.303e+02 3.535e+02, threshold=3.642e+02, percent-clipped=0.0 2023-04-27 02:16:23,982 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=62941.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:17:06,580 INFO [finetune.py:976] (1/7) Epoch 11, batch 5700, loss[loss=0.1389, simple_loss=0.1951, pruned_loss=0.04137, over 4307.00 frames. ], tot_loss[loss=0.1877, simple_loss=0.2548, pruned_loss=0.06027, over 935235.68 frames. ], batch size: 19, lr: 3.67e-03, grad_scale: 32.0 2023-04-27 02:17:06,614 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=62977.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:17:38,294 INFO [finetune.py:976] (1/7) Epoch 12, batch 0, loss[loss=0.1679, simple_loss=0.2409, pruned_loss=0.04745, over 4819.00 frames. ], tot_loss[loss=0.1679, simple_loss=0.2409, pruned_loss=0.04745, over 4819.00 frames. ], batch size: 25, lr: 3.67e-03, grad_scale: 32.0 2023-04-27 02:17:38,294 INFO [finetune.py:1001] (1/7) Computing validation loss 2023-04-27 02:17:47,708 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.2243, 1.4126, 1.6557, 1.8388, 1.7450, 1.9001, 1.6718, 1.7009], device='cuda:1'), covar=tensor([0.4515, 0.6163, 0.5627, 0.4939, 0.6354, 0.8532, 0.6079, 0.5912], device='cuda:1'), in_proj_covar=tensor([0.0327, 0.0379, 0.0315, 0.0327, 0.0339, 0.0400, 0.0359, 0.0324], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 02:17:53,690 INFO [finetune.py:1010] (1/7) Epoch 12, validation: loss=0.1544, simple_loss=0.2267, pruned_loss=0.04099, over 2265189.00 frames. 2023-04-27 02:17:53,691 INFO [finetune.py:1011] (1/7) Maximum memory allocated so far is 6391MB 2023-04-27 02:18:18,012 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=63022.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:18:39,643 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.931e+01 1.655e+02 2.029e+02 2.560e+02 6.942e+02, threshold=4.058e+02, percent-clipped=5.0 2023-04-27 02:18:44,483 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=63047.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:18:54,571 INFO [finetune.py:976] (1/7) Epoch 12, batch 50, loss[loss=0.1669, simple_loss=0.2433, pruned_loss=0.04524, over 4907.00 frames. ], tot_loss[loss=0.1935, simple_loss=0.2601, pruned_loss=0.06345, over 212941.62 frames. ], batch size: 36, lr: 3.67e-03, grad_scale: 32.0 2023-04-27 02:19:04,179 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.1290, 2.4485, 0.9603, 1.3524, 1.8386, 1.2031, 3.4011, 1.7585], device='cuda:1'), covar=tensor([0.0689, 0.0666, 0.0829, 0.1308, 0.0536, 0.1038, 0.0290, 0.0661], device='cuda:1'), in_proj_covar=tensor([0.0051, 0.0066, 0.0049, 0.0047, 0.0050, 0.0052, 0.0077, 0.0051], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0007, 0.0008, 0.0008, 0.0010, 0.0007], device='cuda:1') 2023-04-27 02:19:30,951 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=63095.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:19:37,860 INFO [finetune.py:976] (1/7) Epoch 12, batch 100, loss[loss=0.1923, simple_loss=0.2531, pruned_loss=0.06581, over 4799.00 frames. ], tot_loss[loss=0.1882, simple_loss=0.2545, pruned_loss=0.06096, over 378688.93 frames. ], batch size: 51, lr: 3.67e-03, grad_scale: 32.0 2023-04-27 02:19:54,523 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=63129.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:20:01,184 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.001e+02 1.674e+02 1.936e+02 2.495e+02 3.786e+02, threshold=3.872e+02, percent-clipped=0.0 2023-04-27 02:20:03,605 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.0927, 1.4739, 1.3500, 1.7685, 1.6531, 1.7994, 1.3320, 3.0125], device='cuda:1'), covar=tensor([0.0653, 0.0779, 0.0793, 0.1109, 0.0590, 0.0484, 0.0725, 0.0175], device='cuda:1'), in_proj_covar=tensor([0.0038, 0.0038, 0.0040, 0.0043, 0.0039, 0.0038, 0.0038, 0.0057], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0014, 0.0013, 0.0012, 0.0013, 0.0016], device='cuda:1') 2023-04-27 02:20:11,745 INFO [finetune.py:976] (1/7) Epoch 12, batch 150, loss[loss=0.1811, simple_loss=0.249, pruned_loss=0.05662, over 4939.00 frames. ], tot_loss[loss=0.1849, simple_loss=0.2504, pruned_loss=0.05968, over 507807.85 frames. ], batch size: 38, lr: 3.67e-03, grad_scale: 32.0 2023-04-27 02:20:21,990 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-04-27 02:20:30,379 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.2503, 1.2167, 3.8589, 3.5633, 3.4836, 3.6877, 3.7081, 3.3696], device='cuda:1'), covar=tensor([0.7060, 0.6035, 0.1260, 0.2193, 0.1079, 0.1666, 0.1390, 0.1693], device='cuda:1'), in_proj_covar=tensor([0.0306, 0.0302, 0.0399, 0.0404, 0.0346, 0.0404, 0.0311, 0.0366], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-27 02:20:58,534 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=63201.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:21:01,808 INFO [finetune.py:976] (1/7) Epoch 12, batch 200, loss[loss=0.1688, simple_loss=0.2464, pruned_loss=0.04554, over 4938.00 frames. ], tot_loss[loss=0.1844, simple_loss=0.2498, pruned_loss=0.05947, over 606596.59 frames. ], batch size: 42, lr: 3.67e-03, grad_scale: 32.0 2023-04-27 02:21:13,193 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=63221.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:21:24,494 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.086e+02 1.629e+02 1.966e+02 2.303e+02 3.666e+02, threshold=3.932e+02, percent-clipped=0.0 2023-04-27 02:21:25,233 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=63241.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:21:30,505 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=63249.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:21:34,071 INFO [finetune.py:976] (1/7) Epoch 12, batch 250, loss[loss=0.2072, simple_loss=0.2742, pruned_loss=0.07011, over 4906.00 frames. ], tot_loss[loss=0.1879, simple_loss=0.2536, pruned_loss=0.06112, over 683406.26 frames. ], batch size: 36, lr: 3.67e-03, grad_scale: 32.0 2023-04-27 02:21:52,374 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=63282.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:21:56,552 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=63289.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:21:57,419 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-04-27 02:22:00,259 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.2564, 1.5232, 1.3096, 1.6720, 1.6188, 1.9919, 1.3453, 3.4080], device='cuda:1'), covar=tensor([0.0622, 0.0818, 0.0857, 0.1285, 0.0663, 0.0579, 0.0805, 0.0152], device='cuda:1'), in_proj_covar=tensor([0.0038, 0.0039, 0.0040, 0.0043, 0.0040, 0.0038, 0.0038, 0.0057], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0014, 0.0013, 0.0012, 0.0013, 0.0016], device='cuda:1') 2023-04-27 02:22:11,668 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.1938, 1.9354, 2.3356, 2.5484, 2.1916, 2.0878, 2.1624, 2.1939], device='cuda:1'), covar=tensor([0.5934, 0.8421, 0.8983, 0.7954, 0.7878, 1.0390, 1.0761, 0.9894], device='cuda:1'), in_proj_covar=tensor([0.0408, 0.0409, 0.0496, 0.0513, 0.0439, 0.0459, 0.0466, 0.0468], device='cuda:1'), out_proj_covar=tensor([9.9264e-05, 1.0120e-04, 1.1178e-04, 1.2202e-04, 1.0649e-04, 1.1081e-04, 1.1172e-04, 1.1220e-04], device='cuda:1') 2023-04-27 02:22:12,122 INFO [finetune.py:976] (1/7) Epoch 12, batch 300, loss[loss=0.1133, simple_loss=0.1902, pruned_loss=0.01819, over 4686.00 frames. ], tot_loss[loss=0.1898, simple_loss=0.2565, pruned_loss=0.06155, over 744851.25 frames. ], batch size: 23, lr: 3.67e-03, grad_scale: 32.0 2023-04-27 02:22:36,084 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=63322.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:22:43,410 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.0701, 1.4453, 1.3382, 1.6269, 1.6907, 2.0214, 1.3524, 3.6041], device='cuda:1'), covar=tensor([0.0693, 0.0826, 0.0883, 0.1281, 0.0643, 0.0568, 0.0794, 0.0143], device='cuda:1'), in_proj_covar=tensor([0.0038, 0.0039, 0.0040, 0.0043, 0.0040, 0.0038, 0.0038, 0.0057], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0014, 0.0013, 0.0012, 0.0013, 0.0016], device='cuda:1') 2023-04-27 02:22:46,894 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.130e+02 1.771e+02 2.081e+02 2.585e+02 5.314e+02, threshold=4.161e+02, percent-clipped=4.0 2023-04-27 02:23:01,681 INFO [finetune.py:976] (1/7) Epoch 12, batch 350, loss[loss=0.295, simple_loss=0.3549, pruned_loss=0.1176, over 4924.00 frames. ], tot_loss[loss=0.1908, simple_loss=0.258, pruned_loss=0.06179, over 792781.87 frames. ], batch size: 38, lr: 3.67e-03, grad_scale: 32.0 2023-04-27 02:23:19,002 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=63370.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:23:57,900 INFO [finetune.py:976] (1/7) Epoch 12, batch 400, loss[loss=0.1889, simple_loss=0.27, pruned_loss=0.05395, over 4892.00 frames. ], tot_loss[loss=0.1917, simple_loss=0.2593, pruned_loss=0.06202, over 829917.80 frames. ], batch size: 32, lr: 3.66e-03, grad_scale: 32.0 2023-04-27 02:24:18,826 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=63420.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:24:21,028 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=63422.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:24:31,387 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=63429.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:24:43,211 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.266e+02 1.590e+02 1.875e+02 2.265e+02 5.610e+02, threshold=3.751e+02, percent-clipped=2.0 2023-04-27 02:25:03,610 INFO [finetune.py:976] (1/7) Epoch 12, batch 450, loss[loss=0.1784, simple_loss=0.2418, pruned_loss=0.05748, over 4815.00 frames. ], tot_loss[loss=0.1892, simple_loss=0.257, pruned_loss=0.06068, over 858067.38 frames. ], batch size: 41, lr: 3.66e-03, grad_scale: 32.0 2023-04-27 02:25:38,225 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=63477.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:25:46,553 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=63481.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:25:47,770 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=63483.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:26:12,259 INFO [finetune.py:976] (1/7) Epoch 12, batch 500, loss[loss=0.2029, simple_loss=0.2595, pruned_loss=0.07309, over 4715.00 frames. ], tot_loss[loss=0.1887, simple_loss=0.2556, pruned_loss=0.06095, over 879612.35 frames. ], batch size: 59, lr: 3.66e-03, grad_scale: 32.0 2023-04-27 02:26:41,244 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.737e+01 1.732e+02 1.941e+02 2.281e+02 3.295e+02, threshold=3.881e+02, percent-clipped=0.0 2023-04-27 02:26:50,387 INFO [finetune.py:976] (1/7) Epoch 12, batch 550, loss[loss=0.2031, simple_loss=0.2642, pruned_loss=0.07099, over 4936.00 frames. ], tot_loss[loss=0.187, simple_loss=0.2531, pruned_loss=0.06041, over 897009.54 frames. ], batch size: 38, lr: 3.66e-03, grad_scale: 32.0 2023-04-27 02:26:59,474 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=63569.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:27:05,799 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=63577.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:27:12,221 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=63586.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:27:18,837 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7754, 2.0535, 1.7415, 1.3612, 1.3604, 1.3687, 1.7600, 1.2687], device='cuda:1'), covar=tensor([0.1853, 0.1611, 0.1611, 0.2068, 0.2574, 0.2135, 0.1111, 0.2232], device='cuda:1'), in_proj_covar=tensor([0.0195, 0.0214, 0.0170, 0.0205, 0.0202, 0.0184, 0.0159, 0.0186], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-27 02:27:23,755 INFO [finetune.py:976] (1/7) Epoch 12, batch 600, loss[loss=0.2276, simple_loss=0.29, pruned_loss=0.0826, over 4827.00 frames. ], tot_loss[loss=0.1874, simple_loss=0.2533, pruned_loss=0.06068, over 909849.32 frames. ], batch size: 33, lr: 3.66e-03, grad_scale: 32.0 2023-04-27 02:27:41,112 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=63630.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:27:48,533 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.049e+02 1.750e+02 2.019e+02 2.576e+02 5.185e+02, threshold=4.039e+02, percent-clipped=2.0 2023-04-27 02:27:52,953 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=63647.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:27:57,755 INFO [finetune.py:976] (1/7) Epoch 12, batch 650, loss[loss=0.2286, simple_loss=0.2999, pruned_loss=0.07863, over 4737.00 frames. ], tot_loss[loss=0.1882, simple_loss=0.2549, pruned_loss=0.06079, over 919889.56 frames. ], batch size: 59, lr: 3.66e-03, grad_scale: 32.0 2023-04-27 02:28:41,570 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.3490, 2.1789, 2.5404, 2.7482, 2.7440, 2.1668, 1.7930, 2.4717], device='cuda:1'), covar=tensor([0.0917, 0.0977, 0.0605, 0.0626, 0.0674, 0.1041, 0.0979, 0.0643], device='cuda:1'), in_proj_covar=tensor([0.0197, 0.0203, 0.0184, 0.0175, 0.0179, 0.0189, 0.0160, 0.0184], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 02:28:42,664 INFO [finetune.py:976] (1/7) Epoch 12, batch 700, loss[loss=0.2256, simple_loss=0.2924, pruned_loss=0.07936, over 4919.00 frames. ], tot_loss[loss=0.1911, simple_loss=0.2583, pruned_loss=0.06197, over 929041.25 frames. ], batch size: 33, lr: 3.66e-03, grad_scale: 32.0 2023-04-27 02:29:23,284 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.162e+02 1.645e+02 1.982e+02 2.413e+02 5.979e+02, threshold=3.963e+02, percent-clipped=2.0 2023-04-27 02:29:25,236 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8308, 2.2621, 1.8913, 2.1268, 1.5883, 1.8900, 1.9283, 1.5188], device='cuda:1'), covar=tensor([0.2127, 0.1361, 0.0868, 0.1270, 0.3170, 0.1112, 0.1988, 0.2535], device='cuda:1'), in_proj_covar=tensor([0.0290, 0.0311, 0.0223, 0.0280, 0.0308, 0.0265, 0.0251, 0.0271], device='cuda:1'), out_proj_covar=tensor([1.1690e-04, 1.2467e-04, 8.9204e-05, 1.1185e-04, 1.2575e-04, 1.0613e-04, 1.0206e-04, 1.0834e-04], device='cuda:1') 2023-04-27 02:29:32,955 INFO [finetune.py:976] (1/7) Epoch 12, batch 750, loss[loss=0.1913, simple_loss=0.2688, pruned_loss=0.05689, over 4750.00 frames. ], tot_loss[loss=0.1915, simple_loss=0.2589, pruned_loss=0.062, over 934086.20 frames. ], batch size: 54, lr: 3.66e-03, grad_scale: 32.0 2023-04-27 02:29:45,873 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=63776.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:29:47,617 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=63778.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:30:04,459 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-04-27 02:30:06,648 INFO [finetune.py:976] (1/7) Epoch 12, batch 800, loss[loss=0.2037, simple_loss=0.2719, pruned_loss=0.06778, over 4831.00 frames. ], tot_loss[loss=0.1905, simple_loss=0.2581, pruned_loss=0.06147, over 936941.45 frames. ], batch size: 39, lr: 3.66e-03, grad_scale: 32.0 2023-04-27 02:30:09,840 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=63810.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:30:39,119 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([4.7670, 4.7081, 3.3562, 5.5134, 4.8629, 4.7444, 2.4635, 4.6655], device='cuda:1'), covar=tensor([0.1244, 0.0816, 0.2564, 0.0808, 0.3156, 0.1480, 0.4878, 0.1866], device='cuda:1'), in_proj_covar=tensor([0.0245, 0.0216, 0.0251, 0.0303, 0.0299, 0.0248, 0.0270, 0.0270], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 02:30:40,780 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.251e+02 1.671e+02 1.936e+02 2.427e+02 3.896e+02, threshold=3.873e+02, percent-clipped=0.0 2023-04-27 02:30:49,751 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=63845.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:31:01,223 INFO [finetune.py:976] (1/7) Epoch 12, batch 850, loss[loss=0.1703, simple_loss=0.2381, pruned_loss=0.05123, over 4820.00 frames. ], tot_loss[loss=0.1879, simple_loss=0.2552, pruned_loss=0.06033, over 939008.54 frames. ], batch size: 39, lr: 3.66e-03, grad_scale: 32.0 2023-04-27 02:31:19,398 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7687, 1.4157, 1.8848, 2.1903, 1.8543, 1.7016, 1.7956, 1.8237], device='cuda:1'), covar=tensor([0.4973, 0.7110, 0.7225, 0.6605, 0.6504, 0.8417, 0.8672, 0.8760], device='cuda:1'), in_proj_covar=tensor([0.0408, 0.0408, 0.0496, 0.0513, 0.0440, 0.0459, 0.0466, 0.0468], device='cuda:1'), out_proj_covar=tensor([9.9284e-05, 1.0121e-04, 1.1177e-04, 1.2183e-04, 1.0646e-04, 1.1093e-04, 1.1176e-04, 1.1223e-04], device='cuda:1') 2023-04-27 02:31:21,216 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=63871.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:31:24,881 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=63877.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:32:06,882 INFO [finetune.py:976] (1/7) Epoch 12, batch 900, loss[loss=0.1389, simple_loss=0.2143, pruned_loss=0.03178, over 4869.00 frames. ], tot_loss[loss=0.1845, simple_loss=0.2517, pruned_loss=0.05866, over 944312.93 frames. ], batch size: 31, lr: 3.66e-03, grad_scale: 32.0 2023-04-27 02:32:07,600 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=63906.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:32:19,213 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=63916.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:32:22,894 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8552, 2.1233, 1.1630, 1.5062, 2.4039, 1.7962, 1.6118, 1.7934], device='cuda:1'), covar=tensor([0.0496, 0.0364, 0.0307, 0.0584, 0.0232, 0.0525, 0.0521, 0.0580], device='cuda:1'), in_proj_covar=tensor([0.0030, 0.0025, 0.0023, 0.0030, 0.0020, 0.0029, 0.0029, 0.0029], device='cuda:1'), out_proj_covar=tensor([0.0050, 0.0045, 0.0038, 0.0050, 0.0038, 0.0049, 0.0048, 0.0050], device='cuda:1') 2023-04-27 02:32:24,718 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=63925.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:32:24,741 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=63925.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:32:34,606 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.028e+02 1.661e+02 1.966e+02 2.302e+02 4.372e+02, threshold=3.933e+02, percent-clipped=1.0 2023-04-27 02:32:36,449 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=63942.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:32:36,504 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4348, 2.3982, 2.0549, 2.1571, 2.5241, 2.0254, 3.3650, 1.8054], device='cuda:1'), covar=tensor([0.4276, 0.2307, 0.4654, 0.4029, 0.1977, 0.2806, 0.1570, 0.4888], device='cuda:1'), in_proj_covar=tensor([0.0342, 0.0349, 0.0430, 0.0362, 0.0386, 0.0382, 0.0376, 0.0423], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 02:32:45,688 INFO [finetune.py:976] (1/7) Epoch 12, batch 950, loss[loss=0.1598, simple_loss=0.2299, pruned_loss=0.04484, over 4791.00 frames. ], tot_loss[loss=0.183, simple_loss=0.2499, pruned_loss=0.05803, over 946871.66 frames. ], batch size: 25, lr: 3.66e-03, grad_scale: 32.0 2023-04-27 02:32:49,268 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4785, 3.5215, 1.0927, 1.9675, 1.9622, 2.5206, 2.0572, 1.0508], device='cuda:1'), covar=tensor([0.1461, 0.0821, 0.1934, 0.1169, 0.1057, 0.0977, 0.1501, 0.1884], device='cuda:1'), in_proj_covar=tensor([0.0117, 0.0247, 0.0140, 0.0122, 0.0134, 0.0153, 0.0118, 0.0120], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-27 02:32:59,653 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=63977.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:33:20,847 INFO [finetune.py:976] (1/7) Epoch 12, batch 1000, loss[loss=0.2021, simple_loss=0.2816, pruned_loss=0.06135, over 4935.00 frames. ], tot_loss[loss=0.1859, simple_loss=0.2532, pruned_loss=0.05929, over 947457.79 frames. ], batch size: 38, lr: 3.66e-03, grad_scale: 32.0 2023-04-27 02:33:28,297 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-04-27 02:33:32,467 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=64023.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:33:43,219 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.750e+01 1.665e+02 1.998e+02 2.380e+02 4.049e+02, threshold=3.995e+02, percent-clipped=1.0 2023-04-27 02:33:59,166 INFO [finetune.py:976] (1/7) Epoch 12, batch 1050, loss[loss=0.176, simple_loss=0.2533, pruned_loss=0.04939, over 4895.00 frames. ], tot_loss[loss=0.1892, simple_loss=0.2576, pruned_loss=0.06041, over 949992.45 frames. ], batch size: 43, lr: 3.66e-03, grad_scale: 32.0 2023-04-27 02:34:29,321 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=64076.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:34:30,593 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=64078.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:34:40,109 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=64084.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:34:50,854 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.2289, 2.0580, 2.3281, 2.8050, 2.7075, 2.2031, 1.7809, 2.4497], device='cuda:1'), covar=tensor([0.0962, 0.1049, 0.0706, 0.0518, 0.0638, 0.0902, 0.1026, 0.0555], device='cuda:1'), in_proj_covar=tensor([0.0194, 0.0200, 0.0182, 0.0172, 0.0177, 0.0186, 0.0158, 0.0182], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 02:35:05,823 INFO [finetune.py:976] (1/7) Epoch 12, batch 1100, loss[loss=0.1989, simple_loss=0.2794, pruned_loss=0.05923, over 4813.00 frames. ], tot_loss[loss=0.1914, simple_loss=0.2597, pruned_loss=0.06152, over 952805.65 frames. ], batch size: 45, lr: 3.66e-03, grad_scale: 32.0 2023-04-27 02:35:28,514 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=64124.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:35:29,730 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=64126.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:35:38,340 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.140e+02 1.705e+02 2.099e+02 2.623e+02 4.712e+02, threshold=4.198e+02, percent-clipped=5.0 2023-04-27 02:35:49,424 INFO [finetune.py:976] (1/7) Epoch 12, batch 1150, loss[loss=0.1521, simple_loss=0.2235, pruned_loss=0.04037, over 4778.00 frames. ], tot_loss[loss=0.1907, simple_loss=0.2598, pruned_loss=0.06082, over 954587.49 frames. ], batch size: 25, lr: 3.66e-03, grad_scale: 32.0 2023-04-27 02:35:57,667 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=64166.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:36:02,567 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.51 vs. limit=2.0 2023-04-27 02:36:19,850 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=64201.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:36:22,239 INFO [finetune.py:976] (1/7) Epoch 12, batch 1200, loss[loss=0.1316, simple_loss=0.2, pruned_loss=0.03163, over 4705.00 frames. ], tot_loss[loss=0.1886, simple_loss=0.2571, pruned_loss=0.06006, over 954364.08 frames. ], batch size: 23, lr: 3.66e-03, grad_scale: 32.0 2023-04-27 02:36:35,450 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5290, 3.3956, 0.8679, 1.8982, 1.7959, 2.4303, 1.9784, 1.0508], device='cuda:1'), covar=tensor([0.1281, 0.0919, 0.1970, 0.1209, 0.1117, 0.0970, 0.1351, 0.1913], device='cuda:1'), in_proj_covar=tensor([0.0116, 0.0245, 0.0139, 0.0121, 0.0132, 0.0151, 0.0117, 0.0119], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-27 02:36:36,043 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=64225.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:36:40,395 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=64232.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:36:50,582 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.051e+02 1.610e+02 1.933e+02 2.309e+02 4.139e+02, threshold=3.865e+02, percent-clipped=0.0 2023-04-27 02:36:51,901 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=64242.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:37:10,837 INFO [finetune.py:976] (1/7) Epoch 12, batch 1250, loss[loss=0.1798, simple_loss=0.2375, pruned_loss=0.06105, over 4144.00 frames. ], tot_loss[loss=0.1871, simple_loss=0.2547, pruned_loss=0.05977, over 952777.19 frames. ], batch size: 65, lr: 3.66e-03, grad_scale: 64.0 2023-04-27 02:37:33,223 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=64272.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:37:33,824 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=64273.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:37:44,954 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.2033, 1.7069, 2.0741, 2.6549, 1.9905, 1.6263, 1.4091, 1.9341], device='cuda:1'), covar=tensor([0.3747, 0.3623, 0.1937, 0.2538, 0.3084, 0.3031, 0.4511, 0.2530], device='cuda:1'), in_proj_covar=tensor([0.0284, 0.0247, 0.0220, 0.0314, 0.0212, 0.0227, 0.0228, 0.0184], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-27 02:37:55,439 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=64290.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:37:56,137 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7337, 1.4573, 1.8590, 2.2194, 1.8478, 1.6966, 1.8075, 1.8020], device='cuda:1'), covar=tensor([0.5493, 0.7648, 0.7937, 0.7024, 0.7024, 0.9177, 0.9602, 0.9349], device='cuda:1'), in_proj_covar=tensor([0.0410, 0.0410, 0.0496, 0.0514, 0.0441, 0.0461, 0.0468, 0.0470], device='cuda:1'), out_proj_covar=tensor([9.9802e-05, 1.0148e-04, 1.1187e-04, 1.2216e-04, 1.0681e-04, 1.1140e-04, 1.1224e-04, 1.1255e-04], device='cuda:1') 2023-04-27 02:37:57,330 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=64293.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:38:16,202 INFO [finetune.py:976] (1/7) Epoch 12, batch 1300, loss[loss=0.2054, simple_loss=0.2796, pruned_loss=0.06553, over 4912.00 frames. ], tot_loss[loss=0.1844, simple_loss=0.2519, pruned_loss=0.05842, over 954567.68 frames. ], batch size: 36, lr: 3.66e-03, grad_scale: 32.0 2023-04-27 02:38:40,957 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.098e+02 1.564e+02 1.871e+02 2.311e+02 4.814e+02, threshold=3.742e+02, percent-clipped=4.0 2023-04-27 02:38:43,500 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=64345.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:38:45,323 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0362, 2.5219, 2.1744, 2.3077, 1.7619, 2.1126, 2.2830, 1.6400], device='cuda:1'), covar=tensor([0.2057, 0.1424, 0.0969, 0.1308, 0.3242, 0.1228, 0.1796, 0.2857], device='cuda:1'), in_proj_covar=tensor([0.0294, 0.0314, 0.0226, 0.0283, 0.0313, 0.0268, 0.0254, 0.0275], device='cuda:1'), out_proj_covar=tensor([1.1858e-04, 1.2570e-04, 9.0406e-05, 1.1298e-04, 1.2787e-04, 1.0760e-04, 1.0327e-04, 1.1013e-04], device='cuda:1') 2023-04-27 02:38:49,942 INFO [finetune.py:976] (1/7) Epoch 12, batch 1350, loss[loss=0.2517, simple_loss=0.312, pruned_loss=0.09568, over 4828.00 frames. ], tot_loss[loss=0.1853, simple_loss=0.2524, pruned_loss=0.05912, over 955732.20 frames. ], batch size: 47, lr: 3.66e-03, grad_scale: 32.0 2023-04-27 02:39:01,214 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.1923, 2.1281, 2.4610, 2.6534, 2.6434, 2.1481, 1.7579, 2.4041], device='cuda:1'), covar=tensor([0.1062, 0.0941, 0.0618, 0.0636, 0.0679, 0.0889, 0.0977, 0.0605], device='cuda:1'), in_proj_covar=tensor([0.0194, 0.0200, 0.0182, 0.0172, 0.0177, 0.0186, 0.0158, 0.0181], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 02:39:07,139 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=64379.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:39:22,954 INFO [finetune.py:976] (1/7) Epoch 12, batch 1400, loss[loss=0.1833, simple_loss=0.2659, pruned_loss=0.05038, over 4808.00 frames. ], tot_loss[loss=0.1885, simple_loss=0.256, pruned_loss=0.06047, over 954279.25 frames. ], batch size: 45, lr: 3.66e-03, grad_scale: 32.0 2023-04-27 02:39:23,705 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=64406.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:39:27,618 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5102, 2.5360, 2.2119, 2.3430, 2.8260, 2.5618, 3.6977, 2.1039], device='cuda:1'), covar=tensor([0.4106, 0.2326, 0.4533, 0.3740, 0.1707, 0.2508, 0.1200, 0.4192], device='cuda:1'), in_proj_covar=tensor([0.0339, 0.0346, 0.0426, 0.0357, 0.0382, 0.0380, 0.0372, 0.0418], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 02:39:48,150 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.081e+02 1.746e+02 2.220e+02 2.632e+02 4.800e+02, threshold=4.440e+02, percent-clipped=5.0 2023-04-27 02:40:07,659 INFO [finetune.py:976] (1/7) Epoch 12, batch 1450, loss[loss=0.1759, simple_loss=0.2452, pruned_loss=0.05333, over 4815.00 frames. ], tot_loss[loss=0.1895, simple_loss=0.2577, pruned_loss=0.06068, over 953796.17 frames. ], batch size: 25, lr: 3.66e-03, grad_scale: 32.0 2023-04-27 02:40:21,226 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=64465.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:40:21,841 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=64466.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:40:55,267 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=64501.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:40:57,604 INFO [finetune.py:976] (1/7) Epoch 12, batch 1500, loss[loss=0.1787, simple_loss=0.2499, pruned_loss=0.05377, over 4914.00 frames. ], tot_loss[loss=0.1922, simple_loss=0.2604, pruned_loss=0.06201, over 954675.70 frames. ], batch size: 33, lr: 3.66e-03, grad_scale: 32.0 2023-04-27 02:41:03,987 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=64514.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:41:12,854 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=64526.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:41:22,337 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.091e+02 1.729e+02 2.014e+02 2.511e+02 4.004e+02, threshold=4.028e+02, percent-clipped=0.0 2023-04-27 02:41:27,203 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=64549.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:41:30,767 INFO [finetune.py:976] (1/7) Epoch 12, batch 1550, loss[loss=0.1941, simple_loss=0.2667, pruned_loss=0.06076, over 4838.00 frames. ], tot_loss[loss=0.1922, simple_loss=0.2605, pruned_loss=0.06195, over 954029.47 frames. ], batch size: 44, lr: 3.66e-03, grad_scale: 32.0 2023-04-27 02:41:42,612 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=64572.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:41:53,768 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=64588.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:41:54,936 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=64589.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:42:10,372 INFO [finetune.py:976] (1/7) Epoch 12, batch 1600, loss[loss=0.203, simple_loss=0.26, pruned_loss=0.07301, over 4778.00 frames. ], tot_loss[loss=0.1898, simple_loss=0.2577, pruned_loss=0.06095, over 955070.03 frames. ], batch size: 28, lr: 3.65e-03, grad_scale: 32.0 2023-04-27 02:42:31,724 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=64620.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:42:41,817 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=2.03 vs. limit=2.0 2023-04-27 02:42:46,405 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.047e+02 1.700e+02 1.959e+02 2.289e+02 5.207e+02, threshold=3.917e+02, percent-clipped=1.0 2023-04-27 02:42:56,548 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=64648.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:42:57,751 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=64650.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:43:06,461 INFO [finetune.py:976] (1/7) Epoch 12, batch 1650, loss[loss=0.1733, simple_loss=0.2419, pruned_loss=0.05233, over 4824.00 frames. ], tot_loss[loss=0.1869, simple_loss=0.2541, pruned_loss=0.0598, over 954200.29 frames. ], batch size: 40, lr: 3.65e-03, grad_scale: 32.0 2023-04-27 02:43:36,857 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=64679.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:44:09,945 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=64701.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:44:11,218 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=64703.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:44:12,323 INFO [finetune.py:976] (1/7) Epoch 12, batch 1700, loss[loss=0.2086, simple_loss=0.2719, pruned_loss=0.07271, over 4936.00 frames. ], tot_loss[loss=0.1857, simple_loss=0.2525, pruned_loss=0.0594, over 955020.90 frames. ], batch size: 38, lr: 3.65e-03, grad_scale: 32.0 2023-04-27 02:44:14,915 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=64709.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:44:27,402 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=64727.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:44:37,468 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.062e+02 1.591e+02 1.842e+02 2.298e+02 5.493e+02, threshold=3.684e+02, percent-clipped=1.0 2023-04-27 02:44:40,029 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=64745.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:44:44,741 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=64752.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:44:46,474 INFO [finetune.py:976] (1/7) Epoch 12, batch 1750, loss[loss=0.2339, simple_loss=0.3017, pruned_loss=0.08306, over 4874.00 frames. ], tot_loss[loss=0.1886, simple_loss=0.2559, pruned_loss=0.06066, over 954862.08 frames. ], batch size: 44, lr: 3.65e-03, grad_scale: 32.0 2023-04-27 02:44:51,984 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=64764.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:45:04,251 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4715, 2.7324, 2.3290, 2.4512, 2.5733, 2.3941, 3.7848, 1.9029], device='cuda:1'), covar=tensor([0.4303, 0.1976, 0.3811, 0.3588, 0.2232, 0.2768, 0.1278, 0.4301], device='cuda:1'), in_proj_covar=tensor([0.0340, 0.0347, 0.0430, 0.0358, 0.0385, 0.0382, 0.0375, 0.0422], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 02:45:13,644 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.1336, 2.6026, 1.0355, 1.3866, 1.9777, 1.2111, 3.6848, 1.7570], device='cuda:1'), covar=tensor([0.0693, 0.0644, 0.0828, 0.1406, 0.0558, 0.1125, 0.0245, 0.0703], device='cuda:1'), in_proj_covar=tensor([0.0051, 0.0066, 0.0049, 0.0047, 0.0051, 0.0053, 0.0076, 0.0051], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:1') 2023-04-27 02:45:20,205 INFO [finetune.py:976] (1/7) Epoch 12, batch 1800, loss[loss=0.1897, simple_loss=0.2551, pruned_loss=0.06219, over 4931.00 frames. ], tot_loss[loss=0.1888, simple_loss=0.2563, pruned_loss=0.06067, over 951008.01 frames. ], batch size: 33, lr: 3.65e-03, grad_scale: 32.0 2023-04-27 02:45:20,922 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=64806.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:45:25,156 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=64813.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:45:29,909 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=64821.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:46:00,004 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.230e+02 1.632e+02 2.043e+02 2.586e+02 5.836e+02, threshold=4.087e+02, percent-clipped=7.0 2023-04-27 02:46:21,071 INFO [finetune.py:976] (1/7) Epoch 12, batch 1850, loss[loss=0.1978, simple_loss=0.2698, pruned_loss=0.06294, over 4920.00 frames. ], tot_loss[loss=0.1886, simple_loss=0.2568, pruned_loss=0.06022, over 952419.16 frames. ], batch size: 33, lr: 3.65e-03, grad_scale: 32.0 2023-04-27 02:46:55,920 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.58 vs. limit=5.0 2023-04-27 02:46:58,522 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=64888.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:47:07,273 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4053, 2.5331, 1.9432, 2.1778, 2.3221, 1.9294, 3.2691, 1.5134], device='cuda:1'), covar=tensor([0.4377, 0.1801, 0.4282, 0.3418, 0.2258, 0.3198, 0.1470, 0.5037], device='cuda:1'), in_proj_covar=tensor([0.0340, 0.0347, 0.0429, 0.0358, 0.0385, 0.0382, 0.0375, 0.0423], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 02:47:10,162 INFO [finetune.py:976] (1/7) Epoch 12, batch 1900, loss[loss=0.2176, simple_loss=0.2809, pruned_loss=0.07716, over 4849.00 frames. ], tot_loss[loss=0.1901, simple_loss=0.2587, pruned_loss=0.06073, over 953299.91 frames. ], batch size: 31, lr: 3.65e-03, grad_scale: 32.0 2023-04-27 02:47:21,005 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.3330, 2.6930, 2.4531, 2.5423, 2.0190, 2.2512, 2.3541, 2.0418], device='cuda:1'), covar=tensor([0.1560, 0.1198, 0.0754, 0.0980, 0.2795, 0.1029, 0.1574, 0.2025], device='cuda:1'), in_proj_covar=tensor([0.0290, 0.0311, 0.0223, 0.0280, 0.0310, 0.0264, 0.0251, 0.0272], device='cuda:1'), out_proj_covar=tensor([1.1678e-04, 1.2444e-04, 8.9332e-05, 1.1179e-04, 1.2668e-04, 1.0607e-04, 1.0210e-04, 1.0870e-04], device='cuda:1') 2023-04-27 02:47:21,584 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.3025, 3.0031, 0.8171, 1.6911, 1.5993, 2.1682, 1.6854, 1.0063], device='cuda:1'), covar=tensor([0.1430, 0.1236, 0.2055, 0.1284, 0.1207, 0.0982, 0.1638, 0.1990], device='cuda:1'), in_proj_covar=tensor([0.0119, 0.0249, 0.0141, 0.0123, 0.0135, 0.0154, 0.0119, 0.0121], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-27 02:47:35,068 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=64936.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:47:39,044 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.204e+02 1.650e+02 1.962e+02 2.439e+02 4.068e+02, threshold=3.925e+02, percent-clipped=0.0 2023-04-27 02:47:48,351 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=64945.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:48:00,716 INFO [finetune.py:976] (1/7) Epoch 12, batch 1950, loss[loss=0.1889, simple_loss=0.2558, pruned_loss=0.06098, over 4760.00 frames. ], tot_loss[loss=0.1888, simple_loss=0.2571, pruned_loss=0.06021, over 952648.86 frames. ], batch size: 28, lr: 3.65e-03, grad_scale: 32.0 2023-04-27 02:48:31,531 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=64976.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:48:49,840 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=65001.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:48:50,455 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7929, 1.2987, 5.1976, 4.8370, 4.5960, 4.9627, 4.5382, 4.5547], device='cuda:1'), covar=tensor([0.7347, 0.6650, 0.0946, 0.1784, 0.0988, 0.1115, 0.1363, 0.1539], device='cuda:1'), in_proj_covar=tensor([0.0308, 0.0303, 0.0399, 0.0405, 0.0346, 0.0403, 0.0313, 0.0365], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-27 02:48:51,667 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=65004.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:48:52,188 INFO [finetune.py:976] (1/7) Epoch 12, batch 2000, loss[loss=0.1675, simple_loss=0.2346, pruned_loss=0.05019, over 4762.00 frames. ], tot_loss[loss=0.1863, simple_loss=0.2543, pruned_loss=0.05916, over 953137.45 frames. ], batch size: 28, lr: 3.65e-03, grad_scale: 32.0 2023-04-27 02:49:12,516 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=65037.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 02:49:15,255 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.045e+02 1.562e+02 1.931e+02 2.366e+02 5.248e+02, threshold=3.862e+02, percent-clipped=2.0 2023-04-27 02:49:15,370 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.1971, 1.5495, 1.4465, 1.8990, 1.6640, 1.8135, 1.4622, 3.4322], device='cuda:1'), covar=tensor([0.0682, 0.0788, 0.0865, 0.1208, 0.0697, 0.0533, 0.0778, 0.0185], device='cuda:1'), in_proj_covar=tensor([0.0038, 0.0039, 0.0040, 0.0043, 0.0040, 0.0038, 0.0038, 0.0057], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0014, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:1') 2023-04-27 02:49:21,189 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=65049.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:49:26,350 INFO [finetune.py:976] (1/7) Epoch 12, batch 2050, loss[loss=0.2273, simple_loss=0.2856, pruned_loss=0.0845, over 4712.00 frames. ], tot_loss[loss=0.1837, simple_loss=0.251, pruned_loss=0.05827, over 954089.56 frames. ], batch size: 59, lr: 3.65e-03, grad_scale: 32.0 2023-04-27 02:49:26,471 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=65055.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:49:28,831 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=65059.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:49:56,651 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=65101.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:50:00,020 INFO [finetune.py:976] (1/7) Epoch 12, batch 2100, loss[loss=0.2198, simple_loss=0.2929, pruned_loss=0.07335, over 4865.00 frames. ], tot_loss[loss=0.1843, simple_loss=0.2512, pruned_loss=0.05867, over 954538.37 frames. ], batch size: 44, lr: 3.65e-03, grad_scale: 32.0 2023-04-27 02:50:01,931 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=65108.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:50:06,878 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=65116.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:50:10,307 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=65121.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:50:22,438 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.293e+02 1.658e+02 1.987e+02 2.538e+02 5.198e+02, threshold=3.973e+02, percent-clipped=2.0 2023-04-27 02:50:33,406 INFO [finetune.py:976] (1/7) Epoch 12, batch 2150, loss[loss=0.1738, simple_loss=0.2462, pruned_loss=0.05074, over 4828.00 frames. ], tot_loss[loss=0.1849, simple_loss=0.2528, pruned_loss=0.0585, over 955223.56 frames. ], batch size: 33, lr: 3.65e-03, grad_scale: 32.0 2023-04-27 02:50:40,394 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-04-27 02:50:42,396 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=65169.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:51:05,955 INFO [finetune.py:976] (1/7) Epoch 12, batch 2200, loss[loss=0.2214, simple_loss=0.2818, pruned_loss=0.08048, over 4814.00 frames. ], tot_loss[loss=0.1879, simple_loss=0.2559, pruned_loss=0.05998, over 954977.48 frames. ], batch size: 38, lr: 3.65e-03, grad_scale: 32.0 2023-04-27 02:51:56,236 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.062e+02 1.743e+02 2.057e+02 2.570e+02 5.251e+02, threshold=4.115e+02, percent-clipped=3.0 2023-04-27 02:51:58,787 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=65245.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:52:11,639 INFO [finetune.py:976] (1/7) Epoch 12, batch 2250, loss[loss=0.2284, simple_loss=0.2917, pruned_loss=0.08256, over 4745.00 frames. ], tot_loss[loss=0.19, simple_loss=0.2582, pruned_loss=0.06093, over 956548.68 frames. ], batch size: 59, lr: 3.65e-03, grad_scale: 32.0 2023-04-27 02:52:39,980 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5373, 1.2948, 1.7380, 1.6840, 1.3498, 1.2506, 1.3587, 0.8850], device='cuda:1'), covar=tensor([0.0598, 0.0865, 0.0459, 0.0695, 0.0820, 0.1276, 0.0781, 0.0761], device='cuda:1'), in_proj_covar=tensor([0.0068, 0.0072, 0.0071, 0.0068, 0.0076, 0.0097, 0.0076, 0.0071], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-27 02:52:53,434 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=65285.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:53:03,643 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=65293.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:53:21,400 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=65304.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:53:21,920 INFO [finetune.py:976] (1/7) Epoch 12, batch 2300, loss[loss=0.2066, simple_loss=0.2623, pruned_loss=0.07547, over 4886.00 frames. ], tot_loss[loss=0.1898, simple_loss=0.2586, pruned_loss=0.06046, over 957329.48 frames. ], batch size: 35, lr: 3.65e-03, grad_scale: 32.0 2023-04-27 02:53:24,351 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4020, 1.2536, 4.0772, 3.8084, 3.6516, 3.9267, 3.8342, 3.5400], device='cuda:1'), covar=tensor([0.7154, 0.6004, 0.1089, 0.1741, 0.1123, 0.1694, 0.1876, 0.1536], device='cuda:1'), in_proj_covar=tensor([0.0309, 0.0303, 0.0401, 0.0406, 0.0347, 0.0406, 0.0313, 0.0367], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-27 02:53:56,227 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=65332.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 02:54:07,783 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.005e+02 1.597e+02 1.792e+02 2.067e+02 3.757e+02, threshold=3.584e+02, percent-clipped=0.0 2023-04-27 02:54:16,357 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=65346.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:54:20,390 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=65352.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:54:22,654 INFO [finetune.py:976] (1/7) Epoch 12, batch 2350, loss[loss=0.2012, simple_loss=0.2588, pruned_loss=0.0718, over 4823.00 frames. ], tot_loss[loss=0.1891, simple_loss=0.2571, pruned_loss=0.06054, over 956856.52 frames. ], batch size: 30, lr: 3.65e-03, grad_scale: 32.0 2023-04-27 02:54:25,631 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=65359.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:54:35,789 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-04-27 02:54:51,276 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9015, 2.5614, 1.8432, 1.8402, 1.3758, 1.3713, 1.9813, 1.2278], device='cuda:1'), covar=tensor([0.1633, 0.1294, 0.1482, 0.1712, 0.2362, 0.2044, 0.1012, 0.2037], device='cuda:1'), in_proj_covar=tensor([0.0195, 0.0213, 0.0169, 0.0203, 0.0202, 0.0183, 0.0156, 0.0185], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-27 02:54:52,628 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([0.9288, 1.8034, 1.6990, 1.4099, 1.8366, 1.6082, 2.1890, 1.4599], device='cuda:1'), covar=tensor([0.2980, 0.1298, 0.3697, 0.2340, 0.1349, 0.1841, 0.1456, 0.3983], device='cuda:1'), in_proj_covar=tensor([0.0339, 0.0346, 0.0428, 0.0359, 0.0384, 0.0381, 0.0375, 0.0421], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 02:54:53,205 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=65401.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:54:55,972 INFO [finetune.py:976] (1/7) Epoch 12, batch 2400, loss[loss=0.19, simple_loss=0.2579, pruned_loss=0.06106, over 4918.00 frames. ], tot_loss[loss=0.1858, simple_loss=0.2531, pruned_loss=0.05926, over 954112.71 frames. ], batch size: 36, lr: 3.65e-03, grad_scale: 32.0 2023-04-27 02:54:57,715 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=65407.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:54:57,780 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9829, 1.4283, 1.5254, 1.6685, 2.1393, 1.7827, 1.4919, 1.4865], device='cuda:1'), covar=tensor([0.1567, 0.1652, 0.2184, 0.1464, 0.0902, 0.1524, 0.2309, 0.2353], device='cuda:1'), in_proj_covar=tensor([0.0309, 0.0321, 0.0355, 0.0299, 0.0334, 0.0320, 0.0310, 0.0364], device='cuda:1'), out_proj_covar=tensor([6.4690e-05, 6.7493e-05, 7.6362e-05, 6.1501e-05, 6.9742e-05, 6.8070e-05, 6.6026e-05, 7.8036e-05], device='cuda:1') 2023-04-27 02:54:58,362 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=65408.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:55:00,170 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=65411.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:55:20,648 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.015e+02 1.563e+02 1.962e+02 2.388e+02 3.739e+02, threshold=3.925e+02, percent-clipped=1.0 2023-04-27 02:55:25,583 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=65449.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:55:29,227 INFO [finetune.py:976] (1/7) Epoch 12, batch 2450, loss[loss=0.1803, simple_loss=0.2561, pruned_loss=0.05224, over 4756.00 frames. ], tot_loss[loss=0.1841, simple_loss=0.2508, pruned_loss=0.05869, over 955951.68 frames. ], batch size: 54, lr: 3.65e-03, grad_scale: 32.0 2023-04-27 02:55:30,349 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=65456.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:56:02,906 INFO [finetune.py:976] (1/7) Epoch 12, batch 2500, loss[loss=0.1567, simple_loss=0.2193, pruned_loss=0.0471, over 4754.00 frames. ], tot_loss[loss=0.1865, simple_loss=0.2534, pruned_loss=0.05985, over 956294.56 frames. ], batch size: 26, lr: 3.65e-03, grad_scale: 32.0 2023-04-27 02:56:22,235 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=65532.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 02:56:28,027 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.207e+02 1.847e+02 2.256e+02 2.656e+02 4.911e+02, threshold=4.512e+02, percent-clipped=5.0 2023-04-27 02:56:36,598 INFO [finetune.py:976] (1/7) Epoch 12, batch 2550, loss[loss=0.1726, simple_loss=0.2396, pruned_loss=0.0528, over 4819.00 frames. ], tot_loss[loss=0.19, simple_loss=0.2572, pruned_loss=0.0614, over 953959.97 frames. ], batch size: 30, lr: 3.65e-03, grad_scale: 16.0 2023-04-27 02:56:48,530 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.1421, 2.9718, 0.8888, 1.6938, 1.5503, 2.2753, 1.7343, 0.9989], device='cuda:1'), covar=tensor([0.1708, 0.1441, 0.2238, 0.1482, 0.1264, 0.1152, 0.1707, 0.2316], device='cuda:1'), in_proj_covar=tensor([0.0119, 0.0249, 0.0142, 0.0123, 0.0135, 0.0155, 0.0119, 0.0121], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-27 02:57:02,297 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=65593.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 02:57:10,044 INFO [finetune.py:976] (1/7) Epoch 12, batch 2600, loss[loss=0.2344, simple_loss=0.2948, pruned_loss=0.08699, over 4919.00 frames. ], tot_loss[loss=0.19, simple_loss=0.2574, pruned_loss=0.06127, over 951861.97 frames. ], batch size: 38, lr: 3.65e-03, grad_scale: 16.0 2023-04-27 02:57:29,306 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=65632.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:57:34,707 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=65641.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:57:35,236 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.096e+02 1.708e+02 2.004e+02 2.304e+02 5.079e+02, threshold=4.008e+02, percent-clipped=1.0 2023-04-27 02:57:54,584 INFO [finetune.py:976] (1/7) Epoch 12, batch 2650, loss[loss=0.1814, simple_loss=0.2608, pruned_loss=0.05099, over 4852.00 frames. ], tot_loss[loss=0.1919, simple_loss=0.2596, pruned_loss=0.06213, over 952340.51 frames. ], batch size: 44, lr: 3.65e-03, grad_scale: 16.0 2023-04-27 02:58:26,145 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=65680.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:58:54,676 INFO [finetune.py:976] (1/7) Epoch 12, batch 2700, loss[loss=0.1916, simple_loss=0.2552, pruned_loss=0.06397, over 4862.00 frames. ], tot_loss[loss=0.1899, simple_loss=0.2576, pruned_loss=0.0611, over 950247.31 frames. ], batch size: 34, lr: 3.65e-03, grad_scale: 16.0 2023-04-27 02:59:03,970 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=65711.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:59:29,497 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7556, 1.3913, 1.9735, 2.3237, 1.9043, 1.7741, 1.8702, 1.8092], device='cuda:1'), covar=tensor([0.5088, 0.6985, 0.6975, 0.6378, 0.5962, 0.8302, 0.8306, 0.8656], device='cuda:1'), in_proj_covar=tensor([0.0412, 0.0408, 0.0498, 0.0515, 0.0441, 0.0463, 0.0469, 0.0471], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 02:59:42,189 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.037e+02 1.570e+02 1.817e+02 2.252e+02 5.618e+02, threshold=3.635e+02, percent-clipped=2.0 2023-04-27 02:59:49,576 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.2723, 1.5048, 1.4518, 1.8293, 1.6352, 2.0082, 1.3460, 3.5678], device='cuda:1'), covar=tensor([0.0656, 0.0842, 0.0838, 0.1212, 0.0663, 0.0466, 0.0778, 0.0158], device='cuda:1'), in_proj_covar=tensor([0.0038, 0.0039, 0.0040, 0.0043, 0.0040, 0.0038, 0.0038, 0.0057], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0014, 0.0013, 0.0012, 0.0013, 0.0016], device='cuda:1') 2023-04-27 03:00:01,383 INFO [finetune.py:976] (1/7) Epoch 12, batch 2750, loss[loss=0.193, simple_loss=0.2581, pruned_loss=0.06391, over 4839.00 frames. ], tot_loss[loss=0.1877, simple_loss=0.2548, pruned_loss=0.0603, over 949338.06 frames. ], batch size: 47, lr: 3.65e-03, grad_scale: 16.0 2023-04-27 03:00:03,878 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=65759.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 03:00:35,316 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6005, 2.0300, 1.6561, 1.8191, 1.5504, 1.6616, 1.6013, 1.3146], device='cuda:1'), covar=tensor([0.1731, 0.1108, 0.0905, 0.1204, 0.3163, 0.1083, 0.1803, 0.2422], device='cuda:1'), in_proj_covar=tensor([0.0290, 0.0310, 0.0222, 0.0281, 0.0311, 0.0264, 0.0251, 0.0272], device='cuda:1'), out_proj_covar=tensor([1.1677e-04, 1.2425e-04, 8.8965e-05, 1.1242e-04, 1.2685e-04, 1.0593e-04, 1.0199e-04, 1.0883e-04], device='cuda:1') 2023-04-27 03:01:07,126 INFO [finetune.py:976] (1/7) Epoch 12, batch 2800, loss[loss=0.1598, simple_loss=0.2303, pruned_loss=0.04462, over 4848.00 frames. ], tot_loss[loss=0.1855, simple_loss=0.2522, pruned_loss=0.05944, over 952322.29 frames. ], batch size: 44, lr: 3.64e-03, grad_scale: 16.0 2023-04-27 03:01:18,863 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.5619, 2.2111, 2.5852, 3.0892, 2.9984, 2.5161, 2.0393, 2.5343], device='cuda:1'), covar=tensor([0.0981, 0.1137, 0.0770, 0.0548, 0.0612, 0.0978, 0.0905, 0.0716], device='cuda:1'), in_proj_covar=tensor([0.0195, 0.0203, 0.0184, 0.0175, 0.0179, 0.0189, 0.0159, 0.0183], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 03:01:31,948 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5871, 1.3986, 1.8972, 1.8661, 1.4240, 1.3264, 1.5716, 0.9671], device='cuda:1'), covar=tensor([0.0676, 0.1054, 0.0478, 0.0807, 0.0922, 0.1326, 0.0709, 0.0881], device='cuda:1'), in_proj_covar=tensor([0.0068, 0.0072, 0.0071, 0.0067, 0.0076, 0.0097, 0.0077, 0.0071], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-27 03:01:51,950 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=65838.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 03:01:59,670 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.127e+02 1.664e+02 1.878e+02 2.585e+02 4.047e+02, threshold=3.756e+02, percent-clipped=2.0 2023-04-27 03:02:08,134 INFO [finetune.py:976] (1/7) Epoch 12, batch 2850, loss[loss=0.1543, simple_loss=0.2328, pruned_loss=0.03791, over 4781.00 frames. ], tot_loss[loss=0.1829, simple_loss=0.2494, pruned_loss=0.0582, over 950858.72 frames. ], batch size: 29, lr: 3.64e-03, grad_scale: 16.0 2023-04-27 03:02:29,538 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=65888.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 03:02:38,177 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=65899.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 03:02:41,704 INFO [finetune.py:976] (1/7) Epoch 12, batch 2900, loss[loss=0.1902, simple_loss=0.2682, pruned_loss=0.05608, over 4818.00 frames. ], tot_loss[loss=0.1862, simple_loss=0.2533, pruned_loss=0.05954, over 952518.18 frames. ], batch size: 39, lr: 3.64e-03, grad_scale: 16.0 2023-04-27 03:03:02,660 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.2455, 2.1930, 2.1788, 1.8840, 2.3166, 1.9871, 2.8726, 1.9108], device='cuda:1'), covar=tensor([0.3343, 0.1552, 0.3417, 0.2776, 0.1607, 0.2141, 0.1434, 0.3604], device='cuda:1'), in_proj_covar=tensor([0.0337, 0.0343, 0.0424, 0.0355, 0.0380, 0.0379, 0.0373, 0.0418], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 03:03:05,042 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=65941.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 03:03:05,560 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.180e+02 1.732e+02 2.023e+02 2.445e+02 4.251e+02, threshold=4.047e+02, percent-clipped=2.0 2023-04-27 03:03:15,315 INFO [finetune.py:976] (1/7) Epoch 12, batch 2950, loss[loss=0.2097, simple_loss=0.2724, pruned_loss=0.07353, over 4774.00 frames. ], tot_loss[loss=0.1872, simple_loss=0.2548, pruned_loss=0.05986, over 949885.35 frames. ], batch size: 27, lr: 3.64e-03, grad_scale: 16.0 2023-04-27 03:03:24,061 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.45 vs. limit=5.0 2023-04-27 03:03:36,945 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=65989.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 03:03:49,756 INFO [finetune.py:976] (1/7) Epoch 12, batch 3000, loss[loss=0.1971, simple_loss=0.2694, pruned_loss=0.06241, over 4822.00 frames. ], tot_loss[loss=0.1885, simple_loss=0.256, pruned_loss=0.06055, over 950875.56 frames. ], batch size: 39, lr: 3.64e-03, grad_scale: 16.0 2023-04-27 03:03:49,756 INFO [finetune.py:1001] (1/7) Computing validation loss 2023-04-27 03:03:55,132 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([0.9219, 1.9201, 1.7511, 1.5175, 1.9817, 1.6889, 2.2360, 1.5663], device='cuda:1'), covar=tensor([0.3114, 0.1283, 0.4158, 0.2593, 0.1306, 0.1874, 0.1545, 0.4702], device='cuda:1'), in_proj_covar=tensor([0.0338, 0.0344, 0.0426, 0.0356, 0.0381, 0.0380, 0.0375, 0.0419], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 03:04:00,410 INFO [finetune.py:1010] (1/7) Epoch 12, validation: loss=0.1529, simple_loss=0.2247, pruned_loss=0.04052, over 2265189.00 frames. 2023-04-27 03:04:00,410 INFO [finetune.py:1011] (1/7) Maximum memory allocated so far is 6391MB 2023-04-27 03:04:22,335 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-04-27 03:04:29,438 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.284e+02 1.787e+02 2.243e+02 2.841e+02 1.392e+03, threshold=4.486e+02, percent-clipped=4.0 2023-04-27 03:04:38,274 INFO [finetune.py:976] (1/7) Epoch 12, batch 3050, loss[loss=0.1436, simple_loss=0.2203, pruned_loss=0.0334, over 4793.00 frames. ], tot_loss[loss=0.1891, simple_loss=0.2565, pruned_loss=0.06083, over 950440.84 frames. ], batch size: 45, lr: 3.64e-03, grad_scale: 16.0 2023-04-27 03:04:43,232 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.15 vs. limit=5.0 2023-04-27 03:05:10,512 INFO [finetune.py:976] (1/7) Epoch 12, batch 3100, loss[loss=0.1677, simple_loss=0.2416, pruned_loss=0.04694, over 4824.00 frames. ], tot_loss[loss=0.187, simple_loss=0.2544, pruned_loss=0.05974, over 952313.44 frames. ], batch size: 38, lr: 3.64e-03, grad_scale: 16.0 2023-04-27 03:05:40,809 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.130e+02 1.687e+02 1.900e+02 2.210e+02 4.094e+02, threshold=3.799e+02, percent-clipped=0.0 2023-04-27 03:05:42,203 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=66144.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 03:05:54,641 INFO [finetune.py:976] (1/7) Epoch 12, batch 3150, loss[loss=0.1471, simple_loss=0.2056, pruned_loss=0.04436, over 4716.00 frames. ], tot_loss[loss=0.1849, simple_loss=0.2517, pruned_loss=0.05904, over 950786.71 frames. ], batch size: 23, lr: 3.64e-03, grad_scale: 16.0 2023-04-27 03:05:58,700 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=66160.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 03:06:27,522 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=66188.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 03:06:36,567 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=66194.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 03:06:38,977 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.8723, 2.7485, 2.2709, 3.2905, 2.8513, 2.8496, 1.1241, 2.8253], device='cuda:1'), covar=tensor([0.1948, 0.1782, 0.3503, 0.3063, 0.3619, 0.2162, 0.5679, 0.2684], device='cuda:1'), in_proj_covar=tensor([0.0242, 0.0216, 0.0248, 0.0302, 0.0297, 0.0246, 0.0269, 0.0268], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 03:06:49,144 INFO [finetune.py:976] (1/7) Epoch 12, batch 3200, loss[loss=0.1692, simple_loss=0.2292, pruned_loss=0.0546, over 4913.00 frames. ], tot_loss[loss=0.1828, simple_loss=0.2486, pruned_loss=0.05846, over 949365.98 frames. ], batch size: 36, lr: 3.64e-03, grad_scale: 16.0 2023-04-27 03:06:49,255 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=66205.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 03:07:12,318 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=66221.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 03:07:34,229 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=66236.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 03:07:36,202 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.64 vs. limit=2.0 2023-04-27 03:07:43,559 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.908e+01 1.563e+02 1.889e+02 2.258e+02 3.990e+02, threshold=3.778e+02, percent-clipped=1.0 2023-04-27 03:07:55,641 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.55 vs. limit=2.0 2023-04-27 03:07:57,820 INFO [finetune.py:976] (1/7) Epoch 12, batch 3250, loss[loss=0.1949, simple_loss=0.2543, pruned_loss=0.06779, over 4875.00 frames. ], tot_loss[loss=0.1834, simple_loss=0.249, pruned_loss=0.05886, over 949448.83 frames. ], batch size: 31, lr: 3.64e-03, grad_scale: 16.0 2023-04-27 03:07:58,524 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4957, 1.4904, 4.3354, 4.0969, 3.7787, 4.1850, 4.0799, 3.7460], device='cuda:1'), covar=tensor([0.7032, 0.5577, 0.1043, 0.1711, 0.1137, 0.1772, 0.1208, 0.1412], device='cuda:1'), in_proj_covar=tensor([0.0311, 0.0306, 0.0404, 0.0408, 0.0352, 0.0409, 0.0316, 0.0370], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-27 03:07:59,773 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=66258.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 03:08:29,789 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.1712, 2.7108, 0.9841, 1.2903, 1.9666, 1.2673, 3.5392, 1.7602], device='cuda:1'), covar=tensor([0.0750, 0.0713, 0.0899, 0.1455, 0.0603, 0.1123, 0.0298, 0.0694], device='cuda:1'), in_proj_covar=tensor([0.0051, 0.0066, 0.0049, 0.0047, 0.0051, 0.0052, 0.0077, 0.0051], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0007], device='cuda:1') 2023-04-27 03:08:43,230 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([3.8960, 3.7573, 2.9997, 4.5085, 3.9145, 3.8845, 1.7447, 3.9246], device='cuda:1'), covar=tensor([0.1521, 0.1082, 0.2771, 0.1237, 0.2517, 0.1566, 0.5188, 0.2003], device='cuda:1'), in_proj_covar=tensor([0.0242, 0.0215, 0.0247, 0.0302, 0.0296, 0.0247, 0.0269, 0.0268], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 03:08:47,969 INFO [finetune.py:976] (1/7) Epoch 12, batch 3300, loss[loss=0.2089, simple_loss=0.2821, pruned_loss=0.06781, over 4818.00 frames. ], tot_loss[loss=0.1867, simple_loss=0.2534, pruned_loss=0.05997, over 952031.46 frames. ], batch size: 41, lr: 3.64e-03, grad_scale: 16.0 2023-04-27 03:08:57,543 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=66319.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 03:09:04,050 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.19 vs. limit=5.0 2023-04-27 03:09:13,362 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.300e+02 1.708e+02 2.133e+02 2.444e+02 6.112e+02, threshold=4.266e+02, percent-clipped=2.0 2023-04-27 03:09:14,659 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([4.6468, 4.4725, 3.2865, 5.3348, 4.6346, 4.6190, 2.2274, 4.5411], device='cuda:1'), covar=tensor([0.1467, 0.0864, 0.2813, 0.0925, 0.3223, 0.1689, 0.5129, 0.1982], device='cuda:1'), in_proj_covar=tensor([0.0244, 0.0217, 0.0249, 0.0304, 0.0298, 0.0248, 0.0272, 0.0270], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 03:09:21,689 INFO [finetune.py:976] (1/7) Epoch 12, batch 3350, loss[loss=0.1947, simple_loss=0.2619, pruned_loss=0.06375, over 4849.00 frames. ], tot_loss[loss=0.1871, simple_loss=0.2546, pruned_loss=0.05976, over 954574.86 frames. ], batch size: 44, lr: 3.64e-03, grad_scale: 16.0 2023-04-27 03:09:54,777 INFO [finetune.py:976] (1/7) Epoch 12, batch 3400, loss[loss=0.1813, simple_loss=0.259, pruned_loss=0.0518, over 4818.00 frames. ], tot_loss[loss=0.1899, simple_loss=0.2572, pruned_loss=0.06125, over 954540.42 frames. ], batch size: 38, lr: 3.64e-03, grad_scale: 16.0 2023-04-27 03:10:10,442 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8382, 2.3723, 1.7556, 1.8066, 1.2745, 1.3293, 1.8135, 1.2137], device='cuda:1'), covar=tensor([0.1778, 0.1376, 0.1595, 0.1802, 0.2562, 0.2115, 0.1118, 0.2173], device='cuda:1'), in_proj_covar=tensor([0.0197, 0.0215, 0.0171, 0.0204, 0.0204, 0.0185, 0.0158, 0.0186], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-27 03:10:20,215 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.071e+02 1.592e+02 1.894e+02 2.195e+02 3.276e+02, threshold=3.787e+02, percent-clipped=0.0 2023-04-27 03:10:28,167 INFO [finetune.py:976] (1/7) Epoch 12, batch 3450, loss[loss=0.1788, simple_loss=0.2496, pruned_loss=0.054, over 4786.00 frames. ], tot_loss[loss=0.1898, simple_loss=0.2577, pruned_loss=0.06095, over 955212.39 frames. ], batch size: 51, lr: 3.64e-03, grad_scale: 16.0 2023-04-27 03:10:38,039 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6140, 0.6636, 1.3931, 1.9783, 1.6961, 1.4996, 1.4914, 1.5363], device='cuda:1'), covar=tensor([0.5253, 0.7082, 0.6697, 0.6978, 0.6486, 0.8094, 0.8348, 0.7507], device='cuda:1'), in_proj_covar=tensor([0.0409, 0.0404, 0.0494, 0.0512, 0.0437, 0.0458, 0.0465, 0.0467], device='cuda:1'), out_proj_covar=tensor([9.9325e-05, 1.0025e-04, 1.1129e-04, 1.2144e-04, 1.0585e-04, 1.1060e-04, 1.1132e-04, 1.1175e-04], device='cuda:1') 2023-04-27 03:10:55,255 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=66494.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 03:10:58,890 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=66500.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 03:11:01,857 INFO [finetune.py:976] (1/7) Epoch 12, batch 3500, loss[loss=0.1643, simple_loss=0.2306, pruned_loss=0.04902, over 4914.00 frames. ], tot_loss[loss=0.1886, simple_loss=0.2559, pruned_loss=0.06062, over 956148.07 frames. ], batch size: 46, lr: 3.64e-03, grad_scale: 16.0 2023-04-27 03:11:09,121 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=66516.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 03:11:12,210 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=66521.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 03:11:26,632 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.637e+01 1.554e+02 1.930e+02 2.326e+02 4.798e+02, threshold=3.860e+02, percent-clipped=2.0 2023-04-27 03:11:27,204 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=66542.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 03:11:40,351 INFO [finetune.py:976] (1/7) Epoch 12, batch 3550, loss[loss=0.2114, simple_loss=0.2733, pruned_loss=0.07476, over 4852.00 frames. ], tot_loss[loss=0.1874, simple_loss=0.2537, pruned_loss=0.06048, over 956050.09 frames. ], batch size: 49, lr: 3.64e-03, grad_scale: 16.0 2023-04-27 03:11:50,999 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7990, 1.1205, 1.4457, 1.5379, 1.4848, 1.5950, 1.4969, 1.4363], device='cuda:1'), covar=tensor([0.3503, 0.4316, 0.4035, 0.3680, 0.4700, 0.6453, 0.4205, 0.4235], device='cuda:1'), in_proj_covar=tensor([0.0329, 0.0379, 0.0317, 0.0327, 0.0339, 0.0400, 0.0358, 0.0326], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 03:12:12,784 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=66582.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 03:12:45,654 INFO [finetune.py:976] (1/7) Epoch 12, batch 3600, loss[loss=0.1634, simple_loss=0.2355, pruned_loss=0.04565, over 4747.00 frames. ], tot_loss[loss=0.1861, simple_loss=0.2519, pruned_loss=0.06021, over 954839.52 frames. ], batch size: 54, lr: 3.64e-03, grad_scale: 16.0 2023-04-27 03:12:57,587 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=66614.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 03:13:31,411 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.076e+02 1.620e+02 1.935e+02 2.301e+02 4.046e+02, threshold=3.870e+02, percent-clipped=1.0 2023-04-27 03:13:45,502 INFO [finetune.py:976] (1/7) Epoch 12, batch 3650, loss[loss=0.223, simple_loss=0.2948, pruned_loss=0.07562, over 4874.00 frames. ], tot_loss[loss=0.1867, simple_loss=0.2526, pruned_loss=0.06038, over 953902.34 frames. ], batch size: 34, lr: 3.64e-03, grad_scale: 16.0 2023-04-27 03:13:48,091 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0856, 2.6604, 1.9295, 1.8491, 1.4693, 1.5242, 2.1134, 1.3687], device='cuda:1'), covar=tensor([0.1722, 0.1534, 0.1549, 0.1870, 0.2517, 0.2180, 0.1042, 0.2180], device='cuda:1'), in_proj_covar=tensor([0.0199, 0.0217, 0.0172, 0.0206, 0.0206, 0.0186, 0.0159, 0.0187], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-27 03:14:19,347 INFO [finetune.py:976] (1/7) Epoch 12, batch 3700, loss[loss=0.2145, simple_loss=0.2849, pruned_loss=0.07209, over 4911.00 frames. ], tot_loss[loss=0.1886, simple_loss=0.2556, pruned_loss=0.06075, over 955392.02 frames. ], batch size: 37, lr: 3.64e-03, grad_scale: 16.0 2023-04-27 03:14:21,370 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2023-04-27 03:14:35,811 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=66731.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 03:14:43,337 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.132e+02 1.645e+02 1.962e+02 2.423e+02 5.947e+02, threshold=3.923e+02, percent-clipped=4.0 2023-04-27 03:14:52,768 INFO [finetune.py:976] (1/7) Epoch 12, batch 3750, loss[loss=0.2352, simple_loss=0.2932, pruned_loss=0.08863, over 4717.00 frames. ], tot_loss[loss=0.1906, simple_loss=0.2579, pruned_loss=0.06164, over 956586.02 frames. ], batch size: 59, lr: 3.64e-03, grad_scale: 16.0 2023-04-27 03:15:06,701 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([4.9135, 4.8641, 3.3565, 5.5446, 4.8522, 4.8836, 2.4218, 4.8275], device='cuda:1'), covar=tensor([0.1592, 0.1050, 0.2805, 0.0976, 0.2660, 0.1690, 0.4936, 0.2063], device='cuda:1'), in_proj_covar=tensor([0.0244, 0.0217, 0.0249, 0.0302, 0.0297, 0.0247, 0.0270, 0.0270], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 03:15:16,255 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=66792.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 03:15:21,680 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=66800.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 03:15:25,457 INFO [finetune.py:976] (1/7) Epoch 12, batch 3800, loss[loss=0.1813, simple_loss=0.2642, pruned_loss=0.04924, over 4855.00 frames. ], tot_loss[loss=0.1912, simple_loss=0.2589, pruned_loss=0.06172, over 955331.69 frames. ], batch size: 44, lr: 3.64e-03, grad_scale: 16.0 2023-04-27 03:15:32,280 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=66816.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 03:15:48,321 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.153e+02 1.670e+02 1.892e+02 2.401e+02 5.287e+02, threshold=3.784e+02, percent-clipped=2.0 2023-04-27 03:15:52,971 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=66848.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 03:15:58,164 INFO [finetune.py:976] (1/7) Epoch 12, batch 3850, loss[loss=0.2088, simple_loss=0.2702, pruned_loss=0.07369, over 4915.00 frames. ], tot_loss[loss=0.1895, simple_loss=0.2572, pruned_loss=0.06092, over 954000.20 frames. ], batch size: 37, lr: 3.64e-03, grad_scale: 16.0 2023-04-27 03:16:04,610 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=66864.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 03:16:07,765 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-04-27 03:16:12,476 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=66877.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 03:16:12,530 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7925, 2.3178, 1.7348, 1.6409, 1.2987, 1.3071, 1.8177, 1.1584], device='cuda:1'), covar=tensor([0.1503, 0.1217, 0.1421, 0.1694, 0.2273, 0.1881, 0.0944, 0.2021], device='cuda:1'), in_proj_covar=tensor([0.0196, 0.0215, 0.0171, 0.0204, 0.0204, 0.0185, 0.0158, 0.0185], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-27 03:16:20,958 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=2.17 vs. limit=2.0 2023-04-27 03:16:31,349 INFO [finetune.py:976] (1/7) Epoch 12, batch 3900, loss[loss=0.1847, simple_loss=0.2545, pruned_loss=0.05747, over 4800.00 frames. ], tot_loss[loss=0.1874, simple_loss=0.2542, pruned_loss=0.06031, over 952207.45 frames. ], batch size: 45, lr: 3.64e-03, grad_scale: 16.0 2023-04-27 03:16:32,097 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0631, 2.5483, 2.1766, 2.2705, 1.8377, 2.0117, 2.1103, 1.6310], device='cuda:1'), covar=tensor([0.2247, 0.1202, 0.0825, 0.1380, 0.3241, 0.1273, 0.2102, 0.2800], device='cuda:1'), in_proj_covar=tensor([0.0293, 0.0313, 0.0224, 0.0284, 0.0313, 0.0263, 0.0253, 0.0272], device='cuda:1'), out_proj_covar=tensor([1.1794e-04, 1.2502e-04, 8.9495e-05, 1.1341e-04, 1.2759e-04, 1.0554e-04, 1.0296e-04, 1.0867e-04], device='cuda:1') 2023-04-27 03:16:38,402 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=66914.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 03:17:12,162 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.033e+02 1.631e+02 1.862e+02 2.227e+02 3.932e+02, threshold=3.724e+02, percent-clipped=1.0 2023-04-27 03:17:32,064 INFO [finetune.py:976] (1/7) Epoch 12, batch 3950, loss[loss=0.1739, simple_loss=0.2292, pruned_loss=0.05935, over 4780.00 frames. ], tot_loss[loss=0.1845, simple_loss=0.2509, pruned_loss=0.05902, over 953778.48 frames. ], batch size: 26, lr: 3.64e-03, grad_scale: 16.0 2023-04-27 03:17:33,391 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=66957.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 03:17:42,917 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=66962.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 03:18:39,567 INFO [finetune.py:976] (1/7) Epoch 12, batch 4000, loss[loss=0.1447, simple_loss=0.2155, pruned_loss=0.03698, over 4766.00 frames. ], tot_loss[loss=0.1844, simple_loss=0.2507, pruned_loss=0.05905, over 953109.18 frames. ], batch size: 26, lr: 3.63e-03, grad_scale: 16.0 2023-04-27 03:18:58,942 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=67018.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 03:19:14,014 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.105e+02 1.700e+02 2.045e+02 2.371e+02 4.368e+02, threshold=4.090e+02, percent-clipped=3.0 2023-04-27 03:19:21,912 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([4.5544, 4.4331, 3.2261, 5.1575, 4.5219, 4.4951, 1.8291, 4.4049], device='cuda:1'), covar=tensor([0.1525, 0.0899, 0.3116, 0.0946, 0.3465, 0.1495, 0.5924, 0.2364], device='cuda:1'), in_proj_covar=tensor([0.0243, 0.0217, 0.0249, 0.0304, 0.0297, 0.0247, 0.0271, 0.0271], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 03:19:22,922 INFO [finetune.py:976] (1/7) Epoch 12, batch 4050, loss[loss=0.2104, simple_loss=0.283, pruned_loss=0.06891, over 4808.00 frames. ], tot_loss[loss=0.188, simple_loss=0.2549, pruned_loss=0.06056, over 952486.90 frames. ], batch size: 38, lr: 3.63e-03, grad_scale: 16.0 2023-04-27 03:19:42,050 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=67083.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 03:19:44,506 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=67087.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 03:19:52,463 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.53 vs. limit=5.0 2023-04-27 03:19:56,227 INFO [finetune.py:976] (1/7) Epoch 12, batch 4100, loss[loss=0.1717, simple_loss=0.2459, pruned_loss=0.04879, over 4792.00 frames. ], tot_loss[loss=0.1899, simple_loss=0.2579, pruned_loss=0.06095, over 953930.92 frames. ], batch size: 29, lr: 3.63e-03, grad_scale: 16.0 2023-04-27 03:20:17,669 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7888, 1.6576, 1.9423, 2.2427, 1.9724, 1.7287, 1.8125, 1.7483], device='cuda:1'), covar=tensor([0.4414, 0.6162, 0.6207, 0.5870, 0.5526, 0.8499, 0.8455, 0.8730], device='cuda:1'), in_proj_covar=tensor([0.0412, 0.0407, 0.0498, 0.0516, 0.0441, 0.0462, 0.0468, 0.0471], device='cuda:1'), out_proj_covar=tensor([9.9998e-05, 1.0090e-04, 1.1224e-04, 1.2246e-04, 1.0686e-04, 1.1150e-04, 1.1212e-04, 1.1275e-04], device='cuda:1') 2023-04-27 03:20:21,136 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.154e+02 1.668e+02 1.950e+02 2.451e+02 4.466e+02, threshold=3.899e+02, percent-clipped=3.0 2023-04-27 03:20:22,488 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=67144.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 03:20:29,535 INFO [finetune.py:976] (1/7) Epoch 12, batch 4150, loss[loss=0.1694, simple_loss=0.2293, pruned_loss=0.05482, over 3943.00 frames. ], tot_loss[loss=0.191, simple_loss=0.2588, pruned_loss=0.06158, over 951235.59 frames. ], batch size: 17, lr: 3.63e-03, grad_scale: 16.0 2023-04-27 03:20:45,407 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=67177.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 03:21:03,552 INFO [finetune.py:976] (1/7) Epoch 12, batch 4200, loss[loss=0.1986, simple_loss=0.2708, pruned_loss=0.06323, over 4787.00 frames. ], tot_loss[loss=0.1898, simple_loss=0.2578, pruned_loss=0.06092, over 951239.93 frames. ], batch size: 29, lr: 3.63e-03, grad_scale: 16.0 2023-04-27 03:21:18,094 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=67225.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 03:21:27,723 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5008, 1.4574, 1.8034, 1.8395, 1.3245, 1.2525, 1.5351, 0.9661], device='cuda:1'), covar=tensor([0.0708, 0.0790, 0.0527, 0.0835, 0.1024, 0.1269, 0.0906, 0.0861], device='cuda:1'), in_proj_covar=tensor([0.0068, 0.0071, 0.0071, 0.0067, 0.0075, 0.0096, 0.0076, 0.0071], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-27 03:21:28,835 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.013e+02 1.623e+02 1.873e+02 2.449e+02 3.643e+02, threshold=3.747e+02, percent-clipped=0.0 2023-04-27 03:21:37,113 INFO [finetune.py:976] (1/7) Epoch 12, batch 4250, loss[loss=0.1572, simple_loss=0.2137, pruned_loss=0.05031, over 4270.00 frames. ], tot_loss[loss=0.1883, simple_loss=0.2553, pruned_loss=0.06066, over 951476.46 frames. ], batch size: 18, lr: 3.63e-03, grad_scale: 16.0 2023-04-27 03:21:56,653 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9558, 1.7338, 2.1680, 2.4584, 2.0762, 1.9091, 2.0039, 1.9683], device='cuda:1'), covar=tensor([0.5652, 0.7685, 0.8423, 0.7228, 0.6478, 0.9793, 1.0079, 1.0914], device='cuda:1'), in_proj_covar=tensor([0.0412, 0.0406, 0.0497, 0.0516, 0.0441, 0.0461, 0.0468, 0.0470], device='cuda:1'), out_proj_covar=tensor([9.9899e-05, 1.0079e-04, 1.1196e-04, 1.2228e-04, 1.0682e-04, 1.1130e-04, 1.1210e-04, 1.1251e-04], device='cuda:1') 2023-04-27 03:22:10,185 INFO [finetune.py:976] (1/7) Epoch 12, batch 4300, loss[loss=0.1649, simple_loss=0.2234, pruned_loss=0.05319, over 4702.00 frames. ], tot_loss[loss=0.185, simple_loss=0.2517, pruned_loss=0.05915, over 950008.00 frames. ], batch size: 23, lr: 3.63e-03, grad_scale: 16.0 2023-04-27 03:22:15,625 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=67313.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 03:22:36,119 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.106e+02 1.576e+02 1.859e+02 2.202e+02 4.297e+02, threshold=3.717e+02, percent-clipped=1.0 2023-04-27 03:22:43,956 INFO [finetune.py:976] (1/7) Epoch 12, batch 4350, loss[loss=0.1956, simple_loss=0.2542, pruned_loss=0.0685, over 4760.00 frames. ], tot_loss[loss=0.1833, simple_loss=0.2494, pruned_loss=0.05858, over 950781.43 frames. ], batch size: 28, lr: 3.63e-03, grad_scale: 16.0 2023-04-27 03:23:17,579 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6849, 4.2510, 0.7666, 2.1831, 2.3920, 2.6772, 2.4904, 1.1137], device='cuda:1'), covar=tensor([0.1497, 0.0983, 0.2345, 0.1319, 0.0998, 0.1153, 0.1385, 0.2043], device='cuda:1'), in_proj_covar=tensor([0.0118, 0.0249, 0.0140, 0.0122, 0.0134, 0.0153, 0.0118, 0.0121], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-27 03:23:29,030 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=67387.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 03:23:51,833 INFO [finetune.py:976] (1/7) Epoch 12, batch 4400, loss[loss=0.1366, simple_loss=0.2015, pruned_loss=0.03588, over 4197.00 frames. ], tot_loss[loss=0.1857, simple_loss=0.2513, pruned_loss=0.06006, over 950842.58 frames. ], batch size: 18, lr: 3.63e-03, grad_scale: 16.0 2023-04-27 03:24:11,484 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.62 vs. limit=2.0 2023-04-27 03:24:26,227 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=67435.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 03:24:34,555 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=67439.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 03:24:36,312 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.200e+02 1.735e+02 1.991e+02 2.507e+02 6.130e+02, threshold=3.981e+02, percent-clipped=5.0 2023-04-27 03:24:43,282 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5080, 1.0669, 1.2491, 1.2134, 1.6915, 1.3432, 1.0839, 1.2231], device='cuda:1'), covar=tensor([0.1671, 0.1295, 0.2199, 0.1368, 0.0778, 0.1514, 0.1996, 0.1999], device='cuda:1'), in_proj_covar=tensor([0.0307, 0.0318, 0.0352, 0.0296, 0.0333, 0.0315, 0.0305, 0.0361], device='cuda:1'), out_proj_covar=tensor([6.4100e-05, 6.7014e-05, 7.5542e-05, 6.0708e-05, 6.9476e-05, 6.6942e-05, 6.4953e-05, 7.7353e-05], device='cuda:1') 2023-04-27 03:24:56,594 INFO [finetune.py:976] (1/7) Epoch 12, batch 4450, loss[loss=0.1806, simple_loss=0.2476, pruned_loss=0.05685, over 4885.00 frames. ], tot_loss[loss=0.1883, simple_loss=0.255, pruned_loss=0.06079, over 949539.17 frames. ], batch size: 32, lr: 3.63e-03, grad_scale: 16.0 2023-04-27 03:25:48,054 INFO [finetune.py:976] (1/7) Epoch 12, batch 4500, loss[loss=0.2168, simple_loss=0.2794, pruned_loss=0.07705, over 4197.00 frames. ], tot_loss[loss=0.1907, simple_loss=0.2583, pruned_loss=0.06161, over 951429.50 frames. ], batch size: 65, lr: 3.63e-03, grad_scale: 16.0 2023-04-27 03:25:48,147 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.3232, 3.0344, 1.0992, 1.6368, 2.5199, 1.5280, 4.3632, 2.0791], device='cuda:1'), covar=tensor([0.0688, 0.0734, 0.0890, 0.1267, 0.0481, 0.0999, 0.0179, 0.0621], device='cuda:1'), in_proj_covar=tensor([0.0051, 0.0066, 0.0049, 0.0047, 0.0051, 0.0053, 0.0077, 0.0051], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:1') 2023-04-27 03:25:53,047 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=67513.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 03:26:04,060 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5018, 1.2785, 0.5150, 1.2084, 1.3511, 1.3561, 1.2624, 1.3074], device='cuda:1'), covar=tensor([0.0536, 0.0456, 0.0429, 0.0629, 0.0326, 0.0560, 0.0541, 0.0655], device='cuda:1'), in_proj_covar=tensor([0.0029, 0.0025, 0.0023, 0.0030, 0.0020, 0.0029, 0.0029, 0.0029], device='cuda:1'), out_proj_covar=tensor([0.0050, 0.0045, 0.0038, 0.0050, 0.0038, 0.0049, 0.0049, 0.0050], device='cuda:1') 2023-04-27 03:26:08,741 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6948, 1.6915, 0.8491, 1.3826, 1.7389, 1.5412, 1.4357, 1.5095], device='cuda:1'), covar=tensor([0.0524, 0.0411, 0.0382, 0.0608, 0.0277, 0.0561, 0.0528, 0.0603], device='cuda:1'), in_proj_covar=tensor([0.0029, 0.0025, 0.0023, 0.0030, 0.0020, 0.0029, 0.0029, 0.0029], device='cuda:1'), out_proj_covar=tensor([0.0050, 0.0045, 0.0038, 0.0050, 0.0038, 0.0049, 0.0049, 0.0050], device='cuda:1') 2023-04-27 03:26:12,842 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.294e+02 1.699e+02 1.983e+02 2.527e+02 4.329e+02, threshold=3.965e+02, percent-clipped=1.0 2023-04-27 03:26:22,245 INFO [finetune.py:976] (1/7) Epoch 12, batch 4550, loss[loss=0.2356, simple_loss=0.2909, pruned_loss=0.09012, over 4879.00 frames. ], tot_loss[loss=0.1915, simple_loss=0.259, pruned_loss=0.06195, over 951176.58 frames. ], batch size: 43, lr: 3.63e-03, grad_scale: 32.0 2023-04-27 03:26:34,562 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=67574.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 03:26:56,267 INFO [finetune.py:976] (1/7) Epoch 12, batch 4600, loss[loss=0.1883, simple_loss=0.2557, pruned_loss=0.06047, over 4836.00 frames. ], tot_loss[loss=0.1903, simple_loss=0.258, pruned_loss=0.06131, over 952425.51 frames. ], batch size: 49, lr: 3.63e-03, grad_scale: 32.0 2023-04-27 03:26:56,388 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9456, 1.7038, 1.9723, 2.3824, 1.5945, 1.2807, 1.7489, 0.9563], device='cuda:1'), covar=tensor([0.0672, 0.0732, 0.0675, 0.0859, 0.0840, 0.1776, 0.0885, 0.1027], device='cuda:1'), in_proj_covar=tensor([0.0067, 0.0070, 0.0070, 0.0066, 0.0074, 0.0095, 0.0075, 0.0069], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-27 03:27:01,304 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=67613.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 03:27:01,350 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.9042, 2.2602, 2.0894, 2.2625, 1.9851, 2.1660, 2.1239, 2.1209], device='cuda:1'), covar=tensor([0.4659, 0.7048, 0.6175, 0.5665, 0.6768, 0.8780, 0.7855, 0.6707], device='cuda:1'), in_proj_covar=tensor([0.0329, 0.0376, 0.0315, 0.0326, 0.0338, 0.0399, 0.0357, 0.0324], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 03:27:20,971 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.110e+02 1.683e+02 1.939e+02 2.322e+02 5.503e+02, threshold=3.878e+02, percent-clipped=1.0 2023-04-27 03:27:29,855 INFO [finetune.py:976] (1/7) Epoch 12, batch 4650, loss[loss=0.1513, simple_loss=0.2202, pruned_loss=0.04124, over 4934.00 frames. ], tot_loss[loss=0.1878, simple_loss=0.2551, pruned_loss=0.06022, over 954512.97 frames. ], batch size: 33, lr: 3.63e-03, grad_scale: 16.0 2023-04-27 03:27:34,154 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=67661.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 03:27:37,365 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7483, 1.3758, 1.3969, 1.7220, 2.0218, 1.6783, 1.4359, 1.2858], device='cuda:1'), covar=tensor([0.1899, 0.1525, 0.1994, 0.1366, 0.0902, 0.1848, 0.2354, 0.2301], device='cuda:1'), in_proj_covar=tensor([0.0305, 0.0317, 0.0351, 0.0295, 0.0332, 0.0314, 0.0304, 0.0360], device='cuda:1'), out_proj_covar=tensor([6.3688e-05, 6.6601e-05, 7.5461e-05, 6.0581e-05, 6.9448e-05, 6.6800e-05, 6.4629e-05, 7.7174e-05], device='cuda:1') 2023-04-27 03:27:46,433 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=67680.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 03:27:51,529 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-04-27 03:27:54,846 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=67692.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 03:28:04,214 INFO [finetune.py:976] (1/7) Epoch 12, batch 4700, loss[loss=0.1301, simple_loss=0.2009, pruned_loss=0.02961, over 4902.00 frames. ], tot_loss[loss=0.1858, simple_loss=0.2524, pruned_loss=0.05957, over 953096.40 frames. ], batch size: 32, lr: 3.63e-03, grad_scale: 16.0 2023-04-27 03:28:08,536 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.30 vs. limit=5.0 2023-04-27 03:28:18,321 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8001, 2.1580, 1.6320, 1.3088, 1.3001, 1.2844, 1.6094, 1.1758], device='cuda:1'), covar=tensor([0.1773, 0.1289, 0.1692, 0.2031, 0.2531, 0.2132, 0.1152, 0.2203], device='cuda:1'), in_proj_covar=tensor([0.0196, 0.0214, 0.0171, 0.0204, 0.0203, 0.0185, 0.0158, 0.0186], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-27 03:28:28,335 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.2599, 1.9736, 2.5083, 2.8580, 2.7791, 2.2066, 1.9035, 2.3885], device='cuda:1'), covar=tensor([0.0942, 0.1152, 0.0621, 0.0526, 0.0566, 0.0850, 0.0878, 0.0562], device='cuda:1'), in_proj_covar=tensor([0.0192, 0.0201, 0.0182, 0.0173, 0.0177, 0.0185, 0.0155, 0.0179], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 03:28:30,160 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4681, 3.5774, 1.0711, 1.8425, 1.8532, 2.5304, 1.9587, 1.0371], device='cuda:1'), covar=tensor([0.1578, 0.0966, 0.2061, 0.1435, 0.1212, 0.1095, 0.1704, 0.2065], device='cuda:1'), in_proj_covar=tensor([0.0119, 0.0251, 0.0140, 0.0122, 0.0134, 0.0154, 0.0119, 0.0122], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-27 03:28:37,289 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=67739.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 03:28:38,546 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=67741.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 03:28:40,069 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.078e+02 1.659e+02 1.914e+02 2.232e+02 4.498e+02, threshold=3.829e+02, percent-clipped=2.0 2023-04-27 03:28:51,845 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=67753.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 03:28:59,227 INFO [finetune.py:976] (1/7) Epoch 12, batch 4750, loss[loss=0.1831, simple_loss=0.2533, pruned_loss=0.0565, over 4833.00 frames. ], tot_loss[loss=0.1845, simple_loss=0.2504, pruned_loss=0.05931, over 952615.62 frames. ], batch size: 39, lr: 3.63e-03, grad_scale: 16.0 2023-04-27 03:29:02,244 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.72 vs. limit=2.0 2023-04-27 03:29:42,891 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=67787.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 03:29:58,513 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4070, 1.2773, 1.6344, 1.7020, 1.2713, 1.0276, 1.2735, 0.8801], device='cuda:1'), covar=tensor([0.0730, 0.0747, 0.0514, 0.0640, 0.0816, 0.1867, 0.0762, 0.0811], device='cuda:1'), in_proj_covar=tensor([0.0067, 0.0070, 0.0070, 0.0066, 0.0075, 0.0096, 0.0075, 0.0070], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-27 03:30:06,231 INFO [finetune.py:976] (1/7) Epoch 12, batch 4800, loss[loss=0.1825, simple_loss=0.2526, pruned_loss=0.05616, over 4863.00 frames. ], tot_loss[loss=0.1872, simple_loss=0.253, pruned_loss=0.06066, over 954588.58 frames. ], batch size: 31, lr: 3.63e-03, grad_scale: 16.0 2023-04-27 03:30:24,393 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-04-27 03:30:35,557 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.067e+02 1.720e+02 2.004e+02 2.760e+02 5.659e+02, threshold=4.008e+02, percent-clipped=5.0 2023-04-27 03:30:43,895 INFO [finetune.py:976] (1/7) Epoch 12, batch 4850, loss[loss=0.1971, simple_loss=0.2755, pruned_loss=0.05941, over 4830.00 frames. ], tot_loss[loss=0.1881, simple_loss=0.2553, pruned_loss=0.06041, over 953373.27 frames. ], batch size: 33, lr: 3.63e-03, grad_scale: 16.0 2023-04-27 03:30:49,746 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.4925, 2.1874, 2.8467, 3.1422, 2.9671, 2.5305, 2.0602, 2.6877], device='cuda:1'), covar=tensor([0.1026, 0.1186, 0.0627, 0.0582, 0.0598, 0.0908, 0.0927, 0.0649], device='cuda:1'), in_proj_covar=tensor([0.0192, 0.0202, 0.0182, 0.0174, 0.0178, 0.0185, 0.0155, 0.0180], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 03:30:54,467 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=67869.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 03:31:17,822 INFO [finetune.py:976] (1/7) Epoch 12, batch 4900, loss[loss=0.2053, simple_loss=0.2606, pruned_loss=0.07495, over 4759.00 frames. ], tot_loss[loss=0.1892, simple_loss=0.2567, pruned_loss=0.06088, over 952245.04 frames. ], batch size: 28, lr: 3.63e-03, grad_scale: 16.0 2023-04-27 03:31:43,017 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.676e+01 1.571e+02 1.848e+02 2.234e+02 5.872e+02, threshold=3.696e+02, percent-clipped=2.0 2023-04-27 03:31:51,276 INFO [finetune.py:976] (1/7) Epoch 12, batch 4950, loss[loss=0.1801, simple_loss=0.2567, pruned_loss=0.05178, over 4901.00 frames. ], tot_loss[loss=0.1903, simple_loss=0.2585, pruned_loss=0.06102, over 953817.07 frames. ], batch size: 36, lr: 3.63e-03, grad_scale: 16.0 2023-04-27 03:32:09,683 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=67981.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 03:32:20,592 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=3.67 vs. limit=5.0 2023-04-27 03:32:23,001 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([0.3698, 1.3379, 1.4243, 1.0166, 1.4112, 1.1184, 1.6422, 1.3028], device='cuda:1'), covar=tensor([0.3552, 0.1820, 0.4879, 0.2471, 0.1359, 0.2239, 0.1540, 0.4464], device='cuda:1'), in_proj_covar=tensor([0.0341, 0.0346, 0.0428, 0.0355, 0.0378, 0.0379, 0.0373, 0.0417], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 03:32:26,418 INFO [finetune.py:976] (1/7) Epoch 12, batch 5000, loss[loss=0.1635, simple_loss=0.2378, pruned_loss=0.0446, over 4900.00 frames. ], tot_loss[loss=0.1888, simple_loss=0.2566, pruned_loss=0.06046, over 953610.47 frames. ], batch size: 35, lr: 3.63e-03, grad_scale: 16.0 2023-04-27 03:32:27,158 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5743, 1.3900, 1.7983, 1.8259, 1.3555, 1.2480, 1.4445, 0.9712], device='cuda:1'), covar=tensor([0.0631, 0.0872, 0.0453, 0.0695, 0.0989, 0.1329, 0.0793, 0.0753], device='cuda:1'), in_proj_covar=tensor([0.0067, 0.0070, 0.0070, 0.0067, 0.0075, 0.0096, 0.0075, 0.0070], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-27 03:32:47,883 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=68036.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 03:32:51,617 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=68042.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 03:32:52,121 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.570e+01 1.647e+02 1.940e+02 2.431e+02 4.879e+02, threshold=3.879e+02, percent-clipped=2.0 2023-04-27 03:32:55,685 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=68048.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 03:32:59,805 INFO [finetune.py:976] (1/7) Epoch 12, batch 5050, loss[loss=0.1555, simple_loss=0.2245, pruned_loss=0.04326, over 4751.00 frames. ], tot_loss[loss=0.1869, simple_loss=0.2541, pruned_loss=0.05987, over 954007.09 frames. ], batch size: 27, lr: 3.63e-03, grad_scale: 16.0 2023-04-27 03:33:22,289 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=68087.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 03:33:33,617 INFO [finetune.py:976] (1/7) Epoch 12, batch 5100, loss[loss=0.1686, simple_loss=0.232, pruned_loss=0.05254, over 4074.00 frames. ], tot_loss[loss=0.1833, simple_loss=0.2503, pruned_loss=0.05814, over 954854.83 frames. ], batch size: 65, lr: 3.63e-03, grad_scale: 16.0 2023-04-27 03:34:11,406 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.851e+01 1.609e+02 1.945e+02 2.272e+02 5.546e+02, threshold=3.889e+02, percent-clipped=1.0 2023-04-27 03:34:14,581 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=68148.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 03:34:18,644 INFO [finetune.py:976] (1/7) Epoch 12, batch 5150, loss[loss=0.1707, simple_loss=0.2409, pruned_loss=0.05023, over 4782.00 frames. ], tot_loss[loss=0.1825, simple_loss=0.2494, pruned_loss=0.05775, over 952529.66 frames. ], batch size: 26, lr: 3.63e-03, grad_scale: 16.0 2023-04-27 03:34:34,046 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=68169.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 03:34:34,725 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4400, 2.5855, 2.1996, 2.2934, 2.7192, 2.3116, 3.6202, 1.9556], device='cuda:1'), covar=tensor([0.4514, 0.2444, 0.4815, 0.3600, 0.2082, 0.2872, 0.1884, 0.4321], device='cuda:1'), in_proj_covar=tensor([0.0341, 0.0345, 0.0428, 0.0356, 0.0379, 0.0379, 0.0373, 0.0417], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 03:34:44,064 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.81 vs. limit=2.0 2023-04-27 03:35:12,015 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-04-27 03:35:16,279 INFO [finetune.py:976] (1/7) Epoch 12, batch 5200, loss[loss=0.2972, simple_loss=0.3477, pruned_loss=0.1233, over 4783.00 frames. ], tot_loss[loss=0.1855, simple_loss=0.2528, pruned_loss=0.05913, over 952378.54 frames. ], batch size: 59, lr: 3.62e-03, grad_scale: 16.0 2023-04-27 03:35:34,920 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=68217.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 03:36:09,269 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.144e+02 1.715e+02 2.014e+02 2.436e+02 3.767e+02, threshold=4.027e+02, percent-clipped=0.0 2023-04-27 03:36:18,613 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8946, 1.6555, 4.1048, 3.8733, 3.6725, 3.8037, 3.7004, 3.7053], device='cuda:1'), covar=tensor([0.6151, 0.4865, 0.0982, 0.1440, 0.0947, 0.1730, 0.3281, 0.1238], device='cuda:1'), in_proj_covar=tensor([0.0313, 0.0308, 0.0404, 0.0411, 0.0352, 0.0411, 0.0316, 0.0373], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-27 03:36:22,260 INFO [finetune.py:976] (1/7) Epoch 12, batch 5250, loss[loss=0.1629, simple_loss=0.2481, pruned_loss=0.03884, over 4749.00 frames. ], tot_loss[loss=0.1876, simple_loss=0.2554, pruned_loss=0.0599, over 949372.28 frames. ], batch size: 27, lr: 3.62e-03, grad_scale: 16.0 2023-04-27 03:36:29,958 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.1723, 1.4900, 1.3242, 1.7576, 1.5026, 1.8924, 1.3146, 3.5114], device='cuda:1'), covar=tensor([0.0753, 0.1072, 0.1007, 0.1328, 0.0827, 0.0698, 0.1013, 0.0220], device='cuda:1'), in_proj_covar=tensor([0.0038, 0.0039, 0.0040, 0.0044, 0.0040, 0.0038, 0.0038, 0.0057], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:1') 2023-04-27 03:37:07,992 INFO [finetune.py:976] (1/7) Epoch 12, batch 5300, loss[loss=0.1096, simple_loss=0.18, pruned_loss=0.01957, over 4455.00 frames. ], tot_loss[loss=0.1875, simple_loss=0.2556, pruned_loss=0.05967, over 949144.71 frames. ], batch size: 19, lr: 3.62e-03, grad_scale: 16.0 2023-04-27 03:37:28,992 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-04-27 03:37:29,898 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=68336.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 03:37:30,466 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=68337.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 03:37:34,487 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.347e+01 1.659e+02 1.967e+02 2.376e+02 7.370e+02, threshold=3.933e+02, percent-clipped=4.0 2023-04-27 03:37:37,596 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=68348.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 03:37:41,847 INFO [finetune.py:976] (1/7) Epoch 12, batch 5350, loss[loss=0.1696, simple_loss=0.2482, pruned_loss=0.04547, over 4917.00 frames. ], tot_loss[loss=0.1864, simple_loss=0.2549, pruned_loss=0.05894, over 951308.66 frames. ], batch size: 37, lr: 3.62e-03, grad_scale: 16.0 2023-04-27 03:38:00,875 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=68384.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 03:38:10,025 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=68396.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 03:38:11,911 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=68399.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 03:38:15,658 INFO [finetune.py:976] (1/7) Epoch 12, batch 5400, loss[loss=0.2, simple_loss=0.2615, pruned_loss=0.06928, over 4824.00 frames. ], tot_loss[loss=0.1861, simple_loss=0.2539, pruned_loss=0.05919, over 952329.92 frames. ], batch size: 38, lr: 3.62e-03, grad_scale: 16.0 2023-04-27 03:38:41,371 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.158e+02 1.609e+02 1.919e+02 2.411e+02 6.706e+02, threshold=3.839e+02, percent-clipped=2.0 2023-04-27 03:38:41,452 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=68443.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 03:38:49,104 INFO [finetune.py:976] (1/7) Epoch 12, batch 5450, loss[loss=0.1717, simple_loss=0.2506, pruned_loss=0.04637, over 4824.00 frames. ], tot_loss[loss=0.1844, simple_loss=0.2517, pruned_loss=0.05857, over 953532.40 frames. ], batch size: 41, lr: 3.62e-03, grad_scale: 16.0 2023-04-27 03:38:52,251 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=68460.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 03:39:10,461 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9160, 1.8507, 1.0164, 1.6025, 1.9340, 1.7619, 1.7001, 1.7235], device='cuda:1'), covar=tensor([0.0454, 0.0340, 0.0327, 0.0511, 0.0259, 0.0514, 0.0493, 0.0501], device='cuda:1'), in_proj_covar=tensor([0.0029, 0.0025, 0.0023, 0.0029, 0.0020, 0.0029, 0.0028, 0.0029], device='cuda:1'), out_proj_covar=tensor([0.0049, 0.0044, 0.0037, 0.0050, 0.0037, 0.0048, 0.0048, 0.0050], device='cuda:1') 2023-04-27 03:39:33,462 INFO [finetune.py:976] (1/7) Epoch 12, batch 5500, loss[loss=0.1869, simple_loss=0.2596, pruned_loss=0.0571, over 4809.00 frames. ], tot_loss[loss=0.182, simple_loss=0.2489, pruned_loss=0.05755, over 956229.71 frames. ], batch size: 33, lr: 3.62e-03, grad_scale: 16.0 2023-04-27 03:39:49,487 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0201, 2.7051, 2.2624, 2.4879, 1.8952, 2.2220, 2.4528, 1.7870], device='cuda:1'), covar=tensor([0.2242, 0.1036, 0.0870, 0.1133, 0.2911, 0.1038, 0.1923, 0.2801], device='cuda:1'), in_proj_covar=tensor([0.0294, 0.0314, 0.0226, 0.0286, 0.0314, 0.0265, 0.0256, 0.0274], device='cuda:1'), out_proj_covar=tensor([1.1858e-04, 1.2533e-04, 9.0514e-05, 1.1423e-04, 1.2811e-04, 1.0591e-04, 1.0405e-04, 1.0945e-04], device='cuda:1') 2023-04-27 03:39:58,193 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.108e+02 1.589e+02 1.966e+02 2.318e+02 5.103e+02, threshold=3.932e+02, percent-clipped=5.0 2023-04-27 03:40:06,516 INFO [finetune.py:976] (1/7) Epoch 12, batch 5550, loss[loss=0.194, simple_loss=0.2573, pruned_loss=0.06532, over 4874.00 frames. ], tot_loss[loss=0.1839, simple_loss=0.2505, pruned_loss=0.05865, over 956215.49 frames. ], batch size: 34, lr: 3.62e-03, grad_scale: 16.0 2023-04-27 03:40:21,074 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=68578.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 03:40:54,927 INFO [finetune.py:976] (1/7) Epoch 12, batch 5600, loss[loss=0.1722, simple_loss=0.2536, pruned_loss=0.04537, over 4736.00 frames. ], tot_loss[loss=0.1862, simple_loss=0.2537, pruned_loss=0.05934, over 954426.01 frames. ], batch size: 59, lr: 3.62e-03, grad_scale: 16.0 2023-04-27 03:41:36,475 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=68637.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 03:41:36,509 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0827, 2.6214, 2.2364, 2.4691, 1.7689, 2.1490, 2.2640, 1.8015], device='cuda:1'), covar=tensor([0.2161, 0.1469, 0.0904, 0.1285, 0.3394, 0.1222, 0.1935, 0.2904], device='cuda:1'), in_proj_covar=tensor([0.0293, 0.0313, 0.0226, 0.0284, 0.0313, 0.0264, 0.0256, 0.0273], device='cuda:1'), out_proj_covar=tensor([1.1819e-04, 1.2484e-04, 9.0193e-05, 1.1376e-04, 1.2779e-04, 1.0575e-04, 1.0383e-04, 1.0918e-04], device='cuda:1') 2023-04-27 03:41:37,657 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=68639.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 03:41:46,140 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.206e+02 1.587e+02 1.869e+02 2.457e+02 5.644e+02, threshold=3.738e+02, percent-clipped=6.0 2023-04-27 03:41:58,625 INFO [finetune.py:976] (1/7) Epoch 12, batch 5650, loss[loss=0.1988, simple_loss=0.2723, pruned_loss=0.06259, over 4896.00 frames. ], tot_loss[loss=0.1885, simple_loss=0.2569, pruned_loss=0.06003, over 954661.45 frames. ], batch size: 35, lr: 3.62e-03, grad_scale: 16.0 2023-04-27 03:42:33,429 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=68685.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 03:42:33,473 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=68685.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 03:42:39,468 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8767, 2.1893, 1.8273, 1.8053, 1.5180, 1.5099, 1.8275, 1.4286], device='cuda:1'), covar=tensor([0.0953, 0.1113, 0.0888, 0.1081, 0.1263, 0.1102, 0.0568, 0.1175], device='cuda:1'), in_proj_covar=tensor([0.0196, 0.0213, 0.0169, 0.0204, 0.0202, 0.0185, 0.0157, 0.0185], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-27 03:42:44,212 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([3.1342, 2.8025, 3.1977, 3.7648, 3.0950, 2.6468, 2.6533, 3.0682], device='cuda:1'), covar=tensor([0.2961, 0.2694, 0.1306, 0.2165, 0.2318, 0.2253, 0.3403, 0.1719], device='cuda:1'), in_proj_covar=tensor([0.0287, 0.0249, 0.0224, 0.0317, 0.0215, 0.0229, 0.0232, 0.0185], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-27 03:42:45,294 INFO [finetune.py:976] (1/7) Epoch 12, batch 5700, loss[loss=0.17, simple_loss=0.2268, pruned_loss=0.0566, over 4386.00 frames. ], tot_loss[loss=0.1866, simple_loss=0.2543, pruned_loss=0.05949, over 940267.18 frames. ], batch size: 19, lr: 3.62e-03, grad_scale: 16.0 2023-04-27 03:43:16,427 INFO [finetune.py:976] (1/7) Epoch 13, batch 0, loss[loss=0.2145, simple_loss=0.2834, pruned_loss=0.07277, over 4884.00 frames. ], tot_loss[loss=0.2145, simple_loss=0.2834, pruned_loss=0.07277, over 4884.00 frames. ], batch size: 43, lr: 3.62e-03, grad_scale: 16.0 2023-04-27 03:43:16,427 INFO [finetune.py:1001] (1/7) Computing validation loss 2023-04-27 03:43:32,176 INFO [finetune.py:1010] (1/7) Epoch 13, validation: loss=0.1542, simple_loss=0.2264, pruned_loss=0.04102, over 2265189.00 frames. 2023-04-27 03:43:32,176 INFO [finetune.py:1011] (1/7) Maximum memory allocated so far is 6391MB 2023-04-27 03:43:49,890 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.289e+01 1.578e+02 1.937e+02 2.291e+02 5.419e+02, threshold=3.875e+02, percent-clipped=2.0 2023-04-27 03:43:50,011 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=68743.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 03:43:51,891 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=68746.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 03:43:56,106 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.0975, 0.7402, 0.9374, 0.7893, 1.2264, 0.9589, 0.8920, 0.9637], device='cuda:1'), covar=tensor([0.1604, 0.1411, 0.2069, 0.1415, 0.0937, 0.1362, 0.1769, 0.2226], device='cuda:1'), in_proj_covar=tensor([0.0307, 0.0318, 0.0352, 0.0294, 0.0332, 0.0316, 0.0307, 0.0361], device='cuda:1'), out_proj_covar=tensor([6.4078e-05, 6.6941e-05, 7.5476e-05, 6.0355e-05, 6.9367e-05, 6.7107e-05, 6.5417e-05, 7.7128e-05], device='cuda:1') 2023-04-27 03:43:57,284 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=68755.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 03:44:12,770 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0641, 1.3344, 1.7419, 2.4438, 2.5746, 1.8613, 1.6601, 2.0240], device='cuda:1'), covar=tensor([0.0983, 0.1743, 0.1041, 0.0604, 0.0536, 0.1029, 0.1024, 0.0721], device='cuda:1'), in_proj_covar=tensor([0.0192, 0.0201, 0.0182, 0.0173, 0.0178, 0.0184, 0.0155, 0.0180], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 03:44:15,624 INFO [finetune.py:976] (1/7) Epoch 13, batch 50, loss[loss=0.1631, simple_loss=0.2343, pruned_loss=0.04592, over 4818.00 frames. ], tot_loss[loss=0.1919, simple_loss=0.2604, pruned_loss=0.06173, over 215700.84 frames. ], batch size: 39, lr: 3.62e-03, grad_scale: 16.0 2023-04-27 03:44:21,435 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=68791.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 03:44:22,082 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([0.9981, 1.9936, 1.8739, 1.6640, 2.2299, 1.8463, 2.7873, 1.6213], device='cuda:1'), covar=tensor([0.4027, 0.2139, 0.5134, 0.3378, 0.1934, 0.2660, 0.1747, 0.4837], device='cuda:1'), in_proj_covar=tensor([0.0339, 0.0342, 0.0426, 0.0354, 0.0378, 0.0377, 0.0371, 0.0416], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 03:44:26,398 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9662, 1.5904, 2.0661, 2.3959, 2.0767, 1.9109, 1.9732, 1.9644], device='cuda:1'), covar=tensor([0.5667, 0.7449, 0.7816, 0.7448, 0.7307, 0.9159, 0.9930, 0.9155], device='cuda:1'), in_proj_covar=tensor([0.0412, 0.0407, 0.0499, 0.0514, 0.0442, 0.0462, 0.0470, 0.0473], device='cuda:1'), out_proj_covar=tensor([9.9960e-05, 1.0106e-04, 1.1222e-04, 1.2180e-04, 1.0698e-04, 1.1149e-04, 1.1256e-04, 1.1316e-04], device='cuda:1') 2023-04-27 03:44:46,960 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.1532, 1.5871, 2.0137, 2.3975, 1.9866, 1.5363, 1.3038, 1.8474], device='cuda:1'), covar=tensor([0.2862, 0.3218, 0.1488, 0.2175, 0.2539, 0.2652, 0.4222, 0.2073], device='cuda:1'), in_proj_covar=tensor([0.0286, 0.0248, 0.0223, 0.0317, 0.0215, 0.0229, 0.0231, 0.0185], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-27 03:44:48,040 INFO [finetune.py:976] (1/7) Epoch 13, batch 100, loss[loss=0.1737, simple_loss=0.2474, pruned_loss=0.05002, over 4908.00 frames. ], tot_loss[loss=0.1858, simple_loss=0.253, pruned_loss=0.05931, over 379659.00 frames. ], batch size: 37, lr: 3.62e-03, grad_scale: 16.0 2023-04-27 03:44:55,551 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.308e+02 1.653e+02 1.937e+02 2.262e+02 3.719e+02, threshold=3.874e+02, percent-clipped=0.0 2023-04-27 03:44:57,631 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-04-27 03:45:21,008 INFO [finetune.py:976] (1/7) Epoch 13, batch 150, loss[loss=0.1374, simple_loss=0.2035, pruned_loss=0.03567, over 4773.00 frames. ], tot_loss[loss=0.1812, simple_loss=0.2475, pruned_loss=0.05744, over 509613.47 frames. ], batch size: 26, lr: 3.62e-03, grad_scale: 16.0 2023-04-27 03:45:47,026 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5246, 1.3702, 0.4989, 1.2399, 1.3370, 1.3957, 1.2821, 1.3185], device='cuda:1'), covar=tensor([0.0528, 0.0399, 0.0432, 0.0564, 0.0315, 0.0514, 0.0524, 0.0593], device='cuda:1'), in_proj_covar=tensor([0.0029, 0.0025, 0.0023, 0.0029, 0.0020, 0.0029, 0.0028, 0.0029], device='cuda:1'), out_proj_covar=tensor([0.0050, 0.0044, 0.0037, 0.0050, 0.0038, 0.0048, 0.0048, 0.0050], device='cuda:1') 2023-04-27 03:45:53,957 INFO [finetune.py:976] (1/7) Epoch 13, batch 200, loss[loss=0.1649, simple_loss=0.2343, pruned_loss=0.04775, over 4903.00 frames. ], tot_loss[loss=0.1812, simple_loss=0.247, pruned_loss=0.05772, over 609171.26 frames. ], batch size: 37, lr: 3.62e-03, grad_scale: 16.0 2023-04-27 03:45:55,114 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=68934.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 03:46:01,420 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.726e+01 1.554e+02 1.965e+02 2.387e+02 1.026e+03, threshold=3.930e+02, percent-clipped=4.0 2023-04-27 03:46:31,804 INFO [finetune.py:976] (1/7) Epoch 13, batch 250, loss[loss=0.1767, simple_loss=0.254, pruned_loss=0.04974, over 4799.00 frames. ], tot_loss[loss=0.1848, simple_loss=0.2511, pruned_loss=0.0593, over 687248.66 frames. ], batch size: 41, lr: 3.62e-03, grad_scale: 16.0 2023-04-27 03:47:07,597 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8612, 2.5100, 0.9290, 1.2225, 1.6766, 1.1646, 3.4035, 1.4644], device='cuda:1'), covar=tensor([0.0962, 0.0847, 0.1004, 0.1782, 0.0762, 0.1420, 0.0456, 0.1007], device='cuda:1'), in_proj_covar=tensor([0.0051, 0.0066, 0.0048, 0.0047, 0.0050, 0.0052, 0.0076, 0.0051], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:1') 2023-04-27 03:47:36,760 INFO [finetune.py:976] (1/7) Epoch 13, batch 300, loss[loss=0.1863, simple_loss=0.2576, pruned_loss=0.05749, over 4698.00 frames. ], tot_loss[loss=0.1874, simple_loss=0.2548, pruned_loss=0.06002, over 747212.43 frames. ], batch size: 59, lr: 3.62e-03, grad_scale: 16.0 2023-04-27 03:47:47,287 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=69041.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 03:47:48,431 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.203e+02 1.674e+02 1.942e+02 2.375e+02 4.255e+02, threshold=3.885e+02, percent-clipped=1.0 2023-04-27 03:47:48,553 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=69043.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 03:47:59,620 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9122, 1.7782, 2.2397, 2.3771, 1.7176, 1.5759, 1.8417, 1.0195], device='cuda:1'), covar=tensor([0.0739, 0.0923, 0.0523, 0.0777, 0.0964, 0.1223, 0.0902, 0.0945], device='cuda:1'), in_proj_covar=tensor([0.0068, 0.0071, 0.0071, 0.0067, 0.0075, 0.0096, 0.0075, 0.0070], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-27 03:48:05,314 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=69055.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 03:48:39,070 INFO [finetune.py:976] (1/7) Epoch 13, batch 350, loss[loss=0.191, simple_loss=0.2647, pruned_loss=0.05863, over 4902.00 frames. ], tot_loss[loss=0.1883, simple_loss=0.2559, pruned_loss=0.06034, over 792282.11 frames. ], batch size: 36, lr: 3.62e-03, grad_scale: 16.0 2023-04-27 03:48:59,536 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=69103.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 03:49:00,178 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=69104.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 03:49:17,726 INFO [finetune.py:976] (1/7) Epoch 13, batch 400, loss[loss=0.1731, simple_loss=0.2579, pruned_loss=0.0442, over 4903.00 frames. ], tot_loss[loss=0.1878, simple_loss=0.2561, pruned_loss=0.05969, over 829345.97 frames. ], batch size: 37, lr: 3.61e-03, grad_scale: 16.0 2023-04-27 03:49:24,720 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.134e+02 1.707e+02 2.109e+02 2.400e+02 4.777e+02, threshold=4.219e+02, percent-clipped=1.0 2023-04-27 03:49:38,277 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.13 vs. limit=2.0 2023-04-27 03:49:38,768 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=69162.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 03:49:42,422 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=69168.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 03:49:51,421 INFO [finetune.py:976] (1/7) Epoch 13, batch 450, loss[loss=0.1851, simple_loss=0.2506, pruned_loss=0.05981, over 4818.00 frames. ], tot_loss[loss=0.1863, simple_loss=0.2549, pruned_loss=0.0589, over 857484.53 frames. ], batch size: 40, lr: 3.61e-03, grad_scale: 16.0 2023-04-27 03:49:57,551 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8554, 2.4268, 1.7774, 1.5695, 1.3372, 1.3563, 1.7825, 1.2450], device='cuda:1'), covar=tensor([0.1768, 0.1288, 0.1553, 0.1896, 0.2441, 0.2043, 0.1155, 0.2196], device='cuda:1'), in_proj_covar=tensor([0.0197, 0.0214, 0.0170, 0.0204, 0.0204, 0.0185, 0.0157, 0.0185], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-27 03:49:58,221 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-04-27 03:50:19,111 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=69223.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 03:50:20,341 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=69225.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 03:50:22,769 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=69229.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 03:50:25,119 INFO [finetune.py:976] (1/7) Epoch 13, batch 500, loss[loss=0.2026, simple_loss=0.2518, pruned_loss=0.07674, over 4832.00 frames. ], tot_loss[loss=0.1835, simple_loss=0.2518, pruned_loss=0.05761, over 881585.10 frames. ], batch size: 30, lr: 3.61e-03, grad_scale: 16.0 2023-04-27 03:50:25,820 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=69234.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 03:50:31,193 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.140e+02 1.700e+02 1.950e+02 2.352e+02 3.819e+02, threshold=3.900e+02, percent-clipped=0.0 2023-04-27 03:50:32,587 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.47 vs. limit=2.0 2023-04-27 03:50:32,983 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=69245.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 03:50:37,665 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.1552, 1.3183, 1.5681, 1.7308, 1.6761, 1.8276, 1.6273, 1.6615], device='cuda:1'), covar=tensor([0.4020, 0.5740, 0.4768, 0.4294, 0.5565, 0.7603, 0.5619, 0.5040], device='cuda:1'), in_proj_covar=tensor([0.0327, 0.0377, 0.0316, 0.0328, 0.0341, 0.0398, 0.0358, 0.0325], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 03:50:52,424 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=69273.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 03:50:57,802 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=69282.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 03:50:58,358 INFO [finetune.py:976] (1/7) Epoch 13, batch 550, loss[loss=0.1636, simple_loss=0.2318, pruned_loss=0.04775, over 4812.00 frames. ], tot_loss[loss=0.1809, simple_loss=0.2484, pruned_loss=0.05664, over 899135.58 frames. ], batch size: 25, lr: 3.61e-03, grad_scale: 16.0 2023-04-27 03:51:00,308 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=69286.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 03:51:14,438 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=69306.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 03:51:24,254 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=69320.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 03:51:30,303 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4391, 1.6491, 1.6057, 2.1237, 1.8525, 1.9333, 1.5713, 4.4757], device='cuda:1'), covar=tensor([0.0663, 0.1007, 0.0999, 0.1284, 0.0778, 0.0647, 0.0980, 0.0144], device='cuda:1'), in_proj_covar=tensor([0.0038, 0.0039, 0.0040, 0.0044, 0.0040, 0.0038, 0.0039, 0.0058], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:1') 2023-04-27 03:51:32,073 INFO [finetune.py:976] (1/7) Epoch 13, batch 600, loss[loss=0.2215, simple_loss=0.2826, pruned_loss=0.08023, over 4823.00 frames. ], tot_loss[loss=0.1822, simple_loss=0.2498, pruned_loss=0.05734, over 909153.82 frames. ], batch size: 33, lr: 3.61e-03, grad_scale: 16.0 2023-04-27 03:51:32,819 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=69334.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 03:51:37,050 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=69341.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 03:51:38,139 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.090e+02 1.579e+02 1.985e+02 2.306e+02 4.955e+02, threshold=3.970e+02, percent-clipped=1.0 2023-04-27 03:51:44,210 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.3903, 1.2389, 1.6630, 1.6121, 1.2626, 1.0617, 1.3635, 0.9649], device='cuda:1'), covar=tensor([0.0598, 0.0718, 0.0446, 0.0712, 0.0811, 0.1301, 0.0635, 0.0660], device='cuda:1'), in_proj_covar=tensor([0.0067, 0.0070, 0.0070, 0.0066, 0.0075, 0.0096, 0.0075, 0.0069], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-27 03:51:56,872 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=69368.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 03:52:04,787 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=69381.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 03:52:05,905 INFO [finetune.py:976] (1/7) Epoch 13, batch 650, loss[loss=0.18, simple_loss=0.2625, pruned_loss=0.04878, over 4808.00 frames. ], tot_loss[loss=0.1864, simple_loss=0.254, pruned_loss=0.05942, over 917481.98 frames. ], batch size: 39, lr: 3.61e-03, grad_scale: 16.0 2023-04-27 03:52:09,603 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=69389.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 03:52:12,659 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=69394.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 03:52:15,688 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=69399.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 03:52:36,498 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=69429.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 03:52:44,419 INFO [finetune.py:976] (1/7) Epoch 13, batch 700, loss[loss=0.2281, simple_loss=0.2887, pruned_loss=0.08368, over 4811.00 frames. ], tot_loss[loss=0.1874, simple_loss=0.2554, pruned_loss=0.05973, over 928337.58 frames. ], batch size: 39, lr: 3.61e-03, grad_scale: 16.0 2023-04-27 03:52:56,107 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.279e+02 1.620e+02 1.977e+02 2.375e+02 4.653e+02, threshold=3.954e+02, percent-clipped=1.0 2023-04-27 03:53:09,381 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=69455.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 03:53:51,162 INFO [finetune.py:976] (1/7) Epoch 13, batch 750, loss[loss=0.2342, simple_loss=0.3037, pruned_loss=0.08232, over 4890.00 frames. ], tot_loss[loss=0.1892, simple_loss=0.2574, pruned_loss=0.06048, over 933979.41 frames. ], batch size: 43, lr: 3.61e-03, grad_scale: 16.0 2023-04-27 03:53:59,653 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.51 vs. limit=2.0 2023-04-27 03:54:36,353 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=69518.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 03:54:46,350 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=69524.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 03:54:57,622 INFO [finetune.py:976] (1/7) Epoch 13, batch 800, loss[loss=0.1558, simple_loss=0.2265, pruned_loss=0.04251, over 4866.00 frames. ], tot_loss[loss=0.1887, simple_loss=0.257, pruned_loss=0.06018, over 937548.59 frames. ], batch size: 34, lr: 3.61e-03, grad_scale: 16.0 2023-04-27 03:55:09,401 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.803e+01 1.651e+02 1.914e+02 2.371e+02 3.866e+02, threshold=3.828e+02, percent-clipped=0.0 2023-04-27 03:55:15,798 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 2023-04-27 03:55:18,169 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.50 vs. limit=2.0 2023-04-27 03:55:35,207 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=69581.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 03:55:36,382 INFO [finetune.py:976] (1/7) Epoch 13, batch 850, loss[loss=0.1949, simple_loss=0.261, pruned_loss=0.06439, over 4863.00 frames. ], tot_loss[loss=0.1854, simple_loss=0.2537, pruned_loss=0.05849, over 940664.57 frames. ], batch size: 34, lr: 3.61e-03, grad_scale: 16.0 2023-04-27 03:55:47,512 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=69601.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 03:56:07,721 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=69629.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 03:56:10,108 INFO [finetune.py:976] (1/7) Epoch 13, batch 900, loss[loss=0.1985, simple_loss=0.2564, pruned_loss=0.07029, over 4234.00 frames. ], tot_loss[loss=0.1828, simple_loss=0.2506, pruned_loss=0.05752, over 945432.67 frames. ], batch size: 65, lr: 3.61e-03, grad_scale: 32.0 2023-04-27 03:56:16,242 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.053e+02 1.649e+02 1.999e+02 2.386e+02 5.563e+02, threshold=3.997e+02, percent-clipped=1.0 2023-04-27 03:56:37,651 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=69674.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 03:56:39,834 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=69676.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 03:56:44,052 INFO [finetune.py:976] (1/7) Epoch 13, batch 950, loss[loss=0.1824, simple_loss=0.2582, pruned_loss=0.05334, over 4822.00 frames. ], tot_loss[loss=0.183, simple_loss=0.2503, pruned_loss=0.05781, over 947228.43 frames. ], batch size: 39, lr: 3.61e-03, grad_scale: 32.0 2023-04-27 03:56:49,088 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7361, 1.8712, 0.9535, 1.4105, 1.8498, 1.5875, 1.4694, 1.5768], device='cuda:1'), covar=tensor([0.0485, 0.0360, 0.0352, 0.0572, 0.0270, 0.0556, 0.0544, 0.0581], device='cuda:1'), in_proj_covar=tensor([0.0029, 0.0025, 0.0023, 0.0029, 0.0020, 0.0029, 0.0028, 0.0029], device='cuda:1'), out_proj_covar=tensor([0.0049, 0.0044, 0.0037, 0.0050, 0.0037, 0.0048, 0.0048, 0.0050], device='cuda:1') 2023-04-27 03:56:53,910 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=69699.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 03:57:11,093 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=69724.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 03:57:12,882 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0443, 1.5486, 1.5483, 1.7922, 2.2436, 1.8261, 1.5438, 1.5404], device='cuda:1'), covar=tensor([0.1628, 0.1396, 0.1737, 0.1298, 0.0611, 0.1419, 0.1928, 0.1963], device='cuda:1'), in_proj_covar=tensor([0.0305, 0.0316, 0.0348, 0.0291, 0.0330, 0.0313, 0.0304, 0.0358], device='cuda:1'), out_proj_covar=tensor([6.3594e-05, 6.6397e-05, 7.4635e-05, 5.9490e-05, 6.8842e-05, 6.6584e-05, 6.4590e-05, 7.6506e-05], device='cuda:1') 2023-04-27 03:57:18,035 INFO [finetune.py:976] (1/7) Epoch 13, batch 1000, loss[loss=0.1655, simple_loss=0.2258, pruned_loss=0.05257, over 4718.00 frames. ], tot_loss[loss=0.1858, simple_loss=0.2528, pruned_loss=0.0594, over 948244.95 frames. ], batch size: 23, lr: 3.61e-03, grad_scale: 32.0 2023-04-27 03:57:19,393 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=69735.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 03:57:24,156 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.258e+02 1.699e+02 2.018e+02 2.497e+02 4.277e+02, threshold=4.036e+02, percent-clipped=1.0 2023-04-27 03:57:26,749 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=69747.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 03:57:28,609 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=69750.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 03:57:50,562 INFO [finetune.py:976] (1/7) Epoch 13, batch 1050, loss[loss=0.1838, simple_loss=0.2533, pruned_loss=0.05711, over 4775.00 frames. ], tot_loss[loss=0.1881, simple_loss=0.2566, pruned_loss=0.05982, over 950324.36 frames. ], batch size: 26, lr: 3.61e-03, grad_scale: 32.0 2023-04-27 03:58:10,055 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.1699, 2.6547, 1.1310, 1.5509, 2.1791, 1.3131, 3.5405, 1.7003], device='cuda:1'), covar=tensor([0.0629, 0.0636, 0.0779, 0.1234, 0.0465, 0.0977, 0.0228, 0.0639], device='cuda:1'), in_proj_covar=tensor([0.0051, 0.0067, 0.0049, 0.0047, 0.0051, 0.0052, 0.0076, 0.0051], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:1') 2023-04-27 03:58:13,066 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=69818.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 03:58:16,705 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=69824.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 03:58:23,481 INFO [finetune.py:976] (1/7) Epoch 13, batch 1100, loss[loss=0.1532, simple_loss=0.2203, pruned_loss=0.04312, over 4721.00 frames. ], tot_loss[loss=0.1889, simple_loss=0.2574, pruned_loss=0.06018, over 951586.01 frames. ], batch size: 23, lr: 3.61e-03, grad_scale: 32.0 2023-04-27 03:58:25,255 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([3.9001, 3.8490, 2.7684, 4.5174, 3.9765, 3.9053, 1.7006, 3.9161], device='cuda:1'), covar=tensor([0.1712, 0.1163, 0.3125, 0.1290, 0.2591, 0.1824, 0.5352, 0.2154], device='cuda:1'), in_proj_covar=tensor([0.0241, 0.0212, 0.0247, 0.0300, 0.0295, 0.0244, 0.0267, 0.0264], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 03:58:30,702 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.061e+02 1.653e+02 1.942e+02 2.253e+02 5.634e+02, threshold=3.884e+02, percent-clipped=2.0 2023-04-27 03:58:31,418 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.3350, 1.4227, 3.8866, 3.6658, 3.4660, 3.7386, 3.7121, 3.3712], device='cuda:1'), covar=tensor([0.7027, 0.5604, 0.1193, 0.1704, 0.1141, 0.1989, 0.1416, 0.1514], device='cuda:1'), in_proj_covar=tensor([0.0306, 0.0301, 0.0398, 0.0405, 0.0344, 0.0404, 0.0309, 0.0363], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-27 03:58:50,903 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=69866.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 03:59:00,298 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=69872.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 03:59:12,136 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=69881.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 03:59:13,756 INFO [finetune.py:976] (1/7) Epoch 13, batch 1150, loss[loss=0.2137, simple_loss=0.2772, pruned_loss=0.07511, over 4890.00 frames. ], tot_loss[loss=0.1896, simple_loss=0.2581, pruned_loss=0.0605, over 951530.06 frames. ], batch size: 32, lr: 3.61e-03, grad_scale: 32.0 2023-04-27 03:59:30,847 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.2641, 1.3477, 1.6123, 1.7511, 1.5576, 1.7752, 1.6924, 1.6632], device='cuda:1'), covar=tensor([0.4319, 0.6926, 0.5979, 0.5478, 0.7044, 0.8871, 0.6361, 0.6120], device='cuda:1'), in_proj_covar=tensor([0.0327, 0.0377, 0.0315, 0.0328, 0.0341, 0.0399, 0.0359, 0.0324], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 03:59:41,632 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=69901.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 03:59:42,902 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.2225, 1.6678, 2.1016, 2.6223, 2.0927, 1.5818, 1.3598, 2.0158], device='cuda:1'), covar=tensor([0.3507, 0.3862, 0.1759, 0.2445, 0.2994, 0.2969, 0.4660, 0.2416], device='cuda:1'), in_proj_covar=tensor([0.0286, 0.0248, 0.0223, 0.0316, 0.0213, 0.0228, 0.0230, 0.0184], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-27 04:00:07,491 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0123, 2.5407, 2.0998, 2.3344, 1.7662, 2.0355, 2.2269, 1.7038], device='cuda:1'), covar=tensor([0.2065, 0.1216, 0.0882, 0.1329, 0.3399, 0.1202, 0.1921, 0.2695], device='cuda:1'), in_proj_covar=tensor([0.0295, 0.0315, 0.0228, 0.0288, 0.0318, 0.0267, 0.0258, 0.0276], device='cuda:1'), out_proj_covar=tensor([1.1892e-04, 1.2607e-04, 9.0848e-05, 1.1497e-04, 1.2979e-04, 1.0678e-04, 1.0463e-04, 1.1021e-04], device='cuda:1') 2023-04-27 04:00:15,714 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=69929.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 04:00:15,753 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=69929.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 04:00:18,639 INFO [finetune.py:976] (1/7) Epoch 13, batch 1200, loss[loss=0.1402, simple_loss=0.2153, pruned_loss=0.03256, over 4839.00 frames. ], tot_loss[loss=0.1887, simple_loss=0.2574, pruned_loss=0.06004, over 949855.84 frames. ], batch size: 49, lr: 3.61e-03, grad_scale: 32.0 2023-04-27 04:00:32,298 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.030e+02 1.675e+02 1.952e+02 2.332e+02 4.660e+02, threshold=3.905e+02, percent-clipped=1.0 2023-04-27 04:00:36,032 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=69949.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 04:00:52,473 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=69976.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 04:00:53,025 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=69977.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 04:00:57,153 INFO [finetune.py:976] (1/7) Epoch 13, batch 1250, loss[loss=0.1647, simple_loss=0.2365, pruned_loss=0.04639, over 4893.00 frames. ], tot_loss[loss=0.1858, simple_loss=0.2539, pruned_loss=0.05882, over 951317.38 frames. ], batch size: 35, lr: 3.61e-03, grad_scale: 32.0 2023-04-27 04:01:26,388 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=70024.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 04:01:26,438 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=70024.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 04:01:30,032 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=70030.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 04:01:31,777 INFO [finetune.py:976] (1/7) Epoch 13, batch 1300, loss[loss=0.1484, simple_loss=0.2142, pruned_loss=0.0413, over 4820.00 frames. ], tot_loss[loss=0.1828, simple_loss=0.2506, pruned_loss=0.05753, over 951307.23 frames. ], batch size: 25, lr: 3.61e-03, grad_scale: 32.0 2023-04-27 04:01:39,421 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.185e+02 1.660e+02 1.869e+02 2.265e+02 4.379e+02, threshold=3.739e+02, percent-clipped=1.0 2023-04-27 04:01:45,211 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=70050.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 04:01:58,566 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=70072.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 04:01:59,214 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.2175, 1.5566, 1.4585, 1.8338, 1.6087, 1.6747, 1.4101, 3.1121], device='cuda:1'), covar=tensor([0.0675, 0.0815, 0.0839, 0.1188, 0.0641, 0.0500, 0.0741, 0.0164], device='cuda:1'), in_proj_covar=tensor([0.0038, 0.0039, 0.0040, 0.0044, 0.0040, 0.0038, 0.0039, 0.0058], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:1') 2023-04-27 04:02:05,230 INFO [finetune.py:976] (1/7) Epoch 13, batch 1350, loss[loss=0.2118, simple_loss=0.2755, pruned_loss=0.0741, over 4899.00 frames. ], tot_loss[loss=0.1833, simple_loss=0.2506, pruned_loss=0.05795, over 954507.42 frames. ], batch size: 35, lr: 3.61e-03, grad_scale: 32.0 2023-04-27 04:02:16,792 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=70098.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 04:02:24,496 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=70110.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 04:02:38,384 INFO [finetune.py:976] (1/7) Epoch 13, batch 1400, loss[loss=0.2003, simple_loss=0.2668, pruned_loss=0.06687, over 4764.00 frames. ], tot_loss[loss=0.1858, simple_loss=0.2535, pruned_loss=0.05904, over 955224.66 frames. ], batch size: 27, lr: 3.61e-03, grad_scale: 32.0 2023-04-27 04:02:45,033 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.060e+02 1.629e+02 2.088e+02 2.462e+02 4.428e+02, threshold=4.176e+02, percent-clipped=4.0 2023-04-27 04:02:55,486 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9888, 1.7133, 2.1116, 2.4749, 2.4437, 1.8687, 1.6875, 2.1201], device='cuda:1'), covar=tensor([0.0861, 0.1129, 0.0636, 0.0518, 0.0563, 0.0909, 0.0813, 0.0570], device='cuda:1'), in_proj_covar=tensor([0.0192, 0.0201, 0.0182, 0.0173, 0.0178, 0.0185, 0.0156, 0.0180], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 04:03:04,722 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=70171.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 04:03:11,826 INFO [finetune.py:976] (1/7) Epoch 13, batch 1450, loss[loss=0.1737, simple_loss=0.2424, pruned_loss=0.05246, over 4741.00 frames. ], tot_loss[loss=0.188, simple_loss=0.2566, pruned_loss=0.05964, over 955990.32 frames. ], batch size: 27, lr: 3.61e-03, grad_scale: 32.0 2023-04-27 04:03:18,617 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.1929, 1.4552, 5.2134, 4.9110, 4.5839, 4.9267, 4.5027, 4.6548], device='cuda:1'), covar=tensor([0.6299, 0.6040, 0.0825, 0.1525, 0.0827, 0.1403, 0.1147, 0.1296], device='cuda:1'), in_proj_covar=tensor([0.0308, 0.0303, 0.0399, 0.0407, 0.0345, 0.0404, 0.0310, 0.0364], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-27 04:03:45,319 INFO [finetune.py:976] (1/7) Epoch 13, batch 1500, loss[loss=0.2039, simple_loss=0.2725, pruned_loss=0.06767, over 4890.00 frames. ], tot_loss[loss=0.1899, simple_loss=0.2585, pruned_loss=0.06063, over 955035.78 frames. ], batch size: 35, lr: 3.61e-03, grad_scale: 32.0 2023-04-27 04:03:51,424 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.134e+02 1.692e+02 1.982e+02 2.371e+02 3.829e+02, threshold=3.965e+02, percent-clipped=0.0 2023-04-27 04:04:10,471 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=70260.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 04:04:19,827 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.2211, 1.4621, 1.3271, 1.6335, 1.4998, 1.6421, 1.3256, 3.0589], device='cuda:1'), covar=tensor([0.0632, 0.0823, 0.0863, 0.1270, 0.0677, 0.0538, 0.0781, 0.0145], device='cuda:1'), in_proj_covar=tensor([0.0039, 0.0039, 0.0040, 0.0044, 0.0040, 0.0039, 0.0039, 0.0058], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:1') 2023-04-27 04:04:40,470 INFO [finetune.py:976] (1/7) Epoch 13, batch 1550, loss[loss=0.1875, simple_loss=0.2515, pruned_loss=0.06174, over 4905.00 frames. ], tot_loss[loss=0.1895, simple_loss=0.2582, pruned_loss=0.06037, over 956023.43 frames. ], batch size: 36, lr: 3.61e-03, grad_scale: 32.0 2023-04-27 04:04:51,352 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.5516, 1.7231, 1.8527, 2.0347, 1.8265, 1.9321, 1.9873, 1.8983], device='cuda:1'), covar=tensor([0.3907, 0.6532, 0.5072, 0.4768, 0.5809, 0.8262, 0.5999, 0.5398], device='cuda:1'), in_proj_covar=tensor([0.0329, 0.0378, 0.0317, 0.0330, 0.0342, 0.0400, 0.0359, 0.0325], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 04:05:09,695 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=3.71 vs. limit=5.0 2023-04-27 04:05:17,641 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=70321.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 04:05:28,463 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=70330.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 04:05:30,176 INFO [finetune.py:976] (1/7) Epoch 13, batch 1600, loss[loss=0.205, simple_loss=0.2622, pruned_loss=0.0739, over 4874.00 frames. ], tot_loss[loss=0.1884, simple_loss=0.2567, pruned_loss=0.06006, over 956709.82 frames. ], batch size: 34, lr: 3.60e-03, grad_scale: 32.0 2023-04-27 04:05:38,180 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.61 vs. limit=2.0 2023-04-27 04:05:41,006 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.445e+01 1.654e+02 2.055e+02 2.369e+02 4.520e+02, threshold=4.109e+02, percent-clipped=1.0 2023-04-27 04:06:16,249 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=70378.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 04:06:19,128 INFO [finetune.py:976] (1/7) Epoch 13, batch 1650, loss[loss=0.164, simple_loss=0.2423, pruned_loss=0.04281, over 4758.00 frames. ], tot_loss[loss=0.1868, simple_loss=0.2545, pruned_loss=0.05958, over 956009.32 frames. ], batch size: 27, lr: 3.60e-03, grad_scale: 32.0 2023-04-27 04:06:52,935 INFO [finetune.py:976] (1/7) Epoch 13, batch 1700, loss[loss=0.1759, simple_loss=0.2376, pruned_loss=0.05703, over 4701.00 frames. ], tot_loss[loss=0.1836, simple_loss=0.2508, pruned_loss=0.05818, over 956282.92 frames. ], batch size: 59, lr: 3.60e-03, grad_scale: 32.0 2023-04-27 04:06:59,065 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.006e+02 1.620e+02 1.879e+02 2.417e+02 6.028e+02, threshold=3.758e+02, percent-clipped=1.0 2023-04-27 04:07:16,026 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=70466.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 04:07:26,839 INFO [finetune.py:976] (1/7) Epoch 13, batch 1750, loss[loss=0.2053, simple_loss=0.2655, pruned_loss=0.07255, over 4912.00 frames. ], tot_loss[loss=0.1831, simple_loss=0.2509, pruned_loss=0.05763, over 953718.47 frames. ], batch size: 36, lr: 3.60e-03, grad_scale: 32.0 2023-04-27 04:07:26,993 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0261, 1.7666, 2.1680, 2.5434, 2.1355, 1.9442, 2.0644, 2.0510], device='cuda:1'), covar=tensor([0.5265, 0.7802, 0.8284, 0.6000, 0.6448, 0.9686, 0.9478, 0.9713], device='cuda:1'), in_proj_covar=tensor([0.0411, 0.0406, 0.0496, 0.0515, 0.0442, 0.0463, 0.0468, 0.0473], device='cuda:1'), out_proj_covar=tensor([9.9741e-05, 1.0079e-04, 1.1155e-04, 1.2196e-04, 1.0690e-04, 1.1157e-04, 1.1187e-04, 1.1301e-04], device='cuda:1') 2023-04-27 04:07:59,017 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-04-27 04:08:00,060 INFO [finetune.py:976] (1/7) Epoch 13, batch 1800, loss[loss=0.1892, simple_loss=0.2582, pruned_loss=0.06012, over 4748.00 frames. ], tot_loss[loss=0.1856, simple_loss=0.2539, pruned_loss=0.05863, over 953038.09 frames. ], batch size: 54, lr: 3.60e-03, grad_scale: 32.0 2023-04-27 04:08:03,714 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.8780, 2.8818, 2.2374, 3.2951, 2.9415, 2.8762, 1.1285, 2.8028], device='cuda:1'), covar=tensor([0.1970, 0.1744, 0.3254, 0.2761, 0.3364, 0.2252, 0.5798, 0.2905], device='cuda:1'), in_proj_covar=tensor([0.0241, 0.0213, 0.0247, 0.0301, 0.0295, 0.0246, 0.0268, 0.0267], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 04:08:06,008 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.076e+02 1.577e+02 1.883e+02 2.332e+02 3.915e+02, threshold=3.766e+02, percent-clipped=2.0 2023-04-27 04:08:15,359 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6317, 2.0665, 1.7151, 1.9084, 1.5162, 1.7409, 1.7867, 1.3901], device='cuda:1'), covar=tensor([0.1678, 0.1147, 0.0875, 0.1056, 0.3209, 0.1003, 0.1659, 0.2257], device='cuda:1'), in_proj_covar=tensor([0.0290, 0.0308, 0.0223, 0.0282, 0.0310, 0.0261, 0.0254, 0.0270], device='cuda:1'), out_proj_covar=tensor([1.1683e-04, 1.2319e-04, 8.9054e-05, 1.1254e-04, 1.2668e-04, 1.0456e-04, 1.0284e-04, 1.0803e-04], device='cuda:1') 2023-04-27 04:08:33,360 INFO [finetune.py:976] (1/7) Epoch 13, batch 1850, loss[loss=0.1769, simple_loss=0.2457, pruned_loss=0.05402, over 4818.00 frames. ], tot_loss[loss=0.1866, simple_loss=0.2551, pruned_loss=0.05905, over 953185.54 frames. ], batch size: 40, lr: 3.60e-03, grad_scale: 32.0 2023-04-27 04:08:54,137 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=70616.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 04:09:06,409 INFO [finetune.py:976] (1/7) Epoch 13, batch 1900, loss[loss=0.1354, simple_loss=0.2049, pruned_loss=0.03294, over 4749.00 frames. ], tot_loss[loss=0.1859, simple_loss=0.2552, pruned_loss=0.05835, over 955343.90 frames. ], batch size: 23, lr: 3.60e-03, grad_scale: 32.0 2023-04-27 04:09:12,477 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.157e+02 1.737e+02 2.053e+02 2.436e+02 7.397e+02, threshold=4.106e+02, percent-clipped=5.0 2023-04-27 04:10:02,034 INFO [finetune.py:976] (1/7) Epoch 13, batch 1950, loss[loss=0.1497, simple_loss=0.2276, pruned_loss=0.03589, over 4800.00 frames. ], tot_loss[loss=0.1847, simple_loss=0.2542, pruned_loss=0.05757, over 956555.47 frames. ], batch size: 25, lr: 3.60e-03, grad_scale: 32.0 2023-04-27 04:10:08,984 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.66 vs. limit=2.0 2023-04-27 04:10:39,638 INFO [finetune.py:976] (1/7) Epoch 13, batch 2000, loss[loss=0.1837, simple_loss=0.2514, pruned_loss=0.05802, over 4909.00 frames. ], tot_loss[loss=0.1842, simple_loss=0.2525, pruned_loss=0.05796, over 957196.40 frames. ], batch size: 46, lr: 3.60e-03, grad_scale: 32.0 2023-04-27 04:10:52,181 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.325e+01 1.560e+02 1.780e+02 2.155e+02 4.897e+02, threshold=3.560e+02, percent-clipped=2.0 2023-04-27 04:11:15,877 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=70766.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 04:11:28,145 INFO [finetune.py:976] (1/7) Epoch 13, batch 2050, loss[loss=0.1401, simple_loss=0.2112, pruned_loss=0.03453, over 4868.00 frames. ], tot_loss[loss=0.182, simple_loss=0.2497, pruned_loss=0.05721, over 958089.81 frames. ], batch size: 31, lr: 3.60e-03, grad_scale: 32.0 2023-04-27 04:11:36,612 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.2440, 1.6419, 1.5555, 2.1908, 2.3792, 1.8825, 1.8463, 1.6002], device='cuda:1'), covar=tensor([0.2393, 0.1712, 0.2240, 0.1507, 0.1074, 0.2251, 0.2589, 0.2482], device='cuda:1'), in_proj_covar=tensor([0.0305, 0.0315, 0.0348, 0.0290, 0.0328, 0.0313, 0.0302, 0.0358], device='cuda:1'), out_proj_covar=tensor([6.3531e-05, 6.6133e-05, 7.4632e-05, 5.9350e-05, 6.8398e-05, 6.6486e-05, 6.4253e-05, 7.6562e-05], device='cuda:1') 2023-04-27 04:11:47,932 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.67 vs. limit=2.0 2023-04-27 04:11:48,334 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=70814.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 04:12:01,813 INFO [finetune.py:976] (1/7) Epoch 13, batch 2100, loss[loss=0.173, simple_loss=0.2441, pruned_loss=0.05092, over 4246.00 frames. ], tot_loss[loss=0.1823, simple_loss=0.2497, pruned_loss=0.05745, over 956402.78 frames. ], batch size: 65, lr: 3.60e-03, grad_scale: 32.0 2023-04-27 04:12:08,385 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9531, 2.4876, 1.8720, 1.6425, 1.4301, 1.4664, 1.9244, 1.3889], device='cuda:1'), covar=tensor([0.1551, 0.1471, 0.1536, 0.1904, 0.2360, 0.1944, 0.1023, 0.2044], device='cuda:1'), in_proj_covar=tensor([0.0196, 0.0214, 0.0169, 0.0203, 0.0202, 0.0184, 0.0157, 0.0185], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-27 04:12:08,818 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.125e+02 1.590e+02 1.945e+02 2.393e+02 5.833e+02, threshold=3.889e+02, percent-clipped=3.0 2023-04-27 04:12:14,537 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([0.5082, 1.3953, 1.3417, 1.1150, 1.4207, 1.2381, 1.7152, 1.2461], device='cuda:1'), covar=tensor([0.3102, 0.1514, 0.4660, 0.2289, 0.1373, 0.1830, 0.1536, 0.4207], device='cuda:1'), in_proj_covar=tensor([0.0338, 0.0339, 0.0421, 0.0352, 0.0376, 0.0375, 0.0367, 0.0413], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 04:12:35,717 INFO [finetune.py:976] (1/7) Epoch 13, batch 2150, loss[loss=0.2027, simple_loss=0.2769, pruned_loss=0.06427, over 4869.00 frames. ], tot_loss[loss=0.1842, simple_loss=0.2525, pruned_loss=0.05797, over 956020.30 frames. ], batch size: 34, lr: 3.60e-03, grad_scale: 32.0 2023-04-27 04:12:46,620 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-04-27 04:12:54,500 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.68 vs. limit=5.0 2023-04-27 04:12:56,804 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=70916.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 04:12:59,965 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.50 vs. limit=2.0 2023-04-27 04:13:08,518 INFO [finetune.py:976] (1/7) Epoch 13, batch 2200, loss[loss=0.1827, simple_loss=0.2474, pruned_loss=0.05898, over 4744.00 frames. ], tot_loss[loss=0.1853, simple_loss=0.254, pruned_loss=0.05831, over 955305.89 frames. ], batch size: 27, lr: 3.60e-03, grad_scale: 32.0 2023-04-27 04:13:16,545 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.168e+02 1.648e+02 2.003e+02 2.484e+02 4.937e+02, threshold=4.005e+02, percent-clipped=2.0 2023-04-27 04:13:29,540 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=70964.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 04:13:41,740 INFO [finetune.py:976] (1/7) Epoch 13, batch 2250, loss[loss=0.2368, simple_loss=0.3031, pruned_loss=0.08528, over 4762.00 frames. ], tot_loss[loss=0.1863, simple_loss=0.2552, pruned_loss=0.05869, over 953523.81 frames. ], batch size: 54, lr: 3.60e-03, grad_scale: 32.0 2023-04-27 04:13:49,276 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=70992.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 04:14:13,318 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=71030.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 04:14:14,996 INFO [finetune.py:976] (1/7) Epoch 13, batch 2300, loss[loss=0.2181, simple_loss=0.2743, pruned_loss=0.08092, over 4033.00 frames. ], tot_loss[loss=0.1866, simple_loss=0.2558, pruned_loss=0.05868, over 955009.35 frames. ], batch size: 17, lr: 3.60e-03, grad_scale: 32.0 2023-04-27 04:14:23,497 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.148e+02 1.522e+02 1.891e+02 2.315e+02 3.821e+02, threshold=3.782e+02, percent-clipped=0.0 2023-04-27 04:14:30,589 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=71053.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 04:14:59,467 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5192, 1.4630, 3.9795, 3.7148, 3.4925, 3.7540, 3.6244, 3.4851], device='cuda:1'), covar=tensor([0.6919, 0.5102, 0.1036, 0.1686, 0.1172, 0.1588, 0.2571, 0.1519], device='cuda:1'), in_proj_covar=tensor([0.0311, 0.0305, 0.0402, 0.0407, 0.0348, 0.0406, 0.0312, 0.0368], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-27 04:15:10,017 INFO [finetune.py:976] (1/7) Epoch 13, batch 2350, loss[loss=0.1681, simple_loss=0.2376, pruned_loss=0.04929, over 4828.00 frames. ], tot_loss[loss=0.1851, simple_loss=0.2539, pruned_loss=0.05816, over 956427.02 frames. ], batch size: 49, lr: 3.60e-03, grad_scale: 32.0 2023-04-27 04:15:20,864 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=71091.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 04:15:44,392 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-04-27 04:15:54,944 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.11 vs. limit=2.0 2023-04-27 04:15:57,284 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9173, 2.4547, 2.0811, 2.3855, 1.7428, 2.1093, 2.1845, 1.8363], device='cuda:1'), covar=tensor([0.2006, 0.1218, 0.0845, 0.1044, 0.3078, 0.1003, 0.1774, 0.2307], device='cuda:1'), in_proj_covar=tensor([0.0293, 0.0314, 0.0227, 0.0285, 0.0315, 0.0266, 0.0257, 0.0274], device='cuda:1'), out_proj_covar=tensor([1.1832e-04, 1.2528e-04, 9.0870e-05, 1.1405e-04, 1.2843e-04, 1.0660e-04, 1.0430e-04, 1.0970e-04], device='cuda:1') 2023-04-27 04:16:15,377 INFO [finetune.py:976] (1/7) Epoch 13, batch 2400, loss[loss=0.1529, simple_loss=0.2112, pruned_loss=0.04734, over 4872.00 frames. ], tot_loss[loss=0.1825, simple_loss=0.2508, pruned_loss=0.05704, over 954844.78 frames. ], batch size: 34, lr: 3.60e-03, grad_scale: 32.0 2023-04-27 04:16:26,845 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.075e+02 1.510e+02 1.853e+02 2.211e+02 4.308e+02, threshold=3.705e+02, percent-clipped=1.0 2023-04-27 04:16:46,915 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5090, 0.9485, 1.2907, 1.1641, 1.6312, 1.3216, 1.1183, 1.2571], device='cuda:1'), covar=tensor([0.1616, 0.1674, 0.1805, 0.1463, 0.0792, 0.1432, 0.2063, 0.2089], device='cuda:1'), in_proj_covar=tensor([0.0308, 0.0318, 0.0352, 0.0294, 0.0332, 0.0315, 0.0305, 0.0362], device='cuda:1'), out_proj_covar=tensor([6.4233e-05, 6.6815e-05, 7.5667e-05, 6.0336e-05, 6.9437e-05, 6.6937e-05, 6.4749e-05, 7.7435e-05], device='cuda:1') 2023-04-27 04:16:54,091 INFO [finetune.py:976] (1/7) Epoch 13, batch 2450, loss[loss=0.2099, simple_loss=0.2634, pruned_loss=0.07816, over 4831.00 frames. ], tot_loss[loss=0.1795, simple_loss=0.2476, pruned_loss=0.05573, over 956764.46 frames. ], batch size: 39, lr: 3.60e-03, grad_scale: 32.0 2023-04-27 04:17:28,067 INFO [finetune.py:976] (1/7) Epoch 13, batch 2500, loss[loss=0.1935, simple_loss=0.2661, pruned_loss=0.06047, over 4819.00 frames. ], tot_loss[loss=0.1806, simple_loss=0.2485, pruned_loss=0.05633, over 956846.59 frames. ], batch size: 40, lr: 3.60e-03, grad_scale: 32.0 2023-04-27 04:17:34,107 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.448e+01 1.657e+02 1.843e+02 2.125e+02 3.906e+02, threshold=3.687e+02, percent-clipped=2.0 2023-04-27 04:17:34,256 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5876, 1.7996, 1.4329, 1.1482, 1.1726, 1.1848, 1.4167, 1.1506], device='cuda:1'), covar=tensor([0.1845, 0.1208, 0.1638, 0.1755, 0.2398, 0.2079, 0.1103, 0.2104], device='cuda:1'), in_proj_covar=tensor([0.0195, 0.0212, 0.0168, 0.0201, 0.0201, 0.0182, 0.0156, 0.0184], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-27 04:18:01,395 INFO [finetune.py:976] (1/7) Epoch 13, batch 2550, loss[loss=0.1537, simple_loss=0.2369, pruned_loss=0.0352, over 4779.00 frames. ], tot_loss[loss=0.1839, simple_loss=0.2524, pruned_loss=0.05773, over 956280.67 frames. ], batch size: 29, lr: 3.60e-03, grad_scale: 32.0 2023-04-27 04:18:15,112 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-04-27 04:18:34,379 INFO [finetune.py:976] (1/7) Epoch 13, batch 2600, loss[loss=0.2358, simple_loss=0.2904, pruned_loss=0.09064, over 4802.00 frames. ], tot_loss[loss=0.1863, simple_loss=0.2547, pruned_loss=0.05899, over 956134.07 frames. ], batch size: 40, lr: 3.60e-03, grad_scale: 32.0 2023-04-27 04:18:36,413 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9150, 2.6755, 1.9293, 1.8724, 1.3679, 1.3784, 1.9960, 1.4135], device='cuda:1'), covar=tensor([0.1813, 0.1359, 0.1495, 0.1775, 0.2437, 0.2105, 0.1051, 0.2054], device='cuda:1'), in_proj_covar=tensor([0.0196, 0.0213, 0.0168, 0.0202, 0.0201, 0.0183, 0.0157, 0.0185], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-27 04:18:38,205 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5013, 1.4704, 0.6780, 1.1986, 1.5595, 1.3973, 1.2791, 1.3161], device='cuda:1'), covar=tensor([0.0544, 0.0385, 0.0402, 0.0578, 0.0305, 0.0518, 0.0503, 0.0596], device='cuda:1'), in_proj_covar=tensor([0.0029, 0.0025, 0.0023, 0.0029, 0.0020, 0.0029, 0.0028, 0.0029], device='cuda:1'), out_proj_covar=tensor([0.0050, 0.0044, 0.0038, 0.0050, 0.0037, 0.0048, 0.0049, 0.0050], device='cuda:1') 2023-04-27 04:18:38,812 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.1638, 1.3561, 1.2898, 1.6661, 1.5163, 1.6298, 1.3001, 2.6533], device='cuda:1'), covar=tensor([0.0523, 0.0588, 0.0650, 0.0898, 0.0474, 0.0403, 0.0579, 0.0255], device='cuda:1'), in_proj_covar=tensor([0.0038, 0.0039, 0.0040, 0.0043, 0.0040, 0.0038, 0.0038, 0.0057], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0014, 0.0013, 0.0012, 0.0013, 0.0016], device='cuda:1') 2023-04-27 04:18:40,522 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.287e+02 1.728e+02 1.986e+02 2.411e+02 4.998e+02, threshold=3.972e+02, percent-clipped=3.0 2023-04-27 04:18:43,641 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=71348.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 04:18:55,023 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4852, 1.3361, 1.6386, 1.7034, 1.3580, 1.1860, 1.3448, 1.0306], device='cuda:1'), covar=tensor([0.0549, 0.0561, 0.0583, 0.0418, 0.0671, 0.1163, 0.0518, 0.0717], device='cuda:1'), in_proj_covar=tensor([0.0067, 0.0070, 0.0070, 0.0067, 0.0075, 0.0096, 0.0075, 0.0069], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-27 04:18:56,759 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.59 vs. limit=2.0 2023-04-27 04:19:08,105 INFO [finetune.py:976] (1/7) Epoch 13, batch 2650, loss[loss=0.2068, simple_loss=0.2792, pruned_loss=0.06715, over 4922.00 frames. ], tot_loss[loss=0.1866, simple_loss=0.2556, pruned_loss=0.05885, over 954692.09 frames. ], batch size: 42, lr: 3.60e-03, grad_scale: 32.0 2023-04-27 04:19:10,020 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=71386.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 04:19:37,255 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-04-27 04:19:41,957 INFO [finetune.py:976] (1/7) Epoch 13, batch 2700, loss[loss=0.1731, simple_loss=0.2436, pruned_loss=0.05129, over 4899.00 frames. ], tot_loss[loss=0.1858, simple_loss=0.2551, pruned_loss=0.05828, over 956314.97 frames. ], batch size: 37, lr: 3.60e-03, grad_scale: 32.0 2023-04-27 04:19:47,044 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.4605, 1.6340, 1.7997, 1.9462, 1.8579, 1.9784, 1.9007, 1.8573], device='cuda:1'), covar=tensor([0.4313, 0.6028, 0.5384, 0.5349, 0.5949, 0.7985, 0.5658, 0.5646], device='cuda:1'), in_proj_covar=tensor([0.0330, 0.0378, 0.0318, 0.0331, 0.0342, 0.0400, 0.0359, 0.0326], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 04:19:48,075 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.223e+02 1.575e+02 1.881e+02 2.224e+02 4.491e+02, threshold=3.762e+02, percent-clipped=1.0 2023-04-27 04:20:00,448 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.1921, 1.8274, 2.1236, 2.4942, 2.4509, 1.9302, 1.7459, 2.2566], device='cuda:1'), covar=tensor([0.0773, 0.1031, 0.0650, 0.0564, 0.0630, 0.0823, 0.0808, 0.0491], device='cuda:1'), in_proj_covar=tensor([0.0192, 0.0202, 0.0183, 0.0174, 0.0179, 0.0184, 0.0156, 0.0180], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 04:20:30,302 INFO [finetune.py:976] (1/7) Epoch 13, batch 2750, loss[loss=0.1903, simple_loss=0.2503, pruned_loss=0.06516, over 4771.00 frames. ], tot_loss[loss=0.1835, simple_loss=0.2522, pruned_loss=0.05743, over 957065.74 frames. ], batch size: 28, lr: 3.59e-03, grad_scale: 32.0 2023-04-27 04:20:51,284 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-04-27 04:20:53,077 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6954, 2.0755, 1.6178, 1.3668, 1.2293, 1.2676, 1.6433, 1.2043], device='cuda:1'), covar=tensor([0.1837, 0.1314, 0.1618, 0.1841, 0.2443, 0.2021, 0.1131, 0.2182], device='cuda:1'), in_proj_covar=tensor([0.0197, 0.0214, 0.0169, 0.0204, 0.0202, 0.0184, 0.0157, 0.0185], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-27 04:21:19,161 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.70 vs. limit=2.0 2023-04-27 04:21:21,453 INFO [finetune.py:976] (1/7) Epoch 13, batch 2800, loss[loss=0.1837, simple_loss=0.249, pruned_loss=0.05922, over 4873.00 frames. ], tot_loss[loss=0.1808, simple_loss=0.2486, pruned_loss=0.05648, over 956170.67 frames. ], batch size: 34, lr: 3.59e-03, grad_scale: 32.0 2023-04-27 04:21:33,187 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.066e+02 1.551e+02 1.952e+02 2.336e+02 5.892e+02, threshold=3.903e+02, percent-clipped=4.0 2023-04-27 04:21:40,257 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.1176, 2.6630, 2.1428, 2.0596, 1.6420, 1.4791, 2.2085, 1.4964], device='cuda:1'), covar=tensor([0.1613, 0.1445, 0.1302, 0.1734, 0.2330, 0.1877, 0.1012, 0.1945], device='cuda:1'), in_proj_covar=tensor([0.0195, 0.0212, 0.0168, 0.0202, 0.0201, 0.0183, 0.0156, 0.0184], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-27 04:22:06,642 INFO [finetune.py:976] (1/7) Epoch 13, batch 2850, loss[loss=0.2417, simple_loss=0.3099, pruned_loss=0.08674, over 4896.00 frames. ], tot_loss[loss=0.1803, simple_loss=0.2478, pruned_loss=0.05636, over 956544.06 frames. ], batch size: 37, lr: 3.59e-03, grad_scale: 32.0 2023-04-27 04:22:28,017 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=71616.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 04:22:38,466 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.11 vs. limit=2.0 2023-04-27 04:22:40,693 INFO [finetune.py:976] (1/7) Epoch 13, batch 2900, loss[loss=0.124, simple_loss=0.1892, pruned_loss=0.02937, over 3999.00 frames. ], tot_loss[loss=0.1813, simple_loss=0.2492, pruned_loss=0.0567, over 953427.23 frames. ], batch size: 17, lr: 3.59e-03, grad_scale: 64.0 2023-04-27 04:22:46,796 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.322e+02 1.703e+02 2.022e+02 2.297e+02 3.791e+02, threshold=4.044e+02, percent-clipped=0.0 2023-04-27 04:22:49,910 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=71648.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 04:23:01,449 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.65 vs. limit=5.0 2023-04-27 04:23:09,479 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=71677.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 04:23:13,494 INFO [finetune.py:976] (1/7) Epoch 13, batch 2950, loss[loss=0.1906, simple_loss=0.2592, pruned_loss=0.06093, over 4795.00 frames. ], tot_loss[loss=0.1836, simple_loss=0.2525, pruned_loss=0.05737, over 953736.76 frames. ], batch size: 54, lr: 3.59e-03, grad_scale: 64.0 2023-04-27 04:23:15,379 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=71686.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 04:23:21,434 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=71696.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 04:23:45,589 INFO [finetune.py:976] (1/7) Epoch 13, batch 3000, loss[loss=0.1608, simple_loss=0.2413, pruned_loss=0.04021, over 4762.00 frames. ], tot_loss[loss=0.1853, simple_loss=0.2538, pruned_loss=0.05834, over 953839.00 frames. ], batch size: 28, lr: 3.59e-03, grad_scale: 64.0 2023-04-27 04:23:45,589 INFO [finetune.py:1001] (1/7) Computing validation loss 2023-04-27 04:23:49,181 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4403, 1.7376, 1.6322, 1.9405, 1.8020, 1.9028, 1.5345, 3.0502], device='cuda:1'), covar=tensor([0.0518, 0.0612, 0.0637, 0.0969, 0.0505, 0.0386, 0.0618, 0.0209], device='cuda:1'), in_proj_covar=tensor([0.0038, 0.0039, 0.0040, 0.0043, 0.0040, 0.0038, 0.0038, 0.0057], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0014, 0.0013, 0.0012, 0.0013, 0.0016], device='cuda:1') 2023-04-27 04:23:56,065 INFO [finetune.py:1010] (1/7) Epoch 13, validation: loss=0.1517, simple_loss=0.224, pruned_loss=0.03973, over 2265189.00 frames. 2023-04-27 04:23:56,065 INFO [finetune.py:1011] (1/7) Maximum memory allocated so far is 6391MB 2023-04-27 04:23:56,743 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=71734.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 04:23:59,767 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.3851, 1.8301, 2.3294, 2.8671, 2.2875, 1.8141, 1.7591, 2.2308], device='cuda:1'), covar=tensor([0.3580, 0.3537, 0.1628, 0.2758, 0.2955, 0.2759, 0.4165, 0.2407], device='cuda:1'), in_proj_covar=tensor([0.0285, 0.0247, 0.0221, 0.0315, 0.0214, 0.0229, 0.0230, 0.0184], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-27 04:24:03,110 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.215e+02 1.669e+02 1.992e+02 2.305e+02 4.949e+02, threshold=3.985e+02, percent-clipped=1.0 2023-04-27 04:24:11,953 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.68 vs. limit=5.0 2023-04-27 04:24:27,546 INFO [finetune.py:976] (1/7) Epoch 13, batch 3050, loss[loss=0.1916, simple_loss=0.2651, pruned_loss=0.05902, over 4859.00 frames. ], tot_loss[loss=0.1861, simple_loss=0.2551, pruned_loss=0.05851, over 954480.31 frames. ], batch size: 31, lr: 3.59e-03, grad_scale: 64.0 2023-04-27 04:24:48,527 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-04-27 04:24:49,652 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5243, 1.6852, 1.6316, 1.9962, 1.8295, 2.0515, 1.5289, 3.4070], device='cuda:1'), covar=tensor([0.0575, 0.0720, 0.0774, 0.1091, 0.0583, 0.0530, 0.0694, 0.0260], device='cuda:1'), in_proj_covar=tensor([0.0038, 0.0039, 0.0040, 0.0043, 0.0040, 0.0038, 0.0038, 0.0057], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0014, 0.0013, 0.0012, 0.0013, 0.0016], device='cuda:1') 2023-04-27 04:24:57,037 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.65 vs. limit=2.0 2023-04-27 04:25:00,523 INFO [finetune.py:976] (1/7) Epoch 13, batch 3100, loss[loss=0.2126, simple_loss=0.2745, pruned_loss=0.07534, over 4740.00 frames. ], tot_loss[loss=0.1838, simple_loss=0.2526, pruned_loss=0.05748, over 955301.89 frames. ], batch size: 59, lr: 3.59e-03, grad_scale: 64.0 2023-04-27 04:25:08,971 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 8.880e+01 1.511e+02 1.834e+02 2.152e+02 3.267e+02, threshold=3.669e+02, percent-clipped=0.0 2023-04-27 04:25:14,576 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 2023-04-27 04:25:42,156 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5531, 1.3792, 0.6699, 1.2163, 1.4621, 1.4019, 1.3116, 1.3182], device='cuda:1'), covar=tensor([0.0522, 0.0366, 0.0405, 0.0574, 0.0302, 0.0530, 0.0504, 0.0591], device='cuda:1'), in_proj_covar=tensor([0.0029, 0.0025, 0.0023, 0.0029, 0.0020, 0.0029, 0.0028, 0.0029], device='cuda:1'), out_proj_covar=tensor([0.0050, 0.0044, 0.0037, 0.0050, 0.0037, 0.0048, 0.0049, 0.0050], device='cuda:1') 2023-04-27 04:25:46,563 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-04-27 04:25:54,815 INFO [finetune.py:976] (1/7) Epoch 13, batch 3150, loss[loss=0.1428, simple_loss=0.2112, pruned_loss=0.03718, over 4752.00 frames. ], tot_loss[loss=0.182, simple_loss=0.2503, pruned_loss=0.0569, over 955817.70 frames. ], batch size: 23, lr: 3.59e-03, grad_scale: 64.0 2023-04-27 04:26:48,035 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=71919.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 04:27:02,055 INFO [finetune.py:976] (1/7) Epoch 13, batch 3200, loss[loss=0.1252, simple_loss=0.2085, pruned_loss=0.02097, over 4907.00 frames. ], tot_loss[loss=0.1798, simple_loss=0.2475, pruned_loss=0.05605, over 955916.51 frames. ], batch size: 35, lr: 3.59e-03, grad_scale: 64.0 2023-04-27 04:27:08,182 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.251e+02 1.591e+02 1.837e+02 2.276e+02 4.272e+02, threshold=3.675e+02, percent-clipped=2.0 2023-04-27 04:27:24,801 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 2023-04-27 04:27:26,370 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([3.3438, 3.3415, 2.5844, 3.8785, 3.3650, 3.3678, 1.3460, 3.2742], device='cuda:1'), covar=tensor([0.2125, 0.1453, 0.3306, 0.2171, 0.4345, 0.1998, 0.6267, 0.2972], device='cuda:1'), in_proj_covar=tensor([0.0242, 0.0213, 0.0247, 0.0300, 0.0296, 0.0246, 0.0269, 0.0268], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 04:27:28,806 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=71972.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 04:27:33,750 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=71980.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 04:27:35,466 INFO [finetune.py:976] (1/7) Epoch 13, batch 3250, loss[loss=0.1637, simple_loss=0.2344, pruned_loss=0.04648, over 4786.00 frames. ], tot_loss[loss=0.1817, simple_loss=0.2494, pruned_loss=0.05697, over 954733.61 frames. ], batch size: 26, lr: 3.59e-03, grad_scale: 64.0 2023-04-27 04:27:57,706 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-04-27 04:28:01,981 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.32 vs. limit=5.0 2023-04-27 04:28:10,359 INFO [finetune.py:976] (1/7) Epoch 13, batch 3300, loss[loss=0.1659, simple_loss=0.2494, pruned_loss=0.04117, over 4879.00 frames. ], tot_loss[loss=0.1824, simple_loss=0.2509, pruned_loss=0.05689, over 953770.87 frames. ], batch size: 31, lr: 3.59e-03, grad_scale: 32.0 2023-04-27 04:28:16,989 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.091e+02 1.593e+02 1.877e+02 2.325e+02 3.643e+02, threshold=3.754e+02, percent-clipped=0.0 2023-04-27 04:28:43,989 INFO [finetune.py:976] (1/7) Epoch 13, batch 3350, loss[loss=0.1882, simple_loss=0.2596, pruned_loss=0.05838, over 4861.00 frames. ], tot_loss[loss=0.1847, simple_loss=0.2542, pruned_loss=0.0576, over 953582.35 frames. ], batch size: 44, lr: 3.59e-03, grad_scale: 32.0 2023-04-27 04:29:08,106 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8380, 2.3590, 1.9629, 2.1906, 1.6624, 1.9061, 1.8907, 1.5144], device='cuda:1'), covar=tensor([0.1789, 0.1124, 0.0793, 0.1102, 0.3128, 0.1110, 0.1902, 0.2460], device='cuda:1'), in_proj_covar=tensor([0.0293, 0.0311, 0.0225, 0.0284, 0.0312, 0.0265, 0.0256, 0.0272], device='cuda:1'), out_proj_covar=tensor([1.1797e-04, 1.2431e-04, 9.0051e-05, 1.1346e-04, 1.2717e-04, 1.0614e-04, 1.0383e-04, 1.0889e-04], device='cuda:1') 2023-04-27 04:29:15,408 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=72129.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 04:29:17,727 INFO [finetune.py:976] (1/7) Epoch 13, batch 3400, loss[loss=0.1869, simple_loss=0.2635, pruned_loss=0.05518, over 4918.00 frames. ], tot_loss[loss=0.1863, simple_loss=0.2559, pruned_loss=0.05832, over 955598.93 frames. ], batch size: 33, lr: 3.59e-03, grad_scale: 32.0 2023-04-27 04:29:24,412 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.141e+02 1.571e+02 1.878e+02 2.335e+02 4.983e+02, threshold=3.756e+02, percent-clipped=1.0 2023-04-27 04:29:51,374 INFO [finetune.py:976] (1/7) Epoch 13, batch 3450, loss[loss=0.2145, simple_loss=0.2838, pruned_loss=0.07263, over 4776.00 frames. ], tot_loss[loss=0.1852, simple_loss=0.2549, pruned_loss=0.05778, over 957126.38 frames. ], batch size: 29, lr: 3.59e-03, grad_scale: 32.0 2023-04-27 04:29:55,736 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=72190.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 04:30:01,995 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=3.79 vs. limit=5.0 2023-04-27 04:30:02,523 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.3625, 1.2898, 1.6909, 1.5915, 1.2788, 1.1292, 1.3944, 0.9392], device='cuda:1'), covar=tensor([0.0589, 0.0621, 0.0434, 0.0655, 0.0773, 0.1197, 0.0533, 0.0593], device='cuda:1'), in_proj_covar=tensor([0.0068, 0.0070, 0.0070, 0.0068, 0.0075, 0.0096, 0.0075, 0.0069], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-27 04:30:24,563 INFO [finetune.py:976] (1/7) Epoch 13, batch 3500, loss[loss=0.2042, simple_loss=0.2736, pruned_loss=0.0674, over 4911.00 frames. ], tot_loss[loss=0.1837, simple_loss=0.2525, pruned_loss=0.0574, over 955847.60 frames. ], batch size: 36, lr: 3.59e-03, grad_scale: 32.0 2023-04-27 04:30:31,095 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.136e+02 1.651e+02 2.080e+02 2.507e+02 4.410e+02, threshold=4.160e+02, percent-clipped=6.0 2023-04-27 04:31:01,762 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.71 vs. limit=2.0 2023-04-27 04:31:08,030 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=72272.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 04:31:09,848 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=72275.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 04:31:12,375 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.2175, 1.6564, 1.5957, 2.0175, 1.8919, 1.9996, 1.6440, 4.3051], device='cuda:1'), covar=tensor([0.0591, 0.0794, 0.0790, 0.1186, 0.0614, 0.0551, 0.0746, 0.0130], device='cuda:1'), in_proj_covar=tensor([0.0038, 0.0039, 0.0040, 0.0043, 0.0040, 0.0038, 0.0038, 0.0057], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0014, 0.0013, 0.0012, 0.0013, 0.0016], device='cuda:1') 2023-04-27 04:31:20,404 INFO [finetune.py:976] (1/7) Epoch 13, batch 3550, loss[loss=0.1992, simple_loss=0.2518, pruned_loss=0.07332, over 4834.00 frames. ], tot_loss[loss=0.1821, simple_loss=0.2497, pruned_loss=0.0572, over 956941.55 frames. ], batch size: 38, lr: 3.59e-03, grad_scale: 32.0 2023-04-27 04:31:41,851 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8520, 2.2895, 1.8006, 1.6565, 1.3113, 1.3302, 1.7682, 1.2602], device='cuda:1'), covar=tensor([0.1765, 0.1305, 0.1513, 0.1846, 0.2463, 0.2086, 0.1111, 0.2135], device='cuda:1'), in_proj_covar=tensor([0.0194, 0.0212, 0.0167, 0.0202, 0.0201, 0.0183, 0.0156, 0.0184], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-27 04:31:44,901 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7152, 1.9565, 0.8220, 1.4116, 1.9766, 1.6058, 1.4714, 1.5548], device='cuda:1'), covar=tensor([0.0524, 0.0357, 0.0368, 0.0559, 0.0267, 0.0539, 0.0519, 0.0610], device='cuda:1'), in_proj_covar=tensor([0.0029, 0.0025, 0.0023, 0.0029, 0.0020, 0.0029, 0.0028, 0.0029], device='cuda:1'), out_proj_covar=tensor([0.0050, 0.0044, 0.0038, 0.0050, 0.0037, 0.0048, 0.0048, 0.0050], device='cuda:1') 2023-04-27 04:32:06,131 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=72320.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 04:32:16,787 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=72326.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 04:32:26,338 INFO [finetune.py:976] (1/7) Epoch 13, batch 3600, loss[loss=0.1856, simple_loss=0.2555, pruned_loss=0.05787, over 4806.00 frames. ], tot_loss[loss=0.1809, simple_loss=0.2479, pruned_loss=0.05697, over 956200.76 frames. ], batch size: 41, lr: 3.59e-03, grad_scale: 32.0 2023-04-27 04:32:33,138 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.090e+02 1.673e+02 2.029e+02 2.570e+02 4.003e+02, threshold=4.058e+02, percent-clipped=0.0 2023-04-27 04:32:54,673 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-04-27 04:32:59,900 INFO [finetune.py:976] (1/7) Epoch 13, batch 3650, loss[loss=0.1351, simple_loss=0.2068, pruned_loss=0.03173, over 4782.00 frames. ], tot_loss[loss=0.1822, simple_loss=0.2499, pruned_loss=0.05728, over 952990.83 frames. ], batch size: 26, lr: 3.59e-03, grad_scale: 32.0 2023-04-27 04:33:02,536 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=72387.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 04:33:30,823 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6725, 2.0605, 1.5851, 1.3246, 1.2119, 1.2006, 1.6095, 1.1552], device='cuda:1'), covar=tensor([0.1918, 0.1458, 0.1640, 0.2039, 0.2582, 0.2151, 0.1147, 0.2228], device='cuda:1'), in_proj_covar=tensor([0.0195, 0.0214, 0.0169, 0.0204, 0.0202, 0.0184, 0.0157, 0.0186], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-27 04:33:33,724 INFO [finetune.py:976] (1/7) Epoch 13, batch 3700, loss[loss=0.172, simple_loss=0.2443, pruned_loss=0.04985, over 4817.00 frames. ], tot_loss[loss=0.1863, simple_loss=0.2548, pruned_loss=0.05883, over 954488.05 frames. ], batch size: 51, lr: 3.59e-03, grad_scale: 32.0 2023-04-27 04:33:37,743 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2023-04-27 04:33:40,456 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.072e+02 1.574e+02 1.888e+02 2.267e+02 6.702e+02, threshold=3.776e+02, percent-clipped=3.0 2023-04-27 04:33:47,458 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-04-27 04:34:07,019 INFO [finetune.py:976] (1/7) Epoch 13, batch 3750, loss[loss=0.1572, simple_loss=0.2358, pruned_loss=0.03925, over 4894.00 frames. ], tot_loss[loss=0.1865, simple_loss=0.2556, pruned_loss=0.05874, over 952446.87 frames. ], batch size: 32, lr: 3.59e-03, grad_scale: 32.0 2023-04-27 04:34:08,311 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=72485.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 04:34:18,024 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6643, 1.2617, 1.3031, 1.4848, 1.8870, 1.4811, 1.3000, 1.2183], device='cuda:1'), covar=tensor([0.1500, 0.1466, 0.1976, 0.1292, 0.0969, 0.1505, 0.2018, 0.2132], device='cuda:1'), in_proj_covar=tensor([0.0309, 0.0319, 0.0354, 0.0295, 0.0332, 0.0316, 0.0309, 0.0365], device='cuda:1'), out_proj_covar=tensor([6.4319e-05, 6.6948e-05, 7.6090e-05, 6.0389e-05, 6.9210e-05, 6.7088e-05, 6.5687e-05, 7.8110e-05], device='cuda:1') 2023-04-27 04:34:39,242 INFO [finetune.py:976] (1/7) Epoch 13, batch 3800, loss[loss=0.1679, simple_loss=0.2402, pruned_loss=0.04779, over 4758.00 frames. ], tot_loss[loss=0.1867, simple_loss=0.2564, pruned_loss=0.05852, over 954129.03 frames. ], batch size: 27, lr: 3.59e-03, grad_scale: 32.0 2023-04-27 04:34:46,446 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.067e+02 1.671e+02 1.907e+02 2.254e+02 4.045e+02, threshold=3.813e+02, percent-clipped=1.0 2023-04-27 04:35:05,824 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=72575.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 04:35:11,559 INFO [finetune.py:976] (1/7) Epoch 13, batch 3850, loss[loss=0.1779, simple_loss=0.2376, pruned_loss=0.05908, over 4829.00 frames. ], tot_loss[loss=0.1841, simple_loss=0.2536, pruned_loss=0.05725, over 953548.77 frames. ], batch size: 25, lr: 3.59e-03, grad_scale: 32.0 2023-04-27 04:35:37,008 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=72623.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 04:35:44,026 INFO [finetune.py:976] (1/7) Epoch 13, batch 3900, loss[loss=0.1809, simple_loss=0.2397, pruned_loss=0.06101, over 4825.00 frames. ], tot_loss[loss=0.1824, simple_loss=0.2509, pruned_loss=0.05693, over 954661.93 frames. ], batch size: 30, lr: 3.59e-03, grad_scale: 32.0 2023-04-27 04:35:51,958 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.075e+02 1.669e+02 1.861e+02 2.404e+02 3.488e+02, threshold=3.722e+02, percent-clipped=0.0 2023-04-27 04:36:29,542 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([0.7970, 1.8792, 1.8206, 1.5061, 2.0315, 1.6090, 2.6129, 1.6082], device='cuda:1'), covar=tensor([0.3867, 0.1741, 0.4568, 0.3128, 0.1542, 0.2427, 0.1231, 0.4577], device='cuda:1'), in_proj_covar=tensor([0.0343, 0.0345, 0.0427, 0.0355, 0.0380, 0.0382, 0.0373, 0.0418], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 04:36:32,461 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=72682.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 04:36:33,065 INFO [finetune.py:976] (1/7) Epoch 13, batch 3950, loss[loss=0.1352, simple_loss=0.2055, pruned_loss=0.03249, over 4697.00 frames. ], tot_loss[loss=0.1796, simple_loss=0.2476, pruned_loss=0.05576, over 955811.14 frames. ], batch size: 23, lr: 3.58e-03, grad_scale: 32.0 2023-04-27 04:37:38,716 INFO [finetune.py:976] (1/7) Epoch 13, batch 4000, loss[loss=0.2095, simple_loss=0.2743, pruned_loss=0.07236, over 4824.00 frames. ], tot_loss[loss=0.1801, simple_loss=0.2477, pruned_loss=0.0562, over 956214.06 frames. ], batch size: 39, lr: 3.58e-03, grad_scale: 32.0 2023-04-27 04:37:57,179 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 7.525e+01 1.494e+02 1.807e+02 2.123e+02 3.249e+02, threshold=3.613e+02, percent-clipped=0.0 2023-04-27 04:38:27,849 INFO [finetune.py:976] (1/7) Epoch 13, batch 4050, loss[loss=0.1586, simple_loss=0.2149, pruned_loss=0.05115, over 4037.00 frames. ], tot_loss[loss=0.1838, simple_loss=0.2518, pruned_loss=0.05795, over 953556.32 frames. ], batch size: 17, lr: 3.58e-03, grad_scale: 32.0 2023-04-27 04:38:29,164 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=72785.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 04:38:39,551 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9885, 1.7381, 2.0020, 2.4354, 2.4559, 2.0434, 1.5752, 2.0861], device='cuda:1'), covar=tensor([0.0964, 0.1227, 0.0739, 0.0649, 0.0621, 0.0822, 0.0901, 0.0620], device='cuda:1'), in_proj_covar=tensor([0.0190, 0.0202, 0.0181, 0.0172, 0.0177, 0.0182, 0.0155, 0.0179], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 04:39:01,350 INFO [finetune.py:976] (1/7) Epoch 13, batch 4100, loss[loss=0.1694, simple_loss=0.2436, pruned_loss=0.04761, over 4766.00 frames. ], tot_loss[loss=0.185, simple_loss=0.2539, pruned_loss=0.05802, over 954203.06 frames. ], batch size: 28, lr: 3.58e-03, grad_scale: 32.0 2023-04-27 04:39:01,406 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=72833.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 04:39:09,032 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.148e+02 1.644e+02 1.849e+02 2.325e+02 6.848e+02, threshold=3.698e+02, percent-clipped=3.0 2023-04-27 04:39:34,765 INFO [finetune.py:976] (1/7) Epoch 13, batch 4150, loss[loss=0.1812, simple_loss=0.2516, pruned_loss=0.05544, over 4864.00 frames. ], tot_loss[loss=0.1854, simple_loss=0.2546, pruned_loss=0.05811, over 952665.14 frames. ], batch size: 31, lr: 3.58e-03, grad_scale: 32.0 2023-04-27 04:39:42,051 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=72894.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 04:39:47,754 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.0804, 2.0826, 1.7034, 1.7609, 2.1983, 1.8283, 2.6762, 1.4979], device='cuda:1'), covar=tensor([0.3971, 0.2042, 0.4938, 0.3417, 0.1742, 0.2591, 0.1417, 0.4850], device='cuda:1'), in_proj_covar=tensor([0.0344, 0.0345, 0.0428, 0.0356, 0.0383, 0.0382, 0.0373, 0.0419], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 04:40:08,512 INFO [finetune.py:976] (1/7) Epoch 13, batch 4200, loss[loss=0.2037, simple_loss=0.2729, pruned_loss=0.0672, over 4810.00 frames. ], tot_loss[loss=0.1859, simple_loss=0.2555, pruned_loss=0.05817, over 952392.50 frames. ], batch size: 40, lr: 3.58e-03, grad_scale: 32.0 2023-04-27 04:40:09,242 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7624, 1.7084, 0.7663, 1.4429, 1.7745, 1.5768, 1.4541, 1.5428], device='cuda:1'), covar=tensor([0.0528, 0.0368, 0.0390, 0.0557, 0.0283, 0.0523, 0.0531, 0.0573], device='cuda:1'), in_proj_covar=tensor([0.0029, 0.0025, 0.0023, 0.0029, 0.0020, 0.0029, 0.0028, 0.0029], device='cuda:1'), out_proj_covar=tensor([0.0050, 0.0044, 0.0038, 0.0050, 0.0037, 0.0049, 0.0049, 0.0050], device='cuda:1') 2023-04-27 04:40:15,130 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.034e+02 1.607e+02 1.987e+02 2.411e+02 4.371e+02, threshold=3.975e+02, percent-clipped=4.0 2023-04-27 04:40:23,785 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=72955.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 04:40:30,254 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=72964.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 04:40:30,826 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6304, 3.6912, 0.9847, 1.9051, 2.1777, 2.5714, 2.0458, 1.0205], device='cuda:1'), covar=tensor([0.1315, 0.0869, 0.2045, 0.1247, 0.0975, 0.1045, 0.1502, 0.2005], device='cuda:1'), in_proj_covar=tensor([0.0119, 0.0248, 0.0140, 0.0122, 0.0134, 0.0153, 0.0118, 0.0121], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-27 04:40:41,142 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=72982.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 04:40:41,680 INFO [finetune.py:976] (1/7) Epoch 13, batch 4250, loss[loss=0.1945, simple_loss=0.243, pruned_loss=0.07306, over 4815.00 frames. ], tot_loss[loss=0.1839, simple_loss=0.253, pruned_loss=0.05744, over 952202.09 frames. ], batch size: 30, lr: 3.58e-03, grad_scale: 32.0 2023-04-27 04:41:08,940 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-04-27 04:41:10,631 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=73025.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 04:41:13,601 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=73030.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 04:41:15,357 INFO [finetune.py:976] (1/7) Epoch 13, batch 4300, loss[loss=0.1309, simple_loss=0.1978, pruned_loss=0.032, over 4800.00 frames. ], tot_loss[loss=0.1822, simple_loss=0.2506, pruned_loss=0.05691, over 953301.34 frames. ], batch size: 29, lr: 3.58e-03, grad_scale: 32.0 2023-04-27 04:41:22,016 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.082e+02 1.560e+02 1.991e+02 2.397e+02 3.754e+02, threshold=3.982e+02, percent-clipped=0.0 2023-04-27 04:41:34,863 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5103, 1.1789, 0.3276, 1.2252, 1.1775, 1.4078, 1.2990, 1.2551], device='cuda:1'), covar=tensor([0.0533, 0.0417, 0.0448, 0.0569, 0.0292, 0.0542, 0.0503, 0.0591], device='cuda:1'), in_proj_covar=tensor([0.0029, 0.0025, 0.0023, 0.0029, 0.0020, 0.0029, 0.0028, 0.0029], device='cuda:1'), out_proj_covar=tensor([0.0050, 0.0044, 0.0038, 0.0050, 0.0037, 0.0049, 0.0049, 0.0050], device='cuda:1') 2023-04-27 04:41:51,151 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([0.7786, 1.8511, 1.8515, 1.5082, 2.0694, 1.7017, 2.6312, 1.5724], device='cuda:1'), covar=tensor([0.4333, 0.1841, 0.4700, 0.3304, 0.1657, 0.2560, 0.1356, 0.4846], device='cuda:1'), in_proj_covar=tensor([0.0344, 0.0345, 0.0426, 0.0355, 0.0381, 0.0381, 0.0373, 0.0417], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 04:41:59,413 INFO [finetune.py:976] (1/7) Epoch 13, batch 4350, loss[loss=0.2178, simple_loss=0.2788, pruned_loss=0.07839, over 4753.00 frames. ], tot_loss[loss=0.1788, simple_loss=0.2471, pruned_loss=0.05526, over 954145.66 frames. ], batch size: 59, lr: 3.58e-03, grad_scale: 32.0 2023-04-27 04:42:10,052 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.7740, 2.0827, 2.5113, 3.2966, 2.4369, 1.9709, 1.9638, 2.5005], device='cuda:1'), covar=tensor([0.3354, 0.3484, 0.1656, 0.2389, 0.2995, 0.2732, 0.4115, 0.2458], device='cuda:1'), in_proj_covar=tensor([0.0288, 0.0248, 0.0223, 0.0317, 0.0215, 0.0230, 0.0232, 0.0187], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-27 04:42:13,197 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.13 vs. limit=2.0 2023-04-27 04:42:16,991 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.52 vs. limit=5.0 2023-04-27 04:42:26,071 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9381, 2.7092, 2.0522, 2.1530, 1.5453, 1.3415, 2.1093, 1.4765], device='cuda:1'), covar=tensor([0.1757, 0.1536, 0.1496, 0.1702, 0.2429, 0.2102, 0.1074, 0.2104], device='cuda:1'), in_proj_covar=tensor([0.0195, 0.0213, 0.0169, 0.0203, 0.0201, 0.0183, 0.0157, 0.0186], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-27 04:42:38,505 INFO [finetune.py:976] (1/7) Epoch 13, batch 4400, loss[loss=0.1489, simple_loss=0.2381, pruned_loss=0.02981, over 4865.00 frames. ], tot_loss[loss=0.1798, simple_loss=0.2484, pruned_loss=0.05564, over 954658.81 frames. ], batch size: 44, lr: 3.58e-03, grad_scale: 32.0 2023-04-27 04:42:47,735 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.074e+02 1.646e+02 2.036e+02 2.477e+02 5.410e+02, threshold=4.072e+02, percent-clipped=5.0 2023-04-27 04:43:39,918 INFO [finetune.py:976] (1/7) Epoch 13, batch 4450, loss[loss=0.1832, simple_loss=0.2646, pruned_loss=0.05089, over 4889.00 frames. ], tot_loss[loss=0.184, simple_loss=0.2531, pruned_loss=0.0574, over 951953.53 frames. ], batch size: 32, lr: 3.58e-03, grad_scale: 32.0 2023-04-27 04:44:30,376 INFO [finetune.py:976] (1/7) Epoch 13, batch 4500, loss[loss=0.1606, simple_loss=0.2499, pruned_loss=0.03563, over 4894.00 frames. ], tot_loss[loss=0.1854, simple_loss=0.2544, pruned_loss=0.05821, over 952251.26 frames. ], batch size: 35, lr: 3.58e-03, grad_scale: 32.0 2023-04-27 04:44:33,082 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-04-27 04:44:37,121 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.883e+01 1.575e+02 1.982e+02 2.231e+02 3.715e+02, threshold=3.964e+02, percent-clipped=0.0 2023-04-27 04:44:40,829 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=73250.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 04:44:45,101 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9982, 2.6196, 0.9611, 1.3864, 1.9416, 1.1851, 3.2906, 1.6099], device='cuda:1'), covar=tensor([0.0661, 0.0731, 0.0839, 0.1148, 0.0496, 0.0975, 0.0205, 0.0634], device='cuda:1'), in_proj_covar=tensor([0.0051, 0.0066, 0.0049, 0.0047, 0.0051, 0.0052, 0.0076, 0.0051], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:1') 2023-04-27 04:45:04,241 INFO [finetune.py:976] (1/7) Epoch 13, batch 4550, loss[loss=0.2135, simple_loss=0.2923, pruned_loss=0.06733, over 4741.00 frames. ], tot_loss[loss=0.1865, simple_loss=0.256, pruned_loss=0.05854, over 953252.12 frames. ], batch size: 54, lr: 3.58e-03, grad_scale: 32.0 2023-04-27 04:45:10,363 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=73293.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 04:45:17,026 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.3169, 1.2379, 1.5502, 1.4610, 1.2153, 1.1078, 1.2766, 0.9114], device='cuda:1'), covar=tensor([0.0576, 0.0592, 0.0476, 0.0706, 0.0791, 0.1145, 0.0563, 0.0625], device='cuda:1'), in_proj_covar=tensor([0.0068, 0.0070, 0.0070, 0.0068, 0.0075, 0.0096, 0.0075, 0.0069], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-27 04:45:28,149 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=73320.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 04:45:37,747 INFO [finetune.py:976] (1/7) Epoch 13, batch 4600, loss[loss=0.176, simple_loss=0.2411, pruned_loss=0.05544, over 4830.00 frames. ], tot_loss[loss=0.1849, simple_loss=0.2543, pruned_loss=0.05779, over 953778.46 frames. ], batch size: 38, lr: 3.58e-03, grad_scale: 32.0 2023-04-27 04:45:44,461 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.205e+02 1.640e+02 1.996e+02 2.306e+02 3.998e+02, threshold=3.993e+02, percent-clipped=1.0 2023-04-27 04:45:50,687 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=73354.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 04:46:11,138 INFO [finetune.py:976] (1/7) Epoch 13, batch 4650, loss[loss=0.1599, simple_loss=0.2245, pruned_loss=0.04771, over 4914.00 frames. ], tot_loss[loss=0.1839, simple_loss=0.2524, pruned_loss=0.05769, over 953611.25 frames. ], batch size: 36, lr: 3.58e-03, grad_scale: 32.0 2023-04-27 04:46:19,637 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.2874, 1.7450, 2.2248, 2.5501, 2.1492, 1.6812, 1.4949, 2.0117], device='cuda:1'), covar=tensor([0.3450, 0.3382, 0.1763, 0.2577, 0.2834, 0.2836, 0.4356, 0.2149], device='cuda:1'), in_proj_covar=tensor([0.0288, 0.0248, 0.0223, 0.0317, 0.0215, 0.0230, 0.0232, 0.0186], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-27 04:46:44,503 INFO [finetune.py:976] (1/7) Epoch 13, batch 4700, loss[loss=0.1537, simple_loss=0.223, pruned_loss=0.04222, over 4811.00 frames. ], tot_loss[loss=0.1813, simple_loss=0.2494, pruned_loss=0.05664, over 955767.99 frames. ], batch size: 51, lr: 3.58e-03, grad_scale: 32.0 2023-04-27 04:46:51,528 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.168e+02 1.646e+02 2.000e+02 2.343e+02 4.660e+02, threshold=4.000e+02, percent-clipped=2.0 2023-04-27 04:47:07,211 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5003, 1.4092, 1.8515, 1.7971, 1.3624, 1.1764, 1.4935, 0.9484], device='cuda:1'), covar=tensor([0.0659, 0.0619, 0.0477, 0.0638, 0.0829, 0.1310, 0.0705, 0.0768], device='cuda:1'), in_proj_covar=tensor([0.0068, 0.0070, 0.0070, 0.0067, 0.0075, 0.0096, 0.0075, 0.0069], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-27 04:47:29,985 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.84 vs. limit=5.0 2023-04-27 04:47:33,204 INFO [finetune.py:976] (1/7) Epoch 13, batch 4750, loss[loss=0.1977, simple_loss=0.2601, pruned_loss=0.06762, over 4903.00 frames. ], tot_loss[loss=0.1802, simple_loss=0.2478, pruned_loss=0.05627, over 952826.37 frames. ], batch size: 35, lr: 3.58e-03, grad_scale: 32.0 2023-04-27 04:47:52,975 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-04-27 04:48:21,163 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7992, 1.3938, 1.5052, 1.4720, 1.9155, 1.6323, 1.3586, 1.3926], device='cuda:1'), covar=tensor([0.1570, 0.1404, 0.2214, 0.1569, 0.1028, 0.1453, 0.1973, 0.2323], device='cuda:1'), in_proj_covar=tensor([0.0308, 0.0318, 0.0353, 0.0293, 0.0330, 0.0316, 0.0305, 0.0361], device='cuda:1'), out_proj_covar=tensor([6.4098e-05, 6.6788e-05, 7.5613e-05, 5.9991e-05, 6.8779e-05, 6.7108e-05, 6.4801e-05, 7.7210e-05], device='cuda:1') 2023-04-27 04:48:28,608 INFO [finetune.py:976] (1/7) Epoch 13, batch 4800, loss[loss=0.2134, simple_loss=0.2909, pruned_loss=0.06796, over 4818.00 frames. ], tot_loss[loss=0.1826, simple_loss=0.2503, pruned_loss=0.05747, over 953098.33 frames. ], batch size: 38, lr: 3.58e-03, grad_scale: 32.0 2023-04-27 04:48:34,432 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.0966, 2.3511, 1.0718, 1.4602, 1.3912, 1.9509, 1.5404, 1.0416], device='cuda:1'), covar=tensor([0.1276, 0.1079, 0.1382, 0.1130, 0.1078, 0.0790, 0.1382, 0.1644], device='cuda:1'), in_proj_covar=tensor([0.0118, 0.0247, 0.0139, 0.0122, 0.0132, 0.0153, 0.0118, 0.0121], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-27 04:48:36,789 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.040e+02 1.600e+02 1.891e+02 2.207e+02 3.425e+02, threshold=3.783e+02, percent-clipped=0.0 2023-04-27 04:48:41,071 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=73550.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 04:48:59,570 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.58 vs. limit=2.0 2023-04-27 04:49:02,204 INFO [finetune.py:976] (1/7) Epoch 13, batch 4850, loss[loss=0.1948, simple_loss=0.2658, pruned_loss=0.06184, over 4749.00 frames. ], tot_loss[loss=0.1863, simple_loss=0.2546, pruned_loss=0.05906, over 954718.97 frames. ], batch size: 59, lr: 3.58e-03, grad_scale: 32.0 2023-04-27 04:49:12,464 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=73590.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 04:49:28,459 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=73598.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 04:49:53,458 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=73620.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 04:50:04,209 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([0.3669, 1.3236, 1.3819, 0.9985, 1.4436, 1.1953, 1.7725, 1.3357], device='cuda:1'), covar=tensor([0.3916, 0.1876, 0.5502, 0.2997, 0.1578, 0.2307, 0.1698, 0.4780], device='cuda:1'), in_proj_covar=tensor([0.0344, 0.0349, 0.0429, 0.0357, 0.0385, 0.0385, 0.0374, 0.0420], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 04:50:13,628 INFO [finetune.py:976] (1/7) Epoch 13, batch 4900, loss[loss=0.183, simple_loss=0.2611, pruned_loss=0.05247, over 4892.00 frames. ], tot_loss[loss=0.1876, simple_loss=0.2561, pruned_loss=0.0595, over 955965.39 frames. ], batch size: 36, lr: 3.58e-03, grad_scale: 32.0 2023-04-27 04:50:28,449 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.161e+02 1.796e+02 2.177e+02 2.620e+02 4.604e+02, threshold=4.354e+02, percent-clipped=4.0 2023-04-27 04:50:38,431 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=73649.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 04:50:39,754 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=73651.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 04:50:55,035 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=73668.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 04:50:58,149 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8239, 1.3449, 1.6526, 1.6670, 1.5664, 1.3164, 0.7646, 1.3406], device='cuda:1'), covar=tensor([0.3351, 0.3642, 0.1808, 0.2394, 0.2634, 0.2839, 0.4455, 0.2299], device='cuda:1'), in_proj_covar=tensor([0.0288, 0.0247, 0.0222, 0.0315, 0.0215, 0.0229, 0.0231, 0.0185], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-27 04:51:04,150 INFO [finetune.py:976] (1/7) Epoch 13, batch 4950, loss[loss=0.1647, simple_loss=0.2311, pruned_loss=0.04914, over 4743.00 frames. ], tot_loss[loss=0.188, simple_loss=0.2569, pruned_loss=0.05954, over 955916.64 frames. ], batch size: 23, lr: 3.58e-03, grad_scale: 32.0 2023-04-27 04:51:14,860 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=3.51 vs. limit=5.0 2023-04-27 04:51:36,572 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=73732.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 04:51:37,095 INFO [finetune.py:976] (1/7) Epoch 13, batch 5000, loss[loss=0.1555, simple_loss=0.2212, pruned_loss=0.04488, over 4848.00 frames. ], tot_loss[loss=0.1871, simple_loss=0.2557, pruned_loss=0.05929, over 955887.68 frames. ], batch size: 47, lr: 3.58e-03, grad_scale: 32.0 2023-04-27 04:51:45,208 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.282e+02 1.739e+02 2.081e+02 2.360e+02 4.914e+02, threshold=4.162e+02, percent-clipped=1.0 2023-04-27 04:51:46,546 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.2590, 1.5889, 1.4893, 1.8481, 1.7566, 2.1714, 1.5323, 4.0958], device='cuda:1'), covar=tensor([0.0561, 0.0741, 0.0789, 0.1173, 0.0631, 0.0540, 0.0748, 0.0115], device='cuda:1'), in_proj_covar=tensor([0.0038, 0.0039, 0.0040, 0.0044, 0.0040, 0.0038, 0.0038, 0.0057], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:1') 2023-04-27 04:51:48,321 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.93 vs. limit=5.0 2023-04-27 04:51:54,781 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0389, 1.8618, 2.4271, 2.5910, 1.8235, 1.5537, 1.9248, 0.9984], device='cuda:1'), covar=tensor([0.0780, 0.0861, 0.0555, 0.0824, 0.0885, 0.1359, 0.0955, 0.0939], device='cuda:1'), in_proj_covar=tensor([0.0068, 0.0070, 0.0070, 0.0068, 0.0075, 0.0096, 0.0076, 0.0069], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-27 04:52:11,151 INFO [finetune.py:976] (1/7) Epoch 13, batch 5050, loss[loss=0.1527, simple_loss=0.2076, pruned_loss=0.04883, over 3891.00 frames. ], tot_loss[loss=0.1844, simple_loss=0.2524, pruned_loss=0.05822, over 954146.01 frames. ], batch size: 16, lr: 3.58e-03, grad_scale: 32.0 2023-04-27 04:52:17,787 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=73793.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 04:52:56,560 INFO [finetune.py:976] (1/7) Epoch 13, batch 5100, loss[loss=0.1493, simple_loss=0.2243, pruned_loss=0.03717, over 4802.00 frames. ], tot_loss[loss=0.1819, simple_loss=0.2494, pruned_loss=0.0572, over 955554.80 frames. ], batch size: 25, lr: 3.57e-03, grad_scale: 32.0 2023-04-27 04:52:58,491 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.3636, 1.6704, 2.1668, 2.8301, 2.2592, 1.7308, 1.7584, 2.1378], device='cuda:1'), covar=tensor([0.3359, 0.3964, 0.1836, 0.2639, 0.2958, 0.2557, 0.4259, 0.2458], device='cuda:1'), in_proj_covar=tensor([0.0289, 0.0248, 0.0223, 0.0318, 0.0216, 0.0230, 0.0232, 0.0186], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-27 04:53:03,638 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.084e+02 1.578e+02 1.847e+02 2.242e+02 3.779e+02, threshold=3.694e+02, percent-clipped=0.0 2023-04-27 04:53:14,201 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-04-27 04:53:44,068 INFO [finetune.py:976] (1/7) Epoch 13, batch 5150, loss[loss=0.2071, simple_loss=0.2779, pruned_loss=0.06813, over 4821.00 frames. ], tot_loss[loss=0.1822, simple_loss=0.2497, pruned_loss=0.05738, over 956101.91 frames. ], batch size: 39, lr: 3.57e-03, grad_scale: 32.0 2023-04-27 04:54:09,618 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5147, 1.4505, 1.9007, 1.8608, 1.3762, 1.2073, 1.5482, 0.9409], device='cuda:1'), covar=tensor([0.0736, 0.0844, 0.0534, 0.0839, 0.0882, 0.1219, 0.0845, 0.0787], device='cuda:1'), in_proj_covar=tensor([0.0068, 0.0070, 0.0071, 0.0068, 0.0075, 0.0096, 0.0076, 0.0070], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-27 04:54:13,858 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7067, 1.2175, 4.2257, 3.6594, 3.7394, 3.8983, 3.7582, 3.5679], device='cuda:1'), covar=tensor([0.9307, 0.8920, 0.1641, 0.3374, 0.2280, 0.3729, 0.3064, 0.2788], device='cuda:1'), in_proj_covar=tensor([0.0306, 0.0301, 0.0396, 0.0401, 0.0340, 0.0398, 0.0309, 0.0361], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-27 04:54:18,052 INFO [finetune.py:976] (1/7) Epoch 13, batch 5200, loss[loss=0.2154, simple_loss=0.2824, pruned_loss=0.07418, over 4830.00 frames. ], tot_loss[loss=0.1865, simple_loss=0.2549, pruned_loss=0.05907, over 957143.48 frames. ], batch size: 33, lr: 3.57e-03, grad_scale: 32.0 2023-04-27 04:54:24,746 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.235e+02 1.597e+02 1.940e+02 2.333e+02 4.411e+02, threshold=3.880e+02, percent-clipped=2.0 2023-04-27 04:54:26,016 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=73946.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 04:54:28,372 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=73949.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 04:55:07,119 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-04-27 04:55:07,560 INFO [finetune.py:976] (1/7) Epoch 13, batch 5250, loss[loss=0.1793, simple_loss=0.249, pruned_loss=0.05475, over 4828.00 frames. ], tot_loss[loss=0.186, simple_loss=0.2553, pruned_loss=0.05832, over 958540.09 frames. ], batch size: 39, lr: 3.57e-03, grad_scale: 32.0 2023-04-27 04:55:16,182 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=73997.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 04:55:42,305 INFO [finetune.py:976] (1/7) Epoch 13, batch 5300, loss[loss=0.1814, simple_loss=0.2561, pruned_loss=0.05338, over 4809.00 frames. ], tot_loss[loss=0.1866, simple_loss=0.2558, pruned_loss=0.05866, over 958498.24 frames. ], batch size: 40, lr: 3.57e-03, grad_scale: 64.0 2023-04-27 04:55:54,435 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.148e+02 1.549e+02 1.803e+02 2.231e+02 3.853e+02, threshold=3.607e+02, percent-clipped=0.0 2023-04-27 04:56:19,399 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=74064.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 04:56:35,116 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.2012, 1.5167, 1.4097, 1.7234, 1.6241, 1.9050, 1.4009, 3.4315], device='cuda:1'), covar=tensor([0.0586, 0.0823, 0.0783, 0.1195, 0.0645, 0.0515, 0.0762, 0.0150], device='cuda:1'), in_proj_covar=tensor([0.0038, 0.0039, 0.0040, 0.0043, 0.0040, 0.0038, 0.0038, 0.0057], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:1') 2023-04-27 04:56:36,836 INFO [finetune.py:976] (1/7) Epoch 13, batch 5350, loss[loss=0.2444, simple_loss=0.2928, pruned_loss=0.09802, over 4153.00 frames. ], tot_loss[loss=0.1871, simple_loss=0.2559, pruned_loss=0.05916, over 956312.73 frames. ], batch size: 18, lr: 3.57e-03, grad_scale: 64.0 2023-04-27 04:56:39,358 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9158, 1.4744, 1.7734, 1.8196, 1.7191, 1.4056, 0.8260, 1.4820], device='cuda:1'), covar=tensor([0.3343, 0.3330, 0.1782, 0.2124, 0.2623, 0.2658, 0.3965, 0.2032], device='cuda:1'), in_proj_covar=tensor([0.0288, 0.0247, 0.0222, 0.0316, 0.0215, 0.0229, 0.0230, 0.0185], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-27 04:56:39,879 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=74088.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 04:57:05,264 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=74125.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 04:57:10,417 INFO [finetune.py:976] (1/7) Epoch 13, batch 5400, loss[loss=0.1612, simple_loss=0.2383, pruned_loss=0.04209, over 4923.00 frames. ], tot_loss[loss=0.1853, simple_loss=0.2532, pruned_loss=0.05873, over 954655.10 frames. ], batch size: 38, lr: 3.57e-03, grad_scale: 64.0 2023-04-27 04:57:17,119 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.047e+02 1.609e+02 1.962e+02 2.381e+02 6.341e+02, threshold=3.923e+02, percent-clipped=2.0 2023-04-27 04:57:43,730 INFO [finetune.py:976] (1/7) Epoch 13, batch 5450, loss[loss=0.1544, simple_loss=0.222, pruned_loss=0.04345, over 4717.00 frames. ], tot_loss[loss=0.1823, simple_loss=0.2503, pruned_loss=0.05713, over 955642.36 frames. ], batch size: 23, lr: 3.57e-03, grad_scale: 64.0 2023-04-27 04:57:57,207 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.38 vs. limit=5.0 2023-04-27 04:58:15,603 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.5302, 1.7773, 1.7242, 2.2625, 2.4866, 2.0896, 1.9862, 1.8812], device='cuda:1'), covar=tensor([0.1582, 0.1721, 0.1715, 0.1857, 0.1225, 0.1699, 0.2195, 0.1912], device='cuda:1'), in_proj_covar=tensor([0.0306, 0.0315, 0.0351, 0.0291, 0.0329, 0.0313, 0.0306, 0.0361], device='cuda:1'), out_proj_covar=tensor([6.3609e-05, 6.6209e-05, 7.5182e-05, 5.9405e-05, 6.8589e-05, 6.6458e-05, 6.4863e-05, 7.7183e-05], device='cuda:1') 2023-04-27 04:58:25,836 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.70 vs. limit=5.0 2023-04-27 04:58:33,285 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.1709, 2.6603, 1.0590, 1.4301, 2.1267, 1.2553, 3.4612, 1.9505], device='cuda:1'), covar=tensor([0.0620, 0.0542, 0.0751, 0.1321, 0.0490, 0.1020, 0.0232, 0.0616], device='cuda:1'), in_proj_covar=tensor([0.0050, 0.0065, 0.0048, 0.0046, 0.0050, 0.0051, 0.0075, 0.0050], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0007, 0.0008, 0.0008, 0.0010, 0.0007], device='cuda:1') 2023-04-27 04:58:35,027 INFO [finetune.py:976] (1/7) Epoch 13, batch 5500, loss[loss=0.1851, simple_loss=0.2569, pruned_loss=0.05664, over 4904.00 frames. ], tot_loss[loss=0.1817, simple_loss=0.249, pruned_loss=0.05715, over 956274.22 frames. ], batch size: 43, lr: 3.57e-03, grad_scale: 64.0 2023-04-27 04:58:39,428 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=74240.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 04:58:42,297 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.038e+02 1.578e+02 1.994e+02 2.336e+02 4.147e+02, threshold=3.989e+02, percent-clipped=1.0 2023-04-27 04:58:43,617 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=74246.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 04:59:08,625 INFO [finetune.py:976] (1/7) Epoch 13, batch 5550, loss[loss=0.1801, simple_loss=0.2576, pruned_loss=0.05135, over 4791.00 frames. ], tot_loss[loss=0.1827, simple_loss=0.25, pruned_loss=0.05767, over 956638.24 frames. ], batch size: 29, lr: 3.57e-03, grad_scale: 64.0 2023-04-27 04:59:15,400 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=74294.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 04:59:20,249 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=74301.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 04:59:23,244 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5903, 1.6731, 0.8311, 1.2883, 1.7306, 1.4597, 1.3665, 1.3759], device='cuda:1'), covar=tensor([0.0497, 0.0357, 0.0371, 0.0554, 0.0280, 0.0504, 0.0475, 0.0573], device='cuda:1'), in_proj_covar=tensor([0.0029, 0.0025, 0.0023, 0.0029, 0.0020, 0.0029, 0.0028, 0.0029], device='cuda:1'), out_proj_covar=tensor([0.0050, 0.0045, 0.0038, 0.0051, 0.0038, 0.0049, 0.0049, 0.0050], device='cuda:1') 2023-04-27 04:59:39,647 INFO [finetune.py:976] (1/7) Epoch 13, batch 5600, loss[loss=0.187, simple_loss=0.2574, pruned_loss=0.05826, over 4898.00 frames. ], tot_loss[loss=0.1859, simple_loss=0.2542, pruned_loss=0.05881, over 957451.59 frames. ], batch size: 35, lr: 3.57e-03, grad_scale: 64.0 2023-04-27 04:59:42,091 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5013, 0.6347, 1.3364, 1.8432, 1.5666, 1.3897, 1.3821, 1.4504], device='cuda:1'), covar=tensor([0.4526, 0.6764, 0.6527, 0.6399, 0.5742, 0.7459, 0.7805, 0.8058], device='cuda:1'), in_proj_covar=tensor([0.0411, 0.0406, 0.0493, 0.0511, 0.0442, 0.0462, 0.0469, 0.0473], device='cuda:1'), out_proj_covar=tensor([9.9534e-05, 1.0071e-04, 1.1113e-04, 1.2156e-04, 1.0655e-04, 1.1136e-04, 1.1213e-04, 1.1278e-04], device='cuda:1') 2023-04-27 04:59:46,068 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.158e+02 1.613e+02 1.797e+02 2.128e+02 3.644e+02, threshold=3.594e+02, percent-clipped=1.0 2023-04-27 04:59:50,332 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-04-27 05:00:16,163 INFO [finetune.py:976] (1/7) Epoch 13, batch 5650, loss[loss=0.2113, simple_loss=0.2963, pruned_loss=0.06315, over 4819.00 frames. ], tot_loss[loss=0.1893, simple_loss=0.2582, pruned_loss=0.06018, over 957744.83 frames. ], batch size: 45, lr: 3.57e-03, grad_scale: 64.0 2023-04-27 05:00:25,008 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=74388.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 05:00:31,945 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-04-27 05:00:44,665 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=74420.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 05:00:52,319 INFO [finetune.py:976] (1/7) Epoch 13, batch 5700, loss[loss=0.1784, simple_loss=0.2332, pruned_loss=0.06183, over 4319.00 frames. ], tot_loss[loss=0.1862, simple_loss=0.2536, pruned_loss=0.05938, over 939069.07 frames. ], batch size: 19, lr: 3.57e-03, grad_scale: 64.0 2023-04-27 05:00:54,136 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=74436.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 05:00:58,755 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.054e+02 1.564e+02 1.898e+02 2.359e+02 3.680e+02, threshold=3.797e+02, percent-clipped=1.0 2023-04-27 05:01:23,651 INFO [finetune.py:976] (1/7) Epoch 14, batch 0, loss[loss=0.1775, simple_loss=0.2554, pruned_loss=0.04983, over 4924.00 frames. ], tot_loss[loss=0.1775, simple_loss=0.2554, pruned_loss=0.04983, over 4924.00 frames. ], batch size: 42, lr: 3.57e-03, grad_scale: 64.0 2023-04-27 05:01:23,651 INFO [finetune.py:1001] (1/7) Computing validation loss 2023-04-27 05:01:45,227 INFO [finetune.py:1010] (1/7) Epoch 14, validation: loss=0.1535, simple_loss=0.226, pruned_loss=0.04054, over 2265189.00 frames. 2023-04-27 05:01:45,228 INFO [finetune.py:1011] (1/7) Maximum memory allocated so far is 6391MB 2023-04-27 05:01:57,190 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.1376, 1.3388, 1.2397, 1.6562, 1.4996, 1.6176, 1.2773, 2.4514], device='cuda:1'), covar=tensor([0.0615, 0.0895, 0.0891, 0.1314, 0.0697, 0.0475, 0.0813, 0.0273], device='cuda:1'), in_proj_covar=tensor([0.0038, 0.0039, 0.0040, 0.0043, 0.0040, 0.0038, 0.0038, 0.0057], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0014, 0.0013, 0.0012, 0.0013, 0.0016], device='cuda:1') 2023-04-27 05:02:18,139 INFO [finetune.py:976] (1/7) Epoch 14, batch 50, loss[loss=0.2003, simple_loss=0.2658, pruned_loss=0.06743, over 4864.00 frames. ], tot_loss[loss=0.1932, simple_loss=0.2609, pruned_loss=0.06279, over 217671.42 frames. ], batch size: 34, lr: 3.57e-03, grad_scale: 32.0 2023-04-27 05:02:41,160 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.646e+01 1.673e+02 1.901e+02 2.437e+02 4.267e+02, threshold=3.802e+02, percent-clipped=2.0 2023-04-27 05:02:51,803 INFO [finetune.py:976] (1/7) Epoch 14, batch 100, loss[loss=0.2022, simple_loss=0.2712, pruned_loss=0.06655, over 4899.00 frames. ], tot_loss[loss=0.1854, simple_loss=0.2516, pruned_loss=0.05959, over 380257.12 frames. ], batch size: 37, lr: 3.57e-03, grad_scale: 32.0 2023-04-27 05:02:55,961 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2023-04-27 05:02:56,506 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4255, 3.5818, 0.9971, 1.7765, 1.9606, 2.5700, 1.9531, 0.9894], device='cuda:1'), covar=tensor([0.1514, 0.0920, 0.1988, 0.1357, 0.1141, 0.1004, 0.1579, 0.2078], device='cuda:1'), in_proj_covar=tensor([0.0117, 0.0244, 0.0137, 0.0120, 0.0132, 0.0151, 0.0117, 0.0119], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-27 05:03:29,447 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=74595.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 05:03:30,031 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=74596.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 05:03:51,942 INFO [finetune.py:976] (1/7) Epoch 14, batch 150, loss[loss=0.137, simple_loss=0.2035, pruned_loss=0.03525, over 4816.00 frames. ], tot_loss[loss=0.1794, simple_loss=0.2452, pruned_loss=0.05678, over 506925.92 frames. ], batch size: 41, lr: 3.57e-03, grad_scale: 32.0 2023-04-27 05:04:20,051 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.093e+02 1.548e+02 1.852e+02 2.253e+02 4.630e+02, threshold=3.704e+02, percent-clipped=3.0 2023-04-27 05:04:27,512 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=74656.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 05:04:31,408 INFO [finetune.py:976] (1/7) Epoch 14, batch 200, loss[loss=0.1912, simple_loss=0.2633, pruned_loss=0.05954, over 3992.00 frames. ], tot_loss[loss=0.1774, simple_loss=0.2442, pruned_loss=0.05529, over 606636.98 frames. ], batch size: 65, lr: 3.57e-03, grad_scale: 32.0 2023-04-27 05:04:50,489 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5650, 1.6451, 0.9086, 1.2320, 1.7526, 1.4152, 1.3113, 1.3268], device='cuda:1'), covar=tensor([0.0537, 0.0354, 0.0359, 0.0583, 0.0276, 0.0525, 0.0484, 0.0602], device='cuda:1'), in_proj_covar=tensor([0.0029, 0.0025, 0.0023, 0.0029, 0.0020, 0.0029, 0.0028, 0.0029], device='cuda:1'), out_proj_covar=tensor([0.0050, 0.0045, 0.0038, 0.0051, 0.0038, 0.0049, 0.0049, 0.0050], device='cuda:1') 2023-04-27 05:04:58,315 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5733, 1.7051, 0.8382, 1.2549, 1.8022, 1.4393, 1.3548, 1.3563], device='cuda:1'), covar=tensor([0.0505, 0.0385, 0.0354, 0.0575, 0.0277, 0.0499, 0.0486, 0.0593], device='cuda:1'), in_proj_covar=tensor([0.0029, 0.0025, 0.0023, 0.0029, 0.0020, 0.0029, 0.0028, 0.0029], device='cuda:1'), out_proj_covar=tensor([0.0050, 0.0044, 0.0038, 0.0051, 0.0037, 0.0049, 0.0049, 0.0050], device='cuda:1') 2023-04-27 05:05:05,314 INFO [finetune.py:976] (1/7) Epoch 14, batch 250, loss[loss=0.1698, simple_loss=0.241, pruned_loss=0.04934, over 4813.00 frames. ], tot_loss[loss=0.1836, simple_loss=0.2516, pruned_loss=0.05779, over 685506.73 frames. ], batch size: 51, lr: 3.57e-03, grad_scale: 32.0 2023-04-27 05:05:11,820 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=74720.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 05:05:39,275 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 8.695e+01 1.601e+02 1.840e+02 2.318e+02 4.456e+02, threshold=3.680e+02, percent-clipped=1.0 2023-04-27 05:05:49,930 INFO [finetune.py:976] (1/7) Epoch 14, batch 300, loss[loss=0.2092, simple_loss=0.276, pruned_loss=0.07121, over 4758.00 frames. ], tot_loss[loss=0.1839, simple_loss=0.2527, pruned_loss=0.0576, over 745652.48 frames. ], batch size: 28, lr: 3.56e-03, grad_scale: 32.0 2023-04-27 05:05:50,064 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.2953, 1.6643, 1.5951, 2.1598, 2.4215, 1.8736, 1.8543, 1.6530], device='cuda:1'), covar=tensor([0.2232, 0.1902, 0.1932, 0.1652, 0.1116, 0.2266, 0.2501, 0.2186], device='cuda:1'), in_proj_covar=tensor([0.0307, 0.0316, 0.0351, 0.0292, 0.0329, 0.0314, 0.0305, 0.0362], device='cuda:1'), out_proj_covar=tensor([6.3907e-05, 6.6354e-05, 7.5146e-05, 5.9711e-05, 6.8594e-05, 6.6543e-05, 6.4668e-05, 7.7461e-05], device='cuda:1') 2023-04-27 05:05:54,766 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=74768.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 05:06:23,249 INFO [finetune.py:976] (1/7) Epoch 14, batch 350, loss[loss=0.2222, simple_loss=0.2843, pruned_loss=0.08007, over 4781.00 frames. ], tot_loss[loss=0.1857, simple_loss=0.2547, pruned_loss=0.05836, over 792699.16 frames. ], batch size: 51, lr: 3.56e-03, grad_scale: 32.0 2023-04-27 05:07:09,533 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.209e+02 1.682e+02 1.913e+02 2.290e+02 4.465e+02, threshold=3.826e+02, percent-clipped=1.0 2023-04-27 05:07:25,449 INFO [finetune.py:976] (1/7) Epoch 14, batch 400, loss[loss=0.1544, simple_loss=0.2385, pruned_loss=0.03519, over 4834.00 frames. ], tot_loss[loss=0.1864, simple_loss=0.2562, pruned_loss=0.05827, over 830462.96 frames. ], batch size: 49, lr: 3.56e-03, grad_scale: 32.0 2023-04-27 05:07:47,543 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.36 vs. limit=5.0 2023-04-27 05:07:49,821 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=74896.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 05:07:57,768 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=74909.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 05:07:59,308 INFO [finetune.py:976] (1/7) Epoch 14, batch 450, loss[loss=0.2048, simple_loss=0.27, pruned_loss=0.06976, over 4901.00 frames. ], tot_loss[loss=0.1849, simple_loss=0.2545, pruned_loss=0.05761, over 857056.17 frames. ], batch size: 36, lr: 3.56e-03, grad_scale: 32.0 2023-04-27 05:08:37,891 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=74944.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 05:08:38,409 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.611e+01 1.647e+02 1.819e+02 2.107e+02 4.773e+02, threshold=3.638e+02, percent-clipped=2.0 2023-04-27 05:08:42,160 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=74951.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 05:08:53,659 INFO [finetune.py:976] (1/7) Epoch 14, batch 500, loss[loss=0.1595, simple_loss=0.2335, pruned_loss=0.0427, over 4806.00 frames. ], tot_loss[loss=0.1831, simple_loss=0.2519, pruned_loss=0.05717, over 877074.59 frames. ], batch size: 45, lr: 3.56e-03, grad_scale: 32.0 2023-04-27 05:08:56,090 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=74964.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 05:08:59,768 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=74970.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 05:09:27,833 INFO [finetune.py:976] (1/7) Epoch 14, batch 550, loss[loss=0.1624, simple_loss=0.235, pruned_loss=0.04493, over 4909.00 frames. ], tot_loss[loss=0.1806, simple_loss=0.2486, pruned_loss=0.05625, over 894336.30 frames. ], batch size: 46, lr: 3.56e-03, grad_scale: 32.0 2023-04-27 05:09:37,443 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=75025.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 05:09:39,814 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.0712, 0.7591, 0.9122, 0.7613, 1.2152, 0.9857, 0.8556, 0.9621], device='cuda:1'), covar=tensor([0.1716, 0.1434, 0.2006, 0.1650, 0.1004, 0.1406, 0.1749, 0.2149], device='cuda:1'), in_proj_covar=tensor([0.0309, 0.0318, 0.0353, 0.0295, 0.0332, 0.0316, 0.0308, 0.0364], device='cuda:1'), out_proj_covar=tensor([6.4304e-05, 6.6670e-05, 7.5744e-05, 6.0288e-05, 6.9296e-05, 6.7059e-05, 6.5198e-05, 7.7714e-05], device='cuda:1') 2023-04-27 05:09:51,450 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.023e+02 1.675e+02 1.910e+02 2.434e+02 4.302e+02, threshold=3.821e+02, percent-clipped=4.0 2023-04-27 05:10:01,742 INFO [finetune.py:976] (1/7) Epoch 14, batch 600, loss[loss=0.181, simple_loss=0.2598, pruned_loss=0.05104, over 4751.00 frames. ], tot_loss[loss=0.1816, simple_loss=0.2496, pruned_loss=0.05684, over 906698.75 frames. ], batch size: 54, lr: 3.56e-03, grad_scale: 32.0 2023-04-27 05:10:35,744 INFO [finetune.py:976] (1/7) Epoch 14, batch 650, loss[loss=0.1688, simple_loss=0.2454, pruned_loss=0.04607, over 4744.00 frames. ], tot_loss[loss=0.1867, simple_loss=0.2549, pruned_loss=0.05927, over 917331.25 frames. ], batch size: 54, lr: 3.56e-03, grad_scale: 32.0 2023-04-27 05:10:37,768 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-04-27 05:10:42,569 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.80 vs. limit=2.0 2023-04-27 05:10:59,289 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.138e+02 1.654e+02 2.108e+02 2.537e+02 4.925e+02, threshold=4.216e+02, percent-clipped=6.0 2023-04-27 05:11:09,458 INFO [finetune.py:976] (1/7) Epoch 14, batch 700, loss[loss=0.1948, simple_loss=0.2689, pruned_loss=0.06041, over 4895.00 frames. ], tot_loss[loss=0.1895, simple_loss=0.2582, pruned_loss=0.06035, over 926934.58 frames. ], batch size: 35, lr: 3.56e-03, grad_scale: 32.0 2023-04-27 05:11:34,587 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.3555, 2.0303, 2.4517, 2.7449, 2.8173, 2.2371, 1.8805, 2.2189], device='cuda:1'), covar=tensor([0.0865, 0.1067, 0.0544, 0.0627, 0.0622, 0.0780, 0.0841, 0.0681], device='cuda:1'), in_proj_covar=tensor([0.0192, 0.0205, 0.0184, 0.0174, 0.0180, 0.0185, 0.0157, 0.0182], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 05:11:43,422 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.70 vs. limit=2.0 2023-04-27 05:11:46,326 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4883, 3.8404, 0.7003, 2.0827, 2.0953, 2.5632, 2.3154, 0.8163], device='cuda:1'), covar=tensor([0.1484, 0.0986, 0.2042, 0.1261, 0.1109, 0.1032, 0.1367, 0.2362], device='cuda:1'), in_proj_covar=tensor([0.0118, 0.0245, 0.0137, 0.0121, 0.0133, 0.0152, 0.0118, 0.0120], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-27 05:11:46,819 INFO [finetune.py:976] (1/7) Epoch 14, batch 750, loss[loss=0.175, simple_loss=0.2502, pruned_loss=0.04985, over 4840.00 frames. ], tot_loss[loss=0.1892, simple_loss=0.2585, pruned_loss=0.05997, over 933751.22 frames. ], batch size: 47, lr: 3.56e-03, grad_scale: 32.0 2023-04-27 05:12:09,245 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=75228.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 05:12:31,820 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.094e+02 1.600e+02 1.912e+02 2.289e+02 3.679e+02, threshold=3.824e+02, percent-clipped=0.0 2023-04-27 05:12:35,605 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=75251.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 05:12:39,474 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-04-27 05:12:41,573 INFO [finetune.py:976] (1/7) Epoch 14, batch 800, loss[loss=0.1457, simple_loss=0.221, pruned_loss=0.03522, over 4820.00 frames. ], tot_loss[loss=0.1867, simple_loss=0.2564, pruned_loss=0.05854, over 938378.49 frames. ], batch size: 33, lr: 3.56e-03, grad_scale: 32.0 2023-04-27 05:12:44,556 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=75265.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 05:12:47,626 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=75270.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 05:13:00,656 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=75289.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 05:13:07,658 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=75299.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 05:13:14,861 INFO [finetune.py:976] (1/7) Epoch 14, batch 850, loss[loss=0.1883, simple_loss=0.2477, pruned_loss=0.06446, over 4827.00 frames. ], tot_loss[loss=0.1861, simple_loss=0.2548, pruned_loss=0.05873, over 940871.30 frames. ], batch size: 30, lr: 3.56e-03, grad_scale: 32.0 2023-04-27 05:13:20,841 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=75320.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 05:13:40,168 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=75331.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 05:13:56,952 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.246e+02 1.618e+02 2.006e+02 2.388e+02 8.621e+02, threshold=4.012e+02, percent-clipped=6.0 2023-04-27 05:14:12,833 INFO [finetune.py:976] (1/7) Epoch 14, batch 900, loss[loss=0.1639, simple_loss=0.2364, pruned_loss=0.0457, over 4805.00 frames. ], tot_loss[loss=0.183, simple_loss=0.2512, pruned_loss=0.05739, over 945003.28 frames. ], batch size: 45, lr: 3.56e-03, grad_scale: 32.0 2023-04-27 05:14:57,513 INFO [finetune.py:976] (1/7) Epoch 14, batch 950, loss[loss=0.2014, simple_loss=0.2705, pruned_loss=0.06609, over 4829.00 frames. ], tot_loss[loss=0.1826, simple_loss=0.2504, pruned_loss=0.0574, over 946259.64 frames. ], batch size: 40, lr: 3.56e-03, grad_scale: 32.0 2023-04-27 05:15:02,569 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8786, 1.4705, 1.4840, 1.6350, 2.1275, 1.6961, 1.4146, 1.4024], device='cuda:1'), covar=tensor([0.1514, 0.1356, 0.2054, 0.1331, 0.0765, 0.1441, 0.2290, 0.2170], device='cuda:1'), in_proj_covar=tensor([0.0308, 0.0317, 0.0354, 0.0293, 0.0332, 0.0315, 0.0307, 0.0363], device='cuda:1'), out_proj_covar=tensor([6.4060e-05, 6.6676e-05, 7.5968e-05, 5.9990e-05, 6.9202e-05, 6.6675e-05, 6.4967e-05, 7.7640e-05], device='cuda:1') 2023-04-27 05:15:04,273 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.71 vs. limit=2.0 2023-04-27 05:15:20,049 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.068e+02 1.489e+02 1.870e+02 2.238e+02 3.708e+02, threshold=3.739e+02, percent-clipped=0.0 2023-04-27 05:15:30,323 INFO [finetune.py:976] (1/7) Epoch 14, batch 1000, loss[loss=0.2302, simple_loss=0.2919, pruned_loss=0.08427, over 4918.00 frames. ], tot_loss[loss=0.1854, simple_loss=0.2539, pruned_loss=0.05849, over 948960.39 frames. ], batch size: 36, lr: 3.56e-03, grad_scale: 32.0 2023-04-27 05:16:03,284 INFO [finetune.py:976] (1/7) Epoch 14, batch 1050, loss[loss=0.1446, simple_loss=0.2168, pruned_loss=0.0362, over 4881.00 frames. ], tot_loss[loss=0.1861, simple_loss=0.2549, pruned_loss=0.05868, over 950520.76 frames. ], batch size: 32, lr: 3.56e-03, grad_scale: 32.0 2023-04-27 05:16:25,345 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.198e+02 1.730e+02 1.979e+02 2.287e+02 8.214e+02, threshold=3.958e+02, percent-clipped=1.0 2023-04-27 05:16:36,963 INFO [finetune.py:976] (1/7) Epoch 14, batch 1100, loss[loss=0.1567, simple_loss=0.2276, pruned_loss=0.04291, over 4736.00 frames. ], tot_loss[loss=0.1872, simple_loss=0.2562, pruned_loss=0.05906, over 952210.68 frames. ], batch size: 23, lr: 3.56e-03, grad_scale: 32.0 2023-04-27 05:16:39,533 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=75565.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 05:16:57,395 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=75584.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 05:17:31,382 INFO [finetune.py:976] (1/7) Epoch 14, batch 1150, loss[loss=0.1446, simple_loss=0.2192, pruned_loss=0.03504, over 4788.00 frames. ], tot_loss[loss=0.1885, simple_loss=0.2578, pruned_loss=0.05954, over 953842.16 frames. ], batch size: 26, lr: 3.56e-03, grad_scale: 32.0 2023-04-27 05:17:32,672 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=75613.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 05:17:36,932 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=75620.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 05:17:40,632 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=75626.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 05:17:53,272 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.071e+02 1.637e+02 1.916e+02 2.260e+02 4.790e+02, threshold=3.833e+02, percent-clipped=2.0 2023-04-27 05:18:05,378 INFO [finetune.py:976] (1/7) Epoch 14, batch 1200, loss[loss=0.1423, simple_loss=0.2245, pruned_loss=0.03002, over 4899.00 frames. ], tot_loss[loss=0.1861, simple_loss=0.2551, pruned_loss=0.05857, over 956236.13 frames. ], batch size: 35, lr: 3.56e-03, grad_scale: 32.0 2023-04-27 05:18:09,671 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=75668.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 05:18:24,671 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5440, 3.3555, 0.8740, 2.0132, 2.0506, 2.4279, 2.0184, 0.9742], device='cuda:1'), covar=tensor([0.1278, 0.0864, 0.2024, 0.1115, 0.0925, 0.0984, 0.1349, 0.1936], device='cuda:1'), in_proj_covar=tensor([0.0118, 0.0246, 0.0138, 0.0121, 0.0133, 0.0153, 0.0118, 0.0120], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-27 05:18:43,840 INFO [finetune.py:976] (1/7) Epoch 14, batch 1250, loss[loss=0.1685, simple_loss=0.2496, pruned_loss=0.04367, over 4836.00 frames. ], tot_loss[loss=0.1836, simple_loss=0.2518, pruned_loss=0.05765, over 957224.16 frames. ], batch size: 30, lr: 3.56e-03, grad_scale: 32.0 2023-04-27 05:19:24,354 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.269e+02 1.670e+02 2.009e+02 2.395e+02 6.270e+02, threshold=4.018e+02, percent-clipped=3.0 2023-04-27 05:19:45,790 INFO [finetune.py:976] (1/7) Epoch 14, batch 1300, loss[loss=0.1727, simple_loss=0.2417, pruned_loss=0.05185, over 4751.00 frames. ], tot_loss[loss=0.1815, simple_loss=0.2492, pruned_loss=0.05691, over 957365.21 frames. ], batch size: 27, lr: 3.56e-03, grad_scale: 32.0 2023-04-27 05:20:24,967 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.2955, 1.9437, 2.2531, 2.5606, 2.6935, 2.1397, 1.7833, 2.1798], device='cuda:1'), covar=tensor([0.0785, 0.1147, 0.0613, 0.0648, 0.0603, 0.0794, 0.0829, 0.0648], device='cuda:1'), in_proj_covar=tensor([0.0191, 0.0205, 0.0184, 0.0174, 0.0180, 0.0184, 0.0157, 0.0183], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 05:20:36,081 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5762, 3.5941, 0.6722, 1.9240, 2.1862, 2.5307, 1.9940, 0.9939], device='cuda:1'), covar=tensor([0.1419, 0.0932, 0.2348, 0.1323, 0.0974, 0.1099, 0.1588, 0.2127], device='cuda:1'), in_proj_covar=tensor([0.0119, 0.0248, 0.0140, 0.0122, 0.0134, 0.0154, 0.0119, 0.0121], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-27 05:20:50,514 INFO [finetune.py:976] (1/7) Epoch 14, batch 1350, loss[loss=0.177, simple_loss=0.2279, pruned_loss=0.06306, over 4142.00 frames. ], tot_loss[loss=0.1822, simple_loss=0.2498, pruned_loss=0.05728, over 957246.99 frames. ], batch size: 18, lr: 3.56e-03, grad_scale: 32.0 2023-04-27 05:21:39,793 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.162e+02 1.563e+02 1.910e+02 2.423e+02 5.618e+02, threshold=3.821e+02, percent-clipped=3.0 2023-04-27 05:21:55,189 INFO [finetune.py:976] (1/7) Epoch 14, batch 1400, loss[loss=0.2036, simple_loss=0.2822, pruned_loss=0.06244, over 4741.00 frames. ], tot_loss[loss=0.1832, simple_loss=0.2517, pruned_loss=0.05731, over 958696.45 frames. ], batch size: 59, lr: 3.56e-03, grad_scale: 32.0 2023-04-27 05:22:34,307 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=75884.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 05:22:51,835 INFO [finetune.py:976] (1/7) Epoch 14, batch 1450, loss[loss=0.1652, simple_loss=0.2432, pruned_loss=0.04359, over 4804.00 frames. ], tot_loss[loss=0.185, simple_loss=0.2544, pruned_loss=0.05781, over 958124.69 frames. ], batch size: 40, lr: 3.55e-03, grad_scale: 32.0 2023-04-27 05:22:59,001 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0008, 2.2648, 1.0089, 1.3232, 1.7879, 1.1496, 2.9283, 1.4107], device='cuda:1'), covar=tensor([0.0663, 0.0696, 0.0699, 0.1160, 0.0498, 0.0977, 0.0229, 0.0683], device='cuda:1'), in_proj_covar=tensor([0.0051, 0.0066, 0.0049, 0.0047, 0.0050, 0.0052, 0.0076, 0.0051], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:1') 2023-04-27 05:23:03,092 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=75926.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 05:23:04,354 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0163, 1.4164, 1.8001, 2.1062, 1.7495, 1.4024, 1.0162, 1.5718], device='cuda:1'), covar=tensor([0.3163, 0.3516, 0.1658, 0.2355, 0.2739, 0.2779, 0.4517, 0.2105], device='cuda:1'), in_proj_covar=tensor([0.0287, 0.0247, 0.0222, 0.0314, 0.0215, 0.0229, 0.0230, 0.0185], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-27 05:23:06,746 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=75932.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 05:23:14,580 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.134e+02 1.664e+02 2.009e+02 2.321e+02 4.322e+02, threshold=4.018e+02, percent-clipped=2.0 2023-04-27 05:23:25,298 INFO [finetune.py:976] (1/7) Epoch 14, batch 1500, loss[loss=0.1777, simple_loss=0.2549, pruned_loss=0.05023, over 4832.00 frames. ], tot_loss[loss=0.1864, simple_loss=0.2561, pruned_loss=0.05836, over 959346.19 frames. ], batch size: 49, lr: 3.55e-03, grad_scale: 32.0 2023-04-27 05:23:34,210 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=75974.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 05:24:11,481 INFO [finetune.py:976] (1/7) Epoch 14, batch 1550, loss[loss=0.1673, simple_loss=0.2303, pruned_loss=0.05211, over 4848.00 frames. ], tot_loss[loss=0.1862, simple_loss=0.2556, pruned_loss=0.0584, over 957645.30 frames. ], batch size: 44, lr: 3.55e-03, grad_scale: 32.0 2023-04-27 05:24:17,081 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.63 vs. limit=5.0 2023-04-27 05:24:40,106 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.063e+02 1.630e+02 1.920e+02 2.278e+02 6.745e+02, threshold=3.839e+02, percent-clipped=3.0 2023-04-27 05:25:01,381 INFO [finetune.py:976] (1/7) Epoch 14, batch 1600, loss[loss=0.1446, simple_loss=0.2202, pruned_loss=0.03452, over 4765.00 frames. ], tot_loss[loss=0.1828, simple_loss=0.252, pruned_loss=0.0568, over 956780.37 frames. ], batch size: 27, lr: 3.55e-03, grad_scale: 32.0 2023-04-27 05:25:32,134 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5440, 1.5007, 1.9241, 1.8379, 1.4709, 1.2396, 1.6404, 1.1965], device='cuda:1'), covar=tensor([0.0687, 0.0650, 0.0421, 0.0594, 0.0856, 0.1241, 0.0644, 0.0649], device='cuda:1'), in_proj_covar=tensor([0.0068, 0.0070, 0.0070, 0.0068, 0.0075, 0.0097, 0.0075, 0.0070], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-27 05:25:33,831 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6246, 2.0435, 1.7449, 1.9677, 1.4799, 1.7016, 1.6746, 1.3264], device='cuda:1'), covar=tensor([0.1959, 0.1213, 0.0899, 0.1148, 0.3545, 0.1178, 0.1893, 0.2411], device='cuda:1'), in_proj_covar=tensor([0.0296, 0.0315, 0.0230, 0.0287, 0.0317, 0.0271, 0.0259, 0.0275], device='cuda:1'), out_proj_covar=tensor([1.1912e-04, 1.2558e-04, 9.1876e-05, 1.1446e-04, 1.2928e-04, 1.0833e-04, 1.0503e-04, 1.1011e-04], device='cuda:1') 2023-04-27 05:25:34,942 INFO [finetune.py:976] (1/7) Epoch 14, batch 1650, loss[loss=0.1459, simple_loss=0.2219, pruned_loss=0.03495, over 4911.00 frames. ], tot_loss[loss=0.1811, simple_loss=0.2498, pruned_loss=0.05619, over 956805.93 frames. ], batch size: 37, lr: 3.55e-03, grad_scale: 32.0 2023-04-27 05:25:39,744 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.0930, 1.5738, 1.4222, 1.7242, 1.6612, 1.8871, 1.3735, 3.3428], device='cuda:1'), covar=tensor([0.0731, 0.0796, 0.0847, 0.1222, 0.0644, 0.0534, 0.0755, 0.0147], device='cuda:1'), in_proj_covar=tensor([0.0038, 0.0039, 0.0040, 0.0043, 0.0040, 0.0038, 0.0038, 0.0056], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0014, 0.0013, 0.0012, 0.0013, 0.0016], device='cuda:1') 2023-04-27 05:25:58,459 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.045e+02 1.559e+02 1.837e+02 2.256e+02 6.477e+02, threshold=3.674e+02, percent-clipped=1.0 2023-04-27 05:25:59,483 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-04-27 05:26:08,244 INFO [finetune.py:976] (1/7) Epoch 14, batch 1700, loss[loss=0.1673, simple_loss=0.2345, pruned_loss=0.05007, over 4784.00 frames. ], tot_loss[loss=0.1795, simple_loss=0.2478, pruned_loss=0.0556, over 957962.17 frames. ], batch size: 26, lr: 3.55e-03, grad_scale: 32.0 2023-04-27 05:26:25,504 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.7009, 2.1584, 1.9617, 2.1844, 2.0116, 2.1549, 2.0939, 1.9904], device='cuda:1'), covar=tensor([0.3645, 0.5947, 0.5344, 0.4772, 0.5932, 0.7070, 0.6537, 0.5988], device='cuda:1'), in_proj_covar=tensor([0.0329, 0.0376, 0.0317, 0.0328, 0.0341, 0.0398, 0.0354, 0.0324], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 05:26:42,126 INFO [finetune.py:976] (1/7) Epoch 14, batch 1750, loss[loss=0.231, simple_loss=0.2943, pruned_loss=0.08386, over 4843.00 frames. ], tot_loss[loss=0.1818, simple_loss=0.2504, pruned_loss=0.05662, over 957405.86 frames. ], batch size: 49, lr: 3.55e-03, grad_scale: 32.0 2023-04-27 05:27:06,454 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.142e+02 1.610e+02 1.943e+02 2.415e+02 4.122e+02, threshold=3.886e+02, percent-clipped=3.0 2023-04-27 05:27:14,133 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.77 vs. limit=2.0 2023-04-27 05:27:15,745 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7842, 2.2902, 1.2286, 1.5522, 2.2614, 1.6699, 1.5925, 1.7882], device='cuda:1'), covar=tensor([0.0503, 0.0349, 0.0311, 0.0570, 0.0245, 0.0522, 0.0544, 0.0536], device='cuda:1'), in_proj_covar=tensor([0.0029, 0.0025, 0.0023, 0.0029, 0.0020, 0.0029, 0.0028, 0.0029], device='cuda:1'), out_proj_covar=tensor([0.0050, 0.0045, 0.0038, 0.0051, 0.0037, 0.0049, 0.0049, 0.0050], device='cuda:1') 2023-04-27 05:27:16,246 INFO [finetune.py:976] (1/7) Epoch 14, batch 1800, loss[loss=0.2012, simple_loss=0.2646, pruned_loss=0.0689, over 4776.00 frames. ], tot_loss[loss=0.1842, simple_loss=0.2537, pruned_loss=0.05733, over 954531.53 frames. ], batch size: 29, lr: 3.55e-03, grad_scale: 32.0 2023-04-27 05:27:27,556 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=76269.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 05:28:02,610 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=76294.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 05:28:12,928 INFO [finetune.py:976] (1/7) Epoch 14, batch 1850, loss[loss=0.1995, simple_loss=0.2651, pruned_loss=0.06698, over 4892.00 frames. ], tot_loss[loss=0.1857, simple_loss=0.2551, pruned_loss=0.05818, over 954845.58 frames. ], batch size: 35, lr: 3.55e-03, grad_scale: 32.0 2023-04-27 05:28:20,901 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=76323.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 05:28:23,442 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2023-04-27 05:28:25,668 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=76330.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 05:28:31,066 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.2242, 1.7066, 1.5786, 1.9703, 1.8780, 2.0751, 1.5571, 4.0760], device='cuda:1'), covar=tensor([0.0601, 0.0739, 0.0747, 0.1136, 0.0592, 0.0556, 0.0704, 0.0118], device='cuda:1'), in_proj_covar=tensor([0.0038, 0.0039, 0.0040, 0.0043, 0.0040, 0.0038, 0.0038, 0.0057], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0014, 0.0013, 0.0012, 0.0013, 0.0016], device='cuda:1') 2023-04-27 05:28:36,248 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.095e+02 1.653e+02 2.056e+02 2.488e+02 6.955e+02, threshold=4.112e+02, percent-clipped=5.0 2023-04-27 05:28:43,347 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=76355.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 05:28:46,905 INFO [finetune.py:976] (1/7) Epoch 14, batch 1900, loss[loss=0.1706, simple_loss=0.2403, pruned_loss=0.05045, over 4825.00 frames. ], tot_loss[loss=0.1863, simple_loss=0.2557, pruned_loss=0.0584, over 954103.70 frames. ], batch size: 30, lr: 3.55e-03, grad_scale: 32.0 2023-04-27 05:29:01,991 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=76384.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 05:29:20,625 INFO [finetune.py:976] (1/7) Epoch 14, batch 1950, loss[loss=0.1939, simple_loss=0.2576, pruned_loss=0.0651, over 4826.00 frames. ], tot_loss[loss=0.1844, simple_loss=0.2537, pruned_loss=0.05749, over 955531.32 frames. ], batch size: 30, lr: 3.55e-03, grad_scale: 32.0 2023-04-27 05:29:42,667 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.125e+01 1.567e+02 1.836e+02 2.067e+02 4.304e+02, threshold=3.671e+02, percent-clipped=1.0 2023-04-27 05:29:55,875 INFO [finetune.py:976] (1/7) Epoch 14, batch 2000, loss[loss=0.2142, simple_loss=0.2702, pruned_loss=0.07906, over 4771.00 frames. ], tot_loss[loss=0.1827, simple_loss=0.2514, pruned_loss=0.05705, over 955795.70 frames. ], batch size: 29, lr: 3.55e-03, grad_scale: 32.0 2023-04-27 05:30:58,158 INFO [finetune.py:976] (1/7) Epoch 14, batch 2050, loss[loss=0.1625, simple_loss=0.2264, pruned_loss=0.04932, over 4807.00 frames. ], tot_loss[loss=0.1802, simple_loss=0.2478, pruned_loss=0.05634, over 955150.11 frames. ], batch size: 25, lr: 3.55e-03, grad_scale: 64.0 2023-04-27 05:31:19,795 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.114e+02 1.549e+02 1.874e+02 2.355e+02 5.627e+02, threshold=3.748e+02, percent-clipped=2.0 2023-04-27 05:31:19,925 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6780, 1.5341, 1.7704, 2.0845, 2.0901, 1.6764, 1.3753, 1.8027], device='cuda:1'), covar=tensor([0.0853, 0.1298, 0.0732, 0.0609, 0.0627, 0.0760, 0.0791, 0.0635], device='cuda:1'), in_proj_covar=tensor([0.0191, 0.0203, 0.0182, 0.0173, 0.0178, 0.0182, 0.0154, 0.0181], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 05:31:32,054 INFO [finetune.py:976] (1/7) Epoch 14, batch 2100, loss[loss=0.1531, simple_loss=0.2197, pruned_loss=0.04327, over 4710.00 frames. ], tot_loss[loss=0.1796, simple_loss=0.2475, pruned_loss=0.05587, over 955511.49 frames. ], batch size: 23, lr: 3.55e-03, grad_scale: 64.0 2023-04-27 05:32:05,783 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.86 vs. limit=5.0 2023-04-27 05:32:06,102 INFO [finetune.py:976] (1/7) Epoch 14, batch 2150, loss[loss=0.1721, simple_loss=0.2403, pruned_loss=0.05189, over 4186.00 frames. ], tot_loss[loss=0.1822, simple_loss=0.2511, pruned_loss=0.05667, over 956025.84 frames. ], batch size: 65, lr: 3.55e-03, grad_scale: 64.0 2023-04-27 05:32:14,723 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=76625.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 05:32:27,725 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.061e+02 1.703e+02 1.997e+02 2.488e+02 3.640e+02, threshold=3.993e+02, percent-clipped=1.0 2023-04-27 05:32:30,846 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=76650.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 05:32:33,229 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.3097, 2.7245, 1.0921, 1.4853, 2.2854, 1.2643, 3.7253, 1.9438], device='cuda:1'), covar=tensor([0.0639, 0.0568, 0.0761, 0.1315, 0.0473, 0.1050, 0.0300, 0.0609], device='cuda:1'), in_proj_covar=tensor([0.0051, 0.0066, 0.0049, 0.0047, 0.0050, 0.0052, 0.0076, 0.0051], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:1') 2023-04-27 05:32:44,023 INFO [finetune.py:976] (1/7) Epoch 14, batch 2200, loss[loss=0.1971, simple_loss=0.2597, pruned_loss=0.0673, over 4747.00 frames. ], tot_loss[loss=0.1824, simple_loss=0.252, pruned_loss=0.05642, over 954810.78 frames. ], batch size: 54, lr: 3.55e-03, grad_scale: 64.0 2023-04-27 05:33:06,756 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=76679.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 05:33:08,594 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.3058, 1.5578, 1.3921, 1.5185, 1.2638, 1.2907, 1.3539, 1.0459], device='cuda:1'), covar=tensor([0.1574, 0.1182, 0.0955, 0.1108, 0.3270, 0.1106, 0.1621, 0.2098], device='cuda:1'), in_proj_covar=tensor([0.0294, 0.0313, 0.0227, 0.0287, 0.0314, 0.0268, 0.0257, 0.0273], device='cuda:1'), out_proj_covar=tensor([1.1851e-04, 1.2481e-04, 9.0739e-05, 1.1438e-04, 1.2814e-04, 1.0706e-04, 1.0436e-04, 1.0903e-04], device='cuda:1') 2023-04-27 05:33:46,580 INFO [finetune.py:976] (1/7) Epoch 14, batch 2250, loss[loss=0.1739, simple_loss=0.2308, pruned_loss=0.05854, over 4041.00 frames. ], tot_loss[loss=0.1838, simple_loss=0.2534, pruned_loss=0.05707, over 954880.01 frames. ], batch size: 17, lr: 3.55e-03, grad_scale: 64.0 2023-04-27 05:34:01,055 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7605, 2.4612, 1.8675, 1.8055, 1.2986, 1.2922, 1.9568, 1.2995], device='cuda:1'), covar=tensor([0.1808, 0.1555, 0.1589, 0.1969, 0.2631, 0.2146, 0.1135, 0.2288], device='cuda:1'), in_proj_covar=tensor([0.0196, 0.0214, 0.0169, 0.0205, 0.0202, 0.0184, 0.0157, 0.0187], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-27 05:34:30,285 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.017e+02 1.594e+02 1.898e+02 2.176e+02 4.652e+02, threshold=3.795e+02, percent-clipped=1.0 2023-04-27 05:34:47,190 INFO [finetune.py:976] (1/7) Epoch 14, batch 2300, loss[loss=0.1705, simple_loss=0.2272, pruned_loss=0.05694, over 4018.00 frames. ], tot_loss[loss=0.1845, simple_loss=0.2543, pruned_loss=0.05735, over 955751.60 frames. ], batch size: 17, lr: 3.55e-03, grad_scale: 64.0 2023-04-27 05:35:26,138 INFO [finetune.py:976] (1/7) Epoch 14, batch 2350, loss[loss=0.201, simple_loss=0.2657, pruned_loss=0.06812, over 4894.00 frames. ], tot_loss[loss=0.1818, simple_loss=0.2511, pruned_loss=0.0563, over 954680.83 frames. ], batch size: 32, lr: 3.55e-03, grad_scale: 64.0 2023-04-27 05:36:05,992 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=76839.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 05:36:09,505 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.125e+02 1.656e+02 1.888e+02 2.270e+02 4.545e+02, threshold=3.776e+02, percent-clipped=2.0 2023-04-27 05:36:20,258 INFO [finetune.py:976] (1/7) Epoch 14, batch 2400, loss[loss=0.1677, simple_loss=0.2346, pruned_loss=0.05046, over 4798.00 frames. ], tot_loss[loss=0.1792, simple_loss=0.2482, pruned_loss=0.05515, over 957433.29 frames. ], batch size: 29, lr: 3.55e-03, grad_scale: 64.0 2023-04-27 05:36:46,488 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=76900.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 05:36:54,081 INFO [finetune.py:976] (1/7) Epoch 14, batch 2450, loss[loss=0.1948, simple_loss=0.2564, pruned_loss=0.06661, over 4753.00 frames. ], tot_loss[loss=0.1763, simple_loss=0.2447, pruned_loss=0.05398, over 954939.78 frames. ], batch size: 59, lr: 3.55e-03, grad_scale: 64.0 2023-04-27 05:36:56,574 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5781, 1.3728, 4.3811, 4.1007, 3.8360, 4.1609, 4.0619, 3.8012], device='cuda:1'), covar=tensor([0.6903, 0.5710, 0.0993, 0.1757, 0.1065, 0.1505, 0.1258, 0.1544], device='cuda:1'), in_proj_covar=tensor([0.0305, 0.0302, 0.0397, 0.0398, 0.0341, 0.0398, 0.0310, 0.0359], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-27 05:37:04,096 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=76925.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 05:37:17,025 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.085e+02 1.514e+02 1.972e+02 2.483e+02 3.800e+02, threshold=3.944e+02, percent-clipped=1.0 2023-04-27 05:37:20,136 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=76950.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 05:37:27,666 INFO [finetune.py:976] (1/7) Epoch 14, batch 2500, loss[loss=0.2035, simple_loss=0.271, pruned_loss=0.068, over 4942.00 frames. ], tot_loss[loss=0.1796, simple_loss=0.248, pruned_loss=0.05563, over 954092.26 frames. ], batch size: 33, lr: 3.55e-03, grad_scale: 64.0 2023-04-27 05:37:36,004 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=76973.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 05:37:40,693 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=76979.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 05:37:52,140 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.2113, 2.1595, 1.7087, 1.8132, 2.2087, 1.8449, 2.7680, 1.5489], device='cuda:1'), covar=tensor([0.3870, 0.2003, 0.4906, 0.3216, 0.2026, 0.2596, 0.1607, 0.4734], device='cuda:1'), in_proj_covar=tensor([0.0341, 0.0345, 0.0425, 0.0356, 0.0380, 0.0381, 0.0369, 0.0418], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 05:37:52,649 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=76998.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 05:38:01,548 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5764, 3.3278, 1.0590, 1.8132, 1.9979, 2.3356, 1.9689, 1.0564], device='cuda:1'), covar=tensor([0.1538, 0.1542, 0.1976, 0.1416, 0.1090, 0.1232, 0.1560, 0.1955], device='cuda:1'), in_proj_covar=tensor([0.0118, 0.0245, 0.0139, 0.0121, 0.0132, 0.0153, 0.0118, 0.0120], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-27 05:38:06,234 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5861, 1.7995, 0.7937, 1.3220, 1.8339, 1.4241, 1.3731, 1.4215], device='cuda:1'), covar=tensor([0.0518, 0.0349, 0.0380, 0.0582, 0.0275, 0.0520, 0.0507, 0.0565], device='cuda:1'), in_proj_covar=tensor([0.0029, 0.0025, 0.0023, 0.0029, 0.0020, 0.0029, 0.0028, 0.0029], device='cuda:1'), out_proj_covar=tensor([0.0050, 0.0045, 0.0038, 0.0051, 0.0038, 0.0049, 0.0049, 0.0050], device='cuda:1') 2023-04-27 05:38:07,207 INFO [finetune.py:976] (1/7) Epoch 14, batch 2550, loss[loss=0.194, simple_loss=0.261, pruned_loss=0.06347, over 4813.00 frames. ], tot_loss[loss=0.182, simple_loss=0.2511, pruned_loss=0.05647, over 955008.27 frames. ], batch size: 51, lr: 3.55e-03, grad_scale: 64.0 2023-04-27 05:38:13,569 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-04-27 05:38:18,428 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=77027.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 05:38:30,330 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.907e+01 1.587e+02 1.892e+02 2.346e+02 6.910e+02, threshold=3.784e+02, percent-clipped=5.0 2023-04-27 05:38:40,001 INFO [finetune.py:976] (1/7) Epoch 14, batch 2600, loss[loss=0.2149, simple_loss=0.282, pruned_loss=0.07392, over 4826.00 frames. ], tot_loss[loss=0.1831, simple_loss=0.2528, pruned_loss=0.05677, over 957553.78 frames. ], batch size: 39, lr: 3.55e-03, grad_scale: 64.0 2023-04-27 05:38:41,347 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0323, 3.1071, 2.8486, 2.9166, 3.2422, 2.9871, 3.9640, 2.6254], device='cuda:1'), covar=tensor([0.3412, 0.1630, 0.2808, 0.2746, 0.1646, 0.2120, 0.1465, 0.3200], device='cuda:1'), in_proj_covar=tensor([0.0345, 0.0350, 0.0428, 0.0360, 0.0385, 0.0385, 0.0374, 0.0422], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 05:39:20,529 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=77095.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 05:39:41,102 INFO [finetune.py:976] (1/7) Epoch 14, batch 2650, loss[loss=0.2028, simple_loss=0.2836, pruned_loss=0.06101, over 4814.00 frames. ], tot_loss[loss=0.1846, simple_loss=0.2544, pruned_loss=0.05738, over 956560.36 frames. ], batch size: 38, lr: 3.54e-03, grad_scale: 64.0 2023-04-27 05:40:18,328 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9393, 2.3708, 2.0284, 2.3328, 1.6243, 2.0655, 1.8691, 1.6385], device='cuda:1'), covar=tensor([0.1923, 0.1005, 0.0770, 0.1015, 0.3014, 0.0923, 0.1911, 0.2272], device='cuda:1'), in_proj_covar=tensor([0.0292, 0.0311, 0.0226, 0.0285, 0.0314, 0.0266, 0.0255, 0.0272], device='cuda:1'), out_proj_covar=tensor([1.1772e-04, 1.2386e-04, 9.0105e-05, 1.1355e-04, 1.2782e-04, 1.0625e-04, 1.0341e-04, 1.0845e-04], device='cuda:1') 2023-04-27 05:40:19,432 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.209e+02 1.638e+02 1.948e+02 2.279e+02 4.007e+02, threshold=3.896e+02, percent-clipped=1.0 2023-04-27 05:40:26,271 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=77156.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 05:40:29,225 INFO [finetune.py:976] (1/7) Epoch 14, batch 2700, loss[loss=0.2125, simple_loss=0.2878, pruned_loss=0.06858, over 4765.00 frames. ], tot_loss[loss=0.1838, simple_loss=0.2538, pruned_loss=0.0569, over 956070.12 frames. ], batch size: 54, lr: 3.54e-03, grad_scale: 64.0 2023-04-27 05:40:53,254 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=77195.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 05:41:03,078 INFO [finetune.py:976] (1/7) Epoch 14, batch 2750, loss[loss=0.2006, simple_loss=0.2696, pruned_loss=0.06583, over 4836.00 frames. ], tot_loss[loss=0.1806, simple_loss=0.2503, pruned_loss=0.05543, over 957716.27 frames. ], batch size: 49, lr: 3.54e-03, grad_scale: 64.0 2023-04-27 05:41:03,183 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=77211.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 05:41:06,309 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5912, 0.9607, 1.6795, 2.0588, 1.7026, 1.5650, 1.6718, 1.5993], device='cuda:1'), covar=tensor([0.4711, 0.6613, 0.6043, 0.6325, 0.6011, 0.7439, 0.7514, 0.8010], device='cuda:1'), in_proj_covar=tensor([0.0414, 0.0408, 0.0496, 0.0512, 0.0444, 0.0465, 0.0472, 0.0475], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 05:41:09,316 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7446, 1.6031, 1.7798, 2.0493, 2.0908, 1.8012, 1.3020, 1.8616], device='cuda:1'), covar=tensor([0.0829, 0.1232, 0.0778, 0.0586, 0.0612, 0.0732, 0.0784, 0.0570], device='cuda:1'), in_proj_covar=tensor([0.0191, 0.0202, 0.0182, 0.0172, 0.0178, 0.0181, 0.0154, 0.0180], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 05:41:26,803 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9931, 2.2577, 1.2027, 1.6874, 2.1345, 1.9150, 1.7758, 1.8356], device='cuda:1'), covar=tensor([0.0452, 0.0336, 0.0285, 0.0513, 0.0240, 0.0463, 0.0485, 0.0515], device='cuda:1'), in_proj_covar=tensor([0.0029, 0.0025, 0.0023, 0.0029, 0.0020, 0.0029, 0.0028, 0.0029], device='cuda:1'), out_proj_covar=tensor([0.0050, 0.0045, 0.0038, 0.0051, 0.0038, 0.0049, 0.0049, 0.0050], device='cuda:1') 2023-04-27 05:41:37,789 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.161e+02 1.631e+02 1.907e+02 2.391e+02 4.996e+02, threshold=3.813e+02, percent-clipped=2.0 2023-04-27 05:41:38,578 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-04-27 05:41:48,207 INFO [finetune.py:976] (1/7) Epoch 14, batch 2800, loss[loss=0.1488, simple_loss=0.2155, pruned_loss=0.04112, over 4830.00 frames. ], tot_loss[loss=0.1779, simple_loss=0.2469, pruned_loss=0.05445, over 956690.21 frames. ], batch size: 49, lr: 3.54e-03, grad_scale: 64.0 2023-04-27 05:41:55,160 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=77272.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 05:41:55,749 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=77273.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 05:41:58,037 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6045, 2.6108, 2.1133, 2.3511, 2.7264, 2.1474, 3.5470, 1.8943], device='cuda:1'), covar=tensor([0.4001, 0.2060, 0.4517, 0.3368, 0.1991, 0.3052, 0.1276, 0.4358], device='cuda:1'), in_proj_covar=tensor([0.0342, 0.0348, 0.0427, 0.0356, 0.0383, 0.0385, 0.0372, 0.0421], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 05:41:58,628 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=77277.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 05:42:01,076 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 2023-04-27 05:42:22,267 INFO [finetune.py:976] (1/7) Epoch 14, batch 2850, loss[loss=0.2243, simple_loss=0.304, pruned_loss=0.07231, over 4835.00 frames. ], tot_loss[loss=0.1784, simple_loss=0.2466, pruned_loss=0.05515, over 954591.08 frames. ], batch size: 47, lr: 3.54e-03, grad_scale: 64.0 2023-04-27 05:42:26,642 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4953, 1.3184, 4.4412, 4.1382, 3.8660, 4.1722, 4.0671, 3.8831], device='cuda:1'), covar=tensor([0.7098, 0.6109, 0.0961, 0.1742, 0.1163, 0.1857, 0.1275, 0.1462], device='cuda:1'), in_proj_covar=tensor([0.0311, 0.0307, 0.0403, 0.0407, 0.0346, 0.0405, 0.0315, 0.0365], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-27 05:42:37,447 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=77334.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 05:42:39,844 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=77338.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 05:42:44,407 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.158e+02 1.666e+02 1.986e+02 2.404e+02 4.400e+02, threshold=3.972e+02, percent-clipped=2.0 2023-04-27 05:42:55,656 INFO [finetune.py:976] (1/7) Epoch 14, batch 2900, loss[loss=0.1693, simple_loss=0.2454, pruned_loss=0.04657, over 4778.00 frames. ], tot_loss[loss=0.1827, simple_loss=0.2511, pruned_loss=0.05718, over 956285.09 frames. ], batch size: 28, lr: 3.54e-03, grad_scale: 64.0 2023-04-27 05:43:29,374 INFO [finetune.py:976] (1/7) Epoch 14, batch 2950, loss[loss=0.206, simple_loss=0.2819, pruned_loss=0.06502, over 4747.00 frames. ], tot_loss[loss=0.1853, simple_loss=0.2545, pruned_loss=0.05802, over 957876.43 frames. ], batch size: 54, lr: 3.54e-03, grad_scale: 64.0 2023-04-27 05:43:50,926 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.161e+02 1.615e+02 1.939e+02 2.289e+02 5.440e+02, threshold=3.878e+02, percent-clipped=2.0 2023-04-27 05:43:55,609 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=77451.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 05:44:03,067 INFO [finetune.py:976] (1/7) Epoch 14, batch 3000, loss[loss=0.169, simple_loss=0.2525, pruned_loss=0.04279, over 4815.00 frames. ], tot_loss[loss=0.1863, simple_loss=0.2556, pruned_loss=0.05849, over 956098.78 frames. ], batch size: 33, lr: 3.54e-03, grad_scale: 64.0 2023-04-27 05:44:03,067 INFO [finetune.py:1001] (1/7) Computing validation loss 2023-04-27 05:44:19,534 INFO [finetune.py:1010] (1/7) Epoch 14, validation: loss=0.1527, simple_loss=0.224, pruned_loss=0.04073, over 2265189.00 frames. 2023-04-27 05:44:19,535 INFO [finetune.py:1011] (1/7) Maximum memory allocated so far is 6391MB 2023-04-27 05:44:28,804 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=3.53 vs. limit=5.0 2023-04-27 05:45:02,919 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=77495.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 05:45:24,239 INFO [finetune.py:976] (1/7) Epoch 14, batch 3050, loss[loss=0.1899, simple_loss=0.2479, pruned_loss=0.06592, over 4067.00 frames. ], tot_loss[loss=0.1855, simple_loss=0.255, pruned_loss=0.05804, over 955440.90 frames. ], batch size: 65, lr: 3.54e-03, grad_scale: 64.0 2023-04-27 05:45:45,783 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=77543.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 05:45:46,941 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.289e+02 1.813e+02 2.136e+02 2.467e+02 4.229e+02, threshold=4.273e+02, percent-clipped=1.0 2023-04-27 05:45:57,727 INFO [finetune.py:976] (1/7) Epoch 14, batch 3100, loss[loss=0.146, simple_loss=0.2216, pruned_loss=0.03523, over 4882.00 frames. ], tot_loss[loss=0.1828, simple_loss=0.2525, pruned_loss=0.05659, over 955681.12 frames. ], batch size: 43, lr: 3.54e-03, grad_scale: 64.0 2023-04-27 05:46:01,907 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=77567.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 05:46:10,770 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2023-04-27 05:46:36,287 INFO [finetune.py:976] (1/7) Epoch 14, batch 3150, loss[loss=0.1259, simple_loss=0.1982, pruned_loss=0.02682, over 4819.00 frames. ], tot_loss[loss=0.1829, simple_loss=0.2513, pruned_loss=0.05721, over 955893.17 frames. ], batch size: 39, lr: 3.54e-03, grad_scale: 32.0 2023-04-27 05:46:44,743 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=77616.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 05:47:05,389 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=77629.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 05:47:07,855 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=77633.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 05:47:20,986 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 8.865e+01 1.595e+02 1.944e+02 2.388e+02 4.637e+02, threshold=3.889e+02, percent-clipped=1.0 2023-04-27 05:47:41,184 INFO [finetune.py:976] (1/7) Epoch 14, batch 3200, loss[loss=0.2265, simple_loss=0.2987, pruned_loss=0.0772, over 4153.00 frames. ], tot_loss[loss=0.1802, simple_loss=0.248, pruned_loss=0.05617, over 955634.59 frames. ], batch size: 66, lr: 3.54e-03, grad_scale: 32.0 2023-04-27 05:47:53,140 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.1169, 1.5641, 1.5434, 2.0468, 2.1949, 1.8572, 1.7850, 1.5612], device='cuda:1'), covar=tensor([0.1897, 0.1781, 0.1858, 0.1719, 0.1275, 0.1850, 0.2429, 0.2092], device='cuda:1'), in_proj_covar=tensor([0.0308, 0.0317, 0.0352, 0.0291, 0.0329, 0.0316, 0.0305, 0.0362], device='cuda:1'), out_proj_covar=tensor([6.4076e-05, 6.6553e-05, 7.5318e-05, 5.9374e-05, 6.8477e-05, 6.7004e-05, 6.4605e-05, 7.7227e-05], device='cuda:1') 2023-04-27 05:48:05,081 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=77677.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 05:48:05,091 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=77677.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 05:48:45,407 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5717, 2.7561, 2.3213, 2.5151, 2.9379, 2.5203, 3.6970, 2.1624], device='cuda:1'), covar=tensor([0.4326, 0.2292, 0.4792, 0.3628, 0.1892, 0.2757, 0.1618, 0.4312], device='cuda:1'), in_proj_covar=tensor([0.0342, 0.0346, 0.0427, 0.0354, 0.0381, 0.0383, 0.0372, 0.0419], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 05:48:48,230 INFO [finetune.py:976] (1/7) Epoch 14, batch 3250, loss[loss=0.1752, simple_loss=0.2507, pruned_loss=0.04981, over 4699.00 frames. ], tot_loss[loss=0.1795, simple_loss=0.2474, pruned_loss=0.05581, over 956206.34 frames. ], batch size: 23, lr: 3.54e-03, grad_scale: 32.0 2023-04-27 05:49:21,544 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5657, 0.7027, 1.3841, 1.8444, 1.6316, 1.4660, 1.4498, 1.4871], device='cuda:1'), covar=tensor([0.4851, 0.6782, 0.6403, 0.6335, 0.6387, 0.7437, 0.7473, 0.6732], device='cuda:1'), in_proj_covar=tensor([0.0412, 0.0406, 0.0494, 0.0510, 0.0443, 0.0464, 0.0471, 0.0472], device='cuda:1'), out_proj_covar=tensor([9.9840e-05, 1.0063e-04, 1.1116e-04, 1.2106e-04, 1.0671e-04, 1.1182e-04, 1.1231e-04, 1.1258e-04], device='cuda:1') 2023-04-27 05:49:28,786 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=77738.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 05:49:33,561 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.092e+02 1.660e+02 2.088e+02 2.486e+02 4.990e+02, threshold=4.175e+02, percent-clipped=6.0 2023-04-27 05:49:36,700 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=77751.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 05:49:42,678 INFO [finetune.py:976] (1/7) Epoch 14, batch 3300, loss[loss=0.1823, simple_loss=0.2511, pruned_loss=0.0568, over 4757.00 frames. ], tot_loss[loss=0.1821, simple_loss=0.2507, pruned_loss=0.05676, over 956927.42 frames. ], batch size: 27, lr: 3.54e-03, grad_scale: 32.0 2023-04-27 05:50:05,478 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.2187, 1.5236, 1.6757, 1.7762, 1.7466, 1.8465, 1.7751, 1.7338], device='cuda:1'), covar=tensor([0.3909, 0.5273, 0.4917, 0.4894, 0.5302, 0.7765, 0.4806, 0.5082], device='cuda:1'), in_proj_covar=tensor([0.0329, 0.0373, 0.0315, 0.0328, 0.0340, 0.0397, 0.0352, 0.0323], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 05:50:07,211 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6008, 3.6158, 0.9543, 2.0630, 2.0685, 2.4825, 1.9916, 1.0686], device='cuda:1'), covar=tensor([0.1420, 0.0814, 0.2048, 0.1113, 0.1058, 0.1059, 0.1449, 0.1988], device='cuda:1'), in_proj_covar=tensor([0.0118, 0.0246, 0.0139, 0.0121, 0.0132, 0.0154, 0.0118, 0.0121], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-27 05:50:08,404 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=77799.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 05:50:15,837 INFO [finetune.py:976] (1/7) Epoch 14, batch 3350, loss[loss=0.1688, simple_loss=0.2335, pruned_loss=0.05202, over 4764.00 frames. ], tot_loss[loss=0.1836, simple_loss=0.2527, pruned_loss=0.05727, over 956499.36 frames. ], batch size: 23, lr: 3.54e-03, grad_scale: 32.0 2023-04-27 05:50:39,906 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.052e+02 1.645e+02 1.956e+02 2.306e+02 5.075e+02, threshold=3.912e+02, percent-clipped=1.0 2023-04-27 05:50:49,108 INFO [finetune.py:976] (1/7) Epoch 14, batch 3400, loss[loss=0.1962, simple_loss=0.2743, pruned_loss=0.05906, over 4795.00 frames. ], tot_loss[loss=0.1864, simple_loss=0.2556, pruned_loss=0.05854, over 956087.98 frames. ], batch size: 25, lr: 3.54e-03, grad_scale: 32.0 2023-04-27 05:50:52,845 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=77867.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 05:51:01,207 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5259, 1.3192, 4.5687, 4.2686, 4.0464, 4.4136, 4.2289, 4.0713], device='cuda:1'), covar=tensor([0.7043, 0.6116, 0.0923, 0.1623, 0.1040, 0.1495, 0.1236, 0.1489], device='cuda:1'), in_proj_covar=tensor([0.0314, 0.0310, 0.0407, 0.0410, 0.0350, 0.0409, 0.0317, 0.0369], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-27 05:51:10,414 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.96 vs. limit=2.0 2023-04-27 05:51:11,672 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-04-27 05:51:18,906 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0295, 2.3527, 0.8392, 1.2962, 1.6800, 1.2096, 2.5393, 1.4158], device='cuda:1'), covar=tensor([0.0639, 0.0621, 0.0651, 0.1237, 0.0449, 0.1027, 0.0329, 0.0701], device='cuda:1'), in_proj_covar=tensor([0.0051, 0.0066, 0.0049, 0.0047, 0.0050, 0.0052, 0.0076, 0.0051], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:1') 2023-04-27 05:51:22,436 INFO [finetune.py:976] (1/7) Epoch 14, batch 3450, loss[loss=0.1876, simple_loss=0.2597, pruned_loss=0.05778, over 4807.00 frames. ], tot_loss[loss=0.1858, simple_loss=0.255, pruned_loss=0.05831, over 955004.34 frames. ], batch size: 41, lr: 3.54e-03, grad_scale: 32.0 2023-04-27 05:51:24,870 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=77915.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 05:51:27,964 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6859, 1.8839, 1.1271, 1.4054, 2.0654, 1.5847, 1.5114, 1.5774], device='cuda:1'), covar=tensor([0.0501, 0.0348, 0.0310, 0.0533, 0.0255, 0.0514, 0.0477, 0.0561], device='cuda:1'), in_proj_covar=tensor([0.0029, 0.0025, 0.0023, 0.0029, 0.0020, 0.0029, 0.0028, 0.0029], device='cuda:1'), out_proj_covar=tensor([0.0050, 0.0045, 0.0038, 0.0051, 0.0038, 0.0049, 0.0049, 0.0050], device='cuda:1') 2023-04-27 05:51:33,834 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=77929.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 05:51:36,786 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=77933.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 05:51:36,846 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0026, 1.5963, 1.8412, 1.7122, 1.8374, 1.5320, 0.8007, 1.4653], device='cuda:1'), covar=tensor([0.3632, 0.3392, 0.1808, 0.2461, 0.2647, 0.2754, 0.4380, 0.2276], device='cuda:1'), in_proj_covar=tensor([0.0288, 0.0247, 0.0223, 0.0316, 0.0215, 0.0230, 0.0228, 0.0185], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-27 05:51:43,293 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 2023-04-27 05:51:45,494 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.078e+02 1.602e+02 1.891e+02 2.396e+02 5.235e+02, threshold=3.783e+02, percent-clipped=1.0 2023-04-27 05:51:54,676 INFO [finetune.py:976] (1/7) Epoch 14, batch 3500, loss[loss=0.2024, simple_loss=0.2695, pruned_loss=0.06764, over 4811.00 frames. ], tot_loss[loss=0.1832, simple_loss=0.2521, pruned_loss=0.05713, over 955601.43 frames. ], batch size: 39, lr: 3.54e-03, grad_scale: 32.0 2023-04-27 05:52:01,406 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=77972.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 05:52:04,452 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=77977.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 05:52:07,432 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=77981.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 05:52:24,827 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6102, 0.9783, 1.5680, 1.9469, 1.6323, 1.5124, 1.5360, 1.6546], device='cuda:1'), covar=tensor([0.6005, 0.8211, 0.7989, 0.8826, 0.7710, 0.9704, 0.9474, 0.9999], device='cuda:1'), in_proj_covar=tensor([0.0413, 0.0406, 0.0495, 0.0509, 0.0443, 0.0464, 0.0471, 0.0473], device='cuda:1'), out_proj_covar=tensor([9.9872e-05, 1.0047e-04, 1.1124e-04, 1.2102e-04, 1.0675e-04, 1.1181e-04, 1.1250e-04, 1.1272e-04], device='cuda:1') 2023-04-27 05:52:28,917 INFO [finetune.py:976] (1/7) Epoch 14, batch 3550, loss[loss=0.1953, simple_loss=0.263, pruned_loss=0.06381, over 4834.00 frames. ], tot_loss[loss=0.18, simple_loss=0.2487, pruned_loss=0.05563, over 955252.33 frames. ], batch size: 33, lr: 3.54e-03, grad_scale: 32.0 2023-04-27 05:52:40,029 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.2674, 1.6113, 1.5133, 1.7172, 1.7618, 1.9688, 1.4337, 3.5767], device='cuda:1'), covar=tensor([0.0578, 0.0757, 0.0785, 0.1171, 0.0614, 0.0523, 0.0741, 0.0146], device='cuda:1'), in_proj_covar=tensor([0.0038, 0.0039, 0.0040, 0.0044, 0.0040, 0.0038, 0.0038, 0.0057], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:1') 2023-04-27 05:52:42,973 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=78033.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 05:52:56,888 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.150e+02 1.685e+02 1.967e+02 2.304e+02 5.361e+02, threshold=3.934e+02, percent-clipped=3.0 2023-04-27 05:53:17,516 INFO [finetune.py:976] (1/7) Epoch 14, batch 3600, loss[loss=0.1711, simple_loss=0.2297, pruned_loss=0.05628, over 4741.00 frames. ], tot_loss[loss=0.178, simple_loss=0.246, pruned_loss=0.05496, over 954372.36 frames. ], batch size: 26, lr: 3.54e-03, grad_scale: 32.0 2023-04-27 05:53:29,008 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.4055, 2.0415, 2.5039, 2.7367, 2.8518, 2.3905, 1.8783, 2.3508], device='cuda:1'), covar=tensor([0.0879, 0.1107, 0.0634, 0.0613, 0.0589, 0.0723, 0.0760, 0.0640], device='cuda:1'), in_proj_covar=tensor([0.0192, 0.0203, 0.0183, 0.0173, 0.0179, 0.0183, 0.0154, 0.0181], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 05:53:29,052 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9647, 1.7003, 2.1832, 2.4025, 1.9651, 1.8289, 1.9935, 2.0006], device='cuda:1'), covar=tensor([0.5120, 0.7976, 0.8146, 0.6233, 0.6956, 0.9890, 0.9474, 0.9939], device='cuda:1'), in_proj_covar=tensor([0.0414, 0.0407, 0.0495, 0.0510, 0.0444, 0.0465, 0.0472, 0.0475], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 05:53:30,833 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7418, 1.9089, 0.9837, 1.4320, 2.0262, 1.5616, 1.4711, 1.5706], device='cuda:1'), covar=tensor([0.0535, 0.0386, 0.0370, 0.0599, 0.0264, 0.0547, 0.0556, 0.0606], device='cuda:1'), in_proj_covar=tensor([0.0029, 0.0025, 0.0023, 0.0029, 0.0020, 0.0029, 0.0028, 0.0029], device='cuda:1'), out_proj_covar=tensor([0.0050, 0.0045, 0.0038, 0.0051, 0.0038, 0.0049, 0.0049, 0.0050], device='cuda:1') 2023-04-27 05:53:53,139 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.1796, 1.4323, 1.3679, 1.6368, 1.5149, 1.7761, 1.3074, 2.9260], device='cuda:1'), covar=tensor([0.0603, 0.0746, 0.0751, 0.1111, 0.0612, 0.0536, 0.0757, 0.0203], device='cuda:1'), in_proj_covar=tensor([0.0038, 0.0039, 0.0040, 0.0043, 0.0040, 0.0038, 0.0038, 0.0057], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0014, 0.0013, 0.0012, 0.0013, 0.0016], device='cuda:1') 2023-04-27 05:54:18,583 INFO [finetune.py:976] (1/7) Epoch 14, batch 3650, loss[loss=0.2001, simple_loss=0.2727, pruned_loss=0.06378, over 4809.00 frames. ], tot_loss[loss=0.1799, simple_loss=0.2482, pruned_loss=0.05581, over 954621.41 frames. ], batch size: 41, lr: 3.54e-03, grad_scale: 32.0 2023-04-27 05:54:56,789 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.146e+02 1.566e+02 1.842e+02 2.272e+02 4.887e+02, threshold=3.683e+02, percent-clipped=1.0 2023-04-27 05:54:59,370 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-04-27 05:55:09,389 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.1029, 1.5933, 1.5148, 2.0265, 2.2525, 1.8205, 1.7590, 1.5001], device='cuda:1'), covar=tensor([0.2229, 0.1739, 0.1938, 0.1651, 0.1207, 0.2024, 0.2471, 0.2437], device='cuda:1'), in_proj_covar=tensor([0.0306, 0.0316, 0.0351, 0.0291, 0.0328, 0.0315, 0.0303, 0.0360], device='cuda:1'), out_proj_covar=tensor([6.3605e-05, 6.6295e-05, 7.5333e-05, 5.9430e-05, 6.8340e-05, 6.6708e-05, 6.4168e-05, 7.6974e-05], device='cuda:1') 2023-04-27 05:55:18,204 INFO [finetune.py:976] (1/7) Epoch 14, batch 3700, loss[loss=0.1899, simple_loss=0.2583, pruned_loss=0.06068, over 4891.00 frames. ], tot_loss[loss=0.1815, simple_loss=0.251, pruned_loss=0.05597, over 954948.48 frames. ], batch size: 35, lr: 3.54e-03, grad_scale: 32.0 2023-04-27 05:55:22,009 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6699, 1.4103, 1.3040, 1.4169, 1.8584, 1.4463, 1.2127, 1.2119], device='cuda:1'), covar=tensor([0.1816, 0.1266, 0.1766, 0.1446, 0.0768, 0.1766, 0.2332, 0.2209], device='cuda:1'), in_proj_covar=tensor([0.0305, 0.0315, 0.0350, 0.0290, 0.0327, 0.0314, 0.0302, 0.0359], device='cuda:1'), out_proj_covar=tensor([6.3396e-05, 6.6064e-05, 7.5108e-05, 5.9229e-05, 6.8088e-05, 6.6537e-05, 6.3986e-05, 7.6749e-05], device='cuda:1') 2023-04-27 05:55:56,841 INFO [finetune.py:976] (1/7) Epoch 14, batch 3750, loss[loss=0.2033, simple_loss=0.2684, pruned_loss=0.06908, over 4921.00 frames. ], tot_loss[loss=0.1824, simple_loss=0.2522, pruned_loss=0.05627, over 956364.45 frames. ], batch size: 33, lr: 3.54e-03, grad_scale: 32.0 2023-04-27 05:56:01,886 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.8244, 2.1223, 2.0538, 2.1995, 1.8987, 2.0710, 2.1083, 2.0177], device='cuda:1'), covar=tensor([0.4264, 0.6581, 0.5262, 0.4998, 0.6160, 0.7845, 0.6427, 0.5893], device='cuda:1'), in_proj_covar=tensor([0.0332, 0.0376, 0.0319, 0.0331, 0.0342, 0.0399, 0.0355, 0.0326], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 05:56:09,782 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=78232.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 05:56:18,615 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.305e+02 1.679e+02 1.977e+02 2.535e+02 5.677e+02, threshold=3.954e+02, percent-clipped=2.0 2023-04-27 05:56:30,165 INFO [finetune.py:976] (1/7) Epoch 14, batch 3800, loss[loss=0.2093, simple_loss=0.2859, pruned_loss=0.06632, over 4813.00 frames. ], tot_loss[loss=0.1839, simple_loss=0.2536, pruned_loss=0.05713, over 954417.13 frames. ], batch size: 33, lr: 3.53e-03, grad_scale: 32.0 2023-04-27 05:56:37,456 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=78272.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 05:56:50,270 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=78293.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 05:56:53,326 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-04-27 05:57:03,401 INFO [finetune.py:976] (1/7) Epoch 14, batch 3850, loss[loss=0.1801, simple_loss=0.2511, pruned_loss=0.05455, over 4921.00 frames. ], tot_loss[loss=0.1818, simple_loss=0.2518, pruned_loss=0.05589, over 955280.18 frames. ], batch size: 33, lr: 3.53e-03, grad_scale: 32.0 2023-04-27 05:57:09,828 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=78320.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 05:57:17,832 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=78333.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 05:57:25,786 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.080e+02 1.524e+02 1.884e+02 2.249e+02 3.539e+02, threshold=3.767e+02, percent-clipped=0.0 2023-04-27 05:57:36,828 INFO [finetune.py:976] (1/7) Epoch 14, batch 3900, loss[loss=0.1795, simple_loss=0.249, pruned_loss=0.05505, over 4910.00 frames. ], tot_loss[loss=0.1794, simple_loss=0.2485, pruned_loss=0.05518, over 955090.68 frames. ], batch size: 36, lr: 3.53e-03, grad_scale: 32.0 2023-04-27 05:57:50,413 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=78381.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 05:58:09,805 INFO [finetune.py:976] (1/7) Epoch 14, batch 3950, loss[loss=0.1792, simple_loss=0.2317, pruned_loss=0.06338, over 4864.00 frames. ], tot_loss[loss=0.1777, simple_loss=0.2458, pruned_loss=0.05475, over 955757.58 frames. ], batch size: 44, lr: 3.53e-03, grad_scale: 32.0 2023-04-27 05:58:50,266 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.090e+02 1.573e+02 1.912e+02 2.257e+02 5.641e+02, threshold=3.824e+02, percent-clipped=2.0 2023-04-27 05:58:51,040 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5913, 2.0542, 1.7681, 1.9244, 1.5796, 1.6601, 1.6868, 1.3454], device='cuda:1'), covar=tensor([0.1836, 0.1241, 0.0895, 0.1221, 0.3199, 0.1265, 0.1777, 0.2346], device='cuda:1'), in_proj_covar=tensor([0.0288, 0.0309, 0.0223, 0.0281, 0.0310, 0.0263, 0.0252, 0.0268], device='cuda:1'), out_proj_covar=tensor([1.1625e-04, 1.2275e-04, 8.9029e-05, 1.1206e-04, 1.2656e-04, 1.0534e-04, 1.0203e-04, 1.0686e-04], device='cuda:1') 2023-04-27 05:59:04,626 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=78460.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 05:59:04,812 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.16 vs. limit=5.0 2023-04-27 05:59:05,120 INFO [finetune.py:976] (1/7) Epoch 14, batch 4000, loss[loss=0.1716, simple_loss=0.2418, pruned_loss=0.05066, over 4815.00 frames. ], tot_loss[loss=0.178, simple_loss=0.246, pruned_loss=0.05506, over 954258.71 frames. ], batch size: 25, lr: 3.53e-03, grad_scale: 32.0 2023-04-27 05:59:14,980 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 2023-04-27 05:59:49,592 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.3546, 3.1122, 2.3308, 2.3909, 1.6556, 1.5914, 2.5122, 1.6806], device='cuda:1'), covar=tensor([0.1546, 0.1413, 0.1353, 0.1591, 0.2328, 0.1954, 0.0952, 0.1983], device='cuda:1'), in_proj_covar=tensor([0.0195, 0.0213, 0.0168, 0.0204, 0.0201, 0.0184, 0.0156, 0.0186], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-27 05:59:55,662 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9443, 1.7997, 5.0325, 4.6733, 4.3541, 4.8422, 4.4773, 4.4217], device='cuda:1'), covar=tensor([0.7072, 0.5282, 0.0969, 0.1907, 0.1057, 0.1302, 0.1342, 0.1493], device='cuda:1'), in_proj_covar=tensor([0.0312, 0.0305, 0.0403, 0.0404, 0.0347, 0.0405, 0.0312, 0.0365], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-27 05:59:57,597 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.76 vs. limit=2.0 2023-04-27 06:00:00,327 INFO [finetune.py:976] (1/7) Epoch 14, batch 4050, loss[loss=0.1539, simple_loss=0.2138, pruned_loss=0.04698, over 4035.00 frames. ], tot_loss[loss=0.1814, simple_loss=0.2497, pruned_loss=0.05654, over 953779.48 frames. ], batch size: 17, lr: 3.53e-03, grad_scale: 32.0 2023-04-27 06:00:19,661 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=78521.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:00:51,058 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.238e+02 1.700e+02 2.035e+02 2.375e+02 5.106e+02, threshold=4.071e+02, percent-clipped=4.0 2023-04-27 06:01:03,056 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=78556.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:01:06,003 INFO [finetune.py:976] (1/7) Epoch 14, batch 4100, loss[loss=0.1837, simple_loss=0.2512, pruned_loss=0.05808, over 4926.00 frames. ], tot_loss[loss=0.1829, simple_loss=0.2518, pruned_loss=0.05697, over 953173.83 frames. ], batch size: 33, lr: 3.53e-03, grad_scale: 32.0 2023-04-27 06:01:11,788 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.1658, 1.6546, 1.9846, 2.3945, 2.0609, 1.5756, 1.1861, 1.8229], device='cuda:1'), covar=tensor([0.3378, 0.3332, 0.1821, 0.2404, 0.2468, 0.2717, 0.4492, 0.2178], device='cuda:1'), in_proj_covar=tensor([0.0289, 0.0248, 0.0224, 0.0317, 0.0216, 0.0230, 0.0230, 0.0186], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-27 06:01:45,290 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=78588.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 06:02:05,268 INFO [finetune.py:976] (1/7) Epoch 14, batch 4150, loss[loss=0.1905, simple_loss=0.271, pruned_loss=0.05499, over 4900.00 frames. ], tot_loss[loss=0.184, simple_loss=0.2532, pruned_loss=0.05743, over 952394.93 frames. ], batch size: 37, lr: 3.53e-03, grad_scale: 32.0 2023-04-27 06:02:09,545 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=78617.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:02:27,352 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=3.94 vs. limit=5.0 2023-04-27 06:02:29,493 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.351e+02 1.709e+02 1.960e+02 2.352e+02 3.930e+02, threshold=3.920e+02, percent-clipped=0.0 2023-04-27 06:02:38,717 INFO [finetune.py:976] (1/7) Epoch 14, batch 4200, loss[loss=0.2193, simple_loss=0.2868, pruned_loss=0.07592, over 4827.00 frames. ], tot_loss[loss=0.1839, simple_loss=0.2535, pruned_loss=0.05715, over 953101.51 frames. ], batch size: 49, lr: 3.53e-03, grad_scale: 32.0 2023-04-27 06:03:02,466 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.5082, 1.7910, 1.9031, 2.0244, 1.8461, 1.9784, 1.9622, 1.9366], device='cuda:1'), covar=tensor([0.4153, 0.5787, 0.4964, 0.4763, 0.5930, 0.7829, 0.5990, 0.5229], device='cuda:1'), in_proj_covar=tensor([0.0331, 0.0375, 0.0316, 0.0330, 0.0340, 0.0397, 0.0354, 0.0325], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 06:03:12,003 INFO [finetune.py:976] (1/7) Epoch 14, batch 4250, loss[loss=0.1849, simple_loss=0.2366, pruned_loss=0.06657, over 2917.00 frames. ], tot_loss[loss=0.183, simple_loss=0.2522, pruned_loss=0.05688, over 951156.94 frames. ], batch size: 12, lr: 3.53e-03, grad_scale: 32.0 2023-04-27 06:03:36,179 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.058e+02 1.588e+02 1.905e+02 2.240e+02 4.270e+02, threshold=3.810e+02, percent-clipped=2.0 2023-04-27 06:03:45,440 INFO [finetune.py:976] (1/7) Epoch 14, batch 4300, loss[loss=0.1672, simple_loss=0.2337, pruned_loss=0.05033, over 4750.00 frames. ], tot_loss[loss=0.1802, simple_loss=0.2491, pruned_loss=0.05569, over 950798.78 frames. ], batch size: 26, lr: 3.53e-03, grad_scale: 32.0 2023-04-27 06:04:19,296 INFO [finetune.py:976] (1/7) Epoch 14, batch 4350, loss[loss=0.1536, simple_loss=0.229, pruned_loss=0.03904, over 4817.00 frames. ], tot_loss[loss=0.1768, simple_loss=0.2452, pruned_loss=0.05424, over 950844.93 frames. ], batch size: 39, lr: 3.53e-03, grad_scale: 32.0 2023-04-27 06:04:20,013 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.1001, 1.4949, 1.4522, 1.6984, 1.5760, 1.8158, 1.3589, 3.2564], device='cuda:1'), covar=tensor([0.0658, 0.0821, 0.0789, 0.1272, 0.0667, 0.0495, 0.0767, 0.0153], device='cuda:1'), in_proj_covar=tensor([0.0038, 0.0039, 0.0040, 0.0044, 0.0040, 0.0038, 0.0038, 0.0057], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:1') 2023-04-27 06:04:22,428 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=78816.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:04:43,555 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.082e+02 1.581e+02 1.858e+02 2.218e+02 4.471e+02, threshold=3.716e+02, percent-clipped=1.0 2023-04-27 06:05:04,243 INFO [finetune.py:976] (1/7) Epoch 14, batch 4400, loss[loss=0.2122, simple_loss=0.2902, pruned_loss=0.06712, over 4840.00 frames. ], tot_loss[loss=0.1785, simple_loss=0.2465, pruned_loss=0.05526, over 948907.77 frames. ], batch size: 47, lr: 3.53e-03, grad_scale: 32.0 2023-04-27 06:05:13,177 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5790, 1.4262, 1.6722, 1.9925, 2.0897, 1.5462, 1.1748, 1.7351], device='cuda:1'), covar=tensor([0.0832, 0.1100, 0.0752, 0.0539, 0.0565, 0.0754, 0.0809, 0.0570], device='cuda:1'), in_proj_covar=tensor([0.0191, 0.0202, 0.0182, 0.0172, 0.0178, 0.0183, 0.0154, 0.0180], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 06:05:25,572 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2023-04-27 06:05:38,857 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=78888.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:06:11,507 INFO [finetune.py:976] (1/7) Epoch 14, batch 4450, loss[loss=0.2223, simple_loss=0.283, pruned_loss=0.0808, over 4819.00 frames. ], tot_loss[loss=0.183, simple_loss=0.2516, pruned_loss=0.05717, over 949719.83 frames. ], batch size: 40, lr: 3.53e-03, grad_scale: 32.0 2023-04-27 06:06:12,190 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=78912.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:06:44,170 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=78936.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:07:03,711 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.091e+02 1.730e+02 2.046e+02 2.407e+02 5.260e+02, threshold=4.092e+02, percent-clipped=6.0 2023-04-27 06:07:18,900 INFO [finetune.py:976] (1/7) Epoch 14, batch 4500, loss[loss=0.1641, simple_loss=0.2329, pruned_loss=0.04765, over 4740.00 frames. ], tot_loss[loss=0.1839, simple_loss=0.2528, pruned_loss=0.05746, over 951268.00 frames. ], batch size: 54, lr: 3.53e-03, grad_scale: 32.0 2023-04-27 06:07:21,396 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=78965.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:07:52,194 INFO [finetune.py:976] (1/7) Epoch 14, batch 4550, loss[loss=0.1715, simple_loss=0.239, pruned_loss=0.05206, over 4742.00 frames. ], tot_loss[loss=0.186, simple_loss=0.255, pruned_loss=0.05848, over 952928.85 frames. ], batch size: 59, lr: 3.53e-03, grad_scale: 32.0 2023-04-27 06:08:01,459 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=79026.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:08:14,995 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.614e+01 1.580e+02 1.894e+02 2.383e+02 3.819e+02, threshold=3.787e+02, percent-clipped=0.0 2023-04-27 06:08:26,105 INFO [finetune.py:976] (1/7) Epoch 14, batch 4600, loss[loss=0.2131, simple_loss=0.269, pruned_loss=0.07856, over 4863.00 frames. ], tot_loss[loss=0.1853, simple_loss=0.255, pruned_loss=0.05783, over 953516.20 frames. ], batch size: 34, lr: 3.53e-03, grad_scale: 32.0 2023-04-27 06:08:59,224 INFO [finetune.py:976] (1/7) Epoch 14, batch 4650, loss[loss=0.1358, simple_loss=0.2121, pruned_loss=0.02974, over 4795.00 frames. ], tot_loss[loss=0.1834, simple_loss=0.2525, pruned_loss=0.05713, over 954019.70 frames. ], batch size: 26, lr: 3.53e-03, grad_scale: 32.0 2023-04-27 06:09:02,373 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=79116.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:09:09,120 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.1280, 0.7168, 0.9551, 0.7523, 1.2571, 1.0265, 0.8268, 0.9811], device='cuda:1'), covar=tensor([0.1447, 0.1315, 0.1944, 0.1441, 0.0946, 0.1271, 0.1653, 0.1929], device='cuda:1'), in_proj_covar=tensor([0.0309, 0.0318, 0.0355, 0.0292, 0.0329, 0.0316, 0.0304, 0.0362], device='cuda:1'), out_proj_covar=tensor([6.4175e-05, 6.6596e-05, 7.6267e-05, 5.9709e-05, 6.8398e-05, 6.6846e-05, 6.4251e-05, 7.7389e-05], device='cuda:1') 2023-04-27 06:09:13,510 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=5.00 vs. limit=5.0 2023-04-27 06:09:20,609 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6721, 1.8955, 0.8192, 1.4085, 1.9013, 1.5450, 1.4907, 1.5935], device='cuda:1'), covar=tensor([0.0523, 0.0346, 0.0374, 0.0563, 0.0270, 0.0541, 0.0528, 0.0572], device='cuda:1'), in_proj_covar=tensor([0.0029, 0.0024, 0.0023, 0.0029, 0.0020, 0.0029, 0.0028, 0.0029], device='cuda:1'), out_proj_covar=tensor([0.0050, 0.0044, 0.0038, 0.0050, 0.0037, 0.0049, 0.0048, 0.0050], device='cuda:1') 2023-04-27 06:09:21,599 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.186e+02 1.677e+02 1.942e+02 2.274e+02 5.469e+02, threshold=3.883e+02, percent-clipped=3.0 2023-04-27 06:09:22,948 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.3603, 1.4340, 1.7565, 1.8514, 1.7285, 1.8815, 1.8955, 1.8132], device='cuda:1'), covar=tensor([0.3745, 0.4967, 0.4364, 0.4154, 0.5637, 0.7268, 0.4355, 0.4553], device='cuda:1'), in_proj_covar=tensor([0.0331, 0.0374, 0.0316, 0.0330, 0.0340, 0.0397, 0.0354, 0.0325], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 06:09:32,661 INFO [finetune.py:976] (1/7) Epoch 14, batch 4700, loss[loss=0.1852, simple_loss=0.2564, pruned_loss=0.05704, over 4755.00 frames. ], tot_loss[loss=0.1807, simple_loss=0.2492, pruned_loss=0.05612, over 954195.13 frames. ], batch size: 26, lr: 3.53e-03, grad_scale: 32.0 2023-04-27 06:09:34,546 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=79164.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:10:05,832 INFO [finetune.py:976] (1/7) Epoch 14, batch 4750, loss[loss=0.1579, simple_loss=0.2319, pruned_loss=0.04192, over 4926.00 frames. ], tot_loss[loss=0.1779, simple_loss=0.2462, pruned_loss=0.05482, over 953209.15 frames. ], batch size: 33, lr: 3.53e-03, grad_scale: 32.0 2023-04-27 06:10:06,534 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=79212.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:10:09,415 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.1255, 2.5551, 1.0131, 1.4250, 1.7736, 1.2571, 3.2695, 1.8706], device='cuda:1'), covar=tensor([0.0644, 0.0601, 0.0774, 0.1215, 0.0504, 0.0990, 0.0241, 0.0585], device='cuda:1'), in_proj_covar=tensor([0.0051, 0.0066, 0.0049, 0.0047, 0.0050, 0.0052, 0.0076, 0.0051], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:1') 2023-04-27 06:10:37,980 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.049e+02 1.597e+02 1.966e+02 2.340e+02 3.997e+02, threshold=3.932e+02, percent-clipped=2.0 2023-04-27 06:10:58,673 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=79260.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:10:59,211 INFO [finetune.py:976] (1/7) Epoch 14, batch 4800, loss[loss=0.186, simple_loss=0.259, pruned_loss=0.05646, over 4784.00 frames. ], tot_loss[loss=0.1815, simple_loss=0.2499, pruned_loss=0.05651, over 952365.46 frames. ], batch size: 29, lr: 3.53e-03, grad_scale: 32.0 2023-04-27 06:11:10,708 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.2825, 1.4221, 1.3444, 1.6437, 1.5248, 1.7338, 1.2721, 3.0480], device='cuda:1'), covar=tensor([0.0639, 0.0852, 0.0841, 0.1236, 0.0685, 0.0532, 0.0790, 0.0189], device='cuda:1'), in_proj_covar=tensor([0.0038, 0.0039, 0.0040, 0.0044, 0.0040, 0.0038, 0.0038, 0.0057], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:1') 2023-04-27 06:11:25,277 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=79285.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 06:11:58,655 INFO [finetune.py:976] (1/7) Epoch 14, batch 4850, loss[loss=0.2824, simple_loss=0.3294, pruned_loss=0.1177, over 4821.00 frames. ], tot_loss[loss=0.1856, simple_loss=0.2548, pruned_loss=0.05817, over 954221.45 frames. ], batch size: 40, lr: 3.53e-03, grad_scale: 32.0 2023-04-27 06:12:08,286 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([0.7236, 1.7454, 1.6754, 1.2830, 1.8016, 1.4477, 2.2450, 1.4209], device='cuda:1'), covar=tensor([0.3702, 0.1775, 0.4560, 0.2815, 0.1709, 0.2420, 0.1504, 0.4792], device='cuda:1'), in_proj_covar=tensor([0.0345, 0.0350, 0.0431, 0.0358, 0.0386, 0.0385, 0.0376, 0.0423], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 06:12:10,429 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=79321.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:12:23,698 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=79333.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:12:36,864 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.159e+02 1.635e+02 2.018e+02 2.486e+02 3.725e+02, threshold=4.037e+02, percent-clipped=0.0 2023-04-27 06:12:36,999 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=79346.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 06:12:53,970 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5699, 1.5363, 0.7316, 1.2769, 1.5627, 1.4368, 1.3570, 1.3891], device='cuda:1'), covar=tensor([0.0505, 0.0366, 0.0378, 0.0553, 0.0298, 0.0503, 0.0464, 0.0566], device='cuda:1'), in_proj_covar=tensor([0.0029, 0.0024, 0.0023, 0.0029, 0.0020, 0.0029, 0.0028, 0.0029], device='cuda:1'), out_proj_covar=tensor([0.0050, 0.0045, 0.0038, 0.0050, 0.0038, 0.0049, 0.0048, 0.0050], device='cuda:1') 2023-04-27 06:12:56,921 INFO [finetune.py:976] (1/7) Epoch 14, batch 4900, loss[loss=0.207, simple_loss=0.2847, pruned_loss=0.06468, over 4802.00 frames. ], tot_loss[loss=0.1864, simple_loss=0.2562, pruned_loss=0.05833, over 955459.60 frames. ], batch size: 45, lr: 3.53e-03, grad_scale: 32.0 2023-04-27 06:13:00,242 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=79364.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:13:07,475 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=79375.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 06:13:08,186 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-04-27 06:13:16,770 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.2111, 1.5573, 1.3232, 1.4486, 1.3150, 1.2365, 1.2929, 1.0844], device='cuda:1'), covar=tensor([0.1789, 0.1322, 0.0984, 0.1229, 0.3615, 0.1265, 0.1782, 0.2082], device='cuda:1'), in_proj_covar=tensor([0.0289, 0.0308, 0.0223, 0.0280, 0.0310, 0.0263, 0.0252, 0.0268], device='cuda:1'), out_proj_covar=tensor([1.1635e-04, 1.2281e-04, 8.8757e-05, 1.1164e-04, 1.2623e-04, 1.0497e-04, 1.0203e-04, 1.0690e-04], device='cuda:1') 2023-04-27 06:13:19,221 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=79394.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:13:29,998 INFO [finetune.py:976] (1/7) Epoch 14, batch 4950, loss[loss=0.1897, simple_loss=0.2634, pruned_loss=0.05801, over 4898.00 frames. ], tot_loss[loss=0.1863, simple_loss=0.2564, pruned_loss=0.0581, over 955180.39 frames. ], batch size: 36, lr: 3.52e-03, grad_scale: 32.0 2023-04-27 06:13:40,477 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=79425.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:13:47,624 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=79436.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 06:13:53,546 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.062e+02 1.668e+02 1.946e+02 2.385e+02 4.906e+02, threshold=3.893e+02, percent-clipped=2.0 2023-04-27 06:14:03,213 INFO [finetune.py:976] (1/7) Epoch 14, batch 5000, loss[loss=0.1847, simple_loss=0.2553, pruned_loss=0.0571, over 4793.00 frames. ], tot_loss[loss=0.1841, simple_loss=0.2539, pruned_loss=0.05718, over 953821.22 frames. ], batch size: 51, lr: 3.52e-03, grad_scale: 32.0 2023-04-27 06:14:11,528 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.83 vs. limit=5.0 2023-04-27 06:14:21,139 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8663, 1.4895, 2.1239, 2.4013, 1.9876, 1.9163, 2.0387, 1.9803], device='cuda:1'), covar=tensor([0.4445, 0.5865, 0.5429, 0.5375, 0.5443, 0.7108, 0.6567, 0.6640], device='cuda:1'), in_proj_covar=tensor([0.0417, 0.0409, 0.0497, 0.0512, 0.0447, 0.0468, 0.0474, 0.0478], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 06:14:32,063 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8841, 1.3529, 1.4651, 1.6426, 2.0843, 1.5993, 1.4039, 1.4170], device='cuda:1'), covar=tensor([0.1574, 0.1663, 0.1810, 0.1177, 0.0801, 0.1684, 0.2135, 0.1853], device='cuda:1'), in_proj_covar=tensor([0.0308, 0.0317, 0.0352, 0.0290, 0.0329, 0.0315, 0.0303, 0.0361], device='cuda:1'), out_proj_covar=tensor([6.3898e-05, 6.6410e-05, 7.5605e-05, 5.9303e-05, 6.8408e-05, 6.6727e-05, 6.4165e-05, 7.7077e-05], device='cuda:1') 2023-04-27 06:14:35,550 INFO [finetune.py:976] (1/7) Epoch 14, batch 5050, loss[loss=0.2025, simple_loss=0.2591, pruned_loss=0.07294, over 4901.00 frames. ], tot_loss[loss=0.1819, simple_loss=0.2511, pruned_loss=0.05637, over 954914.28 frames. ], batch size: 32, lr: 3.52e-03, grad_scale: 16.0 2023-04-27 06:14:59,515 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.023e+02 1.519e+02 1.891e+02 2.332e+02 5.445e+02, threshold=3.783e+02, percent-clipped=1.0 2023-04-27 06:15:06,452 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.4038, 1.5996, 1.7643, 1.9246, 1.7730, 1.8878, 1.8464, 1.8010], device='cuda:1'), covar=tensor([0.4111, 0.5945, 0.4731, 0.4793, 0.5583, 0.7813, 0.5678, 0.5457], device='cuda:1'), in_proj_covar=tensor([0.0334, 0.0378, 0.0319, 0.0334, 0.0343, 0.0400, 0.0358, 0.0329], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 06:15:08,088 INFO [finetune.py:976] (1/7) Epoch 14, batch 5100, loss[loss=0.1949, simple_loss=0.2584, pruned_loss=0.06572, over 4904.00 frames. ], tot_loss[loss=0.1783, simple_loss=0.247, pruned_loss=0.05477, over 955535.98 frames. ], batch size: 32, lr: 3.52e-03, grad_scale: 16.0 2023-04-27 06:15:12,444 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-04-27 06:15:25,157 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=79585.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:15:38,536 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9616, 2.2887, 0.8958, 1.2865, 1.6536, 1.2125, 2.5225, 1.5080], device='cuda:1'), covar=tensor([0.0701, 0.0563, 0.0688, 0.1322, 0.0464, 0.1069, 0.0307, 0.0712], device='cuda:1'), in_proj_covar=tensor([0.0051, 0.0066, 0.0049, 0.0047, 0.0050, 0.0052, 0.0076, 0.0051], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:1') 2023-04-27 06:15:41,435 INFO [finetune.py:976] (1/7) Epoch 14, batch 5150, loss[loss=0.1577, simple_loss=0.2345, pruned_loss=0.04041, over 4823.00 frames. ], tot_loss[loss=0.1775, simple_loss=0.2461, pruned_loss=0.05439, over 955103.24 frames. ], batch size: 25, lr: 3.52e-03, grad_scale: 16.0 2023-04-27 06:15:48,510 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=79621.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:16:02,432 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=79641.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 06:16:06,014 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=79646.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:16:06,493 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.050e+02 1.596e+02 1.950e+02 2.248e+02 3.363e+02, threshold=3.899e+02, percent-clipped=0.0 2023-04-27 06:16:12,028 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.2683, 2.9274, 2.3442, 2.6813, 2.0540, 2.4574, 2.5514, 1.8825], device='cuda:1'), covar=tensor([0.2189, 0.1259, 0.0828, 0.1301, 0.3235, 0.1373, 0.2179, 0.2795], device='cuda:1'), in_proj_covar=tensor([0.0289, 0.0309, 0.0223, 0.0281, 0.0311, 0.0263, 0.0253, 0.0268], device='cuda:1'), out_proj_covar=tensor([1.1662e-04, 1.2320e-04, 8.8784e-05, 1.1193e-04, 1.2656e-04, 1.0498e-04, 1.0239e-04, 1.0686e-04], device='cuda:1') 2023-04-27 06:16:12,605 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=79657.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:16:14,970 INFO [finetune.py:976] (1/7) Epoch 14, batch 5200, loss[loss=0.1516, simple_loss=0.2245, pruned_loss=0.03939, over 4770.00 frames. ], tot_loss[loss=0.1796, simple_loss=0.249, pruned_loss=0.05512, over 956630.18 frames. ], batch size: 28, lr: 3.52e-03, grad_scale: 16.0 2023-04-27 06:16:19,850 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=79669.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:16:34,492 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=79689.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:16:54,348 INFO [finetune.py:976] (1/7) Epoch 14, batch 5250, loss[loss=0.1715, simple_loss=0.2384, pruned_loss=0.05229, over 4749.00 frames. ], tot_loss[loss=0.181, simple_loss=0.2508, pruned_loss=0.05563, over 953637.08 frames. ], batch size: 28, lr: 3.52e-03, grad_scale: 16.0 2023-04-27 06:17:03,916 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=79718.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:17:05,049 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=79720.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:17:19,829 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=79731.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 06:17:36,846 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.087e+02 1.641e+02 2.072e+02 2.610e+02 5.253e+02, threshold=4.143e+02, percent-clipped=2.0 2023-04-27 06:17:45,877 INFO [finetune.py:976] (1/7) Epoch 14, batch 5300, loss[loss=0.1497, simple_loss=0.2174, pruned_loss=0.04096, over 4263.00 frames. ], tot_loss[loss=0.1825, simple_loss=0.2523, pruned_loss=0.05639, over 951092.46 frames. ], batch size: 18, lr: 3.52e-03, grad_scale: 16.0 2023-04-27 06:17:46,592 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=79762.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:18:38,535 INFO [finetune.py:976] (1/7) Epoch 14, batch 5350, loss[loss=0.208, simple_loss=0.2714, pruned_loss=0.07233, over 4767.00 frames. ], tot_loss[loss=0.1837, simple_loss=0.2537, pruned_loss=0.05687, over 952110.95 frames. ], batch size: 28, lr: 3.52e-03, grad_scale: 16.0 2023-04-27 06:18:45,990 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=79823.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:19:02,342 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.086e+02 1.530e+02 1.860e+02 2.217e+02 4.388e+02, threshold=3.721e+02, percent-clipped=2.0 2023-04-27 06:19:11,410 INFO [finetune.py:976] (1/7) Epoch 14, batch 5400, loss[loss=0.173, simple_loss=0.2385, pruned_loss=0.05372, over 4899.00 frames. ], tot_loss[loss=0.1819, simple_loss=0.2517, pruned_loss=0.05607, over 953353.43 frames. ], batch size: 36, lr: 3.52e-03, grad_scale: 16.0 2023-04-27 06:19:35,743 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.7361, 2.0903, 1.9342, 2.1136, 1.9308, 2.0732, 1.9208, 1.8993], device='cuda:1'), covar=tensor([0.5369, 0.7272, 0.6294, 0.5504, 0.6546, 0.8115, 0.7634, 0.6993], device='cuda:1'), in_proj_covar=tensor([0.0335, 0.0379, 0.0321, 0.0334, 0.0344, 0.0402, 0.0359, 0.0330], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 06:19:39,953 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6675, 1.1825, 4.1483, 3.6783, 3.6663, 3.6813, 3.7614, 3.5428], device='cuda:1'), covar=tensor([0.8665, 0.8351, 0.1423, 0.2731, 0.1979, 0.3167, 0.3414, 0.2735], device='cuda:1'), in_proj_covar=tensor([0.0310, 0.0304, 0.0401, 0.0402, 0.0347, 0.0402, 0.0310, 0.0362], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-27 06:19:45,248 INFO [finetune.py:976] (1/7) Epoch 14, batch 5450, loss[loss=0.1498, simple_loss=0.2106, pruned_loss=0.04449, over 4917.00 frames. ], tot_loss[loss=0.1812, simple_loss=0.2499, pruned_loss=0.05626, over 953389.44 frames. ], batch size: 36, lr: 3.52e-03, grad_scale: 16.0 2023-04-27 06:20:04,589 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=79941.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:20:04,606 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=79941.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 06:20:09,136 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.082e+02 1.493e+02 1.714e+02 2.129e+02 3.310e+02, threshold=3.428e+02, percent-clipped=0.0 2023-04-27 06:20:18,679 INFO [finetune.py:976] (1/7) Epoch 14, batch 5500, loss[loss=0.1574, simple_loss=0.2141, pruned_loss=0.05028, over 4202.00 frames. ], tot_loss[loss=0.1784, simple_loss=0.2465, pruned_loss=0.05512, over 954179.73 frames. ], batch size: 18, lr: 3.52e-03, grad_scale: 16.0 2023-04-27 06:20:36,321 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=79989.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 06:20:36,338 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=79989.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:20:47,971 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.46 vs. limit=5.0 2023-04-27 06:20:53,195 INFO [finetune.py:976] (1/7) Epoch 14, batch 5550, loss[loss=0.1625, simple_loss=0.2299, pruned_loss=0.04756, over 4730.00 frames. ], tot_loss[loss=0.179, simple_loss=0.2475, pruned_loss=0.05527, over 956159.24 frames. ], batch size: 23, lr: 3.52e-03, grad_scale: 16.0 2023-04-27 06:20:54,495 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=80013.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:20:59,229 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=80020.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:21:06,027 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=80031.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 06:21:06,656 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.2330, 1.6951, 2.0760, 2.4362, 2.0800, 1.6595, 1.2874, 1.8249], device='cuda:1'), covar=tensor([0.3410, 0.3408, 0.1707, 0.2321, 0.2673, 0.2740, 0.4268, 0.2120], device='cuda:1'), in_proj_covar=tensor([0.0291, 0.0250, 0.0225, 0.0318, 0.0217, 0.0232, 0.0232, 0.0187], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-27 06:21:09,682 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=80037.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:21:16,566 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 8.402e+01 1.777e+02 2.079e+02 2.504e+02 5.110e+02, threshold=4.158e+02, percent-clipped=3.0 2023-04-27 06:21:24,824 INFO [finetune.py:976] (1/7) Epoch 14, batch 5600, loss[loss=0.1455, simple_loss=0.2259, pruned_loss=0.03253, over 4925.00 frames. ], tot_loss[loss=0.181, simple_loss=0.2509, pruned_loss=0.05558, over 957529.69 frames. ], batch size: 37, lr: 3.52e-03, grad_scale: 16.0 2023-04-27 06:21:26,096 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.3899, 1.7381, 2.2081, 2.8808, 2.2619, 1.7698, 1.6613, 2.1803], device='cuda:1'), covar=tensor([0.3275, 0.3449, 0.1713, 0.2489, 0.2776, 0.2740, 0.4042, 0.2360], device='cuda:1'), in_proj_covar=tensor([0.0291, 0.0249, 0.0225, 0.0317, 0.0216, 0.0231, 0.0231, 0.0187], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-27 06:21:28,972 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=80068.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:21:35,800 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=80079.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 06:22:01,085 INFO [finetune.py:976] (1/7) Epoch 14, batch 5650, loss[loss=0.2128, simple_loss=0.2755, pruned_loss=0.07504, over 4826.00 frames. ], tot_loss[loss=0.1828, simple_loss=0.2532, pruned_loss=0.05621, over 956803.00 frames. ], batch size: 33, lr: 3.52e-03, grad_scale: 16.0 2023-04-27 06:22:11,112 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=80118.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:22:28,363 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.156e+02 1.596e+02 1.966e+02 2.394e+02 5.142e+02, threshold=3.932e+02, percent-clipped=2.0 2023-04-27 06:22:47,952 INFO [finetune.py:976] (1/7) Epoch 14, batch 5700, loss[loss=0.1468, simple_loss=0.2162, pruned_loss=0.03871, over 4473.00 frames. ], tot_loss[loss=0.1808, simple_loss=0.2502, pruned_loss=0.05572, over 938254.04 frames. ], batch size: 19, lr: 3.52e-03, grad_scale: 16.0 2023-04-27 06:23:36,604 INFO [finetune.py:976] (1/7) Epoch 15, batch 0, loss[loss=0.1831, simple_loss=0.2533, pruned_loss=0.05643, over 4917.00 frames. ], tot_loss[loss=0.1831, simple_loss=0.2533, pruned_loss=0.05643, over 4917.00 frames. ], batch size: 33, lr: 3.52e-03, grad_scale: 16.0 2023-04-27 06:23:36,604 INFO [finetune.py:1001] (1/7) Computing validation loss 2023-04-27 06:23:38,911 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([0.7318, 1.9128, 1.8248, 1.2956, 1.9972, 1.5661, 2.4216, 1.5897], device='cuda:1'), covar=tensor([0.3574, 0.1642, 0.4516, 0.2936, 0.1405, 0.2328, 0.1475, 0.4742], device='cuda:1'), in_proj_covar=tensor([0.0343, 0.0346, 0.0427, 0.0355, 0.0381, 0.0381, 0.0372, 0.0419], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 06:23:53,230 INFO [finetune.py:1010] (1/7) Epoch 15, validation: loss=0.1536, simple_loss=0.2258, pruned_loss=0.04063, over 2265189.00 frames. 2023-04-27 06:23:53,230 INFO [finetune.py:1011] (1/7) Maximum memory allocated so far is 6391MB 2023-04-27 06:23:53,838 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8667, 1.4339, 1.7080, 1.6966, 1.6922, 1.3887, 0.7101, 1.3856], device='cuda:1'), covar=tensor([0.3725, 0.3645, 0.1923, 0.2669, 0.2719, 0.2852, 0.4695, 0.2279], device='cuda:1'), in_proj_covar=tensor([0.0290, 0.0249, 0.0225, 0.0318, 0.0216, 0.0231, 0.0231, 0.0187], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-27 06:24:48,184 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0847, 2.6740, 1.0210, 1.4448, 2.1494, 1.1969, 3.7240, 1.8997], device='cuda:1'), covar=tensor([0.0710, 0.0787, 0.0892, 0.1324, 0.0547, 0.1113, 0.0204, 0.0639], device='cuda:1'), in_proj_covar=tensor([0.0051, 0.0066, 0.0049, 0.0047, 0.0050, 0.0052, 0.0076, 0.0051], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:1') 2023-04-27 06:24:56,493 INFO [finetune.py:976] (1/7) Epoch 15, batch 50, loss[loss=0.1627, simple_loss=0.2416, pruned_loss=0.04192, over 4813.00 frames. ], tot_loss[loss=0.1859, simple_loss=0.2554, pruned_loss=0.05814, over 215311.07 frames. ], batch size: 38, lr: 3.52e-03, grad_scale: 16.0 2023-04-27 06:24:59,882 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=80241.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:24:59,935 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.1099, 1.3509, 1.8744, 2.4067, 1.9973, 1.4983, 1.2187, 1.6956], device='cuda:1'), covar=tensor([0.3750, 0.4447, 0.2032, 0.2896, 0.3015, 0.2962, 0.4997, 0.2552], device='cuda:1'), in_proj_covar=tensor([0.0289, 0.0249, 0.0225, 0.0317, 0.0216, 0.0230, 0.0231, 0.0186], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-27 06:25:08,718 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.035e+02 1.528e+02 1.806e+02 2.187e+02 6.322e+02, threshold=3.611e+02, percent-clipped=3.0 2023-04-27 06:25:08,837 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6260, 1.6169, 0.6551, 1.2997, 1.6405, 1.4456, 1.3351, 1.4352], device='cuda:1'), covar=tensor([0.0544, 0.0415, 0.0403, 0.0630, 0.0301, 0.0593, 0.0563, 0.0639], device='cuda:1'), in_proj_covar=tensor([0.0029, 0.0024, 0.0023, 0.0029, 0.0020, 0.0028, 0.0028, 0.0029], device='cuda:1'), out_proj_covar=tensor([0.0050, 0.0044, 0.0038, 0.0050, 0.0038, 0.0049, 0.0048, 0.0050], device='cuda:1') 2023-04-27 06:25:29,856 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-04-27 06:25:36,397 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=80283.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:25:40,291 INFO [finetune.py:976] (1/7) Epoch 15, batch 100, loss[loss=0.1818, simple_loss=0.253, pruned_loss=0.05532, over 4903.00 frames. ], tot_loss[loss=0.1844, simple_loss=0.2518, pruned_loss=0.05856, over 380461.99 frames. ], batch size: 36, lr: 3.52e-03, grad_scale: 16.0 2023-04-27 06:25:41,453 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=80289.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:25:55,286 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 2023-04-27 06:25:56,996 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=80313.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:26:03,744 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.27 vs. limit=5.0 2023-04-27 06:26:12,989 INFO [finetune.py:976] (1/7) Epoch 15, batch 150, loss[loss=0.1669, simple_loss=0.2404, pruned_loss=0.04669, over 4837.00 frames. ], tot_loss[loss=0.1788, simple_loss=0.2453, pruned_loss=0.05614, over 508921.69 frames. ], batch size: 33, lr: 3.51e-03, grad_scale: 16.0 2023-04-27 06:26:18,199 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=80344.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:26:20,364 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.178e+02 1.629e+02 1.838e+02 2.299e+02 3.632e+02, threshold=3.676e+02, percent-clipped=1.0 2023-04-27 06:26:28,916 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=80361.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:26:34,423 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5465, 1.4449, 1.9123, 1.8788, 1.4730, 1.2415, 1.5589, 1.0692], device='cuda:1'), covar=tensor([0.0581, 0.0741, 0.0409, 0.0636, 0.0746, 0.1242, 0.0579, 0.0713], device='cuda:1'), in_proj_covar=tensor([0.0068, 0.0070, 0.0069, 0.0067, 0.0075, 0.0096, 0.0075, 0.0069], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-27 06:26:46,348 INFO [finetune.py:976] (1/7) Epoch 15, batch 200, loss[loss=0.1864, simple_loss=0.2572, pruned_loss=0.05787, over 4818.00 frames. ], tot_loss[loss=0.1805, simple_loss=0.2465, pruned_loss=0.05724, over 608707.85 frames. ], batch size: 33, lr: 3.51e-03, grad_scale: 16.0 2023-04-27 06:26:49,945 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=80393.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:27:06,645 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=80418.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:27:18,653 INFO [finetune.py:976] (1/7) Epoch 15, batch 250, loss[loss=0.191, simple_loss=0.2725, pruned_loss=0.05472, over 4106.00 frames. ], tot_loss[loss=0.183, simple_loss=0.2492, pruned_loss=0.05839, over 683699.03 frames. ], batch size: 65, lr: 3.51e-03, grad_scale: 16.0 2023-04-27 06:27:25,586 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.144e+02 1.693e+02 2.067e+02 2.411e+02 4.714e+02, threshold=4.133e+02, percent-clipped=3.0 2023-04-27 06:27:31,411 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=80454.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:27:36,877 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=80463.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:27:38,600 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=80466.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:27:40,417 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.3935, 1.3642, 1.6970, 1.6637, 1.3783, 1.1150, 1.4961, 0.9697], device='cuda:1'), covar=tensor([0.0634, 0.0658, 0.0484, 0.0711, 0.0838, 0.1245, 0.0644, 0.0781], device='cuda:1'), in_proj_covar=tensor([0.0068, 0.0071, 0.0070, 0.0068, 0.0076, 0.0097, 0.0076, 0.0069], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-27 06:27:49,029 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.54 vs. limit=5.0 2023-04-27 06:27:51,746 INFO [finetune.py:976] (1/7) Epoch 15, batch 300, loss[loss=0.1737, simple_loss=0.2591, pruned_loss=0.04411, over 4728.00 frames. ], tot_loss[loss=0.1848, simple_loss=0.2522, pruned_loss=0.05873, over 743592.47 frames. ], batch size: 54, lr: 3.51e-03, grad_scale: 16.0 2023-04-27 06:27:55,999 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=80494.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:28:14,220 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4198, 2.4731, 2.0188, 2.1225, 2.6404, 1.9915, 3.3474, 1.9302], device='cuda:1'), covar=tensor([0.3920, 0.2155, 0.4439, 0.3348, 0.1782, 0.2887, 0.1618, 0.4234], device='cuda:1'), in_proj_covar=tensor([0.0345, 0.0350, 0.0432, 0.0357, 0.0385, 0.0384, 0.0375, 0.0424], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 06:28:27,728 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=80524.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:28:28,337 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.1611, 2.8163, 2.1987, 2.6041, 1.8929, 2.4040, 2.4496, 1.8211], device='cuda:1'), covar=tensor([0.1895, 0.0897, 0.0797, 0.1077, 0.2764, 0.1014, 0.1677, 0.2262], device='cuda:1'), in_proj_covar=tensor([0.0288, 0.0308, 0.0222, 0.0279, 0.0310, 0.0261, 0.0252, 0.0267], device='cuda:1'), out_proj_covar=tensor([1.1605e-04, 1.2258e-04, 8.8420e-05, 1.1089e-04, 1.2615e-04, 1.0438e-04, 1.0201e-04, 1.0631e-04], device='cuda:1') 2023-04-27 06:28:36,206 INFO [finetune.py:976] (1/7) Epoch 15, batch 350, loss[loss=0.2691, simple_loss=0.3279, pruned_loss=0.1051, over 4730.00 frames. ], tot_loss[loss=0.1859, simple_loss=0.2546, pruned_loss=0.05859, over 791364.59 frames. ], batch size: 59, lr: 3.51e-03, grad_scale: 16.0 2023-04-27 06:28:37,545 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.2615, 2.9805, 2.2974, 2.7756, 2.0771, 2.5252, 2.5144, 1.8996], device='cuda:1'), covar=tensor([0.2114, 0.1239, 0.0837, 0.1251, 0.2866, 0.1155, 0.2134, 0.2732], device='cuda:1'), in_proj_covar=tensor([0.0287, 0.0307, 0.0221, 0.0278, 0.0309, 0.0261, 0.0252, 0.0266], device='cuda:1'), out_proj_covar=tensor([1.1582e-04, 1.2234e-04, 8.8270e-05, 1.1076e-04, 1.2598e-04, 1.0423e-04, 1.0192e-04, 1.0616e-04], device='cuda:1') 2023-04-27 06:28:42,201 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.160e+02 1.655e+02 1.989e+02 2.509e+02 3.787e+02, threshold=3.978e+02, percent-clipped=0.0 2023-04-27 06:28:48,609 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=80555.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:28:55,113 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-04-27 06:29:15,645 INFO [finetune.py:976] (1/7) Epoch 15, batch 400, loss[loss=0.1842, simple_loss=0.2561, pruned_loss=0.05616, over 4862.00 frames. ], tot_loss[loss=0.1858, simple_loss=0.2549, pruned_loss=0.05833, over 826478.23 frames. ], batch size: 34, lr: 3.51e-03, grad_scale: 16.0 2023-04-27 06:30:07,683 INFO [finetune.py:976] (1/7) Epoch 15, batch 450, loss[loss=0.1681, simple_loss=0.2378, pruned_loss=0.04921, over 4917.00 frames. ], tot_loss[loss=0.1822, simple_loss=0.2512, pruned_loss=0.05661, over 855112.80 frames. ], batch size: 46, lr: 3.51e-03, grad_scale: 16.0 2023-04-27 06:30:13,733 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=80639.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:30:18,556 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.255e+02 1.612e+02 1.974e+02 2.306e+02 4.694e+02, threshold=3.949e+02, percent-clipped=1.0 2023-04-27 06:31:13,676 INFO [finetune.py:976] (1/7) Epoch 15, batch 500, loss[loss=0.1643, simple_loss=0.2289, pruned_loss=0.04982, over 4862.00 frames. ], tot_loss[loss=0.1804, simple_loss=0.2493, pruned_loss=0.05572, over 879285.78 frames. ], batch size: 31, lr: 3.51e-03, grad_scale: 16.0 2023-04-27 06:31:52,662 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4752, 1.5185, 1.7698, 1.8622, 1.4957, 1.1085, 1.4614, 0.9430], device='cuda:1'), covar=tensor([0.0703, 0.0683, 0.0483, 0.0532, 0.0689, 0.1492, 0.0741, 0.0892], device='cuda:1'), in_proj_covar=tensor([0.0068, 0.0070, 0.0069, 0.0068, 0.0075, 0.0096, 0.0075, 0.0069], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-27 06:31:57,445 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.3125, 1.2450, 4.1571, 3.8877, 3.6451, 3.9629, 3.9350, 3.6606], device='cuda:1'), covar=tensor([0.7251, 0.6069, 0.1121, 0.1730, 0.1205, 0.1682, 0.1367, 0.1517], device='cuda:1'), in_proj_covar=tensor([0.0310, 0.0307, 0.0401, 0.0403, 0.0349, 0.0404, 0.0312, 0.0361], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-27 06:32:06,668 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=80736.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:32:07,801 INFO [finetune.py:976] (1/7) Epoch 15, batch 550, loss[loss=0.1695, simple_loss=0.2379, pruned_loss=0.05057, over 4888.00 frames. ], tot_loss[loss=0.1781, simple_loss=0.2463, pruned_loss=0.05495, over 895083.98 frames. ], batch size: 35, lr: 3.51e-03, grad_scale: 16.0 2023-04-27 06:32:08,005 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.55 vs. limit=2.0 2023-04-27 06:32:13,247 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.128e+02 1.610e+02 1.836e+02 2.201e+02 4.321e+02, threshold=3.672e+02, percent-clipped=1.0 2023-04-27 06:32:14,553 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=80749.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:32:32,596 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=80773.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:32:41,567 INFO [finetune.py:976] (1/7) Epoch 15, batch 600, loss[loss=0.1539, simple_loss=0.2256, pruned_loss=0.04114, over 4758.00 frames. ], tot_loss[loss=0.1795, simple_loss=0.2475, pruned_loss=0.05578, over 909283.52 frames. ], batch size: 28, lr: 3.51e-03, grad_scale: 16.0 2023-04-27 06:32:47,132 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=80797.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:33:03,048 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=80819.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:33:12,684 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=80834.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:33:15,001 INFO [finetune.py:976] (1/7) Epoch 15, batch 650, loss[loss=0.1327, simple_loss=0.2083, pruned_loss=0.02859, over 4772.00 frames. ], tot_loss[loss=0.1816, simple_loss=0.2505, pruned_loss=0.05635, over 919027.12 frames. ], batch size: 28, lr: 3.51e-03, grad_scale: 16.0 2023-04-27 06:33:20,401 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.224e+02 1.693e+02 1.997e+02 2.392e+02 3.553e+02, threshold=3.994e+02, percent-clipped=0.0 2023-04-27 06:33:22,322 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=80850.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:33:48,174 INFO [finetune.py:976] (1/7) Epoch 15, batch 700, loss[loss=0.1805, simple_loss=0.2568, pruned_loss=0.05211, over 4769.00 frames. ], tot_loss[loss=0.1821, simple_loss=0.2517, pruned_loss=0.05627, over 927454.42 frames. ], batch size: 28, lr: 3.51e-03, grad_scale: 16.0 2023-04-27 06:34:21,750 INFO [finetune.py:976] (1/7) Epoch 15, batch 750, loss[loss=0.2063, simple_loss=0.2774, pruned_loss=0.06761, over 4712.00 frames. ], tot_loss[loss=0.183, simple_loss=0.2528, pruned_loss=0.0566, over 933362.18 frames. ], batch size: 54, lr: 3.51e-03, grad_scale: 16.0 2023-04-27 06:34:22,437 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=80939.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:34:27,257 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.065e+02 1.649e+02 1.974e+02 2.385e+02 5.578e+02, threshold=3.948e+02, percent-clipped=5.0 2023-04-27 06:34:54,975 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=80987.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:34:55,525 INFO [finetune.py:976] (1/7) Epoch 15, batch 800, loss[loss=0.1764, simple_loss=0.2546, pruned_loss=0.04909, over 4834.00 frames. ], tot_loss[loss=0.1826, simple_loss=0.253, pruned_loss=0.05611, over 937044.07 frames. ], batch size: 47, lr: 3.51e-03, grad_scale: 16.0 2023-04-27 06:35:35,154 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.4323, 2.1742, 2.5076, 2.8334, 2.8706, 2.2337, 2.0440, 2.4493], device='cuda:1'), covar=tensor([0.0797, 0.0944, 0.0500, 0.0545, 0.0570, 0.0823, 0.0685, 0.0510], device='cuda:1'), in_proj_covar=tensor([0.0193, 0.0205, 0.0185, 0.0174, 0.0180, 0.0185, 0.0156, 0.0182], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 06:35:35,657 INFO [finetune.py:976] (1/7) Epoch 15, batch 850, loss[loss=0.1709, simple_loss=0.2471, pruned_loss=0.04737, over 4918.00 frames. ], tot_loss[loss=0.1826, simple_loss=0.2526, pruned_loss=0.05631, over 938275.11 frames. ], batch size: 43, lr: 3.51e-03, grad_scale: 16.0 2023-04-27 06:35:46,469 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.971e+01 1.581e+02 1.945e+02 2.309e+02 3.593e+02, threshold=3.889e+02, percent-clipped=0.0 2023-04-27 06:35:47,791 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=81049.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:36:41,753 INFO [finetune.py:976] (1/7) Epoch 15, batch 900, loss[loss=0.1743, simple_loss=0.2459, pruned_loss=0.05139, over 4859.00 frames. ], tot_loss[loss=0.1805, simple_loss=0.2499, pruned_loss=0.05554, over 943262.98 frames. ], batch size: 44, lr: 3.51e-03, grad_scale: 16.0 2023-04-27 06:36:49,896 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=81092.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:36:52,921 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=81097.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:37:15,226 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=81116.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:37:22,538 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=81119.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:37:34,999 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=81129.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:37:48,250 INFO [finetune.py:976] (1/7) Epoch 15, batch 950, loss[loss=0.1848, simple_loss=0.258, pruned_loss=0.05577, over 4907.00 frames. ], tot_loss[loss=0.1801, simple_loss=0.2489, pruned_loss=0.05566, over 945650.08 frames. ], batch size: 37, lr: 3.51e-03, grad_scale: 16.0 2023-04-27 06:38:04,673 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.127e+02 1.625e+02 2.019e+02 2.368e+02 4.092e+02, threshold=4.038e+02, percent-clipped=2.0 2023-04-27 06:38:06,685 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=81150.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:38:17,042 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=81167.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:38:23,603 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=81177.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:38:32,219 INFO [finetune.py:976] (1/7) Epoch 15, batch 1000, loss[loss=0.2013, simple_loss=0.2661, pruned_loss=0.06821, over 4855.00 frames. ], tot_loss[loss=0.1815, simple_loss=0.2501, pruned_loss=0.05643, over 948499.89 frames. ], batch size: 31, lr: 3.51e-03, grad_scale: 16.0 2023-04-27 06:38:38,869 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=81198.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:38:39,523 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0449, 2.7538, 1.0267, 1.3820, 2.0742, 1.2736, 3.6086, 1.5285], device='cuda:1'), covar=tensor([0.0677, 0.0641, 0.0791, 0.1268, 0.0501, 0.0977, 0.0178, 0.0720], device='cuda:1'), in_proj_covar=tensor([0.0052, 0.0067, 0.0049, 0.0047, 0.0050, 0.0053, 0.0077, 0.0052], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:1') 2023-04-27 06:39:02,527 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.3694, 1.4889, 1.7300, 1.8625, 1.7166, 1.8202, 1.7537, 1.7698], device='cuda:1'), covar=tensor([0.4044, 0.5727, 0.4580, 0.4561, 0.5570, 0.7582, 0.5772, 0.5296], device='cuda:1'), in_proj_covar=tensor([0.0329, 0.0374, 0.0317, 0.0331, 0.0341, 0.0399, 0.0355, 0.0326], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 06:39:05,804 INFO [finetune.py:976] (1/7) Epoch 15, batch 1050, loss[loss=0.1253, simple_loss=0.1833, pruned_loss=0.03366, over 4722.00 frames. ], tot_loss[loss=0.1833, simple_loss=0.2528, pruned_loss=0.05694, over 951084.16 frames. ], batch size: 23, lr: 3.51e-03, grad_scale: 16.0 2023-04-27 06:39:11,771 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.267e+02 1.726e+02 1.946e+02 2.343e+02 3.308e+02, threshold=3.891e+02, percent-clipped=0.0 2023-04-27 06:39:38,631 INFO [finetune.py:976] (1/7) Epoch 15, batch 1100, loss[loss=0.2096, simple_loss=0.2822, pruned_loss=0.06848, over 4800.00 frames. ], tot_loss[loss=0.184, simple_loss=0.2543, pruned_loss=0.05689, over 954417.33 frames. ], batch size: 51, lr: 3.51e-03, grad_scale: 16.0 2023-04-27 06:39:46,743 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=81299.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:39:47,923 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=81301.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:39:53,398 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.3310, 1.2943, 1.4390, 1.6604, 1.6814, 1.4100, 0.8710, 1.5114], device='cuda:1'), covar=tensor([0.0824, 0.1389, 0.0846, 0.0602, 0.0733, 0.0767, 0.0968, 0.0613], device='cuda:1'), in_proj_covar=tensor([0.0192, 0.0204, 0.0184, 0.0173, 0.0180, 0.0184, 0.0155, 0.0182], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 06:40:11,231 INFO [finetune.py:976] (1/7) Epoch 15, batch 1150, loss[loss=0.1684, simple_loss=0.2412, pruned_loss=0.04782, over 4851.00 frames. ], tot_loss[loss=0.1837, simple_loss=0.2543, pruned_loss=0.05656, over 953742.51 frames. ], batch size: 31, lr: 3.51e-03, grad_scale: 16.0 2023-04-27 06:40:18,548 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.118e+02 1.629e+02 1.924e+02 2.399e+02 3.865e+02, threshold=3.848e+02, percent-clipped=0.0 2023-04-27 06:40:27,084 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=81360.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:40:28,311 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=81362.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:40:32,744 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-04-27 06:40:44,964 INFO [finetune.py:976] (1/7) Epoch 15, batch 1200, loss[loss=0.1653, simple_loss=0.2379, pruned_loss=0.04639, over 4745.00 frames. ], tot_loss[loss=0.182, simple_loss=0.2522, pruned_loss=0.05588, over 950294.77 frames. ], batch size: 54, lr: 3.51e-03, grad_scale: 16.0 2023-04-27 06:40:48,476 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=81392.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:40:52,433 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.3843, 1.5429, 1.5077, 1.7916, 1.7232, 1.8807, 1.4343, 3.4615], device='cuda:1'), covar=tensor([0.0649, 0.0879, 0.0840, 0.1229, 0.0656, 0.0505, 0.0766, 0.0171], device='cuda:1'), in_proj_covar=tensor([0.0038, 0.0039, 0.0040, 0.0043, 0.0040, 0.0038, 0.0038, 0.0057], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:1') 2023-04-27 06:41:12,547 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=81429.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:41:18,366 INFO [finetune.py:976] (1/7) Epoch 15, batch 1250, loss[loss=0.1765, simple_loss=0.2376, pruned_loss=0.05771, over 4934.00 frames. ], tot_loss[loss=0.1795, simple_loss=0.2496, pruned_loss=0.05469, over 953170.02 frames. ], batch size: 33, lr: 3.51e-03, grad_scale: 16.0 2023-04-27 06:41:20,055 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=81440.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:41:31,406 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.048e+02 1.659e+02 1.910e+02 2.320e+02 4.141e+02, threshold=3.819e+02, percent-clipped=2.0 2023-04-27 06:42:05,039 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=81472.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:42:08,082 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=81477.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:42:20,541 INFO [finetune.py:976] (1/7) Epoch 15, batch 1300, loss[loss=0.1437, simple_loss=0.2135, pruned_loss=0.03693, over 4772.00 frames. ], tot_loss[loss=0.1766, simple_loss=0.2459, pruned_loss=0.05371, over 953325.23 frames. ], batch size: 28, lr: 3.50e-03, grad_scale: 32.0 2023-04-27 06:43:12,244 INFO [finetune.py:976] (1/7) Epoch 15, batch 1350, loss[loss=0.175, simple_loss=0.2495, pruned_loss=0.0503, over 4833.00 frames. ], tot_loss[loss=0.1777, simple_loss=0.2465, pruned_loss=0.05443, over 953473.19 frames. ], batch size: 51, lr: 3.50e-03, grad_scale: 32.0 2023-04-27 06:43:24,612 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.020e+02 1.644e+02 2.016e+02 2.436e+02 4.078e+02, threshold=4.033e+02, percent-clipped=1.0 2023-04-27 06:44:07,693 INFO [finetune.py:976] (1/7) Epoch 15, batch 1400, loss[loss=0.1768, simple_loss=0.2629, pruned_loss=0.0453, over 4732.00 frames. ], tot_loss[loss=0.1812, simple_loss=0.2507, pruned_loss=0.05589, over 954507.46 frames. ], batch size: 54, lr: 3.50e-03, grad_scale: 32.0 2023-04-27 06:44:10,758 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0882, 2.6730, 1.0633, 1.4667, 2.0342, 1.3016, 3.5308, 1.7523], device='cuda:1'), covar=tensor([0.0666, 0.0689, 0.0787, 0.1274, 0.0476, 0.1016, 0.0202, 0.0668], device='cuda:1'), in_proj_covar=tensor([0.0051, 0.0066, 0.0048, 0.0047, 0.0050, 0.0052, 0.0076, 0.0051], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:1') 2023-04-27 06:44:27,558 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-04-27 06:44:37,161 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-04-27 06:44:41,269 INFO [finetune.py:976] (1/7) Epoch 15, batch 1450, loss[loss=0.2079, simple_loss=0.2643, pruned_loss=0.07577, over 4887.00 frames. ], tot_loss[loss=0.1816, simple_loss=0.2514, pruned_loss=0.05587, over 955003.14 frames. ], batch size: 32, lr: 3.50e-03, grad_scale: 32.0 2023-04-27 06:44:47,202 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.153e+02 1.626e+02 2.006e+02 2.422e+02 4.010e+02, threshold=4.013e+02, percent-clipped=0.0 2023-04-27 06:44:53,606 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=81655.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:44:55,288 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=81657.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:45:04,207 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=81670.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:45:15,082 INFO [finetune.py:976] (1/7) Epoch 15, batch 1500, loss[loss=0.1944, simple_loss=0.2619, pruned_loss=0.0634, over 4311.00 frames. ], tot_loss[loss=0.1835, simple_loss=0.2532, pruned_loss=0.05688, over 954450.69 frames. ], batch size: 65, lr: 3.50e-03, grad_scale: 32.0 2023-04-27 06:45:17,606 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0671, 1.3570, 1.8087, 2.0289, 1.9091, 1.4172, 1.0419, 1.5161], device='cuda:1'), covar=tensor([0.3787, 0.4072, 0.1998, 0.2916, 0.2923, 0.2942, 0.4628, 0.2389], device='cuda:1'), in_proj_covar=tensor([0.0285, 0.0244, 0.0221, 0.0312, 0.0213, 0.0227, 0.0227, 0.0183], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-27 06:45:44,066 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=81731.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 06:45:48,178 INFO [finetune.py:976] (1/7) Epoch 15, batch 1550, loss[loss=0.2459, simple_loss=0.2978, pruned_loss=0.09696, over 4270.00 frames. ], tot_loss[loss=0.1841, simple_loss=0.2541, pruned_loss=0.05708, over 954190.54 frames. ], batch size: 66, lr: 3.50e-03, grad_scale: 32.0 2023-04-27 06:45:53,673 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.064e+02 1.553e+02 1.872e+02 2.288e+02 6.577e+02, threshold=3.744e+02, percent-clipped=2.0 2023-04-27 06:46:04,673 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7601, 1.2166, 1.4508, 1.3947, 1.9147, 1.5641, 1.3032, 1.3953], device='cuda:1'), covar=tensor([0.1428, 0.1440, 0.1634, 0.1367, 0.0769, 0.1330, 0.1773, 0.1975], device='cuda:1'), in_proj_covar=tensor([0.0310, 0.0318, 0.0353, 0.0292, 0.0330, 0.0314, 0.0304, 0.0362], device='cuda:1'), out_proj_covar=tensor([6.4331e-05, 6.6512e-05, 7.5552e-05, 5.9593e-05, 6.8614e-05, 6.6450e-05, 6.4297e-05, 7.7448e-05], device='cuda:1') 2023-04-27 06:46:11,794 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=81772.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:46:21,442 INFO [finetune.py:976] (1/7) Epoch 15, batch 1600, loss[loss=0.1661, simple_loss=0.2387, pruned_loss=0.04679, over 4759.00 frames. ], tot_loss[loss=0.1826, simple_loss=0.2517, pruned_loss=0.05674, over 954286.25 frames. ], batch size: 59, lr: 3.50e-03, grad_scale: 32.0 2023-04-27 06:46:30,142 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.5187, 1.7456, 1.8317, 1.9950, 1.8622, 1.9534, 1.9228, 1.8731], device='cuda:1'), covar=tensor([0.3930, 0.5865, 0.4831, 0.4594, 0.5431, 0.7380, 0.5654, 0.5626], device='cuda:1'), in_proj_covar=tensor([0.0327, 0.0373, 0.0317, 0.0331, 0.0341, 0.0399, 0.0352, 0.0325], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 06:46:42,457 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=81820.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:46:53,851 INFO [finetune.py:976] (1/7) Epoch 15, batch 1650, loss[loss=0.1546, simple_loss=0.2252, pruned_loss=0.04205, over 4858.00 frames. ], tot_loss[loss=0.1787, simple_loss=0.2474, pruned_loss=0.05499, over 955784.27 frames. ], batch size: 47, lr: 3.50e-03, grad_scale: 32.0 2023-04-27 06:47:04,626 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.076e+02 1.625e+02 1.858e+02 2.341e+02 4.893e+02, threshold=3.717e+02, percent-clipped=2.0 2023-04-27 06:47:37,364 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.44 vs. limit=5.0 2023-04-27 06:47:59,312 INFO [finetune.py:976] (1/7) Epoch 15, batch 1700, loss[loss=0.1937, simple_loss=0.2646, pruned_loss=0.06135, over 4854.00 frames. ], tot_loss[loss=0.1775, simple_loss=0.2459, pruned_loss=0.05453, over 955180.19 frames. ], batch size: 49, lr: 3.50e-03, grad_scale: 32.0 2023-04-27 06:49:06,086 INFO [finetune.py:976] (1/7) Epoch 15, batch 1750, loss[loss=0.1809, simple_loss=0.2525, pruned_loss=0.05462, over 4893.00 frames. ], tot_loss[loss=0.18, simple_loss=0.2487, pruned_loss=0.05567, over 954831.36 frames. ], batch size: 35, lr: 3.50e-03, grad_scale: 32.0 2023-04-27 06:49:16,877 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.080e+01 1.647e+02 2.011e+02 2.347e+02 5.729e+02, threshold=4.021e+02, percent-clipped=2.0 2023-04-27 06:49:20,720 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6160, 1.3558, 1.8026, 1.9559, 1.6706, 1.6374, 1.6901, 1.7191], device='cuda:1'), covar=tensor([0.6500, 0.8795, 0.8872, 0.9801, 0.8095, 1.1015, 1.0269, 1.0614], device='cuda:1'), in_proj_covar=tensor([0.0415, 0.0407, 0.0494, 0.0508, 0.0443, 0.0466, 0.0473, 0.0475], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 06:49:26,287 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=81955.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:49:27,538 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=81957.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:49:33,467 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.2287, 1.4438, 1.3499, 1.6494, 1.5793, 2.0012, 1.3191, 3.4909], device='cuda:1'), covar=tensor([0.0553, 0.0811, 0.0778, 0.1198, 0.0660, 0.0533, 0.0750, 0.0126], device='cuda:1'), in_proj_covar=tensor([0.0038, 0.0039, 0.0040, 0.0044, 0.0040, 0.0038, 0.0038, 0.0057], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:1') 2023-04-27 06:49:41,837 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8232, 1.3227, 1.4744, 1.3876, 1.9711, 1.5873, 1.3098, 1.4556], device='cuda:1'), covar=tensor([0.1419, 0.1404, 0.1813, 0.1298, 0.0797, 0.1362, 0.1929, 0.2010], device='cuda:1'), in_proj_covar=tensor([0.0310, 0.0317, 0.0353, 0.0292, 0.0331, 0.0314, 0.0304, 0.0363], device='cuda:1'), out_proj_covar=tensor([6.4301e-05, 6.6447e-05, 7.5562e-05, 5.9591e-05, 6.8832e-05, 6.6445e-05, 6.4336e-05, 7.7514e-05], device='cuda:1') 2023-04-27 06:49:43,631 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0740, 1.5527, 5.4655, 5.1561, 4.7509, 5.2852, 4.7449, 4.9295], device='cuda:1'), covar=tensor([0.6818, 0.6121, 0.1044, 0.1666, 0.0982, 0.1279, 0.1172, 0.1516], device='cuda:1'), in_proj_covar=tensor([0.0313, 0.0308, 0.0403, 0.0406, 0.0350, 0.0406, 0.0315, 0.0364], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-27 06:49:48,431 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=81987.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:49:48,926 INFO [finetune.py:976] (1/7) Epoch 15, batch 1800, loss[loss=0.1691, simple_loss=0.2404, pruned_loss=0.04893, over 4885.00 frames. ], tot_loss[loss=0.1822, simple_loss=0.2518, pruned_loss=0.05628, over 955612.95 frames. ], batch size: 32, lr: 3.50e-03, grad_scale: 32.0 2023-04-27 06:49:59,486 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=82003.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:50:00,739 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=82005.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:50:16,415 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=82026.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 06:50:24,048 INFO [finetune.py:976] (1/7) Epoch 15, batch 1850, loss[loss=0.2174, simple_loss=0.2796, pruned_loss=0.07755, over 4157.00 frames. ], tot_loss[loss=0.1839, simple_loss=0.2538, pruned_loss=0.05704, over 955258.58 frames. ], batch size: 65, lr: 3.50e-03, grad_scale: 32.0 2023-04-27 06:50:29,487 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.115e+02 1.760e+02 1.989e+02 2.247e+02 4.902e+02, threshold=3.979e+02, percent-clipped=1.0 2023-04-27 06:50:30,233 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=82048.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:50:34,521 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=82055.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:50:56,619 INFO [finetune.py:976] (1/7) Epoch 15, batch 1900, loss[loss=0.1453, simple_loss=0.2103, pruned_loss=0.04021, over 4752.00 frames. ], tot_loss[loss=0.1836, simple_loss=0.2535, pruned_loss=0.05681, over 954049.96 frames. ], batch size: 23, lr: 3.50e-03, grad_scale: 32.0 2023-04-27 06:51:14,327 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=82116.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:51:30,519 INFO [finetune.py:976] (1/7) Epoch 15, batch 1950, loss[loss=0.1589, simple_loss=0.2423, pruned_loss=0.03775, over 4870.00 frames. ], tot_loss[loss=0.1819, simple_loss=0.2518, pruned_loss=0.05605, over 953007.47 frames. ], batch size: 34, lr: 3.50e-03, grad_scale: 32.0 2023-04-27 06:51:36,465 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.029e+02 1.535e+02 1.831e+02 2.341e+02 3.747e+02, threshold=3.661e+02, percent-clipped=0.0 2023-04-27 06:51:57,292 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.40 vs. limit=5.0 2023-04-27 06:52:04,204 INFO [finetune.py:976] (1/7) Epoch 15, batch 2000, loss[loss=0.1759, simple_loss=0.2401, pruned_loss=0.05586, over 4840.00 frames. ], tot_loss[loss=0.1803, simple_loss=0.2497, pruned_loss=0.05546, over 953632.05 frames. ], batch size: 47, lr: 3.50e-03, grad_scale: 32.0 2023-04-27 06:52:04,434 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-04-27 06:52:32,394 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-04-27 06:52:37,841 INFO [finetune.py:976] (1/7) Epoch 15, batch 2050, loss[loss=0.1555, simple_loss=0.226, pruned_loss=0.04247, over 4825.00 frames. ], tot_loss[loss=0.1775, simple_loss=0.2463, pruned_loss=0.05432, over 954253.09 frames. ], batch size: 39, lr: 3.50e-03, grad_scale: 32.0 2023-04-27 06:52:43,308 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.160e+02 1.629e+02 1.931e+02 2.371e+02 3.947e+02, threshold=3.861e+02, percent-clipped=2.0 2023-04-27 06:53:21,146 INFO [finetune.py:976] (1/7) Epoch 15, batch 2100, loss[loss=0.2311, simple_loss=0.2996, pruned_loss=0.08128, over 4029.00 frames. ], tot_loss[loss=0.1766, simple_loss=0.2454, pruned_loss=0.05387, over 953343.35 frames. ], batch size: 65, lr: 3.50e-03, grad_scale: 32.0 2023-04-27 06:54:04,937 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=82326.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:54:13,978 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9853, 1.6914, 2.1193, 2.4215, 2.0016, 1.8507, 2.0654, 1.9720], device='cuda:1'), covar=tensor([0.5141, 0.7857, 0.8082, 0.6937, 0.6396, 0.9963, 0.9338, 1.0432], device='cuda:1'), in_proj_covar=tensor([0.0417, 0.0407, 0.0498, 0.0511, 0.0445, 0.0468, 0.0476, 0.0479], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 06:54:24,279 INFO [finetune.py:976] (1/7) Epoch 15, batch 2150, loss[loss=0.1927, simple_loss=0.2662, pruned_loss=0.05956, over 4863.00 frames. ], tot_loss[loss=0.1793, simple_loss=0.2487, pruned_loss=0.05494, over 953257.38 frames. ], batch size: 31, lr: 3.50e-03, grad_scale: 32.0 2023-04-27 06:54:33,145 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=82343.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:54:35,047 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=82346.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:54:35,531 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.316e+02 1.732e+02 1.970e+02 2.465e+02 9.359e+02, threshold=3.940e+02, percent-clipped=1.0 2023-04-27 06:54:37,062 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-04-27 06:54:46,136 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.2966, 3.1905, 0.9123, 1.6557, 1.6831, 2.3200, 1.6997, 0.9874], device='cuda:1'), covar=tensor([0.1503, 0.0983, 0.1908, 0.1303, 0.1155, 0.0957, 0.1676, 0.2110], device='cuda:1'), in_proj_covar=tensor([0.0118, 0.0246, 0.0137, 0.0121, 0.0132, 0.0154, 0.0118, 0.0120], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-27 06:54:58,340 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=82374.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:55:13,104 INFO [finetune.py:976] (1/7) Epoch 15, batch 2200, loss[loss=0.1687, simple_loss=0.2358, pruned_loss=0.05079, over 4816.00 frames. ], tot_loss[loss=0.1816, simple_loss=0.2517, pruned_loss=0.05579, over 954869.63 frames. ], batch size: 25, lr: 3.50e-03, grad_scale: 32.0 2023-04-27 06:55:42,645 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=82407.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:55:45,050 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=82411.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:56:12,256 INFO [finetune.py:976] (1/7) Epoch 15, batch 2250, loss[loss=0.1752, simple_loss=0.236, pruned_loss=0.05722, over 3992.00 frames. ], tot_loss[loss=0.1819, simple_loss=0.2519, pruned_loss=0.0559, over 952043.32 frames. ], batch size: 17, lr: 3.50e-03, grad_scale: 32.0 2023-04-27 06:56:19,194 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.169e+02 1.596e+02 1.887e+02 2.124e+02 5.729e+02, threshold=3.773e+02, percent-clipped=1.0 2023-04-27 06:56:32,318 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.1020, 2.5327, 0.9309, 1.4719, 1.5413, 1.9466, 1.5908, 0.8597], device='cuda:1'), covar=tensor([0.1534, 0.1228, 0.1715, 0.1327, 0.1134, 0.0912, 0.1680, 0.1586], device='cuda:1'), in_proj_covar=tensor([0.0118, 0.0247, 0.0138, 0.0122, 0.0132, 0.0154, 0.0119, 0.0120], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-27 06:56:45,467 INFO [finetune.py:976] (1/7) Epoch 15, batch 2300, loss[loss=0.1761, simple_loss=0.2456, pruned_loss=0.05334, over 4900.00 frames. ], tot_loss[loss=0.1818, simple_loss=0.2524, pruned_loss=0.05562, over 953492.18 frames. ], batch size: 36, lr: 3.50e-03, grad_scale: 32.0 2023-04-27 06:56:49,016 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=82493.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 06:57:00,257 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6730, 1.2544, 1.4153, 1.3567, 1.8706, 1.5087, 1.2200, 1.3709], device='cuda:1'), covar=tensor([0.1428, 0.1312, 0.1690, 0.1222, 0.0776, 0.1354, 0.1652, 0.1927], device='cuda:1'), in_proj_covar=tensor([0.0309, 0.0318, 0.0353, 0.0293, 0.0331, 0.0315, 0.0304, 0.0362], device='cuda:1'), out_proj_covar=tensor([6.4100e-05, 6.6549e-05, 7.5616e-05, 5.9970e-05, 6.8880e-05, 6.6744e-05, 6.4416e-05, 7.7380e-05], device='cuda:1') 2023-04-27 06:57:12,850 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.0273, 1.9333, 1.8119, 1.6909, 2.0810, 1.7581, 2.6461, 1.6491], device='cuda:1'), covar=tensor([0.3992, 0.2136, 0.4847, 0.3019, 0.1941, 0.2559, 0.1467, 0.4622], device='cuda:1'), in_proj_covar=tensor([0.0345, 0.0349, 0.0430, 0.0357, 0.0384, 0.0386, 0.0375, 0.0422], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 06:57:18,140 INFO [finetune.py:976] (1/7) Epoch 15, batch 2350, loss[loss=0.2074, simple_loss=0.2484, pruned_loss=0.08321, over 4749.00 frames. ], tot_loss[loss=0.1806, simple_loss=0.2505, pruned_loss=0.05533, over 953650.95 frames. ], batch size: 54, lr: 3.50e-03, grad_scale: 32.0 2023-04-27 06:57:25,043 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 8.877e+01 1.633e+02 1.887e+02 2.224e+02 3.641e+02, threshold=3.774e+02, percent-clipped=0.0 2023-04-27 06:57:30,342 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=82554.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 06:57:38,098 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=3.98 vs. limit=5.0 2023-04-27 06:57:39,323 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-04-27 06:57:40,471 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=82569.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:57:51,908 INFO [finetune.py:976] (1/7) Epoch 15, batch 2400, loss[loss=0.1762, simple_loss=0.2401, pruned_loss=0.0562, over 4811.00 frames. ], tot_loss[loss=0.1782, simple_loss=0.2475, pruned_loss=0.05447, over 954624.07 frames. ], batch size: 51, lr: 3.50e-03, grad_scale: 32.0 2023-04-27 06:57:52,660 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.1132, 1.5299, 2.0323, 2.4239, 1.9926, 1.5717, 1.1806, 1.7645], device='cuda:1'), covar=tensor([0.3503, 0.3517, 0.1698, 0.2363, 0.2642, 0.2961, 0.4718, 0.2259], device='cuda:1'), in_proj_covar=tensor([0.0292, 0.0249, 0.0225, 0.0318, 0.0217, 0.0231, 0.0232, 0.0186], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-27 06:58:16,888 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-04-27 06:58:21,015 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=82630.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:58:21,595 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.2447, 4.4615, 0.7280, 2.1995, 2.6316, 2.9387, 2.3264, 0.9509], device='cuda:1'), covar=tensor([0.1223, 0.0909, 0.2382, 0.1385, 0.0933, 0.1186, 0.1684, 0.2147], device='cuda:1'), in_proj_covar=tensor([0.0118, 0.0246, 0.0137, 0.0121, 0.0131, 0.0153, 0.0118, 0.0120], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-27 06:58:25,784 INFO [finetune.py:976] (1/7) Epoch 15, batch 2450, loss[loss=0.1695, simple_loss=0.2345, pruned_loss=0.05221, over 4745.00 frames. ], tot_loss[loss=0.1754, simple_loss=0.2443, pruned_loss=0.05321, over 955017.28 frames. ], batch size: 27, lr: 3.49e-03, grad_scale: 32.0 2023-04-27 06:58:28,863 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=82643.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:58:31,172 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.750e+01 1.533e+02 1.809e+02 2.233e+02 3.484e+02, threshold=3.618e+02, percent-clipped=0.0 2023-04-27 06:58:51,166 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([3.9721, 3.9636, 2.8370, 4.5830, 4.0324, 3.9713, 1.7053, 4.0109], device='cuda:1'), covar=tensor([0.1692, 0.1195, 0.3112, 0.1388, 0.3073, 0.1681, 0.5953, 0.2256], device='cuda:1'), in_proj_covar=tensor([0.0242, 0.0213, 0.0248, 0.0300, 0.0296, 0.0245, 0.0268, 0.0271], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 06:59:01,727 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 2023-04-27 06:59:02,102 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4454, 2.4772, 1.9731, 2.1579, 2.2992, 2.0508, 3.2794, 1.8323], device='cuda:1'), covar=tensor([0.4106, 0.2194, 0.4832, 0.3513, 0.2190, 0.2887, 0.1572, 0.4447], device='cuda:1'), in_proj_covar=tensor([0.0345, 0.0348, 0.0430, 0.0356, 0.0384, 0.0385, 0.0373, 0.0422], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 06:59:04,420 INFO [finetune.py:976] (1/7) Epoch 15, batch 2500, loss[loss=0.1581, simple_loss=0.2383, pruned_loss=0.03896, over 4770.00 frames. ], tot_loss[loss=0.178, simple_loss=0.2468, pruned_loss=0.0546, over 956178.27 frames. ], batch size: 27, lr: 3.49e-03, grad_scale: 32.0 2023-04-27 06:59:11,878 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=82691.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:59:24,990 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=82702.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:59:38,014 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=82711.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:00:09,105 INFO [finetune.py:976] (1/7) Epoch 15, batch 2550, loss[loss=0.265, simple_loss=0.3235, pruned_loss=0.1032, over 4830.00 frames. ], tot_loss[loss=0.1815, simple_loss=0.2509, pruned_loss=0.05606, over 955571.21 frames. ], batch size: 47, lr: 3.49e-03, grad_scale: 32.0 2023-04-27 07:00:14,538 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.081e+02 1.661e+02 1.980e+02 2.344e+02 4.065e+02, threshold=3.961e+02, percent-clipped=3.0 2023-04-27 07:00:24,404 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=82759.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:00:47,933 INFO [finetune.py:976] (1/7) Epoch 15, batch 2600, loss[loss=0.1858, simple_loss=0.2615, pruned_loss=0.05508, over 4793.00 frames. ], tot_loss[loss=0.1822, simple_loss=0.2524, pruned_loss=0.05601, over 954767.72 frames. ], batch size: 29, lr: 3.49e-03, grad_scale: 32.0 2023-04-27 07:01:20,017 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-04-27 07:01:33,072 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.2379, 1.2122, 1.2937, 1.6288, 1.6554, 1.3467, 0.9762, 1.5377], device='cuda:1'), covar=tensor([0.0889, 0.1346, 0.0844, 0.0597, 0.0651, 0.0748, 0.0855, 0.0541], device='cuda:1'), in_proj_covar=tensor([0.0191, 0.0204, 0.0183, 0.0173, 0.0179, 0.0184, 0.0155, 0.0181], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 07:01:40,269 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7726, 4.3077, 0.8791, 2.3095, 2.4966, 2.8632, 2.4144, 0.8130], device='cuda:1'), covar=tensor([0.1511, 0.1119, 0.2122, 0.1230, 0.1061, 0.1141, 0.1512, 0.2312], device='cuda:1'), in_proj_covar=tensor([0.0117, 0.0244, 0.0136, 0.0120, 0.0131, 0.0152, 0.0117, 0.0119], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-27 07:01:43,239 INFO [finetune.py:976] (1/7) Epoch 15, batch 2650, loss[loss=0.1813, simple_loss=0.2519, pruned_loss=0.05534, over 4898.00 frames. ], tot_loss[loss=0.1827, simple_loss=0.253, pruned_loss=0.05618, over 953001.92 frames. ], batch size: 36, lr: 3.49e-03, grad_scale: 32.0 2023-04-27 07:01:47,660 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.55 vs. limit=2.0 2023-04-27 07:01:48,666 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.169e+02 1.684e+02 1.897e+02 2.238e+02 3.554e+02, threshold=3.793e+02, percent-clipped=0.0 2023-04-27 07:01:49,959 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=82849.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 07:01:59,029 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-04-27 07:02:33,354 INFO [finetune.py:976] (1/7) Epoch 15, batch 2700, loss[loss=0.1518, simple_loss=0.2239, pruned_loss=0.03987, over 4751.00 frames. ], tot_loss[loss=0.1824, simple_loss=0.2526, pruned_loss=0.05609, over 953352.53 frames. ], batch size: 27, lr: 3.49e-03, grad_scale: 32.0 2023-04-27 07:03:04,630 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=82925.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:03:12,182 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=3.99 vs. limit=5.0 2023-04-27 07:03:12,571 INFO [finetune.py:976] (1/7) Epoch 15, batch 2750, loss[loss=0.1824, simple_loss=0.25, pruned_loss=0.05742, over 4834.00 frames. ], tot_loss[loss=0.1803, simple_loss=0.2499, pruned_loss=0.0554, over 954618.37 frames. ], batch size: 49, lr: 3.49e-03, grad_scale: 32.0 2023-04-27 07:03:18,528 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.092e+02 1.634e+02 2.005e+02 2.343e+02 4.529e+02, threshold=4.010e+02, percent-clipped=3.0 2023-04-27 07:03:28,539 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=82963.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:03:31,611 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.11 vs. limit=2.0 2023-04-27 07:03:46,114 INFO [finetune.py:976] (1/7) Epoch 15, batch 2800, loss[loss=0.1491, simple_loss=0.2106, pruned_loss=0.04377, over 4798.00 frames. ], tot_loss[loss=0.1769, simple_loss=0.2459, pruned_loss=0.05395, over 955812.78 frames. ], batch size: 25, lr: 3.49e-03, grad_scale: 32.0 2023-04-27 07:03:52,202 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5949, 1.4831, 0.5525, 1.3028, 1.4007, 1.5050, 1.3682, 1.4117], device='cuda:1'), covar=tensor([0.0509, 0.0382, 0.0395, 0.0558, 0.0301, 0.0504, 0.0507, 0.0568], device='cuda:1'), in_proj_covar=tensor([0.0029, 0.0024, 0.0023, 0.0029, 0.0020, 0.0029, 0.0028, 0.0029], device='cuda:1'), out_proj_covar=tensor([0.0050, 0.0045, 0.0038, 0.0051, 0.0038, 0.0049, 0.0049, 0.0050], device='cuda:1') 2023-04-27 07:03:55,378 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=83002.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:04:06,870 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.3673, 3.1073, 1.0451, 1.9466, 1.8857, 2.4453, 1.9573, 1.0274], device='cuda:1'), covar=tensor([0.1309, 0.0857, 0.1662, 0.1085, 0.0904, 0.0872, 0.1315, 0.1861], device='cuda:1'), in_proj_covar=tensor([0.0118, 0.0246, 0.0137, 0.0121, 0.0131, 0.0153, 0.0118, 0.0120], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-27 07:04:09,804 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=83024.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:04:19,799 INFO [finetune.py:976] (1/7) Epoch 15, batch 2850, loss[loss=0.1155, simple_loss=0.1916, pruned_loss=0.01972, over 4779.00 frames. ], tot_loss[loss=0.1758, simple_loss=0.2443, pruned_loss=0.05368, over 954583.93 frames. ], batch size: 26, lr: 3.49e-03, grad_scale: 32.0 2023-04-27 07:04:25,724 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 8.603e+01 1.615e+02 1.794e+02 2.149e+02 4.463e+02, threshold=3.588e+02, percent-clipped=2.0 2023-04-27 07:04:27,623 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=83050.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:05:03,759 INFO [finetune.py:976] (1/7) Epoch 15, batch 2900, loss[loss=0.1967, simple_loss=0.2684, pruned_loss=0.06248, over 4749.00 frames. ], tot_loss[loss=0.1782, simple_loss=0.2469, pruned_loss=0.05472, over 953138.52 frames. ], batch size: 59, lr: 3.49e-03, grad_scale: 32.0 2023-04-27 07:05:57,657 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.73 vs. limit=2.0 2023-04-27 07:06:07,041 INFO [finetune.py:976] (1/7) Epoch 15, batch 2950, loss[loss=0.2153, simple_loss=0.2848, pruned_loss=0.07284, over 4758.00 frames. ], tot_loss[loss=0.1798, simple_loss=0.2495, pruned_loss=0.05502, over 954119.99 frames. ], batch size: 54, lr: 3.49e-03, grad_scale: 32.0 2023-04-27 07:06:09,002 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5719, 2.6254, 2.3650, 2.3868, 2.7624, 2.5146, 3.6879, 1.9565], device='cuda:1'), covar=tensor([0.4030, 0.2181, 0.4012, 0.3467, 0.1973, 0.2517, 0.1497, 0.4490], device='cuda:1'), in_proj_covar=tensor([0.0345, 0.0349, 0.0428, 0.0357, 0.0385, 0.0384, 0.0374, 0.0422], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 07:06:18,120 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.270e+02 1.793e+02 2.066e+02 2.627e+02 6.571e+02, threshold=4.132e+02, percent-clipped=5.0 2023-04-27 07:06:19,454 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=83149.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 07:06:40,595 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5757, 0.9928, 1.6162, 2.0292, 1.6832, 1.5093, 1.5822, 1.5707], device='cuda:1'), covar=tensor([0.4296, 0.6250, 0.5404, 0.5971, 0.5480, 0.7223, 0.7201, 0.7826], device='cuda:1'), in_proj_covar=tensor([0.0414, 0.0405, 0.0493, 0.0507, 0.0444, 0.0466, 0.0473, 0.0477], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 07:06:56,294 INFO [finetune.py:976] (1/7) Epoch 15, batch 3000, loss[loss=0.1818, simple_loss=0.252, pruned_loss=0.05581, over 4884.00 frames. ], tot_loss[loss=0.1824, simple_loss=0.2517, pruned_loss=0.05654, over 952323.16 frames. ], batch size: 43, lr: 3.49e-03, grad_scale: 32.0 2023-04-27 07:06:56,294 INFO [finetune.py:1001] (1/7) Computing validation loss 2023-04-27 07:07:03,065 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.3272, 1.2231, 3.8703, 3.5654, 3.5012, 3.7043, 3.8163, 3.4918], device='cuda:1'), covar=tensor([0.7183, 0.5524, 0.1262, 0.2023, 0.1349, 0.1568, 0.0776, 0.1621], device='cuda:1'), in_proj_covar=tensor([0.0311, 0.0307, 0.0401, 0.0406, 0.0348, 0.0404, 0.0312, 0.0360], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-27 07:07:05,061 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.3357, 1.2783, 3.8638, 3.5673, 3.4582, 3.6514, 3.7511, 3.3900], device='cuda:1'), covar=tensor([0.7152, 0.5496, 0.1137, 0.1914, 0.1223, 0.1561, 0.0729, 0.1658], device='cuda:1'), in_proj_covar=tensor([0.0311, 0.0307, 0.0401, 0.0406, 0.0348, 0.0404, 0.0312, 0.0360], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-27 07:07:05,300 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4358, 1.6803, 1.5596, 1.7947, 1.7390, 1.7481, 1.5530, 2.8840], device='cuda:1'), covar=tensor([0.0524, 0.0648, 0.0625, 0.0987, 0.0497, 0.0461, 0.0601, 0.0207], device='cuda:1'), in_proj_covar=tensor([0.0038, 0.0039, 0.0040, 0.0044, 0.0040, 0.0038, 0.0038, 0.0057], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:1') 2023-04-27 07:07:06,875 INFO [finetune.py:1010] (1/7) Epoch 15, validation: loss=0.1516, simple_loss=0.2237, pruned_loss=0.03975, over 2265189.00 frames. 2023-04-27 07:07:06,876 INFO [finetune.py:1011] (1/7) Maximum memory allocated so far is 6391MB 2023-04-27 07:07:08,935 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=6.41 vs. limit=5.0 2023-04-27 07:07:12,338 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=83196.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:07:12,903 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=83197.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 07:07:30,729 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=83225.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:07:35,715 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.62 vs. limit=2.0 2023-04-27 07:07:38,608 INFO [finetune.py:976] (1/7) Epoch 15, batch 3050, loss[loss=0.1775, simple_loss=0.2623, pruned_loss=0.04637, over 4927.00 frames. ], tot_loss[loss=0.183, simple_loss=0.2532, pruned_loss=0.0564, over 954077.19 frames. ], batch size: 42, lr: 3.49e-03, grad_scale: 32.0 2023-04-27 07:07:50,803 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.173e+02 1.583e+02 1.890e+02 2.106e+02 3.614e+02, threshold=3.781e+02, percent-clipped=0.0 2023-04-27 07:08:02,519 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=83257.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:08:24,059 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=83273.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:08:36,019 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=83284.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:08:42,492 INFO [finetune.py:976] (1/7) Epoch 15, batch 3100, loss[loss=0.1879, simple_loss=0.2617, pruned_loss=0.05711, over 4904.00 frames. ], tot_loss[loss=0.1806, simple_loss=0.2507, pruned_loss=0.05525, over 954416.55 frames. ], batch size: 37, lr: 3.49e-03, grad_scale: 32.0 2023-04-27 07:09:15,046 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=83319.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:09:21,336 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-04-27 07:09:23,790 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2023-04-27 07:09:26,582 INFO [finetune.py:976] (1/7) Epoch 15, batch 3150, loss[loss=0.1467, simple_loss=0.2209, pruned_loss=0.03621, over 4911.00 frames. ], tot_loss[loss=0.1793, simple_loss=0.2485, pruned_loss=0.05505, over 952771.73 frames. ], batch size: 43, lr: 3.49e-03, grad_scale: 32.0 2023-04-27 07:09:31,014 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=83345.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:09:32,640 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.218e+02 1.641e+02 1.865e+02 2.202e+02 3.570e+02, threshold=3.730e+02, percent-clipped=0.0 2023-04-27 07:09:59,876 INFO [finetune.py:976] (1/7) Epoch 15, batch 3200, loss[loss=0.1547, simple_loss=0.2312, pruned_loss=0.03912, over 4815.00 frames. ], tot_loss[loss=0.1767, simple_loss=0.2454, pruned_loss=0.05406, over 954786.96 frames. ], batch size: 41, lr: 3.49e-03, grad_scale: 32.0 2023-04-27 07:10:45,247 INFO [finetune.py:976] (1/7) Epoch 15, batch 3250, loss[loss=0.2018, simple_loss=0.2639, pruned_loss=0.06984, over 4805.00 frames. ], tot_loss[loss=0.177, simple_loss=0.2457, pruned_loss=0.05413, over 955149.27 frames. ], batch size: 51, lr: 3.49e-03, grad_scale: 32.0 2023-04-27 07:10:56,130 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.105e+02 1.623e+02 1.932e+02 2.416e+02 4.902e+02, threshold=3.863e+02, percent-clipped=4.0 2023-04-27 07:11:06,590 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5141, 1.4356, 0.5020, 1.2409, 1.4797, 1.3906, 1.2953, 1.3665], device='cuda:1'), covar=tensor([0.0510, 0.0395, 0.0401, 0.0557, 0.0286, 0.0525, 0.0502, 0.0589], device='cuda:1'), in_proj_covar=tensor([0.0029, 0.0024, 0.0023, 0.0029, 0.0020, 0.0029, 0.0028, 0.0029], device='cuda:1'), out_proj_covar=tensor([0.0050, 0.0045, 0.0038, 0.0050, 0.0038, 0.0049, 0.0049, 0.0050], device='cuda:1') 2023-04-27 07:11:37,888 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.2588, 1.7159, 2.1038, 2.4496, 2.1413, 1.6688, 1.2800, 1.8932], device='cuda:1'), covar=tensor([0.3638, 0.3521, 0.1815, 0.2558, 0.2920, 0.2930, 0.4498, 0.2190], device='cuda:1'), in_proj_covar=tensor([0.0290, 0.0246, 0.0224, 0.0318, 0.0216, 0.0229, 0.0230, 0.0184], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-27 07:11:50,145 INFO [finetune.py:976] (1/7) Epoch 15, batch 3300, loss[loss=0.1465, simple_loss=0.2321, pruned_loss=0.03045, over 4767.00 frames. ], tot_loss[loss=0.1801, simple_loss=0.2494, pruned_loss=0.0554, over 955572.01 frames. ], batch size: 54, lr: 3.49e-03, grad_scale: 64.0 2023-04-27 07:11:52,109 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4036, 1.3696, 1.6371, 1.6467, 1.2718, 1.1378, 1.3784, 0.9581], device='cuda:1'), covar=tensor([0.0635, 0.0653, 0.0466, 0.0750, 0.0854, 0.1217, 0.0691, 0.0713], device='cuda:1'), in_proj_covar=tensor([0.0068, 0.0070, 0.0069, 0.0068, 0.0075, 0.0097, 0.0075, 0.0069], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-27 07:11:59,501 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-04-27 07:12:23,030 INFO [finetune.py:976] (1/7) Epoch 15, batch 3350, loss[loss=0.1665, simple_loss=0.2347, pruned_loss=0.04916, over 4829.00 frames. ], tot_loss[loss=0.1829, simple_loss=0.2524, pruned_loss=0.05668, over 954567.43 frames. ], batch size: 30, lr: 3.49e-03, grad_scale: 64.0 2023-04-27 07:12:26,194 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.0887, 2.1418, 1.7239, 1.7167, 2.2347, 1.7693, 2.7899, 1.4796], device='cuda:1'), covar=tensor([0.3991, 0.2045, 0.5372, 0.3185, 0.1772, 0.2638, 0.1082, 0.5033], device='cuda:1'), in_proj_covar=tensor([0.0344, 0.0347, 0.0427, 0.0357, 0.0384, 0.0384, 0.0372, 0.0420], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 07:12:26,726 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([3.7728, 3.7206, 2.6731, 4.3738, 3.7788, 3.7739, 1.9181, 3.7547], device='cuda:1'), covar=tensor([0.1631, 0.1192, 0.3778, 0.1378, 0.3158, 0.1607, 0.5139, 0.2144], device='cuda:1'), in_proj_covar=tensor([0.0244, 0.0214, 0.0250, 0.0301, 0.0298, 0.0247, 0.0269, 0.0269], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 07:12:28,444 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.001e+02 1.803e+02 2.175e+02 2.734e+02 4.804e+02, threshold=4.350e+02, percent-clipped=5.0 2023-04-27 07:12:31,575 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=83552.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:12:49,825 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.07 vs. limit=5.0 2023-04-27 07:12:56,225 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.1377, 1.4637, 1.2794, 1.6975, 1.5830, 1.6878, 1.3289, 3.0265], device='cuda:1'), covar=tensor([0.0616, 0.0780, 0.0804, 0.1164, 0.0633, 0.0522, 0.0760, 0.0187], device='cuda:1'), in_proj_covar=tensor([0.0038, 0.0039, 0.0040, 0.0044, 0.0040, 0.0038, 0.0038, 0.0057], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:1') 2023-04-27 07:12:56,721 INFO [finetune.py:976] (1/7) Epoch 15, batch 3400, loss[loss=0.2141, simple_loss=0.2878, pruned_loss=0.07015, over 4892.00 frames. ], tot_loss[loss=0.1849, simple_loss=0.2544, pruned_loss=0.05769, over 954750.14 frames. ], batch size: 43, lr: 3.49e-03, grad_scale: 64.0 2023-04-27 07:13:09,194 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-04-27 07:13:18,179 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=83619.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:13:28,670 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-04-27 07:13:30,156 INFO [finetune.py:976] (1/7) Epoch 15, batch 3450, loss[loss=0.1981, simple_loss=0.2659, pruned_loss=0.06513, over 4812.00 frames. ], tot_loss[loss=0.1834, simple_loss=0.2533, pruned_loss=0.05672, over 955709.01 frames. ], batch size: 41, lr: 3.49e-03, grad_scale: 64.0 2023-04-27 07:13:31,413 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=83640.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:13:35,582 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.177e+02 1.561e+02 1.911e+02 2.345e+02 4.135e+02, threshold=3.821e+02, percent-clipped=0.0 2023-04-27 07:13:49,255 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=83667.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:14:13,155 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2023-04-27 07:14:13,518 INFO [finetune.py:976] (1/7) Epoch 15, batch 3500, loss[loss=0.1735, simple_loss=0.2324, pruned_loss=0.05726, over 4323.00 frames. ], tot_loss[loss=0.1819, simple_loss=0.2508, pruned_loss=0.05646, over 953889.83 frames. ], batch size: 65, lr: 3.49e-03, grad_scale: 64.0 2023-04-27 07:14:41,715 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5629, 0.9792, 1.2888, 1.1411, 1.6749, 1.3554, 1.1099, 1.2040], device='cuda:1'), covar=tensor([0.1756, 0.1343, 0.1739, 0.1461, 0.0759, 0.1403, 0.1627, 0.2080], device='cuda:1'), in_proj_covar=tensor([0.0310, 0.0318, 0.0354, 0.0294, 0.0332, 0.0315, 0.0307, 0.0365], device='cuda:1'), out_proj_covar=tensor([6.4194e-05, 6.6517e-05, 7.5788e-05, 6.0193e-05, 6.9202e-05, 6.6537e-05, 6.4938e-05, 7.7995e-05], device='cuda:1') 2023-04-27 07:15:04,836 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.1467, 3.1488, 2.4740, 2.8156, 2.1543, 2.6174, 2.6490, 1.7777], device='cuda:1'), covar=tensor([0.2269, 0.1345, 0.0798, 0.1238, 0.2993, 0.1218, 0.2080, 0.3221], device='cuda:1'), in_proj_covar=tensor([0.0290, 0.0310, 0.0222, 0.0284, 0.0314, 0.0266, 0.0253, 0.0270], device='cuda:1'), out_proj_covar=tensor([1.1693e-04, 1.2347e-04, 8.8469e-05, 1.1314e-04, 1.2784e-04, 1.0624e-04, 1.0252e-04, 1.0756e-04], device='cuda:1') 2023-04-27 07:15:15,482 INFO [finetune.py:976] (1/7) Epoch 15, batch 3550, loss[loss=0.155, simple_loss=0.2106, pruned_loss=0.0497, over 4371.00 frames. ], tot_loss[loss=0.1787, simple_loss=0.2472, pruned_loss=0.05508, over 952920.76 frames. ], batch size: 19, lr: 3.49e-03, grad_scale: 64.0 2023-04-27 07:15:21,402 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.049e+02 1.464e+02 1.732e+02 2.080e+02 4.644e+02, threshold=3.463e+02, percent-clipped=1.0 2023-04-27 07:15:29,358 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=83760.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:15:37,022 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.2591, 3.0015, 0.8282, 1.6669, 1.8074, 2.0828, 1.8399, 1.0114], device='cuda:1'), covar=tensor([0.1486, 0.1135, 0.2060, 0.1264, 0.1076, 0.1097, 0.1579, 0.1772], device='cuda:1'), in_proj_covar=tensor([0.0118, 0.0247, 0.0137, 0.0122, 0.0131, 0.0153, 0.0118, 0.0120], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-27 07:15:49,274 INFO [finetune.py:976] (1/7) Epoch 15, batch 3600, loss[loss=0.1606, simple_loss=0.2256, pruned_loss=0.04779, over 4197.00 frames. ], tot_loss[loss=0.1759, simple_loss=0.2439, pruned_loss=0.054, over 951615.67 frames. ], batch size: 18, lr: 3.49e-03, grad_scale: 64.0 2023-04-27 07:16:15,933 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-04-27 07:16:26,198 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=83821.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:16:55,917 INFO [finetune.py:976] (1/7) Epoch 15, batch 3650, loss[loss=0.2048, simple_loss=0.2687, pruned_loss=0.07047, over 4934.00 frames. ], tot_loss[loss=0.1794, simple_loss=0.2474, pruned_loss=0.05571, over 952819.91 frames. ], batch size: 33, lr: 3.48e-03, grad_scale: 64.0 2023-04-27 07:17:01,431 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 8.661e+01 1.633e+02 1.914e+02 2.329e+02 4.921e+02, threshold=3.827e+02, percent-clipped=4.0 2023-04-27 07:17:04,998 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=83852.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:17:29,863 INFO [finetune.py:976] (1/7) Epoch 15, batch 3700, loss[loss=0.1754, simple_loss=0.2525, pruned_loss=0.04912, over 4800.00 frames. ], tot_loss[loss=0.1818, simple_loss=0.2509, pruned_loss=0.05638, over 951452.51 frames. ], batch size: 51, lr: 3.48e-03, grad_scale: 32.0 2023-04-27 07:17:33,030 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.3010, 1.4874, 1.8266, 1.9660, 1.9041, 2.0952, 1.9113, 1.9006], device='cuda:1'), covar=tensor([0.4327, 0.5052, 0.4728, 0.4577, 0.5279, 0.7344, 0.5022, 0.4824], device='cuda:1'), in_proj_covar=tensor([0.0327, 0.0371, 0.0316, 0.0330, 0.0340, 0.0395, 0.0350, 0.0323], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 07:17:37,201 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=83900.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:18:03,516 INFO [finetune.py:976] (1/7) Epoch 15, batch 3750, loss[loss=0.1886, simple_loss=0.2613, pruned_loss=0.05799, over 4886.00 frames. ], tot_loss[loss=0.1833, simple_loss=0.2525, pruned_loss=0.05707, over 949845.53 frames. ], batch size: 35, lr: 3.48e-03, grad_scale: 32.0 2023-04-27 07:18:04,846 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=83940.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:18:09,569 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.106e+02 1.640e+02 2.069e+02 2.492e+02 6.161e+02, threshold=4.139e+02, percent-clipped=2.0 2023-04-27 07:18:10,295 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=83949.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:18:18,984 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-04-27 07:18:34,074 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=83985.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:18:36,294 INFO [finetune.py:976] (1/7) Epoch 15, batch 3800, loss[loss=0.139, simple_loss=0.2149, pruned_loss=0.03157, over 4758.00 frames. ], tot_loss[loss=0.1843, simple_loss=0.254, pruned_loss=0.05729, over 952092.12 frames. ], batch size: 28, lr: 3.48e-03, grad_scale: 32.0 2023-04-27 07:18:36,357 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=83988.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:18:38,094 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7779, 3.5939, 0.9636, 1.8453, 2.2738, 2.4872, 2.0773, 1.0801], device='cuda:1'), covar=tensor([0.1317, 0.1023, 0.2134, 0.1307, 0.0979, 0.1022, 0.1532, 0.2020], device='cuda:1'), in_proj_covar=tensor([0.0117, 0.0245, 0.0137, 0.0121, 0.0130, 0.0152, 0.0118, 0.0119], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-27 07:18:50,814 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.42 vs. limit=5.0 2023-04-27 07:18:51,946 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=84010.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 07:18:56,215 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=3.69 vs. limit=5.0 2023-04-27 07:19:07,418 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4332, 1.7585, 1.6485, 2.0614, 2.0136, 2.0798, 1.6948, 4.2701], device='cuda:1'), covar=tensor([0.0537, 0.0776, 0.0791, 0.1126, 0.0601, 0.0546, 0.0734, 0.0127], device='cuda:1'), in_proj_covar=tensor([0.0038, 0.0039, 0.0040, 0.0044, 0.0040, 0.0038, 0.0038, 0.0056], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:1') 2023-04-27 07:19:10,672 INFO [finetune.py:976] (1/7) Epoch 15, batch 3850, loss[loss=0.175, simple_loss=0.2358, pruned_loss=0.05708, over 4896.00 frames. ], tot_loss[loss=0.1815, simple_loss=0.2512, pruned_loss=0.05587, over 953565.42 frames. ], batch size: 35, lr: 3.48e-03, grad_scale: 32.0 2023-04-27 07:19:21,648 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=84046.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:19:22,756 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.118e+02 1.590e+02 1.914e+02 2.294e+02 5.295e+02, threshold=3.828e+02, percent-clipped=1.0 2023-04-27 07:19:31,925 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.48 vs. limit=2.0 2023-04-27 07:20:15,361 INFO [finetune.py:976] (1/7) Epoch 15, batch 3900, loss[loss=0.1838, simple_loss=0.2549, pruned_loss=0.05639, over 4866.00 frames. ], tot_loss[loss=0.1803, simple_loss=0.2495, pruned_loss=0.05559, over 954089.84 frames. ], batch size: 31, lr: 3.48e-03, grad_scale: 32.0 2023-04-27 07:20:27,928 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=84097.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:20:35,381 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4842, 1.3865, 1.7697, 1.7349, 1.3533, 1.2433, 1.3932, 0.9758], device='cuda:1'), covar=tensor([0.0601, 0.0704, 0.0435, 0.0577, 0.0844, 0.1184, 0.0618, 0.0650], device='cuda:1'), in_proj_covar=tensor([0.0069, 0.0071, 0.0070, 0.0069, 0.0077, 0.0098, 0.0076, 0.0069], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-27 07:20:39,572 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9367, 2.3860, 1.3701, 1.6468, 2.4342, 1.7752, 1.7442, 1.7871], device='cuda:1'), covar=tensor([0.0478, 0.0323, 0.0289, 0.0534, 0.0234, 0.0508, 0.0499, 0.0560], device='cuda:1'), in_proj_covar=tensor([0.0029, 0.0024, 0.0022, 0.0029, 0.0020, 0.0028, 0.0028, 0.0029], device='cuda:1'), out_proj_covar=tensor([0.0050, 0.0045, 0.0038, 0.0050, 0.0038, 0.0049, 0.0049, 0.0050], device='cuda:1') 2023-04-27 07:20:50,571 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=84116.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:21:01,302 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6537, 1.6959, 0.7769, 1.3661, 1.8325, 1.5525, 1.4505, 1.5078], device='cuda:1'), covar=tensor([0.0483, 0.0361, 0.0354, 0.0538, 0.0266, 0.0512, 0.0497, 0.0565], device='cuda:1'), in_proj_covar=tensor([0.0029, 0.0024, 0.0022, 0.0029, 0.0020, 0.0029, 0.0028, 0.0029], device='cuda:1'), out_proj_covar=tensor([0.0050, 0.0045, 0.0038, 0.0050, 0.0038, 0.0049, 0.0049, 0.0050], device='cuda:1') 2023-04-27 07:21:20,858 INFO [finetune.py:976] (1/7) Epoch 15, batch 3950, loss[loss=0.1796, simple_loss=0.2405, pruned_loss=0.05939, over 4918.00 frames. ], tot_loss[loss=0.1777, simple_loss=0.246, pruned_loss=0.05468, over 954556.52 frames. ], batch size: 43, lr: 3.48e-03, grad_scale: 32.0 2023-04-27 07:21:34,733 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 8.122e+01 1.668e+02 1.985e+02 2.633e+02 4.581e+02, threshold=3.971e+02, percent-clipped=4.0 2023-04-27 07:21:41,491 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=84158.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:22:01,414 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.50 vs. limit=2.0 2023-04-27 07:22:12,114 INFO [finetune.py:976] (1/7) Epoch 15, batch 4000, loss[loss=0.2031, simple_loss=0.2755, pruned_loss=0.0653, over 4870.00 frames. ], tot_loss[loss=0.1781, simple_loss=0.2461, pruned_loss=0.05501, over 951706.79 frames. ], batch size: 34, lr: 3.48e-03, grad_scale: 32.0 2023-04-27 07:22:31,715 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=84199.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:22:35,151 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5186, 1.0901, 1.3047, 1.8972, 1.9589, 1.5573, 1.1929, 1.6885], device='cuda:1'), covar=tensor([0.0954, 0.1751, 0.1040, 0.0648, 0.0625, 0.0894, 0.0970, 0.0662], device='cuda:1'), in_proj_covar=tensor([0.0188, 0.0200, 0.0179, 0.0170, 0.0176, 0.0180, 0.0152, 0.0177], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 07:22:40,701 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-04-27 07:22:41,890 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9763, 1.4721, 1.9584, 2.3549, 2.0115, 1.9008, 1.9455, 1.9535], device='cuda:1'), covar=tensor([0.4632, 0.6542, 0.6180, 0.5694, 0.6288, 0.8154, 0.7798, 0.7210], device='cuda:1'), in_proj_covar=tensor([0.0414, 0.0404, 0.0492, 0.0505, 0.0444, 0.0466, 0.0473, 0.0476], device='cuda:1'), out_proj_covar=tensor([9.9995e-05, 9.9913e-05, 1.1076e-04, 1.2020e-04, 1.0665e-04, 1.1213e-04, 1.1257e-04, 1.1305e-04], device='cuda:1') 2023-04-27 07:23:00,607 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=84236.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:23:02,206 INFO [finetune.py:976] (1/7) Epoch 15, batch 4050, loss[loss=0.1888, simple_loss=0.2495, pruned_loss=0.06405, over 4720.00 frames. ], tot_loss[loss=0.1803, simple_loss=0.2486, pruned_loss=0.05594, over 950240.18 frames. ], batch size: 23, lr: 3.48e-03, grad_scale: 32.0 2023-04-27 07:23:09,757 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.868e+01 1.686e+02 2.049e+02 2.449e+02 3.463e+02, threshold=4.099e+02, percent-clipped=0.0 2023-04-27 07:23:18,197 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=84260.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:23:35,692 INFO [finetune.py:976] (1/7) Epoch 15, batch 4100, loss[loss=0.1866, simple_loss=0.2441, pruned_loss=0.06456, over 4892.00 frames. ], tot_loss[loss=0.1829, simple_loss=0.2523, pruned_loss=0.05672, over 950929.18 frames. ], batch size: 35, lr: 3.48e-03, grad_scale: 32.0 2023-04-27 07:23:42,181 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=84297.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:23:48,045 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=84305.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 07:24:00,032 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6699, 1.4446, 1.8319, 1.8636, 1.4457, 1.3326, 1.5003, 1.0574], device='cuda:1'), covar=tensor([0.0527, 0.0729, 0.0416, 0.0523, 0.0752, 0.1151, 0.0566, 0.0626], device='cuda:1'), in_proj_covar=tensor([0.0068, 0.0071, 0.0070, 0.0068, 0.0076, 0.0097, 0.0076, 0.0069], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-27 07:24:08,387 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=84336.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:24:09,505 INFO [finetune.py:976] (1/7) Epoch 15, batch 4150, loss[loss=0.1941, simple_loss=0.2743, pruned_loss=0.0569, over 4822.00 frames. ], tot_loss[loss=0.184, simple_loss=0.2539, pruned_loss=0.05708, over 953965.03 frames. ], batch size: 47, lr: 3.48e-03, grad_scale: 32.0 2023-04-27 07:24:09,645 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.1546, 1.5551, 2.0601, 2.6213, 2.0246, 1.5915, 1.3205, 1.9145], device='cuda:1'), covar=tensor([0.3324, 0.3503, 0.1617, 0.2271, 0.2816, 0.2712, 0.4473, 0.2320], device='cuda:1'), in_proj_covar=tensor([0.0292, 0.0248, 0.0226, 0.0318, 0.0217, 0.0230, 0.0231, 0.0186], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-27 07:24:11,370 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=84341.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:24:16,002 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.201e+02 1.658e+02 1.986e+02 2.406e+02 7.031e+02, threshold=3.971e+02, percent-clipped=3.0 2023-04-27 07:24:27,249 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.11 vs. limit=2.0 2023-04-27 07:24:33,136 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6518, 1.4111, 1.2474, 1.3964, 1.9247, 1.5101, 1.2788, 1.1935], device='cuda:1'), covar=tensor([0.1611, 0.1221, 0.2004, 0.1328, 0.0681, 0.1297, 0.1876, 0.1978], device='cuda:1'), in_proj_covar=tensor([0.0310, 0.0319, 0.0355, 0.0294, 0.0333, 0.0316, 0.0306, 0.0366], device='cuda:1'), out_proj_covar=tensor([6.4326e-05, 6.6661e-05, 7.5950e-05, 6.0009e-05, 6.9282e-05, 6.6737e-05, 6.4668e-05, 7.8073e-05], device='cuda:1') 2023-04-27 07:24:43,282 INFO [finetune.py:976] (1/7) Epoch 15, batch 4200, loss[loss=0.153, simple_loss=0.2193, pruned_loss=0.04338, over 4750.00 frames. ], tot_loss[loss=0.1841, simple_loss=0.2543, pruned_loss=0.05696, over 954017.38 frames. ], batch size: 26, lr: 3.48e-03, grad_scale: 32.0 2023-04-27 07:24:49,273 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=84397.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:25:03,289 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=84416.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:25:10,689 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 2023-04-27 07:25:21,868 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8923, 1.6943, 1.9725, 2.3075, 2.2486, 1.9562, 1.5821, 2.1316], device='cuda:1'), covar=tensor([0.0826, 0.1104, 0.0649, 0.0536, 0.0578, 0.0726, 0.0755, 0.0482], device='cuda:1'), in_proj_covar=tensor([0.0188, 0.0200, 0.0179, 0.0170, 0.0175, 0.0181, 0.0152, 0.0177], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 07:25:22,961 INFO [finetune.py:976] (1/7) Epoch 15, batch 4250, loss[loss=0.2021, simple_loss=0.2637, pruned_loss=0.07025, over 4816.00 frames. ], tot_loss[loss=0.1833, simple_loss=0.2527, pruned_loss=0.05698, over 955367.30 frames. ], batch size: 41, lr: 3.48e-03, grad_scale: 32.0 2023-04-27 07:25:33,108 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.033e+02 1.587e+02 1.989e+02 2.520e+02 5.183e+02, threshold=3.979e+02, percent-clipped=3.0 2023-04-27 07:25:42,003 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=84453.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:25:52,491 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.3508, 1.6345, 1.5081, 2.1491, 2.3442, 1.9584, 1.8459, 1.5874], device='cuda:1'), covar=tensor([0.1701, 0.2032, 0.1841, 0.1757, 0.1196, 0.1933, 0.2467, 0.2378], device='cuda:1'), in_proj_covar=tensor([0.0309, 0.0318, 0.0354, 0.0293, 0.0332, 0.0314, 0.0304, 0.0365], device='cuda:1'), out_proj_covar=tensor([6.4075e-05, 6.6575e-05, 7.5770e-05, 5.9764e-05, 6.9052e-05, 6.6437e-05, 6.4410e-05, 7.7996e-05], device='cuda:1') 2023-04-27 07:25:53,799 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-04-27 07:25:56,126 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=84464.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:26:28,012 INFO [finetune.py:976] (1/7) Epoch 15, batch 4300, loss[loss=0.1742, simple_loss=0.2449, pruned_loss=0.05177, over 4935.00 frames. ], tot_loss[loss=0.1809, simple_loss=0.2498, pruned_loss=0.05595, over 957441.28 frames. ], batch size: 33, lr: 3.48e-03, grad_scale: 32.0 2023-04-27 07:26:47,752 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-04-27 07:27:18,998 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=84525.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:27:32,533 INFO [finetune.py:976] (1/7) Epoch 15, batch 4350, loss[loss=0.2057, simple_loss=0.2526, pruned_loss=0.07941, over 4343.00 frames. ], tot_loss[loss=0.1779, simple_loss=0.2464, pruned_loss=0.05475, over 956726.56 frames. ], batch size: 19, lr: 3.48e-03, grad_scale: 32.0 2023-04-27 07:27:44,743 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.129e+02 1.607e+02 1.900e+02 2.324e+02 5.232e+02, threshold=3.801e+02, percent-clipped=2.0 2023-04-27 07:27:54,645 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=84555.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:28:26,830 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=84586.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:28:27,950 INFO [finetune.py:976] (1/7) Epoch 15, batch 4400, loss[loss=0.2003, simple_loss=0.2828, pruned_loss=0.05893, over 4844.00 frames. ], tot_loss[loss=0.1794, simple_loss=0.2478, pruned_loss=0.05548, over 957124.48 frames. ], batch size: 49, lr: 3.48e-03, grad_scale: 32.0 2023-04-27 07:28:30,485 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=84592.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:28:37,876 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.0709, 0.7516, 0.9250, 0.8004, 1.2276, 0.9880, 0.8991, 0.9226], device='cuda:1'), covar=tensor([0.2388, 0.1691, 0.2335, 0.2087, 0.1141, 0.1561, 0.2045, 0.2574], device='cuda:1'), in_proj_covar=tensor([0.0310, 0.0318, 0.0354, 0.0293, 0.0332, 0.0314, 0.0305, 0.0366], device='cuda:1'), out_proj_covar=tensor([6.4216e-05, 6.6521e-05, 7.5837e-05, 5.9780e-05, 6.9259e-05, 6.6304e-05, 6.4446e-05, 7.8219e-05], device='cuda:1') 2023-04-27 07:28:39,066 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=84605.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:28:59,194 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=84633.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:29:02,146 INFO [finetune.py:976] (1/7) Epoch 15, batch 4450, loss[loss=0.1987, simple_loss=0.2699, pruned_loss=0.06378, over 4918.00 frames. ], tot_loss[loss=0.1822, simple_loss=0.2512, pruned_loss=0.05654, over 954452.28 frames. ], batch size: 36, lr: 3.48e-03, grad_scale: 32.0 2023-04-27 07:29:04,109 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=84641.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:29:08,290 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.195e+02 1.631e+02 1.866e+02 2.330e+02 4.316e+02, threshold=3.732e+02, percent-clipped=2.0 2023-04-27 07:29:11,996 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=84653.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:29:36,301 INFO [finetune.py:976] (1/7) Epoch 15, batch 4500, loss[loss=0.1726, simple_loss=0.2448, pruned_loss=0.05026, over 4824.00 frames. ], tot_loss[loss=0.1815, simple_loss=0.2514, pruned_loss=0.05583, over 954201.96 frames. ], batch size: 39, lr: 3.48e-03, grad_scale: 32.0 2023-04-27 07:29:37,003 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=84689.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:29:38,815 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=84692.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:29:40,083 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=84694.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:29:45,376 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5757, 0.6868, 1.4681, 1.9520, 1.6519, 1.4789, 1.5185, 1.5266], device='cuda:1'), covar=tensor([0.4520, 0.6950, 0.6398, 0.6480, 0.6142, 0.7663, 0.7805, 0.7904], device='cuda:1'), in_proj_covar=tensor([0.0416, 0.0407, 0.0493, 0.0507, 0.0446, 0.0467, 0.0476, 0.0478], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 07:30:09,976 INFO [finetune.py:976] (1/7) Epoch 15, batch 4550, loss[loss=0.2035, simple_loss=0.2741, pruned_loss=0.06642, over 4821.00 frames. ], tot_loss[loss=0.184, simple_loss=0.2535, pruned_loss=0.05729, over 953166.07 frames. ], batch size: 33, lr: 3.48e-03, grad_scale: 32.0 2023-04-27 07:30:16,091 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.003e+02 1.700e+02 1.997e+02 2.528e+02 3.461e+02, threshold=3.994e+02, percent-clipped=0.0 2023-04-27 07:30:19,264 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=84753.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:30:19,295 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.2176, 2.1507, 1.8007, 1.8125, 2.2159, 1.8724, 2.7497, 1.5449], device='cuda:1'), covar=tensor([0.3463, 0.1939, 0.4257, 0.2993, 0.1723, 0.2404, 0.1390, 0.4294], device='cuda:1'), in_proj_covar=tensor([0.0343, 0.0347, 0.0427, 0.0356, 0.0384, 0.0382, 0.0373, 0.0420], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 07:30:43,750 INFO [finetune.py:976] (1/7) Epoch 15, batch 4600, loss[loss=0.1453, simple_loss=0.2299, pruned_loss=0.03036, over 4864.00 frames. ], tot_loss[loss=0.1837, simple_loss=0.2533, pruned_loss=0.05701, over 954145.07 frames. ], batch size: 31, lr: 3.48e-03, grad_scale: 32.0 2023-04-27 07:30:51,273 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=84800.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:30:51,837 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=84801.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:31:21,976 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.67 vs. limit=2.0 2023-04-27 07:31:33,985 INFO [finetune.py:976] (1/7) Epoch 15, batch 4650, loss[loss=0.2271, simple_loss=0.2812, pruned_loss=0.08643, over 4849.00 frames. ], tot_loss[loss=0.182, simple_loss=0.2512, pruned_loss=0.05645, over 956055.35 frames. ], batch size: 49, lr: 3.48e-03, grad_scale: 32.0 2023-04-27 07:31:45,944 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.111e+02 1.628e+02 1.889e+02 2.261e+02 4.327e+02, threshold=3.778e+02, percent-clipped=1.0 2023-04-27 07:31:47,948 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.3548, 1.5238, 1.7713, 1.9147, 1.7947, 1.8589, 1.8293, 1.8214], device='cuda:1'), covar=tensor([0.4271, 0.5600, 0.4533, 0.4456, 0.5486, 0.7613, 0.5486, 0.5134], device='cuda:1'), in_proj_covar=tensor([0.0331, 0.0373, 0.0317, 0.0333, 0.0344, 0.0398, 0.0354, 0.0325], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 07:31:55,726 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=84855.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:31:57,664 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.2852, 1.7338, 2.2096, 2.6761, 2.2277, 1.7324, 1.4764, 1.9677], device='cuda:1'), covar=tensor([0.3530, 0.3546, 0.1683, 0.2463, 0.2681, 0.2708, 0.4443, 0.2175], device='cuda:1'), in_proj_covar=tensor([0.0293, 0.0249, 0.0226, 0.0318, 0.0218, 0.0231, 0.0232, 0.0187], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-27 07:32:05,698 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=84861.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:32:30,054 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=84881.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:32:39,702 INFO [finetune.py:976] (1/7) Epoch 15, batch 4700, loss[loss=0.1342, simple_loss=0.2024, pruned_loss=0.03307, over 4821.00 frames. ], tot_loss[loss=0.1796, simple_loss=0.2481, pruned_loss=0.05557, over 956457.84 frames. ], batch size: 41, lr: 3.48e-03, grad_scale: 32.0 2023-04-27 07:32:40,542 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.74 vs. limit=5.0 2023-04-27 07:32:42,708 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=84892.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:33:00,661 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=84903.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:33:25,802 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9260, 2.5807, 2.1342, 2.4286, 1.6765, 2.0697, 2.1056, 1.6261], device='cuda:1'), covar=tensor([0.2095, 0.1174, 0.0777, 0.1095, 0.3261, 0.1142, 0.1903, 0.2605], device='cuda:1'), in_proj_covar=tensor([0.0287, 0.0309, 0.0220, 0.0282, 0.0313, 0.0264, 0.0251, 0.0267], device='cuda:1'), out_proj_covar=tensor([1.1583e-04, 1.2299e-04, 8.7882e-05, 1.1208e-04, 1.2745e-04, 1.0528e-04, 1.0153e-04, 1.0623e-04], device='cuda:1') 2023-04-27 07:33:46,042 INFO [finetune.py:976] (1/7) Epoch 15, batch 4750, loss[loss=0.1732, simple_loss=0.2485, pruned_loss=0.04898, over 4916.00 frames. ], tot_loss[loss=0.1768, simple_loss=0.2449, pruned_loss=0.05434, over 955349.98 frames. ], batch size: 36, lr: 3.48e-03, grad_scale: 32.0 2023-04-27 07:33:47,831 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=84940.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:33:53,647 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.954e+01 1.625e+02 2.007e+02 2.365e+02 3.914e+02, threshold=4.014e+02, percent-clipped=2.0 2023-04-27 07:33:55,655 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=84951.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:33:59,226 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([3.2667, 3.2857, 2.5042, 3.8577, 3.3741, 3.3151, 1.2352, 3.3250], device='cuda:1'), covar=tensor([0.2229, 0.1353, 0.3581, 0.2249, 0.4134, 0.2061, 0.6439, 0.2749], device='cuda:1'), in_proj_covar=tensor([0.0242, 0.0212, 0.0248, 0.0300, 0.0295, 0.0245, 0.0268, 0.0268], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 07:34:19,536 INFO [finetune.py:976] (1/7) Epoch 15, batch 4800, loss[loss=0.1535, simple_loss=0.2378, pruned_loss=0.03459, over 4832.00 frames. ], tot_loss[loss=0.1791, simple_loss=0.2473, pruned_loss=0.05543, over 954729.77 frames. ], batch size: 47, lr: 3.47e-03, grad_scale: 32.0 2023-04-27 07:34:20,212 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=84989.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:34:20,321 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=3.58 vs. limit=5.0 2023-04-27 07:34:23,046 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=84992.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:34:32,848 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-04-27 07:34:36,386 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=85012.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:34:47,881 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0426, 2.6140, 1.2013, 1.3493, 1.9705, 1.1620, 3.3330, 1.6792], device='cuda:1'), covar=tensor([0.0713, 0.0595, 0.0790, 0.1363, 0.0524, 0.1137, 0.0307, 0.0687], device='cuda:1'), in_proj_covar=tensor([0.0051, 0.0066, 0.0049, 0.0047, 0.0050, 0.0052, 0.0075, 0.0052], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:1') 2023-04-27 07:34:53,702 INFO [finetune.py:976] (1/7) Epoch 15, batch 4850, loss[loss=0.1927, simple_loss=0.2768, pruned_loss=0.05431, over 4317.00 frames. ], tot_loss[loss=0.1818, simple_loss=0.251, pruned_loss=0.05631, over 955178.71 frames. ], batch size: 65, lr: 3.47e-03, grad_scale: 32.0 2023-04-27 07:34:54,928 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=85040.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:35:01,196 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.085e+02 1.698e+02 1.957e+02 2.327e+02 4.552e+02, threshold=3.914e+02, percent-clipped=3.0 2023-04-27 07:35:21,681 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-04-27 07:35:27,222 INFO [finetune.py:976] (1/7) Epoch 15, batch 4900, loss[loss=0.1736, simple_loss=0.247, pruned_loss=0.05009, over 4694.00 frames. ], tot_loss[loss=0.1831, simple_loss=0.2528, pruned_loss=0.0567, over 956495.20 frames. ], batch size: 59, lr: 3.47e-03, grad_scale: 32.0 2023-04-27 07:35:40,316 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=85107.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:35:45,204 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=85115.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:35:51,298 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6011, 0.7386, 1.4964, 1.9821, 1.6789, 1.5094, 1.5329, 1.5494], device='cuda:1'), covar=tensor([0.4585, 0.6553, 0.6047, 0.6196, 0.6097, 0.7368, 0.7583, 0.7650], device='cuda:1'), in_proj_covar=tensor([0.0416, 0.0405, 0.0492, 0.0505, 0.0444, 0.0468, 0.0474, 0.0477], device='cuda:1'), out_proj_covar=tensor([1.0027e-04, 9.9959e-05, 1.1085e-04, 1.2047e-04, 1.0676e-04, 1.1254e-04, 1.1279e-04, 1.1334e-04], device='cuda:1') 2023-04-27 07:36:00,015 INFO [finetune.py:976] (1/7) Epoch 15, batch 4950, loss[loss=0.2037, simple_loss=0.2759, pruned_loss=0.06577, over 4819.00 frames. ], tot_loss[loss=0.1853, simple_loss=0.2551, pruned_loss=0.05772, over 955268.09 frames. ], batch size: 39, lr: 3.47e-03, grad_scale: 32.0 2023-04-27 07:36:07,085 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.205e+02 1.668e+02 1.905e+02 2.304e+02 4.878e+02, threshold=3.810e+02, percent-clipped=1.0 2023-04-27 07:36:12,998 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=85156.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:36:20,330 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=85168.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 07:36:25,212 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=85176.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:36:28,190 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=85181.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:36:29,372 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=3.89 vs. limit=5.0 2023-04-27 07:36:33,339 INFO [finetune.py:976] (1/7) Epoch 15, batch 5000, loss[loss=0.1751, simple_loss=0.2466, pruned_loss=0.05176, over 4817.00 frames. ], tot_loss[loss=0.1838, simple_loss=0.2534, pruned_loss=0.05713, over 954719.39 frames. ], batch size: 30, lr: 3.47e-03, grad_scale: 32.0 2023-04-27 07:37:00,631 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=85229.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:37:12,539 INFO [finetune.py:976] (1/7) Epoch 15, batch 5050, loss[loss=0.1461, simple_loss=0.2152, pruned_loss=0.03847, over 4916.00 frames. ], tot_loss[loss=0.1819, simple_loss=0.2508, pruned_loss=0.05647, over 956516.72 frames. ], batch size: 37, lr: 3.47e-03, grad_scale: 32.0 2023-04-27 07:37:25,142 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.163e+02 1.655e+02 2.048e+02 2.440e+02 4.868e+02, threshold=4.096e+02, percent-clipped=4.0 2023-04-27 07:37:55,449 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.3738, 2.0551, 2.2988, 2.6360, 2.6995, 2.2902, 2.0435, 2.4146], device='cuda:1'), covar=tensor([0.0655, 0.0945, 0.0516, 0.0506, 0.0550, 0.0684, 0.0732, 0.0452], device='cuda:1'), in_proj_covar=tensor([0.0194, 0.0207, 0.0185, 0.0177, 0.0182, 0.0186, 0.0157, 0.0184], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 07:37:57,903 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0747, 2.6405, 0.9732, 1.3357, 2.0812, 1.2547, 3.4255, 1.7680], device='cuda:1'), covar=tensor([0.0671, 0.0618, 0.0895, 0.1292, 0.0478, 0.1005, 0.0338, 0.0621], device='cuda:1'), in_proj_covar=tensor([0.0051, 0.0066, 0.0049, 0.0047, 0.0050, 0.0052, 0.0076, 0.0052], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:1') 2023-04-27 07:38:06,581 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.3048, 1.4793, 1.7370, 1.8376, 1.7555, 1.8474, 1.8040, 1.7539], device='cuda:1'), covar=tensor([0.3908, 0.5363, 0.4460, 0.4285, 0.5504, 0.7041, 0.5219, 0.4984], device='cuda:1'), in_proj_covar=tensor([0.0329, 0.0371, 0.0318, 0.0332, 0.0343, 0.0397, 0.0353, 0.0325], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 07:38:18,000 INFO [finetune.py:976] (1/7) Epoch 15, batch 5100, loss[loss=0.1735, simple_loss=0.2358, pruned_loss=0.05562, over 4897.00 frames. ], tot_loss[loss=0.1793, simple_loss=0.2478, pruned_loss=0.05541, over 957265.55 frames. ], batch size: 32, lr: 3.47e-03, grad_scale: 32.0 2023-04-27 07:38:18,664 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=85289.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:38:41,086 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=85307.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:38:52,364 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0286, 1.4073, 1.6720, 1.6610, 2.0887, 1.8087, 1.4318, 1.5130], device='cuda:1'), covar=tensor([0.1911, 0.1514, 0.1893, 0.1332, 0.0938, 0.1407, 0.2202, 0.2236], device='cuda:1'), in_proj_covar=tensor([0.0310, 0.0319, 0.0355, 0.0293, 0.0332, 0.0314, 0.0303, 0.0367], device='cuda:1'), out_proj_covar=tensor([6.4122e-05, 6.6646e-05, 7.6009e-05, 5.9794e-05, 6.9224e-05, 6.6327e-05, 6.4075e-05, 7.8313e-05], device='cuda:1') 2023-04-27 07:39:06,452 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-04-27 07:39:06,778 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=85337.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:39:07,341 INFO [finetune.py:976] (1/7) Epoch 15, batch 5150, loss[loss=0.1632, simple_loss=0.2348, pruned_loss=0.04586, over 4768.00 frames. ], tot_loss[loss=0.1791, simple_loss=0.2473, pruned_loss=0.05544, over 955611.14 frames. ], batch size: 28, lr: 3.47e-03, grad_scale: 32.0 2023-04-27 07:39:20,279 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.171e+02 1.606e+02 1.874e+02 2.297e+02 4.541e+02, threshold=3.747e+02, percent-clipped=1.0 2023-04-27 07:39:47,578 INFO [finetune.py:976] (1/7) Epoch 15, batch 5200, loss[loss=0.225, simple_loss=0.2933, pruned_loss=0.07833, over 4901.00 frames. ], tot_loss[loss=0.1808, simple_loss=0.2496, pruned_loss=0.05595, over 954083.46 frames. ], batch size: 37, lr: 3.47e-03, grad_scale: 32.0 2023-04-27 07:40:15,432 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=85428.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:40:21,369 INFO [finetune.py:976] (1/7) Epoch 15, batch 5250, loss[loss=0.1651, simple_loss=0.2424, pruned_loss=0.04388, over 4902.00 frames. ], tot_loss[loss=0.1809, simple_loss=0.2502, pruned_loss=0.05579, over 953458.21 frames. ], batch size: 37, lr: 3.47e-03, grad_scale: 32.0 2023-04-27 07:40:23,299 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8824, 1.6359, 1.8211, 2.2411, 2.2396, 1.9066, 1.5328, 2.0984], device='cuda:1'), covar=tensor([0.0779, 0.1176, 0.0759, 0.0517, 0.0549, 0.0751, 0.0860, 0.0491], device='cuda:1'), in_proj_covar=tensor([0.0194, 0.0208, 0.0186, 0.0178, 0.0182, 0.0187, 0.0158, 0.0184], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 07:40:27,421 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.183e+02 1.584e+02 2.032e+02 2.482e+02 6.006e+02, threshold=4.063e+02, percent-clipped=4.0 2023-04-27 07:40:33,312 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=85456.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:40:36,109 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([4.4165, 4.3728, 3.0482, 5.0355, 4.4004, 4.3597, 1.6418, 4.2946], device='cuda:1'), covar=tensor([0.1472, 0.0876, 0.3615, 0.0945, 0.3065, 0.1577, 0.6135, 0.2265], device='cuda:1'), in_proj_covar=tensor([0.0243, 0.0213, 0.0250, 0.0301, 0.0295, 0.0247, 0.0270, 0.0270], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 07:40:38,400 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=85463.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 07:40:44,663 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=85471.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:40:51,976 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5479, 1.3539, 0.6155, 1.2200, 1.3767, 1.4213, 1.3167, 1.3162], device='cuda:1'), covar=tensor([0.0540, 0.0418, 0.0409, 0.0597, 0.0307, 0.0538, 0.0529, 0.0632], device='cuda:1'), in_proj_covar=tensor([0.0029, 0.0024, 0.0023, 0.0029, 0.0020, 0.0028, 0.0028, 0.0029], device='cuda:1'), out_proj_covar=tensor([0.0050, 0.0044, 0.0038, 0.0051, 0.0038, 0.0049, 0.0049, 0.0050], device='cuda:1') 2023-04-27 07:40:51,994 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=85483.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:40:54,957 INFO [finetune.py:976] (1/7) Epoch 15, batch 5300, loss[loss=0.2471, simple_loss=0.3206, pruned_loss=0.08681, over 4917.00 frames. ], tot_loss[loss=0.1814, simple_loss=0.2513, pruned_loss=0.05578, over 954446.92 frames. ], batch size: 42, lr: 3.47e-03, grad_scale: 32.0 2023-04-27 07:40:55,693 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=85489.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:41:04,891 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=85504.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:41:23,610 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=85530.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:41:28,362 INFO [finetune.py:976] (1/7) Epoch 15, batch 5350, loss[loss=0.1384, simple_loss=0.2101, pruned_loss=0.03336, over 4428.00 frames. ], tot_loss[loss=0.1817, simple_loss=0.2516, pruned_loss=0.05587, over 953452.91 frames. ], batch size: 19, lr: 3.47e-03, grad_scale: 32.0 2023-04-27 07:41:32,137 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=85544.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:41:34,431 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.243e+02 1.620e+02 1.904e+02 2.320e+02 4.507e+02, threshold=3.809e+02, percent-clipped=1.0 2023-04-27 07:41:34,750 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2023-04-27 07:41:44,053 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4033, 1.6048, 1.4878, 1.5994, 1.4172, 1.3430, 1.4035, 1.1575], device='cuda:1'), covar=tensor([0.1598, 0.1523, 0.0981, 0.1152, 0.3409, 0.1326, 0.1858, 0.2212], device='cuda:1'), in_proj_covar=tensor([0.0284, 0.0307, 0.0218, 0.0278, 0.0309, 0.0262, 0.0249, 0.0266], device='cuda:1'), out_proj_covar=tensor([1.1429e-04, 1.2198e-04, 8.7069e-05, 1.1060e-04, 1.2593e-04, 1.0432e-04, 1.0072e-04, 1.0607e-04], device='cuda:1') 2023-04-27 07:42:02,174 INFO [finetune.py:976] (1/7) Epoch 15, batch 5400, loss[loss=0.1645, simple_loss=0.2302, pruned_loss=0.04941, over 4821.00 frames. ], tot_loss[loss=0.1812, simple_loss=0.2504, pruned_loss=0.05597, over 955911.09 frames. ], batch size: 33, lr: 3.47e-03, grad_scale: 32.0 2023-04-27 07:42:04,150 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=85591.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:42:13,923 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=85607.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:42:27,861 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=85625.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:42:41,671 INFO [finetune.py:976] (1/7) Epoch 15, batch 5450, loss[loss=0.1418, simple_loss=0.2052, pruned_loss=0.03919, over 4419.00 frames. ], tot_loss[loss=0.179, simple_loss=0.248, pruned_loss=0.05503, over 957219.38 frames. ], batch size: 19, lr: 3.47e-03, grad_scale: 32.0 2023-04-27 07:42:53,051 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.070e+02 1.619e+02 1.980e+02 2.331e+02 6.511e+02, threshold=3.961e+02, percent-clipped=1.0 2023-04-27 07:42:57,387 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=85655.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:43:45,387 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=85686.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 07:43:46,489 INFO [finetune.py:976] (1/7) Epoch 15, batch 5500, loss[loss=0.1593, simple_loss=0.2416, pruned_loss=0.03851, over 4770.00 frames. ], tot_loss[loss=0.1772, simple_loss=0.2458, pruned_loss=0.05431, over 955738.80 frames. ], batch size: 27, lr: 3.47e-03, grad_scale: 32.0 2023-04-27 07:44:17,781 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.4498, 1.7709, 1.8169, 1.9404, 1.7565, 1.8748, 1.9120, 1.8565], device='cuda:1'), covar=tensor([0.4294, 0.5860, 0.5169, 0.4717, 0.5937, 0.7787, 0.6342, 0.5441], device='cuda:1'), in_proj_covar=tensor([0.0331, 0.0374, 0.0319, 0.0333, 0.0344, 0.0400, 0.0354, 0.0326], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 07:44:51,824 INFO [finetune.py:976] (1/7) Epoch 15, batch 5550, loss[loss=0.2271, simple_loss=0.2726, pruned_loss=0.09082, over 4691.00 frames. ], tot_loss[loss=0.1784, simple_loss=0.2471, pruned_loss=0.05488, over 951834.05 frames. ], batch size: 23, lr: 3.47e-03, grad_scale: 32.0 2023-04-27 07:44:54,957 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6542, 1.4490, 0.7475, 1.3424, 1.6179, 1.5373, 1.4478, 1.4552], device='cuda:1'), covar=tensor([0.0500, 0.0404, 0.0375, 0.0578, 0.0280, 0.0504, 0.0516, 0.0571], device='cuda:1'), in_proj_covar=tensor([0.0029, 0.0024, 0.0023, 0.0029, 0.0020, 0.0028, 0.0028, 0.0029], device='cuda:1'), out_proj_covar=tensor([0.0050, 0.0045, 0.0038, 0.0051, 0.0038, 0.0049, 0.0049, 0.0050], device='cuda:1') 2023-04-27 07:45:02,123 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.106e+02 1.632e+02 1.848e+02 2.307e+02 4.705e+02, threshold=3.696e+02, percent-clipped=4.0 2023-04-27 07:45:10,252 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7694, 1.4658, 1.9222, 2.3240, 1.8624, 1.7099, 1.8440, 1.7942], device='cuda:1'), covar=tensor([0.5002, 0.6859, 0.6922, 0.5844, 0.6396, 0.8416, 0.8508, 0.9079], device='cuda:1'), in_proj_covar=tensor([0.0416, 0.0406, 0.0493, 0.0504, 0.0444, 0.0467, 0.0474, 0.0477], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 07:45:11,426 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=85763.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:45:12,676 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5226, 1.8803, 1.9060, 2.3884, 2.0652, 2.2864, 1.7648, 4.6902], device='cuda:1'), covar=tensor([0.0543, 0.0769, 0.0752, 0.1109, 0.0627, 0.0445, 0.0731, 0.0083], device='cuda:1'), in_proj_covar=tensor([0.0038, 0.0039, 0.0040, 0.0044, 0.0040, 0.0038, 0.0038, 0.0056], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:1') 2023-04-27 07:45:16,812 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=85771.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:45:24,798 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=85784.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:45:27,127 INFO [finetune.py:976] (1/7) Epoch 15, batch 5600, loss[loss=0.1446, simple_loss=0.2346, pruned_loss=0.02723, over 4801.00 frames. ], tot_loss[loss=0.1792, simple_loss=0.2489, pruned_loss=0.05473, over 952032.35 frames. ], batch size: 51, lr: 3.47e-03, grad_scale: 32.0 2023-04-27 07:45:30,208 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.78 vs. limit=2.0 2023-04-27 07:45:35,575 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-04-27 07:45:40,676 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=85811.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:45:45,400 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=85819.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:45:57,425 INFO [finetune.py:976] (1/7) Epoch 15, batch 5650, loss[loss=0.2405, simple_loss=0.3024, pruned_loss=0.08932, over 4809.00 frames. ], tot_loss[loss=0.1818, simple_loss=0.2522, pruned_loss=0.05575, over 952157.49 frames. ], batch size: 38, lr: 3.47e-03, grad_scale: 32.0 2023-04-27 07:45:58,080 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=85839.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:46:03,466 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.035e+02 1.549e+02 1.787e+02 2.141e+02 3.216e+02, threshold=3.573e+02, percent-clipped=0.0 2023-04-27 07:46:19,878 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=85875.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:46:26,500 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=85886.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:46:27,658 INFO [finetune.py:976] (1/7) Epoch 15, batch 5700, loss[loss=0.189, simple_loss=0.2315, pruned_loss=0.07322, over 4578.00 frames. ], tot_loss[loss=0.1804, simple_loss=0.2498, pruned_loss=0.05551, over 937860.41 frames. ], batch size: 20, lr: 3.47e-03, grad_scale: 64.0 2023-04-27 07:46:59,231 INFO [finetune.py:976] (1/7) Epoch 16, batch 0, loss[loss=0.2317, simple_loss=0.2859, pruned_loss=0.08871, over 4204.00 frames. ], tot_loss[loss=0.2317, simple_loss=0.2859, pruned_loss=0.08871, over 4204.00 frames. ], batch size: 66, lr: 3.46e-03, grad_scale: 64.0 2023-04-27 07:46:59,231 INFO [finetune.py:1001] (1/7) Computing validation loss 2023-04-27 07:47:06,306 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.1166, 2.5252, 0.9837, 1.4087, 1.9218, 1.2969, 2.9972, 1.6895], device='cuda:1'), covar=tensor([0.0571, 0.0506, 0.0737, 0.1215, 0.0434, 0.0907, 0.0254, 0.0593], device='cuda:1'), in_proj_covar=tensor([0.0051, 0.0066, 0.0048, 0.0046, 0.0050, 0.0052, 0.0075, 0.0051], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:1') 2023-04-27 07:47:15,737 INFO [finetune.py:1010] (1/7) Epoch 16, validation: loss=0.1534, simple_loss=0.2252, pruned_loss=0.04076, over 2265189.00 frames. 2023-04-27 07:47:15,737 INFO [finetune.py:1011] (1/7) Maximum memory allocated so far is 6391MB 2023-04-27 07:47:30,495 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=85930.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 07:47:34,160 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=85936.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:47:41,352 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.024e+02 1.546e+02 1.841e+02 2.213e+02 4.481e+02, threshold=3.682e+02, percent-clipped=3.0 2023-04-27 07:47:43,308 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9871, 4.2036, 0.7710, 2.2549, 2.3521, 2.5698, 2.5454, 0.9520], device='cuda:1'), covar=tensor([0.1143, 0.0633, 0.2078, 0.1138, 0.0854, 0.1064, 0.1183, 0.2089], device='cuda:1'), in_proj_covar=tensor([0.0117, 0.0244, 0.0136, 0.0121, 0.0131, 0.0152, 0.0118, 0.0120], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-27 07:47:46,968 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=85957.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:47:53,857 INFO [finetune.py:976] (1/7) Epoch 16, batch 50, loss[loss=0.1501, simple_loss=0.2251, pruned_loss=0.03757, over 4816.00 frames. ], tot_loss[loss=0.1845, simple_loss=0.2527, pruned_loss=0.05819, over 214284.07 frames. ], batch size: 47, lr: 3.46e-03, grad_scale: 64.0 2023-04-27 07:48:04,122 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=85981.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 07:48:10,279 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=85991.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 07:48:23,358 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=86002.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:48:44,015 INFO [finetune.py:976] (1/7) Epoch 16, batch 100, loss[loss=0.1986, simple_loss=0.2573, pruned_loss=0.06989, over 4855.00 frames. ], tot_loss[loss=0.1813, simple_loss=0.2497, pruned_loss=0.05647, over 381159.68 frames. ], batch size: 31, lr: 3.46e-03, grad_scale: 64.0 2023-04-27 07:48:45,850 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=86018.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:49:27,105 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.062e+02 1.526e+02 1.932e+02 2.355e+02 3.544e+02, threshold=3.863e+02, percent-clipped=0.0 2023-04-27 07:49:34,863 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=86052.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:49:47,407 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=86063.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:49:49,096 INFO [finetune.py:976] (1/7) Epoch 16, batch 150, loss[loss=0.1484, simple_loss=0.2279, pruned_loss=0.0345, over 4766.00 frames. ], tot_loss[loss=0.178, simple_loss=0.2451, pruned_loss=0.05543, over 506398.95 frames. ], batch size: 28, lr: 3.46e-03, grad_scale: 64.0 2023-04-27 07:49:52,047 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=86069.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:50:02,661 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=86084.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:50:13,591 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.2982, 1.5424, 1.4138, 1.7514, 1.6748, 1.8364, 1.3999, 3.3559], device='cuda:1'), covar=tensor([0.0587, 0.0776, 0.0787, 0.1168, 0.0621, 0.0501, 0.0709, 0.0132], device='cuda:1'), in_proj_covar=tensor([0.0037, 0.0039, 0.0040, 0.0043, 0.0040, 0.0038, 0.0038, 0.0056], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:1') 2023-04-27 07:50:37,989 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=86113.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:50:45,305 INFO [finetune.py:976] (1/7) Epoch 16, batch 200, loss[loss=0.2099, simple_loss=0.2707, pruned_loss=0.0745, over 4788.00 frames. ], tot_loss[loss=0.178, simple_loss=0.2448, pruned_loss=0.05564, over 605774.06 frames. ], batch size: 51, lr: 3.46e-03, grad_scale: 64.0 2023-04-27 07:51:06,751 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=86130.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:51:07,905 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=86132.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:51:16,787 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=86139.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:51:22,243 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.150e+02 1.564e+02 1.820e+02 2.219e+02 4.567e+02, threshold=3.639e+02, percent-clipped=1.0 2023-04-27 07:51:33,767 INFO [finetune.py:976] (1/7) Epoch 16, batch 250, loss[loss=0.2014, simple_loss=0.2645, pruned_loss=0.06918, over 4808.00 frames. ], tot_loss[loss=0.1787, simple_loss=0.2464, pruned_loss=0.05544, over 683280.40 frames. ], batch size: 45, lr: 3.46e-03, grad_scale: 64.0 2023-04-27 07:51:41,368 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7945, 1.5170, 1.7257, 2.1961, 2.0767, 1.6955, 1.5344, 1.9716], device='cuda:1'), covar=tensor([0.0852, 0.1239, 0.0782, 0.0597, 0.0633, 0.0905, 0.0788, 0.0595], device='cuda:1'), in_proj_covar=tensor([0.0191, 0.0205, 0.0183, 0.0175, 0.0179, 0.0185, 0.0155, 0.0181], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 07:51:48,354 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=86186.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:51:48,948 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=86187.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:51:52,580 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5934, 1.4589, 4.3413, 4.0983, 3.8138, 4.1370, 3.9648, 3.8088], device='cuda:1'), covar=tensor([0.7151, 0.5654, 0.1011, 0.1525, 0.1099, 0.1483, 0.1781, 0.1461], device='cuda:1'), in_proj_covar=tensor([0.0312, 0.0306, 0.0401, 0.0408, 0.0347, 0.0406, 0.0311, 0.0366], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-27 07:52:06,459 INFO [finetune.py:976] (1/7) Epoch 16, batch 300, loss[loss=0.1341, simple_loss=0.2, pruned_loss=0.03412, over 4802.00 frames. ], tot_loss[loss=0.1825, simple_loss=0.251, pruned_loss=0.05696, over 743081.46 frames. ], batch size: 25, lr: 3.46e-03, grad_scale: 64.0 2023-04-27 07:52:17,508 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=86231.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:52:19,821 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=86234.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:52:20,588 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.81 vs. limit=5.0 2023-04-27 07:52:27,210 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=86246.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:52:28,286 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.076e+02 1.717e+02 1.966e+02 2.307e+02 3.761e+02, threshold=3.931e+02, percent-clipped=1.0 2023-04-27 07:52:30,217 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([3.7124, 3.6693, 2.8211, 4.2884, 3.7353, 3.7163, 1.6092, 3.6762], device='cuda:1'), covar=tensor([0.1728, 0.1128, 0.3372, 0.1504, 0.2755, 0.1751, 0.5667, 0.2305], device='cuda:1'), in_proj_covar=tensor([0.0244, 0.0214, 0.0251, 0.0304, 0.0297, 0.0247, 0.0271, 0.0270], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 07:52:39,294 INFO [finetune.py:976] (1/7) Epoch 16, batch 350, loss[loss=0.2028, simple_loss=0.2797, pruned_loss=0.06298, over 4910.00 frames. ], tot_loss[loss=0.1844, simple_loss=0.2541, pruned_loss=0.05733, over 791318.89 frames. ], batch size: 36, lr: 3.46e-03, grad_scale: 32.0 2023-04-27 07:52:45,433 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.4024, 1.6049, 1.6324, 2.2241, 2.4232, 1.9017, 1.9267, 1.7200], device='cuda:1'), covar=tensor([0.2378, 0.2115, 0.2093, 0.1802, 0.1390, 0.2279, 0.2327, 0.2383], device='cuda:1'), in_proj_covar=tensor([0.0308, 0.0316, 0.0353, 0.0292, 0.0332, 0.0313, 0.0301, 0.0364], device='cuda:1'), out_proj_covar=tensor([6.3701e-05, 6.6055e-05, 7.5404e-05, 5.9583e-05, 6.9171e-05, 6.6080e-05, 6.3631e-05, 7.7644e-05], device='cuda:1') 2023-04-27 07:52:50,053 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=86281.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:52:54,334 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=86286.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 07:53:07,229 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=86307.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:53:10,809 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=86313.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:53:12,610 INFO [finetune.py:976] (1/7) Epoch 16, batch 400, loss[loss=0.1532, simple_loss=0.2322, pruned_loss=0.03714, over 4843.00 frames. ], tot_loss[loss=0.1827, simple_loss=0.2536, pruned_loss=0.05588, over 828106.73 frames. ], batch size: 44, lr: 3.46e-03, grad_scale: 32.0 2023-04-27 07:53:13,318 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0837, 2.5569, 0.9967, 1.4422, 2.0893, 1.1506, 3.3950, 1.6199], device='cuda:1'), covar=tensor([0.0677, 0.0772, 0.0814, 0.1231, 0.0439, 0.1013, 0.0243, 0.0610], device='cuda:1'), in_proj_covar=tensor([0.0051, 0.0066, 0.0048, 0.0046, 0.0050, 0.0052, 0.0075, 0.0051], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:1') 2023-04-27 07:53:21,090 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=86329.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:53:35,103 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7452, 1.3489, 1.3619, 1.5055, 1.9887, 1.5399, 1.2846, 1.2914], device='cuda:1'), covar=tensor([0.1515, 0.1471, 0.1885, 0.1358, 0.0653, 0.1433, 0.2037, 0.2104], device='cuda:1'), in_proj_covar=tensor([0.0306, 0.0314, 0.0351, 0.0290, 0.0330, 0.0311, 0.0299, 0.0361], device='cuda:1'), out_proj_covar=tensor([6.3224e-05, 6.5552e-05, 7.4937e-05, 5.9137e-05, 6.8735e-05, 6.5718e-05, 6.3172e-05, 7.6970e-05], device='cuda:1') 2023-04-27 07:53:35,566 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.079e+02 1.581e+02 1.858e+02 2.136e+02 3.923e+02, threshold=3.716e+02, percent-clipped=0.0 2023-04-27 07:53:41,146 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=86358.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:53:45,900 INFO [finetune.py:976] (1/7) Epoch 16, batch 450, loss[loss=0.1655, simple_loss=0.2286, pruned_loss=0.05115, over 4819.00 frames. ], tot_loss[loss=0.1801, simple_loss=0.2506, pruned_loss=0.05479, over 856128.41 frames. ], batch size: 40, lr: 3.46e-03, grad_scale: 32.0 2023-04-27 07:54:14,885 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=86408.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:54:19,720 INFO [finetune.py:976] (1/7) Epoch 16, batch 500, loss[loss=0.1325, simple_loss=0.2032, pruned_loss=0.03087, over 4850.00 frames. ], tot_loss[loss=0.1787, simple_loss=0.2482, pruned_loss=0.05456, over 876526.77 frames. ], batch size: 49, lr: 3.46e-03, grad_scale: 32.0 2023-04-27 07:54:25,188 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=86425.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:54:42,230 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=86440.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:54:45,741 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=86445.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:54:52,711 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.729e+01 1.556e+02 1.970e+02 2.447e+02 5.301e+02, threshold=3.940e+02, percent-clipped=3.0 2023-04-27 07:55:08,757 INFO [finetune.py:976] (1/7) Epoch 16, batch 550, loss[loss=0.1473, simple_loss=0.2272, pruned_loss=0.03364, over 4783.00 frames. ], tot_loss[loss=0.1779, simple_loss=0.2465, pruned_loss=0.05468, over 890075.86 frames. ], batch size: 26, lr: 3.46e-03, grad_scale: 32.0 2023-04-27 07:56:00,650 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=86501.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:56:09,576 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=86506.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 07:56:20,908 INFO [finetune.py:976] (1/7) Epoch 16, batch 600, loss[loss=0.2275, simple_loss=0.2955, pruned_loss=0.07972, over 4846.00 frames. ], tot_loss[loss=0.1779, simple_loss=0.2464, pruned_loss=0.05466, over 903608.96 frames. ], batch size: 47, lr: 3.46e-03, grad_scale: 32.0 2023-04-27 07:56:35,186 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=86531.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:56:53,562 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0138, 2.5271, 1.0060, 1.3326, 2.1766, 1.2063, 3.2896, 1.5829], device='cuda:1'), covar=tensor([0.0719, 0.0664, 0.0863, 0.1364, 0.0463, 0.1107, 0.0281, 0.0747], device='cuda:1'), in_proj_covar=tensor([0.0051, 0.0066, 0.0048, 0.0046, 0.0050, 0.0052, 0.0075, 0.0051], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0007, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:1') 2023-04-27 07:56:54,677 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.044e+02 1.714e+02 1.969e+02 2.421e+02 4.960e+02, threshold=3.938e+02, percent-clipped=3.0 2023-04-27 07:57:04,540 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4837, 1.4040, 0.5921, 1.1716, 1.3563, 1.3611, 1.2522, 1.2719], device='cuda:1'), covar=tensor([0.0495, 0.0393, 0.0400, 0.0562, 0.0317, 0.0508, 0.0518, 0.0591], device='cuda:1'), in_proj_covar=tensor([0.0029, 0.0024, 0.0022, 0.0029, 0.0020, 0.0028, 0.0028, 0.0028], device='cuda:1'), out_proj_covar=tensor([0.0050, 0.0044, 0.0038, 0.0050, 0.0038, 0.0049, 0.0049, 0.0050], device='cuda:1') 2023-04-27 07:57:05,048 INFO [finetune.py:976] (1/7) Epoch 16, batch 650, loss[loss=0.1988, simple_loss=0.2785, pruned_loss=0.0595, over 4806.00 frames. ], tot_loss[loss=0.1795, simple_loss=0.2487, pruned_loss=0.05514, over 914278.97 frames. ], batch size: 45, lr: 3.46e-03, grad_scale: 32.0 2023-04-27 07:57:12,975 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=86579.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:57:13,045 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5366, 1.4461, 1.7887, 1.7433, 1.3751, 1.2370, 1.5203, 1.0255], device='cuda:1'), covar=tensor([0.0539, 0.0660, 0.0414, 0.0755, 0.0945, 0.1228, 0.0687, 0.0729], device='cuda:1'), in_proj_covar=tensor([0.0068, 0.0070, 0.0070, 0.0068, 0.0076, 0.0097, 0.0076, 0.0069], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-27 07:57:13,605 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=86580.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:57:15,466 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7532, 2.1901, 1.6054, 1.3870, 1.3267, 1.3235, 1.5653, 1.2449], device='cuda:1'), covar=tensor([0.1523, 0.1211, 0.1444, 0.1739, 0.2118, 0.1912, 0.1014, 0.1857], device='cuda:1'), in_proj_covar=tensor([0.0195, 0.0212, 0.0168, 0.0205, 0.0200, 0.0184, 0.0155, 0.0185], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-27 07:57:18,205 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=86586.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 07:57:29,904 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=86602.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:57:36,592 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=86613.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:57:38,331 INFO [finetune.py:976] (1/7) Epoch 16, batch 700, loss[loss=0.191, simple_loss=0.2707, pruned_loss=0.05564, over 4914.00 frames. ], tot_loss[loss=0.1804, simple_loss=0.2502, pruned_loss=0.0553, over 922837.43 frames. ], batch size: 42, lr: 3.46e-03, grad_scale: 32.0 2023-04-27 07:57:49,357 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=86634.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 07:57:54,596 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=86641.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:58:00,292 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.046e+02 1.513e+02 1.841e+02 2.164e+02 3.453e+02, threshold=3.681e+02, percent-clipped=0.0 2023-04-27 07:58:06,420 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=86658.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:58:08,175 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=86661.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:58:11,206 INFO [finetune.py:976] (1/7) Epoch 16, batch 750, loss[loss=0.18, simple_loss=0.2505, pruned_loss=0.05477, over 4796.00 frames. ], tot_loss[loss=0.1825, simple_loss=0.2525, pruned_loss=0.05625, over 929518.36 frames. ], batch size: 45, lr: 3.46e-03, grad_scale: 32.0 2023-04-27 07:58:38,828 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=86706.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:58:40,088 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=86708.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:58:44,816 INFO [finetune.py:976] (1/7) Epoch 16, batch 800, loss[loss=0.1548, simple_loss=0.2279, pruned_loss=0.04088, over 4896.00 frames. ], tot_loss[loss=0.1812, simple_loss=0.2518, pruned_loss=0.05534, over 937181.91 frames. ], batch size: 43, lr: 3.46e-03, grad_scale: 32.0 2023-04-27 07:58:50,324 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=86725.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:58:57,061 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9688, 2.6376, 2.0976, 2.4443, 1.7813, 2.3166, 2.1477, 1.6370], device='cuda:1'), covar=tensor([0.2131, 0.1035, 0.0841, 0.1211, 0.3261, 0.1022, 0.1899, 0.2584], device='cuda:1'), in_proj_covar=tensor([0.0290, 0.0312, 0.0223, 0.0283, 0.0316, 0.0266, 0.0254, 0.0269], device='cuda:1'), out_proj_covar=tensor([1.1680e-04, 1.2396e-04, 8.8776e-05, 1.1251e-04, 1.2878e-04, 1.0595e-04, 1.0264e-04, 1.0697e-04], device='cuda:1') 2023-04-27 07:59:05,803 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.056e+02 1.560e+02 1.911e+02 2.310e+02 3.714e+02, threshold=3.821e+02, percent-clipped=1.0 2023-04-27 07:59:11,118 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=86756.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:59:17,588 INFO [finetune.py:976] (1/7) Epoch 16, batch 850, loss[loss=0.1776, simple_loss=0.2482, pruned_loss=0.05352, over 4768.00 frames. ], tot_loss[loss=0.1811, simple_loss=0.2507, pruned_loss=0.05571, over 942094.13 frames. ], batch size: 28, lr: 3.46e-03, grad_scale: 32.0 2023-04-27 07:59:21,921 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=86773.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:59:36,381 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=86796.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:59:39,986 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=86801.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 07:59:49,841 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=86814.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:59:50,969 INFO [finetune.py:976] (1/7) Epoch 16, batch 900, loss[loss=0.1671, simple_loss=0.2322, pruned_loss=0.05095, over 4907.00 frames. ], tot_loss[loss=0.1771, simple_loss=0.2469, pruned_loss=0.05372, over 946130.87 frames. ], batch size: 43, lr: 3.46e-03, grad_scale: 32.0 2023-04-27 08:00:08,349 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8022, 1.2385, 1.8461, 2.2791, 1.9066, 1.7342, 1.8128, 1.7667], device='cuda:1'), covar=tensor([0.4921, 0.6929, 0.6598, 0.6497, 0.6233, 0.8454, 0.8402, 0.8825], device='cuda:1'), in_proj_covar=tensor([0.0419, 0.0409, 0.0496, 0.0506, 0.0448, 0.0471, 0.0478, 0.0481], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 08:00:22,179 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.095e+02 1.518e+02 1.851e+02 2.220e+02 4.721e+02, threshold=3.701e+02, percent-clipped=1.0 2023-04-27 08:00:52,313 INFO [finetune.py:976] (1/7) Epoch 16, batch 950, loss[loss=0.1894, simple_loss=0.2573, pruned_loss=0.06076, over 4910.00 frames. ], tot_loss[loss=0.1772, simple_loss=0.2463, pruned_loss=0.05401, over 948577.65 frames. ], batch size: 37, lr: 3.46e-03, grad_scale: 32.0 2023-04-27 08:01:03,387 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=86875.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:01:05,171 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=86878.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:01:37,241 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=86902.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:01:58,441 INFO [finetune.py:976] (1/7) Epoch 16, batch 1000, loss[loss=0.2199, simple_loss=0.2917, pruned_loss=0.07403, over 4751.00 frames. ], tot_loss[loss=0.1795, simple_loss=0.2489, pruned_loss=0.05504, over 950639.13 frames. ], batch size: 54, lr: 3.46e-03, grad_scale: 32.0 2023-04-27 08:02:18,017 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=86936.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:02:18,708 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7858, 1.6392, 1.9629, 2.0607, 1.5554, 1.3794, 1.7215, 1.0945], device='cuda:1'), covar=tensor([0.0642, 0.0784, 0.0626, 0.0912, 0.0829, 0.1241, 0.0740, 0.0813], device='cuda:1'), in_proj_covar=tensor([0.0067, 0.0069, 0.0069, 0.0068, 0.0075, 0.0096, 0.0075, 0.0068], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-27 08:02:19,915 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=86939.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:02:31,464 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.164e+02 1.765e+02 1.981e+02 2.453e+02 9.285e+02, threshold=3.962e+02, percent-clipped=3.0 2023-04-27 08:02:37,025 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=86950.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:02:59,852 INFO [finetune.py:976] (1/7) Epoch 16, batch 1050, loss[loss=0.1795, simple_loss=0.2554, pruned_loss=0.05185, over 4924.00 frames. ], tot_loss[loss=0.1811, simple_loss=0.2512, pruned_loss=0.0555, over 952076.54 frames. ], batch size: 42, lr: 3.46e-03, grad_scale: 32.0 2023-04-27 08:03:11,539 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=5.13 vs. limit=5.0 2023-04-27 08:03:17,496 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9755, 1.6800, 1.8794, 2.2958, 2.2188, 1.7605, 1.5805, 2.1888], device='cuda:1'), covar=tensor([0.0923, 0.1305, 0.0759, 0.0616, 0.0636, 0.0962, 0.0833, 0.0515], device='cuda:1'), in_proj_covar=tensor([0.0191, 0.0204, 0.0184, 0.0174, 0.0177, 0.0185, 0.0155, 0.0181], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 08:03:28,650 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=87002.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 08:03:29,293 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.8771, 2.0376, 2.0471, 2.2170, 1.9385, 2.1477, 2.1851, 2.1393], device='cuda:1'), covar=tensor([0.4210, 0.6740, 0.5591, 0.5175, 0.6276, 0.7716, 0.6347, 0.5887], device='cuda:1'), in_proj_covar=tensor([0.0330, 0.0371, 0.0318, 0.0332, 0.0344, 0.0396, 0.0352, 0.0325], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 08:03:41,350 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7346, 1.3320, 1.8514, 2.2222, 1.8232, 1.7431, 1.8081, 1.7610], device='cuda:1'), covar=tensor([0.4837, 0.6956, 0.6808, 0.6015, 0.6413, 0.8574, 0.8431, 0.8740], device='cuda:1'), in_proj_covar=tensor([0.0419, 0.0409, 0.0497, 0.0508, 0.0449, 0.0471, 0.0479, 0.0482], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 08:03:43,716 INFO [finetune.py:976] (1/7) Epoch 16, batch 1100, loss[loss=0.175, simple_loss=0.2479, pruned_loss=0.05101, over 4715.00 frames. ], tot_loss[loss=0.1816, simple_loss=0.2524, pruned_loss=0.05541, over 952551.08 frames. ], batch size: 59, lr: 3.45e-03, grad_scale: 32.0 2023-04-27 08:04:06,134 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.082e+02 1.660e+02 1.994e+02 2.342e+02 3.600e+02, threshold=3.988e+02, percent-clipped=0.0 2023-04-27 08:04:08,077 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6676, 2.1984, 1.5682, 1.4864, 1.1847, 1.2439, 1.6691, 1.1572], device='cuda:1'), covar=tensor([0.1753, 0.1162, 0.1517, 0.1753, 0.2420, 0.1929, 0.1009, 0.2090], device='cuda:1'), in_proj_covar=tensor([0.0195, 0.0211, 0.0167, 0.0203, 0.0199, 0.0182, 0.0154, 0.0185], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-27 08:04:15,693 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=87063.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 08:04:17,376 INFO [finetune.py:976] (1/7) Epoch 16, batch 1150, loss[loss=0.1845, simple_loss=0.2628, pruned_loss=0.05307, over 4798.00 frames. ], tot_loss[loss=0.182, simple_loss=0.2526, pruned_loss=0.05567, over 953662.80 frames. ], batch size: 40, lr: 3.45e-03, grad_scale: 32.0 2023-04-27 08:04:37,609 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=87096.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:04:40,659 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=87101.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 08:04:50,579 INFO [finetune.py:976] (1/7) Epoch 16, batch 1200, loss[loss=0.1559, simple_loss=0.2272, pruned_loss=0.04228, over 4929.00 frames. ], tot_loss[loss=0.1817, simple_loss=0.2515, pruned_loss=0.05598, over 953761.19 frames. ], batch size: 38, lr: 3.45e-03, grad_scale: 32.0 2023-04-27 08:04:51,383 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-04-27 08:05:08,349 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0514, 1.4674, 1.9363, 2.2265, 1.8913, 1.4680, 1.0675, 1.5527], device='cuda:1'), covar=tensor([0.3269, 0.3399, 0.1642, 0.2144, 0.2788, 0.2780, 0.4487, 0.2270], device='cuda:1'), in_proj_covar=tensor([0.0290, 0.0247, 0.0223, 0.0315, 0.0216, 0.0229, 0.0228, 0.0186], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-27 08:05:10,107 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=87144.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:05:13,048 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.133e+02 1.748e+02 2.017e+02 2.271e+02 5.228e+02, threshold=4.034e+02, percent-clipped=1.0 2023-04-27 08:05:13,119 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=87149.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:05:24,512 INFO [finetune.py:976] (1/7) Epoch 16, batch 1250, loss[loss=0.1495, simple_loss=0.2202, pruned_loss=0.03935, over 4838.00 frames. ], tot_loss[loss=0.1793, simple_loss=0.2489, pruned_loss=0.05483, over 953055.87 frames. ], batch size: 47, lr: 3.45e-03, grad_scale: 32.0 2023-04-27 08:05:27,040 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=87170.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:05:52,206 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8977, 1.3938, 1.4901, 1.5260, 2.0401, 1.5966, 1.3339, 1.4672], device='cuda:1'), covar=tensor([0.1425, 0.1365, 0.1906, 0.1338, 0.0899, 0.1820, 0.2010, 0.1865], device='cuda:1'), in_proj_covar=tensor([0.0305, 0.0313, 0.0352, 0.0289, 0.0329, 0.0312, 0.0300, 0.0362], device='cuda:1'), out_proj_covar=tensor([6.3079e-05, 6.5378e-05, 7.5182e-05, 5.8876e-05, 6.8631e-05, 6.5931e-05, 6.3333e-05, 7.7220e-05], device='cuda:1') 2023-04-27 08:05:58,114 INFO [finetune.py:976] (1/7) Epoch 16, batch 1300, loss[loss=0.1772, simple_loss=0.2481, pruned_loss=0.05314, over 4909.00 frames. ], tot_loss[loss=0.1763, simple_loss=0.2455, pruned_loss=0.05359, over 955106.77 frames. ], batch size: 35, lr: 3.45e-03, grad_scale: 32.0 2023-04-27 08:06:08,725 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=87230.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:06:12,072 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=87234.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:06:13,312 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=87236.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:06:21,119 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.142e+02 1.625e+02 1.883e+02 2.267e+02 3.477e+02, threshold=3.766e+02, percent-clipped=0.0 2023-04-27 08:06:26,678 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4136, 1.6297, 1.6572, 2.0951, 1.8444, 2.0702, 1.5455, 4.4273], device='cuda:1'), covar=tensor([0.0561, 0.0780, 0.0780, 0.1130, 0.0657, 0.0574, 0.0772, 0.0107], device='cuda:1'), in_proj_covar=tensor([0.0037, 0.0039, 0.0040, 0.0043, 0.0040, 0.0038, 0.0038, 0.0056], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:1') 2023-04-27 08:06:26,687 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8673, 1.8262, 1.1496, 1.4305, 2.3409, 1.7061, 1.6054, 1.6870], device='cuda:1'), covar=tensor([0.0487, 0.0380, 0.0310, 0.0567, 0.0245, 0.0517, 0.0489, 0.0553], device='cuda:1'), in_proj_covar=tensor([0.0029, 0.0024, 0.0022, 0.0029, 0.0020, 0.0028, 0.0028, 0.0028], device='cuda:1'), out_proj_covar=tensor([0.0050, 0.0045, 0.0038, 0.0050, 0.0038, 0.0049, 0.0049, 0.0050], device='cuda:1') 2023-04-27 08:06:31,975 INFO [finetune.py:976] (1/7) Epoch 16, batch 1350, loss[loss=0.1356, simple_loss=0.208, pruned_loss=0.03162, over 4824.00 frames. ], tot_loss[loss=0.1776, simple_loss=0.2465, pruned_loss=0.05433, over 953332.24 frames. ], batch size: 30, lr: 3.45e-03, grad_scale: 32.0 2023-04-27 08:06:45,594 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=87284.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:06:50,350 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=87291.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:07:16,734 INFO [finetune.py:976] (1/7) Epoch 16, batch 1400, loss[loss=0.192, simple_loss=0.2686, pruned_loss=0.0577, over 4907.00 frames. ], tot_loss[loss=0.18, simple_loss=0.2496, pruned_loss=0.05524, over 952206.24 frames. ], batch size: 36, lr: 3.45e-03, grad_scale: 32.0 2023-04-27 08:07:58,926 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.057e+02 1.760e+02 2.100e+02 2.306e+02 6.240e+02, threshold=4.200e+02, percent-clipped=3.0 2023-04-27 08:08:08,635 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.2309, 2.5682, 1.1230, 1.4845, 2.1236, 1.4270, 3.3799, 1.8279], device='cuda:1'), covar=tensor([0.0619, 0.0605, 0.0790, 0.1291, 0.0472, 0.0989, 0.0408, 0.0642], device='cuda:1'), in_proj_covar=tensor([0.0051, 0.0066, 0.0048, 0.0046, 0.0050, 0.0052, 0.0076, 0.0051], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:1') 2023-04-27 08:08:09,822 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=87358.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 08:08:17,783 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=87362.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:08:20,150 INFO [finetune.py:976] (1/7) Epoch 16, batch 1450, loss[loss=0.1883, simple_loss=0.2624, pruned_loss=0.05712, over 4792.00 frames. ], tot_loss[loss=0.1819, simple_loss=0.2521, pruned_loss=0.05582, over 953762.81 frames. ], batch size: 51, lr: 3.45e-03, grad_scale: 32.0 2023-04-27 08:08:36,907 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.3201, 2.8349, 0.8424, 1.5393, 2.1358, 1.3901, 3.7563, 1.6378], device='cuda:1'), covar=tensor([0.0697, 0.1026, 0.0967, 0.1305, 0.0538, 0.1043, 0.0270, 0.0727], device='cuda:1'), in_proj_covar=tensor([0.0051, 0.0066, 0.0048, 0.0046, 0.0050, 0.0052, 0.0076, 0.0051], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:1') 2023-04-27 08:09:14,067 INFO [finetune.py:976] (1/7) Epoch 16, batch 1500, loss[loss=0.1625, simple_loss=0.2413, pruned_loss=0.04189, over 4800.00 frames. ], tot_loss[loss=0.1835, simple_loss=0.2541, pruned_loss=0.05648, over 954664.09 frames. ], batch size: 25, lr: 3.45e-03, grad_scale: 32.0 2023-04-27 08:09:18,446 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=87423.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:09:36,459 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.033e+02 1.658e+02 1.977e+02 2.415e+02 4.579e+02, threshold=3.953e+02, percent-clipped=1.0 2023-04-27 08:09:43,835 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=87461.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:09:46,778 INFO [finetune.py:976] (1/7) Epoch 16, batch 1550, loss[loss=0.1558, simple_loss=0.2333, pruned_loss=0.0391, over 4867.00 frames. ], tot_loss[loss=0.1838, simple_loss=0.2538, pruned_loss=0.05688, over 952391.44 frames. ], batch size: 34, lr: 3.45e-03, grad_scale: 32.0 2023-04-27 08:09:49,314 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=87470.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:10:30,185 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=87505.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:10:30,203 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0737, 1.4278, 1.8355, 2.1483, 1.9226, 1.5211, 1.1292, 1.5748], device='cuda:1'), covar=tensor([0.3978, 0.3783, 0.2233, 0.2852, 0.2905, 0.2828, 0.4370, 0.2330], device='cuda:1'), in_proj_covar=tensor([0.0289, 0.0245, 0.0223, 0.0314, 0.0216, 0.0229, 0.0227, 0.0185], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-27 08:10:42,326 INFO [finetune.py:976] (1/7) Epoch 16, batch 1600, loss[loss=0.1895, simple_loss=0.2546, pruned_loss=0.06218, over 4816.00 frames. ], tot_loss[loss=0.1817, simple_loss=0.2511, pruned_loss=0.05612, over 952247.07 frames. ], batch size: 41, lr: 3.45e-03, grad_scale: 32.0 2023-04-27 08:10:43,620 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=87518.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:10:46,090 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=87522.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:10:54,408 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=87534.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:10:56,915 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.15 vs. limit=5.0 2023-04-27 08:11:05,883 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.006e+02 1.660e+02 1.902e+02 2.277e+02 5.320e+02, threshold=3.803e+02, percent-clipped=1.0 2023-04-27 08:11:13,912 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.1526, 1.7997, 2.0143, 2.5022, 2.3191, 1.9453, 1.6382, 2.1951], device='cuda:1'), covar=tensor([0.0868, 0.1114, 0.0676, 0.0514, 0.0707, 0.0937, 0.0811, 0.0534], device='cuda:1'), in_proj_covar=tensor([0.0192, 0.0205, 0.0185, 0.0175, 0.0178, 0.0185, 0.0155, 0.0182], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 08:11:16,205 INFO [finetune.py:976] (1/7) Epoch 16, batch 1650, loss[loss=0.1863, simple_loss=0.2503, pruned_loss=0.0611, over 4869.00 frames. ], tot_loss[loss=0.1797, simple_loss=0.2485, pruned_loss=0.05541, over 953859.22 frames. ], batch size: 49, lr: 3.45e-03, grad_scale: 32.0 2023-04-27 08:11:16,339 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=87566.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 08:11:26,481 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=87582.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:11:28,259 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=87584.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:11:29,405 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=87586.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:11:31,300 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=87589.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:11:40,772 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-04-27 08:11:43,110 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=87605.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:11:49,708 INFO [finetune.py:976] (1/7) Epoch 16, batch 1700, loss[loss=0.1337, simple_loss=0.2057, pruned_loss=0.03091, over 4770.00 frames. ], tot_loss[loss=0.1779, simple_loss=0.2467, pruned_loss=0.05457, over 953779.26 frames. ], batch size: 26, lr: 3.45e-03, grad_scale: 32.0 2023-04-27 08:12:08,445 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=87645.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 08:12:12,239 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.051e+02 1.609e+02 2.053e+02 2.465e+02 3.849e+02, threshold=4.106e+02, percent-clipped=1.0 2023-04-27 08:12:12,985 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=87650.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:12:14,644 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([3.4716, 3.5525, 2.8482, 4.0993, 3.4648, 3.4547, 1.8103, 3.5684], device='cuda:1'), covar=tensor([0.1795, 0.1231, 0.3697, 0.1490, 0.3559, 0.1788, 0.5327, 0.2439], device='cuda:1'), in_proj_covar=tensor([0.0243, 0.0212, 0.0250, 0.0302, 0.0298, 0.0247, 0.0270, 0.0272], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 08:12:18,812 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=87658.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 08:12:23,600 INFO [finetune.py:976] (1/7) Epoch 16, batch 1750, loss[loss=0.1764, simple_loss=0.2506, pruned_loss=0.05111, over 4892.00 frames. ], tot_loss[loss=0.1796, simple_loss=0.2485, pruned_loss=0.05538, over 955112.56 frames. ], batch size: 35, lr: 3.45e-03, grad_scale: 32.0 2023-04-27 08:12:23,735 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=87666.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 08:12:55,814 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=87706.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 08:13:07,610 INFO [finetune.py:976] (1/7) Epoch 16, batch 1800, loss[loss=0.1545, simple_loss=0.2269, pruned_loss=0.04107, over 4762.00 frames. ], tot_loss[loss=0.1817, simple_loss=0.2519, pruned_loss=0.05572, over 954969.67 frames. ], batch size: 26, lr: 3.45e-03, grad_scale: 32.0 2023-04-27 08:13:14,247 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=87718.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:13:49,467 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.133e+02 1.657e+02 1.996e+02 2.458e+02 4.294e+02, threshold=3.992e+02, percent-clipped=1.0 2023-04-27 08:14:12,969 INFO [finetune.py:976] (1/7) Epoch 16, batch 1850, loss[loss=0.1857, simple_loss=0.2559, pruned_loss=0.05771, over 4890.00 frames. ], tot_loss[loss=0.1816, simple_loss=0.252, pruned_loss=0.05564, over 953615.44 frames. ], batch size: 35, lr: 3.45e-03, grad_scale: 32.0 2023-04-27 08:14:21,294 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.83 vs. limit=5.0 2023-04-27 08:15:07,315 INFO [finetune.py:976] (1/7) Epoch 16, batch 1900, loss[loss=0.1566, simple_loss=0.2336, pruned_loss=0.0398, over 4822.00 frames. ], tot_loss[loss=0.1826, simple_loss=0.2531, pruned_loss=0.05607, over 952193.77 frames. ], batch size: 39, lr: 3.45e-03, grad_scale: 32.0 2023-04-27 08:15:07,423 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6337, 1.5241, 0.7670, 1.3281, 1.6075, 1.5164, 1.3870, 1.4285], device='cuda:1'), covar=tensor([0.0457, 0.0352, 0.0377, 0.0503, 0.0299, 0.0461, 0.0472, 0.0531], device='cuda:1'), in_proj_covar=tensor([0.0029, 0.0024, 0.0022, 0.0029, 0.0020, 0.0028, 0.0028, 0.0029], device='cuda:1'), out_proj_covar=tensor([0.0050, 0.0044, 0.0038, 0.0051, 0.0038, 0.0049, 0.0049, 0.0050], device='cuda:1') 2023-04-27 08:15:08,020 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=87817.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:15:23,640 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.2156, 1.4773, 1.3709, 1.7667, 1.5794, 1.7181, 1.3910, 3.1070], device='cuda:1'), covar=tensor([0.0644, 0.0834, 0.0833, 0.1234, 0.0692, 0.0510, 0.0758, 0.0178], device='cuda:1'), in_proj_covar=tensor([0.0038, 0.0039, 0.0040, 0.0044, 0.0040, 0.0038, 0.0038, 0.0056], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:1') 2023-04-27 08:15:28,297 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.071e+02 1.628e+02 1.844e+02 2.203e+02 4.331e+02, threshold=3.688e+02, percent-clipped=1.0 2023-04-27 08:15:34,880 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6924, 1.6870, 0.6836, 1.3844, 1.7678, 1.5755, 1.4437, 1.5003], device='cuda:1'), covar=tensor([0.0503, 0.0338, 0.0382, 0.0521, 0.0294, 0.0501, 0.0495, 0.0552], device='cuda:1'), in_proj_covar=tensor([0.0029, 0.0024, 0.0022, 0.0029, 0.0020, 0.0028, 0.0028, 0.0029], device='cuda:1'), out_proj_covar=tensor([0.0050, 0.0045, 0.0038, 0.0051, 0.0038, 0.0049, 0.0049, 0.0050], device='cuda:1') 2023-04-27 08:15:37,112 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=87861.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 08:15:40,073 INFO [finetune.py:976] (1/7) Epoch 16, batch 1950, loss[loss=0.2043, simple_loss=0.2701, pruned_loss=0.0692, over 4818.00 frames. ], tot_loss[loss=0.1804, simple_loss=0.2508, pruned_loss=0.05503, over 951648.26 frames. ], batch size: 38, lr: 3.45e-03, grad_scale: 32.0 2023-04-27 08:16:03,421 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8276, 2.3835, 0.8140, 1.1872, 1.5021, 1.1193, 2.5346, 1.3113], device='cuda:1'), covar=tensor([0.0766, 0.0525, 0.0762, 0.1591, 0.0560, 0.1263, 0.0324, 0.0940], device='cuda:1'), in_proj_covar=tensor([0.0051, 0.0066, 0.0048, 0.0046, 0.0050, 0.0052, 0.0075, 0.0051], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0007, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:1') 2023-04-27 08:16:10,452 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=87886.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:16:47,484 INFO [finetune.py:976] (1/7) Epoch 16, batch 2000, loss[loss=0.1959, simple_loss=0.2525, pruned_loss=0.06962, over 4915.00 frames. ], tot_loss[loss=0.1802, simple_loss=0.2497, pruned_loss=0.05539, over 953638.38 frames. ], batch size: 37, lr: 3.45e-03, grad_scale: 32.0 2023-04-27 08:16:59,001 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=87934.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:17:02,093 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.1117, 1.4667, 1.3641, 1.8166, 1.5510, 1.8347, 1.3995, 3.5906], device='cuda:1'), covar=tensor([0.0743, 0.1055, 0.1051, 0.1318, 0.0866, 0.0630, 0.0980, 0.0250], device='cuda:1'), in_proj_covar=tensor([0.0038, 0.0039, 0.0040, 0.0044, 0.0040, 0.0038, 0.0038, 0.0056], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:1') 2023-04-27 08:17:02,681 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=87940.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 08:17:05,748 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=87945.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:17:08,486 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.110e+02 1.621e+02 1.906e+02 2.380e+02 5.072e+02, threshold=3.811e+02, percent-clipped=4.0 2023-04-27 08:17:17,309 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=87961.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 08:17:21,295 INFO [finetune.py:976] (1/7) Epoch 16, batch 2050, loss[loss=0.1611, simple_loss=0.2315, pruned_loss=0.04533, over 4894.00 frames. ], tot_loss[loss=0.1776, simple_loss=0.2463, pruned_loss=0.05448, over 955693.15 frames. ], batch size: 32, lr: 3.45e-03, grad_scale: 32.0 2023-04-27 08:17:49,200 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.3425, 1.3289, 1.3470, 1.5542, 1.6874, 1.3520, 0.8876, 1.4747], device='cuda:1'), covar=tensor([0.0816, 0.1363, 0.0890, 0.0647, 0.0641, 0.0850, 0.0956, 0.0644], device='cuda:1'), in_proj_covar=tensor([0.0191, 0.0203, 0.0184, 0.0174, 0.0177, 0.0183, 0.0155, 0.0181], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 08:17:55,289 INFO [finetune.py:976] (1/7) Epoch 16, batch 2100, loss[loss=0.1817, simple_loss=0.2416, pruned_loss=0.06089, over 4878.00 frames. ], tot_loss[loss=0.178, simple_loss=0.2462, pruned_loss=0.0549, over 955567.48 frames. ], batch size: 31, lr: 3.45e-03, grad_scale: 32.0 2023-04-27 08:17:56,597 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=88018.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:18:02,982 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.0857, 1.3777, 1.2868, 1.5918, 1.4163, 1.6625, 1.3095, 2.9797], device='cuda:1'), covar=tensor([0.0709, 0.0907, 0.0900, 0.1306, 0.0756, 0.0599, 0.0882, 0.0228], device='cuda:1'), in_proj_covar=tensor([0.0038, 0.0039, 0.0040, 0.0044, 0.0040, 0.0038, 0.0038, 0.0056], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:1') 2023-04-27 08:18:04,217 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6551, 1.2506, 1.2782, 1.3983, 1.8589, 1.5227, 1.2393, 1.2167], device='cuda:1'), covar=tensor([0.1492, 0.1304, 0.1626, 0.1193, 0.0655, 0.1302, 0.1790, 0.1934], device='cuda:1'), in_proj_covar=tensor([0.0304, 0.0310, 0.0349, 0.0289, 0.0327, 0.0310, 0.0298, 0.0359], device='cuda:1'), out_proj_covar=tensor([6.2846e-05, 6.4615e-05, 7.4457e-05, 5.8898e-05, 6.8214e-05, 6.5344e-05, 6.3012e-05, 7.6541e-05], device='cuda:1') 2023-04-27 08:18:16,279 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.262e+02 1.786e+02 2.026e+02 2.569e+02 5.831e+02, threshold=4.052e+02, percent-clipped=4.0 2023-04-27 08:18:28,509 INFO [finetune.py:976] (1/7) Epoch 16, batch 2150, loss[loss=0.2014, simple_loss=0.272, pruned_loss=0.0654, over 4901.00 frames. ], tot_loss[loss=0.1813, simple_loss=0.2498, pruned_loss=0.05643, over 954960.22 frames. ], batch size: 43, lr: 3.45e-03, grad_scale: 32.0 2023-04-27 08:18:28,580 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=88066.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:18:50,528 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=88100.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:19:01,076 INFO [finetune.py:976] (1/7) Epoch 16, batch 2200, loss[loss=0.1592, simple_loss=0.2335, pruned_loss=0.04248, over 4814.00 frames. ], tot_loss[loss=0.1824, simple_loss=0.2518, pruned_loss=0.05651, over 955531.24 frames. ], batch size: 38, lr: 3.45e-03, grad_scale: 32.0 2023-04-27 08:19:02,291 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=88117.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:19:15,303 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.71 vs. limit=2.0 2023-04-27 08:19:34,550 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.043e+02 1.597e+02 1.844e+02 2.202e+02 4.029e+02, threshold=3.688e+02, percent-clipped=0.0 2023-04-27 08:19:39,600 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6031, 1.3303, 1.7965, 1.7888, 1.4479, 1.3134, 1.4602, 0.9905], device='cuda:1'), covar=tensor([0.0502, 0.0732, 0.0434, 0.0596, 0.0646, 0.1216, 0.0572, 0.0733], device='cuda:1'), in_proj_covar=tensor([0.0068, 0.0070, 0.0069, 0.0068, 0.0076, 0.0097, 0.0075, 0.0068], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-27 08:19:47,601 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=88161.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 08:19:47,638 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=88161.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:19:50,024 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=88165.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:19:56,115 INFO [finetune.py:976] (1/7) Epoch 16, batch 2250, loss[loss=0.2039, simple_loss=0.2775, pruned_loss=0.06512, over 4865.00 frames. ], tot_loss[loss=0.1827, simple_loss=0.2526, pruned_loss=0.05637, over 955911.85 frames. ], batch size: 34, lr: 3.45e-03, grad_scale: 32.0 2023-04-27 08:20:50,811 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=88209.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:21:01,645 INFO [finetune.py:976] (1/7) Epoch 16, batch 2300, loss[loss=0.1754, simple_loss=0.2302, pruned_loss=0.06024, over 4711.00 frames. ], tot_loss[loss=0.1833, simple_loss=0.2535, pruned_loss=0.0565, over 956873.72 frames. ], batch size: 23, lr: 3.44e-03, grad_scale: 32.0 2023-04-27 08:21:18,862 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=88240.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:21:21,949 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=88245.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:21:24,254 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.059e+01 1.606e+02 1.877e+02 2.265e+02 8.419e+02, threshold=3.754e+02, percent-clipped=4.0 2023-04-27 08:21:32,154 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=88261.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 08:21:41,584 INFO [finetune.py:976] (1/7) Epoch 16, batch 2350, loss[loss=0.1247, simple_loss=0.1991, pruned_loss=0.02514, over 4703.00 frames. ], tot_loss[loss=0.1799, simple_loss=0.2501, pruned_loss=0.05488, over 956439.59 frames. ], batch size: 23, lr: 3.44e-03, grad_scale: 64.0 2023-04-27 08:22:14,473 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=88288.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:22:16,363 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=88291.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 08:22:17,546 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=88293.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:22:38,197 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=88309.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:22:48,482 INFO [finetune.py:976] (1/7) Epoch 16, batch 2400, loss[loss=0.1831, simple_loss=0.2456, pruned_loss=0.06033, over 4816.00 frames. ], tot_loss[loss=0.1774, simple_loss=0.2471, pruned_loss=0.05386, over 955816.93 frames. ], batch size: 51, lr: 3.44e-03, grad_scale: 64.0 2023-04-27 08:23:09,037 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([3.7113, 3.7592, 2.6800, 4.3309, 3.8412, 3.7016, 1.4732, 3.7294], device='cuda:1'), covar=tensor([0.1881, 0.1068, 0.3336, 0.1674, 0.2955, 0.1810, 0.5929, 0.2237], device='cuda:1'), in_proj_covar=tensor([0.0241, 0.0211, 0.0249, 0.0299, 0.0294, 0.0245, 0.0267, 0.0269], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 08:23:23,049 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.024e+02 1.498e+02 1.832e+02 2.156e+02 3.450e+02, threshold=3.664e+02, percent-clipped=0.0 2023-04-27 08:23:25,022 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=88352.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 08:23:33,493 INFO [finetune.py:976] (1/7) Epoch 16, batch 2450, loss[loss=0.1771, simple_loss=0.2448, pruned_loss=0.05472, over 4823.00 frames. ], tot_loss[loss=0.1754, simple_loss=0.2445, pruned_loss=0.05319, over 955829.43 frames. ], batch size: 38, lr: 3.44e-03, grad_scale: 64.0 2023-04-27 08:23:58,200 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.1726, 1.6547, 1.4466, 1.8133, 1.5860, 1.9130, 1.4505, 3.5687], device='cuda:1'), covar=tensor([0.0656, 0.0788, 0.0813, 0.1149, 0.0698, 0.0516, 0.0750, 0.0141], device='cuda:1'), in_proj_covar=tensor([0.0038, 0.0039, 0.0040, 0.0044, 0.0040, 0.0038, 0.0038, 0.0056], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:1') 2023-04-27 08:24:30,564 INFO [finetune.py:976] (1/7) Epoch 16, batch 2500, loss[loss=0.1841, simple_loss=0.2611, pruned_loss=0.05353, over 4827.00 frames. ], tot_loss[loss=0.1775, simple_loss=0.2462, pruned_loss=0.05433, over 953980.94 frames. ], batch size: 40, lr: 3.44e-03, grad_scale: 32.0 2023-04-27 08:24:54,797 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.230e+02 1.712e+02 2.054e+02 2.521e+02 4.262e+02, threshold=4.109e+02, percent-clipped=2.0 2023-04-27 08:24:58,641 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=88456.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:25:04,746 INFO [finetune.py:976] (1/7) Epoch 16, batch 2550, loss[loss=0.1456, simple_loss=0.2208, pruned_loss=0.03518, over 4879.00 frames. ], tot_loss[loss=0.1803, simple_loss=0.2499, pruned_loss=0.05536, over 954118.81 frames. ], batch size: 32, lr: 3.44e-03, grad_scale: 32.0 2023-04-27 08:25:30,766 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6055, 1.6553, 0.7772, 1.2844, 1.8281, 1.4577, 1.3721, 1.4535], device='cuda:1'), covar=tensor([0.0510, 0.0380, 0.0360, 0.0592, 0.0276, 0.0552, 0.0514, 0.0590], device='cuda:1'), in_proj_covar=tensor([0.0029, 0.0024, 0.0022, 0.0029, 0.0020, 0.0028, 0.0028, 0.0028], device='cuda:1'), out_proj_covar=tensor([0.0050, 0.0044, 0.0037, 0.0050, 0.0038, 0.0049, 0.0049, 0.0050], device='cuda:1') 2023-04-27 08:25:36,260 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.0877, 2.0958, 1.7410, 1.7914, 2.1977, 1.8733, 2.5941, 1.5394], device='cuda:1'), covar=tensor([0.3544, 0.1854, 0.4952, 0.2705, 0.1686, 0.2194, 0.1426, 0.4638], device='cuda:1'), in_proj_covar=tensor([0.0344, 0.0350, 0.0430, 0.0360, 0.0387, 0.0385, 0.0375, 0.0423], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 08:25:38,575 INFO [finetune.py:976] (1/7) Epoch 16, batch 2600, loss[loss=0.2094, simple_loss=0.2918, pruned_loss=0.06348, over 4811.00 frames. ], tot_loss[loss=0.1818, simple_loss=0.2517, pruned_loss=0.05596, over 950776.94 frames. ], batch size: 38, lr: 3.44e-03, grad_scale: 32.0 2023-04-27 08:26:18,046 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.329e+02 1.735e+02 2.047e+02 2.467e+02 4.454e+02, threshold=4.094e+02, percent-clipped=2.0 2023-04-27 08:26:39,229 INFO [finetune.py:976] (1/7) Epoch 16, batch 2650, loss[loss=0.1817, simple_loss=0.2521, pruned_loss=0.0556, over 4891.00 frames. ], tot_loss[loss=0.1824, simple_loss=0.2525, pruned_loss=0.05616, over 951706.84 frames. ], batch size: 35, lr: 3.44e-03, grad_scale: 32.0 2023-04-27 08:26:51,957 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4990, 1.3760, 1.8086, 1.8888, 1.4328, 1.3080, 1.4493, 0.9653], device='cuda:1'), covar=tensor([0.0605, 0.0834, 0.0450, 0.0680, 0.0782, 0.1210, 0.0713, 0.0743], device='cuda:1'), in_proj_covar=tensor([0.0068, 0.0070, 0.0069, 0.0067, 0.0076, 0.0096, 0.0075, 0.0068], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-27 08:27:29,037 INFO [finetune.py:976] (1/7) Epoch 16, batch 2700, loss[loss=0.1681, simple_loss=0.2371, pruned_loss=0.04955, over 4734.00 frames. ], tot_loss[loss=0.181, simple_loss=0.2514, pruned_loss=0.05533, over 953949.58 frames. ], batch size: 54, lr: 3.44e-03, grad_scale: 32.0 2023-04-27 08:28:11,105 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=88647.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 08:28:13,346 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.055e+02 1.464e+02 1.764e+02 2.078e+02 3.620e+02, threshold=3.528e+02, percent-clipped=0.0 2023-04-27 08:28:36,532 INFO [finetune.py:976] (1/7) Epoch 16, batch 2750, loss[loss=0.1577, simple_loss=0.2306, pruned_loss=0.0424, over 4887.00 frames. ], tot_loss[loss=0.1781, simple_loss=0.2477, pruned_loss=0.05428, over 954044.60 frames. ], batch size: 32, lr: 3.44e-03, grad_scale: 32.0 2023-04-27 08:29:16,533 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2023-04-27 08:29:21,108 INFO [finetune.py:976] (1/7) Epoch 16, batch 2800, loss[loss=0.1873, simple_loss=0.257, pruned_loss=0.05878, over 4737.00 frames. ], tot_loss[loss=0.1769, simple_loss=0.2457, pruned_loss=0.05399, over 954654.00 frames. ], batch size: 27, lr: 3.44e-03, grad_scale: 32.0 2023-04-27 08:29:42,735 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.076e+02 1.582e+02 1.928e+02 2.328e+02 4.353e+02, threshold=3.856e+02, percent-clipped=4.0 2023-04-27 08:29:47,476 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=88756.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:29:54,457 INFO [finetune.py:976] (1/7) Epoch 16, batch 2850, loss[loss=0.1919, simple_loss=0.2618, pruned_loss=0.06103, over 4781.00 frames. ], tot_loss[loss=0.1747, simple_loss=0.2436, pruned_loss=0.0529, over 955658.56 frames. ], batch size: 59, lr: 3.44e-03, grad_scale: 32.0 2023-04-27 08:30:01,915 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=5.53 vs. limit=5.0 2023-04-27 08:30:20,173 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=88804.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:30:28,459 INFO [finetune.py:976] (1/7) Epoch 16, batch 2900, loss[loss=0.1858, simple_loss=0.2641, pruned_loss=0.0538, over 4852.00 frames. ], tot_loss[loss=0.1769, simple_loss=0.2461, pruned_loss=0.05385, over 954187.31 frames. ], batch size: 47, lr: 3.44e-03, grad_scale: 32.0 2023-04-27 08:30:42,462 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.2456, 1.5143, 1.4549, 1.8281, 1.6685, 1.6920, 1.4712, 3.1048], device='cuda:1'), covar=tensor([0.0643, 0.0844, 0.0848, 0.1208, 0.0648, 0.0477, 0.0753, 0.0186], device='cuda:1'), in_proj_covar=tensor([0.0038, 0.0039, 0.0040, 0.0044, 0.0040, 0.0038, 0.0038, 0.0056], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:1') 2023-04-27 08:30:47,761 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.1928, 2.6474, 1.2982, 1.4457, 2.3567, 1.2699, 3.6106, 1.9096], device='cuda:1'), covar=tensor([0.0696, 0.0569, 0.0736, 0.1267, 0.0424, 0.1036, 0.0268, 0.0635], device='cuda:1'), in_proj_covar=tensor([0.0052, 0.0066, 0.0049, 0.0046, 0.0050, 0.0052, 0.0075, 0.0052], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:1') 2023-04-27 08:30:50,712 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.204e+02 1.633e+02 2.026e+02 2.312e+02 4.286e+02, threshold=4.053e+02, percent-clipped=1.0 2023-04-27 08:31:02,491 INFO [finetune.py:976] (1/7) Epoch 16, batch 2950, loss[loss=0.1881, simple_loss=0.2615, pruned_loss=0.05733, over 4904.00 frames. ], tot_loss[loss=0.1774, simple_loss=0.2473, pruned_loss=0.05375, over 953886.44 frames. ], batch size: 36, lr: 3.44e-03, grad_scale: 32.0 2023-04-27 08:31:05,705 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6703, 1.8406, 0.9978, 1.3541, 2.1581, 1.5540, 1.4300, 1.5257], device='cuda:1'), covar=tensor([0.0487, 0.0347, 0.0328, 0.0551, 0.0257, 0.0479, 0.0488, 0.0533], device='cuda:1'), in_proj_covar=tensor([0.0029, 0.0024, 0.0022, 0.0029, 0.0020, 0.0028, 0.0028, 0.0028], device='cuda:1'), out_proj_covar=tensor([0.0050, 0.0044, 0.0038, 0.0050, 0.0038, 0.0049, 0.0049, 0.0050], device='cuda:1') 2023-04-27 08:31:58,223 INFO [finetune.py:976] (1/7) Epoch 16, batch 3000, loss[loss=0.2392, simple_loss=0.3083, pruned_loss=0.08511, over 4841.00 frames. ], tot_loss[loss=0.1798, simple_loss=0.2496, pruned_loss=0.05503, over 955077.94 frames. ], batch size: 44, lr: 3.44e-03, grad_scale: 32.0 2023-04-27 08:31:58,223 INFO [finetune.py:1001] (1/7) Computing validation loss 2023-04-27 08:32:14,557 INFO [finetune.py:1010] (1/7) Epoch 16, validation: loss=0.1523, simple_loss=0.2234, pruned_loss=0.04062, over 2265189.00 frames. 2023-04-27 08:32:14,558 INFO [finetune.py:1011] (1/7) Maximum memory allocated so far is 6391MB 2023-04-27 08:32:20,114 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.4485, 2.8976, 1.1697, 1.6374, 2.5652, 1.6450, 4.2098, 2.2521], device='cuda:1'), covar=tensor([0.0684, 0.0761, 0.0844, 0.1293, 0.0470, 0.0970, 0.0217, 0.0582], device='cuda:1'), in_proj_covar=tensor([0.0052, 0.0066, 0.0049, 0.0046, 0.0050, 0.0053, 0.0075, 0.0052], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:1') 2023-04-27 08:32:47,284 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=88947.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 08:32:49,013 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.118e+02 1.654e+02 1.995e+02 2.342e+02 3.743e+02, threshold=3.990e+02, percent-clipped=0.0 2023-04-27 08:32:59,790 INFO [finetune.py:976] (1/7) Epoch 16, batch 3050, loss[loss=0.1695, simple_loss=0.248, pruned_loss=0.04548, over 4822.00 frames. ], tot_loss[loss=0.1806, simple_loss=0.2506, pruned_loss=0.05532, over 954831.66 frames. ], batch size: 38, lr: 3.44e-03, grad_scale: 32.0 2023-04-27 08:33:30,453 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=88995.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 08:33:55,682 INFO [finetune.py:976] (1/7) Epoch 16, batch 3100, loss[loss=0.1582, simple_loss=0.2299, pruned_loss=0.04321, over 4791.00 frames. ], tot_loss[loss=0.18, simple_loss=0.2496, pruned_loss=0.05515, over 955490.13 frames. ], batch size: 29, lr: 3.44e-03, grad_scale: 32.0 2023-04-27 08:34:13,187 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=89027.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:34:17,070 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.13 vs. limit=2.0 2023-04-27 08:34:26,568 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([4.0684, 3.8383, 2.9267, 4.6430, 3.9562, 4.0214, 1.7414, 4.0039], device='cuda:1'), covar=tensor([0.1442, 0.1284, 0.3106, 0.1277, 0.2852, 0.1703, 0.5967, 0.2339], device='cuda:1'), in_proj_covar=tensor([0.0243, 0.0212, 0.0252, 0.0303, 0.0297, 0.0247, 0.0271, 0.0271], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 08:34:38,477 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.056e+02 1.571e+02 1.837e+02 2.281e+02 4.804e+02, threshold=3.674e+02, percent-clipped=2.0 2023-04-27 08:34:59,022 INFO [finetune.py:976] (1/7) Epoch 16, batch 3150, loss[loss=0.1775, simple_loss=0.2508, pruned_loss=0.0521, over 4855.00 frames. ], tot_loss[loss=0.1782, simple_loss=0.2474, pruned_loss=0.05453, over 955851.10 frames. ], batch size: 44, lr: 3.44e-03, grad_scale: 32.0 2023-04-27 08:34:59,104 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.3826, 1.2085, 4.1341, 3.8898, 3.6855, 3.9228, 3.8970, 3.6172], device='cuda:1'), covar=tensor([0.6924, 0.5944, 0.1085, 0.1711, 0.1108, 0.1465, 0.1240, 0.1628], device='cuda:1'), in_proj_covar=tensor([0.0306, 0.0300, 0.0398, 0.0399, 0.0343, 0.0402, 0.0306, 0.0358], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-27 08:35:14,721 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=89088.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:35:29,448 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.61 vs. limit=5.0 2023-04-27 08:35:37,803 INFO [finetune.py:976] (1/7) Epoch 16, batch 3200, loss[loss=0.1171, simple_loss=0.1856, pruned_loss=0.02431, over 4840.00 frames. ], tot_loss[loss=0.1749, simple_loss=0.2441, pruned_loss=0.05289, over 958380.38 frames. ], batch size: 25, lr: 3.44e-03, grad_scale: 32.0 2023-04-27 08:36:23,956 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.134e+02 1.556e+02 1.903e+02 2.206e+02 4.255e+02, threshold=3.807e+02, percent-clipped=1.0 2023-04-27 08:36:44,940 INFO [finetune.py:976] (1/7) Epoch 16, batch 3250, loss[loss=0.2046, simple_loss=0.2714, pruned_loss=0.06891, over 4830.00 frames. ], tot_loss[loss=0.1754, simple_loss=0.2445, pruned_loss=0.05319, over 958087.69 frames. ], batch size: 39, lr: 3.44e-03, grad_scale: 32.0 2023-04-27 08:36:46,910 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=89169.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:37:46,005 INFO [finetune.py:976] (1/7) Epoch 16, batch 3300, loss[loss=0.1533, simple_loss=0.225, pruned_loss=0.04078, over 4755.00 frames. ], tot_loss[loss=0.18, simple_loss=0.2498, pruned_loss=0.05507, over 959048.59 frames. ], batch size: 28, lr: 3.44e-03, grad_scale: 32.0 2023-04-27 08:37:56,935 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-04-27 08:38:07,043 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0219, 2.4136, 0.9924, 1.3303, 1.7996, 1.2609, 3.0712, 1.5545], device='cuda:1'), covar=tensor([0.0716, 0.0595, 0.0790, 0.1318, 0.0508, 0.1025, 0.0283, 0.0686], device='cuda:1'), in_proj_covar=tensor([0.0051, 0.0066, 0.0048, 0.0046, 0.0050, 0.0052, 0.0075, 0.0051], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0007, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:1') 2023-04-27 08:38:07,681 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=89230.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:38:08,866 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5865, 1.3630, 4.4092, 4.0857, 3.8846, 4.1757, 4.0608, 3.8927], device='cuda:1'), covar=tensor([0.7094, 0.5854, 0.1075, 0.1976, 0.1088, 0.2277, 0.1519, 0.1589], device='cuda:1'), in_proj_covar=tensor([0.0308, 0.0301, 0.0398, 0.0401, 0.0343, 0.0403, 0.0306, 0.0359], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-27 08:38:21,747 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.219e+02 1.680e+02 1.944e+02 2.414e+02 4.518e+02, threshold=3.887e+02, percent-clipped=2.0 2023-04-27 08:38:33,914 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9534, 1.7150, 2.1510, 2.4566, 2.0222, 1.8959, 2.0900, 2.0257], device='cuda:1'), covar=tensor([0.5168, 0.7366, 0.7066, 0.6223, 0.6482, 0.9168, 0.9740, 0.9693], device='cuda:1'), in_proj_covar=tensor([0.0418, 0.0407, 0.0493, 0.0507, 0.0448, 0.0473, 0.0478, 0.0479], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 08:38:43,925 INFO [finetune.py:976] (1/7) Epoch 16, batch 3350, loss[loss=0.1578, simple_loss=0.2349, pruned_loss=0.04036, over 4736.00 frames. ], tot_loss[loss=0.1817, simple_loss=0.2519, pruned_loss=0.05576, over 958806.35 frames. ], batch size: 54, lr: 3.44e-03, grad_scale: 32.0 2023-04-27 08:39:07,894 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.12 vs. limit=2.0 2023-04-27 08:39:17,175 INFO [finetune.py:976] (1/7) Epoch 16, batch 3400, loss[loss=0.1582, simple_loss=0.2364, pruned_loss=0.03998, over 4792.00 frames. ], tot_loss[loss=0.1839, simple_loss=0.2546, pruned_loss=0.05662, over 958684.08 frames. ], batch size: 51, lr: 3.44e-03, grad_scale: 32.0 2023-04-27 08:39:40,164 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.038e+02 1.554e+02 1.949e+02 2.299e+02 3.513e+02, threshold=3.898e+02, percent-clipped=0.0 2023-04-27 08:39:53,662 INFO [finetune.py:976] (1/7) Epoch 16, batch 3450, loss[loss=0.1526, simple_loss=0.2196, pruned_loss=0.04279, over 4901.00 frames. ], tot_loss[loss=0.184, simple_loss=0.2546, pruned_loss=0.05666, over 957143.20 frames. ], batch size: 35, lr: 3.43e-03, grad_scale: 32.0 2023-04-27 08:39:57,468 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-04-27 08:40:04,282 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=89374.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:40:16,183 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=89383.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:40:16,207 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5266, 3.5556, 0.8434, 1.7635, 1.9358, 2.5051, 1.9155, 0.9135], device='cuda:1'), covar=tensor([0.1475, 0.0940, 0.2130, 0.1302, 0.1097, 0.1031, 0.1657, 0.2068], device='cuda:1'), in_proj_covar=tensor([0.0118, 0.0246, 0.0138, 0.0121, 0.0131, 0.0154, 0.0119, 0.0120], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-27 08:40:51,384 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4559, 1.4159, 4.0990, 3.8028, 3.6591, 3.8666, 3.8988, 3.6619], device='cuda:1'), covar=tensor([0.6627, 0.5107, 0.1097, 0.1791, 0.1001, 0.1488, 0.1198, 0.1522], device='cuda:1'), in_proj_covar=tensor([0.0308, 0.0301, 0.0398, 0.0400, 0.0343, 0.0403, 0.0306, 0.0358], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-27 08:40:54,851 INFO [finetune.py:976] (1/7) Epoch 16, batch 3500, loss[loss=0.1744, simple_loss=0.2363, pruned_loss=0.05629, over 4825.00 frames. ], tot_loss[loss=0.1828, simple_loss=0.2526, pruned_loss=0.05646, over 955282.17 frames. ], batch size: 33, lr: 3.43e-03, grad_scale: 32.0 2023-04-27 08:41:00,473 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=89425.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:41:07,512 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=89435.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:41:09,969 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=89439.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 08:41:18,073 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.096e+02 1.586e+02 1.877e+02 2.284e+02 3.930e+02, threshold=3.755e+02, percent-clipped=1.0 2023-04-27 08:41:32,332 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7641, 1.6249, 1.7649, 2.1430, 2.0504, 1.6941, 1.4278, 1.9069], device='cuda:1'), covar=tensor([0.0813, 0.1196, 0.0787, 0.0532, 0.0602, 0.0846, 0.0823, 0.0580], device='cuda:1'), in_proj_covar=tensor([0.0189, 0.0202, 0.0182, 0.0174, 0.0178, 0.0182, 0.0154, 0.0181], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 08:41:34,093 INFO [finetune.py:976] (1/7) Epoch 16, batch 3550, loss[loss=0.1901, simple_loss=0.2618, pruned_loss=0.05923, over 4928.00 frames. ], tot_loss[loss=0.1788, simple_loss=0.2481, pruned_loss=0.05477, over 956079.76 frames. ], batch size: 43, lr: 3.43e-03, grad_scale: 32.0 2023-04-27 08:41:58,282 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=89486.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:42:08,139 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=89500.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 08:42:20,186 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=89509.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:42:29,609 INFO [finetune.py:976] (1/7) Epoch 16, batch 3600, loss[loss=0.1486, simple_loss=0.2172, pruned_loss=0.04001, over 4870.00 frames. ], tot_loss[loss=0.176, simple_loss=0.2451, pruned_loss=0.05349, over 958266.22 frames. ], batch size: 34, lr: 3.43e-03, grad_scale: 32.0 2023-04-27 08:42:41,029 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=89525.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:43:05,072 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.63 vs. limit=2.0 2023-04-27 08:43:13,813 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.211e+02 1.668e+02 1.907e+02 2.305e+02 5.911e+02, threshold=3.815e+02, percent-clipped=3.0 2023-04-27 08:43:17,420 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-04-27 08:43:36,162 INFO [finetune.py:976] (1/7) Epoch 16, batch 3650, loss[loss=0.1752, simple_loss=0.2365, pruned_loss=0.05691, over 4698.00 frames. ], tot_loss[loss=0.1772, simple_loss=0.2463, pruned_loss=0.05411, over 955329.28 frames. ], batch size: 23, lr: 3.43e-03, grad_scale: 32.0 2023-04-27 08:43:38,717 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=89570.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:44:07,479 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4612, 1.4680, 1.7867, 1.8048, 1.3637, 1.1604, 1.5455, 0.9798], device='cuda:1'), covar=tensor([0.0625, 0.0676, 0.0421, 0.0679, 0.0755, 0.1158, 0.0663, 0.0696], device='cuda:1'), in_proj_covar=tensor([0.0068, 0.0071, 0.0070, 0.0068, 0.0076, 0.0098, 0.0076, 0.0069], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-27 08:44:42,831 INFO [finetune.py:976] (1/7) Epoch 16, batch 3700, loss[loss=0.1861, simple_loss=0.2648, pruned_loss=0.05373, over 4756.00 frames. ], tot_loss[loss=0.1791, simple_loss=0.2488, pruned_loss=0.05472, over 956577.01 frames. ], batch size: 27, lr: 3.43e-03, grad_scale: 32.0 2023-04-27 08:45:25,811 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-04-27 08:45:26,135 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.109e+02 1.724e+02 2.041e+02 2.428e+02 5.061e+02, threshold=4.082e+02, percent-clipped=2.0 2023-04-27 08:45:48,272 INFO [finetune.py:976] (1/7) Epoch 16, batch 3750, loss[loss=0.2084, simple_loss=0.286, pruned_loss=0.06543, over 4904.00 frames. ], tot_loss[loss=0.1805, simple_loss=0.2507, pruned_loss=0.05514, over 955748.31 frames. ], batch size: 37, lr: 3.43e-03, grad_scale: 32.0 2023-04-27 08:45:51,435 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5283, 1.2171, 4.3454, 4.0840, 3.7998, 4.1054, 3.9602, 3.7912], device='cuda:1'), covar=tensor([0.7085, 0.6014, 0.0908, 0.1499, 0.0966, 0.1477, 0.1514, 0.1486], device='cuda:1'), in_proj_covar=tensor([0.0312, 0.0305, 0.0403, 0.0404, 0.0347, 0.0409, 0.0311, 0.0363], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-27 08:46:02,514 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=89680.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:46:10,269 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=89683.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:46:49,773 INFO [finetune.py:976] (1/7) Epoch 16, batch 3800, loss[loss=0.1446, simple_loss=0.2241, pruned_loss=0.0325, over 4757.00 frames. ], tot_loss[loss=0.1814, simple_loss=0.2518, pruned_loss=0.05547, over 955067.86 frames. ], batch size: 27, lr: 3.43e-03, grad_scale: 32.0 2023-04-27 08:47:09,438 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=89730.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:47:10,031 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=89731.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:47:21,762 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=89741.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:47:31,967 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.072e+02 1.563e+02 1.915e+02 2.320e+02 3.976e+02, threshold=3.830e+02, percent-clipped=0.0 2023-04-27 08:47:33,206 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9134, 1.7668, 2.1113, 2.1672, 1.9990, 1.7619, 1.9325, 1.9362], device='cuda:1'), covar=tensor([0.6184, 0.8579, 0.8974, 0.8878, 0.7737, 1.1967, 1.1214, 1.2161], device='cuda:1'), in_proj_covar=tensor([0.0419, 0.0408, 0.0493, 0.0507, 0.0447, 0.0473, 0.0477, 0.0480], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 08:47:34,999 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0548, 2.5783, 2.1873, 2.4460, 1.8372, 2.2032, 2.2592, 1.6948], device='cuda:1'), covar=tensor([0.1880, 0.1148, 0.0793, 0.1178, 0.3202, 0.1155, 0.1843, 0.2555], device='cuda:1'), in_proj_covar=tensor([0.0288, 0.0312, 0.0223, 0.0282, 0.0314, 0.0265, 0.0253, 0.0268], device='cuda:1'), out_proj_covar=tensor([1.1590e-04, 1.2404e-04, 8.8884e-05, 1.1217e-04, 1.2798e-04, 1.0538e-04, 1.0222e-04, 1.0665e-04], device='cuda:1') 2023-04-27 08:47:39,486 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=3.69 vs. limit=5.0 2023-04-27 08:47:43,655 INFO [finetune.py:976] (1/7) Epoch 16, batch 3850, loss[loss=0.182, simple_loss=0.2451, pruned_loss=0.05941, over 4886.00 frames. ], tot_loss[loss=0.1802, simple_loss=0.2504, pruned_loss=0.055, over 956169.68 frames. ], batch size: 32, lr: 3.43e-03, grad_scale: 32.0 2023-04-27 08:47:58,321 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=89781.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:48:11,238 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0572, 4.2030, 0.8667, 2.1364, 2.3955, 2.8073, 2.4833, 0.9724], device='cuda:1'), covar=tensor([0.1282, 0.0843, 0.2069, 0.1286, 0.0955, 0.1061, 0.1421, 0.2203], device='cuda:1'), in_proj_covar=tensor([0.0117, 0.0245, 0.0138, 0.0121, 0.0131, 0.0153, 0.0118, 0.0120], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-27 08:48:12,385 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=89795.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 08:48:26,725 INFO [finetune.py:976] (1/7) Epoch 16, batch 3900, loss[loss=0.2211, simple_loss=0.2894, pruned_loss=0.07641, over 4823.00 frames. ], tot_loss[loss=0.179, simple_loss=0.2486, pruned_loss=0.05468, over 958718.14 frames. ], batch size: 30, lr: 3.43e-03, grad_scale: 32.0 2023-04-27 08:48:31,443 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([4.0626, 3.9641, 2.8704, 4.6920, 4.1350, 4.0728, 1.6142, 4.0101], device='cuda:1'), covar=tensor([0.1709, 0.0975, 0.3289, 0.1460, 0.2558, 0.1960, 0.6470, 0.2507], device='cuda:1'), in_proj_covar=tensor([0.0248, 0.0215, 0.0254, 0.0307, 0.0302, 0.0252, 0.0276, 0.0276], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 08:48:33,336 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=89825.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:48:48,869 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.776e+01 1.757e+02 2.140e+02 2.503e+02 6.967e+02, threshold=4.281e+02, percent-clipped=3.0 2023-04-27 08:48:52,098 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-04-27 08:48:59,090 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=89865.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:48:59,124 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=89865.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:48:59,621 INFO [finetune.py:976] (1/7) Epoch 16, batch 3950, loss[loss=0.1467, simple_loss=0.2162, pruned_loss=0.03859, over 4896.00 frames. ], tot_loss[loss=0.1758, simple_loss=0.2453, pruned_loss=0.05309, over 959511.96 frames. ], batch size: 32, lr: 3.43e-03, grad_scale: 32.0 2023-04-27 08:49:05,394 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=89873.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:49:33,476 INFO [finetune.py:976] (1/7) Epoch 16, batch 4000, loss[loss=0.2087, simple_loss=0.2743, pruned_loss=0.07154, over 4929.00 frames. ], tot_loss[loss=0.1755, simple_loss=0.2448, pruned_loss=0.05309, over 958761.01 frames. ], batch size: 38, lr: 3.43e-03, grad_scale: 32.0 2023-04-27 08:49:47,312 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=89926.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 08:50:11,839 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.112e+02 1.722e+02 2.012e+02 2.405e+02 4.800e+02, threshold=4.024e+02, percent-clipped=2.0 2023-04-27 08:50:30,451 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.18 vs. limit=5.0 2023-04-27 08:50:33,227 INFO [finetune.py:976] (1/7) Epoch 16, batch 4050, loss[loss=0.2247, simple_loss=0.2958, pruned_loss=0.07678, over 4813.00 frames. ], tot_loss[loss=0.1792, simple_loss=0.2487, pruned_loss=0.05485, over 958694.43 frames. ], batch size: 45, lr: 3.43e-03, grad_scale: 32.0 2023-04-27 08:50:57,142 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.62 vs. limit=5.0 2023-04-27 08:51:17,052 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.04 vs. limit=5.0 2023-04-27 08:51:40,298 INFO [finetune.py:976] (1/7) Epoch 16, batch 4100, loss[loss=0.1479, simple_loss=0.2245, pruned_loss=0.03562, over 4759.00 frames. ], tot_loss[loss=0.1799, simple_loss=0.2503, pruned_loss=0.05471, over 957869.95 frames. ], batch size: 27, lr: 3.43e-03, grad_scale: 32.0 2023-04-27 08:51:41,651 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7923, 1.1061, 1.3843, 1.3474, 1.9082, 1.5309, 1.2192, 1.3073], device='cuda:1'), covar=tensor([0.1758, 0.2144, 0.2491, 0.1813, 0.1122, 0.1955, 0.2480, 0.2700], device='cuda:1'), in_proj_covar=tensor([0.0308, 0.0313, 0.0352, 0.0291, 0.0329, 0.0312, 0.0302, 0.0365], device='cuda:1'), out_proj_covar=tensor([6.3654e-05, 6.5223e-05, 7.5158e-05, 5.9189e-05, 6.8382e-05, 6.5848e-05, 6.3744e-05, 7.8092e-05], device='cuda:1') 2023-04-27 08:52:02,673 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=90030.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:52:12,518 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=90036.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:52:27,089 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.030e+02 1.746e+02 2.004e+02 2.346e+02 5.282e+02, threshold=4.007e+02, percent-clipped=2.0 2023-04-27 08:52:48,665 INFO [finetune.py:976] (1/7) Epoch 16, batch 4150, loss[loss=0.1559, simple_loss=0.2166, pruned_loss=0.04761, over 4127.00 frames. ], tot_loss[loss=0.1811, simple_loss=0.2512, pruned_loss=0.05552, over 956671.61 frames. ], batch size: 17, lr: 3.43e-03, grad_scale: 32.0 2023-04-27 08:53:06,920 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=90078.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:53:09,296 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=90081.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:53:26,048 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=90095.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 08:53:50,593 INFO [finetune.py:976] (1/7) Epoch 16, batch 4200, loss[loss=0.1749, simple_loss=0.2448, pruned_loss=0.05252, over 4881.00 frames. ], tot_loss[loss=0.1811, simple_loss=0.2516, pruned_loss=0.05527, over 956046.81 frames. ], batch size: 32, lr: 3.43e-03, grad_scale: 32.0 2023-04-27 08:53:59,895 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=90129.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:54:00,573 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=90130.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:54:14,870 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=90143.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 08:54:24,333 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.851e+01 1.635e+02 1.941e+02 2.339e+02 3.883e+02, threshold=3.882e+02, percent-clipped=0.0 2023-04-27 08:54:34,085 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=90165.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:54:34,610 INFO [finetune.py:976] (1/7) Epoch 16, batch 4250, loss[loss=0.2201, simple_loss=0.2667, pruned_loss=0.08669, over 4933.00 frames. ], tot_loss[loss=0.1784, simple_loss=0.2486, pruned_loss=0.0541, over 955552.27 frames. ], batch size: 33, lr: 3.43e-03, grad_scale: 32.0 2023-04-27 08:54:52,886 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=90191.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:55:06,208 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=90213.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:55:08,443 INFO [finetune.py:976] (1/7) Epoch 16, batch 4300, loss[loss=0.185, simple_loss=0.2464, pruned_loss=0.06175, over 4137.00 frames. ], tot_loss[loss=0.1776, simple_loss=0.2472, pruned_loss=0.05405, over 953650.86 frames. ], batch size: 65, lr: 3.43e-03, grad_scale: 32.0 2023-04-27 08:55:08,622 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=3.50 vs. limit=5.0 2023-04-27 08:55:11,528 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=90221.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 08:55:19,268 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=90232.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:55:32,136 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.154e+02 1.571e+02 1.935e+02 2.216e+02 3.879e+02, threshold=3.870e+02, percent-clipped=0.0 2023-04-27 08:55:41,845 INFO [finetune.py:976] (1/7) Epoch 16, batch 4350, loss[loss=0.2062, simple_loss=0.2573, pruned_loss=0.0776, over 4821.00 frames. ], tot_loss[loss=0.1751, simple_loss=0.2442, pruned_loss=0.05298, over 955779.61 frames. ], batch size: 30, lr: 3.43e-03, grad_scale: 32.0 2023-04-27 08:56:00,581 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=90293.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:56:15,530 INFO [finetune.py:976] (1/7) Epoch 16, batch 4400, loss[loss=0.1967, simple_loss=0.2827, pruned_loss=0.05533, over 4811.00 frames. ], tot_loss[loss=0.1743, simple_loss=0.2437, pruned_loss=0.05242, over 954547.84 frames. ], batch size: 51, lr: 3.43e-03, grad_scale: 32.0 2023-04-27 08:56:34,939 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=90336.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:56:48,296 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([0.6606, 1.7137, 1.8658, 1.3656, 1.7853, 1.4853, 2.2446, 1.5846], device='cuda:1'), covar=tensor([0.3041, 0.1542, 0.3553, 0.2193, 0.1330, 0.2092, 0.1287, 0.3582], device='cuda:1'), in_proj_covar=tensor([0.0347, 0.0349, 0.0433, 0.0358, 0.0388, 0.0386, 0.0375, 0.0423], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 08:56:55,337 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.166e+02 1.655e+02 1.875e+02 2.353e+02 3.798e+02, threshold=3.751e+02, percent-clipped=0.0 2023-04-27 08:56:56,124 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=90351.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:57:05,587 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=3.79 vs. limit=5.0 2023-04-27 08:57:07,382 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9605, 1.7153, 2.1669, 2.4600, 2.0162, 1.8854, 2.0362, 1.9863], device='cuda:1'), covar=tensor([0.5006, 0.7173, 0.7329, 0.5968, 0.6449, 0.8750, 0.8926, 1.0322], device='cuda:1'), in_proj_covar=tensor([0.0419, 0.0405, 0.0493, 0.0506, 0.0446, 0.0471, 0.0476, 0.0480], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 08:57:10,216 INFO [finetune.py:976] (1/7) Epoch 16, batch 4450, loss[loss=0.2008, simple_loss=0.2709, pruned_loss=0.06538, over 4912.00 frames. ], tot_loss[loss=0.1766, simple_loss=0.2466, pruned_loss=0.05329, over 953898.00 frames. ], batch size: 36, lr: 3.43e-03, grad_scale: 32.0 2023-04-27 08:57:16,603 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-04-27 08:57:33,651 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=90384.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:57:38,506 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7417, 1.2618, 1.7496, 2.3147, 1.8712, 1.6598, 1.7062, 1.7181], device='cuda:1'), covar=tensor([0.4806, 0.7075, 0.6723, 0.5630, 0.5801, 0.7685, 0.8765, 0.9097], device='cuda:1'), in_proj_covar=tensor([0.0420, 0.0406, 0.0494, 0.0507, 0.0447, 0.0472, 0.0477, 0.0481], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 08:58:04,794 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=90412.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:58:07,139 INFO [finetune.py:976] (1/7) Epoch 16, batch 4500, loss[loss=0.1642, simple_loss=0.2417, pruned_loss=0.04333, over 4759.00 frames. ], tot_loss[loss=0.1769, simple_loss=0.2469, pruned_loss=0.05349, over 949710.60 frames. ], batch size: 26, lr: 3.43e-03, grad_scale: 64.0 2023-04-27 08:58:52,083 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.119e+02 1.589e+02 1.862e+02 2.271e+02 3.876e+02, threshold=3.723e+02, percent-clipped=1.0 2023-04-27 08:59:04,433 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=90459.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:59:14,179 INFO [finetune.py:976] (1/7) Epoch 16, batch 4550, loss[loss=0.1737, simple_loss=0.2677, pruned_loss=0.03987, over 4894.00 frames. ], tot_loss[loss=0.1776, simple_loss=0.248, pruned_loss=0.05357, over 948998.31 frames. ], batch size: 36, lr: 3.43e-03, grad_scale: 64.0 2023-04-27 08:59:37,664 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=90486.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:59:44,606 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-04-27 08:59:55,961 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=90496.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:00:12,183 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.1081, 2.6394, 2.2293, 2.4034, 1.8044, 2.1514, 2.2105, 1.7631], device='cuda:1'), covar=tensor([0.1798, 0.1073, 0.0751, 0.1108, 0.3072, 0.1131, 0.1648, 0.2321], device='cuda:1'), in_proj_covar=tensor([0.0287, 0.0310, 0.0222, 0.0281, 0.0314, 0.0263, 0.0252, 0.0266], device='cuda:1'), out_proj_covar=tensor([1.1548e-04, 1.2332e-04, 8.8541e-05, 1.1165e-04, 1.2781e-04, 1.0467e-04, 1.0174e-04, 1.0590e-04], device='cuda:1') 2023-04-27 09:00:15,728 INFO [finetune.py:976] (1/7) Epoch 16, batch 4600, loss[loss=0.17, simple_loss=0.2414, pruned_loss=0.0493, over 4404.00 frames. ], tot_loss[loss=0.1767, simple_loss=0.2477, pruned_loss=0.05283, over 950645.12 frames. ], batch size: 19, lr: 3.43e-03, grad_scale: 64.0 2023-04-27 09:00:18,339 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=90520.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:00:18,933 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=90521.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 09:00:32,581 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0973, 1.8085, 2.3595, 2.5981, 2.1421, 2.0339, 2.1970, 2.1688], device='cuda:1'), covar=tensor([0.5684, 0.7544, 0.8129, 0.6426, 0.6749, 0.9960, 0.9773, 0.9493], device='cuda:1'), in_proj_covar=tensor([0.0422, 0.0408, 0.0497, 0.0510, 0.0449, 0.0475, 0.0479, 0.0484], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 09:00:37,736 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.078e+02 1.675e+02 1.952e+02 2.329e+02 6.121e+02, threshold=3.905e+02, percent-clipped=3.0 2023-04-27 09:00:43,119 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=90557.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:00:47,044 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.1064, 2.6903, 1.0097, 1.5107, 1.9038, 1.4371, 3.6268, 1.8093], device='cuda:1'), covar=tensor([0.0678, 0.0580, 0.0830, 0.1222, 0.0555, 0.0911, 0.0240, 0.0666], device='cuda:1'), in_proj_covar=tensor([0.0051, 0.0066, 0.0048, 0.0046, 0.0050, 0.0052, 0.0075, 0.0052], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0007, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:1') 2023-04-27 09:00:49,341 INFO [finetune.py:976] (1/7) Epoch 16, batch 4650, loss[loss=0.1606, simple_loss=0.2262, pruned_loss=0.04751, over 4766.00 frames. ], tot_loss[loss=0.1768, simple_loss=0.247, pruned_loss=0.05334, over 951197.33 frames. ], batch size: 28, lr: 3.42e-03, grad_scale: 64.0 2023-04-27 09:00:51,263 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=90569.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:00:51,306 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.1653, 1.5013, 1.3343, 1.7092, 1.6111, 1.6949, 1.3518, 2.9077], device='cuda:1'), covar=tensor([0.0638, 0.0707, 0.0771, 0.1098, 0.0551, 0.0546, 0.0717, 0.0203], device='cuda:1'), in_proj_covar=tensor([0.0038, 0.0039, 0.0040, 0.0043, 0.0040, 0.0038, 0.0039, 0.0057], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:1') 2023-04-27 09:00:52,606 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.56 vs. limit=2.0 2023-04-27 09:00:53,177 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.68 vs. limit=2.0 2023-04-27 09:00:57,945 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7160, 1.3490, 1.8633, 2.2513, 1.8434, 1.7247, 1.8060, 1.7731], device='cuda:1'), covar=tensor([0.5328, 0.7162, 0.6997, 0.6473, 0.6355, 0.8245, 0.8240, 0.9457], device='cuda:1'), in_proj_covar=tensor([0.0423, 0.0409, 0.0499, 0.0511, 0.0451, 0.0476, 0.0480, 0.0485], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 09:01:03,238 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=90588.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:01:08,104 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.1980, 2.1532, 2.5560, 2.7897, 2.0076, 1.8450, 2.2850, 1.2152], device='cuda:1'), covar=tensor([0.0606, 0.0707, 0.0422, 0.0669, 0.0746, 0.1110, 0.0668, 0.0841], device='cuda:1'), in_proj_covar=tensor([0.0068, 0.0070, 0.0069, 0.0068, 0.0075, 0.0097, 0.0075, 0.0068], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-27 09:01:23,250 INFO [finetune.py:976] (1/7) Epoch 16, batch 4700, loss[loss=0.1762, simple_loss=0.2416, pruned_loss=0.05543, over 4874.00 frames. ], tot_loss[loss=0.1746, simple_loss=0.2442, pruned_loss=0.05256, over 953833.39 frames. ], batch size: 31, lr: 3.42e-03, grad_scale: 64.0 2023-04-27 09:01:36,757 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 2023-04-27 09:01:41,470 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-04-27 09:01:45,413 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.002e+02 1.583e+02 1.845e+02 2.338e+02 5.124e+02, threshold=3.690e+02, percent-clipped=1.0 2023-04-27 09:01:53,089 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5668, 3.1965, 0.8142, 1.7861, 1.9409, 2.2739, 1.9322, 0.9457], device='cuda:1'), covar=tensor([0.1263, 0.0967, 0.2029, 0.1192, 0.1006, 0.0968, 0.1436, 0.1945], device='cuda:1'), in_proj_covar=tensor([0.0117, 0.0243, 0.0137, 0.0121, 0.0131, 0.0153, 0.0118, 0.0121], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-27 09:02:01,952 INFO [finetune.py:976] (1/7) Epoch 16, batch 4750, loss[loss=0.1495, simple_loss=0.2219, pruned_loss=0.03853, over 4354.00 frames. ], tot_loss[loss=0.1731, simple_loss=0.2423, pruned_loss=0.052, over 952046.08 frames. ], batch size: 19, lr: 3.42e-03, grad_scale: 32.0 2023-04-27 09:02:37,493 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8835, 2.4549, 2.0464, 2.3236, 1.6763, 1.9481, 2.0541, 1.5282], device='cuda:1'), covar=tensor([0.2253, 0.1273, 0.0873, 0.1268, 0.3323, 0.1354, 0.2115, 0.3114], device='cuda:1'), in_proj_covar=tensor([0.0288, 0.0310, 0.0223, 0.0282, 0.0314, 0.0263, 0.0252, 0.0268], device='cuda:1'), out_proj_covar=tensor([1.1566e-04, 1.2339e-04, 8.8680e-05, 1.1186e-04, 1.2788e-04, 1.0474e-04, 1.0186e-04, 1.0637e-04], device='cuda:1') 2023-04-27 09:02:51,143 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=90707.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:03:07,871 INFO [finetune.py:976] (1/7) Epoch 16, batch 4800, loss[loss=0.2593, simple_loss=0.3099, pruned_loss=0.1043, over 4805.00 frames. ], tot_loss[loss=0.1758, simple_loss=0.2452, pruned_loss=0.05326, over 951045.27 frames. ], batch size: 41, lr: 3.42e-03, grad_scale: 32.0 2023-04-27 09:03:10,285 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=3.56 vs. limit=5.0 2023-04-27 09:03:36,199 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.106e+02 1.708e+02 2.005e+02 2.486e+02 4.014e+02, threshold=4.009e+02, percent-clipped=3.0 2023-04-27 09:03:47,214 INFO [finetune.py:976] (1/7) Epoch 16, batch 4850, loss[loss=0.189, simple_loss=0.261, pruned_loss=0.05846, over 4779.00 frames. ], tot_loss[loss=0.1778, simple_loss=0.248, pruned_loss=0.05376, over 952139.08 frames. ], batch size: 28, lr: 3.42e-03, grad_scale: 32.0 2023-04-27 09:03:50,415 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-04-27 09:04:00,357 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=90786.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:04:36,232 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=90815.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:04:36,790 INFO [finetune.py:976] (1/7) Epoch 16, batch 4900, loss[loss=0.1916, simple_loss=0.2556, pruned_loss=0.06374, over 4860.00 frames. ], tot_loss[loss=0.1806, simple_loss=0.2508, pruned_loss=0.05525, over 954019.57 frames. ], batch size: 34, lr: 3.42e-03, grad_scale: 32.0 2023-04-27 09:04:50,424 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-04-27 09:05:01,132 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=90834.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:05:19,767 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-04-27 09:05:21,464 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=90849.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:05:22,604 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.194e+02 1.643e+02 1.978e+02 2.332e+02 3.633e+02, threshold=3.955e+02, percent-clipped=0.0 2023-04-27 09:05:23,821 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=90852.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:05:38,745 INFO [finetune.py:976] (1/7) Epoch 16, batch 4950, loss[loss=0.2088, simple_loss=0.2725, pruned_loss=0.07256, over 4734.00 frames. ], tot_loss[loss=0.1817, simple_loss=0.252, pruned_loss=0.05572, over 952491.00 frames. ], batch size: 59, lr: 3.42e-03, grad_scale: 32.0 2023-04-27 09:05:54,237 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=90888.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:06:02,896 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9131, 2.3304, 1.1850, 1.5917, 2.2563, 1.7815, 1.6760, 1.8461], device='cuda:1'), covar=tensor([0.0467, 0.0326, 0.0302, 0.0543, 0.0242, 0.0512, 0.0543, 0.0517], device='cuda:1'), in_proj_covar=tensor([0.0029, 0.0024, 0.0022, 0.0029, 0.0020, 0.0028, 0.0028, 0.0029], device='cuda:1'), out_proj_covar=tensor([0.0050, 0.0045, 0.0038, 0.0051, 0.0038, 0.0049, 0.0049, 0.0050], device='cuda:1') 2023-04-27 09:06:03,490 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5133, 1.1211, 0.3534, 1.2133, 1.1671, 1.4013, 1.2639, 1.2963], device='cuda:1'), covar=tensor([0.0538, 0.0429, 0.0456, 0.0594, 0.0318, 0.0538, 0.0549, 0.0607], device='cuda:1'), in_proj_covar=tensor([0.0029, 0.0024, 0.0022, 0.0029, 0.0020, 0.0028, 0.0028, 0.0029], device='cuda:1'), out_proj_covar=tensor([0.0050, 0.0045, 0.0038, 0.0051, 0.0038, 0.0049, 0.0049, 0.0050], device='cuda:1') 2023-04-27 09:06:08,379 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=90910.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:06:12,381 INFO [finetune.py:976] (1/7) Epoch 16, batch 5000, loss[loss=0.1508, simple_loss=0.2121, pruned_loss=0.04469, over 4186.00 frames. ], tot_loss[loss=0.1791, simple_loss=0.2494, pruned_loss=0.05438, over 952441.32 frames. ], batch size: 17, lr: 3.42e-03, grad_scale: 32.0 2023-04-27 09:06:27,004 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=90936.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:06:35,698 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-04-27 09:06:36,100 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.088e+02 1.563e+02 1.794e+02 2.191e+02 3.401e+02, threshold=3.587e+02, percent-clipped=0.0 2023-04-27 09:06:46,205 INFO [finetune.py:976] (1/7) Epoch 16, batch 5050, loss[loss=0.1699, simple_loss=0.2348, pruned_loss=0.05253, over 4758.00 frames. ], tot_loss[loss=0.178, simple_loss=0.2477, pruned_loss=0.05416, over 953889.31 frames. ], batch size: 26, lr: 3.42e-03, grad_scale: 32.0 2023-04-27 09:07:13,871 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=91007.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:07:19,797 INFO [finetune.py:976] (1/7) Epoch 16, batch 5100, loss[loss=0.1745, simple_loss=0.2283, pruned_loss=0.06031, over 4838.00 frames. ], tot_loss[loss=0.1749, simple_loss=0.2443, pruned_loss=0.05271, over 954764.94 frames. ], batch size: 30, lr: 3.42e-03, grad_scale: 32.0 2023-04-27 09:07:43,899 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.024e+02 1.518e+02 1.785e+02 2.206e+02 4.900e+02, threshold=3.570e+02, percent-clipped=2.0 2023-04-27 09:07:46,425 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=91055.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:07:53,570 INFO [finetune.py:976] (1/7) Epoch 16, batch 5150, loss[loss=0.2278, simple_loss=0.2907, pruned_loss=0.0825, over 4831.00 frames. ], tot_loss[loss=0.1756, simple_loss=0.2448, pruned_loss=0.05315, over 953257.31 frames. ], batch size: 40, lr: 3.42e-03, grad_scale: 32.0 2023-04-27 09:08:48,477 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=91115.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:08:48,975 INFO [finetune.py:976] (1/7) Epoch 16, batch 5200, loss[loss=0.1893, simple_loss=0.2644, pruned_loss=0.05708, over 4797.00 frames. ], tot_loss[loss=0.1791, simple_loss=0.2486, pruned_loss=0.05481, over 949928.84 frames. ], batch size: 29, lr: 3.42e-03, grad_scale: 32.0 2023-04-27 09:09:38,601 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.048e+02 1.654e+02 2.065e+02 2.493e+02 5.806e+02, threshold=4.130e+02, percent-clipped=4.0 2023-04-27 09:09:39,335 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=91152.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:09:51,477 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=91163.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:09:53,288 INFO [finetune.py:976] (1/7) Epoch 16, batch 5250, loss[loss=0.1525, simple_loss=0.2011, pruned_loss=0.05192, over 4057.00 frames. ], tot_loss[loss=0.1796, simple_loss=0.2496, pruned_loss=0.05483, over 948856.62 frames. ], batch size: 17, lr: 3.42e-03, grad_scale: 32.0 2023-04-27 09:10:44,542 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=91200.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:10:46,029 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-04-27 09:10:47,633 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=91205.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:11:00,424 INFO [finetune.py:976] (1/7) Epoch 16, batch 5300, loss[loss=0.2023, simple_loss=0.2809, pruned_loss=0.06189, over 4846.00 frames. ], tot_loss[loss=0.1795, simple_loss=0.2498, pruned_loss=0.0546, over 948262.12 frames. ], batch size: 44, lr: 3.42e-03, grad_scale: 32.0 2023-04-27 09:11:10,764 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9419, 1.5454, 1.4194, 1.7442, 2.1819, 1.7080, 1.4265, 1.3878], device='cuda:1'), covar=tensor([0.1677, 0.1519, 0.2362, 0.1241, 0.0798, 0.1745, 0.2359, 0.2396], device='cuda:1'), in_proj_covar=tensor([0.0304, 0.0309, 0.0348, 0.0285, 0.0324, 0.0310, 0.0299, 0.0361], device='cuda:1'), out_proj_covar=tensor([6.2771e-05, 6.4377e-05, 7.4233e-05, 5.7977e-05, 6.7381e-05, 6.5304e-05, 6.3173e-05, 7.7096e-05], device='cuda:1') 2023-04-27 09:11:25,281 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.127e+02 1.661e+02 1.899e+02 2.252e+02 3.532e+02, threshold=3.799e+02, percent-clipped=0.0 2023-04-27 09:11:34,433 INFO [finetune.py:976] (1/7) Epoch 16, batch 5350, loss[loss=0.1722, simple_loss=0.243, pruned_loss=0.05073, over 4919.00 frames. ], tot_loss[loss=0.1797, simple_loss=0.2505, pruned_loss=0.0544, over 946826.36 frames. ], batch size: 38, lr: 3.42e-03, grad_scale: 32.0 2023-04-27 09:11:47,132 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=91285.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:12:08,304 INFO [finetune.py:976] (1/7) Epoch 16, batch 5400, loss[loss=0.1727, simple_loss=0.2407, pruned_loss=0.05229, over 4793.00 frames. ], tot_loss[loss=0.1774, simple_loss=0.2478, pruned_loss=0.0535, over 947789.08 frames. ], batch size: 29, lr: 3.42e-03, grad_scale: 32.0 2023-04-27 09:12:28,655 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=91346.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:12:32,094 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.075e+02 1.605e+02 2.003e+02 2.320e+02 6.253e+02, threshold=4.007e+02, percent-clipped=1.0 2023-04-27 09:12:42,146 INFO [finetune.py:976] (1/7) Epoch 16, batch 5450, loss[loss=0.1505, simple_loss=0.2193, pruned_loss=0.04084, over 4937.00 frames. ], tot_loss[loss=0.1776, simple_loss=0.2474, pruned_loss=0.05395, over 950879.52 frames. ], batch size: 38, lr: 3.42e-03, grad_scale: 32.0 2023-04-27 09:13:15,970 INFO [finetune.py:976] (1/7) Epoch 16, batch 5500, loss[loss=0.1567, simple_loss=0.2228, pruned_loss=0.04526, over 4814.00 frames. ], tot_loss[loss=0.1732, simple_loss=0.2427, pruned_loss=0.05185, over 953724.87 frames. ], batch size: 25, lr: 3.42e-03, grad_scale: 32.0 2023-04-27 09:13:26,426 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.1891, 1.6070, 2.0557, 2.3598, 2.0346, 1.5973, 1.1865, 1.6795], device='cuda:1'), covar=tensor([0.3470, 0.3403, 0.1734, 0.2199, 0.2830, 0.2809, 0.4224, 0.2183], device='cuda:1'), in_proj_covar=tensor([0.0292, 0.0247, 0.0228, 0.0316, 0.0218, 0.0231, 0.0229, 0.0187], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-27 09:13:38,281 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.525e+01 1.617e+02 1.882e+02 2.275e+02 4.330e+02, threshold=3.765e+02, percent-clipped=1.0 2023-04-27 09:13:49,958 INFO [finetune.py:976] (1/7) Epoch 16, batch 5550, loss[loss=0.1523, simple_loss=0.2271, pruned_loss=0.03873, over 4779.00 frames. ], tot_loss[loss=0.1756, simple_loss=0.2449, pruned_loss=0.05312, over 953969.67 frames. ], batch size: 28, lr: 3.42e-03, grad_scale: 32.0 2023-04-27 09:14:01,222 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-04-27 09:14:14,709 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=91505.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:14:23,518 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([0.9700, 0.9972, 1.1676, 1.1439, 0.9611, 0.8849, 0.8839, 0.4674], device='cuda:1'), covar=tensor([0.0522, 0.0520, 0.0447, 0.0468, 0.0688, 0.1186, 0.0451, 0.0714], device='cuda:1'), in_proj_covar=tensor([0.0068, 0.0070, 0.0069, 0.0068, 0.0076, 0.0097, 0.0075, 0.0068], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-27 09:14:32,824 INFO [finetune.py:976] (1/7) Epoch 16, batch 5600, loss[loss=0.2201, simple_loss=0.304, pruned_loss=0.06803, over 4910.00 frames. ], tot_loss[loss=0.1771, simple_loss=0.2478, pruned_loss=0.05314, over 956490.07 frames. ], batch size: 36, lr: 3.42e-03, grad_scale: 32.0 2023-04-27 09:15:04,814 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=91542.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 09:15:15,397 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.016e+02 1.634e+02 1.984e+02 2.410e+02 6.572e+02, threshold=3.968e+02, percent-clipped=2.0 2023-04-27 09:15:16,631 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=91553.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:15:19,030 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0990, 1.6680, 1.9025, 2.4847, 1.9346, 1.5477, 1.4155, 1.8324], device='cuda:1'), covar=tensor([0.2675, 0.2814, 0.1643, 0.1791, 0.2250, 0.2435, 0.4117, 0.1875], device='cuda:1'), in_proj_covar=tensor([0.0291, 0.0247, 0.0227, 0.0315, 0.0217, 0.0231, 0.0228, 0.0186], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-27 09:15:28,400 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-04-27 09:15:29,903 INFO [finetune.py:976] (1/7) Epoch 16, batch 5650, loss[loss=0.2397, simple_loss=0.3103, pruned_loss=0.08453, over 4850.00 frames. ], tot_loss[loss=0.1786, simple_loss=0.2503, pruned_loss=0.0535, over 955790.72 frames. ], batch size: 47, lr: 3.42e-03, grad_scale: 32.0 2023-04-27 09:15:40,247 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-04-27 09:16:20,697 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=91603.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 09:16:34,226 INFO [finetune.py:976] (1/7) Epoch 16, batch 5700, loss[loss=0.1358, simple_loss=0.2002, pruned_loss=0.03577, over 4409.00 frames. ], tot_loss[loss=0.1767, simple_loss=0.247, pruned_loss=0.05323, over 937435.17 frames. ], batch size: 19, lr: 3.42e-03, grad_scale: 32.0 2023-04-27 09:17:06,022 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=91641.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:17:19,306 INFO [finetune.py:976] (1/7) Epoch 17, batch 0, loss[loss=0.2156, simple_loss=0.2794, pruned_loss=0.07593, over 4892.00 frames. ], tot_loss[loss=0.2156, simple_loss=0.2794, pruned_loss=0.07593, over 4892.00 frames. ], batch size: 36, lr: 3.41e-03, grad_scale: 32.0 2023-04-27 09:17:19,306 INFO [finetune.py:1001] (1/7) Computing validation loss 2023-04-27 09:17:21,415 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7711, 1.0145, 1.6696, 2.2253, 1.8752, 1.6886, 1.6964, 1.6933], device='cuda:1'), covar=tensor([0.4923, 0.7186, 0.6648, 0.6448, 0.6243, 0.8300, 0.8471, 0.8739], device='cuda:1'), in_proj_covar=tensor([0.0420, 0.0408, 0.0496, 0.0506, 0.0449, 0.0471, 0.0479, 0.0485], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 09:17:21,474 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.3379, 1.3324, 3.8661, 3.5268, 3.4428, 3.6775, 3.7684, 3.3869], device='cuda:1'), covar=tensor([0.7369, 0.5530, 0.1289, 0.2256, 0.1306, 0.1586, 0.0781, 0.1641], device='cuda:1'), in_proj_covar=tensor([0.0308, 0.0303, 0.0400, 0.0400, 0.0345, 0.0408, 0.0307, 0.0361], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-27 09:17:22,006 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4786, 1.3174, 1.7000, 1.6747, 1.3352, 1.2777, 1.3956, 0.9084], device='cuda:1'), covar=tensor([0.0705, 0.0838, 0.0523, 0.0700, 0.1045, 0.1320, 0.0756, 0.0756], device='cuda:1'), in_proj_covar=tensor([0.0068, 0.0070, 0.0069, 0.0068, 0.0075, 0.0097, 0.0075, 0.0067], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-27 09:17:40,813 INFO [finetune.py:1010] (1/7) Epoch 17, validation: loss=0.1535, simple_loss=0.2247, pruned_loss=0.04111, over 2265189.00 frames. 2023-04-27 09:17:40,814 INFO [finetune.py:1011] (1/7) Maximum memory allocated so far is 6391MB 2023-04-27 09:17:45,697 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.062e+02 1.575e+02 1.838e+02 2.202e+02 3.811e+02, threshold=3.676e+02, percent-clipped=0.0 2023-04-27 09:18:20,579 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.62 vs. limit=2.0 2023-04-27 09:18:24,711 INFO [finetune.py:976] (1/7) Epoch 17, batch 50, loss[loss=0.1389, simple_loss=0.2128, pruned_loss=0.03254, over 4762.00 frames. ], tot_loss[loss=0.1827, simple_loss=0.2522, pruned_loss=0.0566, over 215313.74 frames. ], batch size: 27, lr: 3.41e-03, grad_scale: 32.0 2023-04-27 09:18:32,663 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([0.9832, 2.0055, 1.7374, 1.7012, 2.0723, 1.6549, 2.5295, 1.4291], device='cuda:1'), covar=tensor([0.3453, 0.1678, 0.4717, 0.2535, 0.1550, 0.2337, 0.1253, 0.5147], device='cuda:1'), in_proj_covar=tensor([0.0343, 0.0348, 0.0432, 0.0359, 0.0385, 0.0383, 0.0374, 0.0424], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 09:18:39,438 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.56 vs. limit=2.0 2023-04-27 09:18:45,329 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.1412, 3.4197, 0.8893, 1.5969, 1.7577, 2.2006, 1.9397, 1.0025], device='cuda:1'), covar=tensor([0.2193, 0.1747, 0.2679, 0.1987, 0.1472, 0.1546, 0.1879, 0.2289], device='cuda:1'), in_proj_covar=tensor([0.0118, 0.0243, 0.0137, 0.0121, 0.0132, 0.0153, 0.0117, 0.0120], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-27 09:18:57,082 INFO [finetune.py:976] (1/7) Epoch 17, batch 100, loss[loss=0.1576, simple_loss=0.2256, pruned_loss=0.04476, over 4820.00 frames. ], tot_loss[loss=0.1794, simple_loss=0.248, pruned_loss=0.0554, over 378556.60 frames. ], batch size: 40, lr: 3.41e-03, grad_scale: 32.0 2023-04-27 09:19:02,435 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.149e+02 1.653e+02 1.974e+02 2.303e+02 4.426e+02, threshold=3.948e+02, percent-clipped=2.0 2023-04-27 09:19:03,974 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-04-27 09:19:30,053 INFO [finetune.py:976] (1/7) Epoch 17, batch 150, loss[loss=0.1697, simple_loss=0.2557, pruned_loss=0.04188, over 4868.00 frames. ], tot_loss[loss=0.1759, simple_loss=0.2437, pruned_loss=0.05409, over 508867.29 frames. ], batch size: 34, lr: 3.41e-03, grad_scale: 32.0 2023-04-27 09:20:03,540 INFO [finetune.py:976] (1/7) Epoch 17, batch 200, loss[loss=0.1575, simple_loss=0.2256, pruned_loss=0.04466, over 4871.00 frames. ], tot_loss[loss=0.1753, simple_loss=0.2428, pruned_loss=0.05387, over 609879.18 frames. ], batch size: 31, lr: 3.41e-03, grad_scale: 32.0 2023-04-27 09:20:08,959 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.596e+01 1.677e+02 1.899e+02 2.270e+02 6.599e+02, threshold=3.797e+02, percent-clipped=2.0 2023-04-27 09:20:35,891 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=91891.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:20:36,995 INFO [finetune.py:976] (1/7) Epoch 17, batch 250, loss[loss=0.1858, simple_loss=0.246, pruned_loss=0.06276, over 4773.00 frames. ], tot_loss[loss=0.1797, simple_loss=0.2481, pruned_loss=0.05568, over 685568.59 frames. ], batch size: 26, lr: 3.41e-03, grad_scale: 32.0 2023-04-27 09:20:40,144 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=91898.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 09:21:05,113 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.7600, 1.8950, 2.2930, 3.0186, 3.0499, 2.5137, 2.0074, 2.7843], device='cuda:1'), covar=tensor([0.1039, 0.1710, 0.1007, 0.0763, 0.0669, 0.1108, 0.1067, 0.0729], device='cuda:1'), in_proj_covar=tensor([0.0189, 0.0200, 0.0181, 0.0172, 0.0176, 0.0181, 0.0152, 0.0180], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 09:21:08,452 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=91941.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:21:10,055 INFO [finetune.py:976] (1/7) Epoch 17, batch 300, loss[loss=0.2014, simple_loss=0.2757, pruned_loss=0.06354, over 4821.00 frames. ], tot_loss[loss=0.1816, simple_loss=0.251, pruned_loss=0.05608, over 746381.44 frames. ], batch size: 33, lr: 3.41e-03, grad_scale: 32.0 2023-04-27 09:21:14,190 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-04-27 09:21:15,845 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.098e+02 1.651e+02 1.943e+02 2.275e+02 4.588e+02, threshold=3.886e+02, percent-clipped=2.0 2023-04-27 09:21:16,620 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=91952.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:21:17,787 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([3.6735, 3.6349, 2.7129, 4.2522, 3.6819, 3.6859, 1.5977, 3.6461], device='cuda:1'), covar=tensor([0.1921, 0.1334, 0.3101, 0.1862, 0.5518, 0.1918, 0.6221, 0.2663], device='cuda:1'), in_proj_covar=tensor([0.0247, 0.0214, 0.0253, 0.0306, 0.0300, 0.0250, 0.0275, 0.0275], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 09:21:18,447 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=91955.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:21:30,540 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5383, 1.5199, 0.5645, 1.2599, 1.7047, 1.4389, 1.3234, 1.3690], device='cuda:1'), covar=tensor([0.0467, 0.0356, 0.0378, 0.0532, 0.0259, 0.0484, 0.0464, 0.0548], device='cuda:1'), in_proj_covar=tensor([0.0028, 0.0024, 0.0022, 0.0029, 0.0020, 0.0028, 0.0028, 0.0028], device='cuda:1'), out_proj_covar=tensor([0.0050, 0.0044, 0.0038, 0.0051, 0.0038, 0.0049, 0.0049, 0.0050], device='cuda:1') 2023-04-27 09:21:41,950 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4788, 1.3833, 1.8775, 1.7640, 1.3493, 1.1616, 1.4959, 0.8878], device='cuda:1'), covar=tensor([0.0528, 0.0649, 0.0333, 0.0577, 0.0643, 0.1071, 0.0548, 0.0622], device='cuda:1'), in_proj_covar=tensor([0.0067, 0.0069, 0.0068, 0.0067, 0.0074, 0.0096, 0.0074, 0.0067], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-27 09:21:43,850 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=3.86 vs. limit=5.0 2023-04-27 09:21:51,284 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=91981.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 09:21:56,126 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=91989.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:22:03,166 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-04-27 09:22:04,088 INFO [finetune.py:976] (1/7) Epoch 17, batch 350, loss[loss=0.1903, simple_loss=0.2701, pruned_loss=0.05524, over 4828.00 frames. ], tot_loss[loss=0.1814, simple_loss=0.2517, pruned_loss=0.05555, over 793025.58 frames. ], batch size: 30, lr: 3.41e-03, grad_scale: 32.0 2023-04-27 09:22:36,914 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=92016.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:22:57,027 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([4.5417, 4.3959, 3.1766, 5.2435, 4.5445, 4.5358, 2.1228, 4.5735], device='cuda:1'), covar=tensor([0.1653, 0.1060, 0.2946, 0.0964, 0.3523, 0.1619, 0.5212, 0.2060], device='cuda:1'), in_proj_covar=tensor([0.0247, 0.0213, 0.0252, 0.0306, 0.0299, 0.0249, 0.0274, 0.0274], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 09:23:09,177 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=92042.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 09:23:09,648 INFO [finetune.py:976] (1/7) Epoch 17, batch 400, loss[loss=0.1833, simple_loss=0.2637, pruned_loss=0.05145, over 4824.00 frames. ], tot_loss[loss=0.1819, simple_loss=0.2526, pruned_loss=0.05557, over 829104.38 frames. ], batch size: 39, lr: 3.41e-03, grad_scale: 32.0 2023-04-27 09:23:21,383 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.108e+02 1.614e+02 1.906e+02 2.282e+02 3.471e+02, threshold=3.811e+02, percent-clipped=0.0 2023-04-27 09:23:31,583 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-04-27 09:24:03,171 INFO [finetune.py:976] (1/7) Epoch 17, batch 450, loss[loss=0.2008, simple_loss=0.2589, pruned_loss=0.07136, over 4864.00 frames. ], tot_loss[loss=0.1791, simple_loss=0.2498, pruned_loss=0.05414, over 856920.13 frames. ], batch size: 31, lr: 3.41e-03, grad_scale: 32.0 2023-04-27 09:24:36,847 INFO [finetune.py:976] (1/7) Epoch 17, batch 500, loss[loss=0.1524, simple_loss=0.2299, pruned_loss=0.03745, over 4874.00 frames. ], tot_loss[loss=0.1771, simple_loss=0.2475, pruned_loss=0.05335, over 880651.59 frames. ], batch size: 34, lr: 3.41e-03, grad_scale: 32.0 2023-04-27 09:24:42,155 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.066e+02 1.524e+02 1.904e+02 2.252e+02 3.809e+02, threshold=3.808e+02, percent-clipped=0.0 2023-04-27 09:25:10,284 INFO [finetune.py:976] (1/7) Epoch 17, batch 550, loss[loss=0.1842, simple_loss=0.2572, pruned_loss=0.05556, over 4756.00 frames. ], tot_loss[loss=0.1754, simple_loss=0.2451, pruned_loss=0.05288, over 898256.99 frames. ], batch size: 54, lr: 3.41e-03, grad_scale: 32.0 2023-04-27 09:25:13,459 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=92198.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 09:25:44,127 INFO [finetune.py:976] (1/7) Epoch 17, batch 600, loss[loss=0.1495, simple_loss=0.2263, pruned_loss=0.03638, over 4817.00 frames. ], tot_loss[loss=0.1757, simple_loss=0.2451, pruned_loss=0.05318, over 909288.98 frames. ], batch size: 30, lr: 3.41e-03, grad_scale: 32.0 2023-04-27 09:25:46,053 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=92246.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 09:25:46,669 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=92247.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:25:49,030 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.221e+02 1.690e+02 1.979e+02 2.477e+02 6.011e+02, threshold=3.959e+02, percent-clipped=1.0 2023-04-27 09:26:17,387 INFO [finetune.py:976] (1/7) Epoch 17, batch 650, loss[loss=0.1191, simple_loss=0.1904, pruned_loss=0.02387, over 4757.00 frames. ], tot_loss[loss=0.1786, simple_loss=0.2484, pruned_loss=0.05442, over 920321.43 frames. ], batch size: 28, lr: 3.41e-03, grad_scale: 32.0 2023-04-27 09:26:29,965 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=92311.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:26:56,500 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=92337.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 09:27:00,059 INFO [finetune.py:976] (1/7) Epoch 17, batch 700, loss[loss=0.1476, simple_loss=0.2177, pruned_loss=0.03874, over 4793.00 frames. ], tot_loss[loss=0.1792, simple_loss=0.2495, pruned_loss=0.05444, over 927811.64 frames. ], batch size: 25, lr: 3.41e-03, grad_scale: 32.0 2023-04-27 09:27:10,674 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.019e+02 1.654e+02 1.859e+02 2.105e+02 3.715e+02, threshold=3.718e+02, percent-clipped=1.0 2023-04-27 09:27:49,356 INFO [finetune.py:976] (1/7) Epoch 17, batch 750, loss[loss=0.2014, simple_loss=0.2595, pruned_loss=0.07166, over 4740.00 frames. ], tot_loss[loss=0.1797, simple_loss=0.2504, pruned_loss=0.05452, over 936542.14 frames. ], batch size: 54, lr: 3.41e-03, grad_scale: 32.0 2023-04-27 09:28:44,566 INFO [finetune.py:976] (1/7) Epoch 17, batch 800, loss[loss=0.1687, simple_loss=0.2464, pruned_loss=0.04548, over 4769.00 frames. ], tot_loss[loss=0.1786, simple_loss=0.2494, pruned_loss=0.05388, over 940663.64 frames. ], batch size: 26, lr: 3.41e-03, grad_scale: 32.0 2023-04-27 09:28:54,788 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.154e+02 1.529e+02 1.784e+02 2.077e+02 4.006e+02, threshold=3.568e+02, percent-clipped=1.0 2023-04-27 09:29:04,794 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-04-27 09:29:31,910 INFO [finetune.py:976] (1/7) Epoch 17, batch 850, loss[loss=0.117, simple_loss=0.1826, pruned_loss=0.02573, over 4229.00 frames. ], tot_loss[loss=0.1765, simple_loss=0.247, pruned_loss=0.05304, over 945602.31 frames. ], batch size: 18, lr: 3.41e-03, grad_scale: 32.0 2023-04-27 09:29:44,877 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0736, 2.6286, 1.0679, 1.3484, 1.8911, 1.2541, 3.3909, 1.6023], device='cuda:1'), covar=tensor([0.0701, 0.0784, 0.0889, 0.1162, 0.0536, 0.0973, 0.0247, 0.0664], device='cuda:1'), in_proj_covar=tensor([0.0051, 0.0066, 0.0048, 0.0046, 0.0050, 0.0052, 0.0075, 0.0051], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0007, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:1') 2023-04-27 09:29:56,005 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8839, 1.8156, 2.2677, 2.4051, 1.8270, 1.6576, 1.9803, 1.0666], device='cuda:1'), covar=tensor([0.0674, 0.0823, 0.0514, 0.0841, 0.0810, 0.1290, 0.0803, 0.0880], device='cuda:1'), in_proj_covar=tensor([0.0068, 0.0069, 0.0068, 0.0067, 0.0075, 0.0096, 0.0075, 0.0067], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-27 09:30:02,094 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2023-04-27 09:30:05,439 INFO [finetune.py:976] (1/7) Epoch 17, batch 900, loss[loss=0.1502, simple_loss=0.2114, pruned_loss=0.04449, over 4940.00 frames. ], tot_loss[loss=0.1735, simple_loss=0.2434, pruned_loss=0.05175, over 947773.84 frames. ], batch size: 33, lr: 3.41e-03, grad_scale: 32.0 2023-04-27 09:30:07,971 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=92547.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:30:10,301 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.562e+01 1.556e+02 1.849e+02 2.275e+02 6.056e+02, threshold=3.698e+02, percent-clipped=3.0 2023-04-27 09:30:38,498 INFO [finetune.py:976] (1/7) Epoch 17, batch 950, loss[loss=0.1634, simple_loss=0.2431, pruned_loss=0.04189, over 4755.00 frames. ], tot_loss[loss=0.1734, simple_loss=0.2429, pruned_loss=0.05194, over 948019.48 frames. ], batch size: 27, lr: 3.40e-03, grad_scale: 32.0 2023-04-27 09:30:39,819 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=92595.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:30:49,719 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=92611.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:30:59,818 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=92625.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:31:07,674 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=92637.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 09:31:12,292 INFO [finetune.py:976] (1/7) Epoch 17, batch 1000, loss[loss=0.1882, simple_loss=0.2662, pruned_loss=0.05511, over 4743.00 frames. ], tot_loss[loss=0.1773, simple_loss=0.2469, pruned_loss=0.0539, over 948885.28 frames. ], batch size: 54, lr: 3.40e-03, grad_scale: 32.0 2023-04-27 09:31:16,129 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7205, 2.0939, 1.8271, 1.9999, 1.5641, 1.8633, 1.6647, 1.4180], device='cuda:1'), covar=tensor([0.1863, 0.1528, 0.0874, 0.1223, 0.3801, 0.1128, 0.2072, 0.2533], device='cuda:1'), in_proj_covar=tensor([0.0284, 0.0306, 0.0219, 0.0277, 0.0312, 0.0259, 0.0250, 0.0265], device='cuda:1'), out_proj_covar=tensor([1.1404e-04, 1.2161e-04, 8.7013e-05, 1.1013e-04, 1.2690e-04, 1.0302e-04, 1.0084e-04, 1.0537e-04], device='cuda:1') 2023-04-27 09:31:17,225 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.750e+01 1.761e+02 2.088e+02 2.403e+02 5.880e+02, threshold=4.175e+02, percent-clipped=3.0 2023-04-27 09:31:22,262 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=92659.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:31:40,694 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=92685.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 09:31:41,314 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=92686.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:31:45,965 INFO [finetune.py:976] (1/7) Epoch 17, batch 1050, loss[loss=0.1631, simple_loss=0.2246, pruned_loss=0.05077, over 4931.00 frames. ], tot_loss[loss=0.1787, simple_loss=0.2493, pruned_loss=0.05406, over 949889.30 frames. ], batch size: 33, lr: 3.40e-03, grad_scale: 64.0 2023-04-27 09:32:02,616 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.11 vs. limit=2.0 2023-04-27 09:32:03,087 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=92720.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:32:30,324 INFO [finetune.py:976] (1/7) Epoch 17, batch 1100, loss[loss=0.2011, simple_loss=0.2698, pruned_loss=0.06624, over 4819.00 frames. ], tot_loss[loss=0.1809, simple_loss=0.2514, pruned_loss=0.05524, over 950018.66 frames. ], batch size: 33, lr: 3.40e-03, grad_scale: 64.0 2023-04-27 09:32:36,186 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.169e+02 1.706e+02 1.976e+02 2.311e+02 4.775e+02, threshold=3.952e+02, percent-clipped=2.0 2023-04-27 09:32:39,401 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.2836, 1.2232, 1.3268, 1.5715, 1.6312, 1.2574, 0.9260, 1.5211], device='cuda:1'), covar=tensor([0.0884, 0.1332, 0.0865, 0.0607, 0.0672, 0.0827, 0.0883, 0.0604], device='cuda:1'), in_proj_covar=tensor([0.0192, 0.0203, 0.0184, 0.0174, 0.0178, 0.0183, 0.0154, 0.0182], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 09:33:02,815 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0709, 1.5445, 1.9523, 2.5854, 1.8862, 1.5517, 1.3650, 1.8482], device='cuda:1'), covar=tensor([0.3379, 0.3371, 0.1707, 0.2109, 0.2598, 0.2601, 0.4148, 0.2049], device='cuda:1'), in_proj_covar=tensor([0.0289, 0.0245, 0.0224, 0.0314, 0.0216, 0.0229, 0.0228, 0.0186], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-27 09:33:11,514 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=92781.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:33:32,087 INFO [finetune.py:976] (1/7) Epoch 17, batch 1150, loss[loss=0.1995, simple_loss=0.2687, pruned_loss=0.06522, over 4903.00 frames. ], tot_loss[loss=0.1816, simple_loss=0.2523, pruned_loss=0.0555, over 951870.37 frames. ], batch size: 36, lr: 3.40e-03, grad_scale: 32.0 2023-04-27 09:33:59,860 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.1451, 1.6469, 2.0514, 2.3597, 1.9456, 1.5885, 1.1989, 1.7268], device='cuda:1'), covar=tensor([0.3463, 0.3319, 0.1672, 0.2383, 0.2818, 0.2819, 0.4641, 0.2346], device='cuda:1'), in_proj_covar=tensor([0.0289, 0.0245, 0.0224, 0.0313, 0.0216, 0.0229, 0.0228, 0.0185], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-27 09:34:19,565 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=92831.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:34:39,529 INFO [finetune.py:976] (1/7) Epoch 17, batch 1200, loss[loss=0.1699, simple_loss=0.233, pruned_loss=0.0534, over 4929.00 frames. ], tot_loss[loss=0.1807, simple_loss=0.2512, pruned_loss=0.05516, over 949995.46 frames. ], batch size: 33, lr: 3.40e-03, grad_scale: 32.0 2023-04-27 09:34:50,260 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.173e+01 1.694e+02 1.853e+02 2.185e+02 5.052e+02, threshold=3.707e+02, percent-clipped=2.0 2023-04-27 09:35:14,491 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.47 vs. limit=2.0 2023-04-27 09:35:23,523 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.2257, 2.6467, 1.1224, 1.4208, 2.1529, 1.3187, 3.6216, 1.7758], device='cuda:1'), covar=tensor([0.0644, 0.0572, 0.0808, 0.1310, 0.0464, 0.1006, 0.0230, 0.0675], device='cuda:1'), in_proj_covar=tensor([0.0051, 0.0065, 0.0048, 0.0046, 0.0049, 0.0052, 0.0075, 0.0051], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0007, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:1') 2023-04-27 09:35:36,330 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8091, 1.2525, 1.8190, 2.2750, 1.8819, 1.7429, 1.8047, 1.7595], device='cuda:1'), covar=tensor([0.4644, 0.6565, 0.6635, 0.5707, 0.6074, 0.8303, 0.8559, 0.8702], device='cuda:1'), in_proj_covar=tensor([0.0418, 0.0405, 0.0494, 0.0504, 0.0448, 0.0471, 0.0477, 0.0482], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 09:35:44,791 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=92892.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:35:45,254 INFO [finetune.py:976] (1/7) Epoch 17, batch 1250, loss[loss=0.2132, simple_loss=0.2577, pruned_loss=0.08434, over 4728.00 frames. ], tot_loss[loss=0.1783, simple_loss=0.2478, pruned_loss=0.05442, over 950654.33 frames. ], batch size: 23, lr: 3.40e-03, grad_scale: 32.0 2023-04-27 09:35:48,977 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=92899.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:36:18,137 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8998, 1.0710, 1.5736, 1.7109, 1.6619, 1.7274, 1.5578, 1.5661], device='cuda:1'), covar=tensor([0.3724, 0.4778, 0.4109, 0.3899, 0.4989, 0.7020, 0.4598, 0.4243], device='cuda:1'), in_proj_covar=tensor([0.0332, 0.0373, 0.0318, 0.0334, 0.0344, 0.0397, 0.0355, 0.0325], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 09:36:25,912 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=92931.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:36:45,900 INFO [finetune.py:976] (1/7) Epoch 17, batch 1300, loss[loss=0.1451, simple_loss=0.219, pruned_loss=0.03556, over 4823.00 frames. ], tot_loss[loss=0.1759, simple_loss=0.2452, pruned_loss=0.05335, over 951792.34 frames. ], batch size: 25, lr: 3.40e-03, grad_scale: 32.0 2023-04-27 09:36:57,404 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.209e+02 1.510e+02 1.807e+02 2.085e+02 3.413e+02, threshold=3.613e+02, percent-clipped=0.0 2023-04-27 09:37:08,807 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=92960.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:37:23,986 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=92975.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:37:33,567 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=92981.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:37:36,042 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9215, 1.7157, 1.9366, 2.2769, 2.3027, 1.8476, 1.6302, 2.1073], device='cuda:1'), covar=tensor([0.0895, 0.1094, 0.0635, 0.0574, 0.0610, 0.0801, 0.0770, 0.0535], device='cuda:1'), in_proj_covar=tensor([0.0189, 0.0200, 0.0182, 0.0172, 0.0176, 0.0181, 0.0152, 0.0180], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 09:37:40,826 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=92992.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:37:41,318 INFO [finetune.py:976] (1/7) Epoch 17, batch 1350, loss[loss=0.1575, simple_loss=0.2338, pruned_loss=0.04064, over 4909.00 frames. ], tot_loss[loss=0.1761, simple_loss=0.2454, pruned_loss=0.05337, over 953238.38 frames. ], batch size: 36, lr: 3.40e-03, grad_scale: 32.0 2023-04-27 09:37:53,904 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=93010.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:38:16,146 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=93036.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:38:16,156 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=93036.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:38:26,345 INFO [finetune.py:976] (1/7) Epoch 17, batch 1400, loss[loss=0.1653, simple_loss=0.2449, pruned_loss=0.04287, over 4912.00 frames. ], tot_loss[loss=0.1779, simple_loss=0.2484, pruned_loss=0.05372, over 953411.70 frames. ], batch size: 37, lr: 3.40e-03, grad_scale: 32.0 2023-04-27 09:38:44,541 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.119e+02 1.602e+02 1.954e+02 2.219e+02 5.640e+02, threshold=3.909e+02, percent-clipped=3.0 2023-04-27 09:38:57,199 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.2574, 1.4871, 1.6726, 1.8315, 1.7228, 1.7887, 1.7766, 1.7542], device='cuda:1'), covar=tensor([0.4239, 0.6109, 0.5042, 0.4587, 0.5731, 0.7547, 0.5434, 0.5295], device='cuda:1'), in_proj_covar=tensor([0.0331, 0.0372, 0.0318, 0.0333, 0.0344, 0.0396, 0.0354, 0.0324], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 09:39:08,285 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.60 vs. limit=2.0 2023-04-27 09:39:08,819 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=93071.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:39:11,823 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=93076.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:39:28,194 INFO [finetune.py:976] (1/7) Epoch 17, batch 1450, loss[loss=0.147, simple_loss=0.2146, pruned_loss=0.03972, over 4735.00 frames. ], tot_loss[loss=0.1776, simple_loss=0.2486, pruned_loss=0.05329, over 953489.75 frames. ], batch size: 23, lr: 3.40e-03, grad_scale: 32.0 2023-04-27 09:39:37,281 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=93097.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 09:40:01,734 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=93115.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:40:24,592 INFO [finetune.py:976] (1/7) Epoch 17, batch 1500, loss[loss=0.1715, simple_loss=0.2464, pruned_loss=0.04833, over 4749.00 frames. ], tot_loss[loss=0.1797, simple_loss=0.2508, pruned_loss=0.0543, over 953658.28 frames. ], batch size: 27, lr: 3.40e-03, grad_scale: 32.0 2023-04-27 09:40:31,494 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.190e+02 1.667e+02 1.959e+02 2.406e+02 7.498e+02, threshold=3.919e+02, percent-clipped=4.0 2023-04-27 09:40:53,990 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=93176.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:41:06,560 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=93187.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:41:15,565 INFO [finetune.py:976] (1/7) Epoch 17, batch 1550, loss[loss=0.1807, simple_loss=0.2606, pruned_loss=0.05034, over 4792.00 frames. ], tot_loss[loss=0.1802, simple_loss=0.2514, pruned_loss=0.05449, over 954178.48 frames. ], batch size: 45, lr: 3.40e-03, grad_scale: 32.0 2023-04-27 09:41:49,255 INFO [finetune.py:976] (1/7) Epoch 17, batch 1600, loss[loss=0.2336, simple_loss=0.2953, pruned_loss=0.08599, over 4297.00 frames. ], tot_loss[loss=0.1785, simple_loss=0.2492, pruned_loss=0.05394, over 953430.91 frames. ], batch size: 65, lr: 3.40e-03, grad_scale: 32.0 2023-04-27 09:41:54,719 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.149e+02 1.590e+02 1.909e+02 2.207e+02 5.345e+02, threshold=3.818e+02, percent-clipped=1.0 2023-04-27 09:41:57,597 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=93255.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:42:15,815 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=93281.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:42:19,488 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=93287.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:42:22,037 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-04-27 09:42:23,068 INFO [finetune.py:976] (1/7) Epoch 17, batch 1650, loss[loss=0.184, simple_loss=0.2426, pruned_loss=0.06271, over 4818.00 frames. ], tot_loss[loss=0.1771, simple_loss=0.2472, pruned_loss=0.0535, over 955064.48 frames. ], batch size: 45, lr: 3.40e-03, grad_scale: 32.0 2023-04-27 09:43:03,609 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=93329.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:43:04,867 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=93331.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:43:17,621 INFO [finetune.py:976] (1/7) Epoch 17, batch 1700, loss[loss=0.1671, simple_loss=0.2465, pruned_loss=0.04383, over 4886.00 frames. ], tot_loss[loss=0.1757, simple_loss=0.2454, pruned_loss=0.05298, over 955247.75 frames. ], batch size: 32, lr: 3.40e-03, grad_scale: 32.0 2023-04-27 09:43:23,104 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.005e+02 1.559e+02 1.851e+02 2.220e+02 4.230e+02, threshold=3.703e+02, percent-clipped=1.0 2023-04-27 09:43:26,294 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.1208, 1.3944, 1.2972, 1.6512, 1.5694, 1.6314, 1.3006, 2.4852], device='cuda:1'), covar=tensor([0.0611, 0.0811, 0.0799, 0.1197, 0.0642, 0.0459, 0.0795, 0.0229], device='cuda:1'), in_proj_covar=tensor([0.0037, 0.0038, 0.0039, 0.0043, 0.0039, 0.0037, 0.0038, 0.0055], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0014, 0.0013, 0.0012, 0.0013, 0.0016], device='cuda:1') 2023-04-27 09:43:33,690 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=93366.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:43:40,784 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=93376.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:43:51,042 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=93392.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 09:43:51,575 INFO [finetune.py:976] (1/7) Epoch 17, batch 1750, loss[loss=0.2082, simple_loss=0.2761, pruned_loss=0.07016, over 4889.00 frames. ], tot_loss[loss=0.1775, simple_loss=0.2474, pruned_loss=0.05378, over 953760.45 frames. ], batch size: 32, lr: 3.40e-03, grad_scale: 32.0 2023-04-27 09:44:13,252 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=93424.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:44:17,318 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5192, 1.3835, 4.1584, 3.8747, 3.6945, 4.0044, 3.9627, 3.7077], device='cuda:1'), covar=tensor([0.6754, 0.5942, 0.1045, 0.1715, 0.1035, 0.1691, 0.1173, 0.1347], device='cuda:1'), in_proj_covar=tensor([0.0310, 0.0306, 0.0402, 0.0405, 0.0349, 0.0408, 0.0311, 0.0364], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-27 09:44:36,167 INFO [finetune.py:976] (1/7) Epoch 17, batch 1800, loss[loss=0.2117, simple_loss=0.2703, pruned_loss=0.0765, over 4901.00 frames. ], tot_loss[loss=0.1787, simple_loss=0.2491, pruned_loss=0.05421, over 951968.39 frames. ], batch size: 35, lr: 3.40e-03, grad_scale: 32.0 2023-04-27 09:44:38,746 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8498, 2.4534, 2.0024, 2.2305, 1.6943, 2.1119, 1.9426, 1.6243], device='cuda:1'), covar=tensor([0.2126, 0.1400, 0.0943, 0.1262, 0.3315, 0.1235, 0.2025, 0.2491], device='cuda:1'), in_proj_covar=tensor([0.0286, 0.0308, 0.0220, 0.0279, 0.0314, 0.0260, 0.0252, 0.0267], device='cuda:1'), out_proj_covar=tensor([1.1496e-04, 1.2242e-04, 8.7557e-05, 1.1061e-04, 1.2776e-04, 1.0339e-04, 1.0163e-04, 1.0616e-04], device='cuda:1') 2023-04-27 09:44:47,780 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.257e+02 1.710e+02 1.900e+02 2.209e+02 3.616e+02, threshold=3.799e+02, percent-clipped=0.0 2023-04-27 09:45:05,908 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=93471.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:45:06,726 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.76 vs. limit=2.0 2023-04-27 09:45:22,721 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=93487.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:45:26,246 INFO [finetune.py:976] (1/7) Epoch 17, batch 1850, loss[loss=0.2271, simple_loss=0.292, pruned_loss=0.08104, over 4864.00 frames. ], tot_loss[loss=0.1799, simple_loss=0.2503, pruned_loss=0.05475, over 953297.36 frames. ], batch size: 44, lr: 3.40e-03, grad_scale: 32.0 2023-04-27 09:45:46,881 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7276, 1.3504, 1.8674, 2.2516, 1.8571, 1.7301, 1.8002, 1.7685], device='cuda:1'), covar=tensor([0.4906, 0.6985, 0.6665, 0.5864, 0.6100, 0.8541, 0.8430, 0.9049], device='cuda:1'), in_proj_covar=tensor([0.0420, 0.0406, 0.0495, 0.0505, 0.0448, 0.0473, 0.0478, 0.0483], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 09:46:27,191 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=93535.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:46:37,236 INFO [finetune.py:976] (1/7) Epoch 17, batch 1900, loss[loss=0.1583, simple_loss=0.2473, pruned_loss=0.03465, over 4829.00 frames. ], tot_loss[loss=0.1811, simple_loss=0.252, pruned_loss=0.05511, over 954602.99 frames. ], batch size: 47, lr: 3.40e-03, grad_scale: 32.0 2023-04-27 09:46:42,805 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.049e+02 1.573e+02 1.862e+02 2.216e+02 4.322e+02, threshold=3.725e+02, percent-clipped=2.0 2023-04-27 09:46:44,715 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=93555.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:46:55,016 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=3.39 vs. limit=5.0 2023-04-27 09:47:03,673 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6569, 3.4452, 0.9665, 2.0413, 1.9959, 2.4385, 2.0147, 1.0559], device='cuda:1'), covar=tensor([0.1259, 0.0989, 0.1973, 0.1141, 0.0969, 0.1008, 0.1389, 0.1833], device='cuda:1'), in_proj_covar=tensor([0.0117, 0.0241, 0.0136, 0.0120, 0.0131, 0.0152, 0.0116, 0.0119], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-27 09:47:06,561 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=93587.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:47:10,125 INFO [finetune.py:976] (1/7) Epoch 17, batch 1950, loss[loss=0.1655, simple_loss=0.2443, pruned_loss=0.04335, over 4828.00 frames. ], tot_loss[loss=0.1786, simple_loss=0.2501, pruned_loss=0.05361, over 956394.49 frames. ], batch size: 33, lr: 3.40e-03, grad_scale: 32.0 2023-04-27 09:47:16,344 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=93603.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:47:35,237 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=93631.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:47:37,607 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=93635.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:47:48,923 INFO [finetune.py:976] (1/7) Epoch 17, batch 2000, loss[loss=0.1781, simple_loss=0.248, pruned_loss=0.05412, over 4901.00 frames. ], tot_loss[loss=0.1771, simple_loss=0.2478, pruned_loss=0.05319, over 955290.65 frames. ], batch size: 35, lr: 3.40e-03, grad_scale: 32.0 2023-04-27 09:47:54,465 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.097e+02 1.579e+02 1.885e+02 2.263e+02 4.038e+02, threshold=3.769e+02, percent-clipped=1.0 2023-04-27 09:48:03,152 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=93666.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:48:11,984 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=93679.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:48:21,373 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=93692.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:48:21,911 INFO [finetune.py:976] (1/7) Epoch 17, batch 2050, loss[loss=0.1707, simple_loss=0.2338, pruned_loss=0.05382, over 4913.00 frames. ], tot_loss[loss=0.1734, simple_loss=0.2437, pruned_loss=0.05158, over 956172.93 frames. ], batch size: 37, lr: 3.40e-03, grad_scale: 32.0 2023-04-27 09:48:35,564 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=93714.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:48:39,300 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6660, 2.7669, 2.2140, 2.4130, 2.7480, 2.4337, 3.7238, 2.1143], device='cuda:1'), covar=tensor([0.3502, 0.2186, 0.4182, 0.3594, 0.1724, 0.2634, 0.1364, 0.4100], device='cuda:1'), in_proj_covar=tensor([0.0339, 0.0343, 0.0426, 0.0354, 0.0380, 0.0379, 0.0367, 0.0420], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 09:48:53,241 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=93740.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:48:55,038 INFO [finetune.py:976] (1/7) Epoch 17, batch 2100, loss[loss=0.19, simple_loss=0.2566, pruned_loss=0.06168, over 4911.00 frames. ], tot_loss[loss=0.1723, simple_loss=0.2425, pruned_loss=0.05099, over 957792.72 frames. ], batch size: 37, lr: 3.40e-03, grad_scale: 32.0 2023-04-27 09:49:01,810 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.279e+02 1.585e+02 1.847e+02 2.242e+02 6.268e+02, threshold=3.694e+02, percent-clipped=2.0 2023-04-27 09:49:11,153 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7030, 2.0285, 1.0401, 1.4256, 1.9443, 1.5894, 1.4509, 1.5406], device='cuda:1'), covar=tensor([0.0543, 0.0356, 0.0338, 0.0557, 0.0276, 0.0515, 0.0532, 0.0601], device='cuda:1'), in_proj_covar=tensor([0.0028, 0.0024, 0.0022, 0.0029, 0.0020, 0.0028, 0.0028, 0.0028], device='cuda:1'), out_proj_covar=tensor([0.0050, 0.0044, 0.0038, 0.0051, 0.0038, 0.0049, 0.0049, 0.0050], device='cuda:1') 2023-04-27 09:49:12,961 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=93770.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:49:13,577 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=93771.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:49:15,424 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6929, 2.1225, 1.6777, 1.5865, 1.3578, 1.3029, 1.7029, 1.2766], device='cuda:1'), covar=tensor([0.1415, 0.1261, 0.1463, 0.1577, 0.2201, 0.1815, 0.0939, 0.1952], device='cuda:1'), in_proj_covar=tensor([0.0198, 0.0215, 0.0170, 0.0207, 0.0202, 0.0187, 0.0157, 0.0187], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-27 09:49:33,703 INFO [finetune.py:976] (1/7) Epoch 17, batch 2150, loss[loss=0.196, simple_loss=0.2538, pruned_loss=0.06912, over 3992.00 frames. ], tot_loss[loss=0.1768, simple_loss=0.2471, pruned_loss=0.05326, over 954996.23 frames. ], batch size: 17, lr: 3.39e-03, grad_scale: 32.0 2023-04-27 09:49:44,893 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([3.8044, 3.7309, 2.6693, 4.3750, 3.8337, 3.7718, 1.8385, 3.8091], device='cuda:1'), covar=tensor([0.1937, 0.1332, 0.3399, 0.1722, 0.3566, 0.1914, 0.5416, 0.2377], device='cuda:1'), in_proj_covar=tensor([0.0244, 0.0213, 0.0250, 0.0305, 0.0297, 0.0248, 0.0272, 0.0272], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 09:50:13,738 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=93819.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:50:26,753 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=93831.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:50:47,863 INFO [finetune.py:976] (1/7) Epoch 17, batch 2200, loss[loss=0.2443, simple_loss=0.2977, pruned_loss=0.0955, over 4895.00 frames. ], tot_loss[loss=0.179, simple_loss=0.2491, pruned_loss=0.05444, over 953160.07 frames. ], batch size: 35, lr: 3.39e-03, grad_scale: 32.0 2023-04-27 09:50:50,021 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-04-27 09:50:59,348 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.152e+02 1.629e+02 1.916e+02 2.338e+02 4.475e+02, threshold=3.833e+02, percent-clipped=3.0 2023-04-27 09:51:41,201 INFO [finetune.py:976] (1/7) Epoch 17, batch 2250, loss[loss=0.159, simple_loss=0.2383, pruned_loss=0.0398, over 4863.00 frames. ], tot_loss[loss=0.1804, simple_loss=0.251, pruned_loss=0.05488, over 952540.09 frames. ], batch size: 31, lr: 3.39e-03, grad_scale: 32.0 2023-04-27 09:51:48,434 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-04-27 09:51:57,134 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([4.6646, 4.5081, 2.9484, 5.3200, 4.6500, 4.5748, 1.9890, 4.5974], device='cuda:1'), covar=tensor([0.1422, 0.0969, 0.3070, 0.0916, 0.3320, 0.1536, 0.5494, 0.1967], device='cuda:1'), in_proj_covar=tensor([0.0243, 0.0212, 0.0249, 0.0303, 0.0296, 0.0247, 0.0270, 0.0271], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 09:52:14,472 INFO [finetune.py:976] (1/7) Epoch 17, batch 2300, loss[loss=0.1322, simple_loss=0.2092, pruned_loss=0.02754, over 4771.00 frames. ], tot_loss[loss=0.1792, simple_loss=0.2501, pruned_loss=0.05413, over 953090.82 frames. ], batch size: 25, lr: 3.39e-03, grad_scale: 32.0 2023-04-27 09:52:20,959 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.059e+02 1.592e+02 1.958e+02 2.327e+02 4.753e+02, threshold=3.916e+02, percent-clipped=2.0 2023-04-27 09:52:40,182 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=93981.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:52:47,881 INFO [finetune.py:976] (1/7) Epoch 17, batch 2350, loss[loss=0.1413, simple_loss=0.2161, pruned_loss=0.0332, over 4888.00 frames. ], tot_loss[loss=0.1759, simple_loss=0.2471, pruned_loss=0.05236, over 953987.19 frames. ], batch size: 35, lr: 3.39e-03, grad_scale: 32.0 2023-04-27 09:53:08,566 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6584, 1.9560, 1.6376, 1.3841, 1.2180, 1.2497, 1.5961, 1.1611], device='cuda:1'), covar=tensor([0.1725, 0.1441, 0.1480, 0.1902, 0.2505, 0.2086, 0.1081, 0.2211], device='cuda:1'), in_proj_covar=tensor([0.0197, 0.0214, 0.0169, 0.0206, 0.0201, 0.0185, 0.0156, 0.0186], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-27 09:53:48,598 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=94042.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:53:49,094 INFO [finetune.py:976] (1/7) Epoch 17, batch 2400, loss[loss=0.1865, simple_loss=0.2528, pruned_loss=0.06008, over 4856.00 frames. ], tot_loss[loss=0.1749, simple_loss=0.2454, pruned_loss=0.05224, over 954563.01 frames. ], batch size: 49, lr: 3.39e-03, grad_scale: 32.0 2023-04-27 09:53:56,111 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.042e+02 1.528e+02 1.774e+02 2.113e+02 4.800e+02, threshold=3.549e+02, percent-clipped=1.0 2023-04-27 09:54:23,068 INFO [finetune.py:976] (1/7) Epoch 17, batch 2450, loss[loss=0.1792, simple_loss=0.2466, pruned_loss=0.05589, over 4764.00 frames. ], tot_loss[loss=0.1744, simple_loss=0.2439, pruned_loss=0.05245, over 951414.03 frames. ], batch size: 59, lr: 3.39e-03, grad_scale: 32.0 2023-04-27 09:54:46,623 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=94126.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:54:57,072 INFO [finetune.py:976] (1/7) Epoch 17, batch 2500, loss[loss=0.1952, simple_loss=0.2643, pruned_loss=0.06305, over 4810.00 frames. ], tot_loss[loss=0.1756, simple_loss=0.2455, pruned_loss=0.05285, over 950252.20 frames. ], batch size: 38, lr: 3.39e-03, grad_scale: 32.0 2023-04-27 09:55:03,651 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.026e+02 1.665e+02 1.997e+02 2.427e+02 4.291e+02, threshold=3.995e+02, percent-clipped=3.0 2023-04-27 09:55:18,250 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([3.7215, 3.6147, 2.7032, 4.2706, 3.7191, 3.7076, 1.6478, 3.6384], device='cuda:1'), covar=tensor([0.1919, 0.1252, 0.3134, 0.1778, 0.2610, 0.1879, 0.6121, 0.2500], device='cuda:1'), in_proj_covar=tensor([0.0242, 0.0212, 0.0247, 0.0302, 0.0294, 0.0245, 0.0269, 0.0270], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 09:55:30,583 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.5787, 1.8507, 1.8619, 1.9514, 1.8350, 1.8961, 1.9989, 1.8673], device='cuda:1'), covar=tensor([0.3597, 0.5394, 0.4571, 0.4394, 0.5345, 0.7185, 0.5315, 0.5241], device='cuda:1'), in_proj_covar=tensor([0.0330, 0.0371, 0.0317, 0.0332, 0.0342, 0.0394, 0.0354, 0.0324], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 09:55:31,032 INFO [finetune.py:976] (1/7) Epoch 17, batch 2550, loss[loss=0.2374, simple_loss=0.2899, pruned_loss=0.09243, over 4830.00 frames. ], tot_loss[loss=0.1776, simple_loss=0.2481, pruned_loss=0.05357, over 950163.40 frames. ], batch size: 30, lr: 3.39e-03, grad_scale: 32.0 2023-04-27 09:55:52,333 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=94223.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 09:56:06,392 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.73 vs. limit=5.0 2023-04-27 09:56:09,849 INFO [finetune.py:976] (1/7) Epoch 17, batch 2600, loss[loss=0.1503, simple_loss=0.2233, pruned_loss=0.03867, over 4894.00 frames. ], tot_loss[loss=0.1781, simple_loss=0.2488, pruned_loss=0.05367, over 949993.98 frames. ], batch size: 35, lr: 3.39e-03, grad_scale: 32.0 2023-04-27 09:56:21,225 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.193e+02 1.596e+02 1.895e+02 2.398e+02 4.293e+02, threshold=3.790e+02, percent-clipped=2.0 2023-04-27 09:57:00,081 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=94284.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 09:57:05,478 INFO [finetune.py:976] (1/7) Epoch 17, batch 2650, loss[loss=0.1795, simple_loss=0.2551, pruned_loss=0.05196, over 4902.00 frames. ], tot_loss[loss=0.1801, simple_loss=0.2508, pruned_loss=0.05467, over 949718.22 frames. ], batch size: 32, lr: 3.39e-03, grad_scale: 32.0 2023-04-27 09:57:34,904 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=94337.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:57:38,463 INFO [finetune.py:976] (1/7) Epoch 17, batch 2700, loss[loss=0.1909, simple_loss=0.2501, pruned_loss=0.06587, over 4685.00 frames. ], tot_loss[loss=0.1782, simple_loss=0.249, pruned_loss=0.05365, over 950144.76 frames. ], batch size: 23, lr: 3.39e-03, grad_scale: 32.0 2023-04-27 09:57:41,630 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=94348.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:57:43,964 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.118e+02 1.591e+02 1.889e+02 2.175e+02 3.860e+02, threshold=3.779e+02, percent-clipped=1.0 2023-04-27 09:58:17,886 INFO [finetune.py:976] (1/7) Epoch 17, batch 2750, loss[loss=0.1795, simple_loss=0.2467, pruned_loss=0.05619, over 4926.00 frames. ], tot_loss[loss=0.1764, simple_loss=0.2472, pruned_loss=0.05286, over 951701.08 frames. ], batch size: 33, lr: 3.39e-03, grad_scale: 32.0 2023-04-27 09:58:38,905 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=94409.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 09:59:01,607 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=94426.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:59:24,237 INFO [finetune.py:976] (1/7) Epoch 17, batch 2800, loss[loss=0.197, simple_loss=0.2527, pruned_loss=0.07065, over 4828.00 frames. ], tot_loss[loss=0.1735, simple_loss=0.2436, pruned_loss=0.05171, over 952900.42 frames. ], batch size: 33, lr: 3.39e-03, grad_scale: 32.0 2023-04-27 09:59:35,021 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.123e+02 1.496e+02 1.786e+02 2.106e+02 4.249e+02, threshold=3.571e+02, percent-clipped=1.0 2023-04-27 09:59:54,410 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=94474.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:00:04,682 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.12 vs. limit=2.0 2023-04-27 10:00:06,867 INFO [finetune.py:976] (1/7) Epoch 17, batch 2850, loss[loss=0.2022, simple_loss=0.2789, pruned_loss=0.06278, over 4849.00 frames. ], tot_loss[loss=0.1743, simple_loss=0.2438, pruned_loss=0.05234, over 954677.65 frames. ], batch size: 47, lr: 3.39e-03, grad_scale: 32.0 2023-04-27 10:00:21,254 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-04-27 10:00:40,994 INFO [finetune.py:976] (1/7) Epoch 17, batch 2900, loss[loss=0.187, simple_loss=0.2743, pruned_loss=0.04984, over 4792.00 frames. ], tot_loss[loss=0.1763, simple_loss=0.2468, pruned_loss=0.0529, over 954705.88 frames. ], batch size: 51, lr: 3.39e-03, grad_scale: 32.0 2023-04-27 10:00:46,388 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.024e+02 1.635e+02 2.008e+02 2.460e+02 5.439e+02, threshold=4.016e+02, percent-clipped=3.0 2023-04-27 10:00:48,343 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([0.3964, 1.3357, 1.4285, 1.0907, 1.4027, 1.0940, 1.7488, 1.3755], device='cuda:1'), covar=tensor([0.3769, 0.1827, 0.4778, 0.2642, 0.1478, 0.2413, 0.1670, 0.4451], device='cuda:1'), in_proj_covar=tensor([0.0339, 0.0344, 0.0426, 0.0352, 0.0382, 0.0380, 0.0369, 0.0419], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 10:01:02,036 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6248, 2.0134, 1.7591, 1.9261, 1.4618, 1.7622, 1.6372, 1.2622], device='cuda:1'), covar=tensor([0.2037, 0.1429, 0.0854, 0.1236, 0.3684, 0.1197, 0.1914, 0.2737], device='cuda:1'), in_proj_covar=tensor([0.0287, 0.0310, 0.0221, 0.0281, 0.0315, 0.0260, 0.0251, 0.0268], device='cuda:1'), out_proj_covar=tensor([1.1499e-04, 1.2344e-04, 8.7934e-05, 1.1170e-04, 1.2801e-04, 1.0362e-04, 1.0151e-04, 1.0632e-04], device='cuda:1') 2023-04-27 10:01:03,844 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=94579.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 10:01:14,239 INFO [finetune.py:976] (1/7) Epoch 17, batch 2950, loss[loss=0.1941, simple_loss=0.2662, pruned_loss=0.06101, over 4815.00 frames. ], tot_loss[loss=0.1783, simple_loss=0.2493, pruned_loss=0.05366, over 954738.66 frames. ], batch size: 38, lr: 3.39e-03, grad_scale: 32.0 2023-04-27 10:01:25,872 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5939, 1.8237, 0.8594, 1.3174, 1.8770, 1.4811, 1.3575, 1.4624], device='cuda:1'), covar=tensor([0.0503, 0.0334, 0.0342, 0.0529, 0.0263, 0.0494, 0.0493, 0.0569], device='cuda:1'), in_proj_covar=tensor([0.0028, 0.0024, 0.0022, 0.0029, 0.0020, 0.0028, 0.0028, 0.0028], device='cuda:1'), out_proj_covar=tensor([0.0050, 0.0044, 0.0037, 0.0051, 0.0038, 0.0049, 0.0049, 0.0050], device='cuda:1') 2023-04-27 10:01:26,507 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6140, 1.6651, 1.4619, 1.8679, 1.7886, 2.0854, 1.5585, 4.1044], device='cuda:1'), covar=tensor([0.0488, 0.0772, 0.0788, 0.1163, 0.0604, 0.0542, 0.0748, 0.0117], device='cuda:1'), in_proj_covar=tensor([0.0037, 0.0038, 0.0039, 0.0043, 0.0040, 0.0038, 0.0038, 0.0056], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:1') 2023-04-27 10:01:54,131 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=94637.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:02:04,053 INFO [finetune.py:976] (1/7) Epoch 17, batch 3000, loss[loss=0.1625, simple_loss=0.2323, pruned_loss=0.04637, over 4831.00 frames. ], tot_loss[loss=0.181, simple_loss=0.2521, pruned_loss=0.05493, over 956415.66 frames. ], batch size: 30, lr: 3.39e-03, grad_scale: 32.0 2023-04-27 10:02:04,053 INFO [finetune.py:1001] (1/7) Computing validation loss 2023-04-27 10:02:12,314 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7934, 1.6010, 1.7581, 2.0988, 2.1763, 1.6785, 1.5381, 1.9421], device='cuda:1'), covar=tensor([0.0819, 0.1203, 0.0764, 0.0538, 0.0576, 0.0843, 0.0710, 0.0558], device='cuda:1'), in_proj_covar=tensor([0.0189, 0.0201, 0.0182, 0.0171, 0.0176, 0.0180, 0.0152, 0.0179], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 10:02:27,438 INFO [finetune.py:1010] (1/7) Epoch 17, validation: loss=0.1526, simple_loss=0.2233, pruned_loss=0.04089, over 2265189.00 frames. 2023-04-27 10:02:27,438 INFO [finetune.py:1011] (1/7) Maximum memory allocated so far is 6391MB 2023-04-27 10:02:39,392 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.157e+02 1.688e+02 2.099e+02 2.458e+02 3.567e+02, threshold=4.198e+02, percent-clipped=0.0 2023-04-27 10:02:50,318 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.93 vs. limit=5.0 2023-04-27 10:02:50,361 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.72 vs. limit=2.0 2023-04-27 10:03:10,439 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=94685.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:03:15,761 INFO [finetune.py:976] (1/7) Epoch 17, batch 3050, loss[loss=0.1673, simple_loss=0.2533, pruned_loss=0.04062, over 4811.00 frames. ], tot_loss[loss=0.1807, simple_loss=0.2521, pruned_loss=0.05466, over 956974.56 frames. ], batch size: 40, lr: 3.39e-03, grad_scale: 32.0 2023-04-27 10:03:24,034 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=94704.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 10:03:33,956 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([3.8384, 3.8334, 3.0161, 4.4215, 3.9236, 3.8464, 2.1498, 3.9092], device='cuda:1'), covar=tensor([0.1949, 0.1267, 0.2985, 0.1831, 0.3641, 0.2113, 0.5524, 0.2543], device='cuda:1'), in_proj_covar=tensor([0.0244, 0.0214, 0.0249, 0.0305, 0.0297, 0.0248, 0.0272, 0.0271], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 10:03:47,900 INFO [finetune.py:976] (1/7) Epoch 17, batch 3100, loss[loss=0.1506, simple_loss=0.2148, pruned_loss=0.04315, over 4915.00 frames. ], tot_loss[loss=0.1781, simple_loss=0.2489, pruned_loss=0.05361, over 956795.60 frames. ], batch size: 46, lr: 3.39e-03, grad_scale: 32.0 2023-04-27 10:03:53,746 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.19 vs. limit=5.0 2023-04-27 10:03:55,533 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-04-27 10:03:55,871 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.099e+02 1.556e+02 1.873e+02 2.227e+02 4.960e+02, threshold=3.745e+02, percent-clipped=1.0 2023-04-27 10:04:07,403 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-04-27 10:04:21,245 INFO [finetune.py:976] (1/7) Epoch 17, batch 3150, loss[loss=0.1755, simple_loss=0.2459, pruned_loss=0.05251, over 4799.00 frames. ], tot_loss[loss=0.1759, simple_loss=0.2461, pruned_loss=0.05287, over 954839.91 frames. ], batch size: 29, lr: 3.39e-03, grad_scale: 64.0 2023-04-27 10:05:23,148 INFO [finetune.py:976] (1/7) Epoch 17, batch 3200, loss[loss=0.1487, simple_loss=0.22, pruned_loss=0.03876, over 4937.00 frames. ], tot_loss[loss=0.1733, simple_loss=0.2431, pruned_loss=0.05179, over 954796.22 frames. ], batch size: 38, lr: 3.39e-03, grad_scale: 64.0 2023-04-27 10:05:34,599 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.148e+01 1.533e+02 1.770e+02 2.163e+02 3.313e+02, threshold=3.540e+02, percent-clipped=0.0 2023-04-27 10:05:53,664 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0576, 2.5418, 0.9206, 1.3691, 1.8763, 1.2100, 3.4479, 1.7558], device='cuda:1'), covar=tensor([0.0710, 0.0635, 0.0839, 0.1243, 0.0557, 0.1022, 0.0238, 0.0629], device='cuda:1'), in_proj_covar=tensor([0.0051, 0.0065, 0.0048, 0.0046, 0.0050, 0.0052, 0.0074, 0.0051], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0007, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:1') 2023-04-27 10:05:53,675 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=94879.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 10:06:02,202 INFO [finetune.py:976] (1/7) Epoch 17, batch 3250, loss[loss=0.1919, simple_loss=0.2662, pruned_loss=0.0588, over 4913.00 frames. ], tot_loss[loss=0.1752, simple_loss=0.2451, pruned_loss=0.05264, over 956776.09 frames. ], batch size: 43, lr: 3.39e-03, grad_scale: 64.0 2023-04-27 10:06:03,519 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8545, 2.2104, 0.8465, 1.1886, 1.4655, 1.1567, 2.4611, 1.3394], device='cuda:1'), covar=tensor([0.0692, 0.0580, 0.0658, 0.1221, 0.0516, 0.0983, 0.0334, 0.0662], device='cuda:1'), in_proj_covar=tensor([0.0051, 0.0065, 0.0048, 0.0046, 0.0050, 0.0052, 0.0074, 0.0051], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0007, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:1') 2023-04-27 10:06:20,536 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-04-27 10:06:26,490 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=94927.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 10:06:29,637 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.7714, 2.1394, 2.6674, 3.3077, 2.5452, 2.1373, 1.8516, 2.5724], device='cuda:1'), covar=tensor([0.3250, 0.3276, 0.1547, 0.2431, 0.2683, 0.2567, 0.3911, 0.2068], device='cuda:1'), in_proj_covar=tensor([0.0291, 0.0247, 0.0227, 0.0316, 0.0218, 0.0231, 0.0230, 0.0186], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-27 10:06:36,229 INFO [finetune.py:976] (1/7) Epoch 17, batch 3300, loss[loss=0.2013, simple_loss=0.2736, pruned_loss=0.06456, over 4811.00 frames. ], tot_loss[loss=0.1763, simple_loss=0.2466, pruned_loss=0.05297, over 954859.66 frames. ], batch size: 45, lr: 3.38e-03, grad_scale: 64.0 2023-04-27 10:06:47,629 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.257e+02 1.598e+02 1.896e+02 2.226e+02 3.987e+02, threshold=3.793e+02, percent-clipped=1.0 2023-04-27 10:07:30,061 INFO [finetune.py:976] (1/7) Epoch 17, batch 3350, loss[loss=0.1806, simple_loss=0.257, pruned_loss=0.05207, over 4899.00 frames. ], tot_loss[loss=0.1782, simple_loss=0.2487, pruned_loss=0.05388, over 954751.43 frames. ], batch size: 35, lr: 3.38e-03, grad_scale: 64.0 2023-04-27 10:07:36,972 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=95004.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:08:03,818 INFO [finetune.py:976] (1/7) Epoch 17, batch 3400, loss[loss=0.1515, simple_loss=0.2332, pruned_loss=0.03488, over 4781.00 frames. ], tot_loss[loss=0.1786, simple_loss=0.2492, pruned_loss=0.05396, over 953431.39 frames. ], batch size: 29, lr: 3.38e-03, grad_scale: 64.0 2023-04-27 10:08:08,185 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7891, 2.4044, 1.8685, 1.7966, 1.3366, 1.3573, 2.0219, 1.2908], device='cuda:1'), covar=tensor([0.1602, 0.1370, 0.1384, 0.1766, 0.2261, 0.1946, 0.0902, 0.1963], device='cuda:1'), in_proj_covar=tensor([0.0197, 0.0214, 0.0169, 0.0206, 0.0201, 0.0185, 0.0156, 0.0185], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-27 10:08:09,294 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.057e+02 1.633e+02 1.895e+02 2.474e+02 5.536e+02, threshold=3.790e+02, percent-clipped=1.0 2023-04-27 10:08:09,360 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=95052.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:08:21,285 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.67 vs. limit=2.0 2023-04-27 10:08:23,992 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.1973, 1.6909, 1.4785, 2.0839, 2.3332, 1.8061, 1.8367, 1.6337], device='cuda:1'), covar=tensor([0.1729, 0.1713, 0.1879, 0.1368, 0.1069, 0.1996, 0.2150, 0.2232], device='cuda:1'), in_proj_covar=tensor([0.0311, 0.0316, 0.0356, 0.0292, 0.0330, 0.0313, 0.0302, 0.0369], device='cuda:1'), out_proj_covar=tensor([6.4269e-05, 6.5963e-05, 7.5927e-05, 5.9235e-05, 6.8445e-05, 6.5940e-05, 6.3781e-05, 7.8775e-05], device='cuda:1') 2023-04-27 10:08:24,598 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.4896, 1.7879, 1.9027, 1.9974, 1.8521, 1.8960, 2.0347, 1.9758], device='cuda:1'), covar=tensor([0.3813, 0.5796, 0.4628, 0.4797, 0.5576, 0.7938, 0.5594, 0.4697], device='cuda:1'), in_proj_covar=tensor([0.0329, 0.0371, 0.0317, 0.0332, 0.0342, 0.0394, 0.0353, 0.0323], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 10:08:37,247 INFO [finetune.py:976] (1/7) Epoch 17, batch 3450, loss[loss=0.1883, simple_loss=0.2466, pruned_loss=0.06498, over 4736.00 frames. ], tot_loss[loss=0.1791, simple_loss=0.2501, pruned_loss=0.05407, over 953412.54 frames. ], batch size: 54, lr: 3.38e-03, grad_scale: 64.0 2023-04-27 10:08:44,630 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=95105.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 10:09:11,014 INFO [finetune.py:976] (1/7) Epoch 17, batch 3500, loss[loss=0.1859, simple_loss=0.2553, pruned_loss=0.05827, over 4897.00 frames. ], tot_loss[loss=0.177, simple_loss=0.2474, pruned_loss=0.05332, over 954709.17 frames. ], batch size: 32, lr: 3.38e-03, grad_scale: 64.0 2023-04-27 10:09:16,406 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.081e+02 1.623e+02 1.952e+02 2.265e+02 3.860e+02, threshold=3.904e+02, percent-clipped=1.0 2023-04-27 10:09:17,175 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.2409, 1.4529, 2.0536, 2.6053, 2.0946, 1.6385, 1.4481, 1.7917], device='cuda:1'), covar=tensor([0.3301, 0.3698, 0.1824, 0.2572, 0.2761, 0.2815, 0.4395, 0.2302], device='cuda:1'), in_proj_covar=tensor([0.0290, 0.0246, 0.0226, 0.0315, 0.0217, 0.0230, 0.0228, 0.0185], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-27 10:09:25,621 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=95166.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 10:09:44,906 INFO [finetune.py:976] (1/7) Epoch 17, batch 3550, loss[loss=0.1847, simple_loss=0.248, pruned_loss=0.06069, over 4754.00 frames. ], tot_loss[loss=0.1751, simple_loss=0.2449, pruned_loss=0.05267, over 955732.56 frames. ], batch size: 27, lr: 3.38e-03, grad_scale: 64.0 2023-04-27 10:10:42,813 INFO [finetune.py:976] (1/7) Epoch 17, batch 3600, loss[loss=0.2255, simple_loss=0.2722, pruned_loss=0.08936, over 4846.00 frames. ], tot_loss[loss=0.1732, simple_loss=0.2427, pruned_loss=0.05189, over 956952.36 frames. ], batch size: 49, lr: 3.38e-03, grad_scale: 64.0 2023-04-27 10:10:54,066 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.027e+02 1.577e+02 1.958e+02 2.437e+02 7.407e+02, threshold=3.916e+02, percent-clipped=3.0 2023-04-27 10:10:59,895 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=95260.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:11:10,460 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=95268.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:11:33,161 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.82 vs. limit=5.0 2023-04-27 10:11:34,140 INFO [finetune.py:976] (1/7) Epoch 17, batch 3650, loss[loss=0.1297, simple_loss=0.2065, pruned_loss=0.0264, over 4889.00 frames. ], tot_loss[loss=0.174, simple_loss=0.2441, pruned_loss=0.05195, over 956433.28 frames. ], batch size: 32, lr: 3.38e-03, grad_scale: 64.0 2023-04-27 10:11:39,238 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.6473, 1.9620, 1.9273, 2.1040, 1.8418, 1.9185, 2.0476, 1.9747], device='cuda:1'), covar=tensor([0.4033, 0.6354, 0.4680, 0.4568, 0.5926, 0.7767, 0.5961, 0.5841], device='cuda:1'), in_proj_covar=tensor([0.0329, 0.0370, 0.0316, 0.0331, 0.0342, 0.0393, 0.0353, 0.0323], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 10:11:51,438 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=95321.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:11:57,808 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=95329.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:12:05,318 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=95339.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:12:07,680 INFO [finetune.py:976] (1/7) Epoch 17, batch 3700, loss[loss=0.1765, simple_loss=0.2502, pruned_loss=0.05141, over 4744.00 frames. ], tot_loss[loss=0.1757, simple_loss=0.2469, pruned_loss=0.05221, over 955623.70 frames. ], batch size: 59, lr: 3.38e-03, grad_scale: 64.0 2023-04-27 10:12:13,172 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.977e+01 1.742e+02 1.979e+02 2.292e+02 4.077e+02, threshold=3.957e+02, percent-clipped=1.0 2023-04-27 10:12:41,442 INFO [finetune.py:976] (1/7) Epoch 17, batch 3750, loss[loss=0.1788, simple_loss=0.2609, pruned_loss=0.04829, over 4887.00 frames. ], tot_loss[loss=0.1772, simple_loss=0.2488, pruned_loss=0.05285, over 955146.38 frames. ], batch size: 43, lr: 3.38e-03, grad_scale: 32.0 2023-04-27 10:12:46,018 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=95400.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:13:06,436 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.2886, 1.7885, 2.0706, 2.5559, 2.4870, 2.0629, 1.7740, 2.3338], device='cuda:1'), covar=tensor([0.0782, 0.1173, 0.0666, 0.0571, 0.0622, 0.0867, 0.0749, 0.0554], device='cuda:1'), in_proj_covar=tensor([0.0188, 0.0199, 0.0180, 0.0169, 0.0175, 0.0179, 0.0151, 0.0178], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 10:13:14,397 INFO [finetune.py:976] (1/7) Epoch 17, batch 3800, loss[loss=0.1892, simple_loss=0.2656, pruned_loss=0.05644, over 4868.00 frames. ], tot_loss[loss=0.179, simple_loss=0.2506, pruned_loss=0.05375, over 954931.19 frames. ], batch size: 34, lr: 3.38e-03, grad_scale: 32.0 2023-04-27 10:13:20,913 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.076e+02 1.628e+02 1.968e+02 2.270e+02 5.105e+02, threshold=3.935e+02, percent-clipped=1.0 2023-04-27 10:13:22,943 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.68 vs. limit=5.0 2023-04-27 10:13:25,881 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=95461.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 10:13:33,856 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-04-27 10:13:48,140 INFO [finetune.py:976] (1/7) Epoch 17, batch 3850, loss[loss=0.1573, simple_loss=0.2374, pruned_loss=0.0386, over 4839.00 frames. ], tot_loss[loss=0.1789, simple_loss=0.25, pruned_loss=0.05384, over 955796.81 frames. ], batch size: 49, lr: 3.38e-03, grad_scale: 32.0 2023-04-27 10:14:20,427 INFO [finetune.py:976] (1/7) Epoch 17, batch 3900, loss[loss=0.1436, simple_loss=0.2236, pruned_loss=0.0318, over 4748.00 frames. ], tot_loss[loss=0.1771, simple_loss=0.2478, pruned_loss=0.05319, over 956866.30 frames. ], batch size: 27, lr: 3.38e-03, grad_scale: 32.0 2023-04-27 10:14:27,544 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.967e+01 1.562e+02 1.786e+02 2.211e+02 6.075e+02, threshold=3.573e+02, percent-clipped=1.0 2023-04-27 10:14:33,307 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-04-27 10:14:52,350 INFO [finetune.py:976] (1/7) Epoch 17, batch 3950, loss[loss=0.1607, simple_loss=0.2373, pruned_loss=0.0421, over 4830.00 frames. ], tot_loss[loss=0.1752, simple_loss=0.2457, pruned_loss=0.05236, over 956562.21 frames. ], batch size: 39, lr: 3.38e-03, grad_scale: 32.0 2023-04-27 10:15:04,298 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7661, 4.0972, 0.6853, 2.2478, 2.5471, 2.5589, 2.5442, 0.9677], device='cuda:1'), covar=tensor([0.1321, 0.0846, 0.2184, 0.1178, 0.0875, 0.1143, 0.1344, 0.2022], device='cuda:1'), in_proj_covar=tensor([0.0116, 0.0240, 0.0136, 0.0119, 0.0131, 0.0151, 0.0116, 0.0119], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-27 10:15:08,539 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=95616.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:15:09,191 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=95617.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:15:13,451 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=95624.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:15:31,266 INFO [finetune.py:976] (1/7) Epoch 17, batch 4000, loss[loss=0.1832, simple_loss=0.2505, pruned_loss=0.05797, over 4920.00 frames. ], tot_loss[loss=0.1741, simple_loss=0.2437, pruned_loss=0.05225, over 951430.08 frames. ], batch size: 38, lr: 3.38e-03, grad_scale: 32.0 2023-04-27 10:15:43,871 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.207e+01 1.538e+02 1.849e+02 2.258e+02 4.125e+02, threshold=3.697e+02, percent-clipped=1.0 2023-04-27 10:16:16,162 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=95678.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:16:27,321 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.2795, 1.3308, 1.4018, 1.5746, 1.6457, 1.3861, 0.9695, 1.5231], device='cuda:1'), covar=tensor([0.0910, 0.1287, 0.0919, 0.0664, 0.0722, 0.0798, 0.0922, 0.0644], device='cuda:1'), in_proj_covar=tensor([0.0189, 0.0201, 0.0182, 0.0171, 0.0176, 0.0179, 0.0152, 0.0179], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 10:16:36,028 INFO [finetune.py:976] (1/7) Epoch 17, batch 4050, loss[loss=0.2213, simple_loss=0.2929, pruned_loss=0.07482, over 4861.00 frames. ], tot_loss[loss=0.1768, simple_loss=0.247, pruned_loss=0.0533, over 952327.14 frames. ], batch size: 44, lr: 3.38e-03, grad_scale: 32.0 2023-04-27 10:16:37,796 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=95695.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:16:59,112 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.3724, 3.2904, 0.8678, 1.8691, 1.9249, 2.3847, 1.9852, 1.0382], device='cuda:1'), covar=tensor([0.1440, 0.1093, 0.2002, 0.1203, 0.1037, 0.1000, 0.1401, 0.1844], device='cuda:1'), in_proj_covar=tensor([0.0116, 0.0242, 0.0137, 0.0119, 0.0131, 0.0152, 0.0116, 0.0119], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-27 10:16:59,728 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.2443, 3.3220, 1.0732, 1.7827, 1.7890, 2.4638, 1.9601, 1.0600], device='cuda:1'), covar=tensor([0.1686, 0.1265, 0.1977, 0.1484, 0.1215, 0.1124, 0.1558, 0.2020], device='cuda:1'), in_proj_covar=tensor([0.0116, 0.0242, 0.0137, 0.0119, 0.0131, 0.0152, 0.0116, 0.0119], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-27 10:17:36,636 INFO [finetune.py:976] (1/7) Epoch 17, batch 4100, loss[loss=0.1527, simple_loss=0.2196, pruned_loss=0.04288, over 3997.00 frames. ], tot_loss[loss=0.1789, simple_loss=0.2499, pruned_loss=0.05399, over 951752.72 frames. ], batch size: 17, lr: 3.38e-03, grad_scale: 32.0 2023-04-27 10:17:43,702 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.341e+02 1.651e+02 1.954e+02 2.364e+02 4.876e+02, threshold=3.909e+02, percent-clipped=1.0 2023-04-27 10:17:50,102 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=95761.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 10:18:09,896 INFO [finetune.py:976] (1/7) Epoch 17, batch 4150, loss[loss=0.1326, simple_loss=0.205, pruned_loss=0.03008, over 4723.00 frames. ], tot_loss[loss=0.1814, simple_loss=0.2518, pruned_loss=0.05546, over 952697.39 frames. ], batch size: 23, lr: 3.38e-03, grad_scale: 32.0 2023-04-27 10:18:21,840 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=95809.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 10:18:21,852 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4090, 1.1347, 4.0406, 3.4974, 3.5707, 3.7632, 3.7517, 3.3786], device='cuda:1'), covar=tensor([0.7874, 0.7758, 0.1674, 0.3076, 0.2179, 0.2637, 0.2905, 0.2856], device='cuda:1'), in_proj_covar=tensor([0.0307, 0.0305, 0.0401, 0.0404, 0.0346, 0.0404, 0.0308, 0.0364], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-27 10:18:24,194 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=95812.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 10:18:43,505 INFO [finetune.py:976] (1/7) Epoch 17, batch 4200, loss[loss=0.1873, simple_loss=0.2615, pruned_loss=0.05657, over 4816.00 frames. ], tot_loss[loss=0.1819, simple_loss=0.2527, pruned_loss=0.05556, over 954111.19 frames. ], batch size: 39, lr: 3.38e-03, grad_scale: 32.0 2023-04-27 10:18:46,042 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=95847.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:18:46,668 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9473, 2.3533, 0.9751, 1.2821, 1.6914, 1.1955, 2.9477, 1.5876], device='cuda:1'), covar=tensor([0.0668, 0.0580, 0.0732, 0.1181, 0.0510, 0.0960, 0.0264, 0.0612], device='cuda:1'), in_proj_covar=tensor([0.0051, 0.0065, 0.0048, 0.0046, 0.0050, 0.0052, 0.0074, 0.0051], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:1') 2023-04-27 10:18:49,672 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 8.908e+01 1.602e+02 1.963e+02 2.412e+02 3.992e+02, threshold=3.927e+02, percent-clipped=2.0 2023-04-27 10:19:04,201 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=95873.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 10:19:05,412 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.4332, 3.3595, 2.6811, 3.0041, 2.3179, 2.7790, 2.9562, 2.1725], device='cuda:1'), covar=tensor([0.2220, 0.1030, 0.0771, 0.1118, 0.2873, 0.1066, 0.1753, 0.2637], device='cuda:1'), in_proj_covar=tensor([0.0283, 0.0305, 0.0218, 0.0278, 0.0311, 0.0258, 0.0248, 0.0264], device='cuda:1'), out_proj_covar=tensor([1.1379e-04, 1.2119e-04, 8.6655e-05, 1.1027e-04, 1.2628e-04, 1.0250e-04, 1.0034e-04, 1.0490e-04], device='cuda:1') 2023-04-27 10:19:16,336 INFO [finetune.py:976] (1/7) Epoch 17, batch 4250, loss[loss=0.1693, simple_loss=0.2342, pruned_loss=0.05213, over 4728.00 frames. ], tot_loss[loss=0.1797, simple_loss=0.2501, pruned_loss=0.05464, over 953505.54 frames. ], batch size: 54, lr: 3.38e-03, grad_scale: 32.0 2023-04-27 10:19:19,441 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.2342, 1.2433, 3.7625, 3.4843, 3.3493, 3.5730, 3.5312, 3.3419], device='cuda:1'), covar=tensor([0.7045, 0.5740, 0.1289, 0.2014, 0.1311, 0.2045, 0.2043, 0.1566], device='cuda:1'), in_proj_covar=tensor([0.0309, 0.0306, 0.0401, 0.0405, 0.0347, 0.0405, 0.0309, 0.0364], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-27 10:19:26,043 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=95908.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 10:19:31,948 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=95916.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:19:36,871 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=95924.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:19:47,590 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9716, 2.6719, 2.1614, 2.4560, 1.7931, 2.1687, 2.2392, 1.6239], device='cuda:1'), covar=tensor([0.2078, 0.1252, 0.0878, 0.1001, 0.3258, 0.1071, 0.1690, 0.2550], device='cuda:1'), in_proj_covar=tensor([0.0282, 0.0304, 0.0217, 0.0277, 0.0310, 0.0256, 0.0247, 0.0263], device='cuda:1'), out_proj_covar=tensor([1.1330e-04, 1.2073e-04, 8.6219e-05, 1.0981e-04, 1.2590e-04, 1.0199e-04, 9.9898e-05, 1.0465e-04], device='cuda:1') 2023-04-27 10:19:48,685 INFO [finetune.py:976] (1/7) Epoch 17, batch 4300, loss[loss=0.1435, simple_loss=0.2126, pruned_loss=0.0372, over 4722.00 frames. ], tot_loss[loss=0.177, simple_loss=0.2472, pruned_loss=0.05338, over 953320.40 frames. ], batch size: 54, lr: 3.38e-03, grad_scale: 32.0 2023-04-27 10:19:54,839 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.107e+02 1.623e+02 1.872e+02 2.208e+02 5.069e+02, threshold=3.743e+02, percent-clipped=1.0 2023-04-27 10:20:03,624 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=95964.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:20:08,325 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 2023-04-27 10:20:09,300 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=95972.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:20:09,895 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=95973.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:20:14,620 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=95980.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:20:22,488 INFO [finetune.py:976] (1/7) Epoch 17, batch 4350, loss[loss=0.1482, simple_loss=0.216, pruned_loss=0.04022, over 4825.00 frames. ], tot_loss[loss=0.1735, simple_loss=0.2434, pruned_loss=0.05181, over 954766.64 frames. ], batch size: 30, lr: 3.38e-03, grad_scale: 32.0 2023-04-27 10:20:23,767 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=95995.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:20:29,481 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.22 vs. limit=5.0 2023-04-27 10:21:00,670 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.0652, 2.0079, 1.7359, 1.7593, 2.1603, 1.7827, 2.6107, 1.5714], device='cuda:1'), covar=tensor([0.3551, 0.1694, 0.4582, 0.2541, 0.1591, 0.2268, 0.1156, 0.4412], device='cuda:1'), in_proj_covar=tensor([0.0334, 0.0342, 0.0421, 0.0351, 0.0375, 0.0376, 0.0365, 0.0414], device='cuda:1'), out_proj_covar=tensor([9.9748e-05, 1.0300e-04, 1.2820e-04, 1.0623e-04, 1.1218e-04, 1.1281e-04, 1.0753e-04, 1.2556e-04], device='cuda:1') 2023-04-27 10:21:07,895 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=96041.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:21:08,983 INFO [finetune.py:976] (1/7) Epoch 17, batch 4400, loss[loss=0.1672, simple_loss=0.2421, pruned_loss=0.04615, over 4901.00 frames. ], tot_loss[loss=0.1744, simple_loss=0.2446, pruned_loss=0.05211, over 956068.84 frames. ], batch size: 32, lr: 3.38e-03, grad_scale: 32.0 2023-04-27 10:21:09,043 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=96043.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:21:15,039 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.713e+01 1.567e+02 1.808e+02 2.181e+02 3.991e+02, threshold=3.617e+02, percent-clipped=1.0 2023-04-27 10:21:34,156 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.2807, 1.1796, 1.5645, 1.4639, 1.1960, 1.0837, 1.2631, 0.6378], device='cuda:1'), covar=tensor([0.0569, 0.0655, 0.0383, 0.0588, 0.0797, 0.1145, 0.0524, 0.0680], device='cuda:1'), in_proj_covar=tensor([0.0069, 0.0070, 0.0068, 0.0068, 0.0075, 0.0096, 0.0075, 0.0067], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-27 10:21:34,770 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=96080.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:21:48,839 INFO [finetune.py:976] (1/7) Epoch 17, batch 4450, loss[loss=0.1404, simple_loss=0.2193, pruned_loss=0.03074, over 4752.00 frames. ], tot_loss[loss=0.1774, simple_loss=0.2482, pruned_loss=0.05332, over 954566.20 frames. ], batch size: 28, lr: 3.38e-03, grad_scale: 32.0 2023-04-27 10:22:53,809 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=96141.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:22:54,920 INFO [finetune.py:976] (1/7) Epoch 17, batch 4500, loss[loss=0.1719, simple_loss=0.2552, pruned_loss=0.04433, over 4817.00 frames. ], tot_loss[loss=0.1795, simple_loss=0.2504, pruned_loss=0.05431, over 952631.51 frames. ], batch size: 38, lr: 3.37e-03, grad_scale: 32.0 2023-04-27 10:23:02,750 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.142e+02 1.688e+02 1.990e+02 2.416e+02 4.917e+02, threshold=3.981e+02, percent-clipped=4.0 2023-04-27 10:23:16,193 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=96165.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:23:17,979 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=96168.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 10:23:25,377 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-04-27 10:23:58,142 INFO [finetune.py:976] (1/7) Epoch 17, batch 4550, loss[loss=0.2054, simple_loss=0.2657, pruned_loss=0.07254, over 4853.00 frames. ], tot_loss[loss=0.1794, simple_loss=0.2503, pruned_loss=0.05425, over 954292.72 frames. ], batch size: 31, lr: 3.37e-03, grad_scale: 32.0 2023-04-27 10:24:10,612 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=96203.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 10:24:42,303 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=96226.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:24:54,976 INFO [finetune.py:976] (1/7) Epoch 17, batch 4600, loss[loss=0.1522, simple_loss=0.2297, pruned_loss=0.03732, over 4747.00 frames. ], tot_loss[loss=0.1783, simple_loss=0.2494, pruned_loss=0.0536, over 955269.64 frames. ], batch size: 54, lr: 3.37e-03, grad_scale: 32.0 2023-04-27 10:24:55,821 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.14 vs. limit=2.0 2023-04-27 10:25:01,050 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.153e+02 1.538e+02 1.819e+02 2.307e+02 5.540e+02, threshold=3.639e+02, percent-clipped=2.0 2023-04-27 10:25:13,335 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=96273.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:25:27,832 INFO [finetune.py:976] (1/7) Epoch 17, batch 4650, loss[loss=0.1654, simple_loss=0.2363, pruned_loss=0.04722, over 4816.00 frames. ], tot_loss[loss=0.1778, simple_loss=0.2481, pruned_loss=0.05373, over 956081.50 frames. ], batch size: 38, lr: 3.37e-03, grad_scale: 32.0 2023-04-27 10:25:44,880 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-04-27 10:25:45,389 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=96321.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:25:47,231 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=96324.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:25:56,382 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=96336.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:26:01,179 INFO [finetune.py:976] (1/7) Epoch 17, batch 4700, loss[loss=0.183, simple_loss=0.2359, pruned_loss=0.06506, over 4256.00 frames. ], tot_loss[loss=0.1759, simple_loss=0.2455, pruned_loss=0.05318, over 955352.78 frames. ], batch size: 66, lr: 3.37e-03, grad_scale: 32.0 2023-04-27 10:26:07,154 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.22 vs. limit=5.0 2023-04-27 10:26:07,579 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.173e+02 1.665e+02 1.923e+02 2.344e+02 4.487e+02, threshold=3.847e+02, percent-clipped=4.0 2023-04-27 10:26:38,842 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=96385.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:26:48,039 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=96390.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:26:55,843 INFO [finetune.py:976] (1/7) Epoch 17, batch 4750, loss[loss=0.1972, simple_loss=0.2601, pruned_loss=0.06709, over 4823.00 frames. ], tot_loss[loss=0.1752, simple_loss=0.2442, pruned_loss=0.05313, over 957209.63 frames. ], batch size: 30, lr: 3.37e-03, grad_scale: 32.0 2023-04-27 10:27:08,574 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2023-04-27 10:27:39,946 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.2810, 1.5642, 1.4982, 1.8587, 1.8040, 1.9286, 1.4781, 3.5621], device='cuda:1'), covar=tensor([0.0572, 0.0763, 0.0753, 0.1158, 0.0550, 0.0472, 0.0737, 0.0139], device='cuda:1'), in_proj_covar=tensor([0.0037, 0.0038, 0.0039, 0.0043, 0.0039, 0.0038, 0.0038, 0.0056], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:1') 2023-04-27 10:27:43,628 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.3548, 1.6763, 1.7587, 1.8605, 1.6743, 1.7644, 1.7878, 1.7461], device='cuda:1'), covar=tensor([0.4114, 0.5779, 0.5006, 0.4801, 0.5970, 0.7786, 0.6070, 0.5688], device='cuda:1'), in_proj_covar=tensor([0.0331, 0.0371, 0.0316, 0.0334, 0.0344, 0.0394, 0.0355, 0.0324], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 10:27:44,798 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=96436.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:27:50,990 INFO [finetune.py:976] (1/7) Epoch 17, batch 4800, loss[loss=0.2121, simple_loss=0.2758, pruned_loss=0.07422, over 4901.00 frames. ], tot_loss[loss=0.1777, simple_loss=0.2473, pruned_loss=0.05402, over 958351.73 frames. ], batch size: 32, lr: 3.37e-03, grad_scale: 32.0 2023-04-27 10:27:56,400 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=96451.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:27:57,983 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.109e+02 1.553e+02 2.009e+02 2.369e+02 4.346e+02, threshold=4.018e+02, percent-clipped=1.0 2023-04-27 10:28:07,160 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=96468.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 10:28:34,487 INFO [finetune.py:976] (1/7) Epoch 17, batch 4850, loss[loss=0.1834, simple_loss=0.247, pruned_loss=0.05996, over 4836.00 frames. ], tot_loss[loss=0.1805, simple_loss=0.2506, pruned_loss=0.05521, over 957805.51 frames. ], batch size: 30, lr: 3.37e-03, grad_scale: 32.0 2023-04-27 10:28:41,607 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=96503.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:28:49,910 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=96516.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 10:28:52,927 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=96521.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:29:07,105 INFO [finetune.py:976] (1/7) Epoch 17, batch 4900, loss[loss=0.2149, simple_loss=0.2751, pruned_loss=0.07733, over 4885.00 frames. ], tot_loss[loss=0.1809, simple_loss=0.2511, pruned_loss=0.05529, over 956629.23 frames. ], batch size: 35, lr: 3.37e-03, grad_scale: 32.0 2023-04-27 10:29:13,476 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=96551.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:29:15,085 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.788e+01 1.547e+02 1.882e+02 2.260e+02 4.956e+02, threshold=3.763e+02, percent-clipped=3.0 2023-04-27 10:29:29,080 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6885, 1.1355, 1.7001, 2.1568, 1.7757, 1.6385, 1.6937, 1.6564], device='cuda:1'), covar=tensor([0.4363, 0.6605, 0.5906, 0.5677, 0.5259, 0.7217, 0.7500, 0.8621], device='cuda:1'), in_proj_covar=tensor([0.0421, 0.0408, 0.0497, 0.0505, 0.0450, 0.0476, 0.0482, 0.0486], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 10:29:39,873 INFO [finetune.py:976] (1/7) Epoch 17, batch 4950, loss[loss=0.1756, simple_loss=0.2404, pruned_loss=0.05542, over 4758.00 frames. ], tot_loss[loss=0.1817, simple_loss=0.252, pruned_loss=0.05569, over 954077.92 frames. ], batch size: 26, lr: 3.37e-03, grad_scale: 32.0 2023-04-27 10:30:09,368 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=96636.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:30:13,518 INFO [finetune.py:976] (1/7) Epoch 17, batch 5000, loss[loss=0.1453, simple_loss=0.2168, pruned_loss=0.03687, over 4866.00 frames. ], tot_loss[loss=0.1802, simple_loss=0.2507, pruned_loss=0.05483, over 952390.84 frames. ], batch size: 34, lr: 3.37e-03, grad_scale: 32.0 2023-04-27 10:30:21,474 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 8.257e+01 1.582e+02 1.781e+02 2.065e+02 3.108e+02, threshold=3.561e+02, percent-clipped=0.0 2023-04-27 10:30:38,902 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=96680.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:30:41,301 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=96684.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:30:46,661 INFO [finetune.py:976] (1/7) Epoch 17, batch 5050, loss[loss=0.1885, simple_loss=0.2532, pruned_loss=0.06188, over 4916.00 frames. ], tot_loss[loss=0.1787, simple_loss=0.2484, pruned_loss=0.05451, over 955050.98 frames. ], batch size: 36, lr: 3.37e-03, grad_scale: 32.0 2023-04-27 10:30:50,469 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-04-27 10:30:50,901 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6565, 1.3472, 4.3471, 4.0248, 3.7573, 4.1155, 4.0124, 3.7735], device='cuda:1'), covar=tensor([0.6476, 0.5903, 0.1025, 0.1727, 0.1215, 0.1551, 0.1521, 0.1497], device='cuda:1'), in_proj_covar=tensor([0.0309, 0.0306, 0.0402, 0.0405, 0.0349, 0.0405, 0.0312, 0.0365], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-27 10:31:15,672 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=96736.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:31:19,861 INFO [finetune.py:976] (1/7) Epoch 17, batch 5100, loss[loss=0.1301, simple_loss=0.2069, pruned_loss=0.02668, over 4755.00 frames. ], tot_loss[loss=0.1764, simple_loss=0.2457, pruned_loss=0.05358, over 955566.79 frames. ], batch size: 27, lr: 3.37e-03, grad_scale: 32.0 2023-04-27 10:31:21,744 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=96746.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:31:25,905 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.147e+02 1.614e+02 1.893e+02 2.231e+02 4.752e+02, threshold=3.786e+02, percent-clipped=4.0 2023-04-27 10:31:54,175 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=96776.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:32:04,480 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=96784.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:32:15,344 INFO [finetune.py:976] (1/7) Epoch 17, batch 5150, loss[loss=0.1808, simple_loss=0.2573, pruned_loss=0.05218, over 4820.00 frames. ], tot_loss[loss=0.1781, simple_loss=0.2468, pruned_loss=0.05474, over 956145.36 frames. ], batch size: 30, lr: 3.37e-03, grad_scale: 32.0 2023-04-27 10:32:46,397 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=96821.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:33:03,945 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=96837.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:33:12,850 INFO [finetune.py:976] (1/7) Epoch 17, batch 5200, loss[loss=0.1875, simple_loss=0.2592, pruned_loss=0.05789, over 4750.00 frames. ], tot_loss[loss=0.1811, simple_loss=0.25, pruned_loss=0.05606, over 955011.38 frames. ], batch size: 59, lr: 3.37e-03, grad_scale: 32.0 2023-04-27 10:33:24,247 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6715, 3.6538, 0.9837, 1.9022, 2.1083, 2.4558, 1.9188, 1.1077], device='cuda:1'), covar=tensor([0.1392, 0.0793, 0.2018, 0.1195, 0.1001, 0.1079, 0.1627, 0.2022], device='cuda:1'), in_proj_covar=tensor([0.0117, 0.0243, 0.0138, 0.0121, 0.0133, 0.0154, 0.0119, 0.0121], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-27 10:33:24,748 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.802e+01 1.740e+02 2.121e+02 2.596e+02 7.000e+02, threshold=4.242e+02, percent-clipped=4.0 2023-04-27 10:33:47,093 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=96869.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:34:02,258 INFO [finetune.py:976] (1/7) Epoch 17, batch 5250, loss[loss=0.1766, simple_loss=0.2502, pruned_loss=0.05153, over 4910.00 frames. ], tot_loss[loss=0.183, simple_loss=0.2526, pruned_loss=0.05673, over 955118.97 frames. ], batch size: 37, lr: 3.37e-03, grad_scale: 32.0 2023-04-27 10:34:35,558 INFO [finetune.py:976] (1/7) Epoch 17, batch 5300, loss[loss=0.1697, simple_loss=0.2459, pruned_loss=0.04681, over 4904.00 frames. ], tot_loss[loss=0.1836, simple_loss=0.2539, pruned_loss=0.05664, over 955787.83 frames. ], batch size: 36, lr: 3.37e-03, grad_scale: 32.0 2023-04-27 10:34:37,581 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.57 vs. limit=2.0 2023-04-27 10:34:41,675 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.055e+02 1.524e+02 1.845e+02 2.298e+02 4.151e+02, threshold=3.690e+02, percent-clipped=0.0 2023-04-27 10:35:00,614 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=96980.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:35:08,783 INFO [finetune.py:976] (1/7) Epoch 17, batch 5350, loss[loss=0.1475, simple_loss=0.2211, pruned_loss=0.03693, over 4796.00 frames. ], tot_loss[loss=0.1822, simple_loss=0.2527, pruned_loss=0.05584, over 954366.59 frames. ], batch size: 25, lr: 3.37e-03, grad_scale: 32.0 2023-04-27 10:35:13,339 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=97000.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:35:22,968 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-04-27 10:35:33,058 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=97028.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:35:42,645 INFO [finetune.py:976] (1/7) Epoch 17, batch 5400, loss[loss=0.1327, simple_loss=0.2116, pruned_loss=0.0269, over 4744.00 frames. ], tot_loss[loss=0.1789, simple_loss=0.2486, pruned_loss=0.0546, over 953214.08 frames. ], batch size: 23, lr: 3.37e-03, grad_scale: 32.0 2023-04-27 10:35:44,572 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=97046.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:35:48,747 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.116e+02 1.600e+02 1.805e+02 2.105e+02 4.244e+02, threshold=3.609e+02, percent-clipped=1.0 2023-04-27 10:35:53,814 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=97061.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:35:56,223 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.5555, 1.8932, 2.3868, 3.0223, 2.4239, 1.8743, 1.7691, 2.3445], device='cuda:1'), covar=tensor([0.3678, 0.3825, 0.1846, 0.2496, 0.3040, 0.2808, 0.4260, 0.2153], device='cuda:1'), in_proj_covar=tensor([0.0293, 0.0248, 0.0228, 0.0317, 0.0219, 0.0232, 0.0230, 0.0186], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-27 10:36:15,423 INFO [finetune.py:976] (1/7) Epoch 17, batch 5450, loss[loss=0.1386, simple_loss=0.2036, pruned_loss=0.03682, over 4773.00 frames. ], tot_loss[loss=0.1766, simple_loss=0.2459, pruned_loss=0.05366, over 956257.94 frames. ], batch size: 27, lr: 3.37e-03, grad_scale: 32.0 2023-04-27 10:36:16,112 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=97094.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:36:18,604 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4664, 1.7387, 1.6004, 1.9753, 1.8961, 2.1196, 1.5615, 3.8871], device='cuda:1'), covar=tensor([0.0555, 0.0774, 0.0787, 0.1120, 0.0605, 0.0439, 0.0728, 0.0130], device='cuda:1'), in_proj_covar=tensor([0.0037, 0.0038, 0.0039, 0.0043, 0.0039, 0.0038, 0.0038, 0.0056], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0014, 0.0013, 0.0012, 0.0013, 0.0016], device='cuda:1') 2023-04-27 10:36:21,645 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0054, 2.6518, 2.0902, 2.1443, 1.6446, 1.6238, 2.1043, 1.5313], device='cuda:1'), covar=tensor([0.1362, 0.1230, 0.1272, 0.1447, 0.1975, 0.1637, 0.0828, 0.1767], device='cuda:1'), in_proj_covar=tensor([0.0198, 0.0214, 0.0169, 0.0206, 0.0202, 0.0185, 0.0156, 0.0186], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-27 10:36:40,610 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=97132.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:36:47,765 INFO [finetune.py:976] (1/7) Epoch 17, batch 5500, loss[loss=0.1654, simple_loss=0.2421, pruned_loss=0.04438, over 4888.00 frames. ], tot_loss[loss=0.1741, simple_loss=0.2433, pruned_loss=0.05244, over 957309.00 frames. ], batch size: 43, lr: 3.37e-03, grad_scale: 32.0 2023-04-27 10:36:54,340 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.109e+02 1.627e+02 1.973e+02 2.284e+02 4.082e+02, threshold=3.945e+02, percent-clipped=1.0 2023-04-27 10:37:37,876 INFO [finetune.py:976] (1/7) Epoch 17, batch 5550, loss[loss=0.2657, simple_loss=0.3128, pruned_loss=0.1093, over 4054.00 frames. ], tot_loss[loss=0.1774, simple_loss=0.2465, pruned_loss=0.05411, over 954817.67 frames. ], batch size: 65, lr: 3.37e-03, grad_scale: 32.0 2023-04-27 10:38:05,588 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([0.3127, 1.3122, 1.4151, 1.0684, 1.3733, 1.1897, 1.7441, 1.4250], device='cuda:1'), covar=tensor([0.3576, 0.1857, 0.4974, 0.2483, 0.1377, 0.2095, 0.1369, 0.4278], device='cuda:1'), in_proj_covar=tensor([0.0335, 0.0342, 0.0422, 0.0351, 0.0378, 0.0376, 0.0367, 0.0414], device='cuda:1'), out_proj_covar=tensor([9.9977e-05, 1.0312e-04, 1.2845e-04, 1.0627e-04, 1.1294e-04, 1.1284e-04, 1.0823e-04, 1.2566e-04], device='cuda:1') 2023-04-27 10:38:07,286 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-04-27 10:38:15,510 INFO [finetune.py:976] (1/7) Epoch 17, batch 5600, loss[loss=0.2161, simple_loss=0.2862, pruned_loss=0.07298, over 4910.00 frames. ], tot_loss[loss=0.1803, simple_loss=0.25, pruned_loss=0.0553, over 955669.99 frames. ], batch size: 36, lr: 3.37e-03, grad_scale: 32.0 2023-04-27 10:38:26,564 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.133e+02 1.758e+02 2.119e+02 2.585e+02 7.806e+02, threshold=4.239e+02, percent-clipped=5.0 2023-04-27 10:38:37,084 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5873, 1.8310, 1.7208, 2.0541, 2.0348, 2.2737, 1.7617, 3.8287], device='cuda:1'), covar=tensor([0.0502, 0.0707, 0.0733, 0.1038, 0.0558, 0.0428, 0.0689, 0.0155], device='cuda:1'), in_proj_covar=tensor([0.0037, 0.0038, 0.0039, 0.0043, 0.0039, 0.0038, 0.0038, 0.0056], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:1') 2023-04-27 10:38:37,099 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=97260.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:38:49,396 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.51 vs. limit=5.0 2023-04-27 10:39:12,277 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6621, 1.5129, 1.7253, 1.9885, 2.0559, 1.5804, 1.2289, 1.8339], device='cuda:1'), covar=tensor([0.0869, 0.1193, 0.0738, 0.0616, 0.0613, 0.0861, 0.0859, 0.0599], device='cuda:1'), in_proj_covar=tensor([0.0186, 0.0199, 0.0180, 0.0169, 0.0175, 0.0178, 0.0150, 0.0177], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 10:39:14,493 INFO [finetune.py:976] (1/7) Epoch 17, batch 5650, loss[loss=0.2002, simple_loss=0.2815, pruned_loss=0.05949, over 4900.00 frames. ], tot_loss[loss=0.1825, simple_loss=0.2532, pruned_loss=0.05594, over 956380.17 frames. ], batch size: 37, lr: 3.37e-03, grad_scale: 32.0 2023-04-27 10:39:47,534 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=97321.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:40:16,397 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=97342.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:40:16,919 INFO [finetune.py:976] (1/7) Epoch 17, batch 5700, loss[loss=0.1456, simple_loss=0.198, pruned_loss=0.04667, over 4371.00 frames. ], tot_loss[loss=0.179, simple_loss=0.2484, pruned_loss=0.05474, over 937787.10 frames. ], batch size: 19, lr: 3.36e-03, grad_scale: 32.0 2023-04-27 10:40:19,377 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=97347.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:40:22,861 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 8.291e+01 1.370e+02 1.630e+02 1.988e+02 3.241e+02, threshold=3.261e+02, percent-clipped=0.0 2023-04-27 10:40:24,680 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=97356.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:40:46,576 INFO [finetune.py:976] (1/7) Epoch 18, batch 0, loss[loss=0.169, simple_loss=0.2479, pruned_loss=0.04507, over 4919.00 frames. ], tot_loss[loss=0.169, simple_loss=0.2479, pruned_loss=0.04507, over 4919.00 frames. ], batch size: 38, lr: 3.36e-03, grad_scale: 64.0 2023-04-27 10:40:46,576 INFO [finetune.py:1001] (1/7) Computing validation loss 2023-04-27 10:41:03,166 INFO [finetune.py:1010] (1/7) Epoch 18, validation: loss=0.1537, simple_loss=0.225, pruned_loss=0.04121, over 2265189.00 frames. 2023-04-27 10:41:03,166 INFO [finetune.py:1011] (1/7) Maximum memory allocated so far is 6391MB 2023-04-27 10:41:05,487 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=97374.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:41:28,963 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=97403.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 10:41:32,009 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=97408.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:41:40,837 INFO [finetune.py:976] (1/7) Epoch 18, batch 50, loss[loss=0.2005, simple_loss=0.2712, pruned_loss=0.06489, over 4783.00 frames. ], tot_loss[loss=0.1716, simple_loss=0.2439, pruned_loss=0.04967, over 213603.15 frames. ], batch size: 51, lr: 3.36e-03, grad_scale: 64.0 2023-04-27 10:41:49,635 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=97432.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:41:51,466 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=97435.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:42:02,221 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.680e+01 1.493e+02 1.785e+02 2.115e+02 3.636e+02, threshold=3.569e+02, percent-clipped=3.0 2023-04-27 10:42:14,117 INFO [finetune.py:976] (1/7) Epoch 18, batch 100, loss[loss=0.179, simple_loss=0.2477, pruned_loss=0.05511, over 4867.00 frames. ], tot_loss[loss=0.1717, simple_loss=0.2415, pruned_loss=0.05092, over 378905.33 frames. ], batch size: 34, lr: 3.36e-03, grad_scale: 64.0 2023-04-27 10:42:22,013 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=97480.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:42:23,973 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=6.30 vs. limit=5.0 2023-04-27 10:42:25,741 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=3.54 vs. limit=5.0 2023-04-27 10:42:30,675 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-04-27 10:42:47,722 INFO [finetune.py:976] (1/7) Epoch 18, batch 150, loss[loss=0.1591, simple_loss=0.2343, pruned_loss=0.04197, over 4830.00 frames. ], tot_loss[loss=0.1694, simple_loss=0.2379, pruned_loss=0.05045, over 507490.53 frames. ], batch size: 30, lr: 3.36e-03, grad_scale: 64.0 2023-04-27 10:43:03,123 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([4.8162, 4.6782, 3.2209, 5.5491, 4.9254, 4.7804, 2.0053, 4.8677], device='cuda:1'), covar=tensor([0.1445, 0.1089, 0.2943, 0.0872, 0.2944, 0.1591, 0.6097, 0.1902], device='cuda:1'), in_proj_covar=tensor([0.0243, 0.0212, 0.0247, 0.0303, 0.0297, 0.0249, 0.0272, 0.0271], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 10:43:08,515 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.039e+02 1.668e+02 1.928e+02 2.305e+02 4.422e+02, threshold=3.856e+02, percent-clipped=2.0 2023-04-27 10:43:09,852 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=97555.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:43:19,854 INFO [finetune.py:976] (1/7) Epoch 18, batch 200, loss[loss=0.1974, simple_loss=0.2602, pruned_loss=0.06729, over 4774.00 frames. ], tot_loss[loss=0.1715, simple_loss=0.2397, pruned_loss=0.05172, over 607292.48 frames. ], batch size: 54, lr: 3.36e-03, grad_scale: 64.0 2023-04-27 10:43:32,234 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.3267, 2.9143, 0.8949, 1.6605, 2.1076, 1.4153, 3.9838, 2.1266], device='cuda:1'), covar=tensor([0.0698, 0.0778, 0.0949, 0.1205, 0.0546, 0.1012, 0.0158, 0.0579], device='cuda:1'), in_proj_covar=tensor([0.0051, 0.0066, 0.0049, 0.0047, 0.0050, 0.0052, 0.0074, 0.0052], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:1') 2023-04-27 10:43:55,458 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=97616.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:43:55,513 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=97616.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:43:58,890 INFO [finetune.py:976] (1/7) Epoch 18, batch 250, loss[loss=0.1803, simple_loss=0.2607, pruned_loss=0.04994, over 4920.00 frames. ], tot_loss[loss=0.1744, simple_loss=0.2437, pruned_loss=0.05256, over 685734.32 frames. ], batch size: 42, lr: 3.36e-03, grad_scale: 64.0 2023-04-27 10:44:22,367 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.1929, 2.5362, 0.8300, 1.5035, 1.4904, 1.9281, 1.5648, 0.8498], device='cuda:1'), covar=tensor([0.1497, 0.1383, 0.1814, 0.1245, 0.1167, 0.0902, 0.1541, 0.1739], device='cuda:1'), in_proj_covar=tensor([0.0117, 0.0243, 0.0137, 0.0121, 0.0133, 0.0153, 0.0118, 0.0121], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-27 10:44:29,795 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9815, 2.4444, 0.9343, 1.3804, 1.8496, 1.1463, 3.3168, 1.6220], device='cuda:1'), covar=tensor([0.0721, 0.0729, 0.0841, 0.1225, 0.0513, 0.1044, 0.0213, 0.0660], device='cuda:1'), in_proj_covar=tensor([0.0051, 0.0066, 0.0049, 0.0046, 0.0050, 0.0052, 0.0074, 0.0051], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:1') 2023-04-27 10:44:42,422 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.134e+02 1.696e+02 2.035e+02 2.379e+02 5.416e+02, threshold=4.070e+02, percent-clipped=1.0 2023-04-27 10:44:49,675 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=97656.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:45:03,745 INFO [finetune.py:976] (1/7) Epoch 18, batch 300, loss[loss=0.1621, simple_loss=0.2365, pruned_loss=0.04386, over 4748.00 frames. ], tot_loss[loss=0.1766, simple_loss=0.2475, pruned_loss=0.05289, over 746714.07 frames. ], batch size: 27, lr: 3.36e-03, grad_scale: 64.0 2023-04-27 10:45:09,983 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=97674.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 10:45:43,984 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=97698.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 10:45:47,049 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=97703.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:45:47,652 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=97704.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:46:07,661 INFO [finetune.py:976] (1/7) Epoch 18, batch 350, loss[loss=0.169, simple_loss=0.2548, pruned_loss=0.04157, over 4860.00 frames. ], tot_loss[loss=0.1784, simple_loss=0.2492, pruned_loss=0.05376, over 793995.17 frames. ], batch size: 49, lr: 3.36e-03, grad_scale: 64.0 2023-04-27 10:46:18,790 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=97730.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:46:28,144 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=97735.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 10:46:48,151 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.8404, 2.4456, 2.8405, 3.4532, 2.7811, 2.3397, 2.1467, 2.8203], device='cuda:1'), covar=tensor([0.3386, 0.3126, 0.1575, 0.2540, 0.2832, 0.2777, 0.3805, 0.1945], device='cuda:1'), in_proj_covar=tensor([0.0292, 0.0247, 0.0228, 0.0316, 0.0218, 0.0232, 0.0229, 0.0186], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-27 10:46:51,633 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.284e+02 1.687e+02 1.979e+02 2.374e+02 3.417e+02, threshold=3.957e+02, percent-clipped=0.0 2023-04-27 10:47:08,343 INFO [finetune.py:976] (1/7) Epoch 18, batch 400, loss[loss=0.1677, simple_loss=0.2369, pruned_loss=0.04924, over 4877.00 frames. ], tot_loss[loss=0.1786, simple_loss=0.2498, pruned_loss=0.05364, over 829477.39 frames. ], batch size: 43, lr: 3.36e-03, grad_scale: 64.0 2023-04-27 10:47:10,266 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.1244, 2.6066, 1.0142, 1.5488, 1.9329, 1.2825, 3.3967, 1.8351], device='cuda:1'), covar=tensor([0.0624, 0.0721, 0.0851, 0.1104, 0.0438, 0.0931, 0.0165, 0.0545], device='cuda:1'), in_proj_covar=tensor([0.0051, 0.0066, 0.0049, 0.0046, 0.0050, 0.0052, 0.0074, 0.0051], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:1') 2023-04-27 10:47:16,926 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=3.81 vs. limit=5.0 2023-04-27 10:47:33,116 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-04-27 10:47:42,003 INFO [finetune.py:976] (1/7) Epoch 18, batch 450, loss[loss=0.1439, simple_loss=0.2061, pruned_loss=0.04083, over 4259.00 frames. ], tot_loss[loss=0.1776, simple_loss=0.2486, pruned_loss=0.05333, over 857660.57 frames. ], batch size: 18, lr: 3.36e-03, grad_scale: 32.0 2023-04-27 10:47:58,391 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.3585, 1.8061, 2.3113, 2.6792, 2.2284, 1.7386, 1.3767, 2.0018], device='cuda:1'), covar=tensor([0.2826, 0.2943, 0.1524, 0.2081, 0.2559, 0.2561, 0.4115, 0.1911], device='cuda:1'), in_proj_covar=tensor([0.0291, 0.0246, 0.0226, 0.0314, 0.0217, 0.0231, 0.0227, 0.0185], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-27 10:48:05,036 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.073e+02 1.551e+02 1.835e+02 2.204e+02 3.781e+02, threshold=3.670e+02, percent-clipped=0.0 2023-04-27 10:48:15,396 INFO [finetune.py:976] (1/7) Epoch 18, batch 500, loss[loss=0.1754, simple_loss=0.2454, pruned_loss=0.05269, over 4905.00 frames. ], tot_loss[loss=0.1759, simple_loss=0.2463, pruned_loss=0.05274, over 879788.36 frames. ], batch size: 37, lr: 3.36e-03, grad_scale: 32.0 2023-04-27 10:48:26,746 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=97888.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:48:27,590 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.97 vs. limit=2.0 2023-04-27 10:48:42,842 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=97911.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:48:45,957 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=97916.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:48:48,922 INFO [finetune.py:976] (1/7) Epoch 18, batch 550, loss[loss=0.1397, simple_loss=0.2147, pruned_loss=0.03232, over 4765.00 frames. ], tot_loss[loss=0.174, simple_loss=0.2436, pruned_loss=0.05216, over 897530.52 frames. ], batch size: 26, lr: 3.36e-03, grad_scale: 32.0 2023-04-27 10:49:09,141 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=97949.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:49:12,032 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.221e+01 1.571e+02 1.914e+02 2.299e+02 4.493e+02, threshold=3.828e+02, percent-clipped=2.0 2023-04-27 10:49:18,169 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=97964.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:49:22,478 INFO [finetune.py:976] (1/7) Epoch 18, batch 600, loss[loss=0.1637, simple_loss=0.2362, pruned_loss=0.04561, over 4811.00 frames. ], tot_loss[loss=0.1742, simple_loss=0.2436, pruned_loss=0.05236, over 908202.71 frames. ], batch size: 39, lr: 3.36e-03, grad_scale: 32.0 2023-04-27 10:49:35,977 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-04-27 10:49:41,018 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=97998.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:49:46,925 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=98003.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:49:57,856 INFO [finetune.py:976] (1/7) Epoch 18, batch 650, loss[loss=0.2204, simple_loss=0.2951, pruned_loss=0.07287, over 4908.00 frames. ], tot_loss[loss=0.1759, simple_loss=0.2462, pruned_loss=0.05278, over 919729.04 frames. ], batch size: 43, lr: 3.36e-03, grad_scale: 32.0 2023-04-27 10:50:03,551 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=98030.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 10:50:03,563 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=98030.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:50:14,375 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=98046.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:50:17,925 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=98051.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:50:26,530 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.176e+02 1.729e+02 2.021e+02 2.442e+02 4.739e+02, threshold=4.043e+02, percent-clipped=3.0 2023-04-27 10:50:48,619 INFO [finetune.py:976] (1/7) Epoch 18, batch 700, loss[loss=0.1727, simple_loss=0.2536, pruned_loss=0.04594, over 4804.00 frames. ], tot_loss[loss=0.1781, simple_loss=0.2492, pruned_loss=0.05355, over 930293.75 frames. ], batch size: 39, lr: 3.36e-03, grad_scale: 32.0 2023-04-27 10:50:58,804 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=98078.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:51:14,125 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.50 vs. limit=2.0 2023-04-27 10:51:33,409 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=98106.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:51:45,519 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([3.8851, 3.8171, 2.6982, 4.5177, 3.9336, 3.8794, 1.6434, 3.8731], device='cuda:1'), covar=tensor([0.1574, 0.1228, 0.3015, 0.1564, 0.3441, 0.1889, 0.6016, 0.2328], device='cuda:1'), in_proj_covar=tensor([0.0239, 0.0210, 0.0245, 0.0301, 0.0294, 0.0246, 0.0268, 0.0267], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 10:51:53,714 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6082, 1.1920, 1.3150, 1.2170, 1.7005, 1.3893, 1.1408, 1.2505], device='cuda:1'), covar=tensor([0.1521, 0.1466, 0.2107, 0.1472, 0.0864, 0.1517, 0.1975, 0.2305], device='cuda:1'), in_proj_covar=tensor([0.0308, 0.0312, 0.0351, 0.0287, 0.0329, 0.0310, 0.0300, 0.0366], device='cuda:1'), out_proj_covar=tensor([6.3701e-05, 6.5053e-05, 7.4740e-05, 5.8196e-05, 6.8415e-05, 6.5210e-05, 6.3188e-05, 7.8106e-05], device='cuda:1') 2023-04-27 10:51:54,812 INFO [finetune.py:976] (1/7) Epoch 18, batch 750, loss[loss=0.1717, simple_loss=0.245, pruned_loss=0.04914, over 4815.00 frames. ], tot_loss[loss=0.1786, simple_loss=0.2501, pruned_loss=0.05356, over 937435.44 frames. ], batch size: 33, lr: 3.36e-03, grad_scale: 32.0 2023-04-27 10:52:31,990 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.015e+02 1.562e+02 1.753e+02 1.961e+02 3.537e+02, threshold=3.506e+02, percent-clipped=0.0 2023-04-27 10:52:42,186 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=98167.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:52:44,510 INFO [finetune.py:976] (1/7) Epoch 18, batch 800, loss[loss=0.1531, simple_loss=0.2334, pruned_loss=0.03633, over 4787.00 frames. ], tot_loss[loss=0.1773, simple_loss=0.2491, pruned_loss=0.05273, over 940573.48 frames. ], batch size: 29, lr: 3.36e-03, grad_scale: 32.0 2023-04-27 10:53:10,396 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=98211.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:53:17,963 INFO [finetune.py:976] (1/7) Epoch 18, batch 850, loss[loss=0.1621, simple_loss=0.2318, pruned_loss=0.04618, over 4933.00 frames. ], tot_loss[loss=0.1751, simple_loss=0.2466, pruned_loss=0.05184, over 944279.85 frames. ], batch size: 33, lr: 3.35e-03, grad_scale: 32.0 2023-04-27 10:53:32,029 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=98244.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:53:38,453 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.063e+02 1.514e+02 1.837e+02 2.117e+02 3.312e+02, threshold=3.674e+02, percent-clipped=0.0 2023-04-27 10:53:42,073 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=98259.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:53:46,703 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8124, 2.2109, 1.7410, 1.5618, 1.3287, 1.3647, 1.8620, 1.2656], device='cuda:1'), covar=tensor([0.1649, 0.1400, 0.1473, 0.1782, 0.2159, 0.1841, 0.0950, 0.1998], device='cuda:1'), in_proj_covar=tensor([0.0199, 0.0214, 0.0169, 0.0206, 0.0201, 0.0185, 0.0157, 0.0186], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-27 10:53:51,234 INFO [finetune.py:976] (1/7) Epoch 18, batch 900, loss[loss=0.1497, simple_loss=0.2228, pruned_loss=0.03828, over 4834.00 frames. ], tot_loss[loss=0.1745, simple_loss=0.245, pruned_loss=0.05203, over 945258.28 frames. ], batch size: 39, lr: 3.35e-03, grad_scale: 32.0 2023-04-27 10:54:24,150 INFO [finetune.py:976] (1/7) Epoch 18, batch 950, loss[loss=0.2193, simple_loss=0.2709, pruned_loss=0.0838, over 4824.00 frames. ], tot_loss[loss=0.1744, simple_loss=0.244, pruned_loss=0.05234, over 947806.17 frames. ], batch size: 30, lr: 3.35e-03, grad_scale: 32.0 2023-04-27 10:54:30,296 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=98330.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 10:54:44,943 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.018e+02 1.562e+02 1.785e+02 2.020e+02 3.278e+02, threshold=3.570e+02, percent-clipped=0.0 2023-04-27 10:54:57,898 INFO [finetune.py:976] (1/7) Epoch 18, batch 1000, loss[loss=0.1688, simple_loss=0.2592, pruned_loss=0.03923, over 4817.00 frames. ], tot_loss[loss=0.1751, simple_loss=0.2449, pruned_loss=0.05266, over 948829.86 frames. ], batch size: 51, lr: 3.35e-03, grad_scale: 32.0 2023-04-27 10:55:02,793 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=98378.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 10:55:22,444 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=98410.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:55:25,925 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.1744, 1.3688, 1.2761, 1.7199, 1.5747, 1.6970, 1.3211, 3.0417], device='cuda:1'), covar=tensor([0.0630, 0.0832, 0.0829, 0.1189, 0.0667, 0.0501, 0.0782, 0.0145], device='cuda:1'), in_proj_covar=tensor([0.0037, 0.0038, 0.0039, 0.0043, 0.0040, 0.0038, 0.0038, 0.0056], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:1') 2023-04-27 10:55:29,960 INFO [finetune.py:976] (1/7) Epoch 18, batch 1050, loss[loss=0.2, simple_loss=0.2822, pruned_loss=0.05889, over 4899.00 frames. ], tot_loss[loss=0.1774, simple_loss=0.2486, pruned_loss=0.05307, over 949934.97 frames. ], batch size: 43, lr: 3.35e-03, grad_scale: 32.0 2023-04-27 10:55:40,012 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9705, 1.5081, 1.8363, 1.7350, 1.7803, 1.5019, 0.8358, 1.4253], device='cuda:1'), covar=tensor([0.3521, 0.3303, 0.1852, 0.2364, 0.2734, 0.2764, 0.4407, 0.2256], device='cuda:1'), in_proj_covar=tensor([0.0289, 0.0245, 0.0225, 0.0313, 0.0217, 0.0230, 0.0227, 0.0183], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-27 10:55:52,030 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.114e+02 1.610e+02 1.888e+02 2.252e+02 4.818e+02, threshold=3.776e+02, percent-clipped=3.0 2023-04-27 10:55:56,967 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=98462.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:56:08,982 INFO [finetune.py:976] (1/7) Epoch 18, batch 1100, loss[loss=0.2456, simple_loss=0.2924, pruned_loss=0.09936, over 4919.00 frames. ], tot_loss[loss=0.1778, simple_loss=0.2494, pruned_loss=0.05314, over 952745.69 frames. ], batch size: 38, lr: 3.35e-03, grad_scale: 32.0 2023-04-27 10:56:09,120 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=98471.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:56:43,236 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([0.9211, 0.8955, 1.0770, 1.0117, 0.8991, 0.8214, 0.9163, 0.5690], device='cuda:1'), covar=tensor([0.0552, 0.0444, 0.0531, 0.0441, 0.0589, 0.0928, 0.0433, 0.0687], device='cuda:1'), in_proj_covar=tensor([0.0068, 0.0069, 0.0068, 0.0067, 0.0075, 0.0095, 0.0074, 0.0067], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-27 10:56:54,542 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6872, 1.2304, 1.8268, 2.2026, 1.7971, 1.7163, 1.7796, 1.7583], device='cuda:1'), covar=tensor([0.4630, 0.6660, 0.6684, 0.5777, 0.6369, 0.8277, 0.7963, 0.8554], device='cuda:1'), in_proj_covar=tensor([0.0417, 0.0404, 0.0493, 0.0499, 0.0447, 0.0474, 0.0480, 0.0484], device='cuda:1'), out_proj_covar=tensor([1.0042e-04, 9.9326e-05, 1.1085e-04, 1.1927e-04, 1.0716e-04, 1.1388e-04, 1.1396e-04, 1.1425e-04], device='cuda:1') 2023-04-27 10:57:13,272 INFO [finetune.py:976] (1/7) Epoch 18, batch 1150, loss[loss=0.16, simple_loss=0.2388, pruned_loss=0.0406, over 4736.00 frames. ], tot_loss[loss=0.1797, simple_loss=0.2514, pruned_loss=0.05394, over 954421.73 frames. ], batch size: 27, lr: 3.35e-03, grad_scale: 32.0 2023-04-27 10:57:45,480 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=98543.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:57:46,067 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=98544.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:57:57,932 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.919e+01 1.656e+02 1.874e+02 2.209e+02 6.337e+02, threshold=3.749e+02, percent-clipped=3.0 2023-04-27 10:58:14,178 INFO [finetune.py:976] (1/7) Epoch 18, batch 1200, loss[loss=0.1668, simple_loss=0.2364, pruned_loss=0.04855, over 4844.00 frames. ], tot_loss[loss=0.1787, simple_loss=0.25, pruned_loss=0.05374, over 955663.56 frames. ], batch size: 49, lr: 3.35e-03, grad_scale: 32.0 2023-04-27 10:58:15,587 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0041, 1.7553, 2.1865, 2.4548, 2.0565, 1.9377, 2.0424, 2.0779], device='cuda:1'), covar=tensor([0.5088, 0.7532, 0.7768, 0.6024, 0.6432, 0.9924, 1.0210, 1.0131], device='cuda:1'), in_proj_covar=tensor([0.0418, 0.0405, 0.0493, 0.0500, 0.0447, 0.0475, 0.0480, 0.0485], device='cuda:1'), out_proj_covar=tensor([1.0056e-04, 9.9488e-05, 1.1097e-04, 1.1944e-04, 1.0714e-04, 1.1405e-04, 1.1397e-04, 1.1441e-04], device='cuda:1') 2023-04-27 10:58:28,912 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=98592.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:58:36,398 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=98604.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:58:46,674 INFO [finetune.py:976] (1/7) Epoch 18, batch 1250, loss[loss=0.1787, simple_loss=0.2467, pruned_loss=0.05528, over 4898.00 frames. ], tot_loss[loss=0.1767, simple_loss=0.2476, pruned_loss=0.05295, over 956994.49 frames. ], batch size: 35, lr: 3.35e-03, grad_scale: 32.0 2023-04-27 10:59:10,194 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.095e+02 1.558e+02 1.895e+02 2.216e+02 4.222e+02, threshold=3.789e+02, percent-clipped=2.0 2023-04-27 10:59:20,476 INFO [finetune.py:976] (1/7) Epoch 18, batch 1300, loss[loss=0.1719, simple_loss=0.2373, pruned_loss=0.05323, over 4222.00 frames. ], tot_loss[loss=0.1739, simple_loss=0.2439, pruned_loss=0.05189, over 957310.00 frames. ], batch size: 66, lr: 3.35e-03, grad_scale: 32.0 2023-04-27 10:59:38,653 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6696, 2.7164, 2.3815, 2.4017, 2.7752, 2.5182, 3.7415, 1.9988], device='cuda:1'), covar=tensor([0.3636, 0.2123, 0.3567, 0.3267, 0.1744, 0.2352, 0.1211, 0.4152], device='cuda:1'), in_proj_covar=tensor([0.0335, 0.0342, 0.0422, 0.0351, 0.0377, 0.0376, 0.0367, 0.0414], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 10:59:50,729 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.5268, 2.3204, 2.5682, 2.9997, 2.8643, 2.4278, 2.0860, 2.7424], device='cuda:1'), covar=tensor([0.0888, 0.0985, 0.0644, 0.0642, 0.0594, 0.0898, 0.0777, 0.0566], device='cuda:1'), in_proj_covar=tensor([0.0186, 0.0198, 0.0179, 0.0170, 0.0175, 0.0179, 0.0149, 0.0176], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 10:59:53,701 INFO [finetune.py:976] (1/7) Epoch 18, batch 1350, loss[loss=0.244, simple_loss=0.3037, pruned_loss=0.09216, over 4758.00 frames. ], tot_loss[loss=0.1738, simple_loss=0.2437, pruned_loss=0.05199, over 955481.91 frames. ], batch size: 54, lr: 3.35e-03, grad_scale: 32.0 2023-04-27 10:59:58,510 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6854, 1.5166, 0.5935, 1.3383, 1.4502, 1.5446, 1.4270, 1.4574], device='cuda:1'), covar=tensor([0.0495, 0.0396, 0.0406, 0.0590, 0.0288, 0.0539, 0.0512, 0.0593], device='cuda:1'), in_proj_covar=tensor([0.0028, 0.0024, 0.0022, 0.0029, 0.0020, 0.0028, 0.0028, 0.0028], device='cuda:1'), out_proj_covar=tensor([0.0050, 0.0045, 0.0038, 0.0051, 0.0038, 0.0049, 0.0049, 0.0050], device='cuda:1') 2023-04-27 11:00:17,035 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.199e+02 1.614e+02 2.005e+02 2.367e+02 4.015e+02, threshold=4.011e+02, percent-clipped=1.0 2023-04-27 11:00:22,049 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=98762.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:00:24,435 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=98766.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:00:24,461 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.1428, 1.4850, 1.3143, 1.6676, 1.6074, 1.9311, 1.3855, 3.4095], device='cuda:1'), covar=tensor([0.0610, 0.0803, 0.0788, 0.1186, 0.0603, 0.0573, 0.0782, 0.0138], device='cuda:1'), in_proj_covar=tensor([0.0037, 0.0039, 0.0040, 0.0044, 0.0040, 0.0038, 0.0038, 0.0056], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:1') 2023-04-27 11:00:27,467 INFO [finetune.py:976] (1/7) Epoch 18, batch 1400, loss[loss=0.2154, simple_loss=0.2877, pruned_loss=0.07155, over 4901.00 frames. ], tot_loss[loss=0.1768, simple_loss=0.2468, pruned_loss=0.05341, over 955071.87 frames. ], batch size: 43, lr: 3.35e-03, grad_scale: 32.0 2023-04-27 11:00:28,868 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-04-27 11:00:47,550 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0048, 2.3128, 2.0620, 2.3441, 1.7052, 2.1325, 2.1364, 1.7703], device='cuda:1'), covar=tensor([0.1598, 0.1077, 0.0652, 0.0840, 0.2650, 0.0860, 0.1669, 0.1952], device='cuda:1'), in_proj_covar=tensor([0.0287, 0.0307, 0.0221, 0.0281, 0.0315, 0.0261, 0.0251, 0.0268], device='cuda:1'), out_proj_covar=tensor([1.1537e-04, 1.2227e-04, 8.7837e-05, 1.1143e-04, 1.2796e-04, 1.0380e-04, 1.0172e-04, 1.0636e-04], device='cuda:1') 2023-04-27 11:00:48,299 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.3788, 1.3341, 1.6908, 1.6295, 1.3090, 1.1928, 1.4339, 1.0433], device='cuda:1'), covar=tensor([0.0604, 0.0727, 0.0415, 0.0638, 0.0769, 0.1082, 0.0554, 0.0565], device='cuda:1'), in_proj_covar=tensor([0.0069, 0.0069, 0.0068, 0.0068, 0.0075, 0.0096, 0.0074, 0.0068], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-27 11:00:54,321 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=98810.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:00:56,167 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5944, 1.5148, 1.9037, 1.9338, 1.4380, 1.3476, 1.6721, 1.1215], device='cuda:1'), covar=tensor([0.0539, 0.0718, 0.0399, 0.0531, 0.0744, 0.1034, 0.0551, 0.0671], device='cuda:1'), in_proj_covar=tensor([0.0068, 0.0069, 0.0068, 0.0068, 0.0075, 0.0096, 0.0074, 0.0067], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-27 11:01:00,975 INFO [finetune.py:976] (1/7) Epoch 18, batch 1450, loss[loss=0.1688, simple_loss=0.2436, pruned_loss=0.04704, over 4763.00 frames. ], tot_loss[loss=0.1776, simple_loss=0.2484, pruned_loss=0.05337, over 955771.91 frames. ], batch size: 28, lr: 3.35e-03, grad_scale: 32.0 2023-04-27 11:01:02,318 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.5093, 1.9844, 2.4869, 2.8992, 2.4164, 1.9200, 1.8230, 2.3523], device='cuda:1'), covar=tensor([0.3327, 0.3290, 0.1712, 0.2457, 0.2829, 0.2862, 0.3925, 0.1994], device='cuda:1'), in_proj_covar=tensor([0.0289, 0.0245, 0.0226, 0.0315, 0.0218, 0.0230, 0.0227, 0.0184], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-27 11:01:24,438 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.190e+02 1.637e+02 1.928e+02 2.404e+02 4.442e+02, threshold=3.855e+02, percent-clipped=1.0 2023-04-27 11:01:34,709 INFO [finetune.py:976] (1/7) Epoch 18, batch 1500, loss[loss=0.1761, simple_loss=0.2527, pruned_loss=0.04976, over 4807.00 frames. ], tot_loss[loss=0.1781, simple_loss=0.2493, pruned_loss=0.05346, over 955984.41 frames. ], batch size: 40, lr: 3.35e-03, grad_scale: 32.0 2023-04-27 11:01:35,468 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7625, 2.2466, 1.7245, 1.7944, 1.3213, 1.3369, 1.7797, 1.2445], device='cuda:1'), covar=tensor([0.1577, 0.1294, 0.1405, 0.1561, 0.2228, 0.1927, 0.0932, 0.1983], device='cuda:1'), in_proj_covar=tensor([0.0197, 0.0212, 0.0168, 0.0204, 0.0199, 0.0184, 0.0156, 0.0185], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-27 11:01:40,266 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7300, 1.2579, 1.4054, 1.4265, 1.8603, 1.5644, 1.2667, 1.3631], device='cuda:1'), covar=tensor([0.1710, 0.1465, 0.2225, 0.1546, 0.0972, 0.1492, 0.2001, 0.2207], device='cuda:1'), in_proj_covar=tensor([0.0308, 0.0313, 0.0350, 0.0286, 0.0328, 0.0309, 0.0299, 0.0365], device='cuda:1'), out_proj_covar=tensor([6.3618e-05, 6.5122e-05, 7.4475e-05, 5.8086e-05, 6.8177e-05, 6.5049e-05, 6.3077e-05, 7.7820e-05], device='cuda:1') 2023-04-27 11:01:52,807 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=98890.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:02:02,807 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=98895.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:02:11,048 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=98899.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:02:35,628 INFO [finetune.py:976] (1/7) Epoch 18, batch 1550, loss[loss=0.2041, simple_loss=0.2882, pruned_loss=0.06, over 4882.00 frames. ], tot_loss[loss=0.1782, simple_loss=0.2495, pruned_loss=0.05346, over 956208.88 frames. ], batch size: 35, lr: 3.35e-03, grad_scale: 32.0 2023-04-27 11:03:17,444 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=98951.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:03:19,154 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.300e+01 1.636e+02 1.920e+02 2.366e+02 5.799e+02, threshold=3.840e+02, percent-clipped=2.0 2023-04-27 11:03:25,930 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=98956.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:03:41,674 INFO [finetune.py:976] (1/7) Epoch 18, batch 1600, loss[loss=0.1897, simple_loss=0.2531, pruned_loss=0.06311, over 4817.00 frames. ], tot_loss[loss=0.1769, simple_loss=0.2476, pruned_loss=0.05312, over 956924.09 frames. ], batch size: 41, lr: 3.35e-03, grad_scale: 32.0 2023-04-27 11:03:41,819 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=98971.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:04:20,540 INFO [finetune.py:976] (1/7) Epoch 18, batch 1650, loss[loss=0.1524, simple_loss=0.2287, pruned_loss=0.03803, over 4815.00 frames. ], tot_loss[loss=0.1752, simple_loss=0.2456, pruned_loss=0.05241, over 956103.76 frames. ], batch size: 41, lr: 3.35e-03, grad_scale: 32.0 2023-04-27 11:04:27,361 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=99032.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:04:40,839 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=99051.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:04:42,972 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.087e+02 1.609e+02 1.883e+02 2.336e+02 5.190e+02, threshold=3.766e+02, percent-clipped=4.0 2023-04-27 11:04:50,873 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=99066.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:04:53,813 INFO [finetune.py:976] (1/7) Epoch 18, batch 1700, loss[loss=0.183, simple_loss=0.2413, pruned_loss=0.06237, over 4794.00 frames. ], tot_loss[loss=0.174, simple_loss=0.2438, pruned_loss=0.05206, over 956736.93 frames. ], batch size: 29, lr: 3.35e-03, grad_scale: 32.0 2023-04-27 11:05:21,697 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=99112.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:05:22,857 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=99114.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:05:27,557 INFO [finetune.py:976] (1/7) Epoch 18, batch 1750, loss[loss=0.2385, simple_loss=0.3066, pruned_loss=0.08515, over 4827.00 frames. ], tot_loss[loss=0.1763, simple_loss=0.2464, pruned_loss=0.05308, over 956570.39 frames. ], batch size: 39, lr: 3.35e-03, grad_scale: 32.0 2023-04-27 11:05:50,008 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.166e+02 1.654e+02 1.955e+02 2.443e+02 4.969e+02, threshold=3.909e+02, percent-clipped=5.0 2023-04-27 11:06:01,215 INFO [finetune.py:976] (1/7) Epoch 18, batch 1800, loss[loss=0.1767, simple_loss=0.2465, pruned_loss=0.05349, over 4752.00 frames. ], tot_loss[loss=0.1783, simple_loss=0.2492, pruned_loss=0.05369, over 955126.99 frames. ], batch size: 54, lr: 3.35e-03, grad_scale: 32.0 2023-04-27 11:06:18,769 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=99199.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:06:34,181 INFO [finetune.py:976] (1/7) Epoch 18, batch 1850, loss[loss=0.1858, simple_loss=0.2563, pruned_loss=0.05769, over 4809.00 frames. ], tot_loss[loss=0.1807, simple_loss=0.2513, pruned_loss=0.05508, over 955346.07 frames. ], batch size: 39, lr: 3.35e-03, grad_scale: 32.0 2023-04-27 11:06:39,647 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=99229.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:06:49,925 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=99246.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:06:50,531 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=99247.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:06:53,443 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=99251.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:06:55,634 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.025e+02 1.735e+02 2.090e+02 2.546e+02 5.570e+02, threshold=4.180e+02, percent-clipped=4.0 2023-04-27 11:07:04,300 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6555, 2.0030, 1.0029, 1.4010, 2.1674, 1.5483, 1.4658, 1.5303], device='cuda:1'), covar=tensor([0.0506, 0.0361, 0.0312, 0.0557, 0.0239, 0.0514, 0.0505, 0.0564], device='cuda:1'), in_proj_covar=tensor([0.0028, 0.0024, 0.0022, 0.0029, 0.0020, 0.0028, 0.0028, 0.0028], device='cuda:1'), out_proj_covar=tensor([0.0050, 0.0045, 0.0038, 0.0051, 0.0038, 0.0049, 0.0049, 0.0050], device='cuda:1') 2023-04-27 11:07:06,158 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=99268.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:07:07,875 INFO [finetune.py:976] (1/7) Epoch 18, batch 1900, loss[loss=0.1708, simple_loss=0.2465, pruned_loss=0.04754, over 4814.00 frames. ], tot_loss[loss=0.1807, simple_loss=0.2519, pruned_loss=0.05482, over 955788.74 frames. ], batch size: 45, lr: 3.35e-03, grad_scale: 32.0 2023-04-27 11:07:20,139 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=99290.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:07:34,027 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-04-27 11:08:03,838 INFO [finetune.py:976] (1/7) Epoch 18, batch 1950, loss[loss=0.1953, simple_loss=0.274, pruned_loss=0.05829, over 4905.00 frames. ], tot_loss[loss=0.1774, simple_loss=0.2485, pruned_loss=0.05311, over 954220.59 frames. ], batch size: 36, lr: 3.35e-03, grad_scale: 32.0 2023-04-27 11:08:05,831 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=99324.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:08:07,585 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=99327.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:08:09,366 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=99329.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:08:30,192 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.408e+01 1.522e+02 1.826e+02 2.209e+02 4.570e+02, threshold=3.652e+02, percent-clipped=1.0 2023-04-27 11:08:52,739 INFO [finetune.py:976] (1/7) Epoch 18, batch 2000, loss[loss=0.1758, simple_loss=0.2495, pruned_loss=0.05106, over 4891.00 frames. ], tot_loss[loss=0.1766, simple_loss=0.2469, pruned_loss=0.05317, over 955763.47 frames. ], batch size: 35, lr: 3.34e-03, grad_scale: 32.0 2023-04-27 11:09:12,997 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=99385.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:09:26,268 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=3.93 vs. limit=5.0 2023-04-27 11:09:44,016 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=99407.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:09:54,356 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=99420.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:09:54,850 INFO [finetune.py:976] (1/7) Epoch 18, batch 2050, loss[loss=0.1708, simple_loss=0.2424, pruned_loss=0.0496, over 4908.00 frames. ], tot_loss[loss=0.1738, simple_loss=0.2438, pruned_loss=0.05195, over 956293.17 frames. ], batch size: 35, lr: 3.34e-03, grad_scale: 32.0 2023-04-27 11:10:15,956 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.033e+02 1.635e+02 1.952e+02 2.334e+02 5.427e+02, threshold=3.904e+02, percent-clipped=2.0 2023-04-27 11:10:26,037 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=99468.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 11:10:28,728 INFO [finetune.py:976] (1/7) Epoch 18, batch 2100, loss[loss=0.1796, simple_loss=0.2424, pruned_loss=0.05837, over 4871.00 frames. ], tot_loss[loss=0.1744, simple_loss=0.244, pruned_loss=0.05241, over 955153.11 frames. ], batch size: 34, lr: 3.34e-03, grad_scale: 32.0 2023-04-27 11:10:35,444 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=99481.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:10:43,925 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9817, 2.6758, 2.0897, 2.0770, 1.4406, 1.4241, 2.0830, 1.3734], device='cuda:1'), covar=tensor([0.1662, 0.1436, 0.1446, 0.1686, 0.2351, 0.2035, 0.1023, 0.2119], device='cuda:1'), in_proj_covar=tensor([0.0198, 0.0213, 0.0168, 0.0205, 0.0201, 0.0185, 0.0156, 0.0187], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-27 11:10:57,627 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7314, 2.0181, 1.3131, 1.4589, 2.1807, 1.5932, 1.5144, 1.5965], device='cuda:1'), covar=tensor([0.0482, 0.0347, 0.0267, 0.0525, 0.0233, 0.0496, 0.0518, 0.0558], device='cuda:1'), in_proj_covar=tensor([0.0028, 0.0024, 0.0022, 0.0029, 0.0019, 0.0028, 0.0028, 0.0028], device='cuda:1'), out_proj_covar=tensor([0.0050, 0.0044, 0.0038, 0.0051, 0.0038, 0.0049, 0.0049, 0.0050], device='cuda:1') 2023-04-27 11:11:02,725 INFO [finetune.py:976] (1/7) Epoch 18, batch 2150, loss[loss=0.1779, simple_loss=0.2548, pruned_loss=0.05052, over 4822.00 frames. ], tot_loss[loss=0.1777, simple_loss=0.2479, pruned_loss=0.05371, over 955529.30 frames. ], batch size: 51, lr: 3.34e-03, grad_scale: 32.0 2023-04-27 11:11:08,207 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=99529.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 11:11:11,756 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.3048, 2.3178, 2.0413, 1.9670, 2.5477, 1.8977, 3.0416, 1.8173], device='cuda:1'), covar=tensor([0.3695, 0.1821, 0.4673, 0.3101, 0.1580, 0.2870, 0.1600, 0.4376], device='cuda:1'), in_proj_covar=tensor([0.0340, 0.0348, 0.0428, 0.0355, 0.0383, 0.0381, 0.0371, 0.0421], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 11:11:18,913 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=99546.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:11:21,401 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.1841, 1.8278, 1.5433, 2.1002, 2.3081, 1.9436, 1.8536, 1.6062], device='cuda:1'), covar=tensor([0.1821, 0.1510, 0.1772, 0.1511, 0.1326, 0.1671, 0.2329, 0.2355], device='cuda:1'), in_proj_covar=tensor([0.0311, 0.0314, 0.0352, 0.0289, 0.0329, 0.0310, 0.0302, 0.0368], device='cuda:1'), out_proj_covar=tensor([6.4265e-05, 6.5370e-05, 7.4739e-05, 5.8666e-05, 6.8415e-05, 6.5246e-05, 6.3553e-05, 7.8372e-05], device='cuda:1') 2023-04-27 11:11:21,997 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=99551.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:11:23,739 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.043e+02 1.609e+02 1.810e+02 2.358e+02 5.458e+02, threshold=3.621e+02, percent-clipped=4.0 2023-04-27 11:11:32,857 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=99568.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:11:35,034 INFO [finetune.py:976] (1/7) Epoch 18, batch 2200, loss[loss=0.1418, simple_loss=0.2219, pruned_loss=0.03083, over 4806.00 frames. ], tot_loss[loss=0.1809, simple_loss=0.2515, pruned_loss=0.05521, over 954867.34 frames. ], batch size: 25, lr: 3.34e-03, grad_scale: 32.0 2023-04-27 11:11:42,644 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0667, 1.5174, 1.9049, 2.2658, 1.8669, 1.4853, 1.0316, 1.5820], device='cuda:1'), covar=tensor([0.3218, 0.3285, 0.1632, 0.2149, 0.2528, 0.2772, 0.4570, 0.2165], device='cuda:1'), in_proj_covar=tensor([0.0291, 0.0246, 0.0227, 0.0316, 0.0218, 0.0231, 0.0229, 0.0185], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-27 11:11:45,026 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=99585.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:11:47,519 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=99589.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:11:48,161 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.4462, 1.8575, 1.8284, 1.9444, 1.7351, 1.8060, 1.9124, 1.8301], device='cuda:1'), covar=tensor([0.3641, 0.5842, 0.5009, 0.4655, 0.5890, 0.7435, 0.5998, 0.5846], device='cuda:1'), in_proj_covar=tensor([0.0330, 0.0371, 0.0320, 0.0331, 0.0343, 0.0393, 0.0354, 0.0325], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 11:11:51,021 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=99594.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:11:51,047 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9903, 1.3996, 4.9949, 4.6725, 4.3330, 4.7185, 4.4383, 4.4313], device='cuda:1'), covar=tensor([0.6664, 0.5909, 0.1018, 0.1807, 0.1012, 0.1405, 0.1271, 0.1550], device='cuda:1'), in_proj_covar=tensor([0.0308, 0.0308, 0.0403, 0.0407, 0.0349, 0.0404, 0.0311, 0.0366], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-27 11:11:54,027 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=99599.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:12:08,620 INFO [finetune.py:976] (1/7) Epoch 18, batch 2250, loss[loss=0.2247, simple_loss=0.2952, pruned_loss=0.07712, over 4895.00 frames. ], tot_loss[loss=0.1814, simple_loss=0.2525, pruned_loss=0.05515, over 955472.21 frames. ], batch size: 36, lr: 3.34e-03, grad_scale: 32.0 2023-04-27 11:12:11,006 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=99624.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:12:13,833 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=99627.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:12:15,099 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=99629.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:12:28,879 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=99650.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:12:31,119 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.508e+01 1.589e+02 1.867e+02 2.262e+02 4.430e+02, threshold=3.734e+02, percent-clipped=1.0 2023-04-27 11:12:32,502 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.2628, 1.7362, 2.1851, 2.6885, 2.1286, 1.6935, 1.3962, 1.9115], device='cuda:1'), covar=tensor([0.3018, 0.3072, 0.1635, 0.2180, 0.2639, 0.2680, 0.4162, 0.2059], device='cuda:1'), in_proj_covar=tensor([0.0289, 0.0245, 0.0226, 0.0314, 0.0217, 0.0230, 0.0227, 0.0184], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-27 11:12:41,913 INFO [finetune.py:976] (1/7) Epoch 18, batch 2300, loss[loss=0.1536, simple_loss=0.2315, pruned_loss=0.03785, over 4775.00 frames. ], tot_loss[loss=0.1797, simple_loss=0.251, pruned_loss=0.05417, over 953735.05 frames. ], batch size: 26, lr: 3.34e-03, grad_scale: 32.0 2023-04-27 11:12:44,909 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=99675.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:12:49,367 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=99680.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:12:49,397 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=99680.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:13:12,569 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=99707.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:13:26,444 INFO [finetune.py:976] (1/7) Epoch 18, batch 2350, loss[loss=0.1883, simple_loss=0.2524, pruned_loss=0.06213, over 4928.00 frames. ], tot_loss[loss=0.1775, simple_loss=0.2481, pruned_loss=0.05345, over 951219.57 frames. ], batch size: 33, lr: 3.34e-03, grad_scale: 32.0 2023-04-27 11:13:56,977 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=99741.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:14:08,292 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-04-27 11:14:08,590 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.1486, 2.8419, 0.9631, 1.4934, 2.1880, 1.3124, 3.5312, 1.9198], device='cuda:1'), covar=tensor([0.0659, 0.0610, 0.0847, 0.1201, 0.0476, 0.0919, 0.0231, 0.0553], device='cuda:1'), in_proj_covar=tensor([0.0051, 0.0065, 0.0048, 0.0046, 0.0049, 0.0052, 0.0073, 0.0051], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0007, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:1') 2023-04-27 11:14:16,632 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.069e+02 1.602e+02 1.875e+02 2.250e+02 5.277e+02, threshold=3.750e+02, percent-clipped=1.0 2023-04-27 11:14:17,850 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=99755.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:14:33,758 INFO [finetune.py:976] (1/7) Epoch 18, batch 2400, loss[loss=0.1383, simple_loss=0.2146, pruned_loss=0.03095, over 4755.00 frames. ], tot_loss[loss=0.173, simple_loss=0.2433, pruned_loss=0.05138, over 952945.45 frames. ], batch size: 28, lr: 3.34e-03, grad_scale: 32.0 2023-04-27 11:14:36,880 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=99776.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:15:08,014 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 2023-04-27 11:15:25,045 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.3902, 1.5624, 1.3789, 1.5485, 1.3690, 1.3265, 1.4337, 1.0853], device='cuda:1'), covar=tensor([0.1526, 0.1238, 0.0914, 0.1092, 0.3513, 0.1106, 0.1601, 0.2089], device='cuda:1'), in_proj_covar=tensor([0.0288, 0.0307, 0.0220, 0.0281, 0.0313, 0.0261, 0.0250, 0.0267], device='cuda:1'), out_proj_covar=tensor([1.1557e-04, 1.2186e-04, 8.7393e-05, 1.1169e-04, 1.2714e-04, 1.0357e-04, 1.0115e-04, 1.0609e-04], device='cuda:1') 2023-04-27 11:15:27,489 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.3466, 2.0281, 2.3627, 2.8402, 2.7385, 2.2519, 2.0020, 2.4511], device='cuda:1'), covar=tensor([0.0695, 0.0985, 0.0556, 0.0410, 0.0519, 0.0778, 0.0681, 0.0498], device='cuda:1'), in_proj_covar=tensor([0.0188, 0.0201, 0.0182, 0.0171, 0.0178, 0.0181, 0.0151, 0.0179], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 11:15:35,837 INFO [finetune.py:976] (1/7) Epoch 18, batch 2450, loss[loss=0.1805, simple_loss=0.2528, pruned_loss=0.05409, over 4909.00 frames. ], tot_loss[loss=0.1702, simple_loss=0.2402, pruned_loss=0.05014, over 954247.41 frames. ], batch size: 36, lr: 3.34e-03, grad_scale: 64.0 2023-04-27 11:15:37,156 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.1219, 2.5197, 1.0427, 1.4807, 2.0926, 1.3244, 3.4419, 1.8216], device='cuda:1'), covar=tensor([0.0670, 0.0755, 0.0843, 0.1263, 0.0501, 0.1008, 0.0218, 0.0625], device='cuda:1'), in_proj_covar=tensor([0.0051, 0.0065, 0.0048, 0.0046, 0.0049, 0.0052, 0.0073, 0.0051], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0007, 0.0008, 0.0008, 0.0010, 0.0007], device='cuda:1') 2023-04-27 11:15:37,732 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=99824.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 11:16:08,394 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.453e+01 1.597e+02 1.870e+02 2.238e+02 4.750e+02, threshold=3.741e+02, percent-clipped=1.0 2023-04-27 11:16:19,143 INFO [finetune.py:976] (1/7) Epoch 18, batch 2500, loss[loss=0.1948, simple_loss=0.2664, pruned_loss=0.06165, over 4817.00 frames. ], tot_loss[loss=0.1724, simple_loss=0.2422, pruned_loss=0.05126, over 951596.63 frames. ], batch size: 40, lr: 3.34e-03, grad_scale: 64.0 2023-04-27 11:16:28,669 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=99885.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:16:36,190 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([0.9767, 1.0122, 3.0154, 2.5684, 2.6470, 2.8583, 2.8265, 2.5363], device='cuda:1'), covar=tensor([0.9841, 0.8388, 0.2982, 0.4643, 0.3375, 0.3897, 0.6298, 0.4105], device='cuda:1'), in_proj_covar=tensor([0.0310, 0.0310, 0.0407, 0.0410, 0.0352, 0.0407, 0.0314, 0.0369], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-27 11:16:52,960 INFO [finetune.py:976] (1/7) Epoch 18, batch 2550, loss[loss=0.1717, simple_loss=0.2559, pruned_loss=0.0437, over 4913.00 frames. ], tot_loss[loss=0.1762, simple_loss=0.2468, pruned_loss=0.05283, over 952117.03 frames. ], batch size: 35, lr: 3.34e-03, grad_scale: 64.0 2023-04-27 11:16:54,876 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=99924.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:16:54,894 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=99924.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:17:00,810 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=99933.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:17:10,139 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=99945.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:17:14,383 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.4516, 3.1362, 2.6137, 2.9664, 2.2765, 2.7059, 2.9435, 2.0783], device='cuda:1'), covar=tensor([0.2161, 0.1445, 0.0838, 0.1148, 0.2905, 0.1390, 0.1881, 0.2917], device='cuda:1'), in_proj_covar=tensor([0.0284, 0.0304, 0.0217, 0.0279, 0.0311, 0.0258, 0.0248, 0.0264], device='cuda:1'), out_proj_covar=tensor([1.1418e-04, 1.2087e-04, 8.6411e-05, 1.1067e-04, 1.2626e-04, 1.0223e-04, 1.0011e-04, 1.0494e-04], device='cuda:1') 2023-04-27 11:17:16,096 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.051e+02 1.590e+02 1.888e+02 2.384e+02 3.870e+02, threshold=3.776e+02, percent-clipped=2.0 2023-04-27 11:17:22,954 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8555, 2.3292, 1.8909, 1.7143, 1.3719, 1.3886, 2.0022, 1.3043], device='cuda:1'), covar=tensor([0.1882, 0.1581, 0.1462, 0.1896, 0.2403, 0.2065, 0.1006, 0.2163], device='cuda:1'), in_proj_covar=tensor([0.0198, 0.0212, 0.0168, 0.0205, 0.0200, 0.0184, 0.0157, 0.0187], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-27 11:17:26,891 INFO [finetune.py:976] (1/7) Epoch 18, batch 2600, loss[loss=0.1764, simple_loss=0.2501, pruned_loss=0.05139, over 4811.00 frames. ], tot_loss[loss=0.1764, simple_loss=0.248, pruned_loss=0.05239, over 952515.28 frames. ], batch size: 51, lr: 3.34e-03, grad_scale: 64.0 2023-04-27 11:17:27,553 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=99972.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:17:32,465 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=99980.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:17:52,662 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=100007.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 11:18:01,562 INFO [finetune.py:976] (1/7) Epoch 18, batch 2650, loss[loss=0.1226, simple_loss=0.1963, pruned_loss=0.02445, over 4756.00 frames. ], tot_loss[loss=0.1766, simple_loss=0.2482, pruned_loss=0.05249, over 951314.10 frames. ], batch size: 23, lr: 3.34e-03, grad_scale: 64.0 2023-04-27 11:18:05,840 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=100028.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:18:10,666 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=100036.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:18:18,721 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.3574, 1.9863, 2.1767, 2.8122, 2.4460, 2.2858, 1.7679, 2.4040], device='cuda:1'), covar=tensor([0.0815, 0.1219, 0.0696, 0.0505, 0.0655, 0.0825, 0.0865, 0.0577], device='cuda:1'), in_proj_covar=tensor([0.0190, 0.0203, 0.0184, 0.0173, 0.0180, 0.0183, 0.0153, 0.0180], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 11:18:24,399 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.122e+02 1.666e+02 1.842e+02 2.203e+02 3.197e+02, threshold=3.685e+02, percent-clipped=0.0 2023-04-27 11:18:33,084 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=100068.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 11:18:35,203 INFO [finetune.py:976] (1/7) Epoch 18, batch 2700, loss[loss=0.1679, simple_loss=0.2454, pruned_loss=0.04516, over 4890.00 frames. ], tot_loss[loss=0.1759, simple_loss=0.2477, pruned_loss=0.0521, over 952663.13 frames. ], batch size: 32, lr: 3.34e-03, grad_scale: 64.0 2023-04-27 11:18:38,369 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=100076.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:18:40,952 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=3.95 vs. limit=5.0 2023-04-27 11:19:03,524 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.76 vs. limit=5.0 2023-04-27 11:19:37,243 INFO [finetune.py:976] (1/7) Epoch 18, batch 2750, loss[loss=0.1774, simple_loss=0.2336, pruned_loss=0.06067, over 4870.00 frames. ], tot_loss[loss=0.1741, simple_loss=0.2448, pruned_loss=0.05172, over 953197.64 frames. ], batch size: 31, lr: 3.34e-03, grad_scale: 64.0 2023-04-27 11:19:39,079 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=100124.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:19:39,127 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=100124.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 11:20:11,925 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.977e+01 1.603e+02 1.951e+02 2.448e+02 3.827e+02, threshold=3.902e+02, percent-clipped=2.0 2023-04-27 11:20:28,162 INFO [finetune.py:976] (1/7) Epoch 18, batch 2800, loss[loss=0.1502, simple_loss=0.2216, pruned_loss=0.03936, over 4049.00 frames. ], tot_loss[loss=0.1716, simple_loss=0.2418, pruned_loss=0.05072, over 954062.80 frames. ], batch size: 17, lr: 3.34e-03, grad_scale: 32.0 2023-04-27 11:20:34,234 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=100172.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 11:20:48,159 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-04-27 11:20:59,036 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.2346, 2.3117, 2.1808, 1.9915, 2.4768, 2.0972, 3.0777, 1.8909], device='cuda:1'), covar=tensor([0.3572, 0.1710, 0.4022, 0.2908, 0.1495, 0.2374, 0.1165, 0.3851], device='cuda:1'), in_proj_covar=tensor([0.0341, 0.0350, 0.0431, 0.0356, 0.0385, 0.0382, 0.0374, 0.0423], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 11:21:07,287 INFO [finetune.py:976] (1/7) Epoch 18, batch 2850, loss[loss=0.1695, simple_loss=0.2383, pruned_loss=0.0504, over 4760.00 frames. ], tot_loss[loss=0.1714, simple_loss=0.241, pruned_loss=0.05087, over 953265.68 frames. ], batch size: 27, lr: 3.34e-03, grad_scale: 32.0 2023-04-27 11:21:09,239 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=100224.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:21:29,439 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=100238.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:21:39,948 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=100245.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:21:52,184 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.041e+02 1.601e+02 1.939e+02 2.362e+02 3.802e+02, threshold=3.878e+02, percent-clipped=0.0 2023-04-27 11:22:14,793 INFO [finetune.py:976] (1/7) Epoch 18, batch 2900, loss[loss=0.1514, simple_loss=0.2342, pruned_loss=0.03428, over 4819.00 frames. ], tot_loss[loss=0.1748, simple_loss=0.2453, pruned_loss=0.05217, over 953414.41 frames. ], batch size: 45, lr: 3.34e-03, grad_scale: 32.0 2023-04-27 11:22:15,465 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=100272.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:22:28,729 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=100293.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:22:32,440 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=100299.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:22:37,822 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.11 vs. limit=2.0 2023-04-27 11:22:48,676 INFO [finetune.py:976] (1/7) Epoch 18, batch 2950, loss[loss=0.1508, simple_loss=0.2195, pruned_loss=0.04108, over 4301.00 frames. ], tot_loss[loss=0.1762, simple_loss=0.248, pruned_loss=0.05218, over 954283.55 frames. ], batch size: 65, lr: 3.34e-03, grad_scale: 32.0 2023-04-27 11:22:54,835 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.3215, 3.1679, 0.9575, 1.9179, 1.7363, 2.3402, 1.8625, 1.1673], device='cuda:1'), covar=tensor([0.1407, 0.0929, 0.1862, 0.1167, 0.1161, 0.0941, 0.1384, 0.1944], device='cuda:1'), in_proj_covar=tensor([0.0116, 0.0242, 0.0136, 0.0120, 0.0132, 0.0152, 0.0117, 0.0119], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-27 11:22:57,898 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=100336.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:23:01,193 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.44 vs. limit=5.0 2023-04-27 11:23:09,851 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.143e+02 1.535e+02 1.848e+02 2.418e+02 4.913e+02, threshold=3.697e+02, percent-clipped=1.0 2023-04-27 11:23:15,825 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=100363.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 11:23:21,637 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-04-27 11:23:22,041 INFO [finetune.py:976] (1/7) Epoch 18, batch 3000, loss[loss=0.1724, simple_loss=0.2637, pruned_loss=0.04054, over 4848.00 frames. ], tot_loss[loss=0.1763, simple_loss=0.2481, pruned_loss=0.05225, over 954220.18 frames. ], batch size: 44, lr: 3.34e-03, grad_scale: 32.0 2023-04-27 11:23:22,041 INFO [finetune.py:1001] (1/7) Computing validation loss 2023-04-27 11:23:32,627 INFO [finetune.py:1010] (1/7) Epoch 18, validation: loss=0.1524, simple_loss=0.2231, pruned_loss=0.04086, over 2265189.00 frames. 2023-04-27 11:23:32,628 INFO [finetune.py:1011] (1/7) Maximum memory allocated so far is 6391MB 2023-04-27 11:23:41,187 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=100384.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:23:52,511 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=100402.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:23:56,783 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.10 vs. limit=2.0 2023-04-27 11:24:04,563 INFO [finetune.py:976] (1/7) Epoch 18, batch 3050, loss[loss=0.1722, simple_loss=0.2516, pruned_loss=0.04638, over 4898.00 frames. ], tot_loss[loss=0.1764, simple_loss=0.2483, pruned_loss=0.05226, over 954722.18 frames. ], batch size: 43, lr: 3.34e-03, grad_scale: 32.0 2023-04-27 11:24:43,546 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.094e+02 1.525e+02 1.858e+02 2.098e+02 3.467e+02, threshold=3.716e+02, percent-clipped=0.0 2023-04-27 11:24:55,032 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=100463.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 11:24:59,807 INFO [finetune.py:976] (1/7) Epoch 18, batch 3100, loss[loss=0.1608, simple_loss=0.2304, pruned_loss=0.0456, over 4754.00 frames. ], tot_loss[loss=0.1748, simple_loss=0.2465, pruned_loss=0.05156, over 956139.54 frames. ], batch size: 28, lr: 3.34e-03, grad_scale: 32.0 2023-04-27 11:25:39,581 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-04-27 11:26:02,765 INFO [finetune.py:976] (1/7) Epoch 18, batch 3150, loss[loss=0.1514, simple_loss=0.2384, pruned_loss=0.0322, over 4795.00 frames. ], tot_loss[loss=0.1733, simple_loss=0.2444, pruned_loss=0.05112, over 956238.90 frames. ], batch size: 29, lr: 3.34e-03, grad_scale: 32.0 2023-04-27 11:26:17,410 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9843, 1.4659, 1.7924, 1.7470, 1.7294, 1.4608, 0.8654, 1.3563], device='cuda:1'), covar=tensor([0.3246, 0.3168, 0.1608, 0.2098, 0.2592, 0.2553, 0.4010, 0.2145], device='cuda:1'), in_proj_covar=tensor([0.0288, 0.0245, 0.0225, 0.0314, 0.0218, 0.0230, 0.0227, 0.0184], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-27 11:26:34,116 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5124, 1.6445, 1.6057, 1.8790, 1.6600, 2.0935, 1.4781, 3.7043], device='cuda:1'), covar=tensor([0.0575, 0.0834, 0.0809, 0.1156, 0.0675, 0.0474, 0.0762, 0.0148], device='cuda:1'), in_proj_covar=tensor([0.0037, 0.0039, 0.0040, 0.0043, 0.0040, 0.0038, 0.0038, 0.0056], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:1') 2023-04-27 11:26:37,705 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.912e+01 1.520e+02 1.805e+02 2.268e+02 8.838e+02, threshold=3.610e+02, percent-clipped=4.0 2023-04-27 11:26:59,498 INFO [finetune.py:976] (1/7) Epoch 18, batch 3200, loss[loss=0.1809, simple_loss=0.2478, pruned_loss=0.05699, over 4769.00 frames. ], tot_loss[loss=0.1729, simple_loss=0.2433, pruned_loss=0.05124, over 957667.53 frames. ], batch size: 28, lr: 3.33e-03, grad_scale: 32.0 2023-04-27 11:27:33,044 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=100594.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:28:06,109 INFO [finetune.py:976] (1/7) Epoch 18, batch 3250, loss[loss=0.2075, simple_loss=0.2759, pruned_loss=0.06958, over 4866.00 frames. ], tot_loss[loss=0.1735, simple_loss=0.2436, pruned_loss=0.0517, over 956094.68 frames. ], batch size: 44, lr: 3.33e-03, grad_scale: 32.0 2023-04-27 11:28:08,110 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.3298, 2.0352, 2.3017, 2.9614, 2.8407, 2.3728, 2.0401, 2.5752], device='cuda:1'), covar=tensor([0.0934, 0.1083, 0.0703, 0.0561, 0.0570, 0.0844, 0.0701, 0.0558], device='cuda:1'), in_proj_covar=tensor([0.0189, 0.0201, 0.0182, 0.0172, 0.0178, 0.0181, 0.0151, 0.0179], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 11:28:30,300 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.424e+01 1.529e+02 1.873e+02 2.219e+02 4.824e+02, threshold=3.746e+02, percent-clipped=4.0 2023-04-27 11:28:35,348 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=100663.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 11:28:40,634 INFO [finetune.py:976] (1/7) Epoch 18, batch 3300, loss[loss=0.1937, simple_loss=0.2745, pruned_loss=0.05649, over 4815.00 frames. ], tot_loss[loss=0.1763, simple_loss=0.2475, pruned_loss=0.0526, over 956201.14 frames. ], batch size: 38, lr: 3.33e-03, grad_scale: 32.0 2023-04-27 11:29:07,773 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=100711.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 11:29:13,775 INFO [finetune.py:976] (1/7) Epoch 18, batch 3350, loss[loss=0.1433, simple_loss=0.2182, pruned_loss=0.03419, over 4742.00 frames. ], tot_loss[loss=0.1768, simple_loss=0.2484, pruned_loss=0.05262, over 955839.70 frames. ], batch size: 27, lr: 3.33e-03, grad_scale: 32.0 2023-04-27 11:29:15,838 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-04-27 11:29:19,898 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=100730.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 11:29:37,231 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.219e+01 1.703e+02 2.119e+02 2.646e+02 1.102e+03, threshold=4.237e+02, percent-clipped=5.0 2023-04-27 11:29:39,117 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=100758.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 11:29:44,126 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8787, 1.1012, 1.5850, 1.7111, 1.6759, 1.7027, 1.5985, 1.5922], device='cuda:1'), covar=tensor([0.3998, 0.5228, 0.4582, 0.4600, 0.5444, 0.7294, 0.4909, 0.4801], device='cuda:1'), in_proj_covar=tensor([0.0330, 0.0372, 0.0318, 0.0332, 0.0342, 0.0392, 0.0354, 0.0326], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 11:29:47,005 INFO [finetune.py:976] (1/7) Epoch 18, batch 3400, loss[loss=0.1449, simple_loss=0.2263, pruned_loss=0.03181, over 4752.00 frames. ], tot_loss[loss=0.1779, simple_loss=0.2494, pruned_loss=0.05326, over 953821.68 frames. ], batch size: 27, lr: 3.33e-03, grad_scale: 32.0 2023-04-27 11:29:58,644 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([4.1503, 4.0226, 2.8663, 4.7795, 4.1492, 4.1609, 1.7507, 4.1246], device='cuda:1'), covar=tensor([0.1514, 0.1315, 0.2878, 0.1297, 0.3200, 0.1817, 0.5635, 0.2006], device='cuda:1'), in_proj_covar=tensor([0.0243, 0.0213, 0.0248, 0.0302, 0.0294, 0.0247, 0.0269, 0.0270], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 11:30:00,848 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=100791.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 11:30:20,268 INFO [finetune.py:976] (1/7) Epoch 18, batch 3450, loss[loss=0.1847, simple_loss=0.25, pruned_loss=0.05971, over 4891.00 frames. ], tot_loss[loss=0.1773, simple_loss=0.2492, pruned_loss=0.05265, over 954251.17 frames. ], batch size: 35, lr: 3.33e-03, grad_scale: 32.0 2023-04-27 11:30:21,005 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.2118, 2.0834, 2.4261, 2.6608, 1.9791, 1.7585, 2.1153, 1.0532], device='cuda:1'), covar=tensor([0.0479, 0.0905, 0.0439, 0.0692, 0.0672, 0.1118, 0.0709, 0.0870], device='cuda:1'), in_proj_covar=tensor([0.0067, 0.0068, 0.0067, 0.0067, 0.0074, 0.0095, 0.0073, 0.0066], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-27 11:30:54,161 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.079e+02 1.615e+02 1.948e+02 2.378e+02 4.172e+02, threshold=3.896e+02, percent-clipped=0.0 2023-04-27 11:31:08,674 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=3.28 vs. limit=5.0 2023-04-27 11:31:09,716 INFO [finetune.py:976] (1/7) Epoch 18, batch 3500, loss[loss=0.1626, simple_loss=0.2299, pruned_loss=0.04769, over 4824.00 frames. ], tot_loss[loss=0.1756, simple_loss=0.2469, pruned_loss=0.05216, over 956177.74 frames. ], batch size: 39, lr: 3.33e-03, grad_scale: 32.0 2023-04-27 11:31:30,615 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=100894.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:31:49,439 INFO [finetune.py:976] (1/7) Epoch 18, batch 3550, loss[loss=0.1629, simple_loss=0.2328, pruned_loss=0.04647, over 4902.00 frames. ], tot_loss[loss=0.1744, simple_loss=0.245, pruned_loss=0.05196, over 956294.77 frames. ], batch size: 35, lr: 3.33e-03, grad_scale: 32.0 2023-04-27 11:31:55,149 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.48 vs. limit=2.0 2023-04-27 11:32:02,731 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=100942.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:32:23,486 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.154e+02 1.524e+02 1.758e+02 2.158e+02 5.191e+02, threshold=3.516e+02, percent-clipped=1.0 2023-04-27 11:32:39,051 INFO [finetune.py:976] (1/7) Epoch 18, batch 3600, loss[loss=0.1383, simple_loss=0.2077, pruned_loss=0.03441, over 4811.00 frames. ], tot_loss[loss=0.1725, simple_loss=0.2423, pruned_loss=0.05129, over 956858.14 frames. ], batch size: 25, lr: 3.33e-03, grad_scale: 32.0 2023-04-27 11:33:07,696 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.68 vs. limit=2.0 2023-04-27 11:33:40,662 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4606, 1.3159, 4.1757, 3.8565, 3.6133, 3.9600, 3.8915, 3.6808], device='cuda:1'), covar=tensor([0.7424, 0.6178, 0.1196, 0.2166, 0.1326, 0.1749, 0.1577, 0.1562], device='cuda:1'), in_proj_covar=tensor([0.0309, 0.0307, 0.0405, 0.0409, 0.0351, 0.0405, 0.0313, 0.0367], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-27 11:33:50,171 INFO [finetune.py:976] (1/7) Epoch 18, batch 3650, loss[loss=0.1931, simple_loss=0.2696, pruned_loss=0.05833, over 4845.00 frames. ], tot_loss[loss=0.1743, simple_loss=0.2443, pruned_loss=0.05211, over 957129.33 frames. ], batch size: 49, lr: 3.33e-03, grad_scale: 32.0 2023-04-27 11:34:31,992 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.015e+02 1.646e+02 1.981e+02 2.447e+02 4.488e+02, threshold=3.961e+02, percent-clipped=2.0 2023-04-27 11:34:34,872 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=101058.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:34:48,424 INFO [finetune.py:976] (1/7) Epoch 18, batch 3700, loss[loss=0.1796, simple_loss=0.2695, pruned_loss=0.04488, over 4817.00 frames. ], tot_loss[loss=0.1779, simple_loss=0.2483, pruned_loss=0.05373, over 955649.16 frames. ], batch size: 45, lr: 3.33e-03, grad_scale: 32.0 2023-04-27 11:34:57,618 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=101086.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 11:35:05,550 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6121, 2.0605, 1.6856, 1.9250, 1.4379, 1.7081, 1.6063, 1.3805], device='cuda:1'), covar=tensor([0.1831, 0.1150, 0.0814, 0.1126, 0.3236, 0.1093, 0.1935, 0.2330], device='cuda:1'), in_proj_covar=tensor([0.0288, 0.0306, 0.0220, 0.0281, 0.0313, 0.0262, 0.0251, 0.0268], device='cuda:1'), out_proj_covar=tensor([1.1560e-04, 1.2166e-04, 8.7281e-05, 1.1155e-04, 1.2719e-04, 1.0380e-04, 1.0166e-04, 1.0650e-04], device='cuda:1') 2023-04-27 11:35:10,799 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=101106.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:35:11,915 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=101107.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:35:15,553 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9228, 1.6756, 2.1218, 2.3419, 1.9946, 1.8318, 1.9797, 1.9621], device='cuda:1'), covar=tensor([0.4555, 0.6545, 0.6589, 0.5539, 0.5986, 0.8705, 0.8590, 0.9619], device='cuda:1'), in_proj_covar=tensor([0.0420, 0.0405, 0.0497, 0.0501, 0.0451, 0.0476, 0.0482, 0.0486], device='cuda:1'), out_proj_covar=tensor([1.0125e-04, 9.9481e-05, 1.1168e-04, 1.1942e-04, 1.0804e-04, 1.1413e-04, 1.1437e-04, 1.1466e-04], device='cuda:1') 2023-04-27 11:35:21,745 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.1702, 4.6084, 0.8635, 2.4618, 2.6763, 2.9523, 2.7185, 0.9420], device='cuda:1'), covar=tensor([0.1349, 0.1019, 0.2334, 0.1232, 0.1045, 0.1062, 0.1396, 0.2207], device='cuda:1'), in_proj_covar=tensor([0.0117, 0.0244, 0.0137, 0.0121, 0.0132, 0.0152, 0.0117, 0.0120], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-27 11:35:22,248 INFO [finetune.py:976] (1/7) Epoch 18, batch 3750, loss[loss=0.1278, simple_loss=0.2012, pruned_loss=0.02722, over 4744.00 frames. ], tot_loss[loss=0.1802, simple_loss=0.251, pruned_loss=0.05468, over 958428.71 frames. ], batch size: 23, lr: 3.33e-03, grad_scale: 32.0 2023-04-27 11:35:43,440 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.081e+02 1.572e+02 1.895e+02 2.251e+02 5.219e+02, threshold=3.790e+02, percent-clipped=2.0 2023-04-27 11:35:54,451 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=101168.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:35:56,188 INFO [finetune.py:976] (1/7) Epoch 18, batch 3800, loss[loss=0.1585, simple_loss=0.2333, pruned_loss=0.04185, over 4752.00 frames. ], tot_loss[loss=0.1819, simple_loss=0.2528, pruned_loss=0.0555, over 957448.13 frames. ], batch size: 27, lr: 3.33e-03, grad_scale: 32.0 2023-04-27 11:36:30,056 INFO [finetune.py:976] (1/7) Epoch 18, batch 3850, loss[loss=0.173, simple_loss=0.2396, pruned_loss=0.05318, over 4812.00 frames. ], tot_loss[loss=0.1793, simple_loss=0.2505, pruned_loss=0.0541, over 956321.59 frames. ], batch size: 41, lr: 3.33e-03, grad_scale: 32.0 2023-04-27 11:36:45,927 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.0266, 1.0484, 1.2129, 1.1893, 1.0123, 0.9416, 0.9553, 0.5533], device='cuda:1'), covar=tensor([0.0611, 0.0601, 0.0493, 0.0497, 0.0737, 0.1357, 0.0487, 0.0706], device='cuda:1'), in_proj_covar=tensor([0.0067, 0.0069, 0.0068, 0.0067, 0.0074, 0.0095, 0.0073, 0.0066], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-27 11:36:50,613 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.073e+02 1.513e+02 1.818e+02 2.217e+02 6.339e+02, threshold=3.636e+02, percent-clipped=4.0 2023-04-27 11:37:01,722 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6785, 1.4768, 1.8585, 1.9201, 1.4960, 1.3587, 1.5638, 1.0319], device='cuda:1'), covar=tensor([0.0459, 0.0686, 0.0416, 0.0548, 0.0678, 0.1221, 0.0617, 0.0650], device='cuda:1'), in_proj_covar=tensor([0.0068, 0.0069, 0.0068, 0.0067, 0.0074, 0.0096, 0.0073, 0.0066], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-27 11:37:02,694 INFO [finetune.py:976] (1/7) Epoch 18, batch 3900, loss[loss=0.1517, simple_loss=0.2249, pruned_loss=0.03927, over 4835.00 frames. ], tot_loss[loss=0.1773, simple_loss=0.2478, pruned_loss=0.0534, over 956888.29 frames. ], batch size: 25, lr: 3.33e-03, grad_scale: 32.0 2023-04-27 11:37:35,479 INFO [finetune.py:976] (1/7) Epoch 18, batch 3950, loss[loss=0.1219, simple_loss=0.189, pruned_loss=0.02741, over 4742.00 frames. ], tot_loss[loss=0.1737, simple_loss=0.244, pruned_loss=0.0517, over 957414.89 frames. ], batch size: 23, lr: 3.33e-03, grad_scale: 32.0 2023-04-27 11:38:09,082 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.783e+01 1.521e+02 1.787e+02 2.150e+02 4.001e+02, threshold=3.574e+02, percent-clipped=1.0 2023-04-27 11:38:30,024 INFO [finetune.py:976] (1/7) Epoch 18, batch 4000, loss[loss=0.1843, simple_loss=0.2604, pruned_loss=0.05413, over 4900.00 frames. ], tot_loss[loss=0.1736, simple_loss=0.2437, pruned_loss=0.05179, over 957423.97 frames. ], batch size: 35, lr: 3.33e-03, grad_scale: 32.0 2023-04-27 11:38:30,169 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6948, 1.0323, 1.6685, 2.1598, 1.7968, 1.6590, 1.6914, 1.7014], device='cuda:1'), covar=tensor([0.4331, 0.6068, 0.5756, 0.5572, 0.5651, 0.7011, 0.7354, 0.7947], device='cuda:1'), in_proj_covar=tensor([0.0421, 0.0406, 0.0497, 0.0501, 0.0451, 0.0476, 0.0484, 0.0487], device='cuda:1'), out_proj_covar=tensor([1.0140e-04, 9.9766e-05, 1.1179e-04, 1.1951e-04, 1.0812e-04, 1.1422e-04, 1.1462e-04, 1.1492e-04], device='cuda:1') 2023-04-27 11:38:52,344 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=101386.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 11:39:35,325 INFO [finetune.py:976] (1/7) Epoch 18, batch 4050, loss[loss=0.1738, simple_loss=0.2548, pruned_loss=0.04642, over 4916.00 frames. ], tot_loss[loss=0.178, simple_loss=0.2485, pruned_loss=0.05373, over 958747.82 frames. ], batch size: 38, lr: 3.33e-03, grad_scale: 32.0 2023-04-27 11:39:43,280 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9390, 2.6582, 1.9948, 2.1087, 1.4526, 1.3982, 2.1081, 1.4041], device='cuda:1'), covar=tensor([0.1633, 0.1342, 0.1356, 0.1642, 0.2235, 0.1869, 0.0943, 0.1983], device='cuda:1'), in_proj_covar=tensor([0.0199, 0.0214, 0.0170, 0.0206, 0.0202, 0.0186, 0.0157, 0.0187], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-27 11:39:55,121 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=101434.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 11:40:07,832 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.179e+02 1.675e+02 1.988e+02 2.422e+02 4.320e+02, threshold=3.976e+02, percent-clipped=3.0 2023-04-27 11:40:12,808 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=101463.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:40:18,124 INFO [finetune.py:976] (1/7) Epoch 18, batch 4100, loss[loss=0.1823, simple_loss=0.2545, pruned_loss=0.05505, over 4902.00 frames. ], tot_loss[loss=0.1803, simple_loss=0.2514, pruned_loss=0.05457, over 958521.12 frames. ], batch size: 36, lr: 3.33e-03, grad_scale: 32.0 2023-04-27 11:40:36,951 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-04-27 11:40:51,379 INFO [finetune.py:976] (1/7) Epoch 18, batch 4150, loss[loss=0.1806, simple_loss=0.2479, pruned_loss=0.05664, over 4844.00 frames. ], tot_loss[loss=0.1801, simple_loss=0.251, pruned_loss=0.05459, over 957765.05 frames. ], batch size: 30, lr: 3.33e-03, grad_scale: 32.0 2023-04-27 11:41:14,404 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.437e+01 1.627e+02 2.025e+02 2.401e+02 3.721e+02, threshold=4.051e+02, percent-clipped=0.0 2023-04-27 11:41:19,546 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.11 vs. limit=2.0 2023-04-27 11:41:24,217 INFO [finetune.py:976] (1/7) Epoch 18, batch 4200, loss[loss=0.1619, simple_loss=0.2297, pruned_loss=0.04704, over 4817.00 frames. ], tot_loss[loss=0.1796, simple_loss=0.2502, pruned_loss=0.05448, over 954826.76 frames. ], batch size: 33, lr: 3.33e-03, grad_scale: 32.0 2023-04-27 11:41:55,250 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.53 vs. limit=5.0 2023-04-27 11:41:58,014 INFO [finetune.py:976] (1/7) Epoch 18, batch 4250, loss[loss=0.1537, simple_loss=0.2359, pruned_loss=0.03573, over 4833.00 frames. ], tot_loss[loss=0.1757, simple_loss=0.2464, pruned_loss=0.05245, over 955266.57 frames. ], batch size: 49, lr: 3.33e-03, grad_scale: 32.0 2023-04-27 11:42:21,983 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.254e+01 1.472e+02 1.774e+02 2.178e+02 3.033e+02, threshold=3.548e+02, percent-clipped=0.0 2023-04-27 11:42:31,623 INFO [finetune.py:976] (1/7) Epoch 18, batch 4300, loss[loss=0.1525, simple_loss=0.2199, pruned_loss=0.04253, over 4825.00 frames. ], tot_loss[loss=0.1745, simple_loss=0.2446, pruned_loss=0.05218, over 955341.84 frames. ], batch size: 30, lr: 3.33e-03, grad_scale: 32.0 2023-04-27 11:42:40,118 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=101684.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:42:50,114 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0561, 1.4877, 1.5229, 1.7702, 2.2096, 1.7462, 1.4712, 1.5028], device='cuda:1'), covar=tensor([0.1428, 0.1755, 0.1989, 0.1407, 0.1001, 0.1969, 0.2444, 0.2198], device='cuda:1'), in_proj_covar=tensor([0.0309, 0.0314, 0.0352, 0.0290, 0.0331, 0.0309, 0.0302, 0.0370], device='cuda:1'), out_proj_covar=tensor([6.3879e-05, 6.5204e-05, 7.4800e-05, 5.8740e-05, 6.8732e-05, 6.4940e-05, 6.3415e-05, 7.8772e-05], device='cuda:1') 2023-04-27 11:42:56,698 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.3554, 1.7377, 1.5963, 2.1413, 1.9256, 2.3140, 1.6526, 4.4826], device='cuda:1'), covar=tensor([0.0550, 0.0811, 0.0818, 0.1125, 0.0627, 0.0461, 0.0720, 0.0102], device='cuda:1'), in_proj_covar=tensor([0.0037, 0.0038, 0.0039, 0.0043, 0.0039, 0.0037, 0.0038, 0.0055], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0014, 0.0013, 0.0012, 0.0013, 0.0016], device='cuda:1') 2023-04-27 11:43:00,522 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.53 vs. limit=2.0 2023-04-27 11:43:03,998 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8992, 1.6606, 1.8633, 2.2489, 2.2988, 1.8681, 1.5414, 2.0241], device='cuda:1'), covar=tensor([0.0873, 0.1145, 0.0792, 0.0558, 0.0554, 0.0885, 0.0763, 0.0586], device='cuda:1'), in_proj_covar=tensor([0.0189, 0.0201, 0.0183, 0.0171, 0.0177, 0.0181, 0.0150, 0.0179], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 11:43:04,507 INFO [finetune.py:976] (1/7) Epoch 18, batch 4350, loss[loss=0.1969, simple_loss=0.2561, pruned_loss=0.06882, over 4825.00 frames. ], tot_loss[loss=0.1713, simple_loss=0.2414, pruned_loss=0.0506, over 956526.14 frames. ], batch size: 30, lr: 3.33e-03, grad_scale: 32.0 2023-04-27 11:43:20,706 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=101745.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:43:32,959 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 8.731e+01 1.464e+02 1.839e+02 2.155e+02 7.335e+02, threshold=3.679e+02, percent-clipped=2.0 2023-04-27 11:43:43,340 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=101763.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:43:48,147 INFO [finetune.py:976] (1/7) Epoch 18, batch 4400, loss[loss=0.1509, simple_loss=0.2212, pruned_loss=0.04033, over 4761.00 frames. ], tot_loss[loss=0.1729, simple_loss=0.2431, pruned_loss=0.0514, over 954092.80 frames. ], batch size: 54, lr: 3.32e-03, grad_scale: 32.0 2023-04-27 11:44:23,367 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.74 vs. limit=2.0 2023-04-27 11:44:37,288 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=101811.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:44:49,150 INFO [finetune.py:976] (1/7) Epoch 18, batch 4450, loss[loss=0.1506, simple_loss=0.2321, pruned_loss=0.0346, over 4758.00 frames. ], tot_loss[loss=0.1759, simple_loss=0.2471, pruned_loss=0.0524, over 954835.16 frames. ], batch size: 59, lr: 3.32e-03, grad_scale: 32.0 2023-04-27 11:44:58,811 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0288, 2.5514, 1.0366, 1.4629, 1.9326, 1.2593, 3.4829, 1.7465], device='cuda:1'), covar=tensor([0.0682, 0.0623, 0.0822, 0.1283, 0.0556, 0.1029, 0.0228, 0.0657], device='cuda:1'), in_proj_covar=tensor([0.0051, 0.0064, 0.0047, 0.0046, 0.0049, 0.0051, 0.0073, 0.0050], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0009, 0.0007, 0.0007, 0.0008, 0.0008, 0.0010, 0.0007], device='cuda:1') 2023-04-27 11:45:19,939 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([3.2905, 3.2895, 2.7990, 3.8921, 3.2488, 3.2828, 1.6618, 3.3148], device='cuda:1'), covar=tensor([0.1989, 0.1380, 0.4163, 0.1864, 0.2743, 0.2072, 0.5402, 0.2576], device='cuda:1'), in_proj_covar=tensor([0.0245, 0.0215, 0.0251, 0.0305, 0.0299, 0.0250, 0.0274, 0.0273], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 11:45:32,213 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.150e+02 1.703e+02 1.979e+02 2.478e+02 5.839e+02, threshold=3.957e+02, percent-clipped=3.0 2023-04-27 11:45:42,461 INFO [finetune.py:976] (1/7) Epoch 18, batch 4500, loss[loss=0.1856, simple_loss=0.2525, pruned_loss=0.05932, over 4751.00 frames. ], tot_loss[loss=0.1784, simple_loss=0.2498, pruned_loss=0.05347, over 956548.59 frames. ], batch size: 27, lr: 3.32e-03, grad_scale: 32.0 2023-04-27 11:45:43,312 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-04-27 11:45:50,534 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.67 vs. limit=2.0 2023-04-27 11:45:58,512 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7096, 1.1665, 1.2156, 1.4705, 1.7864, 1.4118, 1.2833, 1.1970], device='cuda:1'), covar=tensor([0.1441, 0.1791, 0.1644, 0.1188, 0.0946, 0.1894, 0.1920, 0.2139], device='cuda:1'), in_proj_covar=tensor([0.0310, 0.0314, 0.0353, 0.0291, 0.0331, 0.0310, 0.0303, 0.0372], device='cuda:1'), out_proj_covar=tensor([6.3976e-05, 6.5353e-05, 7.4983e-05, 5.9083e-05, 6.8863e-05, 6.5064e-05, 6.3745e-05, 7.9258e-05], device='cuda:1') 2023-04-27 11:46:04,100 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-04-27 11:46:15,937 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=101920.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 11:46:16,436 INFO [finetune.py:976] (1/7) Epoch 18, batch 4550, loss[loss=0.1955, simple_loss=0.2651, pruned_loss=0.06298, over 4872.00 frames. ], tot_loss[loss=0.1793, simple_loss=0.2511, pruned_loss=0.05377, over 956223.19 frames. ], batch size: 34, lr: 3.32e-03, grad_scale: 32.0 2023-04-27 11:46:22,724 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.8258, 2.0404, 1.9203, 2.0848, 1.8748, 2.0071, 2.1033, 2.0063], device='cuda:1'), covar=tensor([0.4050, 0.6817, 0.5540, 0.4767, 0.6046, 0.7520, 0.6088, 0.6047], device='cuda:1'), in_proj_covar=tensor([0.0330, 0.0372, 0.0318, 0.0331, 0.0342, 0.0391, 0.0356, 0.0324], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 11:46:27,002 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7514, 1.2259, 1.8336, 2.1968, 1.8234, 1.7152, 1.7702, 1.7532], device='cuda:1'), covar=tensor([0.4764, 0.7224, 0.7271, 0.6321, 0.6306, 0.8205, 0.8502, 0.9315], device='cuda:1'), in_proj_covar=tensor([0.0423, 0.0407, 0.0499, 0.0503, 0.0452, 0.0477, 0.0484, 0.0489], device='cuda:1'), out_proj_covar=tensor([1.0185e-04, 9.9923e-05, 1.1221e-04, 1.1986e-04, 1.0824e-04, 1.1444e-04, 1.1477e-04, 1.1532e-04], device='cuda:1') 2023-04-27 11:46:38,530 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.069e+02 1.683e+02 1.915e+02 2.320e+02 4.613e+02, threshold=3.831e+02, percent-clipped=2.0 2023-04-27 11:46:49,874 INFO [finetune.py:976] (1/7) Epoch 18, batch 4600, loss[loss=0.1581, simple_loss=0.2388, pruned_loss=0.03876, over 4897.00 frames. ], tot_loss[loss=0.1793, simple_loss=0.251, pruned_loss=0.05383, over 956307.72 frames. ], batch size: 35, lr: 3.32e-03, grad_scale: 32.0 2023-04-27 11:46:51,873 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.1683, 1.9460, 2.3830, 2.4071, 2.1033, 2.0764, 2.1885, 2.1969], device='cuda:1'), covar=tensor([0.6504, 0.8472, 0.9349, 0.8984, 0.8153, 1.1541, 1.1408, 1.1973], device='cuda:1'), in_proj_covar=tensor([0.0423, 0.0407, 0.0500, 0.0503, 0.0452, 0.0478, 0.0485, 0.0489], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 11:46:56,093 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=101981.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 11:46:59,718 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=101987.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:47:05,308 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.45 vs. limit=5.0 2023-04-27 11:47:14,147 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.5098, 1.7695, 1.8347, 1.9463, 1.8298, 1.8771, 1.9510, 1.9199], device='cuda:1'), covar=tensor([0.4214, 0.5651, 0.5019, 0.4799, 0.5810, 0.7507, 0.5784, 0.5178], device='cuda:1'), in_proj_covar=tensor([0.0330, 0.0373, 0.0318, 0.0332, 0.0343, 0.0392, 0.0356, 0.0325], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 11:47:24,660 INFO [finetune.py:976] (1/7) Epoch 18, batch 4650, loss[loss=0.183, simple_loss=0.2476, pruned_loss=0.05922, over 4929.00 frames. ], tot_loss[loss=0.1773, simple_loss=0.2483, pruned_loss=0.05312, over 956642.81 frames. ], batch size: 33, lr: 3.32e-03, grad_scale: 32.0 2023-04-27 11:47:36,311 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=102040.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:47:41,223 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=102048.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:47:45,410 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.010e+02 1.596e+02 1.902e+02 2.192e+02 3.719e+02, threshold=3.804e+02, percent-clipped=0.0 2023-04-27 11:47:50,186 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-04-27 11:47:51,850 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.3675, 1.5235, 1.4949, 1.8104, 1.6233, 1.9306, 1.4644, 3.7162], device='cuda:1'), covar=tensor([0.0601, 0.0907, 0.0870, 0.1231, 0.0716, 0.0523, 0.0850, 0.0149], device='cuda:1'), in_proj_covar=tensor([0.0037, 0.0038, 0.0039, 0.0043, 0.0039, 0.0038, 0.0038, 0.0055], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0014, 0.0013, 0.0012, 0.0013, 0.0016], device='cuda:1') 2023-04-27 11:47:58,102 INFO [finetune.py:976] (1/7) Epoch 18, batch 4700, loss[loss=0.1769, simple_loss=0.243, pruned_loss=0.05536, over 4825.00 frames. ], tot_loss[loss=0.1744, simple_loss=0.2447, pruned_loss=0.05205, over 957185.86 frames. ], batch size: 39, lr: 3.32e-03, grad_scale: 32.0 2023-04-27 11:48:10,356 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5296, 1.2071, 4.1394, 3.8934, 3.5658, 3.8856, 3.7769, 3.6497], device='cuda:1'), covar=tensor([0.7299, 0.6081, 0.0991, 0.1541, 0.1194, 0.1599, 0.2363, 0.1442], device='cuda:1'), in_proj_covar=tensor([0.0309, 0.0308, 0.0406, 0.0407, 0.0350, 0.0406, 0.0314, 0.0367], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-27 11:48:30,318 INFO [finetune.py:976] (1/7) Epoch 18, batch 4750, loss[loss=0.1878, simple_loss=0.2541, pruned_loss=0.06069, over 4908.00 frames. ], tot_loss[loss=0.1727, simple_loss=0.2423, pruned_loss=0.0515, over 955619.09 frames. ], batch size: 43, lr: 3.32e-03, grad_scale: 32.0 2023-04-27 11:48:34,627 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.98 vs. limit=2.0 2023-04-27 11:48:51,431 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([0.9109, 1.2876, 3.3082, 3.0731, 2.9100, 3.2047, 3.1804, 2.8732], device='cuda:1'), covar=tensor([0.7251, 0.5398, 0.1541, 0.2237, 0.1530, 0.2422, 0.1543, 0.1818], device='cuda:1'), in_proj_covar=tensor([0.0307, 0.0307, 0.0404, 0.0406, 0.0349, 0.0405, 0.0313, 0.0365], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-27 11:48:51,943 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.089e+02 1.496e+02 1.838e+02 2.058e+02 4.267e+02, threshold=3.676e+02, percent-clipped=2.0 2023-04-27 11:49:09,300 INFO [finetune.py:976] (1/7) Epoch 18, batch 4800, loss[loss=0.2098, simple_loss=0.278, pruned_loss=0.07085, over 4773.00 frames. ], tot_loss[loss=0.1748, simple_loss=0.2452, pruned_loss=0.05223, over 954373.01 frames. ], batch size: 28, lr: 3.32e-03, grad_scale: 64.0 2023-04-27 11:50:15,445 INFO [finetune.py:976] (1/7) Epoch 18, batch 4850, loss[loss=0.1875, simple_loss=0.2628, pruned_loss=0.05607, over 4916.00 frames. ], tot_loss[loss=0.1776, simple_loss=0.2489, pruned_loss=0.05312, over 953769.97 frames. ], batch size: 38, lr: 3.32e-03, grad_scale: 64.0 2023-04-27 11:50:27,700 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.2030, 2.6136, 2.2693, 2.4684, 1.8723, 2.2637, 2.1935, 1.7399], device='cuda:1'), covar=tensor([0.1647, 0.1231, 0.0662, 0.1029, 0.3194, 0.1100, 0.1664, 0.2454], device='cuda:1'), in_proj_covar=tensor([0.0285, 0.0304, 0.0217, 0.0279, 0.0310, 0.0258, 0.0248, 0.0265], device='cuda:1'), out_proj_covar=tensor([1.1464e-04, 1.2096e-04, 8.6074e-05, 1.1083e-04, 1.2593e-04, 1.0237e-04, 1.0044e-04, 1.0502e-04], device='cuda:1') 2023-04-27 11:50:58,311 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.240e+02 1.715e+02 2.079e+02 2.399e+02 4.707e+02, threshold=4.157e+02, percent-clipped=1.0 2023-04-27 11:51:08,944 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.2379, 1.6307, 2.0699, 2.5693, 2.0535, 1.6052, 1.3575, 1.9113], device='cuda:1'), covar=tensor([0.2972, 0.3239, 0.1562, 0.2094, 0.2445, 0.2660, 0.4105, 0.1990], device='cuda:1'), in_proj_covar=tensor([0.0289, 0.0246, 0.0226, 0.0314, 0.0218, 0.0230, 0.0227, 0.0184], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-27 11:51:19,609 INFO [finetune.py:976] (1/7) Epoch 18, batch 4900, loss[loss=0.1747, simple_loss=0.242, pruned_loss=0.05372, over 4694.00 frames. ], tot_loss[loss=0.1799, simple_loss=0.2515, pruned_loss=0.05413, over 954364.14 frames. ], batch size: 23, lr: 3.32e-03, grad_scale: 64.0 2023-04-27 11:51:22,001 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([4.2480, 4.1390, 2.8693, 4.9157, 4.2883, 4.1763, 1.9574, 4.1853], device='cuda:1'), covar=tensor([0.1664, 0.1040, 0.4300, 0.1093, 0.2709, 0.1847, 0.5685, 0.2456], device='cuda:1'), in_proj_covar=tensor([0.0243, 0.0214, 0.0249, 0.0303, 0.0297, 0.0249, 0.0272, 0.0272], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 11:51:23,226 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=102276.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 11:51:55,246 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([4.1077, 3.8723, 2.9037, 4.7063, 3.9583, 4.0652, 1.7518, 4.0464], device='cuda:1'), covar=tensor([0.1609, 0.1038, 0.4120, 0.1037, 0.2275, 0.1613, 0.5540, 0.2331], device='cuda:1'), in_proj_covar=tensor([0.0244, 0.0215, 0.0250, 0.0304, 0.0298, 0.0250, 0.0272, 0.0273], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 11:52:15,983 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 2023-04-27 11:52:25,540 INFO [finetune.py:976] (1/7) Epoch 18, batch 4950, loss[loss=0.1531, simple_loss=0.2393, pruned_loss=0.03351, over 4776.00 frames. ], tot_loss[loss=0.1781, simple_loss=0.2502, pruned_loss=0.053, over 953828.67 frames. ], batch size: 29, lr: 3.32e-03, grad_scale: 64.0 2023-04-27 11:52:35,548 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=5.06 vs. limit=5.0 2023-04-27 11:52:50,483 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=102340.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:52:57,565 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=102343.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:53:04,800 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.066e+02 1.629e+02 1.929e+02 2.326e+02 4.533e+02, threshold=3.857e+02, percent-clipped=1.0 2023-04-27 11:53:14,474 INFO [finetune.py:976] (1/7) Epoch 18, batch 5000, loss[loss=0.1846, simple_loss=0.2477, pruned_loss=0.06073, over 4887.00 frames. ], tot_loss[loss=0.1771, simple_loss=0.2487, pruned_loss=0.05277, over 954903.60 frames. ], batch size: 35, lr: 3.32e-03, grad_scale: 64.0 2023-04-27 11:53:20,198 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7203, 1.3900, 1.3515, 1.5655, 1.9499, 1.5762, 1.3436, 1.2500], device='cuda:1'), covar=tensor([0.1356, 0.1294, 0.1574, 0.1208, 0.0662, 0.1345, 0.1894, 0.1947], device='cuda:1'), in_proj_covar=tensor([0.0308, 0.0313, 0.0353, 0.0291, 0.0331, 0.0310, 0.0303, 0.0371], device='cuda:1'), out_proj_covar=tensor([6.3538e-05, 6.4982e-05, 7.4860e-05, 5.9034e-05, 6.8879e-05, 6.5067e-05, 6.3652e-05, 7.9106e-05], device='cuda:1') 2023-04-27 11:53:28,308 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=102388.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:53:35,016 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([0.9313, 1.2910, 3.1057, 2.9231, 2.7942, 2.9939, 2.9915, 2.7514], device='cuda:1'), covar=tensor([0.6496, 0.4702, 0.1450, 0.1840, 0.1335, 0.1921, 0.1368, 0.1538], device='cuda:1'), in_proj_covar=tensor([0.0307, 0.0307, 0.0406, 0.0408, 0.0350, 0.0407, 0.0314, 0.0365], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-27 11:53:48,401 INFO [finetune.py:976] (1/7) Epoch 18, batch 5050, loss[loss=0.15, simple_loss=0.2258, pruned_loss=0.03713, over 4748.00 frames. ], tot_loss[loss=0.1742, simple_loss=0.2451, pruned_loss=0.05167, over 955497.23 frames. ], batch size: 27, lr: 3.32e-03, grad_scale: 64.0 2023-04-27 11:54:23,901 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.725e+01 1.630e+02 1.921e+02 2.242e+02 3.810e+02, threshold=3.842e+02, percent-clipped=0.0 2023-04-27 11:54:45,116 INFO [finetune.py:976] (1/7) Epoch 18, batch 5100, loss[loss=0.1254, simple_loss=0.1988, pruned_loss=0.02597, over 4789.00 frames. ], tot_loss[loss=0.1713, simple_loss=0.2418, pruned_loss=0.05039, over 956222.90 frames. ], batch size: 28, lr: 3.32e-03, grad_scale: 64.0 2023-04-27 11:54:46,470 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.0441, 1.4160, 1.3011, 1.6520, 1.5496, 1.5862, 1.3453, 2.4251], device='cuda:1'), covar=tensor([0.0641, 0.0800, 0.0840, 0.1210, 0.0652, 0.0489, 0.0755, 0.0208], device='cuda:1'), in_proj_covar=tensor([0.0037, 0.0038, 0.0039, 0.0043, 0.0040, 0.0038, 0.0038, 0.0055], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:1') 2023-04-27 11:54:46,527 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6784, 0.7358, 1.5271, 2.0623, 1.7167, 1.5893, 1.5978, 1.5757], device='cuda:1'), covar=tensor([0.4431, 0.5855, 0.6031, 0.5830, 0.5663, 0.6967, 0.7216, 0.7566], device='cuda:1'), in_proj_covar=tensor([0.0424, 0.0407, 0.0499, 0.0502, 0.0450, 0.0476, 0.0484, 0.0489], device='cuda:1'), out_proj_covar=tensor([1.0209e-04, 9.9986e-05, 1.1202e-04, 1.1958e-04, 1.0783e-04, 1.1421e-04, 1.1474e-04, 1.1532e-04], device='cuda:1') 2023-04-27 11:55:04,003 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-04-27 11:55:18,947 INFO [finetune.py:976] (1/7) Epoch 18, batch 5150, loss[loss=0.1439, simple_loss=0.2202, pruned_loss=0.03382, over 4772.00 frames. ], tot_loss[loss=0.1725, simple_loss=0.243, pruned_loss=0.05103, over 956157.20 frames. ], batch size: 28, lr: 3.32e-03, grad_scale: 64.0 2023-04-27 11:55:52,831 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.038e+02 1.715e+02 2.103e+02 2.437e+02 3.988e+02, threshold=4.205e+02, percent-clipped=0.0 2023-04-27 11:56:08,073 INFO [finetune.py:976] (1/7) Epoch 18, batch 5200, loss[loss=0.1971, simple_loss=0.271, pruned_loss=0.06158, over 4204.00 frames. ], tot_loss[loss=0.1771, simple_loss=0.2481, pruned_loss=0.05307, over 956344.51 frames. ], batch size: 65, lr: 3.32e-03, grad_scale: 64.0 2023-04-27 11:56:11,161 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=102576.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 11:56:42,078 INFO [finetune.py:976] (1/7) Epoch 18, batch 5250, loss[loss=0.1845, simple_loss=0.2602, pruned_loss=0.05437, over 4215.00 frames. ], tot_loss[loss=0.1776, simple_loss=0.2492, pruned_loss=0.05296, over 954756.63 frames. ], batch size: 65, lr: 3.32e-03, grad_scale: 64.0 2023-04-27 11:56:42,848 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.3714, 1.5901, 1.7018, 1.7783, 1.6243, 1.7735, 1.8846, 1.7907], device='cuda:1'), covar=tensor([0.4001, 0.5480, 0.4578, 0.4334, 0.5752, 0.7474, 0.4756, 0.4976], device='cuda:1'), in_proj_covar=tensor([0.0330, 0.0372, 0.0320, 0.0331, 0.0342, 0.0392, 0.0355, 0.0326], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 11:56:43,956 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=102624.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 11:56:56,938 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=102643.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:56:56,995 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9127, 1.4672, 1.7671, 1.8249, 1.7653, 1.4073, 0.8971, 1.4005], device='cuda:1'), covar=tensor([0.3439, 0.3425, 0.1855, 0.2179, 0.2574, 0.2728, 0.4075, 0.2203], device='cuda:1'), in_proj_covar=tensor([0.0289, 0.0246, 0.0226, 0.0313, 0.0219, 0.0230, 0.0227, 0.0184], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-27 11:57:01,559 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9573, 1.6581, 2.1453, 2.3117, 1.6911, 1.5234, 1.8601, 1.2386], device='cuda:1'), covar=tensor([0.0484, 0.0834, 0.0421, 0.0470, 0.0668, 0.1153, 0.0592, 0.0726], device='cuda:1'), in_proj_covar=tensor([0.0068, 0.0069, 0.0068, 0.0067, 0.0075, 0.0096, 0.0074, 0.0066], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-27 11:57:11,272 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.104e+02 1.529e+02 1.797e+02 2.306e+02 3.318e+02, threshold=3.594e+02, percent-clipped=0.0 2023-04-27 11:57:26,878 INFO [finetune.py:976] (1/7) Epoch 18, batch 5300, loss[loss=0.1936, simple_loss=0.2738, pruned_loss=0.05668, over 4918.00 frames. ], tot_loss[loss=0.1784, simple_loss=0.2503, pruned_loss=0.05323, over 954750.24 frames. ], batch size: 33, lr: 3.32e-03, grad_scale: 64.0 2023-04-27 11:57:55,087 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=102691.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:58:28,798 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9110, 1.4849, 1.5063, 1.7488, 2.0747, 1.7077, 1.4765, 1.4180], device='cuda:1'), covar=tensor([0.1576, 0.1497, 0.2178, 0.1226, 0.0951, 0.1512, 0.2010, 0.2525], device='cuda:1'), in_proj_covar=tensor([0.0310, 0.0312, 0.0353, 0.0291, 0.0330, 0.0310, 0.0303, 0.0370], device='cuda:1'), out_proj_covar=tensor([6.3969e-05, 6.4880e-05, 7.4836e-05, 5.8981e-05, 6.8430e-05, 6.5132e-05, 6.3826e-05, 7.8882e-05], device='cuda:1') 2023-04-27 11:58:31,734 INFO [finetune.py:976] (1/7) Epoch 18, batch 5350, loss[loss=0.1405, simple_loss=0.2207, pruned_loss=0.03013, over 4376.00 frames. ], tot_loss[loss=0.1771, simple_loss=0.2495, pruned_loss=0.05233, over 955155.13 frames. ], batch size: 65, lr: 3.32e-03, grad_scale: 64.0 2023-04-27 11:58:41,794 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7134, 2.0147, 1.7239, 1.8916, 1.4542, 1.6945, 1.7046, 1.3736], device='cuda:1'), covar=tensor([0.1536, 0.1001, 0.0734, 0.1034, 0.3297, 0.1027, 0.1579, 0.2140], device='cuda:1'), in_proj_covar=tensor([0.0287, 0.0304, 0.0218, 0.0280, 0.0311, 0.0259, 0.0250, 0.0266], device='cuda:1'), out_proj_covar=tensor([1.1521e-04, 1.2087e-04, 8.6451e-05, 1.1104e-04, 1.2641e-04, 1.0270e-04, 1.0101e-04, 1.0549e-04], device='cuda:1') 2023-04-27 11:59:01,540 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6573, 0.9958, 1.6497, 2.1258, 1.7460, 1.5740, 1.6395, 1.6022], device='cuda:1'), covar=tensor([0.4628, 0.6618, 0.6552, 0.5949, 0.6097, 0.7841, 0.7546, 0.8351], device='cuda:1'), in_proj_covar=tensor([0.0424, 0.0409, 0.0500, 0.0504, 0.0451, 0.0478, 0.0485, 0.0490], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 11:59:22,079 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 8.383e+01 1.595e+02 1.837e+02 2.151e+02 4.228e+02, threshold=3.674e+02, percent-clipped=3.0 2023-04-27 11:59:32,589 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.2103, 2.1671, 1.8218, 1.7965, 2.3579, 1.8412, 2.7311, 1.5575], device='cuda:1'), covar=tensor([0.3155, 0.1801, 0.4045, 0.2903, 0.1500, 0.2172, 0.1295, 0.4348], device='cuda:1'), in_proj_covar=tensor([0.0342, 0.0348, 0.0431, 0.0358, 0.0382, 0.0384, 0.0373, 0.0423], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 11:59:37,392 INFO [finetune.py:976] (1/7) Epoch 18, batch 5400, loss[loss=0.1522, simple_loss=0.227, pruned_loss=0.03874, over 4822.00 frames. ], tot_loss[loss=0.176, simple_loss=0.2474, pruned_loss=0.05232, over 955657.30 frames. ], batch size: 30, lr: 3.32e-03, grad_scale: 32.0 2023-04-27 11:59:54,481 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([3.3952, 3.2568, 2.4812, 3.9241, 3.4087, 3.4291, 1.2915, 3.3733], device='cuda:1'), covar=tensor([0.1845, 0.1394, 0.2906, 0.2203, 0.3029, 0.1902, 0.5733, 0.2546], device='cuda:1'), in_proj_covar=tensor([0.0242, 0.0213, 0.0248, 0.0303, 0.0297, 0.0248, 0.0269, 0.0271], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 12:00:05,397 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-04-27 12:00:49,559 INFO [finetune.py:976] (1/7) Epoch 18, batch 5450, loss[loss=0.1884, simple_loss=0.2569, pruned_loss=0.05993, over 4796.00 frames. ], tot_loss[loss=0.1739, simple_loss=0.2442, pruned_loss=0.05175, over 955370.27 frames. ], batch size: 51, lr: 3.32e-03, grad_scale: 32.0 2023-04-27 12:00:51,064 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.81 vs. limit=2.0 2023-04-27 12:01:23,562 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=102851.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:01:33,555 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.061e+02 1.567e+02 1.920e+02 2.316e+02 4.548e+02, threshold=3.840e+02, percent-clipped=3.0 2023-04-27 12:01:44,220 INFO [finetune.py:976] (1/7) Epoch 18, batch 5500, loss[loss=0.1037, simple_loss=0.1795, pruned_loss=0.01396, over 4772.00 frames. ], tot_loss[loss=0.171, simple_loss=0.2409, pruned_loss=0.05054, over 955805.00 frames. ], batch size: 28, lr: 3.32e-03, grad_scale: 32.0 2023-04-27 12:01:59,980 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=102887.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:02:17,500 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=102912.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:02:23,854 INFO [finetune.py:976] (1/7) Epoch 18, batch 5550, loss[loss=0.1839, simple_loss=0.2635, pruned_loss=0.05218, over 4863.00 frames. ], tot_loss[loss=0.1738, simple_loss=0.2438, pruned_loss=0.0519, over 953819.62 frames. ], batch size: 31, lr: 3.32e-03, grad_scale: 32.0 2023-04-27 12:02:25,777 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5069, 1.5198, 3.9507, 3.7261, 3.4381, 3.7043, 3.6524, 3.4658], device='cuda:1'), covar=tensor([0.6872, 0.5302, 0.1045, 0.1491, 0.1186, 0.1734, 0.2579, 0.1431], device='cuda:1'), in_proj_covar=tensor([0.0307, 0.0308, 0.0407, 0.0409, 0.0351, 0.0408, 0.0314, 0.0366], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-27 12:02:40,496 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=102948.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:02:45,149 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.004e+02 1.541e+02 1.915e+02 2.352e+02 3.794e+02, threshold=3.830e+02, percent-clipped=0.0 2023-04-27 12:02:54,416 INFO [finetune.py:976] (1/7) Epoch 18, batch 5600, loss[loss=0.1808, simple_loss=0.2609, pruned_loss=0.05035, over 4907.00 frames. ], tot_loss[loss=0.1772, simple_loss=0.2485, pruned_loss=0.05298, over 954491.79 frames. ], batch size: 43, lr: 3.32e-03, grad_scale: 32.0 2023-04-27 12:03:23,042 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.1587, 1.3854, 1.3282, 1.6686, 1.4711, 1.6366, 1.3184, 3.0348], device='cuda:1'), covar=tensor([0.0665, 0.0937, 0.0887, 0.1272, 0.0748, 0.0547, 0.0908, 0.0185], device='cuda:1'), in_proj_covar=tensor([0.0037, 0.0038, 0.0039, 0.0043, 0.0040, 0.0038, 0.0038, 0.0055], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:1') 2023-04-27 12:03:24,715 INFO [finetune.py:976] (1/7) Epoch 18, batch 5650, loss[loss=0.1773, simple_loss=0.2575, pruned_loss=0.04861, over 4912.00 frames. ], tot_loss[loss=0.1784, simple_loss=0.2508, pruned_loss=0.053, over 954060.41 frames. ], batch size: 37, lr: 3.31e-03, grad_scale: 32.0 2023-04-27 12:03:31,648 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.1179, 1.2870, 1.1768, 1.5563, 1.3775, 1.5028, 1.2442, 2.4617], device='cuda:1'), covar=tensor([0.0593, 0.0846, 0.0842, 0.1173, 0.0668, 0.0505, 0.0816, 0.0210], device='cuda:1'), in_proj_covar=tensor([0.0037, 0.0038, 0.0039, 0.0043, 0.0040, 0.0038, 0.0038, 0.0055], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:1') 2023-04-27 12:03:46,252 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.096e+02 1.540e+02 1.863e+02 2.247e+02 6.465e+02, threshold=3.725e+02, percent-clipped=2.0 2023-04-27 12:03:52,320 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.6714, 2.1878, 2.5611, 3.1089, 2.5845, 2.1509, 2.0009, 2.4068], device='cuda:1'), covar=tensor([0.3166, 0.3154, 0.1574, 0.2142, 0.2562, 0.2467, 0.3723, 0.1891], device='cuda:1'), in_proj_covar=tensor([0.0291, 0.0246, 0.0227, 0.0314, 0.0219, 0.0231, 0.0228, 0.0185], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-27 12:03:55,213 INFO [finetune.py:976] (1/7) Epoch 18, batch 5700, loss[loss=0.1663, simple_loss=0.234, pruned_loss=0.04934, over 4264.00 frames. ], tot_loss[loss=0.176, simple_loss=0.2466, pruned_loss=0.05272, over 934955.65 frames. ], batch size: 18, lr: 3.31e-03, grad_scale: 32.0 2023-04-27 12:04:00,095 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([3.2640, 2.7761, 2.9233, 3.5926, 3.2636, 2.9955, 2.5547, 3.2790], device='cuda:1'), covar=tensor([0.0664, 0.0881, 0.0557, 0.0480, 0.0569, 0.0757, 0.0687, 0.0435], device='cuda:1'), in_proj_covar=tensor([0.0187, 0.0199, 0.0180, 0.0170, 0.0176, 0.0180, 0.0149, 0.0176], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 12:04:00,822 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.77 vs. limit=2.0 2023-04-27 12:04:03,792 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-04-27 12:04:24,051 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=103098.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:04:24,531 INFO [finetune.py:976] (1/7) Epoch 19, batch 0, loss[loss=0.1876, simple_loss=0.2582, pruned_loss=0.05857, over 4779.00 frames. ], tot_loss[loss=0.1876, simple_loss=0.2582, pruned_loss=0.05857, over 4779.00 frames. ], batch size: 51, lr: 3.31e-03, grad_scale: 32.0 2023-04-27 12:04:24,531 INFO [finetune.py:1001] (1/7) Computing validation loss 2023-04-27 12:04:35,097 INFO [finetune.py:1010] (1/7) Epoch 19, validation: loss=0.1545, simple_loss=0.2248, pruned_loss=0.04209, over 2265189.00 frames. 2023-04-27 12:04:35,097 INFO [finetune.py:1011] (1/7) Maximum memory allocated so far is 6391MB 2023-04-27 12:04:56,252 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=103132.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:05:02,846 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-04-27 12:05:07,368 INFO [finetune.py:976] (1/7) Epoch 19, batch 50, loss[loss=0.1399, simple_loss=0.2158, pruned_loss=0.03203, over 4817.00 frames. ], tot_loss[loss=0.18, simple_loss=0.2515, pruned_loss=0.05423, over 216983.21 frames. ], batch size: 33, lr: 3.31e-03, grad_scale: 32.0 2023-04-27 12:05:13,147 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-04-27 12:05:13,475 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.041e+02 1.545e+02 1.855e+02 2.320e+02 4.324e+02, threshold=3.710e+02, percent-clipped=2.0 2023-04-27 12:05:15,437 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=103159.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:05:57,686 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=103193.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:06:01,235 INFO [finetune.py:976] (1/7) Epoch 19, batch 100, loss[loss=0.1676, simple_loss=0.2262, pruned_loss=0.05455, over 4840.00 frames. ], tot_loss[loss=0.1726, simple_loss=0.2426, pruned_loss=0.0513, over 382709.52 frames. ], batch size: 47, lr: 3.31e-03, grad_scale: 32.0 2023-04-27 12:06:13,376 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=103207.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:07:01,894 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=103243.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:07:06,071 INFO [finetune.py:976] (1/7) Epoch 19, batch 150, loss[loss=0.1539, simple_loss=0.2211, pruned_loss=0.04328, over 4896.00 frames. ], tot_loss[loss=0.1692, simple_loss=0.2388, pruned_loss=0.04977, over 512269.71 frames. ], batch size: 35, lr: 3.31e-03, grad_scale: 32.0 2023-04-27 12:07:16,418 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.139e+02 1.567e+02 1.850e+02 2.251e+02 4.056e+02, threshold=3.701e+02, percent-clipped=1.0 2023-04-27 12:07:35,077 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.2309, 1.4864, 1.4562, 1.7768, 1.6655, 1.7888, 1.4255, 3.3600], device='cuda:1'), covar=tensor([0.0635, 0.0785, 0.0784, 0.1178, 0.0636, 0.0535, 0.0727, 0.0156], device='cuda:1'), in_proj_covar=tensor([0.0037, 0.0038, 0.0039, 0.0043, 0.0040, 0.0038, 0.0038, 0.0055], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:1') 2023-04-27 12:07:45,051 INFO [finetune.py:976] (1/7) Epoch 19, batch 200, loss[loss=0.1655, simple_loss=0.243, pruned_loss=0.04395, over 4817.00 frames. ], tot_loss[loss=0.1704, simple_loss=0.2395, pruned_loss=0.05068, over 611591.93 frames. ], batch size: 45, lr: 3.31e-03, grad_scale: 32.0 2023-04-27 12:07:48,066 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.5907, 2.1595, 2.6105, 3.1518, 2.6485, 2.1058, 2.0165, 2.4881], device='cuda:1'), covar=tensor([0.3009, 0.2865, 0.1398, 0.1755, 0.2374, 0.2485, 0.3376, 0.1616], device='cuda:1'), in_proj_covar=tensor([0.0290, 0.0246, 0.0226, 0.0313, 0.0219, 0.0231, 0.0227, 0.0184], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-27 12:08:14,339 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0424, 2.6443, 1.1188, 1.5682, 1.9828, 1.2358, 3.4038, 1.9008], device='cuda:1'), covar=tensor([0.0658, 0.0562, 0.0763, 0.1126, 0.0484, 0.0978, 0.0223, 0.0524], device='cuda:1'), in_proj_covar=tensor([0.0050, 0.0064, 0.0047, 0.0046, 0.0049, 0.0052, 0.0073, 0.0050], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0009, 0.0007, 0.0007, 0.0008, 0.0008, 0.0010, 0.0007], device='cuda:1') 2023-04-27 12:08:33,911 INFO [finetune.py:976] (1/7) Epoch 19, batch 250, loss[loss=0.1225, simple_loss=0.1882, pruned_loss=0.02842, over 4819.00 frames. ], tot_loss[loss=0.1715, simple_loss=0.2417, pruned_loss=0.05066, over 687416.51 frames. ], batch size: 25, lr: 3.31e-03, grad_scale: 32.0 2023-04-27 12:08:44,150 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.956e+01 1.630e+02 1.978e+02 2.388e+02 4.596e+02, threshold=3.957e+02, percent-clipped=1.0 2023-04-27 12:08:56,544 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=103366.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 12:08:58,390 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.0214, 1.3255, 1.2655, 1.6314, 1.4546, 1.5933, 1.3068, 3.0091], device='cuda:1'), covar=tensor([0.0762, 0.0960, 0.0953, 0.1299, 0.0768, 0.0581, 0.0914, 0.0213], device='cuda:1'), in_proj_covar=tensor([0.0037, 0.0038, 0.0039, 0.0043, 0.0039, 0.0038, 0.0038, 0.0055], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0014, 0.0013, 0.0012, 0.0013, 0.0016], device='cuda:1') 2023-04-27 12:09:14,218 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.67 vs. limit=5.0 2023-04-27 12:09:22,446 INFO [finetune.py:976] (1/7) Epoch 19, batch 300, loss[loss=0.132, simple_loss=0.1976, pruned_loss=0.03323, over 4792.00 frames. ], tot_loss[loss=0.1748, simple_loss=0.2458, pruned_loss=0.05185, over 746676.29 frames. ], batch size: 25, lr: 3.31e-03, grad_scale: 32.0 2023-04-27 12:09:42,600 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=103427.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 12:09:55,322 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4078, 1.2403, 1.6137, 1.5996, 1.2672, 1.1588, 1.2575, 0.7799], device='cuda:1'), covar=tensor([0.0514, 0.0709, 0.0396, 0.0577, 0.0711, 0.1235, 0.0596, 0.0611], device='cuda:1'), in_proj_covar=tensor([0.0068, 0.0068, 0.0067, 0.0067, 0.0075, 0.0094, 0.0073, 0.0065], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-27 12:09:55,811 INFO [finetune.py:976] (1/7) Epoch 19, batch 350, loss[loss=0.1729, simple_loss=0.2556, pruned_loss=0.04513, over 4788.00 frames. ], tot_loss[loss=0.1773, simple_loss=0.2485, pruned_loss=0.05302, over 792905.84 frames. ], batch size: 29, lr: 3.31e-03, grad_scale: 32.0 2023-04-27 12:09:58,942 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=103454.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:10:00,581 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.010e+02 1.607e+02 1.938e+02 2.366e+02 5.284e+02, threshold=3.875e+02, percent-clipped=4.0 2023-04-27 12:10:22,423 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=103488.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:10:23,692 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0073, 1.5391, 1.5382, 1.7651, 2.2031, 1.8046, 1.5116, 1.4350], device='cuda:1'), covar=tensor([0.1428, 0.1472, 0.1880, 0.1266, 0.0745, 0.1415, 0.1760, 0.2258], device='cuda:1'), in_proj_covar=tensor([0.0306, 0.0310, 0.0349, 0.0288, 0.0328, 0.0306, 0.0301, 0.0367], device='cuda:1'), out_proj_covar=tensor([6.3112e-05, 6.4461e-05, 7.3942e-05, 5.8483e-05, 6.8205e-05, 6.4228e-05, 6.3217e-05, 7.8292e-05], device='cuda:1') 2023-04-27 12:10:24,895 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4671, 1.4071, 1.8018, 1.7225, 1.3330, 1.2569, 1.4686, 0.9161], device='cuda:1'), covar=tensor([0.0611, 0.0543, 0.0345, 0.0544, 0.0705, 0.1126, 0.0559, 0.0596], device='cuda:1'), in_proj_covar=tensor([0.0068, 0.0068, 0.0068, 0.0067, 0.0075, 0.0095, 0.0073, 0.0066], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-27 12:10:29,037 INFO [finetune.py:976] (1/7) Epoch 19, batch 400, loss[loss=0.1379, simple_loss=0.2082, pruned_loss=0.03381, over 4855.00 frames. ], tot_loss[loss=0.1788, simple_loss=0.2503, pruned_loss=0.05362, over 827912.66 frames. ], batch size: 31, lr: 3.31e-03, grad_scale: 32.0 2023-04-27 12:10:34,535 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=103507.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:10:58,904 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=103543.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:10:58,936 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.0993, 0.7357, 0.9241, 0.7864, 1.2188, 0.9736, 0.8521, 1.0163], device='cuda:1'), covar=tensor([0.1591, 0.1510, 0.2133, 0.1521, 0.1013, 0.1364, 0.1689, 0.2332], device='cuda:1'), in_proj_covar=tensor([0.0306, 0.0311, 0.0349, 0.0289, 0.0329, 0.0306, 0.0301, 0.0368], device='cuda:1'), out_proj_covar=tensor([6.3190e-05, 6.4551e-05, 7.4044e-05, 5.8551e-05, 6.8299e-05, 6.4365e-05, 6.3264e-05, 7.8413e-05], device='cuda:1') 2023-04-27 12:11:02,472 INFO [finetune.py:976] (1/7) Epoch 19, batch 450, loss[loss=0.1755, simple_loss=0.2452, pruned_loss=0.05289, over 4886.00 frames. ], tot_loss[loss=0.176, simple_loss=0.2479, pruned_loss=0.05203, over 858443.80 frames. ], batch size: 32, lr: 3.31e-03, grad_scale: 32.0 2023-04-27 12:11:06,182 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=103555.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:11:06,725 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.103e+02 1.516e+02 1.781e+02 2.073e+02 5.569e+02, threshold=3.563e+02, percent-clipped=1.0 2023-04-27 12:11:31,864 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.2422, 2.6336, 2.3405, 2.5764, 1.8268, 2.1867, 2.2491, 1.6605], device='cuda:1'), covar=tensor([0.1843, 0.1080, 0.0768, 0.1019, 0.3197, 0.1439, 0.1914, 0.2718], device='cuda:1'), in_proj_covar=tensor([0.0286, 0.0303, 0.0218, 0.0279, 0.0311, 0.0259, 0.0251, 0.0265], device='cuda:1'), out_proj_covar=tensor([1.1488e-04, 1.2061e-04, 8.6466e-05, 1.1062e-04, 1.2621e-04, 1.0268e-04, 1.0130e-04, 1.0518e-04], device='cuda:1') 2023-04-27 12:11:42,148 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=103591.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:11:46,990 INFO [finetune.py:976] (1/7) Epoch 19, batch 500, loss[loss=0.1531, simple_loss=0.2205, pruned_loss=0.04291, over 4747.00 frames. ], tot_loss[loss=0.1751, simple_loss=0.2459, pruned_loss=0.05212, over 879926.48 frames. ], batch size: 27, lr: 3.31e-03, grad_scale: 32.0 2023-04-27 12:11:53,261 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.2433, 2.7096, 1.2517, 1.6156, 2.0442, 1.4543, 3.2101, 1.7247], device='cuda:1'), covar=tensor([0.0579, 0.0669, 0.0747, 0.0969, 0.0412, 0.0829, 0.0214, 0.0554], device='cuda:1'), in_proj_covar=tensor([0.0051, 0.0065, 0.0048, 0.0046, 0.0050, 0.0052, 0.0073, 0.0051], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:1') 2023-04-27 12:12:13,940 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9404, 2.4676, 1.9895, 1.8294, 1.4544, 1.4756, 2.0046, 1.3574], device='cuda:1'), covar=tensor([0.1715, 0.1357, 0.1500, 0.1785, 0.2518, 0.2087, 0.1033, 0.2161], device='cuda:1'), in_proj_covar=tensor([0.0197, 0.0213, 0.0168, 0.0205, 0.0200, 0.0185, 0.0156, 0.0186], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-27 12:12:30,580 INFO [finetune.py:976] (1/7) Epoch 19, batch 550, loss[loss=0.155, simple_loss=0.2303, pruned_loss=0.03981, over 4802.00 frames. ], tot_loss[loss=0.1738, simple_loss=0.244, pruned_loss=0.05177, over 896545.39 frames. ], batch size: 51, lr: 3.31e-03, grad_scale: 32.0 2023-04-27 12:12:34,853 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.077e+02 1.551e+02 1.854e+02 2.248e+02 4.068e+02, threshold=3.707e+02, percent-clipped=1.0 2023-04-27 12:13:04,201 INFO [finetune.py:976] (1/7) Epoch 19, batch 600, loss[loss=0.1549, simple_loss=0.2252, pruned_loss=0.04225, over 4790.00 frames. ], tot_loss[loss=0.1742, simple_loss=0.2448, pruned_loss=0.05179, over 908925.33 frames. ], batch size: 29, lr: 3.31e-03, grad_scale: 32.0 2023-04-27 12:13:19,608 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=103722.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 12:13:43,193 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=103740.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:13:53,493 INFO [finetune.py:976] (1/7) Epoch 19, batch 650, loss[loss=0.1829, simple_loss=0.2567, pruned_loss=0.0546, over 4904.00 frames. ], tot_loss[loss=0.1762, simple_loss=0.2476, pruned_loss=0.05239, over 917754.58 frames. ], batch size: 35, lr: 3.31e-03, grad_scale: 32.0 2023-04-27 12:14:01,801 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=103754.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:14:02,875 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.117e+02 1.593e+02 1.866e+02 2.212e+02 4.115e+02, threshold=3.733e+02, percent-clipped=2.0 2023-04-27 12:14:05,384 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4471, 1.3560, 1.7538, 1.7156, 1.3805, 1.2025, 1.4729, 0.9709], device='cuda:1'), covar=tensor([0.0622, 0.0565, 0.0431, 0.0558, 0.0727, 0.1120, 0.0574, 0.0628], device='cuda:1'), in_proj_covar=tensor([0.0067, 0.0068, 0.0067, 0.0066, 0.0074, 0.0094, 0.0073, 0.0065], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-27 12:14:14,197 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=103765.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:14:30,663 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=103788.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:14:37,245 INFO [finetune.py:976] (1/7) Epoch 19, batch 700, loss[loss=0.1777, simple_loss=0.2458, pruned_loss=0.05479, over 4901.00 frames. ], tot_loss[loss=0.1766, simple_loss=0.2486, pruned_loss=0.05237, over 928405.81 frames. ], batch size: 35, lr: 3.31e-03, grad_scale: 32.0 2023-04-27 12:14:38,724 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=103801.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:14:39,255 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=103802.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:14:46,828 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2023-04-27 12:14:55,329 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=103826.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:15:02,440 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=103836.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:15:10,877 INFO [finetune.py:976] (1/7) Epoch 19, batch 750, loss[loss=0.1489, simple_loss=0.2316, pruned_loss=0.03313, over 4835.00 frames. ], tot_loss[loss=0.1786, simple_loss=0.251, pruned_loss=0.05313, over 936497.93 frames. ], batch size: 47, lr: 3.30e-03, grad_scale: 32.0 2023-04-27 12:15:15,074 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.139e+02 1.609e+02 1.947e+02 2.389e+02 3.942e+02, threshold=3.894e+02, percent-clipped=2.0 2023-04-27 12:15:28,406 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=103876.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:15:44,006 INFO [finetune.py:976] (1/7) Epoch 19, batch 800, loss[loss=0.1838, simple_loss=0.2537, pruned_loss=0.05696, over 4793.00 frames. ], tot_loss[loss=0.1775, simple_loss=0.2503, pruned_loss=0.0524, over 939552.12 frames. ], batch size: 51, lr: 3.30e-03, grad_scale: 32.0 2023-04-27 12:16:09,160 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=103937.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 12:16:17,416 INFO [finetune.py:976] (1/7) Epoch 19, batch 850, loss[loss=0.1369, simple_loss=0.223, pruned_loss=0.02536, over 4775.00 frames. ], tot_loss[loss=0.1755, simple_loss=0.2477, pruned_loss=0.05162, over 945050.99 frames. ], batch size: 26, lr: 3.30e-03, grad_scale: 32.0 2023-04-27 12:16:21,649 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.943e+01 1.500e+02 1.730e+02 2.145e+02 3.862e+02, threshold=3.461e+02, percent-clipped=0.0 2023-04-27 12:16:27,894 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=3.21 vs. limit=5.0 2023-04-27 12:16:31,045 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2023-04-27 12:16:53,704 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=5.09 vs. limit=5.0 2023-04-27 12:16:59,790 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=103990.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:17:05,170 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=103997.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:17:10,831 INFO [finetune.py:976] (1/7) Epoch 19, batch 900, loss[loss=0.148, simple_loss=0.2113, pruned_loss=0.04231, over 4712.00 frames. ], tot_loss[loss=0.173, simple_loss=0.2448, pruned_loss=0.05058, over 946546.75 frames. ], batch size: 23, lr: 3.30e-03, grad_scale: 32.0 2023-04-27 12:17:37,465 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=104022.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 12:17:57,244 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-04-27 12:18:05,341 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.6432, 2.0934, 2.5247, 3.1317, 2.5413, 2.0361, 2.0548, 2.3436], device='cuda:1'), covar=tensor([0.3463, 0.3261, 0.1776, 0.2651, 0.2786, 0.2767, 0.3871, 0.2268], device='cuda:1'), in_proj_covar=tensor([0.0292, 0.0247, 0.0227, 0.0315, 0.0219, 0.0232, 0.0228, 0.0186], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-27 12:18:18,255 INFO [finetune.py:976] (1/7) Epoch 19, batch 950, loss[loss=0.1745, simple_loss=0.2419, pruned_loss=0.05355, over 4685.00 frames. ], tot_loss[loss=0.1713, simple_loss=0.2423, pruned_loss=0.05017, over 949284.23 frames. ], batch size: 23, lr: 3.30e-03, grad_scale: 32.0 2023-04-27 12:18:18,362 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=104049.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:18:19,613 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=104051.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:18:27,839 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.269e+02 1.627e+02 1.900e+02 2.232e+02 4.587e+02, threshold=3.799e+02, percent-clipped=1.0 2023-04-27 12:18:29,218 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=104058.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:18:30,542 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 2023-04-27 12:18:41,643 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=104070.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 12:19:02,240 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4599, 1.7383, 1.6069, 1.8755, 1.7838, 1.9889, 1.6731, 3.1744], device='cuda:1'), covar=tensor([0.0581, 0.0670, 0.0671, 0.1040, 0.0528, 0.0584, 0.0671, 0.0187], device='cuda:1'), in_proj_covar=tensor([0.0037, 0.0039, 0.0039, 0.0043, 0.0040, 0.0038, 0.0038, 0.0056], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:1') 2023-04-27 12:19:14,757 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=104096.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:19:21,310 INFO [finetune.py:976] (1/7) Epoch 19, batch 1000, loss[loss=0.1676, simple_loss=0.2453, pruned_loss=0.04497, over 4870.00 frames. ], tot_loss[loss=0.1728, simple_loss=0.2441, pruned_loss=0.05077, over 952557.10 frames. ], batch size: 34, lr: 3.30e-03, grad_scale: 32.0 2023-04-27 12:19:33,854 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=104110.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:19:51,328 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=104121.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:19:55,056 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6052, 1.5552, 1.9230, 1.9332, 1.5514, 1.3534, 1.6283, 1.0859], device='cuda:1'), covar=tensor([0.0653, 0.0625, 0.0499, 0.0753, 0.0679, 0.0961, 0.0682, 0.0726], device='cuda:1'), in_proj_covar=tensor([0.0068, 0.0068, 0.0067, 0.0067, 0.0074, 0.0095, 0.0073, 0.0066], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-27 12:20:05,156 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9401, 2.3483, 0.9999, 1.3982, 1.7569, 1.3022, 2.8906, 1.6380], device='cuda:1'), covar=tensor([0.0662, 0.0598, 0.0772, 0.1252, 0.0479, 0.0944, 0.0271, 0.0628], device='cuda:1'), in_proj_covar=tensor([0.0050, 0.0065, 0.0048, 0.0046, 0.0050, 0.0052, 0.0074, 0.0051], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0007], device='cuda:1') 2023-04-27 12:20:15,912 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.68 vs. limit=5.0 2023-04-27 12:20:27,135 INFO [finetune.py:976] (1/7) Epoch 19, batch 1050, loss[loss=0.1911, simple_loss=0.2754, pruned_loss=0.05337, over 4910.00 frames. ], tot_loss[loss=0.1747, simple_loss=0.2471, pruned_loss=0.05114, over 954493.02 frames. ], batch size: 43, lr: 3.30e-03, grad_scale: 32.0 2023-04-27 12:20:37,627 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.418e+01 1.570e+02 1.795e+02 2.236e+02 4.469e+02, threshold=3.589e+02, percent-clipped=1.0 2023-04-27 12:20:38,389 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.3816, 1.6554, 1.5184, 1.9504, 1.7446, 1.9329, 1.4833, 4.1122], device='cuda:1'), covar=tensor([0.0553, 0.0782, 0.0795, 0.1171, 0.0655, 0.0564, 0.0780, 0.0124], device='cuda:1'), in_proj_covar=tensor([0.0037, 0.0039, 0.0039, 0.0043, 0.0040, 0.0038, 0.0038, 0.0056], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:1') 2023-04-27 12:21:05,398 INFO [finetune.py:976] (1/7) Epoch 19, batch 1100, loss[loss=0.1806, simple_loss=0.2545, pruned_loss=0.05332, over 4745.00 frames. ], tot_loss[loss=0.1749, simple_loss=0.2474, pruned_loss=0.05117, over 952627.87 frames. ], batch size: 59, lr: 3.30e-03, grad_scale: 32.0 2023-04-27 12:21:27,607 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=104232.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 12:21:34,195 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5368, 1.3362, 4.1449, 3.8916, 3.5902, 3.9071, 3.8322, 3.6655], device='cuda:1'), covar=tensor([0.7195, 0.5815, 0.1069, 0.1602, 0.1081, 0.1685, 0.1819, 0.1438], device='cuda:1'), in_proj_covar=tensor([0.0306, 0.0307, 0.0403, 0.0407, 0.0350, 0.0407, 0.0312, 0.0366], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-27 12:21:39,372 INFO [finetune.py:976] (1/7) Epoch 19, batch 1150, loss[loss=0.2028, simple_loss=0.2811, pruned_loss=0.06221, over 4891.00 frames. ], tot_loss[loss=0.1767, simple_loss=0.2491, pruned_loss=0.05216, over 953155.18 frames. ], batch size: 37, lr: 3.30e-03, grad_scale: 32.0 2023-04-27 12:21:39,468 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5373, 3.5049, 1.0551, 1.9154, 1.9194, 2.4484, 1.9195, 0.9795], device='cuda:1'), covar=tensor([0.1444, 0.0938, 0.1973, 0.1238, 0.1095, 0.1025, 0.1612, 0.2119], device='cuda:1'), in_proj_covar=tensor([0.0116, 0.0241, 0.0136, 0.0120, 0.0131, 0.0151, 0.0116, 0.0119], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-27 12:21:44,616 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.137e+01 1.582e+02 1.911e+02 2.328e+02 3.877e+02, threshold=3.822e+02, percent-clipped=1.0 2023-04-27 12:22:12,706 INFO [finetune.py:976] (1/7) Epoch 19, batch 1200, loss[loss=0.1988, simple_loss=0.2569, pruned_loss=0.07033, over 4818.00 frames. ], tot_loss[loss=0.1751, simple_loss=0.2473, pruned_loss=0.05149, over 953346.14 frames. ], batch size: 39, lr: 3.30e-03, grad_scale: 32.0 2023-04-27 12:22:15,770 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8948, 2.5389, 1.9191, 1.8254, 1.3350, 1.4077, 1.8933, 1.3111], device='cuda:1'), covar=tensor([0.1831, 0.1324, 0.1476, 0.1820, 0.2564, 0.2103, 0.1071, 0.2198], device='cuda:1'), in_proj_covar=tensor([0.0197, 0.0212, 0.0168, 0.0206, 0.0201, 0.0185, 0.0156, 0.0186], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-27 12:22:29,899 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-04-27 12:22:44,236 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=104346.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:22:46,459 INFO [finetune.py:976] (1/7) Epoch 19, batch 1250, loss[loss=0.182, simple_loss=0.2574, pruned_loss=0.05333, over 4867.00 frames. ], tot_loss[loss=0.1741, simple_loss=0.2455, pruned_loss=0.05138, over 952823.67 frames. ], batch size: 31, lr: 3.30e-03, grad_scale: 32.0 2023-04-27 12:22:49,468 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=104353.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:22:51,237 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 8.445e+01 1.483e+02 1.801e+02 2.223e+02 4.756e+02, threshold=3.603e+02, percent-clipped=1.0 2023-04-27 12:23:45,731 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=104396.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:23:47,995 INFO [finetune.py:976] (1/7) Epoch 19, batch 1300, loss[loss=0.1918, simple_loss=0.2646, pruned_loss=0.05947, over 4841.00 frames. ], tot_loss[loss=0.1712, simple_loss=0.2419, pruned_loss=0.05025, over 954503.59 frames. ], batch size: 47, lr: 3.30e-03, grad_scale: 32.0 2023-04-27 12:23:56,156 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=104405.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:24:18,871 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=104421.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:24:44,516 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=104444.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:24:53,501 INFO [finetune.py:976] (1/7) Epoch 19, batch 1350, loss[loss=0.2521, simple_loss=0.3017, pruned_loss=0.1013, over 4902.00 frames. ], tot_loss[loss=0.172, simple_loss=0.2421, pruned_loss=0.05092, over 953361.28 frames. ], batch size: 43, lr: 3.30e-03, grad_scale: 32.0 2023-04-27 12:25:03,938 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.093e+02 1.562e+02 1.851e+02 2.195e+02 3.087e+02, threshold=3.702e+02, percent-clipped=0.0 2023-04-27 12:25:15,281 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.1269, 1.8390, 2.0526, 2.3822, 2.4322, 2.0168, 1.5687, 2.0949], device='cuda:1'), covar=tensor([0.0885, 0.1118, 0.0705, 0.0614, 0.0609, 0.0907, 0.0843, 0.0674], device='cuda:1'), in_proj_covar=tensor([0.0185, 0.0198, 0.0179, 0.0169, 0.0175, 0.0178, 0.0150, 0.0176], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 12:25:23,558 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=104469.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:25:57,941 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5745, 1.5927, 1.4518, 1.1350, 1.2423, 1.2170, 1.4263, 1.1728], device='cuda:1'), covar=tensor([0.1479, 0.1295, 0.1287, 0.1560, 0.1994, 0.1566, 0.0954, 0.1748], device='cuda:1'), in_proj_covar=tensor([0.0197, 0.0212, 0.0168, 0.0205, 0.0199, 0.0184, 0.0156, 0.0185], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-27 12:25:58,402 INFO [finetune.py:976] (1/7) Epoch 19, batch 1400, loss[loss=0.1699, simple_loss=0.2232, pruned_loss=0.05829, over 4107.00 frames. ], tot_loss[loss=0.1732, simple_loss=0.2443, pruned_loss=0.05105, over 953263.14 frames. ], batch size: 17, lr: 3.30e-03, grad_scale: 32.0 2023-04-27 12:26:30,550 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=104532.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 12:26:34,509 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.51 vs. limit=2.0 2023-04-27 12:26:41,373 INFO [finetune.py:976] (1/7) Epoch 19, batch 1450, loss[loss=0.1887, simple_loss=0.2749, pruned_loss=0.05124, over 4868.00 frames. ], tot_loss[loss=0.1752, simple_loss=0.2468, pruned_loss=0.05178, over 952453.68 frames. ], batch size: 34, lr: 3.30e-03, grad_scale: 32.0 2023-04-27 12:26:44,065 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.50 vs. limit=5.0 2023-04-27 12:26:46,126 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.128e+02 1.618e+02 1.883e+02 2.290e+02 5.063e+02, threshold=3.766e+02, percent-clipped=2.0 2023-04-27 12:27:03,310 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=104580.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:27:14,737 INFO [finetune.py:976] (1/7) Epoch 19, batch 1500, loss[loss=0.1957, simple_loss=0.2671, pruned_loss=0.06216, over 4852.00 frames. ], tot_loss[loss=0.1763, simple_loss=0.248, pruned_loss=0.05228, over 953876.89 frames. ], batch size: 31, lr: 3.30e-03, grad_scale: 32.0 2023-04-27 12:27:46,358 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-04-27 12:27:46,868 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=104646.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:27:48,606 INFO [finetune.py:976] (1/7) Epoch 19, batch 1550, loss[loss=0.1616, simple_loss=0.2099, pruned_loss=0.05664, over 4179.00 frames. ], tot_loss[loss=0.1771, simple_loss=0.2486, pruned_loss=0.05284, over 954894.03 frames. ], batch size: 18, lr: 3.30e-03, grad_scale: 32.0 2023-04-27 12:27:51,147 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=104653.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:27:53,360 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.152e+02 1.681e+02 1.907e+02 2.272e+02 3.950e+02, threshold=3.814e+02, percent-clipped=3.0 2023-04-27 12:28:35,129 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=104694.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:28:43,386 INFO [finetune.py:976] (1/7) Epoch 19, batch 1600, loss[loss=0.175, simple_loss=0.2351, pruned_loss=0.05744, over 4856.00 frames. ], tot_loss[loss=0.1743, simple_loss=0.2454, pruned_loss=0.05159, over 955545.65 frames. ], batch size: 44, lr: 3.30e-03, grad_scale: 32.0 2023-04-27 12:28:44,682 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=104701.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:28:47,656 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=104705.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:29:06,232 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.3718, 2.9968, 2.5595, 2.8080, 2.1182, 2.5161, 2.7841, 1.9259], device='cuda:1'), covar=tensor([0.2091, 0.1234, 0.0749, 0.1245, 0.3128, 0.1314, 0.1967, 0.2745], device='cuda:1'), in_proj_covar=tensor([0.0289, 0.0304, 0.0218, 0.0281, 0.0312, 0.0259, 0.0253, 0.0266], device='cuda:1'), out_proj_covar=tensor([1.1606e-04, 1.2078e-04, 8.6293e-05, 1.1143e-04, 1.2689e-04, 1.0287e-04, 1.0196e-04, 1.0567e-04], device='cuda:1') 2023-04-27 12:29:27,150 INFO [finetune.py:976] (1/7) Epoch 19, batch 1650, loss[loss=0.1983, simple_loss=0.2535, pruned_loss=0.07153, over 4860.00 frames. ], tot_loss[loss=0.1729, simple_loss=0.2434, pruned_loss=0.05118, over 955975.23 frames. ], batch size: 49, lr: 3.30e-03, grad_scale: 64.0 2023-04-27 12:29:27,361 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.28 vs. limit=5.0 2023-04-27 12:29:29,682 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=104753.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:29:31,435 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.063e+01 1.479e+02 1.732e+02 2.090e+02 3.270e+02, threshold=3.465e+02, percent-clipped=0.0 2023-04-27 12:30:01,077 INFO [finetune.py:976] (1/7) Epoch 19, batch 1700, loss[loss=0.1814, simple_loss=0.2349, pruned_loss=0.06392, over 4188.00 frames. ], tot_loss[loss=0.171, simple_loss=0.2411, pruned_loss=0.05051, over 955312.83 frames. ], batch size: 18, lr: 3.30e-03, grad_scale: 64.0 2023-04-27 12:30:08,378 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0633, 1.1911, 4.6045, 4.2855, 4.0460, 4.3926, 4.2143, 4.1610], device='cuda:1'), covar=tensor([0.6582, 0.6382, 0.1229, 0.2074, 0.1295, 0.1463, 0.2019, 0.1583], device='cuda:1'), in_proj_covar=tensor([0.0306, 0.0305, 0.0404, 0.0407, 0.0349, 0.0407, 0.0313, 0.0366], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-27 12:30:41,224 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-04-27 12:30:48,746 INFO [finetune.py:976] (1/7) Epoch 19, batch 1750, loss[loss=0.1449, simple_loss=0.2242, pruned_loss=0.03282, over 4769.00 frames. ], tot_loss[loss=0.1731, simple_loss=0.2431, pruned_loss=0.05152, over 955230.56 frames. ], batch size: 27, lr: 3.30e-03, grad_scale: 64.0 2023-04-27 12:30:50,809 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-04-27 12:30:53,007 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.077e+02 1.623e+02 1.966e+02 2.378e+02 4.217e+02, threshold=3.932e+02, percent-clipped=2.0 2023-04-27 12:31:37,540 INFO [finetune.py:976] (1/7) Epoch 19, batch 1800, loss[loss=0.181, simple_loss=0.2543, pruned_loss=0.05381, over 4805.00 frames. ], tot_loss[loss=0.1749, simple_loss=0.2455, pruned_loss=0.05217, over 953166.36 frames. ], batch size: 51, lr: 3.30e-03, grad_scale: 64.0 2023-04-27 12:32:10,885 INFO [finetune.py:976] (1/7) Epoch 19, batch 1850, loss[loss=0.1815, simple_loss=0.2348, pruned_loss=0.06408, over 4174.00 frames. ], tot_loss[loss=0.1751, simple_loss=0.2458, pruned_loss=0.05223, over 950594.15 frames. ], batch size: 17, lr: 3.30e-03, grad_scale: 64.0 2023-04-27 12:32:15,623 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.138e+02 1.634e+02 1.902e+02 2.257e+02 6.149e+02, threshold=3.804e+02, percent-clipped=1.0 2023-04-27 12:32:44,645 INFO [finetune.py:976] (1/7) Epoch 19, batch 1900, loss[loss=0.1453, simple_loss=0.2233, pruned_loss=0.03368, over 4792.00 frames. ], tot_loss[loss=0.1763, simple_loss=0.2478, pruned_loss=0.05241, over 952327.13 frames. ], batch size: 25, lr: 3.30e-03, grad_scale: 64.0 2023-04-27 12:33:18,499 INFO [finetune.py:976] (1/7) Epoch 19, batch 1950, loss[loss=0.1409, simple_loss=0.215, pruned_loss=0.03334, over 4765.00 frames. ], tot_loss[loss=0.175, simple_loss=0.2468, pruned_loss=0.05159, over 954403.72 frames. ], batch size: 27, lr: 3.30e-03, grad_scale: 64.0 2023-04-27 12:33:22,765 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.216e+02 1.554e+02 1.843e+02 2.166e+02 4.298e+02, threshold=3.686e+02, percent-clipped=1.0 2023-04-27 12:34:18,466 INFO [finetune.py:976] (1/7) Epoch 19, batch 2000, loss[loss=0.207, simple_loss=0.2584, pruned_loss=0.07778, over 4939.00 frames. ], tot_loss[loss=0.1739, simple_loss=0.2448, pruned_loss=0.05154, over 955911.42 frames. ], batch size: 33, lr: 3.29e-03, grad_scale: 64.0 2023-04-27 12:34:56,985 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4547, 1.6408, 1.5715, 1.8338, 1.7607, 1.9555, 1.5701, 3.1099], device='cuda:1'), covar=tensor([0.0579, 0.0657, 0.0662, 0.1001, 0.0516, 0.0541, 0.0655, 0.0178], device='cuda:1'), in_proj_covar=tensor([0.0037, 0.0038, 0.0039, 0.0043, 0.0040, 0.0038, 0.0038, 0.0055], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:1') 2023-04-27 12:35:14,512 INFO [finetune.py:976] (1/7) Epoch 19, batch 2050, loss[loss=0.1453, simple_loss=0.2122, pruned_loss=0.0392, over 4832.00 frames. ], tot_loss[loss=0.1721, simple_loss=0.2425, pruned_loss=0.05082, over 956187.09 frames. ], batch size: 30, lr: 3.29e-03, grad_scale: 64.0 2023-04-27 12:35:18,788 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.655e+01 1.540e+02 1.837e+02 2.232e+02 4.472e+02, threshold=3.673e+02, percent-clipped=5.0 2023-04-27 12:35:19,632 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.12 vs. limit=2.0 2023-04-27 12:35:30,919 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.0503, 1.3573, 1.3208, 1.7019, 1.4989, 1.7833, 1.3318, 3.2737], device='cuda:1'), covar=tensor([0.0692, 0.0892, 0.0901, 0.1267, 0.0718, 0.0554, 0.0860, 0.0202], device='cuda:1'), in_proj_covar=tensor([0.0037, 0.0038, 0.0039, 0.0043, 0.0040, 0.0038, 0.0038, 0.0055], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:1') 2023-04-27 12:35:52,485 INFO [finetune.py:976] (1/7) Epoch 19, batch 2100, loss[loss=0.1658, simple_loss=0.2488, pruned_loss=0.0414, over 4815.00 frames. ], tot_loss[loss=0.1716, simple_loss=0.242, pruned_loss=0.0506, over 956031.29 frames. ], batch size: 38, lr: 3.29e-03, grad_scale: 32.0 2023-04-27 12:36:37,773 INFO [finetune.py:976] (1/7) Epoch 19, batch 2150, loss[loss=0.2267, simple_loss=0.3106, pruned_loss=0.07136, over 4856.00 frames. ], tot_loss[loss=0.1749, simple_loss=0.2456, pruned_loss=0.05217, over 953137.96 frames. ], batch size: 44, lr: 3.29e-03, grad_scale: 32.0 2023-04-27 12:36:48,984 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.234e+01 1.652e+02 1.972e+02 2.377e+02 5.881e+02, threshold=3.945e+02, percent-clipped=2.0 2023-04-27 12:37:36,678 INFO [finetune.py:976] (1/7) Epoch 19, batch 2200, loss[loss=0.2027, simple_loss=0.267, pruned_loss=0.0692, over 4830.00 frames. ], tot_loss[loss=0.1756, simple_loss=0.2473, pruned_loss=0.05198, over 954195.11 frames. ], batch size: 30, lr: 3.29e-03, grad_scale: 32.0 2023-04-27 12:38:49,116 INFO [finetune.py:976] (1/7) Epoch 19, batch 2250, loss[loss=0.1649, simple_loss=0.2439, pruned_loss=0.04294, over 4763.00 frames. ], tot_loss[loss=0.1766, simple_loss=0.249, pruned_loss=0.05211, over 955363.01 frames. ], batch size: 27, lr: 3.29e-03, grad_scale: 32.0 2023-04-27 12:38:59,481 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.067e+02 1.635e+02 1.938e+02 2.374e+02 3.739e+02, threshold=3.876e+02, percent-clipped=0.0 2023-04-27 12:39:34,809 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=105388.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:39:53,436 INFO [finetune.py:976] (1/7) Epoch 19, batch 2300, loss[loss=0.1766, simple_loss=0.2386, pruned_loss=0.0573, over 4220.00 frames. ], tot_loss[loss=0.1762, simple_loss=0.2492, pruned_loss=0.05162, over 954348.80 frames. ], batch size: 18, lr: 3.29e-03, grad_scale: 32.0 2023-04-27 12:40:03,682 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-04-27 12:40:49,871 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.81 vs. limit=2.0 2023-04-27 12:40:58,676 INFO [finetune.py:976] (1/7) Epoch 19, batch 2350, loss[loss=0.1377, simple_loss=0.2197, pruned_loss=0.02783, over 4722.00 frames. ], tot_loss[loss=0.1738, simple_loss=0.2461, pruned_loss=0.05079, over 954770.96 frames. ], batch size: 23, lr: 3.29e-03, grad_scale: 32.0 2023-04-27 12:40:58,800 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=105449.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 12:41:09,882 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.005e+02 1.591e+02 1.880e+02 2.234e+02 4.098e+02, threshold=3.760e+02, percent-clipped=1.0 2023-04-27 12:41:52,226 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=105486.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:42:06,554 INFO [finetune.py:976] (1/7) Epoch 19, batch 2400, loss[loss=0.187, simple_loss=0.2419, pruned_loss=0.06604, over 4908.00 frames. ], tot_loss[loss=0.1723, simple_loss=0.2437, pruned_loss=0.05045, over 955624.00 frames. ], batch size: 32, lr: 3.29e-03, grad_scale: 32.0 2023-04-27 12:42:08,508 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=105502.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:42:39,439 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=105547.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:42:40,531 INFO [finetune.py:976] (1/7) Epoch 19, batch 2450, loss[loss=0.1827, simple_loss=0.2423, pruned_loss=0.06154, over 4928.00 frames. ], tot_loss[loss=0.1713, simple_loss=0.2417, pruned_loss=0.05046, over 957321.50 frames. ], batch size: 38, lr: 3.29e-03, grad_scale: 32.0 2023-04-27 12:42:45,849 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.021e+02 1.599e+02 1.963e+02 2.373e+02 4.323e+02, threshold=3.926e+02, percent-clipped=3.0 2023-04-27 12:42:50,131 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=105563.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:43:10,630 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2023-04-27 12:43:14,474 INFO [finetune.py:976] (1/7) Epoch 19, batch 2500, loss[loss=0.1709, simple_loss=0.25, pruned_loss=0.04589, over 4759.00 frames. ], tot_loss[loss=0.1732, simple_loss=0.244, pruned_loss=0.05123, over 957852.97 frames. ], batch size: 27, lr: 3.29e-03, grad_scale: 32.0 2023-04-27 12:43:47,976 INFO [finetune.py:976] (1/7) Epoch 19, batch 2550, loss[loss=0.1555, simple_loss=0.2406, pruned_loss=0.03522, over 4802.00 frames. ], tot_loss[loss=0.1751, simple_loss=0.2471, pruned_loss=0.05151, over 956821.08 frames. ], batch size: 45, lr: 3.29e-03, grad_scale: 32.0 2023-04-27 12:43:53,308 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.125e+02 1.536e+02 1.856e+02 2.246e+02 3.753e+02, threshold=3.711e+02, percent-clipped=0.0 2023-04-27 12:43:57,033 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5928, 1.1839, 1.8025, 2.1129, 1.7024, 1.5650, 1.6841, 1.6707], device='cuda:1'), covar=tensor([0.4192, 0.5812, 0.5734, 0.5268, 0.5581, 0.7122, 0.7269, 0.7495], device='cuda:1'), in_proj_covar=tensor([0.0427, 0.0410, 0.0502, 0.0505, 0.0456, 0.0482, 0.0489, 0.0493], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 12:44:27,706 INFO [finetune.py:976] (1/7) Epoch 19, batch 2600, loss[loss=0.1954, simple_loss=0.2746, pruned_loss=0.0581, over 4818.00 frames. ], tot_loss[loss=0.178, simple_loss=0.25, pruned_loss=0.05302, over 954900.75 frames. ], batch size: 51, lr: 3.29e-03, grad_scale: 32.0 2023-04-27 12:44:36,885 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 2023-04-27 12:44:58,142 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=105744.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 12:45:01,148 INFO [finetune.py:976] (1/7) Epoch 19, batch 2650, loss[loss=0.1706, simple_loss=0.2469, pruned_loss=0.04711, over 4885.00 frames. ], tot_loss[loss=0.1794, simple_loss=0.2517, pruned_loss=0.05358, over 953744.68 frames. ], batch size: 43, lr: 3.29e-03, grad_scale: 32.0 2023-04-27 12:45:06,401 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.206e+02 1.583e+02 1.907e+02 2.231e+02 5.599e+02, threshold=3.814e+02, percent-clipped=2.0 2023-04-27 12:45:30,661 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=105792.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:45:34,813 INFO [finetune.py:976] (1/7) Epoch 19, batch 2700, loss[loss=0.1409, simple_loss=0.2113, pruned_loss=0.03526, over 4724.00 frames. ], tot_loss[loss=0.1774, simple_loss=0.2492, pruned_loss=0.05282, over 954204.87 frames. ], batch size: 54, lr: 3.29e-03, grad_scale: 32.0 2023-04-27 12:45:34,934 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=105799.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:46:05,562 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8304, 1.7395, 2.0879, 2.3086, 1.7192, 1.4593, 1.8246, 0.8832], device='cuda:1'), covar=tensor([0.0726, 0.0673, 0.0481, 0.0674, 0.0767, 0.1180, 0.0703, 0.0738], device='cuda:1'), in_proj_covar=tensor([0.0067, 0.0067, 0.0066, 0.0066, 0.0073, 0.0094, 0.0072, 0.0065], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-27 12:46:16,757 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.2671, 1.3570, 1.6988, 1.8128, 1.7577, 1.8791, 1.7733, 1.7906], device='cuda:1'), covar=tensor([0.3781, 0.5018, 0.4151, 0.3976, 0.5200, 0.6853, 0.4455, 0.4477], device='cuda:1'), in_proj_covar=tensor([0.0335, 0.0371, 0.0318, 0.0332, 0.0344, 0.0393, 0.0356, 0.0327], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 12:46:36,273 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=105842.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:46:40,481 INFO [finetune.py:976] (1/7) Epoch 19, batch 2750, loss[loss=0.1514, simple_loss=0.2243, pruned_loss=0.03927, over 4815.00 frames. ], tot_loss[loss=0.1756, simple_loss=0.2469, pruned_loss=0.05215, over 955063.20 frames. ], batch size: 41, lr: 3.29e-03, grad_scale: 32.0 2023-04-27 12:46:43,062 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=105853.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:46:45,896 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.998e+01 1.495e+02 1.843e+02 2.455e+02 4.470e+02, threshold=3.686e+02, percent-clipped=1.0 2023-04-27 12:46:46,567 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=105858.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:46:47,886 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=105860.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:47:35,994 INFO [finetune.py:976] (1/7) Epoch 19, batch 2800, loss[loss=0.143, simple_loss=0.2106, pruned_loss=0.03776, over 4892.00 frames. ], tot_loss[loss=0.1716, simple_loss=0.2424, pruned_loss=0.05041, over 954570.47 frames. ], batch size: 35, lr: 3.29e-03, grad_scale: 32.0 2023-04-27 12:47:46,774 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7242, 1.4865, 1.3511, 1.6005, 1.9777, 1.6048, 1.3545, 1.2846], device='cuda:1'), covar=tensor([0.1751, 0.1392, 0.1760, 0.1370, 0.0931, 0.1771, 0.2091, 0.2306], device='cuda:1'), in_proj_covar=tensor([0.0311, 0.0315, 0.0354, 0.0293, 0.0330, 0.0312, 0.0303, 0.0373], device='cuda:1'), out_proj_covar=tensor([6.4079e-05, 6.5286e-05, 7.5043e-05, 5.9520e-05, 6.8467e-05, 6.5564e-05, 6.3635e-05, 7.9460e-05], device='cuda:1') 2023-04-27 12:48:19,707 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-04-27 12:48:43,335 INFO [finetune.py:976] (1/7) Epoch 19, batch 2850, loss[loss=0.1643, simple_loss=0.2228, pruned_loss=0.0529, over 4149.00 frames. ], tot_loss[loss=0.1722, simple_loss=0.2423, pruned_loss=0.05103, over 953551.38 frames. ], batch size: 17, lr: 3.29e-03, grad_scale: 32.0 2023-04-27 12:48:53,758 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.059e+02 1.431e+02 1.804e+02 2.150e+02 4.441e+02, threshold=3.609e+02, percent-clipped=4.0 2023-04-27 12:48:56,285 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([4.1003, 3.8341, 2.9981, 4.7175, 3.9697, 4.0731, 1.9309, 4.0669], device='cuda:1'), covar=tensor([0.1638, 0.1060, 0.3902, 0.1056, 0.2777, 0.1657, 0.5173, 0.2254], device='cuda:1'), in_proj_covar=tensor([0.0243, 0.0215, 0.0247, 0.0306, 0.0298, 0.0247, 0.0270, 0.0270], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 12:49:47,047 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.75 vs. limit=5.0 2023-04-27 12:49:49,881 INFO [finetune.py:976] (1/7) Epoch 19, batch 2900, loss[loss=0.1616, simple_loss=0.2206, pruned_loss=0.05128, over 4193.00 frames. ], tot_loss[loss=0.1734, simple_loss=0.2441, pruned_loss=0.05132, over 952510.66 frames. ], batch size: 18, lr: 3.29e-03, grad_scale: 32.0 2023-04-27 12:50:13,506 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=2.01 vs. limit=2.0 2023-04-27 12:50:14,545 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4010, 1.6860, 1.5266, 1.8672, 1.8745, 1.9557, 1.6021, 4.2141], device='cuda:1'), covar=tensor([0.0532, 0.0757, 0.0778, 0.1197, 0.0601, 0.0581, 0.0727, 0.0119], device='cuda:1'), in_proj_covar=tensor([0.0037, 0.0039, 0.0039, 0.0043, 0.0040, 0.0038, 0.0038, 0.0056], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:1') 2023-04-27 12:50:40,769 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=106044.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:50:43,675 INFO [finetune.py:976] (1/7) Epoch 19, batch 2950, loss[loss=0.173, simple_loss=0.2448, pruned_loss=0.05058, over 4813.00 frames. ], tot_loss[loss=0.1748, simple_loss=0.2465, pruned_loss=0.05154, over 954580.57 frames. ], batch size: 25, lr: 3.29e-03, grad_scale: 32.0 2023-04-27 12:50:48,560 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.113e+02 1.617e+02 1.806e+02 2.204e+02 5.278e+02, threshold=3.611e+02, percent-clipped=1.0 2023-04-27 12:50:55,161 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([4.0021, 3.8125, 3.1256, 4.5524, 3.7935, 4.0302, 1.8671, 3.9507], device='cuda:1'), covar=tensor([0.1593, 0.1045, 0.3402, 0.1014, 0.2950, 0.1459, 0.4946, 0.1838], device='cuda:1'), in_proj_covar=tensor([0.0247, 0.0218, 0.0250, 0.0309, 0.0302, 0.0249, 0.0273, 0.0273], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 12:51:01,763 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=106077.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:51:04,763 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.0167, 1.9702, 1.7115, 1.6688, 2.0201, 1.7153, 2.4173, 1.4918], device='cuda:1'), covar=tensor([0.3288, 0.1737, 0.4339, 0.2505, 0.1581, 0.2292, 0.1402, 0.4382], device='cuda:1'), in_proj_covar=tensor([0.0335, 0.0344, 0.0425, 0.0351, 0.0380, 0.0375, 0.0370, 0.0416], device='cuda:1'), out_proj_covar=tensor([9.9650e-05, 1.0320e-04, 1.2913e-04, 1.0593e-04, 1.1356e-04, 1.1211e-04, 1.0911e-04, 1.2591e-04], device='cuda:1') 2023-04-27 12:51:12,297 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=106092.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:51:17,510 INFO [finetune.py:976] (1/7) Epoch 19, batch 3000, loss[loss=0.1427, simple_loss=0.2299, pruned_loss=0.02774, over 4820.00 frames. ], tot_loss[loss=0.175, simple_loss=0.247, pruned_loss=0.05152, over 952970.25 frames. ], batch size: 38, lr: 3.29e-03, grad_scale: 32.0 2023-04-27 12:51:17,510 INFO [finetune.py:1001] (1/7) Computing validation loss 2023-04-27 12:51:22,471 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8948, 2.1789, 1.9092, 2.1207, 1.7275, 1.8646, 1.9316, 1.4749], device='cuda:1'), covar=tensor([0.1397, 0.0957, 0.0746, 0.0993, 0.2869, 0.1055, 0.1485, 0.2121], device='cuda:1'), in_proj_covar=tensor([0.0286, 0.0303, 0.0216, 0.0279, 0.0312, 0.0258, 0.0250, 0.0265], device='cuda:1'), out_proj_covar=tensor([1.1453e-04, 1.2016e-04, 8.5849e-05, 1.1062e-04, 1.2653e-04, 1.0233e-04, 1.0096e-04, 1.0501e-04], device='cuda:1') 2023-04-27 12:51:33,614 INFO [finetune.py:1010] (1/7) Epoch 19, validation: loss=0.1523, simple_loss=0.2226, pruned_loss=0.04099, over 2265189.00 frames. 2023-04-27 12:51:33,614 INFO [finetune.py:1011] (1/7) Maximum memory allocated so far is 6391MB 2023-04-27 12:52:18,485 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=106138.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:52:26,597 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=106142.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:52:36,642 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=106148.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:52:37,150 INFO [finetune.py:976] (1/7) Epoch 19, batch 3050, loss[loss=0.1402, simple_loss=0.2195, pruned_loss=0.03044, over 4757.00 frames. ], tot_loss[loss=0.175, simple_loss=0.2478, pruned_loss=0.0511, over 953802.72 frames. ], batch size: 28, lr: 3.29e-03, grad_scale: 32.0 2023-04-27 12:52:40,349 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-04-27 12:52:46,797 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=106155.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:52:48,481 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.007e+02 1.467e+02 1.745e+02 2.097e+02 4.231e+02, threshold=3.490e+02, percent-clipped=3.0 2023-04-27 12:52:49,235 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=106158.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:53:31,155 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=106190.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:53:42,135 INFO [finetune.py:976] (1/7) Epoch 19, batch 3100, loss[loss=0.1358, simple_loss=0.209, pruned_loss=0.03128, over 4862.00 frames. ], tot_loss[loss=0.1744, simple_loss=0.2467, pruned_loss=0.05108, over 955932.55 frames. ], batch size: 31, lr: 3.29e-03, grad_scale: 32.0 2023-04-27 12:53:51,784 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=106206.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:54:35,394 INFO [finetune.py:976] (1/7) Epoch 19, batch 3150, loss[loss=0.1742, simple_loss=0.2517, pruned_loss=0.04838, over 4821.00 frames. ], tot_loss[loss=0.1728, simple_loss=0.2441, pruned_loss=0.0507, over 957934.73 frames. ], batch size: 41, lr: 3.29e-03, grad_scale: 32.0 2023-04-27 12:54:40,719 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.012e+02 1.609e+02 1.934e+02 2.307e+02 5.875e+02, threshold=3.868e+02, percent-clipped=2.0 2023-04-27 12:54:49,453 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=106268.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:55:23,648 INFO [finetune.py:976] (1/7) Epoch 19, batch 3200, loss[loss=0.1556, simple_loss=0.2239, pruned_loss=0.04368, over 4727.00 frames. ], tot_loss[loss=0.1703, simple_loss=0.2409, pruned_loss=0.04981, over 957107.50 frames. ], batch size: 54, lr: 3.28e-03, grad_scale: 32.0 2023-04-27 12:55:47,933 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0587, 2.6266, 1.1096, 1.3737, 2.0301, 1.2869, 3.2823, 1.5991], device='cuda:1'), covar=tensor([0.0666, 0.0692, 0.0740, 0.1143, 0.0421, 0.0882, 0.0204, 0.0632], device='cuda:1'), in_proj_covar=tensor([0.0051, 0.0065, 0.0048, 0.0047, 0.0050, 0.0052, 0.0074, 0.0051], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:1') 2023-04-27 12:55:49,180 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7523, 2.3944, 1.8728, 1.9073, 1.2863, 1.3519, 1.9965, 1.2544], device='cuda:1'), covar=tensor([0.1714, 0.1377, 0.1447, 0.1691, 0.2352, 0.1997, 0.0971, 0.2117], device='cuda:1'), in_proj_covar=tensor([0.0196, 0.0212, 0.0168, 0.0203, 0.0199, 0.0184, 0.0155, 0.0185], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-27 12:55:50,418 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=106329.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:55:54,735 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=106336.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:55:55,918 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0901, 2.5641, 0.9980, 1.3227, 1.9377, 1.3930, 3.1313, 1.7108], device='cuda:1'), covar=tensor([0.0659, 0.0605, 0.0744, 0.1303, 0.0466, 0.0956, 0.0327, 0.0654], device='cuda:1'), in_proj_covar=tensor([0.0051, 0.0065, 0.0048, 0.0047, 0.0050, 0.0052, 0.0074, 0.0051], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:1') 2023-04-27 12:55:55,928 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.3488, 1.5492, 1.4089, 1.7676, 1.6676, 1.7967, 1.3882, 3.1696], device='cuda:1'), covar=tensor([0.0580, 0.0758, 0.0796, 0.1206, 0.0600, 0.0459, 0.0742, 0.0166], device='cuda:1'), in_proj_covar=tensor([0.0037, 0.0038, 0.0039, 0.0043, 0.0039, 0.0038, 0.0038, 0.0055], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:1') 2023-04-27 12:56:03,600 INFO [finetune.py:976] (1/7) Epoch 19, batch 3250, loss[loss=0.1611, simple_loss=0.2394, pruned_loss=0.04143, over 4814.00 frames. ], tot_loss[loss=0.17, simple_loss=0.2404, pruned_loss=0.04978, over 952472.42 frames. ], batch size: 45, lr: 3.28e-03, grad_scale: 32.0 2023-04-27 12:56:08,529 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.061e+02 1.638e+02 1.854e+02 2.216e+02 4.860e+02, threshold=3.708e+02, percent-clipped=2.0 2023-04-27 12:56:23,517 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.6067, 2.6253, 2.0013, 3.0063, 2.5120, 2.6705, 1.1148, 2.5530], device='cuda:1'), covar=tensor([0.2080, 0.1404, 0.2787, 0.2346, 0.3329, 0.1847, 0.5240, 0.2711], device='cuda:1'), in_proj_covar=tensor([0.0246, 0.0216, 0.0249, 0.0309, 0.0300, 0.0248, 0.0272, 0.0271], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 12:56:30,998 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7999, 1.7663, 2.1378, 2.3226, 1.6433, 1.4065, 1.8474, 1.1517], device='cuda:1'), covar=tensor([0.0672, 0.0736, 0.0546, 0.0977, 0.0890, 0.1344, 0.0861, 0.0775], device='cuda:1'), in_proj_covar=tensor([0.0068, 0.0068, 0.0067, 0.0067, 0.0075, 0.0096, 0.0073, 0.0066], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-27 12:56:35,791 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=106397.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:56:37,423 INFO [finetune.py:976] (1/7) Epoch 19, batch 3300, loss[loss=0.2096, simple_loss=0.2884, pruned_loss=0.06542, over 4237.00 frames. ], tot_loss[loss=0.1728, simple_loss=0.2439, pruned_loss=0.05088, over 952501.41 frames. ], batch size: 66, lr: 3.28e-03, grad_scale: 32.0 2023-04-27 12:57:00,443 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=106433.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:57:01,843 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 2023-04-27 12:57:09,665 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=106448.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:57:10,217 INFO [finetune.py:976] (1/7) Epoch 19, batch 3350, loss[loss=0.1372, simple_loss=0.2202, pruned_loss=0.0271, over 4838.00 frames. ], tot_loss[loss=0.173, simple_loss=0.2445, pruned_loss=0.05075, over 951589.27 frames. ], batch size: 49, lr: 3.28e-03, grad_scale: 32.0 2023-04-27 12:57:14,359 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=106454.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:57:14,934 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=106455.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:57:16,069 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.216e+02 1.680e+02 1.880e+02 2.311e+02 4.481e+02, threshold=3.759e+02, percent-clipped=2.0 2023-04-27 12:57:42,140 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=106496.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:57:43,907 INFO [finetune.py:976] (1/7) Epoch 19, batch 3400, loss[loss=0.1637, simple_loss=0.2271, pruned_loss=0.05012, over 4684.00 frames. ], tot_loss[loss=0.1741, simple_loss=0.2459, pruned_loss=0.05111, over 952651.50 frames. ], batch size: 23, lr: 3.28e-03, grad_scale: 32.0 2023-04-27 12:57:45,891 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7883, 2.1618, 0.8929, 1.2027, 1.6720, 1.1932, 2.4628, 1.3712], device='cuda:1'), covar=tensor([0.0728, 0.0589, 0.0670, 0.1211, 0.0429, 0.0945, 0.0307, 0.0693], device='cuda:1'), in_proj_covar=tensor([0.0051, 0.0066, 0.0048, 0.0047, 0.0050, 0.0052, 0.0074, 0.0051], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:1') 2023-04-27 12:57:46,912 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=106503.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:57:49,928 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8392, 1.4213, 1.4909, 1.6662, 2.0227, 1.6089, 1.3582, 1.3712], device='cuda:1'), covar=tensor([0.1858, 0.1549, 0.1914, 0.1351, 0.0924, 0.1794, 0.2211, 0.2532], device='cuda:1'), in_proj_covar=tensor([0.0308, 0.0312, 0.0350, 0.0289, 0.0328, 0.0308, 0.0301, 0.0370], device='cuda:1'), out_proj_covar=tensor([6.3414e-05, 6.4597e-05, 7.4091e-05, 5.8565e-05, 6.8057e-05, 6.4728e-05, 6.3069e-05, 7.8895e-05], device='cuda:1') 2023-04-27 12:57:54,776 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=106515.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:58:17,776 INFO [finetune.py:976] (1/7) Epoch 19, batch 3450, loss[loss=0.1728, simple_loss=0.2343, pruned_loss=0.05571, over 4879.00 frames. ], tot_loss[loss=0.1741, simple_loss=0.2461, pruned_loss=0.0511, over 955180.31 frames. ], batch size: 35, lr: 3.28e-03, grad_scale: 32.0 2023-04-27 12:58:23,137 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.135e+02 1.592e+02 1.859e+02 2.190e+02 3.296e+02, threshold=3.717e+02, percent-clipped=0.0 2023-04-27 12:58:57,661 INFO [finetune.py:976] (1/7) Epoch 19, batch 3500, loss[loss=0.2017, simple_loss=0.2672, pruned_loss=0.06812, over 4852.00 frames. ], tot_loss[loss=0.1735, simple_loss=0.2446, pruned_loss=0.0512, over 955994.21 frames. ], batch size: 44, lr: 3.28e-03, grad_scale: 32.0 2023-04-27 12:59:14,552 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=106624.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:59:36,355 INFO [finetune.py:976] (1/7) Epoch 19, batch 3550, loss[loss=0.1729, simple_loss=0.2353, pruned_loss=0.05529, over 4757.00 frames. ], tot_loss[loss=0.1717, simple_loss=0.2421, pruned_loss=0.05062, over 956598.23 frames. ], batch size: 27, lr: 3.28e-03, grad_scale: 32.0 2023-04-27 12:59:41,156 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.386e+01 1.482e+02 1.819e+02 2.170e+02 5.506e+02, threshold=3.638e+02, percent-clipped=1.0 2023-04-27 13:00:05,459 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=106692.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:00:14,310 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-04-27 13:00:14,734 INFO [finetune.py:976] (1/7) Epoch 19, batch 3600, loss[loss=0.2178, simple_loss=0.2841, pruned_loss=0.07574, over 4938.00 frames. ], tot_loss[loss=0.1702, simple_loss=0.2402, pruned_loss=0.05011, over 957061.46 frames. ], batch size: 38, lr: 3.28e-03, grad_scale: 32.0 2023-04-27 13:00:58,215 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=106733.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:01:06,581 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5200, 3.7564, 0.9531, 1.8580, 2.0091, 2.6539, 2.1227, 1.1246], device='cuda:1'), covar=tensor([0.1395, 0.0829, 0.2011, 0.1238, 0.1056, 0.0916, 0.1475, 0.1900], device='cuda:1'), in_proj_covar=tensor([0.0116, 0.0240, 0.0137, 0.0119, 0.0131, 0.0151, 0.0115, 0.0118], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-27 13:01:14,588 INFO [finetune.py:976] (1/7) Epoch 19, batch 3650, loss[loss=0.1681, simple_loss=0.2494, pruned_loss=0.04341, over 4836.00 frames. ], tot_loss[loss=0.1722, simple_loss=0.2429, pruned_loss=0.05076, over 956253.86 frames. ], batch size: 49, lr: 3.28e-03, grad_scale: 32.0 2023-04-27 13:01:19,420 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.799e+01 1.618e+02 2.033e+02 2.562e+02 4.341e+02, threshold=4.066e+02, percent-clipped=3.0 2023-04-27 13:01:35,794 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=106781.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:01:48,135 INFO [finetune.py:976] (1/7) Epoch 19, batch 3700, loss[loss=0.174, simple_loss=0.2551, pruned_loss=0.04641, over 4818.00 frames. ], tot_loss[loss=0.1736, simple_loss=0.2453, pruned_loss=0.05095, over 956393.44 frames. ], batch size: 40, lr: 3.28e-03, grad_scale: 32.0 2023-04-27 13:01:55,079 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=106810.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:02:10,605 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6940, 1.3499, 1.8281, 2.2031, 1.8286, 1.6801, 1.7607, 1.7215], device='cuda:1'), covar=tensor([0.4537, 0.6772, 0.6329, 0.5562, 0.5649, 0.8345, 0.8142, 0.8705], device='cuda:1'), in_proj_covar=tensor([0.0427, 0.0409, 0.0502, 0.0506, 0.0455, 0.0483, 0.0489, 0.0495], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 13:02:22,251 INFO [finetune.py:976] (1/7) Epoch 19, batch 3750, loss[loss=0.1918, simple_loss=0.2579, pruned_loss=0.06289, over 4811.00 frames. ], tot_loss[loss=0.1761, simple_loss=0.2478, pruned_loss=0.05222, over 956858.93 frames. ], batch size: 38, lr: 3.28e-03, grad_scale: 32.0 2023-04-27 13:02:27,098 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.601e+01 1.567e+02 1.830e+02 2.266e+02 4.856e+02, threshold=3.659e+02, percent-clipped=1.0 2023-04-27 13:02:47,406 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.1078, 2.5533, 0.8557, 1.4640, 1.4990, 1.8512, 1.6346, 0.7912], device='cuda:1'), covar=tensor([0.1649, 0.1235, 0.1855, 0.1389, 0.1218, 0.1070, 0.1687, 0.1803], device='cuda:1'), in_proj_covar=tensor([0.0116, 0.0240, 0.0137, 0.0119, 0.0131, 0.0151, 0.0115, 0.0119], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-27 13:02:55,379 INFO [finetune.py:976] (1/7) Epoch 19, batch 3800, loss[loss=0.2149, simple_loss=0.2884, pruned_loss=0.07072, over 4725.00 frames. ], tot_loss[loss=0.1783, simple_loss=0.25, pruned_loss=0.05332, over 956675.40 frames. ], batch size: 54, lr: 3.28e-03, grad_scale: 32.0 2023-04-27 13:03:06,019 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2023-04-27 13:03:10,668 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=106924.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:03:20,089 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-04-27 13:03:26,500 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([3.2842, 3.2256, 2.3876, 3.7985, 3.1882, 3.2777, 1.3365, 3.2876], device='cuda:1'), covar=tensor([0.1936, 0.1395, 0.3524, 0.2401, 0.3132, 0.1869, 0.5565, 0.2428], device='cuda:1'), in_proj_covar=tensor([0.0244, 0.0215, 0.0248, 0.0307, 0.0297, 0.0247, 0.0270, 0.0270], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 13:03:27,654 INFO [finetune.py:976] (1/7) Epoch 19, batch 3850, loss[loss=0.1694, simple_loss=0.2463, pruned_loss=0.04628, over 4815.00 frames. ], tot_loss[loss=0.1767, simple_loss=0.2481, pruned_loss=0.05262, over 956478.10 frames. ], batch size: 38, lr: 3.28e-03, grad_scale: 32.0 2023-04-27 13:03:33,087 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.149e+02 1.551e+02 1.789e+02 2.154e+02 4.176e+02, threshold=3.578e+02, percent-clipped=2.0 2023-04-27 13:03:42,311 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=106972.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:03:56,025 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=106992.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:04:00,617 INFO [finetune.py:976] (1/7) Epoch 19, batch 3900, loss[loss=0.1571, simple_loss=0.2217, pruned_loss=0.04623, over 4791.00 frames. ], tot_loss[loss=0.174, simple_loss=0.2448, pruned_loss=0.05157, over 955900.05 frames. ], batch size: 29, lr: 3.28e-03, grad_scale: 32.0 2023-04-27 13:04:33,281 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=107040.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:04:33,295 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([3.9663, 4.0809, 2.8033, 4.6373, 4.0804, 4.0275, 1.8666, 3.9252], device='cuda:1'), covar=tensor([0.1745, 0.1330, 0.3080, 0.1634, 0.3337, 0.1773, 0.6152, 0.2327], device='cuda:1'), in_proj_covar=tensor([0.0245, 0.0216, 0.0249, 0.0307, 0.0299, 0.0247, 0.0271, 0.0270], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 13:04:41,523 INFO [finetune.py:976] (1/7) Epoch 19, batch 3950, loss[loss=0.1744, simple_loss=0.2448, pruned_loss=0.05199, over 4913.00 frames. ], tot_loss[loss=0.1715, simple_loss=0.2419, pruned_loss=0.05057, over 956671.37 frames. ], batch size: 35, lr: 3.28e-03, grad_scale: 32.0 2023-04-27 13:04:54,081 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.559e+01 1.455e+02 1.767e+02 2.125e+02 3.704e+02, threshold=3.534e+02, percent-clipped=1.0 2023-04-27 13:05:26,567 INFO [finetune.py:976] (1/7) Epoch 19, batch 4000, loss[loss=0.2405, simple_loss=0.3095, pruned_loss=0.08581, over 4837.00 frames. ], tot_loss[loss=0.1716, simple_loss=0.2416, pruned_loss=0.05076, over 955002.30 frames. ], batch size: 49, lr: 3.28e-03, grad_scale: 32.0 2023-04-27 13:05:35,403 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=107110.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:06:09,674 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0404, 1.7263, 2.0429, 2.3328, 2.4452, 1.9573, 1.5542, 2.1170], device='cuda:1'), covar=tensor([0.0818, 0.1110, 0.0601, 0.0595, 0.0551, 0.0871, 0.0796, 0.0602], device='cuda:1'), in_proj_covar=tensor([0.0186, 0.0201, 0.0182, 0.0173, 0.0175, 0.0181, 0.0151, 0.0179], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 13:06:15,749 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=107140.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 13:06:28,916 INFO [finetune.py:976] (1/7) Epoch 19, batch 4050, loss[loss=0.1706, simple_loss=0.248, pruned_loss=0.04663, over 4814.00 frames. ], tot_loss[loss=0.1751, simple_loss=0.2458, pruned_loss=0.05221, over 955615.48 frames. ], batch size: 51, lr: 3.28e-03, grad_scale: 32.0 2023-04-27 13:06:30,808 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=107152.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:06:35,807 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.479e+01 1.702e+02 1.942e+02 2.364e+02 4.795e+02, threshold=3.885e+02, percent-clipped=4.0 2023-04-27 13:06:36,471 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=107158.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:06:49,948 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5686, 1.2348, 0.3164, 1.2325, 1.0955, 1.4508, 1.3374, 1.3553], device='cuda:1'), covar=tensor([0.0532, 0.0397, 0.0432, 0.0596, 0.0308, 0.0522, 0.0499, 0.0561], device='cuda:1'), in_proj_covar=tensor([0.0028, 0.0024, 0.0022, 0.0029, 0.0019, 0.0028, 0.0028, 0.0028], device='cuda:1'), out_proj_covar=tensor([0.0050, 0.0045, 0.0038, 0.0051, 0.0038, 0.0049, 0.0049, 0.0050], device='cuda:1') 2023-04-27 13:06:59,448 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.2905, 2.8691, 1.1633, 1.5706, 2.5109, 1.3992, 3.7718, 2.0780], device='cuda:1'), covar=tensor([0.0617, 0.0581, 0.0761, 0.1240, 0.0456, 0.0995, 0.0255, 0.0617], device='cuda:1'), in_proj_covar=tensor([0.0051, 0.0065, 0.0048, 0.0047, 0.0050, 0.0052, 0.0074, 0.0051], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:1') 2023-04-27 13:07:15,574 INFO [finetune.py:976] (1/7) Epoch 19, batch 4100, loss[loss=0.168, simple_loss=0.2448, pruned_loss=0.04561, over 4897.00 frames. ], tot_loss[loss=0.1768, simple_loss=0.2478, pruned_loss=0.05288, over 954568.49 frames. ], batch size: 43, lr: 3.28e-03, grad_scale: 64.0 2023-04-27 13:07:17,380 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=107201.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 13:07:26,168 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=107213.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 13:07:34,501 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.2702, 1.3576, 1.3730, 2.1820, 2.1779, 1.8271, 1.8955, 1.3557], device='cuda:1'), covar=tensor([0.1573, 0.1788, 0.1768, 0.1227, 0.1032, 0.1796, 0.1751, 0.2315], device='cuda:1'), in_proj_covar=tensor([0.0306, 0.0310, 0.0347, 0.0288, 0.0326, 0.0306, 0.0297, 0.0368], device='cuda:1'), out_proj_covar=tensor([6.3017e-05, 6.4373e-05, 7.3557e-05, 5.8316e-05, 6.7604e-05, 6.4262e-05, 6.2342e-05, 7.8451e-05], device='cuda:1') 2023-04-27 13:07:48,307 INFO [finetune.py:976] (1/7) Epoch 19, batch 4150, loss[loss=0.1833, simple_loss=0.2446, pruned_loss=0.06095, over 4875.00 frames. ], tot_loss[loss=0.1786, simple_loss=0.2493, pruned_loss=0.05391, over 954805.11 frames. ], batch size: 34, lr: 3.28e-03, grad_scale: 64.0 2023-04-27 13:07:54,579 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.951e+01 1.595e+02 1.845e+02 2.201e+02 3.820e+02, threshold=3.690e+02, percent-clipped=0.0 2023-04-27 13:08:22,019 INFO [finetune.py:976] (1/7) Epoch 19, batch 4200, loss[loss=0.1887, simple_loss=0.2637, pruned_loss=0.05688, over 4291.00 frames. ], tot_loss[loss=0.178, simple_loss=0.2495, pruned_loss=0.05326, over 954607.45 frames. ], batch size: 66, lr: 3.28e-03, grad_scale: 64.0 2023-04-27 13:08:28,357 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.1955, 1.4127, 1.3416, 1.6368, 1.6118, 1.6764, 1.3521, 3.0525], device='cuda:1'), covar=tensor([0.0631, 0.0804, 0.0783, 0.1247, 0.0637, 0.0506, 0.0742, 0.0156], device='cuda:1'), in_proj_covar=tensor([0.0037, 0.0038, 0.0039, 0.0043, 0.0039, 0.0038, 0.0038, 0.0056], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:1') 2023-04-27 13:08:30,224 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([0.9218, 1.2266, 3.3085, 3.0613, 2.9746, 3.2656, 3.1798, 2.9075], device='cuda:1'), covar=tensor([0.7224, 0.5645, 0.1412, 0.2146, 0.1408, 0.2158, 0.1749, 0.1614], device='cuda:1'), in_proj_covar=tensor([0.0305, 0.0301, 0.0404, 0.0402, 0.0346, 0.0402, 0.0310, 0.0363], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-27 13:08:42,624 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9108, 1.6306, 1.4667, 1.8711, 2.2109, 1.7888, 1.5078, 1.3692], device='cuda:1'), covar=tensor([0.1737, 0.1420, 0.1827, 0.1312, 0.0954, 0.1502, 0.2004, 0.2445], device='cuda:1'), in_proj_covar=tensor([0.0305, 0.0310, 0.0347, 0.0287, 0.0325, 0.0305, 0.0297, 0.0367], device='cuda:1'), out_proj_covar=tensor([6.2790e-05, 6.4244e-05, 7.3438e-05, 5.8045e-05, 6.7460e-05, 6.4165e-05, 6.2161e-05, 7.8231e-05], device='cuda:1') 2023-04-27 13:08:53,617 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4334, 3.4057, 0.8337, 1.6688, 1.8186, 2.4865, 1.8909, 1.0912], device='cuda:1'), covar=tensor([0.1485, 0.1012, 0.2142, 0.1412, 0.1169, 0.0974, 0.1479, 0.1975], device='cuda:1'), in_proj_covar=tensor([0.0116, 0.0242, 0.0138, 0.0119, 0.0132, 0.0152, 0.0116, 0.0120], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-27 13:08:54,801 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([0.9330, 1.1698, 3.2652, 3.0174, 2.9536, 3.1612, 3.1341, 2.8847], device='cuda:1'), covar=tensor([0.7073, 0.5494, 0.1525, 0.2280, 0.1446, 0.1897, 0.2145, 0.1831], device='cuda:1'), in_proj_covar=tensor([0.0306, 0.0302, 0.0404, 0.0403, 0.0347, 0.0403, 0.0311, 0.0364], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-27 13:08:55,963 INFO [finetune.py:976] (1/7) Epoch 19, batch 4250, loss[loss=0.1694, simple_loss=0.2377, pruned_loss=0.05053, over 4796.00 frames. ], tot_loss[loss=0.1753, simple_loss=0.2466, pruned_loss=0.052, over 954760.40 frames. ], batch size: 25, lr: 3.28e-03, grad_scale: 64.0 2023-04-27 13:09:01,365 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.574e+01 1.612e+02 1.786e+02 2.196e+02 3.792e+02, threshold=3.571e+02, percent-clipped=1.0 2023-04-27 13:09:29,662 INFO [finetune.py:976] (1/7) Epoch 19, batch 4300, loss[loss=0.1896, simple_loss=0.2475, pruned_loss=0.06583, over 4787.00 frames. ], tot_loss[loss=0.1746, simple_loss=0.2448, pruned_loss=0.05215, over 955509.81 frames. ], batch size: 29, lr: 3.28e-03, grad_scale: 32.0 2023-04-27 13:09:32,918 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4856, 1.0711, 0.3669, 1.1762, 1.0077, 1.3699, 1.3001, 1.2317], device='cuda:1'), covar=tensor([0.0514, 0.0409, 0.0416, 0.0566, 0.0298, 0.0512, 0.0501, 0.0576], device='cuda:1'), in_proj_covar=tensor([0.0028, 0.0024, 0.0022, 0.0029, 0.0019, 0.0028, 0.0028, 0.0028], device='cuda:1'), out_proj_covar=tensor([0.0050, 0.0045, 0.0038, 0.0052, 0.0038, 0.0049, 0.0049, 0.0051], device='cuda:1') 2023-04-27 13:10:17,267 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=5.18 vs. limit=5.0 2023-04-27 13:10:28,297 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5809, 1.8882, 1.8162, 2.1381, 2.2309, 2.0988, 1.7422, 4.4653], device='cuda:1'), covar=tensor([0.0467, 0.0743, 0.0705, 0.1128, 0.0529, 0.0490, 0.0692, 0.0100], device='cuda:1'), in_proj_covar=tensor([0.0037, 0.0039, 0.0039, 0.0043, 0.0039, 0.0038, 0.0038, 0.0056], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:1') 2023-04-27 13:10:28,322 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5034, 2.5657, 2.1798, 2.3320, 2.5334, 2.3121, 3.4936, 1.9465], device='cuda:1'), covar=tensor([0.3596, 0.2243, 0.3921, 0.3085, 0.1897, 0.2493, 0.1475, 0.4120], device='cuda:1'), in_proj_covar=tensor([0.0336, 0.0343, 0.0426, 0.0351, 0.0379, 0.0375, 0.0370, 0.0415], device='cuda:1'), out_proj_covar=tensor([9.9781e-05, 1.0300e-04, 1.2947e-04, 1.0585e-04, 1.1323e-04, 1.1211e-04, 1.0899e-04, 1.2577e-04], device='cuda:1') 2023-04-27 13:10:31,834 INFO [finetune.py:976] (1/7) Epoch 19, batch 4350, loss[loss=0.2313, simple_loss=0.2892, pruned_loss=0.08667, over 4809.00 frames. ], tot_loss[loss=0.1713, simple_loss=0.2413, pruned_loss=0.05068, over 955145.43 frames. ], batch size: 40, lr: 3.28e-03, grad_scale: 32.0 2023-04-27 13:10:37,327 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.020e+02 1.596e+02 1.862e+02 2.295e+02 4.192e+02, threshold=3.723e+02, percent-clipped=1.0 2023-04-27 13:11:03,028 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=107496.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 13:11:04,783 INFO [finetune.py:976] (1/7) Epoch 19, batch 4400, loss[loss=0.2179, simple_loss=0.261, pruned_loss=0.08741, over 3930.00 frames. ], tot_loss[loss=0.1714, simple_loss=0.2414, pruned_loss=0.05073, over 955082.17 frames. ], batch size: 17, lr: 3.27e-03, grad_scale: 32.0 2023-04-27 13:11:10,415 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=107508.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 13:11:11,650 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=107510.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:11:49,532 INFO [finetune.py:976] (1/7) Epoch 19, batch 4450, loss[loss=0.2205, simple_loss=0.2876, pruned_loss=0.07669, over 4858.00 frames. ], tot_loss[loss=0.1743, simple_loss=0.245, pruned_loss=0.0518, over 954828.34 frames. ], batch size: 49, lr: 3.27e-03, grad_scale: 32.0 2023-04-27 13:12:00,085 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.014e+02 1.632e+02 1.928e+02 2.326e+02 3.887e+02, threshold=3.856e+02, percent-clipped=1.0 2023-04-27 13:12:07,862 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.2198, 1.5016, 5.4567, 5.1614, 4.7190, 5.2609, 4.8020, 4.8476], device='cuda:1'), covar=tensor([0.6206, 0.5993, 0.0890, 0.1621, 0.0983, 0.1946, 0.1036, 0.1385], device='cuda:1'), in_proj_covar=tensor([0.0307, 0.0303, 0.0405, 0.0403, 0.0348, 0.0405, 0.0312, 0.0364], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-27 13:12:18,640 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=107571.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:12:43,863 INFO [finetune.py:976] (1/7) Epoch 19, batch 4500, loss[loss=0.1813, simple_loss=0.2484, pruned_loss=0.05707, over 4890.00 frames. ], tot_loss[loss=0.1757, simple_loss=0.2472, pruned_loss=0.05214, over 954995.33 frames. ], batch size: 32, lr: 3.27e-03, grad_scale: 32.0 2023-04-27 13:13:13,496 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.2644, 1.6381, 2.1493, 2.6861, 2.1157, 1.6897, 1.4107, 2.0547], device='cuda:1'), covar=tensor([0.3297, 0.3257, 0.1566, 0.2187, 0.2779, 0.2612, 0.4258, 0.1976], device='cuda:1'), in_proj_covar=tensor([0.0294, 0.0248, 0.0227, 0.0317, 0.0219, 0.0233, 0.0228, 0.0185], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-27 13:13:17,593 INFO [finetune.py:976] (1/7) Epoch 19, batch 4550, loss[loss=0.1627, simple_loss=0.2358, pruned_loss=0.04481, over 4746.00 frames. ], tot_loss[loss=0.176, simple_loss=0.2475, pruned_loss=0.05223, over 952380.85 frames. ], batch size: 27, lr: 3.27e-03, grad_scale: 32.0 2023-04-27 13:13:20,751 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4790, 1.8145, 1.4023, 1.2796, 1.2145, 1.1630, 1.4220, 1.1427], device='cuda:1'), covar=tensor([0.1312, 0.1134, 0.1213, 0.1495, 0.1961, 0.1598, 0.0846, 0.1742], device='cuda:1'), in_proj_covar=tensor([0.0196, 0.0209, 0.0167, 0.0202, 0.0199, 0.0183, 0.0155, 0.0184], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-27 13:13:23,034 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.130e+02 1.628e+02 1.814e+02 2.190e+02 5.311e+02, threshold=3.629e+02, percent-clipped=2.0 2023-04-27 13:13:23,738 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=107659.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:13:33,270 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.3194, 2.8714, 1.0357, 1.5630, 2.5094, 1.4871, 4.2375, 2.1395], device='cuda:1'), covar=tensor([0.0699, 0.0723, 0.0824, 0.1325, 0.0461, 0.1030, 0.0197, 0.0617], device='cuda:1'), in_proj_covar=tensor([0.0051, 0.0065, 0.0048, 0.0046, 0.0050, 0.0052, 0.0074, 0.0051], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:1') 2023-04-27 13:13:42,989 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6041, 1.5100, 1.9293, 1.9784, 1.4936, 1.3714, 1.6438, 0.9755], device='cuda:1'), covar=tensor([0.0508, 0.0662, 0.0371, 0.0621, 0.0813, 0.1071, 0.0578, 0.0704], device='cuda:1'), in_proj_covar=tensor([0.0068, 0.0069, 0.0067, 0.0067, 0.0075, 0.0096, 0.0074, 0.0066], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-27 13:13:45,798 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8810, 2.4826, 1.9858, 1.9108, 1.4120, 1.4361, 2.0744, 1.3544], device='cuda:1'), covar=tensor([0.1711, 0.1305, 0.1403, 0.1723, 0.2266, 0.1884, 0.0937, 0.2011], device='cuda:1'), in_proj_covar=tensor([0.0195, 0.0209, 0.0167, 0.0202, 0.0199, 0.0183, 0.0154, 0.0184], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-27 13:13:51,126 INFO [finetune.py:976] (1/7) Epoch 19, batch 4600, loss[loss=0.1412, simple_loss=0.1958, pruned_loss=0.0433, over 4248.00 frames. ], tot_loss[loss=0.1763, simple_loss=0.2476, pruned_loss=0.05248, over 951572.65 frames. ], batch size: 18, lr: 3.27e-03, grad_scale: 32.0 2023-04-27 13:13:59,970 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-04-27 13:14:04,015 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=107720.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:14:04,058 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9776, 1.7625, 2.1812, 2.5200, 2.0299, 1.9867, 2.0684, 2.0147], device='cuda:1'), covar=tensor([0.5008, 0.7276, 0.8147, 0.5806, 0.6380, 0.8838, 0.9128, 0.9995], device='cuda:1'), in_proj_covar=tensor([0.0429, 0.0411, 0.0505, 0.0506, 0.0455, 0.0483, 0.0490, 0.0497], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 13:14:08,153 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.2862, 3.0353, 0.8433, 1.6625, 1.7269, 2.1050, 1.7664, 0.9935], device='cuda:1'), covar=tensor([0.1552, 0.1156, 0.2041, 0.1346, 0.1180, 0.1076, 0.1556, 0.1880], device='cuda:1'), in_proj_covar=tensor([0.0117, 0.0242, 0.0138, 0.0119, 0.0132, 0.0152, 0.0117, 0.0120], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-27 13:14:11,826 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-04-27 13:14:16,206 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=107737.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:14:20,926 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-04-27 13:14:22,086 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-04-27 13:14:24,297 INFO [finetune.py:976] (1/7) Epoch 19, batch 4650, loss[loss=0.1602, simple_loss=0.2338, pruned_loss=0.04326, over 4713.00 frames. ], tot_loss[loss=0.1746, simple_loss=0.2451, pruned_loss=0.05201, over 952688.40 frames. ], batch size: 59, lr: 3.27e-03, grad_scale: 32.0 2023-04-27 13:14:29,768 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.154e+02 1.588e+02 1.876e+02 2.187e+02 4.173e+02, threshold=3.752e+02, percent-clipped=2.0 2023-04-27 13:15:06,673 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=107796.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 13:15:07,908 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=107798.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 13:15:08,401 INFO [finetune.py:976] (1/7) Epoch 19, batch 4700, loss[loss=0.1731, simple_loss=0.2417, pruned_loss=0.05226, over 4843.00 frames. ], tot_loss[loss=0.1716, simple_loss=0.2416, pruned_loss=0.05083, over 952837.29 frames. ], batch size: 30, lr: 3.27e-03, grad_scale: 32.0 2023-04-27 13:15:19,599 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=107808.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:15:44,959 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.3002, 3.0515, 2.5152, 2.8222, 2.1615, 2.4495, 2.6420, 1.8966], device='cuda:1'), covar=tensor([0.2198, 0.1196, 0.0740, 0.1165, 0.3035, 0.1208, 0.1934, 0.2930], device='cuda:1'), in_proj_covar=tensor([0.0289, 0.0303, 0.0217, 0.0279, 0.0313, 0.0260, 0.0252, 0.0267], device='cuda:1'), out_proj_covar=tensor([1.1558e-04, 1.2038e-04, 8.5945e-05, 1.1059e-04, 1.2705e-04, 1.0309e-04, 1.0166e-04, 1.0573e-04], device='cuda:1') 2023-04-27 13:15:48,345 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=107844.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 13:15:52,874 INFO [finetune.py:976] (1/7) Epoch 19, batch 4750, loss[loss=0.2055, simple_loss=0.2613, pruned_loss=0.07481, over 4130.00 frames. ], tot_loss[loss=0.1709, simple_loss=0.2404, pruned_loss=0.05069, over 953030.67 frames. ], batch size: 65, lr: 3.27e-03, grad_scale: 32.0 2023-04-27 13:15:57,099 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=107856.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:15:58,278 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.851e+01 1.548e+02 1.806e+02 2.304e+02 4.398e+02, threshold=3.612e+02, percent-clipped=1.0 2023-04-27 13:16:09,257 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=107866.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:16:12,400 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7229, 1.3177, 1.8882, 2.2167, 1.8164, 1.7693, 1.8251, 1.8210], device='cuda:1'), covar=tensor([0.4719, 0.6830, 0.6766, 0.6209, 0.6264, 0.7863, 0.8258, 0.8602], device='cuda:1'), in_proj_covar=tensor([0.0429, 0.0411, 0.0505, 0.0506, 0.0456, 0.0483, 0.0490, 0.0498], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 13:16:53,228 INFO [finetune.py:976] (1/7) Epoch 19, batch 4800, loss[loss=0.1648, simple_loss=0.2397, pruned_loss=0.04494, over 4822.00 frames. ], tot_loss[loss=0.171, simple_loss=0.2415, pruned_loss=0.05024, over 953047.48 frames. ], batch size: 51, lr: 3.27e-03, grad_scale: 32.0 2023-04-27 13:16:53,903 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5892, 1.2156, 1.7641, 2.1139, 1.6994, 1.6457, 1.6921, 1.6593], device='cuda:1'), covar=tensor([0.4443, 0.6388, 0.5812, 0.5533, 0.5469, 0.7374, 0.7301, 0.8146], device='cuda:1'), in_proj_covar=tensor([0.0428, 0.0410, 0.0504, 0.0506, 0.0456, 0.0483, 0.0489, 0.0497], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 13:17:31,844 INFO [finetune.py:976] (1/7) Epoch 19, batch 4850, loss[loss=0.1396, simple_loss=0.2024, pruned_loss=0.03838, over 4381.00 frames. ], tot_loss[loss=0.1753, simple_loss=0.2463, pruned_loss=0.0522, over 954506.57 frames. ], batch size: 19, lr: 3.27e-03, grad_scale: 32.0 2023-04-27 13:17:37,759 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.098e+02 1.607e+02 1.917e+02 2.218e+02 6.571e+02, threshold=3.834e+02, percent-clipped=2.0 2023-04-27 13:17:39,085 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.1800, 2.1125, 1.8495, 1.9470, 2.3133, 1.7983, 2.7350, 1.6153], device='cuda:1'), covar=tensor([0.3988, 0.2223, 0.4331, 0.2915, 0.1688, 0.2650, 0.1599, 0.3907], device='cuda:1'), in_proj_covar=tensor([0.0337, 0.0344, 0.0423, 0.0351, 0.0379, 0.0375, 0.0370, 0.0413], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 13:18:04,093 INFO [finetune.py:976] (1/7) Epoch 19, batch 4900, loss[loss=0.19, simple_loss=0.2588, pruned_loss=0.06054, over 4793.00 frames. ], tot_loss[loss=0.1762, simple_loss=0.2473, pruned_loss=0.05258, over 955142.84 frames. ], batch size: 25, lr: 3.27e-03, grad_scale: 32.0 2023-04-27 13:18:16,701 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=108015.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:18:38,323 INFO [finetune.py:976] (1/7) Epoch 19, batch 4950, loss[loss=0.1863, simple_loss=0.2465, pruned_loss=0.06308, over 4234.00 frames. ], tot_loss[loss=0.176, simple_loss=0.2474, pruned_loss=0.05229, over 953564.64 frames. ], batch size: 65, lr: 3.27e-03, grad_scale: 32.0 2023-04-27 13:18:45,329 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.105e+02 1.581e+02 1.882e+02 2.406e+02 4.521e+02, threshold=3.765e+02, percent-clipped=4.0 2023-04-27 13:19:01,896 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=3.90 vs. limit=5.0 2023-04-27 13:19:07,753 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=108093.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 13:19:11,334 INFO [finetune.py:976] (1/7) Epoch 19, batch 5000, loss[loss=0.1715, simple_loss=0.2405, pruned_loss=0.0513, over 4753.00 frames. ], tot_loss[loss=0.1741, simple_loss=0.2455, pruned_loss=0.0513, over 954575.43 frames. ], batch size: 26, lr: 3.27e-03, grad_scale: 32.0 2023-04-27 13:19:37,020 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.9441, 2.1154, 2.0766, 2.1788, 2.0100, 2.0756, 2.2501, 2.1452], device='cuda:1'), covar=tensor([0.3683, 0.5594, 0.4641, 0.4210, 0.5233, 0.6789, 0.5311, 0.4938], device='cuda:1'), in_proj_covar=tensor([0.0337, 0.0373, 0.0322, 0.0335, 0.0345, 0.0394, 0.0358, 0.0328], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 13:19:44,759 INFO [finetune.py:976] (1/7) Epoch 19, batch 5050, loss[loss=0.178, simple_loss=0.244, pruned_loss=0.05604, over 4918.00 frames. ], tot_loss[loss=0.1735, simple_loss=0.2442, pruned_loss=0.0514, over 954683.49 frames. ], batch size: 37, lr: 3.27e-03, grad_scale: 32.0 2023-04-27 13:19:46,674 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.0810, 2.5249, 0.8427, 1.5096, 1.6005, 1.8512, 1.6414, 0.8239], device='cuda:1'), covar=tensor([0.1566, 0.1256, 0.1901, 0.1304, 0.1123, 0.0968, 0.1605, 0.1806], device='cuda:1'), in_proj_covar=tensor([0.0118, 0.0244, 0.0139, 0.0121, 0.0133, 0.0153, 0.0118, 0.0121], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-27 13:19:51,163 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.142e+01 1.502e+02 1.806e+02 2.114e+02 4.959e+02, threshold=3.611e+02, percent-clipped=2.0 2023-04-27 13:19:57,073 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=108166.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:20:05,468 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5663, 3.3115, 0.9396, 1.7747, 2.0338, 2.2976, 1.9613, 0.9417], device='cuda:1'), covar=tensor([0.1391, 0.0854, 0.2037, 0.1295, 0.0957, 0.1001, 0.1459, 0.1963], device='cuda:1'), in_proj_covar=tensor([0.0118, 0.0244, 0.0139, 0.0121, 0.0133, 0.0153, 0.0118, 0.0121], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-27 13:20:29,632 INFO [finetune.py:976] (1/7) Epoch 19, batch 5100, loss[loss=0.1718, simple_loss=0.2375, pruned_loss=0.05306, over 4821.00 frames. ], tot_loss[loss=0.1714, simple_loss=0.2418, pruned_loss=0.05048, over 955576.46 frames. ], batch size: 30, lr: 3.27e-03, grad_scale: 32.0 2023-04-27 13:20:39,721 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.1392, 2.6837, 2.1934, 2.5281, 1.9472, 2.3258, 2.2739, 1.8124], device='cuda:1'), covar=tensor([0.1835, 0.1329, 0.0764, 0.1214, 0.2868, 0.1092, 0.1881, 0.2594], device='cuda:1'), in_proj_covar=tensor([0.0287, 0.0302, 0.0217, 0.0277, 0.0311, 0.0259, 0.0250, 0.0264], device='cuda:1'), out_proj_covar=tensor([1.1502e-04, 1.1981e-04, 8.5851e-05, 1.0996e-04, 1.2633e-04, 1.0250e-04, 1.0098e-04, 1.0465e-04], device='cuda:1') 2023-04-27 13:20:50,602 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=108214.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:21:34,996 INFO [finetune.py:976] (1/7) Epoch 19, batch 5150, loss[loss=0.1336, simple_loss=0.2085, pruned_loss=0.02939, over 4815.00 frames. ], tot_loss[loss=0.1714, simple_loss=0.2421, pruned_loss=0.05035, over 956813.50 frames. ], batch size: 25, lr: 3.27e-03, grad_scale: 32.0 2023-04-27 13:21:44,948 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 8.855e+01 1.529e+02 1.881e+02 2.347e+02 3.720e+02, threshold=3.762e+02, percent-clipped=1.0 2023-04-27 13:22:28,059 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.3652, 1.6208, 1.6700, 1.7736, 1.6470, 1.7045, 1.7787, 1.7782], device='cuda:1'), covar=tensor([0.3630, 0.5217, 0.4761, 0.4256, 0.5676, 0.7254, 0.5106, 0.4998], device='cuda:1'), in_proj_covar=tensor([0.0337, 0.0373, 0.0322, 0.0335, 0.0346, 0.0395, 0.0358, 0.0329], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 13:22:29,162 INFO [finetune.py:976] (1/7) Epoch 19, batch 5200, loss[loss=0.1975, simple_loss=0.2573, pruned_loss=0.06885, over 4350.00 frames. ], tot_loss[loss=0.1765, simple_loss=0.2477, pruned_loss=0.0527, over 954816.05 frames. ], batch size: 19, lr: 3.27e-03, grad_scale: 32.0 2023-04-27 13:22:32,962 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9106, 1.2056, 1.5098, 1.6318, 1.6374, 1.6622, 1.5633, 1.5649], device='cuda:1'), covar=tensor([0.3948, 0.4602, 0.3850, 0.3848, 0.4794, 0.6761, 0.4305, 0.4221], device='cuda:1'), in_proj_covar=tensor([0.0337, 0.0373, 0.0322, 0.0335, 0.0345, 0.0395, 0.0358, 0.0329], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 13:22:51,800 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=108315.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:22:53,084 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.5160, 1.7954, 1.8655, 1.9575, 1.8296, 1.8463, 1.9415, 1.9076], device='cuda:1'), covar=tensor([0.4014, 0.5762, 0.4363, 0.4397, 0.5456, 0.7240, 0.5564, 0.5078], device='cuda:1'), in_proj_covar=tensor([0.0336, 0.0372, 0.0322, 0.0335, 0.0345, 0.0394, 0.0357, 0.0328], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 13:23:05,694 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=108325.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:23:16,923 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.0776, 1.3623, 1.2550, 1.6096, 1.4829, 1.5762, 1.3013, 2.4861], device='cuda:1'), covar=tensor([0.0658, 0.0810, 0.0878, 0.1284, 0.0675, 0.0503, 0.0795, 0.0263], device='cuda:1'), in_proj_covar=tensor([0.0038, 0.0039, 0.0040, 0.0044, 0.0040, 0.0038, 0.0038, 0.0056], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0013, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:1') 2023-04-27 13:23:36,574 INFO [finetune.py:976] (1/7) Epoch 19, batch 5250, loss[loss=0.1666, simple_loss=0.2412, pruned_loss=0.04602, over 4756.00 frames. ], tot_loss[loss=0.1773, simple_loss=0.2489, pruned_loss=0.05283, over 955837.29 frames. ], batch size: 54, lr: 3.27e-03, grad_scale: 32.0 2023-04-27 13:23:44,339 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.109e+02 1.684e+02 2.011e+02 2.395e+02 3.683e+02, threshold=4.022e+02, percent-clipped=0.0 2023-04-27 13:23:53,060 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=108363.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:24:19,499 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=108386.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:24:23,688 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=108393.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:24:27,186 INFO [finetune.py:976] (1/7) Epoch 19, batch 5300, loss[loss=0.1802, simple_loss=0.2534, pruned_loss=0.05353, over 4821.00 frames. ], tot_loss[loss=0.1772, simple_loss=0.249, pruned_loss=0.05269, over 955045.39 frames. ], batch size: 38, lr: 3.27e-03, grad_scale: 32.0 2023-04-27 13:24:45,715 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([4.7184, 4.6548, 2.9400, 5.4191, 4.7758, 4.6825, 1.8578, 4.5122], device='cuda:1'), covar=tensor([0.1552, 0.0954, 0.2910, 0.0796, 0.2343, 0.1531, 0.5694, 0.2212], device='cuda:1'), in_proj_covar=tensor([0.0245, 0.0215, 0.0249, 0.0305, 0.0297, 0.0248, 0.0272, 0.0270], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 13:24:50,807 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=108432.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:24:50,843 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7763, 1.3340, 1.9555, 2.2897, 1.8564, 1.7808, 1.8344, 1.7926], device='cuda:1'), covar=tensor([0.5129, 0.7322, 0.6928, 0.5998, 0.6841, 0.8578, 0.8863, 0.8536], device='cuda:1'), in_proj_covar=tensor([0.0428, 0.0410, 0.0504, 0.0505, 0.0456, 0.0483, 0.0489, 0.0496], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 13:24:56,173 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=108441.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:25:01,038 INFO [finetune.py:976] (1/7) Epoch 19, batch 5350, loss[loss=0.2567, simple_loss=0.3115, pruned_loss=0.101, over 4804.00 frames. ], tot_loss[loss=0.1772, simple_loss=0.2491, pruned_loss=0.0526, over 953663.91 frames. ], batch size: 40, lr: 3.27e-03, grad_scale: 32.0 2023-04-27 13:25:06,496 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.103e+02 1.659e+02 1.988e+02 2.329e+02 5.055e+02, threshold=3.977e+02, percent-clipped=1.0 2023-04-27 13:25:27,047 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.3790, 1.6785, 1.6719, 2.2349, 2.3922, 1.9510, 1.8989, 1.6675], device='cuda:1'), covar=tensor([0.1986, 0.2003, 0.2007, 0.1569, 0.1365, 0.2022, 0.2429, 0.2377], device='cuda:1'), in_proj_covar=tensor([0.0310, 0.0313, 0.0350, 0.0290, 0.0328, 0.0308, 0.0300, 0.0371], device='cuda:1'), out_proj_covar=tensor([6.3765e-05, 6.4894e-05, 7.4143e-05, 5.8609e-05, 6.7968e-05, 6.4730e-05, 6.2859e-05, 7.9086e-05], device='cuda:1') 2023-04-27 13:25:31,248 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=108493.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 13:25:31,848 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.3645, 1.2961, 1.5845, 1.5889, 1.2712, 1.2116, 1.3418, 0.9027], device='cuda:1'), covar=tensor([0.0603, 0.0611, 0.0373, 0.0561, 0.0792, 0.1154, 0.0538, 0.0593], device='cuda:1'), in_proj_covar=tensor([0.0068, 0.0069, 0.0067, 0.0067, 0.0075, 0.0097, 0.0073, 0.0066], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-27 13:25:34,737 INFO [finetune.py:976] (1/7) Epoch 19, batch 5400, loss[loss=0.1738, simple_loss=0.2534, pruned_loss=0.0471, over 4779.00 frames. ], tot_loss[loss=0.1747, simple_loss=0.2461, pruned_loss=0.05168, over 953776.83 frames. ], batch size: 28, lr: 3.27e-03, grad_scale: 32.0 2023-04-27 13:25:36,122 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5472, 1.1297, 1.2706, 1.2517, 1.6797, 1.3127, 1.1201, 1.2265], device='cuda:1'), covar=tensor([0.1415, 0.1372, 0.1849, 0.1381, 0.0910, 0.1370, 0.1691, 0.2078], device='cuda:1'), in_proj_covar=tensor([0.0309, 0.0312, 0.0350, 0.0289, 0.0327, 0.0308, 0.0299, 0.0371], device='cuda:1'), out_proj_covar=tensor([6.3657e-05, 6.4780e-05, 7.4033e-05, 5.8519e-05, 6.7895e-05, 6.4649e-05, 6.2732e-05, 7.8940e-05], device='cuda:1') 2023-04-27 13:26:05,887 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-04-27 13:26:09,395 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6065, 1.3867, 1.7048, 1.8181, 1.3457, 1.2925, 1.4349, 0.9367], device='cuda:1'), covar=tensor([0.0480, 0.0765, 0.0381, 0.0587, 0.0862, 0.1096, 0.0621, 0.0608], device='cuda:1'), in_proj_covar=tensor([0.0068, 0.0068, 0.0067, 0.0067, 0.0074, 0.0096, 0.0073, 0.0065], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-27 13:26:19,344 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([0.9087, 1.9173, 1.7978, 1.5847, 2.0722, 1.5855, 2.4444, 1.4975], device='cuda:1'), covar=tensor([0.3564, 0.1884, 0.4113, 0.2918, 0.1392, 0.2472, 0.1519, 0.4397], device='cuda:1'), in_proj_covar=tensor([0.0338, 0.0345, 0.0426, 0.0354, 0.0382, 0.0377, 0.0372, 0.0417], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 13:26:31,589 INFO [finetune.py:976] (1/7) Epoch 19, batch 5450, loss[loss=0.1789, simple_loss=0.244, pruned_loss=0.0569, over 4808.00 frames. ], tot_loss[loss=0.1717, simple_loss=0.2428, pruned_loss=0.05032, over 954842.14 frames. ], batch size: 51, lr: 3.27e-03, grad_scale: 32.0 2023-04-27 13:26:42,380 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.852e+01 1.533e+02 1.903e+02 2.214e+02 4.592e+02, threshold=3.806e+02, percent-clipped=2.0 2023-04-27 13:26:54,694 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([0.8467, 1.8960, 1.8789, 1.5951, 2.0819, 1.5615, 2.5831, 1.5798], device='cuda:1'), covar=tensor([0.4426, 0.2222, 0.4893, 0.3332, 0.1860, 0.2927, 0.1528, 0.5101], device='cuda:1'), in_proj_covar=tensor([0.0339, 0.0345, 0.0427, 0.0354, 0.0383, 0.0378, 0.0373, 0.0418], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 13:27:04,685 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=5.60 vs. limit=5.0 2023-04-27 13:27:06,380 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=108577.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:27:25,457 INFO [finetune.py:976] (1/7) Epoch 19, batch 5500, loss[loss=0.1503, simple_loss=0.2118, pruned_loss=0.04442, over 4769.00 frames. ], tot_loss[loss=0.1693, simple_loss=0.2398, pruned_loss=0.04945, over 954535.82 frames. ], batch size: 28, lr: 3.27e-03, grad_scale: 32.0 2023-04-27 13:27:35,612 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.65 vs. limit=5.0 2023-04-27 13:27:51,043 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=108638.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:27:58,828 INFO [finetune.py:976] (1/7) Epoch 19, batch 5550, loss[loss=0.1872, simple_loss=0.2635, pruned_loss=0.05548, over 4801.00 frames. ], tot_loss[loss=0.169, simple_loss=0.2397, pruned_loss=0.0491, over 951104.72 frames. ], batch size: 45, lr: 3.27e-03, grad_scale: 32.0 2023-04-27 13:28:09,998 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.131e+01 1.592e+02 1.937e+02 2.305e+02 3.019e+02, threshold=3.873e+02, percent-clipped=0.0 2023-04-27 13:28:41,414 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=108681.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:29:03,284 INFO [finetune.py:976] (1/7) Epoch 19, batch 5600, loss[loss=0.1675, simple_loss=0.24, pruned_loss=0.04755, over 4862.00 frames. ], tot_loss[loss=0.1721, simple_loss=0.2437, pruned_loss=0.05025, over 952930.67 frames. ], batch size: 31, lr: 3.27e-03, grad_scale: 32.0 2023-04-27 13:29:12,621 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6900, 1.4040, 1.6332, 2.0549, 2.0145, 1.6813, 1.3100, 1.8150], device='cuda:1'), covar=tensor([0.0833, 0.1221, 0.0789, 0.0534, 0.0633, 0.0858, 0.0732, 0.0560], device='cuda:1'), in_proj_covar=tensor([0.0189, 0.0203, 0.0184, 0.0173, 0.0180, 0.0183, 0.0153, 0.0181], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 13:29:26,960 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-04-27 13:30:00,561 INFO [finetune.py:976] (1/7) Epoch 19, batch 5650, loss[loss=0.1696, simple_loss=0.2552, pruned_loss=0.04198, over 4926.00 frames. ], tot_loss[loss=0.1743, simple_loss=0.2469, pruned_loss=0.05081, over 954622.27 frames. ], batch size: 38, lr: 3.26e-03, grad_scale: 32.0 2023-04-27 13:30:17,473 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.047e+02 1.527e+02 1.868e+02 2.111e+02 3.831e+02, threshold=3.735e+02, percent-clipped=0.0 2023-04-27 13:30:39,806 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6809, 2.0210, 1.7459, 2.0013, 1.6819, 1.7587, 1.7409, 1.4231], device='cuda:1'), covar=tensor([0.1613, 0.1315, 0.0823, 0.1077, 0.3176, 0.1220, 0.1726, 0.2360], device='cuda:1'), in_proj_covar=tensor([0.0288, 0.0303, 0.0217, 0.0280, 0.0314, 0.0260, 0.0251, 0.0265], device='cuda:1'), out_proj_covar=tensor([1.1547e-04, 1.2046e-04, 8.6164e-05, 1.1102e-04, 1.2733e-04, 1.0304e-04, 1.0145e-04, 1.0509e-04], device='cuda:1') 2023-04-27 13:30:52,596 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=108788.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 13:30:54,417 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=108791.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:31:04,843 INFO [finetune.py:976] (1/7) Epoch 19, batch 5700, loss[loss=0.1565, simple_loss=0.2159, pruned_loss=0.04856, over 4179.00 frames. ], tot_loss[loss=0.1729, simple_loss=0.2438, pruned_loss=0.05098, over 933899.18 frames. ], batch size: 18, lr: 3.26e-03, grad_scale: 32.0 2023-04-27 13:31:21,344 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([3.3697, 2.8058, 2.6416, 2.7722, 2.5830, 2.8646, 2.7541, 2.6561], device='cuda:1'), covar=tensor([0.3606, 0.5493, 0.4553, 0.4672, 0.5296, 0.6082, 0.5467, 0.4898], device='cuda:1'), in_proj_covar=tensor([0.0338, 0.0372, 0.0321, 0.0334, 0.0345, 0.0394, 0.0357, 0.0328], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 13:31:40,718 INFO [finetune.py:976] (1/7) Epoch 20, batch 0, loss[loss=0.1854, simple_loss=0.2585, pruned_loss=0.05617, over 4665.00 frames. ], tot_loss[loss=0.1854, simple_loss=0.2585, pruned_loss=0.05617, over 4665.00 frames. ], batch size: 59, lr: 3.26e-03, grad_scale: 32.0 2023-04-27 13:31:40,719 INFO [finetune.py:1001] (1/7) Computing validation loss 2023-04-27 13:31:48,048 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.1437, 2.6190, 1.0665, 1.4626, 1.7915, 1.3748, 2.9877, 1.7651], device='cuda:1'), covar=tensor([0.0621, 0.0583, 0.0681, 0.1092, 0.0439, 0.0812, 0.0283, 0.0528], device='cuda:1'), in_proj_covar=tensor([0.0051, 0.0065, 0.0048, 0.0047, 0.0050, 0.0052, 0.0074, 0.0051], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:1') 2023-04-27 13:31:49,416 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.3139, 1.4946, 1.7767, 1.9558, 1.8661, 2.0109, 1.8064, 1.8654], device='cuda:1'), covar=tensor([0.3944, 0.5668, 0.4969, 0.4328, 0.5598, 0.6995, 0.5605, 0.4972], device='cuda:1'), in_proj_covar=tensor([0.0338, 0.0373, 0.0321, 0.0334, 0.0345, 0.0394, 0.0357, 0.0328], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 13:31:57,318 INFO [finetune.py:1010] (1/7) Epoch 20, validation: loss=0.1536, simple_loss=0.2249, pruned_loss=0.04109, over 2265189.00 frames. 2023-04-27 13:31:57,318 INFO [finetune.py:1011] (1/7) Maximum memory allocated so far is 6391MB 2023-04-27 13:31:59,108 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6053, 1.3062, 1.3335, 1.4470, 1.7811, 1.4161, 1.1570, 1.2573], device='cuda:1'), covar=tensor([0.1581, 0.1243, 0.1546, 0.1266, 0.0754, 0.1495, 0.1645, 0.2107], device='cuda:1'), in_proj_covar=tensor([0.0304, 0.0307, 0.0345, 0.0285, 0.0323, 0.0304, 0.0296, 0.0366], device='cuda:1'), out_proj_covar=tensor([6.2631e-05, 6.3684e-05, 7.2927e-05, 5.7675e-05, 6.6904e-05, 6.3848e-05, 6.2092e-05, 7.7926e-05], device='cuda:1') 2023-04-27 13:32:03,774 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.8252, 2.1182, 2.0051, 2.2185, 1.9443, 2.0924, 2.0464, 2.0101], device='cuda:1'), covar=tensor([0.3989, 0.6558, 0.4780, 0.4460, 0.6136, 0.7072, 0.6481, 0.6206], device='cuda:1'), in_proj_covar=tensor([0.0338, 0.0372, 0.0321, 0.0334, 0.0345, 0.0393, 0.0357, 0.0328], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 13:32:13,988 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=108852.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:32:15,899 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.14 vs. limit=5.0 2023-04-27 13:32:16,950 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.1835, 1.5856, 1.4241, 1.8492, 1.6415, 2.0739, 1.4539, 3.6424], device='cuda:1'), covar=tensor([0.0632, 0.0810, 0.0827, 0.1175, 0.0671, 0.0451, 0.0724, 0.0128], device='cuda:1'), in_proj_covar=tensor([0.0037, 0.0039, 0.0039, 0.0043, 0.0040, 0.0038, 0.0038, 0.0056], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:1') 2023-04-27 13:32:17,458 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.016e+02 1.429e+02 1.789e+02 2.182e+02 4.169e+02, threshold=3.578e+02, percent-clipped=1.0 2023-04-27 13:32:30,438 INFO [finetune.py:976] (1/7) Epoch 20, batch 50, loss[loss=0.181, simple_loss=0.2522, pruned_loss=0.05483, over 4818.00 frames. ], tot_loss[loss=0.1793, simple_loss=0.2513, pruned_loss=0.05362, over 216435.33 frames. ], batch size: 38, lr: 3.26e-03, grad_scale: 32.0 2023-04-27 13:32:54,746 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9323, 1.6967, 4.6800, 4.4044, 4.1523, 4.5106, 4.3735, 4.1754], device='cuda:1'), covar=tensor([0.6238, 0.5322, 0.1109, 0.1791, 0.0977, 0.1549, 0.1044, 0.1589], device='cuda:1'), in_proj_covar=tensor([0.0306, 0.0299, 0.0403, 0.0402, 0.0346, 0.0404, 0.0310, 0.0363], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-27 13:33:03,608 INFO [finetune.py:976] (1/7) Epoch 20, batch 100, loss[loss=0.1406, simple_loss=0.2078, pruned_loss=0.03668, over 4719.00 frames. ], tot_loss[loss=0.1726, simple_loss=0.2426, pruned_loss=0.05127, over 380580.66 frames. ], batch size: 54, lr: 3.26e-03, grad_scale: 32.0 2023-04-27 13:33:08,741 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=108933.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:33:20,505 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.72 vs. limit=2.0 2023-04-27 13:33:23,933 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.275e+01 1.453e+02 1.763e+02 2.128e+02 3.737e+02, threshold=3.527e+02, percent-clipped=2.0 2023-04-27 13:33:36,956 INFO [finetune.py:976] (1/7) Epoch 20, batch 150, loss[loss=0.1729, simple_loss=0.2395, pruned_loss=0.05312, over 4914.00 frames. ], tot_loss[loss=0.171, simple_loss=0.2395, pruned_loss=0.05124, over 508892.41 frames. ], batch size: 36, lr: 3.26e-03, grad_scale: 32.0 2023-04-27 13:33:39,424 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=108981.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:34:02,755 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4302, 1.3941, 1.7411, 1.7412, 1.3033, 1.2008, 1.4271, 0.9385], device='cuda:1'), covar=tensor([0.0629, 0.0570, 0.0376, 0.0561, 0.0751, 0.1175, 0.0593, 0.0623], device='cuda:1'), in_proj_covar=tensor([0.0068, 0.0068, 0.0067, 0.0067, 0.0074, 0.0096, 0.0073, 0.0065], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-27 13:34:09,750 INFO [finetune.py:976] (1/7) Epoch 20, batch 200, loss[loss=0.1396, simple_loss=0.2179, pruned_loss=0.03068, over 4756.00 frames. ], tot_loss[loss=0.1705, simple_loss=0.2391, pruned_loss=0.05097, over 610648.05 frames. ], batch size: 28, lr: 3.26e-03, grad_scale: 32.0 2023-04-27 13:34:09,842 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4520, 3.6528, 0.7817, 1.8134, 1.9407, 2.4612, 2.0078, 1.0735], device='cuda:1'), covar=tensor([0.1478, 0.0850, 0.2175, 0.1307, 0.1116, 0.1057, 0.1573, 0.1951], device='cuda:1'), in_proj_covar=tensor([0.0117, 0.0242, 0.0139, 0.0119, 0.0132, 0.0152, 0.0117, 0.0120], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-27 13:34:11,041 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=109029.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:34:16,937 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=109037.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 13:34:29,599 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.062e+02 1.602e+02 1.908e+02 2.213e+02 3.365e+02, threshold=3.817e+02, percent-clipped=0.0 2023-04-27 13:34:30,328 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.1668, 2.5411, 1.0370, 1.3157, 1.8716, 1.3497, 3.0701, 1.7202], device='cuda:1'), covar=tensor([0.0708, 0.0553, 0.0734, 0.1385, 0.0508, 0.1002, 0.0351, 0.0655], device='cuda:1'), in_proj_covar=tensor([0.0051, 0.0066, 0.0048, 0.0047, 0.0050, 0.0052, 0.0075, 0.0051], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:1') 2023-04-27 13:34:39,428 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8798, 2.4360, 1.0256, 1.2639, 1.7129, 1.1559, 2.9526, 1.5370], device='cuda:1'), covar=tensor([0.0750, 0.0596, 0.0739, 0.1237, 0.0535, 0.1006, 0.0248, 0.0657], device='cuda:1'), in_proj_covar=tensor([0.0051, 0.0066, 0.0048, 0.0047, 0.0050, 0.0052, 0.0074, 0.0051], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:1') 2023-04-27 13:34:42,777 INFO [finetune.py:976] (1/7) Epoch 20, batch 250, loss[loss=0.2235, simple_loss=0.2981, pruned_loss=0.0744, over 4811.00 frames. ], tot_loss[loss=0.174, simple_loss=0.2434, pruned_loss=0.05234, over 688780.47 frames. ], batch size: 45, lr: 3.26e-03, grad_scale: 32.0 2023-04-27 13:35:02,385 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=109088.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:35:13,162 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=109098.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 13:35:31,290 INFO [finetune.py:976] (1/7) Epoch 20, batch 300, loss[loss=0.1749, simple_loss=0.2496, pruned_loss=0.05006, over 4918.00 frames. ], tot_loss[loss=0.1753, simple_loss=0.2456, pruned_loss=0.05249, over 747522.26 frames. ], batch size: 38, lr: 3.26e-03, grad_scale: 32.0 2023-04-27 13:35:49,172 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=109136.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:35:56,827 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=109147.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:36:03,479 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.124e+02 1.663e+02 2.046e+02 2.623e+02 4.738e+02, threshold=4.093e+02, percent-clipped=3.0 2023-04-27 13:36:25,687 INFO [finetune.py:976] (1/7) Epoch 20, batch 350, loss[loss=0.1565, simple_loss=0.2331, pruned_loss=0.03994, over 4847.00 frames. ], tot_loss[loss=0.1772, simple_loss=0.248, pruned_loss=0.05326, over 793408.23 frames. ], batch size: 49, lr: 3.26e-03, grad_scale: 32.0 2023-04-27 13:36:47,064 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=109194.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:37:21,418 INFO [finetune.py:976] (1/7) Epoch 20, batch 400, loss[loss=0.1559, simple_loss=0.2313, pruned_loss=0.04026, over 4791.00 frames. ], tot_loss[loss=0.1773, simple_loss=0.2485, pruned_loss=0.05302, over 829514.79 frames. ], batch size: 29, lr: 3.26e-03, grad_scale: 32.0 2023-04-27 13:37:31,590 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=109233.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:38:03,877 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=109255.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:38:05,569 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.152e+02 1.658e+02 1.948e+02 2.358e+02 4.073e+02, threshold=3.896e+02, percent-clipped=0.0 2023-04-27 13:38:22,934 INFO [finetune.py:976] (1/7) Epoch 20, batch 450, loss[loss=0.1673, simple_loss=0.2353, pruned_loss=0.04965, over 4887.00 frames. ], tot_loss[loss=0.1756, simple_loss=0.2466, pruned_loss=0.05233, over 855947.22 frames. ], batch size: 32, lr: 3.26e-03, grad_scale: 32.0 2023-04-27 13:38:25,925 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=109281.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:38:34,948 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=109294.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:38:50,577 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=109318.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:38:55,898 INFO [finetune.py:976] (1/7) Epoch 20, batch 500, loss[loss=0.2276, simple_loss=0.2775, pruned_loss=0.08887, over 4712.00 frames. ], tot_loss[loss=0.174, simple_loss=0.2445, pruned_loss=0.05175, over 877947.05 frames. ], batch size: 23, lr: 3.26e-03, grad_scale: 32.0 2023-04-27 13:39:09,687 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7895, 1.3973, 1.5150, 1.4489, 1.9382, 1.5907, 1.2800, 1.4533], device='cuda:1'), covar=tensor([0.1620, 0.1474, 0.2005, 0.1416, 0.0827, 0.1478, 0.2086, 0.2312], device='cuda:1'), in_proj_covar=tensor([0.0303, 0.0306, 0.0344, 0.0285, 0.0320, 0.0303, 0.0295, 0.0364], device='cuda:1'), out_proj_covar=tensor([6.2297e-05, 6.3454e-05, 7.2779e-05, 5.7600e-05, 6.6279e-05, 6.3504e-05, 6.1910e-05, 7.7446e-05], device='cuda:1') 2023-04-27 13:39:16,055 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=109355.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:39:17,760 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 8.900e+01 1.534e+02 1.939e+02 2.475e+02 3.939e+02, threshold=3.877e+02, percent-clipped=2.0 2023-04-27 13:39:27,719 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8850, 1.3032, 1.9368, 2.3915, 1.9746, 1.8171, 1.8544, 1.8777], device='cuda:1'), covar=tensor([0.4777, 0.6975, 0.6694, 0.5737, 0.6170, 0.8471, 0.8825, 0.7827], device='cuda:1'), in_proj_covar=tensor([0.0429, 0.0410, 0.0505, 0.0506, 0.0456, 0.0484, 0.0491, 0.0497], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 13:39:29,358 INFO [finetune.py:976] (1/7) Epoch 20, batch 550, loss[loss=0.1651, simple_loss=0.2195, pruned_loss=0.05528, over 4903.00 frames. ], tot_loss[loss=0.1719, simple_loss=0.2421, pruned_loss=0.05088, over 896840.98 frames. ], batch size: 32, lr: 3.26e-03, grad_scale: 32.0 2023-04-27 13:39:30,711 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=109379.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 13:39:40,651 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=109393.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 13:39:58,712 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.1632, 1.5060, 1.7992, 2.5608, 2.6936, 2.0961, 1.7339, 2.3146], device='cuda:1'), covar=tensor([0.0794, 0.1564, 0.0905, 0.0510, 0.0468, 0.0910, 0.0799, 0.0534], device='cuda:1'), in_proj_covar=tensor([0.0185, 0.0199, 0.0180, 0.0170, 0.0176, 0.0179, 0.0150, 0.0177], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 13:40:02,930 INFO [finetune.py:976] (1/7) Epoch 20, batch 600, loss[loss=0.1833, simple_loss=0.2631, pruned_loss=0.05173, over 4913.00 frames. ], tot_loss[loss=0.1736, simple_loss=0.2436, pruned_loss=0.0518, over 908914.37 frames. ], batch size: 43, lr: 3.26e-03, grad_scale: 64.0 2023-04-27 13:40:16,606 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=109447.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:40:24,270 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.064e+02 1.599e+02 2.024e+02 2.481e+02 5.741e+02, threshold=4.048e+02, percent-clipped=1.0 2023-04-27 13:40:30,366 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.1032, 2.2213, 2.0262, 1.8486, 2.3486, 1.9330, 2.8861, 1.7297], device='cuda:1'), covar=tensor([0.3432, 0.1640, 0.4195, 0.2467, 0.1568, 0.2043, 0.1122, 0.4015], device='cuda:1'), in_proj_covar=tensor([0.0333, 0.0341, 0.0419, 0.0348, 0.0377, 0.0371, 0.0367, 0.0411], device='cuda:1'), out_proj_covar=tensor([9.8949e-05, 1.0213e-04, 1.2742e-04, 1.0499e-04, 1.1238e-04, 1.1076e-04, 1.0815e-04, 1.2443e-04], device='cuda:1') 2023-04-27 13:40:36,281 INFO [finetune.py:976] (1/7) Epoch 20, batch 650, loss[loss=0.2151, simple_loss=0.281, pruned_loss=0.07457, over 4815.00 frames. ], tot_loss[loss=0.1751, simple_loss=0.2459, pruned_loss=0.05215, over 919000.31 frames. ], batch size: 39, lr: 3.26e-03, grad_scale: 64.0 2023-04-27 13:40:48,776 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=109495.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:40:53,638 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([0.8817, 1.2712, 3.2323, 2.9862, 2.9137, 3.1519, 3.1319, 2.8364], device='cuda:1'), covar=tensor([0.7467, 0.5386, 0.1527, 0.2173, 0.1537, 0.2292, 0.2173, 0.1877], device='cuda:1'), in_proj_covar=tensor([0.0309, 0.0302, 0.0404, 0.0405, 0.0350, 0.0408, 0.0312, 0.0366], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-27 13:41:10,002 INFO [finetune.py:976] (1/7) Epoch 20, batch 700, loss[loss=0.1758, simple_loss=0.2428, pruned_loss=0.05438, over 4810.00 frames. ], tot_loss[loss=0.1761, simple_loss=0.2473, pruned_loss=0.05246, over 926334.31 frames. ], batch size: 39, lr: 3.26e-03, grad_scale: 64.0 2023-04-27 13:41:25,575 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=109550.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:41:30,317 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.167e+02 1.590e+02 1.898e+02 2.337e+02 3.520e+02, threshold=3.796e+02, percent-clipped=0.0 2023-04-27 13:41:43,810 INFO [finetune.py:976] (1/7) Epoch 20, batch 750, loss[loss=0.1452, simple_loss=0.2241, pruned_loss=0.03311, over 4820.00 frames. ], tot_loss[loss=0.1758, simple_loss=0.2481, pruned_loss=0.0518, over 934318.76 frames. ], batch size: 25, lr: 3.26e-03, grad_scale: 64.0 2023-04-27 13:41:45,085 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=109579.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:42:49,901 INFO [finetune.py:976] (1/7) Epoch 20, batch 800, loss[loss=0.1643, simple_loss=0.2426, pruned_loss=0.04297, over 4820.00 frames. ], tot_loss[loss=0.1754, simple_loss=0.2472, pruned_loss=0.05174, over 937385.48 frames. ], batch size: 41, lr: 3.25e-03, grad_scale: 64.0 2023-04-27 13:43:07,640 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=109640.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:43:19,838 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=109650.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:43:27,912 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.098e+02 1.469e+02 1.785e+02 2.276e+02 4.254e+02, threshold=3.571e+02, percent-clipped=2.0 2023-04-27 13:43:28,671 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=109659.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:43:50,499 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=109674.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 13:43:52,232 INFO [finetune.py:976] (1/7) Epoch 20, batch 850, loss[loss=0.2308, simple_loss=0.2873, pruned_loss=0.08715, over 4768.00 frames. ], tot_loss[loss=0.1744, simple_loss=0.2456, pruned_loss=0.05156, over 943433.50 frames. ], batch size: 26, lr: 3.25e-03, grad_scale: 64.0 2023-04-27 13:44:05,026 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8499, 1.9647, 1.1203, 1.5332, 2.1913, 1.7301, 1.5908, 1.6359], device='cuda:1'), covar=tensor([0.0436, 0.0310, 0.0305, 0.0484, 0.0251, 0.0454, 0.0434, 0.0487], device='cuda:1'), in_proj_covar=tensor([0.0028, 0.0024, 0.0022, 0.0029, 0.0019, 0.0028, 0.0028, 0.0028], device='cuda:1'), out_proj_covar=tensor([0.0050, 0.0045, 0.0038, 0.0051, 0.0038, 0.0049, 0.0049, 0.0051], device='cuda:1') 2023-04-27 13:44:08,048 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=109693.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 13:44:26,910 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=109720.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:44:31,595 INFO [finetune.py:976] (1/7) Epoch 20, batch 900, loss[loss=0.1316, simple_loss=0.2058, pruned_loss=0.02869, over 4851.00 frames. ], tot_loss[loss=0.1726, simple_loss=0.243, pruned_loss=0.05117, over 946226.36 frames. ], batch size: 47, lr: 3.25e-03, grad_scale: 64.0 2023-04-27 13:44:37,337 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-04-27 13:44:40,146 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=109741.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 13:44:50,920 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.301e+02 1.595e+02 1.882e+02 2.269e+02 5.455e+02, threshold=3.764e+02, percent-clipped=0.0 2023-04-27 13:45:03,938 INFO [finetune.py:976] (1/7) Epoch 20, batch 950, loss[loss=0.2492, simple_loss=0.3059, pruned_loss=0.09628, over 4743.00 frames. ], tot_loss[loss=0.1717, simple_loss=0.2417, pruned_loss=0.05086, over 949129.32 frames. ], batch size: 59, lr: 3.25e-03, grad_scale: 32.0 2023-04-27 13:45:20,630 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.67 vs. limit=2.0 2023-04-27 13:45:26,541 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 2023-04-27 13:45:33,082 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6375, 0.7195, 1.4934, 2.0184, 1.7047, 1.5384, 1.5489, 1.6084], device='cuda:1'), covar=tensor([0.4553, 0.6507, 0.6517, 0.6627, 0.6192, 0.7800, 0.7592, 0.7760], device='cuda:1'), in_proj_covar=tensor([0.0430, 0.0412, 0.0506, 0.0507, 0.0456, 0.0485, 0.0493, 0.0498], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 13:45:38,010 INFO [finetune.py:976] (1/7) Epoch 20, batch 1000, loss[loss=0.1968, simple_loss=0.2683, pruned_loss=0.06269, over 4922.00 frames. ], tot_loss[loss=0.1724, simple_loss=0.2426, pruned_loss=0.05114, over 950888.90 frames. ], batch size: 38, lr: 3.25e-03, grad_scale: 32.0 2023-04-27 13:45:40,606 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([0.9299, 0.9513, 1.0749, 1.0533, 0.8871, 0.8221, 0.8503, 0.5425], device='cuda:1'), covar=tensor([0.0471, 0.0640, 0.0540, 0.0518, 0.0592, 0.1057, 0.0491, 0.0604], device='cuda:1'), in_proj_covar=tensor([0.0068, 0.0068, 0.0067, 0.0067, 0.0074, 0.0096, 0.0073, 0.0065], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-27 13:45:52,566 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=109850.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:45:58,961 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.126e+02 1.623e+02 1.911e+02 2.262e+02 3.883e+02, threshold=3.822e+02, percent-clipped=2.0 2023-04-27 13:46:00,290 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9767, 1.6138, 4.0362, 3.8157, 3.6109, 3.7524, 3.6629, 3.6302], device='cuda:1'), covar=tensor([0.6259, 0.5269, 0.1152, 0.1580, 0.1154, 0.1623, 0.4261, 0.1599], device='cuda:1'), in_proj_covar=tensor([0.0310, 0.0304, 0.0406, 0.0406, 0.0350, 0.0408, 0.0313, 0.0367], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-27 13:46:11,418 INFO [finetune.py:976] (1/7) Epoch 20, batch 1050, loss[loss=0.2024, simple_loss=0.2754, pruned_loss=0.06465, over 4751.00 frames. ], tot_loss[loss=0.1749, simple_loss=0.2463, pruned_loss=0.05176, over 950999.63 frames. ], batch size: 28, lr: 3.25e-03, grad_scale: 32.0 2023-04-27 13:46:25,234 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=109898.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:46:25,306 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5027, 1.3834, 1.7815, 1.8102, 1.3371, 1.2712, 1.5096, 1.0560], device='cuda:1'), covar=tensor([0.0534, 0.0743, 0.0382, 0.0484, 0.0698, 0.1065, 0.0557, 0.0580], device='cuda:1'), in_proj_covar=tensor([0.0068, 0.0068, 0.0067, 0.0067, 0.0074, 0.0096, 0.0073, 0.0065], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-27 13:46:43,735 INFO [finetune.py:976] (1/7) Epoch 20, batch 1100, loss[loss=0.21, simple_loss=0.2758, pruned_loss=0.07212, over 4868.00 frames. ], tot_loss[loss=0.176, simple_loss=0.2476, pruned_loss=0.05214, over 949209.67 frames. ], batch size: 31, lr: 3.25e-03, grad_scale: 32.0 2023-04-27 13:46:50,147 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=109935.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:46:59,873 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=109950.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:47:05,248 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.994e+01 1.686e+02 1.942e+02 2.391e+02 4.510e+02, threshold=3.883e+02, percent-clipped=3.0 2023-04-27 13:47:16,127 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=109974.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:47:17,879 INFO [finetune.py:976] (1/7) Epoch 20, batch 1150, loss[loss=0.1716, simple_loss=0.2408, pruned_loss=0.05121, over 4771.00 frames. ], tot_loss[loss=0.1763, simple_loss=0.2484, pruned_loss=0.05209, over 951882.63 frames. ], batch size: 27, lr: 3.25e-03, grad_scale: 32.0 2023-04-27 13:47:40,685 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=109998.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:47:52,883 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=110015.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:47:54,548 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=110017.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:47:55,138 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0418, 1.2417, 5.0617, 4.7900, 4.4886, 4.9002, 4.4820, 4.4642], device='cuda:1'), covar=tensor([0.6572, 0.6124, 0.0965, 0.1531, 0.0979, 0.1032, 0.1523, 0.1575], device='cuda:1'), in_proj_covar=tensor([0.0310, 0.0305, 0.0406, 0.0408, 0.0350, 0.0409, 0.0315, 0.0369], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-27 13:47:57,539 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=110022.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:48:00,998 INFO [finetune.py:976] (1/7) Epoch 20, batch 1200, loss[loss=0.1604, simple_loss=0.2247, pruned_loss=0.04802, over 4920.00 frames. ], tot_loss[loss=0.1755, simple_loss=0.2473, pruned_loss=0.05179, over 953861.16 frames. ], batch size: 33, lr: 3.25e-03, grad_scale: 32.0 2023-04-27 13:48:15,948 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.5122, 2.8221, 1.2159, 1.7363, 2.3055, 1.6126, 3.5933, 2.1663], device='cuda:1'), covar=tensor([0.0569, 0.0487, 0.0690, 0.1078, 0.0433, 0.0850, 0.0324, 0.0501], device='cuda:1'), in_proj_covar=tensor([0.0051, 0.0065, 0.0048, 0.0046, 0.0049, 0.0052, 0.0074, 0.0051], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:1') 2023-04-27 13:48:17,801 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=110044.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:48:37,464 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.027e+02 1.685e+02 1.958e+02 2.263e+02 7.161e+02, threshold=3.916e+02, percent-clipped=1.0 2023-04-27 13:48:47,301 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9871, 2.3314, 2.0658, 1.8617, 1.5272, 1.5740, 1.9885, 1.5310], device='cuda:1'), covar=tensor([0.1344, 0.1372, 0.1183, 0.1487, 0.2112, 0.1613, 0.0912, 0.1707], device='cuda:1'), in_proj_covar=tensor([0.0194, 0.0209, 0.0167, 0.0202, 0.0199, 0.0183, 0.0154, 0.0184], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-27 13:49:01,339 INFO [finetune.py:976] (1/7) Epoch 20, batch 1250, loss[loss=0.1482, simple_loss=0.2137, pruned_loss=0.04138, over 4910.00 frames. ], tot_loss[loss=0.1731, simple_loss=0.2443, pruned_loss=0.05092, over 953467.39 frames. ], batch size: 36, lr: 3.25e-03, grad_scale: 32.0 2023-04-27 13:49:02,086 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=110078.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:49:08,373 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=3.99 vs. limit=5.0 2023-04-27 13:49:25,656 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=110105.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:49:28,695 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=110110.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:49:31,772 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=110115.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:49:40,409 INFO [finetune.py:976] (1/7) Epoch 20, batch 1300, loss[loss=0.1813, simple_loss=0.2497, pruned_loss=0.05646, over 4815.00 frames. ], tot_loss[loss=0.1712, simple_loss=0.2419, pruned_loss=0.05022, over 953348.29 frames. ], batch size: 41, lr: 3.25e-03, grad_scale: 32.0 2023-04-27 13:49:45,457 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9202, 1.8546, 2.1717, 2.4733, 1.8624, 1.5623, 1.9048, 1.0884], device='cuda:1'), covar=tensor([0.0643, 0.0720, 0.0459, 0.0697, 0.0615, 0.1048, 0.0722, 0.0761], device='cuda:1'), in_proj_covar=tensor([0.0068, 0.0068, 0.0067, 0.0067, 0.0074, 0.0096, 0.0073, 0.0065], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-27 13:50:01,794 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.082e+02 1.513e+02 1.801e+02 2.249e+02 8.027e+02, threshold=3.601e+02, percent-clipped=2.0 2023-04-27 13:50:09,730 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=110171.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:50:11,552 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-04-27 13:50:13,289 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=110176.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:50:13,762 INFO [finetune.py:976] (1/7) Epoch 20, batch 1350, loss[loss=0.1733, simple_loss=0.251, pruned_loss=0.0478, over 4809.00 frames. ], tot_loss[loss=0.1704, simple_loss=0.241, pruned_loss=0.04994, over 952219.91 frames. ], batch size: 38, lr: 3.25e-03, grad_scale: 32.0 2023-04-27 13:50:47,121 INFO [finetune.py:976] (1/7) Epoch 20, batch 1400, loss[loss=0.1846, simple_loss=0.2613, pruned_loss=0.05391, over 4907.00 frames. ], tot_loss[loss=0.1731, simple_loss=0.2444, pruned_loss=0.05088, over 952418.52 frames. ], batch size: 35, lr: 3.25e-03, grad_scale: 32.0 2023-04-27 13:50:52,580 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=110235.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:50:53,277 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.22 vs. limit=5.0 2023-04-27 13:51:09,028 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.151e+02 1.619e+02 1.811e+02 2.225e+02 6.587e+02, threshold=3.621e+02, percent-clipped=5.0 2023-04-27 13:51:12,287 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-04-27 13:51:19,999 INFO [finetune.py:976] (1/7) Epoch 20, batch 1450, loss[loss=0.1554, simple_loss=0.2381, pruned_loss=0.03631, over 4832.00 frames. ], tot_loss[loss=0.1738, simple_loss=0.2462, pruned_loss=0.05073, over 951171.41 frames. ], batch size: 30, lr: 3.25e-03, grad_scale: 32.0 2023-04-27 13:51:25,116 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=110283.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:51:25,161 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=110283.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:51:34,614 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([4.4464, 4.4268, 3.1303, 5.1624, 4.4237, 4.4400, 1.9681, 4.4374], device='cuda:1'), covar=tensor([0.1812, 0.1028, 0.3113, 0.0924, 0.3222, 0.1531, 0.5689, 0.2145], device='cuda:1'), in_proj_covar=tensor([0.0244, 0.0215, 0.0249, 0.0304, 0.0295, 0.0246, 0.0272, 0.0271], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 13:51:38,782 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=110303.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:51:46,128 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.49 vs. limit=2.0 2023-04-27 13:51:46,547 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=110315.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:51:48,403 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=110318.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:51:53,750 INFO [finetune.py:976] (1/7) Epoch 20, batch 1500, loss[loss=0.1519, simple_loss=0.2272, pruned_loss=0.03833, over 4907.00 frames. ], tot_loss[loss=0.1737, simple_loss=0.2466, pruned_loss=0.05037, over 953542.12 frames. ], batch size: 36, lr: 3.25e-03, grad_scale: 32.0 2023-04-27 13:52:05,637 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=110344.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:52:16,146 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.004e+02 1.560e+02 1.895e+02 2.185e+02 4.822e+02, threshold=3.791e+02, percent-clipped=2.0 2023-04-27 13:52:19,125 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=110363.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:52:19,790 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=110364.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:52:25,265 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=110373.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:52:27,673 INFO [finetune.py:976] (1/7) Epoch 20, batch 1550, loss[loss=0.1549, simple_loss=0.2226, pruned_loss=0.04358, over 4930.00 frames. ], tot_loss[loss=0.1743, simple_loss=0.2473, pruned_loss=0.05068, over 954132.77 frames. ], batch size: 33, lr: 3.25e-03, grad_scale: 32.0 2023-04-27 13:52:29,062 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=110379.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:53:00,008 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=110400.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:53:22,805 INFO [finetune.py:976] (1/7) Epoch 20, batch 1600, loss[loss=0.1598, simple_loss=0.2303, pruned_loss=0.04468, over 4861.00 frames. ], tot_loss[loss=0.1727, simple_loss=0.245, pruned_loss=0.05024, over 953078.38 frames. ], batch size: 44, lr: 3.25e-03, grad_scale: 32.0 2023-04-27 13:54:08,137 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.013e+02 1.546e+02 1.968e+02 2.341e+02 5.862e+02, threshold=3.936e+02, percent-clipped=2.0 2023-04-27 13:54:17,900 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=110466.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:54:21,384 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=110471.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:54:30,173 INFO [finetune.py:976] (1/7) Epoch 20, batch 1650, loss[loss=0.1658, simple_loss=0.2326, pruned_loss=0.04952, over 4830.00 frames. ], tot_loss[loss=0.1711, simple_loss=0.2425, pruned_loss=0.04983, over 953670.54 frames. ], batch size: 30, lr: 3.25e-03, grad_scale: 32.0 2023-04-27 13:54:51,132 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([0.9651, 1.2729, 3.2887, 3.0211, 2.9538, 3.2544, 3.2526, 2.9093], device='cuda:1'), covar=tensor([0.7225, 0.5631, 0.1496, 0.2287, 0.1557, 0.2175, 0.1687, 0.1769], device='cuda:1'), in_proj_covar=tensor([0.0307, 0.0301, 0.0402, 0.0404, 0.0347, 0.0405, 0.0311, 0.0364], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-27 13:54:59,479 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.68 vs. limit=5.0 2023-04-27 13:55:07,692 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.2435, 2.2032, 1.7817, 1.9169, 2.2655, 1.8128, 2.8094, 1.5415], device='cuda:1'), covar=tensor([0.3696, 0.1979, 0.4810, 0.3160, 0.1816, 0.2587, 0.1314, 0.4792], device='cuda:1'), in_proj_covar=tensor([0.0338, 0.0346, 0.0424, 0.0350, 0.0381, 0.0374, 0.0371, 0.0418], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 13:55:09,360 INFO [finetune.py:976] (1/7) Epoch 20, batch 1700, loss[loss=0.176, simple_loss=0.2583, pruned_loss=0.04691, over 4936.00 frames. ], tot_loss[loss=0.1701, simple_loss=0.2409, pruned_loss=0.04964, over 954490.73 frames. ], batch size: 38, lr: 3.25e-03, grad_scale: 32.0 2023-04-27 13:55:25,901 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([3.9158, 3.7686, 2.7488, 4.5714, 3.8623, 3.9704, 1.6844, 3.8419], device='cuda:1'), covar=tensor([0.1648, 0.1282, 0.3224, 0.1260, 0.3574, 0.1578, 0.5968, 0.2400], device='cuda:1'), in_proj_covar=tensor([0.0244, 0.0215, 0.0249, 0.0303, 0.0294, 0.0246, 0.0271, 0.0270], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 13:55:31,160 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.338e+01 1.550e+02 1.876e+02 2.294e+02 4.137e+02, threshold=3.752e+02, percent-clipped=2.0 2023-04-27 13:55:42,008 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.3205, 1.8349, 1.6292, 2.0856, 2.0147, 2.1612, 1.6924, 4.4557], device='cuda:1'), covar=tensor([0.0544, 0.0782, 0.0755, 0.1140, 0.0560, 0.0475, 0.0697, 0.0103], device='cuda:1'), in_proj_covar=tensor([0.0038, 0.0039, 0.0040, 0.0043, 0.0040, 0.0038, 0.0038, 0.0056], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0013, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:1') 2023-04-27 13:55:43,125 INFO [finetune.py:976] (1/7) Epoch 20, batch 1750, loss[loss=0.1321, simple_loss=0.1982, pruned_loss=0.03296, over 4089.00 frames. ], tot_loss[loss=0.1723, simple_loss=0.2433, pruned_loss=0.05064, over 955405.24 frames. ], batch size: 17, lr: 3.25e-03, grad_scale: 32.0 2023-04-27 13:56:06,132 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=110611.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:56:17,235 INFO [finetune.py:976] (1/7) Epoch 20, batch 1800, loss[loss=0.1895, simple_loss=0.2721, pruned_loss=0.05346, over 4892.00 frames. ], tot_loss[loss=0.1717, simple_loss=0.2438, pruned_loss=0.04981, over 955402.99 frames. ], batch size: 35, lr: 3.25e-03, grad_scale: 32.0 2023-04-27 13:56:17,340 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6337, 1.6699, 0.8371, 1.2804, 1.8842, 1.4784, 1.3643, 1.4438], device='cuda:1'), covar=tensor([0.0482, 0.0356, 0.0344, 0.0552, 0.0267, 0.0523, 0.0509, 0.0537], device='cuda:1'), in_proj_covar=tensor([0.0028, 0.0024, 0.0022, 0.0029, 0.0020, 0.0028, 0.0028, 0.0028], device='cuda:1'), out_proj_covar=tensor([0.0051, 0.0045, 0.0038, 0.0052, 0.0038, 0.0049, 0.0050, 0.0051], device='cuda:1') 2023-04-27 13:56:24,701 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=110639.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:56:38,212 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.091e+02 1.583e+02 1.893e+02 2.198e+02 3.811e+02, threshold=3.786e+02, percent-clipped=1.0 2023-04-27 13:56:38,297 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=110659.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:56:47,734 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=110672.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:56:48,331 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=110673.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:56:48,911 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=110674.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:56:50,674 INFO [finetune.py:976] (1/7) Epoch 20, batch 1850, loss[loss=0.1621, simple_loss=0.2161, pruned_loss=0.0541, over 3832.00 frames. ], tot_loss[loss=0.1726, simple_loss=0.2448, pruned_loss=0.05016, over 953100.34 frames. ], batch size: 16, lr: 3.25e-03, grad_scale: 32.0 2023-04-27 13:57:06,460 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=110700.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:57:21,095 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=110721.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:57:24,741 INFO [finetune.py:976] (1/7) Epoch 20, batch 1900, loss[loss=0.2039, simple_loss=0.2746, pruned_loss=0.06664, over 4924.00 frames. ], tot_loss[loss=0.1741, simple_loss=0.2463, pruned_loss=0.05096, over 952671.48 frames. ], batch size: 41, lr: 3.25e-03, grad_scale: 32.0 2023-04-27 13:57:30,858 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=110736.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:57:38,120 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=110748.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:57:42,285 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9013, 1.1199, 1.6156, 1.7084, 1.7026, 1.7412, 1.6237, 1.6547], device='cuda:1'), covar=tensor([0.3483, 0.4907, 0.4183, 0.4409, 0.5080, 0.6799, 0.4520, 0.4210], device='cuda:1'), in_proj_covar=tensor([0.0337, 0.0373, 0.0323, 0.0336, 0.0345, 0.0395, 0.0357, 0.0329], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 13:57:46,355 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.050e+01 1.575e+02 1.913e+02 2.383e+02 8.631e+02, threshold=3.825e+02, percent-clipped=5.0 2023-04-27 13:57:51,451 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=110766.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:57:55,035 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=110771.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:57:59,103 INFO [finetune.py:976] (1/7) Epoch 20, batch 1950, loss[loss=0.1833, simple_loss=0.2428, pruned_loss=0.06187, over 4718.00 frames. ], tot_loss[loss=0.1727, simple_loss=0.2452, pruned_loss=0.05006, over 954961.61 frames. ], batch size: 59, lr: 3.25e-03, grad_scale: 32.0 2023-04-27 13:58:29,597 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=110797.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 13:58:52,669 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=110814.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:59:01,944 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=110819.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:59:11,704 INFO [finetune.py:976] (1/7) Epoch 20, batch 2000, loss[loss=0.2077, simple_loss=0.2678, pruned_loss=0.07381, over 4824.00 frames. ], tot_loss[loss=0.1703, simple_loss=0.2426, pruned_loss=0.04898, over 957215.69 frames. ], batch size: 30, lr: 3.25e-03, grad_scale: 32.0 2023-04-27 13:59:33,917 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.034e+02 1.549e+02 1.813e+02 2.155e+02 3.715e+02, threshold=3.626e+02, percent-clipped=0.0 2023-04-27 13:59:45,762 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=110868.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:59:56,921 INFO [finetune.py:976] (1/7) Epoch 20, batch 2050, loss[loss=0.1483, simple_loss=0.2265, pruned_loss=0.03509, over 4872.00 frames. ], tot_loss[loss=0.1681, simple_loss=0.2399, pruned_loss=0.0482, over 955849.77 frames. ], batch size: 31, lr: 3.24e-03, grad_scale: 32.0 2023-04-27 14:00:53,302 INFO [finetune.py:976] (1/7) Epoch 20, batch 2100, loss[loss=0.1599, simple_loss=0.2232, pruned_loss=0.04831, over 4713.00 frames. ], tot_loss[loss=0.1694, simple_loss=0.2409, pruned_loss=0.04891, over 953638.35 frames. ], batch size: 23, lr: 3.24e-03, grad_scale: 32.0 2023-04-27 14:00:54,662 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=110929.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:01:01,675 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=110939.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:01:13,893 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=110958.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:01:14,358 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.165e+02 1.601e+02 1.866e+02 2.314e+02 3.983e+02, threshold=3.732e+02, percent-clipped=4.0 2023-04-27 14:01:14,953 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=110959.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:01:14,981 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.2524, 2.1133, 2.2621, 2.7034, 2.7409, 2.1739, 1.8822, 2.2424], device='cuda:1'), covar=tensor([0.0816, 0.0988, 0.0671, 0.0524, 0.0627, 0.0835, 0.0800, 0.0607], device='cuda:1'), in_proj_covar=tensor([0.0189, 0.0202, 0.0184, 0.0174, 0.0180, 0.0182, 0.0154, 0.0181], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 14:01:20,715 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=110967.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:01:25,044 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=110974.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:01:27,215 INFO [finetune.py:976] (1/7) Epoch 20, batch 2150, loss[loss=0.2229, simple_loss=0.2732, pruned_loss=0.08628, over 4768.00 frames. ], tot_loss[loss=0.1724, simple_loss=0.2442, pruned_loss=0.05033, over 953667.85 frames. ], batch size: 54, lr: 3.24e-03, grad_scale: 32.0 2023-04-27 14:01:34,419 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=110987.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:01:47,107 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=111007.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:01:55,864 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=111019.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:01:57,630 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=111022.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:02:00,609 INFO [finetune.py:976] (1/7) Epoch 20, batch 2200, loss[loss=0.2317, simple_loss=0.2938, pruned_loss=0.08478, over 4826.00 frames. ], tot_loss[loss=0.1751, simple_loss=0.2474, pruned_loss=0.05141, over 953824.54 frames. ], batch size: 39, lr: 3.24e-03, grad_scale: 32.0 2023-04-27 14:02:11,890 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8904, 4.1243, 0.6415, 2.0521, 2.3278, 2.5616, 2.4620, 0.9461], device='cuda:1'), covar=tensor([0.1194, 0.0776, 0.2056, 0.1289, 0.0905, 0.1046, 0.1358, 0.2040], device='cuda:1'), in_proj_covar=tensor([0.0118, 0.0243, 0.0139, 0.0121, 0.0134, 0.0153, 0.0118, 0.0120], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-27 14:02:21,773 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.09 vs. limit=5.0 2023-04-27 14:02:22,144 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.079e+02 1.590e+02 1.963e+02 2.328e+02 5.181e+02, threshold=3.927e+02, percent-clipped=3.0 2023-04-27 14:02:34,616 INFO [finetune.py:976] (1/7) Epoch 20, batch 2250, loss[loss=0.1987, simple_loss=0.2806, pruned_loss=0.05838, over 4820.00 frames. ], tot_loss[loss=0.1765, simple_loss=0.2489, pruned_loss=0.05204, over 953600.69 frames. ], batch size: 51, lr: 3.24e-03, grad_scale: 32.0 2023-04-27 14:02:44,858 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=111092.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 14:03:08,137 INFO [finetune.py:976] (1/7) Epoch 20, batch 2300, loss[loss=0.1807, simple_loss=0.2658, pruned_loss=0.04785, over 4850.00 frames. ], tot_loss[loss=0.1759, simple_loss=0.2486, pruned_loss=0.05157, over 952104.68 frames. ], batch size: 49, lr: 3.24e-03, grad_scale: 32.0 2023-04-27 14:03:29,169 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.334e+01 1.540e+02 1.823e+02 2.154e+02 3.608e+02, threshold=3.645e+02, percent-clipped=0.0 2023-04-27 14:03:32,305 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([4.1367, 4.1063, 2.8657, 4.7606, 4.1576, 4.0645, 1.9452, 4.0758], device='cuda:1'), covar=tensor([0.1687, 0.1045, 0.2806, 0.1281, 0.3370, 0.1562, 0.5328, 0.2133], device='cuda:1'), in_proj_covar=tensor([0.0246, 0.0215, 0.0250, 0.0303, 0.0297, 0.0248, 0.0274, 0.0271], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 14:03:36,405 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0223, 1.9484, 2.0334, 2.5007, 2.4575, 2.0866, 1.7090, 2.2493], device='cuda:1'), covar=tensor([0.0831, 0.0978, 0.0662, 0.0527, 0.0542, 0.0763, 0.0732, 0.0500], device='cuda:1'), in_proj_covar=tensor([0.0189, 0.0202, 0.0185, 0.0174, 0.0180, 0.0182, 0.0154, 0.0181], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 14:03:41,037 INFO [finetune.py:976] (1/7) Epoch 20, batch 2350, loss[loss=0.185, simple_loss=0.2548, pruned_loss=0.05764, over 4820.00 frames. ], tot_loss[loss=0.1731, simple_loss=0.2452, pruned_loss=0.05053, over 950600.97 frames. ], batch size: 39, lr: 3.24e-03, grad_scale: 32.0 2023-04-27 14:03:42,219 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.2387, 2.5475, 0.7917, 1.4228, 1.5669, 1.9077, 1.6556, 0.8027], device='cuda:1'), covar=tensor([0.1387, 0.1088, 0.1798, 0.1384, 0.1079, 0.0932, 0.1614, 0.1748], device='cuda:1'), in_proj_covar=tensor([0.0118, 0.0243, 0.0138, 0.0121, 0.0134, 0.0153, 0.0118, 0.0120], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-27 14:03:51,277 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.49 vs. limit=5.0 2023-04-27 14:03:55,875 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.89 vs. limit=5.0 2023-04-27 14:04:27,659 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=111224.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:04:28,315 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.2101, 2.6969, 1.0353, 1.4366, 1.9409, 1.3132, 3.6412, 1.9143], device='cuda:1'), covar=tensor([0.0642, 0.0662, 0.0817, 0.1247, 0.0538, 0.0997, 0.0253, 0.0617], device='cuda:1'), in_proj_covar=tensor([0.0051, 0.0065, 0.0048, 0.0047, 0.0050, 0.0052, 0.0074, 0.0051], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:1') 2023-04-27 14:04:28,339 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=111225.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:04:29,454 INFO [finetune.py:976] (1/7) Epoch 20, batch 2400, loss[loss=0.1822, simple_loss=0.243, pruned_loss=0.06073, over 4802.00 frames. ], tot_loss[loss=0.1718, simple_loss=0.2431, pruned_loss=0.05026, over 952325.48 frames. ], batch size: 25, lr: 3.24e-03, grad_scale: 32.0 2023-04-27 14:04:53,367 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6908, 1.6555, 1.8947, 2.0673, 1.6483, 1.3888, 1.6454, 1.0268], device='cuda:1'), covar=tensor([0.0664, 0.0561, 0.0608, 0.0606, 0.0690, 0.1012, 0.0746, 0.0718], device='cuda:1'), in_proj_covar=tensor([0.0069, 0.0069, 0.0067, 0.0068, 0.0075, 0.0096, 0.0073, 0.0065], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-27 14:05:02,139 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.830e+01 1.579e+02 1.923e+02 2.296e+02 7.513e+02, threshold=3.847e+02, percent-clipped=3.0 2023-04-27 14:05:07,626 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=111267.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:05:19,546 INFO [finetune.py:976] (1/7) Epoch 20, batch 2450, loss[loss=0.1558, simple_loss=0.226, pruned_loss=0.04285, over 4821.00 frames. ], tot_loss[loss=0.1697, simple_loss=0.2408, pruned_loss=0.04928, over 953010.62 frames. ], batch size: 25, lr: 3.24e-03, grad_scale: 32.0 2023-04-27 14:05:36,573 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=111286.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:06:05,489 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=111314.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:06:06,087 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=111315.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:06:13,837 INFO [finetune.py:976] (1/7) Epoch 20, batch 2500, loss[loss=0.1603, simple_loss=0.241, pruned_loss=0.03984, over 4836.00 frames. ], tot_loss[loss=0.1729, simple_loss=0.2437, pruned_loss=0.05105, over 953763.33 frames. ], batch size: 39, lr: 3.24e-03, grad_scale: 32.0 2023-04-27 14:06:19,345 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([3.9037, 3.8445, 2.8528, 4.5447, 3.9121, 3.8157, 2.1340, 3.8617], device='cuda:1'), covar=tensor([0.1952, 0.1274, 0.3550, 0.1446, 0.3628, 0.1994, 0.5286, 0.2600], device='cuda:1'), in_proj_covar=tensor([0.0244, 0.0214, 0.0249, 0.0302, 0.0294, 0.0247, 0.0272, 0.0270], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 14:06:22,432 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-04-27 14:06:25,857 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7348, 2.8121, 2.3777, 2.6027, 2.9402, 2.4095, 3.8176, 2.2115], device='cuda:1'), covar=tensor([0.3934, 0.2041, 0.3801, 0.3007, 0.1907, 0.2426, 0.1376, 0.3889], device='cuda:1'), in_proj_covar=tensor([0.0341, 0.0350, 0.0429, 0.0354, 0.0384, 0.0376, 0.0375, 0.0419], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 14:06:35,391 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.107e+02 1.671e+02 1.878e+02 2.269e+02 4.413e+02, threshold=3.755e+02, percent-clipped=2.0 2023-04-27 14:06:46,745 INFO [finetune.py:976] (1/7) Epoch 20, batch 2550, loss[loss=0.1879, simple_loss=0.267, pruned_loss=0.05441, over 4906.00 frames. ], tot_loss[loss=0.1744, simple_loss=0.2454, pruned_loss=0.05173, over 951741.83 frames. ], batch size: 43, lr: 3.24e-03, grad_scale: 32.0 2023-04-27 14:06:57,273 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=111392.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 14:07:20,615 INFO [finetune.py:976] (1/7) Epoch 20, batch 2600, loss[loss=0.1528, simple_loss=0.2001, pruned_loss=0.05282, over 4080.00 frames. ], tot_loss[loss=0.1756, simple_loss=0.2469, pruned_loss=0.05213, over 952686.04 frames. ], batch size: 17, lr: 3.24e-03, grad_scale: 32.0 2023-04-27 14:07:29,126 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=111440.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:07:42,988 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.116e+02 1.630e+02 1.917e+02 2.393e+02 4.047e+02, threshold=3.834e+02, percent-clipped=2.0 2023-04-27 14:07:48,674 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=111468.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:07:54,483 INFO [finetune.py:976] (1/7) Epoch 20, batch 2650, loss[loss=0.2235, simple_loss=0.3091, pruned_loss=0.06895, over 4808.00 frames. ], tot_loss[loss=0.176, simple_loss=0.2478, pruned_loss=0.05211, over 953155.21 frames. ], batch size: 45, lr: 3.24e-03, grad_scale: 32.0 2023-04-27 14:08:05,306 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.48 vs. limit=2.0 2023-04-27 14:08:09,402 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.2068, 1.5927, 1.8276, 1.9191, 2.3274, 1.9518, 1.6787, 1.6065], device='cuda:1'), covar=tensor([0.1208, 0.1363, 0.1747, 0.1170, 0.0852, 0.1249, 0.1618, 0.1984], device='cuda:1'), in_proj_covar=tensor([0.0311, 0.0311, 0.0351, 0.0288, 0.0327, 0.0307, 0.0301, 0.0371], device='cuda:1'), out_proj_covar=tensor([6.4077e-05, 6.4388e-05, 7.4268e-05, 5.8301e-05, 6.7760e-05, 6.4373e-05, 6.3157e-05, 7.8944e-05], device='cuda:1') 2023-04-27 14:08:26,341 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=111524.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:08:28,081 INFO [finetune.py:976] (1/7) Epoch 20, batch 2700, loss[loss=0.1958, simple_loss=0.2813, pruned_loss=0.05515, over 4885.00 frames. ], tot_loss[loss=0.177, simple_loss=0.2486, pruned_loss=0.05265, over 950972.95 frames. ], batch size: 43, lr: 3.24e-03, grad_scale: 32.0 2023-04-27 14:08:29,428 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=111529.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:08:49,485 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.017e+02 1.491e+02 1.860e+02 2.271e+02 3.945e+02, threshold=3.719e+02, percent-clipped=1.0 2023-04-27 14:08:58,397 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=111572.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:09:01,908 INFO [finetune.py:976] (1/7) Epoch 20, batch 2750, loss[loss=0.1477, simple_loss=0.2199, pruned_loss=0.03772, over 4903.00 frames. ], tot_loss[loss=0.1757, simple_loss=0.2467, pruned_loss=0.05232, over 952994.68 frames. ], batch size: 35, lr: 3.24e-03, grad_scale: 32.0 2023-04-27 14:09:04,324 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=111581.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:09:09,231 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5403, 1.3617, 0.5709, 1.2136, 1.4343, 1.4082, 1.2928, 1.3400], device='cuda:1'), covar=tensor([0.0508, 0.0371, 0.0387, 0.0543, 0.0290, 0.0482, 0.0490, 0.0564], device='cuda:1'), in_proj_covar=tensor([0.0028, 0.0024, 0.0022, 0.0029, 0.0019, 0.0028, 0.0028, 0.0028], device='cuda:1'), out_proj_covar=tensor([0.0051, 0.0045, 0.0038, 0.0052, 0.0038, 0.0049, 0.0049, 0.0051], device='cuda:1') 2023-04-27 14:09:37,586 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=111614.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:09:57,466 INFO [finetune.py:976] (1/7) Epoch 20, batch 2800, loss[loss=0.1879, simple_loss=0.2589, pruned_loss=0.0585, over 4842.00 frames. ], tot_loss[loss=0.1733, simple_loss=0.2436, pruned_loss=0.0515, over 953782.70 frames. ], batch size: 44, lr: 3.24e-03, grad_scale: 32.0 2023-04-27 14:10:24,458 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.707e+01 1.587e+02 1.829e+02 2.177e+02 6.182e+02, threshold=3.659e+02, percent-clipped=2.0 2023-04-27 14:10:26,838 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=111662.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:10:28,731 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9183, 1.3670, 4.9905, 4.6431, 4.3198, 4.7789, 4.3979, 4.3501], device='cuda:1'), covar=tensor([0.6842, 0.6203, 0.0946, 0.1835, 0.1200, 0.1704, 0.1688, 0.1687], device='cuda:1'), in_proj_covar=tensor([0.0311, 0.0307, 0.0408, 0.0408, 0.0353, 0.0410, 0.0316, 0.0367], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-27 14:10:36,931 INFO [finetune.py:976] (1/7) Epoch 20, batch 2850, loss[loss=0.1958, simple_loss=0.2586, pruned_loss=0.06649, over 4873.00 frames. ], tot_loss[loss=0.1719, simple_loss=0.2419, pruned_loss=0.05095, over 955145.36 frames. ], batch size: 34, lr: 3.24e-03, grad_scale: 32.0 2023-04-27 14:11:31,378 INFO [finetune.py:976] (1/7) Epoch 20, batch 2900, loss[loss=0.1924, simple_loss=0.2806, pruned_loss=0.05217, over 4826.00 frames. ], tot_loss[loss=0.174, simple_loss=0.2446, pruned_loss=0.05166, over 955632.73 frames. ], batch size: 33, lr: 3.24e-03, grad_scale: 32.0 2023-04-27 14:12:12,492 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.123e+02 1.619e+02 1.984e+02 2.443e+02 6.162e+02, threshold=3.968e+02, percent-clipped=7.0 2023-04-27 14:12:36,595 INFO [finetune.py:976] (1/7) Epoch 20, batch 2950, loss[loss=0.1884, simple_loss=0.2623, pruned_loss=0.0573, over 4792.00 frames. ], tot_loss[loss=0.1762, simple_loss=0.248, pruned_loss=0.05221, over 955815.88 frames. ], batch size: 25, lr: 3.24e-03, grad_scale: 64.0 2023-04-27 14:12:56,801 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=111791.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 14:13:27,620 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8646, 2.5994, 1.8359, 2.1022, 1.5668, 1.5212, 1.9143, 1.5074], device='cuda:1'), covar=tensor([0.1512, 0.1198, 0.1425, 0.1448, 0.1950, 0.1986, 0.0840, 0.1761], device='cuda:1'), in_proj_covar=tensor([0.0197, 0.0211, 0.0169, 0.0204, 0.0200, 0.0186, 0.0155, 0.0185], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-27 14:13:29,430 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.2626, 1.6728, 2.1348, 2.5311, 2.1090, 1.6726, 1.3778, 1.9363], device='cuda:1'), covar=tensor([0.3285, 0.3280, 0.1612, 0.2388, 0.2575, 0.2663, 0.4209, 0.1967], device='cuda:1'), in_proj_covar=tensor([0.0291, 0.0245, 0.0226, 0.0314, 0.0218, 0.0232, 0.0228, 0.0185], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-27 14:13:29,943 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=111824.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:13:32,106 INFO [finetune.py:976] (1/7) Epoch 20, batch 3000, loss[loss=0.1785, simple_loss=0.2563, pruned_loss=0.05033, over 4824.00 frames. ], tot_loss[loss=0.1766, simple_loss=0.2488, pruned_loss=0.05221, over 953875.27 frames. ], batch size: 30, lr: 3.24e-03, grad_scale: 64.0 2023-04-27 14:13:32,106 INFO [finetune.py:1001] (1/7) Computing validation loss 2023-04-27 14:13:49,407 INFO [finetune.py:1010] (1/7) Epoch 20, validation: loss=0.1527, simple_loss=0.2229, pruned_loss=0.04123, over 2265189.00 frames. 2023-04-27 14:13:49,408 INFO [finetune.py:1011] (1/7) Maximum memory allocated so far is 6391MB 2023-04-27 14:14:18,542 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-04-27 14:14:20,689 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=111852.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 14:14:29,830 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.150e+02 1.660e+02 1.972e+02 2.446e+02 4.506e+02, threshold=3.944e+02, percent-clipped=2.0 2023-04-27 14:14:58,618 INFO [finetune.py:976] (1/7) Epoch 20, batch 3050, loss[loss=0.167, simple_loss=0.2357, pruned_loss=0.04921, over 4824.00 frames. ], tot_loss[loss=0.1774, simple_loss=0.25, pruned_loss=0.05238, over 955448.42 frames. ], batch size: 25, lr: 3.24e-03, grad_scale: 64.0 2023-04-27 14:15:01,119 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=111881.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:15:48,560 INFO [finetune.py:976] (1/7) Epoch 20, batch 3100, loss[loss=0.1624, simple_loss=0.2275, pruned_loss=0.04859, over 4789.00 frames. ], tot_loss[loss=0.1754, simple_loss=0.2475, pruned_loss=0.0516, over 957183.83 frames. ], batch size: 29, lr: 3.24e-03, grad_scale: 32.0 2023-04-27 14:15:50,328 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=111929.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:15:59,427 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.2685, 1.7271, 2.1326, 2.5891, 2.1425, 1.7040, 1.5119, 1.9118], device='cuda:1'), covar=tensor([0.2969, 0.3064, 0.1520, 0.2281, 0.2410, 0.2561, 0.4128, 0.2119], device='cuda:1'), in_proj_covar=tensor([0.0291, 0.0245, 0.0226, 0.0313, 0.0218, 0.0232, 0.0228, 0.0185], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-27 14:16:16,415 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.135e+02 1.458e+02 1.721e+02 2.109e+02 3.576e+02, threshold=3.441e+02, percent-clipped=0.0 2023-04-27 14:16:21,342 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6726, 1.6483, 0.6867, 1.4235, 1.8205, 1.5535, 1.4345, 1.5648], device='cuda:1'), covar=tensor([0.0479, 0.0368, 0.0333, 0.0527, 0.0249, 0.0487, 0.0477, 0.0529], device='cuda:1'), in_proj_covar=tensor([0.0028, 0.0024, 0.0022, 0.0028, 0.0019, 0.0028, 0.0028, 0.0028], device='cuda:1'), out_proj_covar=tensor([0.0050, 0.0045, 0.0038, 0.0051, 0.0038, 0.0049, 0.0049, 0.0050], device='cuda:1') 2023-04-27 14:16:27,776 INFO [finetune.py:976] (1/7) Epoch 20, batch 3150, loss[loss=0.1423, simple_loss=0.2168, pruned_loss=0.03385, over 4833.00 frames. ], tot_loss[loss=0.1731, simple_loss=0.2442, pruned_loss=0.051, over 955385.69 frames. ], batch size: 33, lr: 3.24e-03, grad_scale: 32.0 2023-04-27 14:17:02,005 INFO [finetune.py:976] (1/7) Epoch 20, batch 3200, loss[loss=0.2107, simple_loss=0.2777, pruned_loss=0.07181, over 4743.00 frames. ], tot_loss[loss=0.1697, simple_loss=0.2405, pruned_loss=0.04946, over 954962.20 frames. ], batch size: 59, lr: 3.24e-03, grad_scale: 32.0 2023-04-27 14:17:14,096 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.4175, 1.8256, 1.8784, 1.9328, 1.8523, 1.9594, 1.9133, 1.9085], device='cuda:1'), covar=tensor([0.4095, 0.4680, 0.4231, 0.4125, 0.4972, 0.6208, 0.4830, 0.4543], device='cuda:1'), in_proj_covar=tensor([0.0335, 0.0372, 0.0320, 0.0334, 0.0344, 0.0392, 0.0356, 0.0327], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 14:17:35,524 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.097e+02 1.584e+02 1.831e+02 2.274e+02 4.743e+02, threshold=3.663e+02, percent-clipped=4.0 2023-04-27 14:17:39,737 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=112066.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:17:46,878 INFO [finetune.py:976] (1/7) Epoch 20, batch 3250, loss[loss=0.1366, simple_loss=0.2235, pruned_loss=0.02483, over 4751.00 frames. ], tot_loss[loss=0.1704, simple_loss=0.2413, pruned_loss=0.04979, over 952749.29 frames. ], batch size: 27, lr: 3.23e-03, grad_scale: 32.0 2023-04-27 14:18:18,408 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=112124.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:18:20,135 INFO [finetune.py:976] (1/7) Epoch 20, batch 3300, loss[loss=0.1926, simple_loss=0.2637, pruned_loss=0.06071, over 4764.00 frames. ], tot_loss[loss=0.1732, simple_loss=0.2444, pruned_loss=0.05095, over 951310.48 frames. ], batch size: 28, lr: 3.23e-03, grad_scale: 32.0 2023-04-27 14:18:20,258 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=112127.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:18:44,825 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=112147.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 14:18:52,672 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.675e+01 1.599e+02 1.882e+02 2.331e+02 3.289e+02, threshold=3.764e+02, percent-clipped=0.0 2023-04-27 14:19:00,558 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=112172.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:19:04,158 INFO [finetune.py:976] (1/7) Epoch 20, batch 3350, loss[loss=0.199, simple_loss=0.2569, pruned_loss=0.07049, over 4924.00 frames. ], tot_loss[loss=0.1751, simple_loss=0.2468, pruned_loss=0.05177, over 952283.88 frames. ], batch size: 33, lr: 3.23e-03, grad_scale: 32.0 2023-04-27 14:19:25,812 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.10 vs. limit=2.0 2023-04-27 14:19:34,128 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6692, 1.8177, 0.9377, 1.3802, 1.9813, 1.5265, 1.4373, 1.5061], device='cuda:1'), covar=tensor([0.0473, 0.0348, 0.0345, 0.0554, 0.0244, 0.0509, 0.0491, 0.0538], device='cuda:1'), in_proj_covar=tensor([0.0028, 0.0024, 0.0022, 0.0029, 0.0019, 0.0028, 0.0028, 0.0028], device='cuda:1'), out_proj_covar=tensor([0.0050, 0.0045, 0.0038, 0.0051, 0.0038, 0.0049, 0.0049, 0.0051], device='cuda:1') 2023-04-27 14:19:35,235 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.3471, 1.6213, 1.8113, 1.8792, 1.7562, 1.8458, 1.8635, 1.8445], device='cuda:1'), covar=tensor([0.4403, 0.5610, 0.4478, 0.4465, 0.5617, 0.7369, 0.5373, 0.5051], device='cuda:1'), in_proj_covar=tensor([0.0336, 0.0372, 0.0321, 0.0335, 0.0345, 0.0393, 0.0356, 0.0327], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 14:19:37,509 INFO [finetune.py:976] (1/7) Epoch 20, batch 3400, loss[loss=0.2049, simple_loss=0.2734, pruned_loss=0.0682, over 4909.00 frames. ], tot_loss[loss=0.1754, simple_loss=0.2472, pruned_loss=0.05174, over 952344.82 frames. ], batch size: 33, lr: 3.23e-03, grad_scale: 32.0 2023-04-27 14:20:15,438 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.071e+02 1.600e+02 1.895e+02 2.268e+02 5.639e+02, threshold=3.789e+02, percent-clipped=3.0 2023-04-27 14:20:35,316 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.71 vs. limit=2.0 2023-04-27 14:20:37,464 INFO [finetune.py:976] (1/7) Epoch 20, batch 3450, loss[loss=0.142, simple_loss=0.2257, pruned_loss=0.02914, over 4811.00 frames. ], tot_loss[loss=0.1755, simple_loss=0.2475, pruned_loss=0.05171, over 953871.59 frames. ], batch size: 40, lr: 3.23e-03, grad_scale: 32.0 2023-04-27 14:21:20,379 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0227, 1.5027, 5.0587, 4.7530, 4.4028, 4.8072, 4.4367, 4.4915], device='cuda:1'), covar=tensor([0.6564, 0.5517, 0.0930, 0.1735, 0.0909, 0.1179, 0.1542, 0.1673], device='cuda:1'), in_proj_covar=tensor([0.0310, 0.0308, 0.0408, 0.0407, 0.0352, 0.0410, 0.0315, 0.0367], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-27 14:21:35,777 INFO [finetune.py:976] (1/7) Epoch 20, batch 3500, loss[loss=0.1769, simple_loss=0.2418, pruned_loss=0.05601, over 4854.00 frames. ], tot_loss[loss=0.1735, simple_loss=0.245, pruned_loss=0.05096, over 953439.34 frames. ], batch size: 49, lr: 3.23e-03, grad_scale: 32.0 2023-04-27 14:21:53,126 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5451, 0.7328, 1.4026, 1.8465, 1.5847, 1.4547, 1.4717, 1.4953], device='cuda:1'), covar=tensor([0.5016, 0.6984, 0.7061, 0.7882, 0.6168, 0.8593, 0.7789, 0.8627], device='cuda:1'), in_proj_covar=tensor([0.0430, 0.0412, 0.0507, 0.0507, 0.0456, 0.0486, 0.0493, 0.0497], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 14:21:59,355 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.140e+02 1.563e+02 1.868e+02 2.164e+02 4.202e+02, threshold=3.735e+02, percent-clipped=1.0 2023-04-27 14:22:09,766 INFO [finetune.py:976] (1/7) Epoch 20, batch 3550, loss[loss=0.1433, simple_loss=0.2278, pruned_loss=0.0294, over 4833.00 frames. ], tot_loss[loss=0.1706, simple_loss=0.2416, pruned_loss=0.04982, over 952226.35 frames. ], batch size: 40, lr: 3.23e-03, grad_scale: 32.0 2023-04-27 14:22:18,191 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=112389.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:22:40,993 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=112422.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:22:44,029 INFO [finetune.py:976] (1/7) Epoch 20, batch 3600, loss[loss=0.1291, simple_loss=0.2042, pruned_loss=0.02696, over 4864.00 frames. ], tot_loss[loss=0.1692, simple_loss=0.2399, pruned_loss=0.04932, over 953513.68 frames. ], batch size: 31, lr: 3.23e-03, grad_scale: 32.0 2023-04-27 14:22:55,018 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6878, 1.4122, 1.7972, 1.8827, 1.4967, 1.3702, 1.5091, 1.0139], device='cuda:1'), covar=tensor([0.0430, 0.0617, 0.0395, 0.0462, 0.0680, 0.1109, 0.0506, 0.0619], device='cuda:1'), in_proj_covar=tensor([0.0069, 0.0069, 0.0068, 0.0069, 0.0076, 0.0098, 0.0074, 0.0066], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-27 14:22:57,511 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=112447.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 14:22:59,360 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=112450.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:22:59,375 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.5718, 2.0235, 2.3638, 3.1455, 2.3515, 1.9423, 2.1357, 2.3282], device='cuda:1'), covar=tensor([0.3173, 0.3219, 0.1652, 0.2291, 0.2855, 0.2548, 0.3467, 0.2150], device='cuda:1'), in_proj_covar=tensor([0.0291, 0.0246, 0.0226, 0.0313, 0.0219, 0.0233, 0.0228, 0.0185], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-27 14:23:06,296 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.099e+02 1.532e+02 1.802e+02 2.141e+02 3.664e+02, threshold=3.603e+02, percent-clipped=0.0 2023-04-27 14:23:18,445 INFO [finetune.py:976] (1/7) Epoch 20, batch 3650, loss[loss=0.1378, simple_loss=0.2089, pruned_loss=0.03332, over 4761.00 frames. ], tot_loss[loss=0.1708, simple_loss=0.2416, pruned_loss=0.05001, over 954237.82 frames. ], batch size: 26, lr: 3.23e-03, grad_scale: 32.0 2023-04-27 14:23:30,564 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=112495.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 14:23:58,093 INFO [finetune.py:976] (1/7) Epoch 20, batch 3700, loss[loss=0.1473, simple_loss=0.2276, pruned_loss=0.03354, over 4770.00 frames. ], tot_loss[loss=0.1724, simple_loss=0.2441, pruned_loss=0.05036, over 953725.90 frames. ], batch size: 28, lr: 3.23e-03, grad_scale: 32.0 2023-04-27 14:24:37,722 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.3684, 1.2735, 1.5978, 1.5730, 1.2643, 1.1919, 1.2879, 0.8796], device='cuda:1'), covar=tensor([0.0533, 0.0624, 0.0350, 0.0533, 0.0814, 0.1136, 0.0584, 0.0563], device='cuda:1'), in_proj_covar=tensor([0.0069, 0.0069, 0.0068, 0.0068, 0.0076, 0.0097, 0.0074, 0.0066], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-27 14:24:41,780 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.085e+02 1.566e+02 1.857e+02 2.073e+02 3.044e+02, threshold=3.713e+02, percent-clipped=0.0 2023-04-27 14:24:43,151 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=112562.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:25:05,134 INFO [finetune.py:976] (1/7) Epoch 20, batch 3750, loss[loss=0.1919, simple_loss=0.2666, pruned_loss=0.05865, over 4811.00 frames. ], tot_loss[loss=0.1749, simple_loss=0.2467, pruned_loss=0.05159, over 956492.62 frames. ], batch size: 40, lr: 3.23e-03, grad_scale: 32.0 2023-04-27 14:25:05,268 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=112577.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:26:00,359 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=112617.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:26:10,693 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=112623.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:26:19,268 INFO [finetune.py:976] (1/7) Epoch 20, batch 3800, loss[loss=0.1697, simple_loss=0.2483, pruned_loss=0.04555, over 4789.00 frames. ], tot_loss[loss=0.1744, simple_loss=0.2469, pruned_loss=0.05094, over 955449.03 frames. ], batch size: 29, lr: 3.23e-03, grad_scale: 32.0 2023-04-27 14:26:31,074 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=112638.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 14:26:31,719 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7429, 1.2890, 1.8560, 2.1051, 1.7809, 1.7086, 1.7642, 1.7935], device='cuda:1'), covar=tensor([0.5875, 0.7277, 0.7444, 0.8270, 0.7101, 0.9450, 0.9048, 0.9101], device='cuda:1'), in_proj_covar=tensor([0.0429, 0.0411, 0.0506, 0.0507, 0.0456, 0.0486, 0.0492, 0.0497], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 14:27:02,949 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.113e+02 1.520e+02 1.863e+02 2.243e+02 5.060e+02, threshold=3.726e+02, percent-clipped=6.0 2023-04-27 14:27:25,229 INFO [finetune.py:976] (1/7) Epoch 20, batch 3850, loss[loss=0.2094, simple_loss=0.2841, pruned_loss=0.06732, over 4920.00 frames. ], tot_loss[loss=0.1725, simple_loss=0.2451, pruned_loss=0.04997, over 956314.16 frames. ], batch size: 42, lr: 3.23e-03, grad_scale: 32.0 2023-04-27 14:27:25,430 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-04-27 14:27:25,961 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=112678.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 14:27:57,616 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.67 vs. limit=5.0 2023-04-27 14:28:09,555 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=112722.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:28:11,397 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 2023-04-27 14:28:13,013 INFO [finetune.py:976] (1/7) Epoch 20, batch 3900, loss[loss=0.1731, simple_loss=0.245, pruned_loss=0.05065, over 4861.00 frames. ], tot_loss[loss=0.172, simple_loss=0.2436, pruned_loss=0.05021, over 957280.83 frames. ], batch size: 34, lr: 3.23e-03, grad_scale: 32.0 2023-04-27 14:28:17,109 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([3.1013, 2.4186, 2.2810, 2.4229, 2.2244, 2.2853, 2.3452, 2.3113], device='cuda:1'), covar=tensor([0.3419, 0.5615, 0.4778, 0.4711, 0.5513, 0.7070, 0.6157, 0.5709], device='cuda:1'), in_proj_covar=tensor([0.0335, 0.0371, 0.0320, 0.0334, 0.0344, 0.0393, 0.0354, 0.0327], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 14:28:24,995 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=112745.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:28:34,454 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.220e+01 1.530e+02 1.887e+02 2.253e+02 4.348e+02, threshold=3.774e+02, percent-clipped=2.0 2023-04-27 14:28:41,693 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=112770.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:28:45,926 INFO [finetune.py:976] (1/7) Epoch 20, batch 3950, loss[loss=0.1512, simple_loss=0.2196, pruned_loss=0.04141, over 4803.00 frames. ], tot_loss[loss=0.17, simple_loss=0.241, pruned_loss=0.04952, over 958645.70 frames. ], batch size: 29, lr: 3.23e-03, grad_scale: 32.0 2023-04-27 14:28:59,091 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.69 vs. limit=5.0 2023-04-27 14:29:04,077 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9406, 4.3840, 0.8766, 2.4149, 2.6963, 2.9292, 2.6421, 1.1186], device='cuda:1'), covar=tensor([0.1371, 0.0811, 0.2126, 0.1141, 0.0957, 0.0990, 0.1286, 0.1933], device='cuda:1'), in_proj_covar=tensor([0.0117, 0.0241, 0.0137, 0.0120, 0.0134, 0.0152, 0.0117, 0.0119], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-27 14:29:16,560 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.4305, 1.5902, 1.8185, 1.9053, 1.7838, 1.8484, 1.9044, 1.9015], device='cuda:1'), covar=tensor([0.3900, 0.5538, 0.4259, 0.4148, 0.5306, 0.7063, 0.5206, 0.4871], device='cuda:1'), in_proj_covar=tensor([0.0336, 0.0372, 0.0320, 0.0334, 0.0344, 0.0393, 0.0354, 0.0326], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 14:29:20,146 INFO [finetune.py:976] (1/7) Epoch 20, batch 4000, loss[loss=0.1927, simple_loss=0.2599, pruned_loss=0.0628, over 4824.00 frames. ], tot_loss[loss=0.1693, simple_loss=0.24, pruned_loss=0.04933, over 960252.56 frames. ], batch size: 51, lr: 3.23e-03, grad_scale: 32.0 2023-04-27 14:29:42,039 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.029e+02 1.556e+02 1.866e+02 2.319e+02 5.001e+02, threshold=3.732e+02, percent-clipped=1.0 2023-04-27 14:29:53,761 INFO [finetune.py:976] (1/7) Epoch 20, batch 4050, loss[loss=0.2699, simple_loss=0.3098, pruned_loss=0.115, over 4051.00 frames. ], tot_loss[loss=0.1735, simple_loss=0.2443, pruned_loss=0.05132, over 959712.86 frames. ], batch size: 65, lr: 3.23e-03, grad_scale: 32.0 2023-04-27 14:30:21,646 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=112918.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:30:23,588 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=112921.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:30:27,669 INFO [finetune.py:976] (1/7) Epoch 20, batch 4100, loss[loss=0.1827, simple_loss=0.2626, pruned_loss=0.05144, over 4815.00 frames. ], tot_loss[loss=0.1757, simple_loss=0.2475, pruned_loss=0.05199, over 959981.14 frames. ], batch size: 38, lr: 3.23e-03, grad_scale: 32.0 2023-04-27 14:30:31,351 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=112933.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 14:31:11,350 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.236e+02 1.611e+02 1.943e+02 2.315e+02 4.443e+02, threshold=3.885e+02, percent-clipped=3.0 2023-04-27 14:31:13,983 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.9320, 2.2562, 2.1589, 2.3233, 2.0682, 2.2038, 2.2315, 2.1864], device='cuda:1'), covar=tensor([0.4004, 0.5823, 0.4908, 0.4266, 0.5726, 0.7227, 0.6007, 0.5359], device='cuda:1'), in_proj_covar=tensor([0.0336, 0.0372, 0.0321, 0.0335, 0.0344, 0.0393, 0.0355, 0.0326], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 14:31:25,622 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=112973.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 14:31:33,459 INFO [finetune.py:976] (1/7) Epoch 20, batch 4150, loss[loss=0.1772, simple_loss=0.2531, pruned_loss=0.05064, over 4819.00 frames. ], tot_loss[loss=0.1758, simple_loss=0.2477, pruned_loss=0.05198, over 958221.95 frames. ], batch size: 30, lr: 3.23e-03, grad_scale: 32.0 2023-04-27 14:31:42,858 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=112982.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:32:29,692 INFO [finetune.py:976] (1/7) Epoch 20, batch 4200, loss[loss=0.1783, simple_loss=0.2489, pruned_loss=0.05386, over 4811.00 frames. ], tot_loss[loss=0.1763, simple_loss=0.2482, pruned_loss=0.05217, over 956259.23 frames. ], batch size: 38, lr: 3.23e-03, grad_scale: 32.0 2023-04-27 14:32:41,736 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=113045.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:32:57,027 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.082e+02 1.553e+02 1.872e+02 2.220e+02 4.301e+02, threshold=3.745e+02, percent-clipped=1.0 2023-04-27 14:33:18,845 INFO [finetune.py:976] (1/7) Epoch 20, batch 4250, loss[loss=0.1872, simple_loss=0.2507, pruned_loss=0.06184, over 4905.00 frames. ], tot_loss[loss=0.1753, simple_loss=0.2464, pruned_loss=0.05212, over 953358.62 frames. ], batch size: 32, lr: 3.23e-03, grad_scale: 32.0 2023-04-27 14:33:33,944 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.5355, 2.1871, 2.4828, 2.9182, 2.5160, 2.1178, 1.8805, 2.3840], device='cuda:1'), covar=tensor([0.2859, 0.2765, 0.1517, 0.1923, 0.2246, 0.2256, 0.3485, 0.1523], device='cuda:1'), in_proj_covar=tensor([0.0292, 0.0245, 0.0227, 0.0313, 0.0219, 0.0233, 0.0227, 0.0185], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-27 14:33:35,094 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=113093.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:34:04,770 INFO [finetune.py:976] (1/7) Epoch 20, batch 4300, loss[loss=0.134, simple_loss=0.2009, pruned_loss=0.03353, over 4827.00 frames. ], tot_loss[loss=0.1734, simple_loss=0.2438, pruned_loss=0.05148, over 952914.30 frames. ], batch size: 25, lr: 3.23e-03, grad_scale: 32.0 2023-04-27 14:34:16,362 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.54 vs. limit=2.0 2023-04-27 14:34:27,840 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.100e+02 1.602e+02 1.982e+02 2.267e+02 5.350e+02, threshold=3.963e+02, percent-clipped=3.0 2023-04-27 14:34:38,150 INFO [finetune.py:976] (1/7) Epoch 20, batch 4350, loss[loss=0.1058, simple_loss=0.181, pruned_loss=0.01534, over 4812.00 frames. ], tot_loss[loss=0.1691, simple_loss=0.2394, pruned_loss=0.0494, over 955580.78 frames. ], batch size: 25, lr: 3.23e-03, grad_scale: 32.0 2023-04-27 14:34:50,025 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.3688, 1.5000, 1.3376, 1.7220, 1.5445, 1.7634, 1.3817, 3.0167], device='cuda:1'), covar=tensor([0.0569, 0.0779, 0.0805, 0.1125, 0.0619, 0.0448, 0.0733, 0.0180], device='cuda:1'), in_proj_covar=tensor([0.0037, 0.0039, 0.0039, 0.0043, 0.0040, 0.0038, 0.0038, 0.0056], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:1') 2023-04-27 14:34:58,472 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0896, 2.4676, 0.9581, 1.2698, 1.6923, 1.2018, 3.0001, 1.4882], device='cuda:1'), covar=tensor([0.0675, 0.0594, 0.0784, 0.1365, 0.0536, 0.1079, 0.0355, 0.0711], device='cuda:1'), in_proj_covar=tensor([0.0051, 0.0065, 0.0048, 0.0046, 0.0049, 0.0052, 0.0074, 0.0051], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:1') 2023-04-27 14:35:06,251 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=113218.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:35:11,655 INFO [finetune.py:976] (1/7) Epoch 20, batch 4400, loss[loss=0.1355, simple_loss=0.1992, pruned_loss=0.03586, over 4122.00 frames. ], tot_loss[loss=0.1698, simple_loss=0.2404, pruned_loss=0.0496, over 955272.50 frames. ], batch size: 17, lr: 3.23e-03, grad_scale: 32.0 2023-04-27 14:35:15,420 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=113233.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:35:16,043 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=113234.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:35:34,720 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.136e+02 1.671e+02 1.942e+02 2.362e+02 3.791e+02, threshold=3.884e+02, percent-clipped=0.0 2023-04-27 14:35:38,468 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=113266.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:35:42,675 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=113273.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 14:35:45,062 INFO [finetune.py:976] (1/7) Epoch 20, batch 4450, loss[loss=0.1596, simple_loss=0.2356, pruned_loss=0.04174, over 4811.00 frames. ], tot_loss[loss=0.173, simple_loss=0.2446, pruned_loss=0.05065, over 956470.36 frames. ], batch size: 25, lr: 3.23e-03, grad_scale: 32.0 2023-04-27 14:35:45,128 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=113277.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:35:47,576 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=113281.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:35:56,993 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=113295.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:35:57,560 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.3625, 2.9763, 1.0366, 1.7680, 1.7244, 2.2275, 1.7719, 1.0525], device='cuda:1'), covar=tensor([0.1339, 0.1059, 0.1757, 0.1216, 0.1070, 0.0912, 0.1396, 0.1794], device='cuda:1'), in_proj_covar=tensor([0.0117, 0.0240, 0.0137, 0.0119, 0.0133, 0.0152, 0.0116, 0.0119], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-27 14:36:15,055 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=113321.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:36:18,632 INFO [finetune.py:976] (1/7) Epoch 20, batch 4500, loss[loss=0.1812, simple_loss=0.256, pruned_loss=0.05321, over 4894.00 frames. ], tot_loss[loss=0.1752, simple_loss=0.2469, pruned_loss=0.05173, over 955320.49 frames. ], batch size: 35, lr: 3.22e-03, grad_scale: 32.0 2023-04-27 14:36:30,254 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.70 vs. limit=2.0 2023-04-27 14:37:03,577 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.082e+01 1.625e+02 1.957e+02 2.322e+02 4.205e+02, threshold=3.914e+02, percent-clipped=1.0 2023-04-27 14:37:25,453 INFO [finetune.py:976] (1/7) Epoch 20, batch 4550, loss[loss=0.2234, simple_loss=0.3076, pruned_loss=0.06963, over 4838.00 frames. ], tot_loss[loss=0.1762, simple_loss=0.2489, pruned_loss=0.05174, over 957202.65 frames. ], batch size: 47, lr: 3.22e-03, grad_scale: 32.0 2023-04-27 14:38:24,622 INFO [finetune.py:976] (1/7) Epoch 20, batch 4600, loss[loss=0.1544, simple_loss=0.2286, pruned_loss=0.04014, over 4882.00 frames. ], tot_loss[loss=0.1749, simple_loss=0.2477, pruned_loss=0.0511, over 956139.70 frames. ], batch size: 32, lr: 3.22e-03, grad_scale: 32.0 2023-04-27 14:39:00,865 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.917e+01 1.466e+02 1.708e+02 2.051e+02 4.040e+02, threshold=3.416e+02, percent-clipped=1.0 2023-04-27 14:39:13,176 INFO [finetune.py:976] (1/7) Epoch 20, batch 4650, loss[loss=0.1593, simple_loss=0.2273, pruned_loss=0.04563, over 4870.00 frames. ], tot_loss[loss=0.1746, simple_loss=0.2463, pruned_loss=0.05149, over 955652.65 frames. ], batch size: 31, lr: 3.22e-03, grad_scale: 32.0 2023-04-27 14:39:22,493 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.4244, 1.6900, 1.9098, 2.0120, 1.8764, 1.9472, 1.9491, 1.9707], device='cuda:1'), covar=tensor([0.4275, 0.5848, 0.4757, 0.4379, 0.5785, 0.7228, 0.5583, 0.4922], device='cuda:1'), in_proj_covar=tensor([0.0336, 0.0373, 0.0320, 0.0335, 0.0345, 0.0393, 0.0356, 0.0327], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 14:39:47,027 INFO [finetune.py:976] (1/7) Epoch 20, batch 4700, loss[loss=0.1748, simple_loss=0.2443, pruned_loss=0.05272, over 4925.00 frames. ], tot_loss[loss=0.172, simple_loss=0.2428, pruned_loss=0.05056, over 956471.31 frames. ], batch size: 37, lr: 3.22e-03, grad_scale: 32.0 2023-04-27 14:40:08,116 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 8.331e+01 1.541e+02 1.809e+02 2.201e+02 3.607e+02, threshold=3.618e+02, percent-clipped=1.0 2023-04-27 14:40:19,997 INFO [finetune.py:976] (1/7) Epoch 20, batch 4750, loss[loss=0.1602, simple_loss=0.2323, pruned_loss=0.04408, over 4818.00 frames. ], tot_loss[loss=0.1698, simple_loss=0.2402, pruned_loss=0.04975, over 956820.26 frames. ], batch size: 39, lr: 3.22e-03, grad_scale: 32.0 2023-04-27 14:40:20,087 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=113577.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:40:29,029 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=113590.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:40:43,913 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9568, 4.2816, 0.9845, 2.1904, 2.5732, 2.9079, 2.4899, 0.9744], device='cuda:1'), covar=tensor([0.1361, 0.0797, 0.1910, 0.1316, 0.0996, 0.1021, 0.1489, 0.2245], device='cuda:1'), in_proj_covar=tensor([0.0116, 0.0240, 0.0137, 0.0119, 0.0133, 0.0152, 0.0116, 0.0119], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-27 14:40:46,848 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=113617.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:40:52,161 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=113625.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:40:53,787 INFO [finetune.py:976] (1/7) Epoch 20, batch 4800, loss[loss=0.1857, simple_loss=0.2637, pruned_loss=0.05391, over 4903.00 frames. ], tot_loss[loss=0.1707, simple_loss=0.2416, pruned_loss=0.04988, over 954162.43 frames. ], batch size: 36, lr: 3.22e-03, grad_scale: 32.0 2023-04-27 14:41:06,495 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=113646.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:41:15,405 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.167e+02 1.571e+02 1.898e+02 2.197e+02 3.707e+02, threshold=3.796e+02, percent-clipped=1.0 2023-04-27 14:41:27,086 INFO [finetune.py:976] (1/7) Epoch 20, batch 4850, loss[loss=0.2317, simple_loss=0.3005, pruned_loss=0.08143, over 4262.00 frames. ], tot_loss[loss=0.1732, simple_loss=0.2446, pruned_loss=0.05087, over 953500.53 frames. ], batch size: 66, lr: 3.22e-03, grad_scale: 32.0 2023-04-27 14:41:27,813 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=113678.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:41:37,978 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6230, 1.0011, 1.6291, 2.0632, 1.6871, 1.5475, 1.6055, 1.6036], device='cuda:1'), covar=tensor([0.4225, 0.6305, 0.5348, 0.5489, 0.5293, 0.6929, 0.6821, 0.7968], device='cuda:1'), in_proj_covar=tensor([0.0427, 0.0408, 0.0501, 0.0502, 0.0455, 0.0482, 0.0490, 0.0496], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 14:41:46,449 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=113707.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:42:00,310 INFO [finetune.py:976] (1/7) Epoch 20, batch 4900, loss[loss=0.1618, simple_loss=0.241, pruned_loss=0.04126, over 4744.00 frames. ], tot_loss[loss=0.1736, simple_loss=0.2454, pruned_loss=0.05092, over 951974.69 frames. ], batch size: 27, lr: 3.22e-03, grad_scale: 32.0 2023-04-27 14:42:27,118 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.158e+02 1.699e+02 1.953e+02 2.350e+02 6.319e+02, threshold=3.906e+02, percent-clipped=4.0 2023-04-27 14:42:38,976 INFO [finetune.py:976] (1/7) Epoch 20, batch 4950, loss[loss=0.1636, simple_loss=0.2387, pruned_loss=0.04429, over 4780.00 frames. ], tot_loss[loss=0.1759, simple_loss=0.2478, pruned_loss=0.05202, over 953202.68 frames. ], batch size: 26, lr: 3.22e-03, grad_scale: 32.0 2023-04-27 14:43:27,564 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0199, 2.7138, 2.1267, 2.1544, 1.4434, 1.4448, 2.2793, 1.4681], device='cuda:1'), covar=tensor([0.1587, 0.1413, 0.1357, 0.1641, 0.2235, 0.1882, 0.0938, 0.1997], device='cuda:1'), in_proj_covar=tensor([0.0198, 0.0211, 0.0168, 0.0204, 0.0199, 0.0185, 0.0155, 0.0186], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-27 14:43:30,853 INFO [finetune.py:976] (1/7) Epoch 20, batch 5000, loss[loss=0.1955, simple_loss=0.2605, pruned_loss=0.06528, over 4851.00 frames. ], tot_loss[loss=0.1742, simple_loss=0.2461, pruned_loss=0.0511, over 954647.70 frames. ], batch size: 44, lr: 3.22e-03, grad_scale: 32.0 2023-04-27 14:44:14,003 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.052e+02 1.541e+02 1.777e+02 2.201e+02 3.232e+02, threshold=3.554e+02, percent-clipped=0.0 2023-04-27 14:44:30,266 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7042, 1.7353, 0.8155, 1.3882, 1.7590, 1.5843, 1.4389, 1.5357], device='cuda:1'), covar=tensor([0.0481, 0.0375, 0.0357, 0.0547, 0.0271, 0.0537, 0.0522, 0.0559], device='cuda:1'), in_proj_covar=tensor([0.0028, 0.0024, 0.0022, 0.0029, 0.0019, 0.0028, 0.0028, 0.0028], device='cuda:1'), out_proj_covar=tensor([0.0050, 0.0045, 0.0038, 0.0052, 0.0038, 0.0050, 0.0050, 0.0051], device='cuda:1') 2023-04-27 14:44:32,041 INFO [finetune.py:976] (1/7) Epoch 20, batch 5050, loss[loss=0.1478, simple_loss=0.2199, pruned_loss=0.03782, over 4819.00 frames. ], tot_loss[loss=0.171, simple_loss=0.2426, pruned_loss=0.04973, over 956978.79 frames. ], batch size: 25, lr: 3.22e-03, grad_scale: 32.0 2023-04-27 14:44:42,631 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=3.65 vs. limit=5.0 2023-04-27 14:44:44,791 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([3.3284, 3.2122, 2.5018, 3.8600, 3.2714, 3.3187, 1.3781, 3.2680], device='cuda:1'), covar=tensor([0.2223, 0.1445, 0.3380, 0.2271, 0.2745, 0.2043, 0.6168, 0.2585], device='cuda:1'), in_proj_covar=tensor([0.0247, 0.0216, 0.0253, 0.0307, 0.0297, 0.0248, 0.0275, 0.0272], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 14:44:52,857 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=113890.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:44:59,612 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9419, 2.2868, 0.9890, 1.2924, 1.6363, 1.1938, 2.4595, 1.3616], device='cuda:1'), covar=tensor([0.0665, 0.0572, 0.0603, 0.1268, 0.0430, 0.1010, 0.0308, 0.0711], device='cuda:1'), in_proj_covar=tensor([0.0050, 0.0065, 0.0047, 0.0046, 0.0049, 0.0052, 0.0073, 0.0051], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:1') 2023-04-27 14:45:34,210 INFO [finetune.py:976] (1/7) Epoch 20, batch 5100, loss[loss=0.1658, simple_loss=0.2375, pruned_loss=0.04707, over 4795.00 frames. ], tot_loss[loss=0.1685, simple_loss=0.2395, pruned_loss=0.04869, over 957310.29 frames. ], batch size: 29, lr: 3.22e-03, grad_scale: 64.0 2023-04-27 14:45:42,067 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=113938.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:45:57,340 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.243e+01 1.576e+02 1.837e+02 2.238e+02 4.174e+02, threshold=3.673e+02, percent-clipped=2.0 2023-04-27 14:46:05,378 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=113973.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:46:05,401 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.2455, 3.1478, 0.8397, 1.7480, 1.8118, 2.3582, 1.9157, 1.0759], device='cuda:1'), covar=tensor([0.1441, 0.0955, 0.1965, 0.1261, 0.1044, 0.0909, 0.1360, 0.1896], device='cuda:1'), in_proj_covar=tensor([0.0117, 0.0242, 0.0138, 0.0120, 0.0134, 0.0153, 0.0117, 0.0119], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-27 14:46:08,236 INFO [finetune.py:976] (1/7) Epoch 20, batch 5150, loss[loss=0.1864, simple_loss=0.2685, pruned_loss=0.05221, over 4718.00 frames. ], tot_loss[loss=0.1699, simple_loss=0.2404, pruned_loss=0.04966, over 957213.29 frames. ], batch size: 59, lr: 3.22e-03, grad_scale: 64.0 2023-04-27 14:46:27,893 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=114002.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:46:43,540 INFO [finetune.py:976] (1/7) Epoch 20, batch 5200, loss[loss=0.1756, simple_loss=0.2436, pruned_loss=0.05383, over 4754.00 frames. ], tot_loss[loss=0.1726, simple_loss=0.2439, pruned_loss=0.05067, over 953503.44 frames. ], batch size: 28, lr: 3.22e-03, grad_scale: 64.0 2023-04-27 14:47:06,606 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 8.860e+01 1.682e+02 1.942e+02 2.331e+02 4.447e+02, threshold=3.884e+02, percent-clipped=1.0 2023-04-27 14:47:16,853 INFO [finetune.py:976] (1/7) Epoch 20, batch 5250, loss[loss=0.1369, simple_loss=0.1979, pruned_loss=0.03797, over 4719.00 frames. ], tot_loss[loss=0.1736, simple_loss=0.2452, pruned_loss=0.05105, over 951424.42 frames. ], batch size: 23, lr: 3.22e-03, grad_scale: 64.0 2023-04-27 14:47:47,848 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-04-27 14:47:50,695 INFO [finetune.py:976] (1/7) Epoch 20, batch 5300, loss[loss=0.1495, simple_loss=0.2324, pruned_loss=0.03324, over 4747.00 frames. ], tot_loss[loss=0.1744, simple_loss=0.2459, pruned_loss=0.05143, over 952317.03 frames. ], batch size: 26, lr: 3.22e-03, grad_scale: 64.0 2023-04-27 14:47:50,816 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=114127.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:48:13,280 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.642e+01 1.650e+02 1.908e+02 2.204e+02 4.976e+02, threshold=3.816e+02, percent-clipped=2.0 2023-04-27 14:48:15,089 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8501, 4.3308, 0.8691, 2.0634, 2.6619, 2.8832, 2.5006, 1.0551], device='cuda:1'), covar=tensor([0.1406, 0.0861, 0.2184, 0.1264, 0.0916, 0.1030, 0.1449, 0.2076], device='cuda:1'), in_proj_covar=tensor([0.0117, 0.0241, 0.0137, 0.0119, 0.0133, 0.0152, 0.0117, 0.0119], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-27 14:48:24,237 INFO [finetune.py:976] (1/7) Epoch 20, batch 5350, loss[loss=0.1533, simple_loss=0.2233, pruned_loss=0.04167, over 4751.00 frames. ], tot_loss[loss=0.1738, simple_loss=0.2458, pruned_loss=0.05085, over 950014.56 frames. ], batch size: 27, lr: 3.22e-03, grad_scale: 64.0 2023-04-27 14:48:31,412 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=114188.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:48:37,365 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.3343, 1.5702, 1.4488, 1.6564, 1.6436, 1.9414, 1.4247, 3.4203], device='cuda:1'), covar=tensor([0.0597, 0.0760, 0.0752, 0.1181, 0.0593, 0.0524, 0.0716, 0.0135], device='cuda:1'), in_proj_covar=tensor([0.0038, 0.0039, 0.0040, 0.0043, 0.0040, 0.0038, 0.0038, 0.0056], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0013, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:1') 2023-04-27 14:48:52,725 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-04-27 14:48:57,551 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.6493, 2.9882, 0.9904, 1.7556, 2.2943, 1.5370, 4.1316, 2.0167], device='cuda:1'), covar=tensor([0.0557, 0.0825, 0.0902, 0.1266, 0.0506, 0.0970, 0.0270, 0.0623], device='cuda:1'), in_proj_covar=tensor([0.0050, 0.0065, 0.0048, 0.0046, 0.0049, 0.0052, 0.0073, 0.0052], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:1') 2023-04-27 14:48:58,063 INFO [finetune.py:976] (1/7) Epoch 20, batch 5400, loss[loss=0.167, simple_loss=0.2308, pruned_loss=0.05162, over 4897.00 frames. ], tot_loss[loss=0.1719, simple_loss=0.2434, pruned_loss=0.05025, over 950148.06 frames. ], batch size: 35, lr: 3.22e-03, grad_scale: 64.0 2023-04-27 14:49:33,868 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.776e+01 1.553e+02 1.821e+02 2.286e+02 4.099e+02, threshold=3.642e+02, percent-clipped=1.0 2023-04-27 14:49:34,016 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.1263, 2.7168, 2.1602, 2.1477, 1.5602, 1.5640, 2.1795, 1.4915], device='cuda:1'), covar=tensor([0.1513, 0.1364, 0.1267, 0.1596, 0.2136, 0.1784, 0.0926, 0.1935], device='cuda:1'), in_proj_covar=tensor([0.0198, 0.0212, 0.0169, 0.0204, 0.0200, 0.0185, 0.0156, 0.0186], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-27 14:49:53,935 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=114273.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:49:56,280 INFO [finetune.py:976] (1/7) Epoch 20, batch 5450, loss[loss=0.1475, simple_loss=0.2294, pruned_loss=0.03279, over 4901.00 frames. ], tot_loss[loss=0.1718, simple_loss=0.2422, pruned_loss=0.05073, over 951634.99 frames. ], batch size: 32, lr: 3.22e-03, grad_scale: 64.0 2023-04-27 14:50:06,098 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-04-27 14:50:27,132 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.49 vs. limit=5.0 2023-04-27 14:50:28,217 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=114302.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:50:30,549 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.1679, 2.7150, 2.3373, 2.4766, 1.9050, 2.2668, 2.1794, 1.8410], device='cuda:1'), covar=tensor([0.1897, 0.1177, 0.0613, 0.1047, 0.3355, 0.1161, 0.1978, 0.2504], device='cuda:1'), in_proj_covar=tensor([0.0289, 0.0303, 0.0218, 0.0279, 0.0315, 0.0258, 0.0251, 0.0264], device='cuda:1'), out_proj_covar=tensor([1.1572e-04, 1.2056e-04, 8.6455e-05, 1.1050e-04, 1.2767e-04, 1.0226e-04, 1.0164e-04, 1.0456e-04], device='cuda:1') 2023-04-27 14:50:58,433 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=114321.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:51:02,106 INFO [finetune.py:976] (1/7) Epoch 20, batch 5500, loss[loss=0.1681, simple_loss=0.216, pruned_loss=0.06006, over 4196.00 frames. ], tot_loss[loss=0.1704, simple_loss=0.2406, pruned_loss=0.05009, over 954174.33 frames. ], batch size: 18, lr: 3.22e-03, grad_scale: 64.0 2023-04-27 14:51:11,998 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=3.43 vs. limit=5.0 2023-04-27 14:51:22,114 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=114350.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:51:28,689 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.001e+02 1.538e+02 1.938e+02 2.403e+02 5.552e+02, threshold=3.877e+02, percent-clipped=2.0 2023-04-27 14:51:34,779 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-04-27 14:51:37,618 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4673, 1.4710, 1.8898, 1.8786, 1.4158, 1.2781, 1.5815, 1.0889], device='cuda:1'), covar=tensor([0.0532, 0.0690, 0.0365, 0.0706, 0.0726, 0.1039, 0.0546, 0.0572], device='cuda:1'), in_proj_covar=tensor([0.0069, 0.0069, 0.0067, 0.0068, 0.0076, 0.0097, 0.0073, 0.0065], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-27 14:51:41,081 INFO [finetune.py:976] (1/7) Epoch 20, batch 5550, loss[loss=0.1615, simple_loss=0.2412, pruned_loss=0.04092, over 4799.00 frames. ], tot_loss[loss=0.172, simple_loss=0.2426, pruned_loss=0.05071, over 954761.03 frames. ], batch size: 45, lr: 3.22e-03, grad_scale: 64.0 2023-04-27 14:51:41,898 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.48 vs. limit=2.0 2023-04-27 14:52:12,738 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.1925, 1.3118, 1.1864, 1.4605, 1.3642, 1.4094, 1.2095, 2.2107], device='cuda:1'), covar=tensor([0.0529, 0.0695, 0.0730, 0.1038, 0.0549, 0.0579, 0.0680, 0.0285], device='cuda:1'), in_proj_covar=tensor([0.0038, 0.0039, 0.0040, 0.0043, 0.0040, 0.0038, 0.0038, 0.0056], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0013, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:1') 2023-04-27 14:52:13,272 INFO [finetune.py:976] (1/7) Epoch 20, batch 5600, loss[loss=0.2298, simple_loss=0.2984, pruned_loss=0.08059, over 4062.00 frames. ], tot_loss[loss=0.175, simple_loss=0.2462, pruned_loss=0.05188, over 952561.40 frames. ], batch size: 65, lr: 3.22e-03, grad_scale: 64.0 2023-04-27 14:52:16,232 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5666, 1.7161, 0.8397, 1.2643, 1.8197, 1.4569, 1.3851, 1.4349], device='cuda:1'), covar=tensor([0.0521, 0.0373, 0.0340, 0.0566, 0.0270, 0.0524, 0.0490, 0.0582], device='cuda:1'), in_proj_covar=tensor([0.0028, 0.0024, 0.0022, 0.0029, 0.0019, 0.0028, 0.0028, 0.0028], device='cuda:1'), out_proj_covar=tensor([0.0050, 0.0045, 0.0038, 0.0052, 0.0038, 0.0050, 0.0050, 0.0051], device='cuda:1') 2023-04-27 14:52:20,936 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7130, 1.9081, 1.0883, 1.4326, 2.0727, 1.6061, 1.5156, 1.5659], device='cuda:1'), covar=tensor([0.0488, 0.0345, 0.0300, 0.0526, 0.0258, 0.0485, 0.0480, 0.0539], device='cuda:1'), in_proj_covar=tensor([0.0028, 0.0024, 0.0022, 0.0029, 0.0019, 0.0028, 0.0028, 0.0028], device='cuda:1'), out_proj_covar=tensor([0.0051, 0.0045, 0.0038, 0.0052, 0.0038, 0.0050, 0.0050, 0.0051], device='cuda:1') 2023-04-27 14:52:32,508 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.241e+02 1.607e+02 1.931e+02 2.416e+02 4.718e+02, threshold=3.861e+02, percent-clipped=3.0 2023-04-27 14:52:42,449 INFO [finetune.py:976] (1/7) Epoch 20, batch 5650, loss[loss=0.1579, simple_loss=0.23, pruned_loss=0.04289, over 4828.00 frames. ], tot_loss[loss=0.1755, simple_loss=0.2477, pruned_loss=0.05166, over 954686.65 frames. ], batch size: 30, lr: 3.22e-03, grad_scale: 64.0 2023-04-27 14:52:46,428 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=114483.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:53:01,367 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=3.89 vs. limit=5.0 2023-04-27 14:53:12,741 INFO [finetune.py:976] (1/7) Epoch 20, batch 5700, loss[loss=0.125, simple_loss=0.1881, pruned_loss=0.03089, over 3994.00 frames. ], tot_loss[loss=0.1729, simple_loss=0.2444, pruned_loss=0.0507, over 942994.38 frames. ], batch size: 17, lr: 3.22e-03, grad_scale: 64.0 2023-04-27 14:53:39,071 INFO [finetune.py:976] (1/7) Epoch 21, batch 0, loss[loss=0.14, simple_loss=0.2207, pruned_loss=0.0296, over 4769.00 frames. ], tot_loss[loss=0.14, simple_loss=0.2207, pruned_loss=0.0296, over 4769.00 frames. ], batch size: 29, lr: 3.21e-03, grad_scale: 64.0 2023-04-27 14:53:39,071 INFO [finetune.py:1001] (1/7) Computing validation loss 2023-04-27 14:53:45,612 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.3926, 1.4012, 3.8609, 3.5960, 3.4803, 3.7103, 3.8119, 3.4153], device='cuda:1'), covar=tensor([0.6568, 0.5268, 0.1213, 0.1930, 0.1172, 0.1309, 0.0738, 0.1543], device='cuda:1'), in_proj_covar=tensor([0.0309, 0.0306, 0.0403, 0.0403, 0.0347, 0.0407, 0.0311, 0.0365], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-27 14:53:56,091 INFO [finetune.py:1010] (1/7) Epoch 21, validation: loss=0.1544, simple_loss=0.2245, pruned_loss=0.04212, over 2265189.00 frames. 2023-04-27 14:53:56,092 INFO [finetune.py:1011] (1/7) Maximum memory allocated so far is 6391MB 2023-04-27 14:54:02,564 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 7.383e+01 1.462e+02 1.753e+02 2.110e+02 4.375e+02, threshold=3.507e+02, percent-clipped=1.0 2023-04-27 14:54:24,908 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.1461, 1.8405, 2.3183, 2.6654, 2.1589, 2.0447, 2.1864, 2.1460], device='cuda:1'), covar=tensor([0.4508, 0.7257, 0.6974, 0.5540, 0.6039, 0.8683, 0.8862, 0.9065], device='cuda:1'), in_proj_covar=tensor([0.0427, 0.0410, 0.0504, 0.0503, 0.0455, 0.0483, 0.0493, 0.0496], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 14:54:37,746 INFO [finetune.py:976] (1/7) Epoch 21, batch 50, loss[loss=0.1715, simple_loss=0.2552, pruned_loss=0.04389, over 4922.00 frames. ], tot_loss[loss=0.1745, simple_loss=0.2457, pruned_loss=0.05165, over 215891.53 frames. ], batch size: 33, lr: 3.21e-03, grad_scale: 64.0 2023-04-27 14:55:05,756 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.4350, 1.7457, 1.8196, 1.9885, 1.8301, 1.8627, 1.9051, 1.9320], device='cuda:1'), covar=tensor([0.4201, 0.4986, 0.4347, 0.4070, 0.5061, 0.7104, 0.4836, 0.4589], device='cuda:1'), in_proj_covar=tensor([0.0340, 0.0376, 0.0322, 0.0337, 0.0348, 0.0395, 0.0358, 0.0329], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 14:55:26,964 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-04-27 14:55:37,893 INFO [finetune.py:976] (1/7) Epoch 21, batch 100, loss[loss=0.141, simple_loss=0.2195, pruned_loss=0.03122, over 4725.00 frames. ], tot_loss[loss=0.1693, simple_loss=0.2396, pruned_loss=0.04948, over 380864.09 frames. ], batch size: 54, lr: 3.21e-03, grad_scale: 64.0 2023-04-27 14:55:46,620 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-04-27 14:55:48,247 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.033e+02 1.526e+02 1.765e+02 2.101e+02 5.147e+02, threshold=3.531e+02, percent-clipped=4.0 2023-04-27 14:56:44,428 INFO [finetune.py:976] (1/7) Epoch 21, batch 150, loss[loss=0.171, simple_loss=0.2348, pruned_loss=0.05363, over 4871.00 frames. ], tot_loss[loss=0.1695, simple_loss=0.2385, pruned_loss=0.05026, over 510157.35 frames. ], batch size: 31, lr: 3.21e-03, grad_scale: 64.0 2023-04-27 14:56:57,806 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=114716.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:56:58,595 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.81 vs. limit=2.0 2023-04-27 14:57:12,223 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.1040, 2.2570, 1.8845, 1.8580, 2.2639, 1.8591, 2.7105, 1.6992], device='cuda:1'), covar=tensor([0.4151, 0.1758, 0.4298, 0.3027, 0.1748, 0.2420, 0.1545, 0.4135], device='cuda:1'), in_proj_covar=tensor([0.0344, 0.0350, 0.0427, 0.0354, 0.0383, 0.0374, 0.0374, 0.0420], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 14:57:22,715 INFO [finetune.py:976] (1/7) Epoch 21, batch 200, loss[loss=0.1753, simple_loss=0.2569, pruned_loss=0.04679, over 4778.00 frames. ], tot_loss[loss=0.1686, simple_loss=0.2382, pruned_loss=0.04957, over 609014.97 frames. ], batch size: 54, lr: 3.21e-03, grad_scale: 64.0 2023-04-27 14:57:26,735 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.046e+02 1.480e+02 1.757e+02 1.981e+02 3.579e+02, threshold=3.513e+02, percent-clipped=1.0 2023-04-27 14:57:38,718 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=114777.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:57:42,327 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=114783.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:57:56,199 INFO [finetune.py:976] (1/7) Epoch 21, batch 250, loss[loss=0.1875, simple_loss=0.2482, pruned_loss=0.06344, over 4890.00 frames. ], tot_loss[loss=0.1737, simple_loss=0.2439, pruned_loss=0.05176, over 685166.70 frames. ], batch size: 32, lr: 3.21e-03, grad_scale: 32.0 2023-04-27 14:58:00,576 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-04-27 14:58:15,123 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=114831.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:58:30,014 INFO [finetune.py:976] (1/7) Epoch 21, batch 300, loss[loss=0.1814, simple_loss=0.2677, pruned_loss=0.04757, over 4813.00 frames. ], tot_loss[loss=0.1752, simple_loss=0.2467, pruned_loss=0.05187, over 745612.95 frames. ], batch size: 41, lr: 3.21e-03, grad_scale: 32.0 2023-04-27 14:58:32,534 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.4716, 2.1533, 2.3940, 2.8478, 2.8335, 2.1897, 1.7783, 2.4080], device='cuda:1'), covar=tensor([0.0815, 0.1085, 0.0755, 0.0520, 0.0593, 0.0864, 0.0802, 0.0601], device='cuda:1'), in_proj_covar=tensor([0.0191, 0.0205, 0.0186, 0.0174, 0.0180, 0.0183, 0.0154, 0.0181], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 14:58:34,664 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 8.745e+01 1.667e+02 1.887e+02 2.264e+02 4.948e+02, threshold=3.774e+02, percent-clipped=2.0 2023-04-27 14:58:42,279 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=114871.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:58:43,514 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9632, 2.4222, 1.8753, 1.7741, 1.5721, 1.5693, 1.8946, 1.5175], device='cuda:1'), covar=tensor([0.1785, 0.1339, 0.1571, 0.1805, 0.2435, 0.2159, 0.1085, 0.1965], device='cuda:1'), in_proj_covar=tensor([0.0199, 0.0212, 0.0170, 0.0204, 0.0201, 0.0186, 0.0157, 0.0187], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-27 14:59:03,087 INFO [finetune.py:976] (1/7) Epoch 21, batch 350, loss[loss=0.1395, simple_loss=0.211, pruned_loss=0.03405, over 4221.00 frames. ], tot_loss[loss=0.1777, simple_loss=0.2491, pruned_loss=0.05315, over 789859.30 frames. ], batch size: 18, lr: 3.21e-03, grad_scale: 32.0 2023-04-27 14:59:22,176 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=114932.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:59:36,225 INFO [finetune.py:976] (1/7) Epoch 21, batch 400, loss[loss=0.1422, simple_loss=0.2339, pruned_loss=0.0253, over 4772.00 frames. ], tot_loss[loss=0.1758, simple_loss=0.2478, pruned_loss=0.05187, over 826089.04 frames. ], batch size: 26, lr: 3.21e-03, grad_scale: 32.0 2023-04-27 14:59:40,904 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.017e+02 1.660e+02 1.983e+02 2.165e+02 4.861e+02, threshold=3.966e+02, percent-clipped=1.0 2023-04-27 15:00:10,133 INFO [finetune.py:976] (1/7) Epoch 21, batch 450, loss[loss=0.1647, simple_loss=0.2274, pruned_loss=0.05102, over 4822.00 frames. ], tot_loss[loss=0.175, simple_loss=0.2471, pruned_loss=0.05144, over 854543.04 frames. ], batch size: 41, lr: 3.21e-03, grad_scale: 32.0 2023-04-27 15:00:25,639 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.4036, 2.2076, 2.5658, 2.9099, 2.8618, 2.2005, 1.8355, 2.4588], device='cuda:1'), covar=tensor([0.0808, 0.1041, 0.0577, 0.0481, 0.0575, 0.0904, 0.0811, 0.0564], device='cuda:1'), in_proj_covar=tensor([0.0189, 0.0204, 0.0185, 0.0173, 0.0179, 0.0182, 0.0153, 0.0180], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 15:00:59,032 INFO [finetune.py:976] (1/7) Epoch 21, batch 500, loss[loss=0.1495, simple_loss=0.2267, pruned_loss=0.03621, over 4799.00 frames. ], tot_loss[loss=0.1726, simple_loss=0.2443, pruned_loss=0.05044, over 875920.41 frames. ], batch size: 29, lr: 3.21e-03, grad_scale: 32.0 2023-04-27 15:01:09,614 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 8.974e+01 1.571e+02 1.812e+02 2.235e+02 3.288e+02, threshold=3.623e+02, percent-clipped=0.0 2023-04-27 15:01:16,897 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=115072.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:01:32,715 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=115086.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:01:54,522 INFO [finetune.py:976] (1/7) Epoch 21, batch 550, loss[loss=0.1991, simple_loss=0.2618, pruned_loss=0.06817, over 4901.00 frames. ], tot_loss[loss=0.171, simple_loss=0.2421, pruned_loss=0.04997, over 895133.43 frames. ], batch size: 32, lr: 3.21e-03, grad_scale: 32.0 2023-04-27 15:02:09,093 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.2094, 2.7458, 2.1824, 2.3160, 1.5402, 1.5288, 2.2729, 1.4896], device='cuda:1'), covar=tensor([0.1694, 0.1491, 0.1369, 0.1524, 0.2343, 0.1907, 0.0992, 0.2002], device='cuda:1'), in_proj_covar=tensor([0.0198, 0.0211, 0.0169, 0.0203, 0.0199, 0.0185, 0.0156, 0.0187], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-27 15:02:53,400 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=115147.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:03:02,665 INFO [finetune.py:976] (1/7) Epoch 21, batch 600, loss[loss=0.197, simple_loss=0.264, pruned_loss=0.06503, over 4867.00 frames. ], tot_loss[loss=0.1733, simple_loss=0.2439, pruned_loss=0.0514, over 908588.84 frames. ], batch size: 34, lr: 3.21e-03, grad_scale: 32.0 2023-04-27 15:03:03,478 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-04-27 15:03:12,536 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.066e+02 1.494e+02 1.729e+02 2.315e+02 4.392e+02, threshold=3.458e+02, percent-clipped=4.0 2023-04-27 15:03:56,685 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-04-27 15:04:02,673 INFO [finetune.py:976] (1/7) Epoch 21, batch 650, loss[loss=0.1955, simple_loss=0.2623, pruned_loss=0.06436, over 4861.00 frames. ], tot_loss[loss=0.1756, simple_loss=0.2469, pruned_loss=0.05216, over 919391.47 frames. ], batch size: 44, lr: 3.21e-03, grad_scale: 32.0 2023-04-27 15:04:05,207 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.1838, 1.3884, 1.3158, 1.6725, 1.5197, 1.5735, 1.3209, 2.5217], device='cuda:1'), covar=tensor([0.0570, 0.0850, 0.0847, 0.1216, 0.0666, 0.0449, 0.0787, 0.0221], device='cuda:1'), in_proj_covar=tensor([0.0038, 0.0038, 0.0039, 0.0043, 0.0040, 0.0038, 0.0038, 0.0056], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:1') 2023-04-27 15:04:17,775 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=115227.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:04:17,817 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.2399, 2.9469, 2.4146, 2.6618, 1.9952, 2.5135, 2.5627, 1.9171], device='cuda:1'), covar=tensor([0.2155, 0.0964, 0.0818, 0.1208, 0.2968, 0.1161, 0.1869, 0.2722], device='cuda:1'), in_proj_covar=tensor([0.0287, 0.0301, 0.0216, 0.0277, 0.0312, 0.0255, 0.0249, 0.0263], device='cuda:1'), out_proj_covar=tensor([1.1482e-04, 1.1968e-04, 8.5674e-05, 1.0962e-04, 1.2634e-04, 1.0101e-04, 1.0063e-04, 1.0423e-04], device='cuda:1') 2023-04-27 15:04:36,594 INFO [finetune.py:976] (1/7) Epoch 21, batch 700, loss[loss=0.2234, simple_loss=0.2752, pruned_loss=0.08586, over 4895.00 frames. ], tot_loss[loss=0.1759, simple_loss=0.2478, pruned_loss=0.05194, over 927619.75 frames. ], batch size: 43, lr: 3.21e-03, grad_scale: 32.0 2023-04-27 15:04:40,862 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.165e+02 1.680e+02 1.934e+02 2.254e+02 3.960e+02, threshold=3.868e+02, percent-clipped=2.0 2023-04-27 15:05:10,553 INFO [finetune.py:976] (1/7) Epoch 21, batch 750, loss[loss=0.1768, simple_loss=0.2604, pruned_loss=0.04665, over 4813.00 frames. ], tot_loss[loss=0.176, simple_loss=0.2484, pruned_loss=0.05177, over 934082.58 frames. ], batch size: 33, lr: 3.21e-03, grad_scale: 32.0 2023-04-27 15:05:11,285 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.2909, 1.7798, 2.1913, 2.6139, 2.2212, 1.7693, 1.3850, 2.0724], device='cuda:1'), covar=tensor([0.3490, 0.3332, 0.1803, 0.2355, 0.2745, 0.2877, 0.4274, 0.1887], device='cuda:1'), in_proj_covar=tensor([0.0291, 0.0244, 0.0225, 0.0312, 0.0219, 0.0232, 0.0225, 0.0183], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-27 15:05:11,404 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.74 vs. limit=2.0 2023-04-27 15:05:44,410 INFO [finetune.py:976] (1/7) Epoch 21, batch 800, loss[loss=0.1909, simple_loss=0.2621, pruned_loss=0.05983, over 4805.00 frames. ], tot_loss[loss=0.1762, simple_loss=0.2483, pruned_loss=0.05201, over 938292.00 frames. ], batch size: 45, lr: 3.21e-03, grad_scale: 32.0 2023-04-27 15:05:48,601 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.653e+01 1.488e+02 1.739e+02 2.068e+02 3.121e+02, threshold=3.478e+02, percent-clipped=0.0 2023-04-27 15:05:54,623 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=115370.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:05:55,244 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=115371.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:05:55,820 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=115372.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:06:18,145 INFO [finetune.py:976] (1/7) Epoch 21, batch 850, loss[loss=0.1532, simple_loss=0.2279, pruned_loss=0.0393, over 4830.00 frames. ], tot_loss[loss=0.1745, simple_loss=0.2459, pruned_loss=0.05153, over 943233.28 frames. ], batch size: 33, lr: 3.21e-03, grad_scale: 32.0 2023-04-27 15:06:27,932 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=115420.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:06:29,720 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=3.52 vs. limit=5.0 2023-04-27 15:06:35,548 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=115431.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:06:36,163 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=115432.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:06:37,971 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6829, 1.7406, 0.7301, 1.3870, 1.7237, 1.5541, 1.4717, 1.4983], device='cuda:1'), covar=tensor([0.0506, 0.0372, 0.0369, 0.0565, 0.0281, 0.0533, 0.0517, 0.0562], device='cuda:1'), in_proj_covar=tensor([0.0028, 0.0024, 0.0022, 0.0028, 0.0019, 0.0028, 0.0028, 0.0028], device='cuda:1'), out_proj_covar=tensor([0.0050, 0.0045, 0.0038, 0.0051, 0.0038, 0.0049, 0.0050, 0.0051], device='cuda:1') 2023-04-27 15:06:40,880 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9521, 2.4769, 2.0204, 2.3138, 1.7100, 2.1380, 2.0003, 1.5189], device='cuda:1'), covar=tensor([0.2024, 0.1209, 0.0904, 0.1340, 0.3418, 0.1171, 0.2119, 0.3059], device='cuda:1'), in_proj_covar=tensor([0.0284, 0.0298, 0.0215, 0.0274, 0.0309, 0.0252, 0.0247, 0.0261], device='cuda:1'), out_proj_covar=tensor([1.1373e-04, 1.1846e-04, 8.5029e-05, 1.0856e-04, 1.2521e-04, 1.0005e-04, 9.9686e-05, 1.0351e-04], device='cuda:1') 2023-04-27 15:06:42,653 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=115442.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:07:01,737 INFO [finetune.py:976] (1/7) Epoch 21, batch 900, loss[loss=0.1531, simple_loss=0.2228, pruned_loss=0.04169, over 4890.00 frames. ], tot_loss[loss=0.1715, simple_loss=0.2427, pruned_loss=0.05012, over 946525.36 frames. ], batch size: 35, lr: 3.20e-03, grad_scale: 32.0 2023-04-27 15:07:06,010 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.637e+01 1.508e+02 1.757e+02 2.120e+02 3.037e+02, threshold=3.515e+02, percent-clipped=0.0 2023-04-27 15:07:34,516 INFO [finetune.py:976] (1/7) Epoch 21, batch 950, loss[loss=0.2058, simple_loss=0.2635, pruned_loss=0.07404, over 4824.00 frames. ], tot_loss[loss=0.1717, simple_loss=0.2423, pruned_loss=0.05058, over 947392.85 frames. ], batch size: 30, lr: 3.20e-03, grad_scale: 32.0 2023-04-27 15:08:06,296 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=115527.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:08:39,545 INFO [finetune.py:976] (1/7) Epoch 21, batch 1000, loss[loss=0.2161, simple_loss=0.2844, pruned_loss=0.07385, over 4746.00 frames. ], tot_loss[loss=0.1716, simple_loss=0.2427, pruned_loss=0.05027, over 948012.71 frames. ], batch size: 54, lr: 3.20e-03, grad_scale: 32.0 2023-04-27 15:08:49,079 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.133e+02 1.678e+02 2.025e+02 2.586e+02 4.511e+02, threshold=4.050e+02, percent-clipped=4.0 2023-04-27 15:09:00,576 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=3.14 vs. limit=5.0 2023-04-27 15:09:08,294 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=115575.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:09:09,630 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.3850, 1.6074, 1.6859, 1.8254, 1.6597, 1.7829, 1.7280, 1.7865], device='cuda:1'), covar=tensor([0.3535, 0.5156, 0.4121, 0.4093, 0.5444, 0.7112, 0.4905, 0.4709], device='cuda:1'), in_proj_covar=tensor([0.0336, 0.0374, 0.0321, 0.0334, 0.0344, 0.0393, 0.0355, 0.0326], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 15:09:20,649 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5446, 1.4146, 1.8908, 1.8971, 1.3721, 1.2837, 1.4992, 0.9304], device='cuda:1'), covar=tensor([0.0596, 0.0715, 0.0383, 0.0701, 0.0756, 0.1228, 0.0618, 0.0638], device='cuda:1'), in_proj_covar=tensor([0.0069, 0.0069, 0.0067, 0.0068, 0.0075, 0.0097, 0.0073, 0.0066], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-27 15:09:41,556 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.10 vs. limit=2.0 2023-04-27 15:09:45,484 INFO [finetune.py:976] (1/7) Epoch 21, batch 1050, loss[loss=0.1378, simple_loss=0.2106, pruned_loss=0.03252, over 4437.00 frames. ], tot_loss[loss=0.1735, simple_loss=0.2459, pruned_loss=0.05058, over 950433.09 frames. ], batch size: 19, lr: 3.20e-03, grad_scale: 32.0 2023-04-27 15:10:18,955 INFO [finetune.py:976] (1/7) Epoch 21, batch 1100, loss[loss=0.2364, simple_loss=0.2914, pruned_loss=0.09064, over 4929.00 frames. ], tot_loss[loss=0.1741, simple_loss=0.2472, pruned_loss=0.05051, over 951790.47 frames. ], batch size: 42, lr: 3.20e-03, grad_scale: 32.0 2023-04-27 15:10:23,752 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.026e+02 1.551e+02 1.797e+02 2.277e+02 4.571e+02, threshold=3.594e+02, percent-clipped=1.0 2023-04-27 15:10:44,998 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.54 vs. limit=5.0 2023-04-27 15:10:52,442 INFO [finetune.py:976] (1/7) Epoch 21, batch 1150, loss[loss=0.1878, simple_loss=0.2613, pruned_loss=0.05718, over 4868.00 frames. ], tot_loss[loss=0.1734, simple_loss=0.2471, pruned_loss=0.04989, over 953378.17 frames. ], batch size: 34, lr: 3.20e-03, grad_scale: 32.0 2023-04-27 15:11:07,907 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=115726.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:11:08,523 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=115727.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:11:18,229 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=115742.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:11:25,942 INFO [finetune.py:976] (1/7) Epoch 21, batch 1200, loss[loss=0.1262, simple_loss=0.1951, pruned_loss=0.02863, over 3969.00 frames. ], tot_loss[loss=0.1721, simple_loss=0.2457, pruned_loss=0.04924, over 954321.94 frames. ], batch size: 17, lr: 3.20e-03, grad_scale: 32.0 2023-04-27 15:11:27,773 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.5351, 1.9704, 2.4538, 2.9989, 2.4158, 1.9009, 1.8275, 2.3457], device='cuda:1'), covar=tensor([0.3228, 0.3290, 0.1614, 0.2429, 0.2824, 0.2827, 0.3909, 0.2119], device='cuda:1'), in_proj_covar=tensor([0.0292, 0.0245, 0.0226, 0.0314, 0.0219, 0.0232, 0.0227, 0.0184], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-27 15:11:31,125 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.085e+02 1.519e+02 1.757e+02 2.035e+02 5.645e+02, threshold=3.514e+02, percent-clipped=1.0 2023-04-27 15:11:49,068 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.1964, 1.4792, 1.3436, 1.7473, 1.5578, 1.7627, 1.3519, 3.4338], device='cuda:1'), covar=tensor([0.0706, 0.1066, 0.1041, 0.1352, 0.0804, 0.0615, 0.0989, 0.0231], device='cuda:1'), in_proj_covar=tensor([0.0038, 0.0039, 0.0040, 0.0043, 0.0040, 0.0038, 0.0038, 0.0056], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0013, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:1') 2023-04-27 15:11:50,225 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=115790.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:11:59,754 INFO [finetune.py:976] (1/7) Epoch 21, batch 1250, loss[loss=0.1502, simple_loss=0.2271, pruned_loss=0.03664, over 4650.00 frames. ], tot_loss[loss=0.1705, simple_loss=0.2435, pruned_loss=0.04878, over 956295.99 frames. ], batch size: 23, lr: 3.20e-03, grad_scale: 32.0 2023-04-27 15:12:22,645 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8111, 2.1741, 1.8864, 2.1574, 1.6537, 1.8655, 1.7157, 1.4816], device='cuda:1'), covar=tensor([0.1514, 0.1167, 0.0698, 0.0911, 0.3056, 0.1055, 0.1758, 0.2080], device='cuda:1'), in_proj_covar=tensor([0.0285, 0.0299, 0.0215, 0.0275, 0.0309, 0.0253, 0.0247, 0.0261], device='cuda:1'), out_proj_covar=tensor([1.1388e-04, 1.1864e-04, 8.4956e-05, 1.0898e-04, 1.2540e-04, 1.0022e-04, 9.9941e-05, 1.0358e-04], device='cuda:1') 2023-04-27 15:12:55,528 INFO [finetune.py:976] (1/7) Epoch 21, batch 1300, loss[loss=0.126, simple_loss=0.1975, pruned_loss=0.02724, over 4915.00 frames. ], tot_loss[loss=0.1683, simple_loss=0.2402, pruned_loss=0.0482, over 955492.90 frames. ], batch size: 35, lr: 3.20e-03, grad_scale: 32.0 2023-04-27 15:12:59,788 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.963e+01 1.535e+02 1.763e+02 2.248e+02 3.829e+02, threshold=3.527e+02, percent-clipped=1.0 2023-04-27 15:13:16,469 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=3.73 vs. limit=5.0 2023-04-27 15:13:21,230 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=115892.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 15:13:29,449 INFO [finetune.py:976] (1/7) Epoch 21, batch 1350, loss[loss=0.2028, simple_loss=0.2676, pruned_loss=0.06898, over 4837.00 frames. ], tot_loss[loss=0.1696, simple_loss=0.2413, pruned_loss=0.04898, over 956423.42 frames. ], batch size: 33, lr: 3.20e-03, grad_scale: 32.0 2023-04-27 15:14:34,412 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=115953.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 15:14:34,901 INFO [finetune.py:976] (1/7) Epoch 21, batch 1400, loss[loss=0.1977, simple_loss=0.2721, pruned_loss=0.06168, over 4915.00 frames. ], tot_loss[loss=0.1705, simple_loss=0.2427, pruned_loss=0.04921, over 955302.98 frames. ], batch size: 37, lr: 3.20e-03, grad_scale: 32.0 2023-04-27 15:14:44,906 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.084e+02 1.645e+02 1.882e+02 2.243e+02 4.994e+02, threshold=3.764e+02, percent-clipped=1.0 2023-04-27 15:15:26,316 INFO [finetune.py:976] (1/7) Epoch 21, batch 1450, loss[loss=0.1808, simple_loss=0.2623, pruned_loss=0.04963, over 4909.00 frames. ], tot_loss[loss=0.1733, simple_loss=0.2458, pruned_loss=0.05038, over 957190.78 frames. ], batch size: 37, lr: 3.20e-03, grad_scale: 32.0 2023-04-27 15:15:33,447 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=116014.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:15:42,219 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=116026.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:15:42,826 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=116027.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:16:00,013 INFO [finetune.py:976] (1/7) Epoch 21, batch 1500, loss[loss=0.173, simple_loss=0.2466, pruned_loss=0.04974, over 4848.00 frames. ], tot_loss[loss=0.1754, simple_loss=0.2479, pruned_loss=0.05146, over 955048.42 frames. ], batch size: 44, lr: 3.20e-03, grad_scale: 32.0 2023-04-27 15:16:05,187 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.070e+02 1.610e+02 1.919e+02 2.360e+02 3.995e+02, threshold=3.837e+02, percent-clipped=2.0 2023-04-27 15:16:13,680 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=116074.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:16:14,288 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=116075.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:16:14,354 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=116075.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 15:16:33,606 INFO [finetune.py:976] (1/7) Epoch 21, batch 1550, loss[loss=0.1769, simple_loss=0.2526, pruned_loss=0.05059, over 4778.00 frames. ], tot_loss[loss=0.1769, simple_loss=0.2493, pruned_loss=0.05225, over 953728.01 frames. ], batch size: 51, lr: 3.20e-03, grad_scale: 32.0 2023-04-27 15:16:34,325 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=116105.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:16:42,000 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=116116.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:16:53,941 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.24 vs. limit=5.0 2023-04-27 15:17:06,680 INFO [finetune.py:976] (1/7) Epoch 21, batch 1600, loss[loss=0.1553, simple_loss=0.2331, pruned_loss=0.03872, over 4824.00 frames. ], tot_loss[loss=0.1755, simple_loss=0.2472, pruned_loss=0.0519, over 953273.53 frames. ], batch size: 33, lr: 3.20e-03, grad_scale: 32.0 2023-04-27 15:17:10,952 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.072e+02 1.579e+02 1.851e+02 2.317e+02 5.378e+02, threshold=3.702e+02, percent-clipped=3.0 2023-04-27 15:17:14,639 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=116166.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:17:21,304 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=116177.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:17:39,784 INFO [finetune.py:976] (1/7) Epoch 21, batch 1650, loss[loss=0.1345, simple_loss=0.1977, pruned_loss=0.03568, over 4916.00 frames. ], tot_loss[loss=0.1727, simple_loss=0.2439, pruned_loss=0.05074, over 955809.38 frames. ], batch size: 43, lr: 3.20e-03, grad_scale: 32.0 2023-04-27 15:17:46,558 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.1701, 1.5812, 5.2462, 4.9632, 4.4864, 5.0021, 4.5506, 4.6565], device='cuda:1'), covar=tensor([0.6857, 0.6045, 0.1186, 0.1756, 0.1071, 0.1657, 0.1198, 0.1601], device='cuda:1'), in_proj_covar=tensor([0.0310, 0.0307, 0.0405, 0.0405, 0.0347, 0.0408, 0.0312, 0.0367], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-27 15:17:50,662 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-04-27 15:18:04,973 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.64 vs. limit=2.0 2023-04-27 15:18:14,846 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=116248.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 15:18:18,419 INFO [finetune.py:976] (1/7) Epoch 21, batch 1700, loss[loss=0.1332, simple_loss=0.2067, pruned_loss=0.02984, over 4812.00 frames. ], tot_loss[loss=0.1706, simple_loss=0.2415, pruned_loss=0.04985, over 957165.54 frames. ], batch size: 25, lr: 3.20e-03, grad_scale: 32.0 2023-04-27 15:18:28,122 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.038e+02 1.499e+02 1.782e+02 2.142e+02 3.522e+02, threshold=3.563e+02, percent-clipped=0.0 2023-04-27 15:19:10,478 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.1866, 1.3960, 1.6368, 1.7609, 1.6696, 1.7793, 1.7245, 1.7356], device='cuda:1'), covar=tensor([0.3481, 0.4807, 0.4147, 0.4102, 0.5167, 0.6627, 0.4581, 0.4499], device='cuda:1'), in_proj_covar=tensor([0.0338, 0.0376, 0.0321, 0.0336, 0.0346, 0.0394, 0.0357, 0.0328], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 15:19:31,551 INFO [finetune.py:976] (1/7) Epoch 21, batch 1750, loss[loss=0.128, simple_loss=0.2068, pruned_loss=0.02466, over 4766.00 frames. ], tot_loss[loss=0.1723, simple_loss=0.2435, pruned_loss=0.05051, over 957147.65 frames. ], batch size: 28, lr: 3.20e-03, grad_scale: 32.0 2023-04-27 15:19:33,091 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 2023-04-27 15:19:55,828 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=116323.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:20:33,794 INFO [finetune.py:976] (1/7) Epoch 21, batch 1800, loss[loss=0.1784, simple_loss=0.2578, pruned_loss=0.04946, over 4821.00 frames. ], tot_loss[loss=0.175, simple_loss=0.2471, pruned_loss=0.05151, over 955621.02 frames. ], batch size: 38, lr: 3.20e-03, grad_scale: 32.0 2023-04-27 15:20:38,071 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.160e+02 1.681e+02 1.987e+02 2.405e+02 5.932e+02, threshold=3.974e+02, percent-clipped=5.0 2023-04-27 15:20:43,630 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=116370.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 15:20:53,282 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=116384.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:20:54,059 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.70 vs. limit=5.0 2023-04-27 15:20:56,767 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([4.6725, 4.5876, 3.1171, 5.2835, 4.6155, 4.5325, 1.9645, 4.5884], device='cuda:1'), covar=tensor([0.1632, 0.1007, 0.3265, 0.0979, 0.2617, 0.1568, 0.5406, 0.2034], device='cuda:1'), in_proj_covar=tensor([0.0249, 0.0218, 0.0255, 0.0312, 0.0302, 0.0249, 0.0279, 0.0276], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 15:21:07,564 INFO [finetune.py:976] (1/7) Epoch 21, batch 1850, loss[loss=0.2015, simple_loss=0.266, pruned_loss=0.06851, over 4865.00 frames. ], tot_loss[loss=0.177, simple_loss=0.2492, pruned_loss=0.05233, over 955300.44 frames. ], batch size: 34, lr: 3.20e-03, grad_scale: 32.0 2023-04-27 15:21:35,314 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=116446.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:21:40,090 INFO [finetune.py:976] (1/7) Epoch 21, batch 1900, loss[loss=0.1498, simple_loss=0.2315, pruned_loss=0.03401, over 4868.00 frames. ], tot_loss[loss=0.1777, simple_loss=0.2502, pruned_loss=0.05258, over 955400.95 frames. ], batch size: 34, lr: 3.20e-03, grad_scale: 32.0 2023-04-27 15:21:45,220 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.938e+01 1.604e+02 1.932e+02 2.429e+02 3.655e+02, threshold=3.864e+02, percent-clipped=0.0 2023-04-27 15:21:45,317 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=116461.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:21:52,091 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=116472.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:22:13,561 INFO [finetune.py:976] (1/7) Epoch 21, batch 1950, loss[loss=0.1264, simple_loss=0.2096, pruned_loss=0.02157, over 4706.00 frames. ], tot_loss[loss=0.1747, simple_loss=0.2475, pruned_loss=0.05101, over 956921.92 frames. ], batch size: 23, lr: 3.20e-03, grad_scale: 32.0 2023-04-27 15:22:14,293 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=116505.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:22:15,450 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9468, 1.4959, 2.0588, 2.3767, 2.0053, 1.9481, 1.9957, 1.9521], device='cuda:1'), covar=tensor([0.4507, 0.6897, 0.6534, 0.5850, 0.6222, 0.7459, 0.8304, 0.9336], device='cuda:1'), in_proj_covar=tensor([0.0431, 0.0413, 0.0509, 0.0508, 0.0459, 0.0489, 0.0498, 0.0502], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 15:22:16,018 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=116507.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 15:22:24,983 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.3132, 1.5060, 3.8638, 3.6093, 3.3431, 3.7095, 3.7472, 3.3959], device='cuda:1'), covar=tensor([0.7394, 0.5440, 0.1178, 0.1924, 0.1347, 0.1912, 0.1250, 0.1614], device='cuda:1'), in_proj_covar=tensor([0.0309, 0.0306, 0.0402, 0.0403, 0.0347, 0.0407, 0.0310, 0.0366], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-27 15:22:38,199 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7958, 1.4362, 1.9027, 2.3091, 1.9172, 1.7714, 1.8384, 1.8036], device='cuda:1'), covar=tensor([0.4415, 0.6429, 0.6124, 0.5199, 0.5754, 0.7680, 0.7723, 0.8686], device='cuda:1'), in_proj_covar=tensor([0.0431, 0.0412, 0.0509, 0.0508, 0.0459, 0.0488, 0.0498, 0.0502], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 15:22:42,823 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=116548.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 15:22:46,839 INFO [finetune.py:976] (1/7) Epoch 21, batch 2000, loss[loss=0.1865, simple_loss=0.2535, pruned_loss=0.05979, over 4901.00 frames. ], tot_loss[loss=0.1718, simple_loss=0.2442, pruned_loss=0.04975, over 958171.43 frames. ], batch size: 43, lr: 3.20e-03, grad_scale: 32.0 2023-04-27 15:22:50,465 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9501, 1.4529, 1.6102, 1.6303, 2.0672, 1.7455, 1.4292, 1.4980], device='cuda:1'), covar=tensor([0.1694, 0.1653, 0.2073, 0.1467, 0.0832, 0.1645, 0.2209, 0.2341], device='cuda:1'), in_proj_covar=tensor([0.0312, 0.0312, 0.0353, 0.0290, 0.0329, 0.0312, 0.0305, 0.0375], device='cuda:1'), out_proj_covar=tensor([6.3993e-05, 6.4536e-05, 7.4502e-05, 5.8655e-05, 6.8037e-05, 6.5362e-05, 6.3884e-05, 7.9631e-05], device='cuda:1') 2023-04-27 15:22:51,558 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.079e+02 1.496e+02 1.815e+02 2.157e+02 3.594e+02, threshold=3.630e+02, percent-clipped=0.0 2023-04-27 15:22:55,218 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=116566.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:22:55,398 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-04-27 15:23:14,412 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=116596.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 15:23:20,571 INFO [finetune.py:976] (1/7) Epoch 21, batch 2050, loss[loss=0.1849, simple_loss=0.2427, pruned_loss=0.06355, over 4892.00 frames. ], tot_loss[loss=0.1693, simple_loss=0.2407, pruned_loss=0.04897, over 957327.17 frames. ], batch size: 32, lr: 3.20e-03, grad_scale: 32.0 2023-04-27 15:23:30,113 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=116618.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:23:41,567 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([3.3366, 3.2398, 2.4694, 3.7983, 3.2842, 3.2742, 1.4214, 3.3113], device='cuda:1'), covar=tensor([0.1749, 0.1418, 0.3452, 0.2361, 0.3954, 0.1959, 0.5577, 0.2723], device='cuda:1'), in_proj_covar=tensor([0.0247, 0.0217, 0.0255, 0.0309, 0.0301, 0.0248, 0.0277, 0.0275], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 15:23:44,572 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9695, 1.8427, 2.4119, 2.5346, 1.8808, 1.6476, 1.9304, 0.9683], device='cuda:1'), covar=tensor([0.0481, 0.0714, 0.0338, 0.0752, 0.0795, 0.1074, 0.0733, 0.0775], device='cuda:1'), in_proj_covar=tensor([0.0068, 0.0067, 0.0066, 0.0067, 0.0074, 0.0095, 0.0072, 0.0064], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-27 15:23:45,170 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.3090, 1.7140, 1.6440, 2.2264, 2.3895, 1.9956, 1.9463, 1.6455], device='cuda:1'), covar=tensor([0.1725, 0.1698, 0.1857, 0.1506, 0.1127, 0.1809, 0.2120, 0.2286], device='cuda:1'), in_proj_covar=tensor([0.0311, 0.0312, 0.0352, 0.0291, 0.0329, 0.0311, 0.0304, 0.0374], device='cuda:1'), out_proj_covar=tensor([6.3934e-05, 6.4600e-05, 7.4441e-05, 5.8666e-05, 6.7977e-05, 6.5233e-05, 6.3764e-05, 7.9611e-05], device='cuda:1') 2023-04-27 15:23:59,069 INFO [finetune.py:976] (1/7) Epoch 21, batch 2100, loss[loss=0.1627, simple_loss=0.2329, pruned_loss=0.04627, over 4768.00 frames. ], tot_loss[loss=0.1694, simple_loss=0.2407, pruned_loss=0.049, over 957851.21 frames. ], batch size: 29, lr: 3.20e-03, grad_scale: 32.0 2023-04-27 15:24:03,926 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.802e+01 1.616e+02 1.834e+02 2.363e+02 4.673e+02, threshold=3.668e+02, percent-clipped=1.0 2023-04-27 15:24:11,043 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0926, 2.7077, 2.2929, 2.4352, 1.9093, 2.3420, 2.2842, 1.8219], device='cuda:1'), covar=tensor([0.1958, 0.1417, 0.0800, 0.1416, 0.3156, 0.1137, 0.1861, 0.2573], device='cuda:1'), in_proj_covar=tensor([0.0286, 0.0300, 0.0215, 0.0277, 0.0312, 0.0253, 0.0247, 0.0261], device='cuda:1'), out_proj_covar=tensor([1.1432e-04, 1.1913e-04, 8.5216e-05, 1.0963e-04, 1.2627e-04, 1.0034e-04, 9.9649e-05, 1.0370e-04], device='cuda:1') 2023-04-27 15:24:21,185 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=116670.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 15:24:32,887 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=116679.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:24:32,944 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=116679.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:25:00,267 INFO [finetune.py:976] (1/7) Epoch 21, batch 2150, loss[loss=0.1763, simple_loss=0.2507, pruned_loss=0.051, over 4840.00 frames. ], tot_loss[loss=0.1722, simple_loss=0.2437, pruned_loss=0.05035, over 956442.92 frames. ], batch size: 33, lr: 3.20e-03, grad_scale: 32.0 2023-04-27 15:25:00,405 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=116704.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:25:21,294 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=116718.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:25:21,462 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-04-27 15:26:05,620 INFO [finetune.py:976] (1/7) Epoch 21, batch 2200, loss[loss=0.2121, simple_loss=0.2825, pruned_loss=0.07085, over 4862.00 frames. ], tot_loss[loss=0.1737, simple_loss=0.2456, pruned_loss=0.05092, over 956900.00 frames. ], batch size: 44, lr: 3.19e-03, grad_scale: 32.0 2023-04-27 15:26:10,849 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.212e+01 1.600e+02 1.964e+02 2.414e+02 3.602e+02, threshold=3.928e+02, percent-clipped=0.0 2023-04-27 15:26:10,951 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=116761.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:26:13,882 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=116765.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:26:18,738 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=116772.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:26:38,510 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=116802.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 15:26:38,536 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=116802.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:26:39,689 INFO [finetune.py:976] (1/7) Epoch 21, batch 2250, loss[loss=0.201, simple_loss=0.2735, pruned_loss=0.06429, over 4888.00 frames. ], tot_loss[loss=0.1749, simple_loss=0.247, pruned_loss=0.05139, over 957623.58 frames. ], batch size: 43, lr: 3.19e-03, grad_scale: 64.0 2023-04-27 15:26:42,838 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=116809.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:26:44,856 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-04-27 15:26:50,606 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=116820.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:27:07,336 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8852, 1.2791, 4.5807, 4.3192, 3.9805, 4.2699, 4.0438, 4.0130], device='cuda:1'), covar=tensor([0.7107, 0.6039, 0.1033, 0.1735, 0.1149, 0.1538, 0.2446, 0.1494], device='cuda:1'), in_proj_covar=tensor([0.0311, 0.0307, 0.0405, 0.0404, 0.0348, 0.0408, 0.0310, 0.0366], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-27 15:27:13,183 INFO [finetune.py:976] (1/7) Epoch 21, batch 2300, loss[loss=0.179, simple_loss=0.2487, pruned_loss=0.05469, over 4809.00 frames. ], tot_loss[loss=0.1736, simple_loss=0.246, pruned_loss=0.05063, over 958370.39 frames. ], batch size: 41, lr: 3.19e-03, grad_scale: 64.0 2023-04-27 15:27:17,447 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.071e+02 1.557e+02 1.911e+02 2.274e+02 3.749e+02, threshold=3.822e+02, percent-clipped=0.0 2023-04-27 15:27:17,531 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=116861.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:27:18,811 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=116863.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:27:25,868 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-04-27 15:27:46,961 INFO [finetune.py:976] (1/7) Epoch 21, batch 2350, loss[loss=0.1358, simple_loss=0.2087, pruned_loss=0.0315, over 4815.00 frames. ], tot_loss[loss=0.1734, simple_loss=0.2449, pruned_loss=0.05096, over 956500.60 frames. ], batch size: 38, lr: 3.19e-03, grad_scale: 64.0 2023-04-27 15:27:54,459 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9495, 2.5491, 1.9995, 2.0386, 1.4404, 1.4585, 2.0704, 1.3955], device='cuda:1'), covar=tensor([0.1727, 0.1405, 0.1415, 0.1539, 0.2385, 0.1918, 0.0980, 0.2085], device='cuda:1'), in_proj_covar=tensor([0.0197, 0.0212, 0.0170, 0.0204, 0.0201, 0.0186, 0.0157, 0.0187], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-27 15:28:20,897 INFO [finetune.py:976] (1/7) Epoch 21, batch 2400, loss[loss=0.1768, simple_loss=0.2435, pruned_loss=0.05504, over 4764.00 frames. ], tot_loss[loss=0.1708, simple_loss=0.2417, pruned_loss=0.05001, over 955934.25 frames. ], batch size: 54, lr: 3.19e-03, grad_scale: 64.0 2023-04-27 15:28:25,119 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.043e+02 1.594e+02 1.806e+02 2.173e+02 4.519e+02, threshold=3.612e+02, percent-clipped=1.0 2023-04-27 15:28:34,745 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=116974.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:28:37,807 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=116979.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:28:54,617 INFO [finetune.py:976] (1/7) Epoch 21, batch 2450, loss[loss=0.1808, simple_loss=0.2489, pruned_loss=0.05637, over 4805.00 frames. ], tot_loss[loss=0.1696, simple_loss=0.2399, pruned_loss=0.04965, over 957319.24 frames. ], batch size: 45, lr: 3.19e-03, grad_scale: 64.0 2023-04-27 15:29:06,873 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5978, 3.7828, 0.7312, 1.8385, 1.8931, 2.4785, 2.1729, 0.9686], device='cuda:1'), covar=tensor([0.1773, 0.1921, 0.2598, 0.1846, 0.1396, 0.1495, 0.1830, 0.2398], device='cuda:1'), in_proj_covar=tensor([0.0117, 0.0239, 0.0136, 0.0119, 0.0132, 0.0152, 0.0116, 0.0118], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-27 15:29:10,688 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=117027.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:29:28,088 INFO [finetune.py:976] (1/7) Epoch 21, batch 2500, loss[loss=0.207, simple_loss=0.2675, pruned_loss=0.0732, over 4395.00 frames. ], tot_loss[loss=0.1714, simple_loss=0.2421, pruned_loss=0.05035, over 955454.87 frames. ], batch size: 19, lr: 3.19e-03, grad_scale: 64.0 2023-04-27 15:29:32,764 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=117060.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:29:33,286 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.097e+02 1.592e+02 1.814e+02 2.143e+02 3.626e+02, threshold=3.628e+02, percent-clipped=1.0 2023-04-27 15:29:55,003 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([3.0210, 2.4215, 2.2078, 2.3211, 2.1032, 2.2958, 2.2965, 2.2280], device='cuda:1'), covar=tensor([0.4059, 0.5760, 0.5444, 0.5330, 0.6106, 0.7481, 0.6158, 0.5515], device='cuda:1'), in_proj_covar=tensor([0.0339, 0.0376, 0.0324, 0.0337, 0.0348, 0.0396, 0.0359, 0.0330], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 15:30:29,013 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=117102.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:30:30,127 INFO [finetune.py:976] (1/7) Epoch 21, batch 2550, loss[loss=0.1444, simple_loss=0.223, pruned_loss=0.03291, over 4307.00 frames. ], tot_loss[loss=0.1755, simple_loss=0.2467, pruned_loss=0.05217, over 955018.61 frames. ], batch size: 19, lr: 3.19e-03, grad_scale: 64.0 2023-04-27 15:31:10,425 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7273, 2.2714, 1.7787, 1.5996, 1.2824, 1.2905, 1.8374, 1.2711], device='cuda:1'), covar=tensor([0.1625, 0.1263, 0.1442, 0.1757, 0.2252, 0.1950, 0.0961, 0.2041], device='cuda:1'), in_proj_covar=tensor([0.0199, 0.0213, 0.0171, 0.0206, 0.0202, 0.0187, 0.0158, 0.0188], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-27 15:31:33,696 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=117150.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:31:41,891 INFO [finetune.py:976] (1/7) Epoch 21, batch 2600, loss[loss=0.2097, simple_loss=0.2959, pruned_loss=0.06173, over 4861.00 frames. ], tot_loss[loss=0.1765, simple_loss=0.2478, pruned_loss=0.05258, over 952328.15 frames. ], batch size: 44, lr: 3.19e-03, grad_scale: 64.0 2023-04-27 15:31:44,421 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=117158.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:31:52,560 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.120e+02 1.651e+02 1.953e+02 2.341e+02 4.282e+02, threshold=3.906e+02, percent-clipped=1.0 2023-04-27 15:31:52,661 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=117161.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:32:23,914 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5044, 1.3269, 1.8074, 1.7377, 1.3699, 1.2405, 1.3821, 0.8961], device='cuda:1'), covar=tensor([0.0558, 0.0643, 0.0358, 0.0576, 0.0826, 0.1263, 0.0648, 0.0629], device='cuda:1'), in_proj_covar=tensor([0.0068, 0.0067, 0.0066, 0.0067, 0.0074, 0.0095, 0.0073, 0.0065], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-27 15:32:27,548 INFO [finetune.py:976] (1/7) Epoch 21, batch 2650, loss[loss=0.1697, simple_loss=0.2542, pruned_loss=0.04264, over 4817.00 frames. ], tot_loss[loss=0.1764, simple_loss=0.2483, pruned_loss=0.05226, over 953848.68 frames. ], batch size: 38, lr: 3.19e-03, grad_scale: 64.0 2023-04-27 15:32:30,679 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=117209.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:32:39,135 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=117221.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:33:01,272 INFO [finetune.py:976] (1/7) Epoch 21, batch 2700, loss[loss=0.1938, simple_loss=0.2541, pruned_loss=0.06674, over 4864.00 frames. ], tot_loss[loss=0.1754, simple_loss=0.2474, pruned_loss=0.05163, over 956311.59 frames. ], batch size: 31, lr: 3.19e-03, grad_scale: 64.0 2023-04-27 15:33:06,003 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 8.896e+01 1.368e+02 1.680e+02 2.087e+02 5.894e+02, threshold=3.359e+02, percent-clipped=1.0 2023-04-27 15:33:15,076 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=117274.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:33:18,146 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0561, 2.5030, 2.1358, 2.3503, 1.8305, 2.1457, 2.1264, 1.6321], device='cuda:1'), covar=tensor([0.1807, 0.1075, 0.0696, 0.1190, 0.2940, 0.1033, 0.1872, 0.2564], device='cuda:1'), in_proj_covar=tensor([0.0286, 0.0299, 0.0215, 0.0277, 0.0310, 0.0252, 0.0246, 0.0261], device='cuda:1'), out_proj_covar=tensor([1.1451e-04, 1.1842e-04, 8.5410e-05, 1.0956e-04, 1.2568e-04, 1.0007e-04, 9.9289e-05, 1.0371e-04], device='cuda:1') 2023-04-27 15:33:19,980 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=117282.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:33:35,172 INFO [finetune.py:976] (1/7) Epoch 21, batch 2750, loss[loss=0.1369, simple_loss=0.2208, pruned_loss=0.02654, over 4760.00 frames. ], tot_loss[loss=0.1733, simple_loss=0.2449, pruned_loss=0.05085, over 958015.76 frames. ], batch size: 27, lr: 3.19e-03, grad_scale: 64.0 2023-04-27 15:33:47,712 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=117322.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:33:52,062 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=117329.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:34:08,696 INFO [finetune.py:976] (1/7) Epoch 21, batch 2800, loss[loss=0.1758, simple_loss=0.2368, pruned_loss=0.05745, over 4749.00 frames. ], tot_loss[loss=0.1702, simple_loss=0.241, pruned_loss=0.04968, over 956057.49 frames. ], batch size: 54, lr: 3.19e-03, grad_scale: 64.0 2023-04-27 15:34:09,395 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([0.7341, 1.7554, 1.6866, 1.4481, 1.7663, 1.4599, 2.2210, 1.4561], device='cuda:1'), covar=tensor([0.3712, 0.1924, 0.4818, 0.2562, 0.1727, 0.2290, 0.1542, 0.4608], device='cuda:1'), in_proj_covar=tensor([0.0340, 0.0346, 0.0422, 0.0350, 0.0380, 0.0373, 0.0368, 0.0417], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 15:34:12,916 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=117360.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:34:13,414 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.017e+02 1.550e+02 1.821e+02 2.376e+02 5.325e+02, threshold=3.642e+02, percent-clipped=3.0 2023-04-27 15:34:31,558 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=117390.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:34:41,632 INFO [finetune.py:976] (1/7) Epoch 21, batch 2850, loss[loss=0.1328, simple_loss=0.2034, pruned_loss=0.03109, over 4774.00 frames. ], tot_loss[loss=0.169, simple_loss=0.2394, pruned_loss=0.04934, over 955636.95 frames. ], batch size: 27, lr: 3.19e-03, grad_scale: 64.0 2023-04-27 15:34:44,561 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=117408.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:35:11,268 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.2286, 2.7769, 2.1355, 2.6156, 1.8064, 2.2763, 2.5818, 1.9230], device='cuda:1'), covar=tensor([0.1974, 0.1138, 0.0928, 0.1153, 0.3778, 0.1196, 0.1777, 0.2781], device='cuda:1'), in_proj_covar=tensor([0.0287, 0.0300, 0.0217, 0.0278, 0.0311, 0.0254, 0.0247, 0.0262], device='cuda:1'), out_proj_covar=tensor([1.1491e-04, 1.1883e-04, 8.5912e-05, 1.0997e-04, 1.2609e-04, 1.0053e-04, 9.9845e-05, 1.0417e-04], device='cuda:1') 2023-04-27 15:35:14,767 INFO [finetune.py:976] (1/7) Epoch 21, batch 2900, loss[loss=0.1825, simple_loss=0.2535, pruned_loss=0.05579, over 4824.00 frames. ], tot_loss[loss=0.1717, simple_loss=0.2424, pruned_loss=0.05051, over 956828.97 frames. ], batch size: 30, lr: 3.19e-03, grad_scale: 64.0 2023-04-27 15:35:17,801 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=117458.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:35:19,564 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.085e+02 1.585e+02 1.858e+02 2.394e+02 5.975e+02, threshold=3.717e+02, percent-clipped=5.0 2023-04-27 15:36:16,931 INFO [finetune.py:976] (1/7) Epoch 21, batch 2950, loss[loss=0.2379, simple_loss=0.3015, pruned_loss=0.08713, over 4226.00 frames. ], tot_loss[loss=0.1739, simple_loss=0.2455, pruned_loss=0.05117, over 957206.57 frames. ], batch size: 65, lr: 3.19e-03, grad_scale: 64.0 2023-04-27 15:36:18,222 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=117506.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:37:09,249 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9446, 2.6164, 2.0878, 1.9653, 1.3748, 1.4278, 2.1895, 1.4012], device='cuda:1'), covar=tensor([0.1868, 0.1633, 0.1551, 0.1895, 0.2468, 0.2121, 0.1034, 0.2164], device='cuda:1'), in_proj_covar=tensor([0.0197, 0.0211, 0.0169, 0.0203, 0.0199, 0.0185, 0.0156, 0.0185], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-27 15:37:23,420 INFO [finetune.py:976] (1/7) Epoch 21, batch 3000, loss[loss=0.1599, simple_loss=0.2413, pruned_loss=0.03924, over 4839.00 frames. ], tot_loss[loss=0.1765, simple_loss=0.2486, pruned_loss=0.05221, over 958487.65 frames. ], batch size: 44, lr: 3.19e-03, grad_scale: 64.0 2023-04-27 15:37:23,420 INFO [finetune.py:1001] (1/7) Computing validation loss 2023-04-27 15:37:30,175 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5009, 3.0406, 0.9827, 1.7197, 1.9199, 2.2392, 1.7628, 0.9651], device='cuda:1'), covar=tensor([0.1350, 0.1045, 0.1937, 0.1222, 0.1083, 0.0965, 0.1649, 0.1820], device='cuda:1'), in_proj_covar=tensor([0.0116, 0.0238, 0.0135, 0.0119, 0.0132, 0.0151, 0.0115, 0.0118], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-27 15:37:43,669 INFO [finetune.py:1010] (1/7) Epoch 21, validation: loss=0.1531, simple_loss=0.2228, pruned_loss=0.04164, over 2265189.00 frames. 2023-04-27 15:37:43,670 INFO [finetune.py:1011] (1/7) Maximum memory allocated so far is 6391MB 2023-04-27 15:37:54,334 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.170e+02 1.590e+02 1.926e+02 2.493e+02 6.945e+02, threshold=3.852e+02, percent-clipped=2.0 2023-04-27 15:38:15,507 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=117577.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:38:49,964 INFO [finetune.py:976] (1/7) Epoch 21, batch 3050, loss[loss=0.147, simple_loss=0.2351, pruned_loss=0.02941, over 4904.00 frames. ], tot_loss[loss=0.1761, simple_loss=0.2492, pruned_loss=0.0515, over 958733.27 frames. ], batch size: 37, lr: 3.19e-03, grad_scale: 64.0 2023-04-27 15:39:07,049 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9344, 1.1971, 1.6253, 1.7553, 1.7248, 1.7820, 1.6508, 1.6756], device='cuda:1'), covar=tensor([0.3912, 0.4945, 0.4293, 0.4327, 0.5296, 0.7188, 0.4446, 0.4486], device='cuda:1'), in_proj_covar=tensor([0.0339, 0.0375, 0.0324, 0.0337, 0.0348, 0.0395, 0.0357, 0.0329], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 15:39:15,603 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-04-27 15:39:20,762 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=3.72 vs. limit=5.0 2023-04-27 15:39:28,406 INFO [finetune.py:976] (1/7) Epoch 21, batch 3100, loss[loss=0.1237, simple_loss=0.1979, pruned_loss=0.02474, over 4826.00 frames. ], tot_loss[loss=0.1738, simple_loss=0.2464, pruned_loss=0.05058, over 958034.29 frames. ], batch size: 39, lr: 3.19e-03, grad_scale: 64.0 2023-04-27 15:39:33,616 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.782e+01 1.629e+02 1.831e+02 2.140e+02 4.594e+02, threshold=3.661e+02, percent-clipped=1.0 2023-04-27 15:39:39,797 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.2354, 1.5976, 1.6298, 1.8889, 1.8430, 1.9665, 1.4697, 3.7800], device='cuda:1'), covar=tensor([0.0650, 0.0870, 0.0836, 0.1217, 0.0655, 0.0527, 0.0808, 0.0188], device='cuda:1'), in_proj_covar=tensor([0.0037, 0.0038, 0.0039, 0.0043, 0.0040, 0.0038, 0.0038, 0.0056], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:1') 2023-04-27 15:39:49,470 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=117685.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:39:51,928 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.56 vs. limit=5.0 2023-04-27 15:40:01,887 INFO [finetune.py:976] (1/7) Epoch 21, batch 3150, loss[loss=0.2003, simple_loss=0.2714, pruned_loss=0.06461, over 4917.00 frames. ], tot_loss[loss=0.1719, simple_loss=0.2435, pruned_loss=0.05014, over 956546.18 frames. ], batch size: 46, lr: 3.19e-03, grad_scale: 64.0 2023-04-27 15:40:08,469 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9927, 1.7757, 2.2198, 2.4623, 2.0683, 1.8452, 2.0428, 2.0290], device='cuda:1'), covar=tensor([0.5324, 0.7573, 0.7063, 0.6342, 0.6514, 0.9820, 0.9671, 1.0392], device='cuda:1'), in_proj_covar=tensor([0.0429, 0.0412, 0.0505, 0.0506, 0.0458, 0.0487, 0.0494, 0.0501], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 15:40:34,883 INFO [finetune.py:976] (1/7) Epoch 21, batch 3200, loss[loss=0.1718, simple_loss=0.2367, pruned_loss=0.05347, over 4878.00 frames. ], tot_loss[loss=0.1694, simple_loss=0.2406, pruned_loss=0.04912, over 956487.07 frames. ], batch size: 31, lr: 3.19e-03, grad_scale: 32.0 2023-04-27 15:40:40,630 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.053e+02 1.488e+02 1.740e+02 2.134e+02 4.816e+02, threshold=3.479e+02, percent-clipped=1.0 2023-04-27 15:41:31,202 INFO [finetune.py:976] (1/7) Epoch 21, batch 3250, loss[loss=0.2162, simple_loss=0.2704, pruned_loss=0.08103, over 4919.00 frames. ], tot_loss[loss=0.1691, simple_loss=0.2401, pruned_loss=0.0491, over 956390.33 frames. ], batch size: 36, lr: 3.19e-03, grad_scale: 32.0 2023-04-27 15:42:33,313 INFO [finetune.py:976] (1/7) Epoch 21, batch 3300, loss[loss=0.2008, simple_loss=0.253, pruned_loss=0.07429, over 3827.00 frames. ], tot_loss[loss=0.1718, simple_loss=0.2434, pruned_loss=0.05012, over 955742.91 frames. ], batch size: 16, lr: 3.19e-03, grad_scale: 32.0 2023-04-27 15:42:45,019 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.168e+02 1.642e+02 1.977e+02 2.287e+02 4.163e+02, threshold=3.954e+02, percent-clipped=4.0 2023-04-27 15:43:01,367 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=117877.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:43:01,450 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=3.76 vs. limit=5.0 2023-04-27 15:43:06,697 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.2810, 1.4814, 1.7188, 1.7752, 1.7016, 1.8114, 1.8234, 1.8496], device='cuda:1'), covar=tensor([0.4228, 0.5251, 0.4652, 0.4655, 0.5441, 0.7204, 0.4808, 0.4418], device='cuda:1'), in_proj_covar=tensor([0.0337, 0.0373, 0.0322, 0.0335, 0.0345, 0.0392, 0.0355, 0.0327], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 15:43:18,017 INFO [finetune.py:976] (1/7) Epoch 21, batch 3350, loss[loss=0.1431, simple_loss=0.2188, pruned_loss=0.03365, over 4176.00 frames. ], tot_loss[loss=0.1741, simple_loss=0.2463, pruned_loss=0.05093, over 955613.51 frames. ], batch size: 18, lr: 3.19e-03, grad_scale: 32.0 2023-04-27 15:43:37,870 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=117925.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:44:01,216 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.4205, 1.8867, 2.2752, 2.7168, 2.2884, 1.7875, 1.5440, 2.1511], device='cuda:1'), covar=tensor([0.3360, 0.2995, 0.1719, 0.2557, 0.2764, 0.2553, 0.3888, 0.1834], device='cuda:1'), in_proj_covar=tensor([0.0292, 0.0243, 0.0226, 0.0314, 0.0218, 0.0231, 0.0227, 0.0183], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-27 15:44:02,872 INFO [finetune.py:976] (1/7) Epoch 21, batch 3400, loss[loss=0.1807, simple_loss=0.26, pruned_loss=0.05075, over 4899.00 frames. ], tot_loss[loss=0.1744, simple_loss=0.2472, pruned_loss=0.05078, over 955220.41 frames. ], batch size: 37, lr: 3.19e-03, grad_scale: 32.0 2023-04-27 15:44:13,575 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.012e+02 1.581e+02 1.978e+02 2.353e+02 4.308e+02, threshold=3.955e+02, percent-clipped=3.0 2023-04-27 15:44:44,678 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=117985.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:44:57,990 INFO [finetune.py:976] (1/7) Epoch 21, batch 3450, loss[loss=0.2165, simple_loss=0.2704, pruned_loss=0.0813, over 4922.00 frames. ], tot_loss[loss=0.1745, simple_loss=0.2476, pruned_loss=0.05072, over 956558.67 frames. ], batch size: 42, lr: 3.18e-03, grad_scale: 32.0 2023-04-27 15:45:17,646 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=118033.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:45:22,456 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=118040.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:45:31,328 INFO [finetune.py:976] (1/7) Epoch 21, batch 3500, loss[loss=0.1636, simple_loss=0.2331, pruned_loss=0.04705, over 4931.00 frames. ], tot_loss[loss=0.1729, simple_loss=0.2453, pruned_loss=0.05025, over 956432.54 frames. ], batch size: 33, lr: 3.18e-03, grad_scale: 32.0 2023-04-27 15:45:33,872 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4607, 1.4490, 1.7902, 1.7664, 1.3824, 1.2493, 1.5096, 0.9367], device='cuda:1'), covar=tensor([0.0610, 0.0571, 0.0419, 0.0649, 0.0798, 0.1079, 0.0624, 0.0591], device='cuda:1'), in_proj_covar=tensor([0.0069, 0.0068, 0.0067, 0.0068, 0.0075, 0.0096, 0.0073, 0.0065], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-27 15:45:36,183 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.106e+01 1.494e+02 1.798e+02 2.113e+02 5.768e+02, threshold=3.596e+02, percent-clipped=1.0 2023-04-27 15:46:03,430 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=118101.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:46:05,140 INFO [finetune.py:976] (1/7) Epoch 21, batch 3550, loss[loss=0.1314, simple_loss=0.2053, pruned_loss=0.02874, over 4915.00 frames. ], tot_loss[loss=0.1728, simple_loss=0.2442, pruned_loss=0.05075, over 957254.71 frames. ], batch size: 43, lr: 3.18e-03, grad_scale: 32.0 2023-04-27 15:46:20,836 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5855, 1.7210, 1.4101, 1.0419, 1.1853, 1.1485, 1.3797, 1.1541], device='cuda:1'), covar=tensor([0.1794, 0.1316, 0.1621, 0.1860, 0.2288, 0.2122, 0.1050, 0.2070], device='cuda:1'), in_proj_covar=tensor([0.0198, 0.0212, 0.0170, 0.0205, 0.0199, 0.0185, 0.0157, 0.0186], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-27 15:46:33,160 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.3534, 1.6161, 1.5625, 1.8552, 1.7995, 1.7928, 1.4627, 3.4744], device='cuda:1'), covar=tensor([0.0552, 0.0746, 0.0729, 0.1098, 0.0581, 0.0439, 0.0693, 0.0123], device='cuda:1'), in_proj_covar=tensor([0.0038, 0.0039, 0.0040, 0.0044, 0.0040, 0.0038, 0.0038, 0.0056], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0013, 0.0013, 0.0015, 0.0014, 0.0013, 0.0013, 0.0016], device='cuda:1') 2023-04-27 15:46:39,040 INFO [finetune.py:976] (1/7) Epoch 21, batch 3600, loss[loss=0.2228, simple_loss=0.2835, pruned_loss=0.08109, over 4730.00 frames. ], tot_loss[loss=0.1709, simple_loss=0.2414, pruned_loss=0.05021, over 955403.51 frames. ], batch size: 59, lr: 3.18e-03, grad_scale: 32.0 2023-04-27 15:46:40,503 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 2023-04-27 15:46:43,924 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.120e+02 1.483e+02 1.865e+02 2.338e+02 3.870e+02, threshold=3.730e+02, percent-clipped=2.0 2023-04-27 15:46:47,627 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=118168.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:47:39,451 INFO [finetune.py:976] (1/7) Epoch 21, batch 3650, loss[loss=0.2048, simple_loss=0.2743, pruned_loss=0.06761, over 4759.00 frames. ], tot_loss[loss=0.174, simple_loss=0.2447, pruned_loss=0.05168, over 955713.09 frames. ], batch size: 59, lr: 3.18e-03, grad_scale: 32.0 2023-04-27 15:48:08,209 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9319, 2.3119, 2.0120, 2.2384, 1.5737, 2.0637, 2.1159, 1.5009], device='cuda:1'), covar=tensor([0.1855, 0.1031, 0.0826, 0.1145, 0.3459, 0.1146, 0.1669, 0.2532], device='cuda:1'), in_proj_covar=tensor([0.0282, 0.0297, 0.0213, 0.0276, 0.0308, 0.0250, 0.0243, 0.0261], device='cuda:1'), out_proj_covar=tensor([1.1299e-04, 1.1771e-04, 8.4472e-05, 1.0930e-04, 1.2488e-04, 9.9196e-05, 9.8253e-05, 1.0350e-04], device='cuda:1') 2023-04-27 15:48:09,469 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=118229.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 15:48:43,767 INFO [finetune.py:976] (1/7) Epoch 21, batch 3700, loss[loss=0.1926, simple_loss=0.2679, pruned_loss=0.05868, over 4865.00 frames. ], tot_loss[loss=0.1762, simple_loss=0.2479, pruned_loss=0.05227, over 955014.53 frames. ], batch size: 31, lr: 3.18e-03, grad_scale: 32.0 2023-04-27 15:48:54,002 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.129e+02 1.556e+02 1.839e+02 2.205e+02 3.757e+02, threshold=3.678e+02, percent-clipped=1.0 2023-04-27 15:48:58,371 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=118269.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:49:16,474 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0944, 4.2167, 0.8363, 2.3616, 2.6305, 2.7911, 2.5829, 0.9912], device='cuda:1'), covar=tensor([0.1173, 0.0821, 0.1877, 0.1118, 0.0827, 0.1033, 0.1260, 0.2011], device='cuda:1'), in_proj_covar=tensor([0.0117, 0.0239, 0.0136, 0.0119, 0.0132, 0.0152, 0.0116, 0.0118], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-27 15:49:16,511 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=118289.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:49:27,410 INFO [finetune.py:976] (1/7) Epoch 21, batch 3750, loss[loss=0.2199, simple_loss=0.2818, pruned_loss=0.07896, over 4855.00 frames. ], tot_loss[loss=0.1769, simple_loss=0.2489, pruned_loss=0.05247, over 954716.89 frames. ], batch size: 31, lr: 3.18e-03, grad_scale: 32.0 2023-04-27 15:49:40,380 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-04-27 15:49:43,210 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=118329.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:49:43,834 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=118330.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:50:10,267 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=118350.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:50:12,621 INFO [finetune.py:976] (1/7) Epoch 21, batch 3800, loss[loss=0.1414, simple_loss=0.2087, pruned_loss=0.03701, over 4721.00 frames. ], tot_loss[loss=0.1782, simple_loss=0.2503, pruned_loss=0.05306, over 952442.13 frames. ], batch size: 23, lr: 3.18e-03, grad_scale: 32.0 2023-04-27 15:50:23,124 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.061e+02 1.645e+02 2.008e+02 2.325e+02 4.479e+02, threshold=4.015e+02, percent-clipped=4.0 2023-04-27 15:50:52,599 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=118390.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:50:56,189 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=118396.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:51:02,180 INFO [finetune.py:976] (1/7) Epoch 21, batch 3850, loss[loss=0.2158, simple_loss=0.2699, pruned_loss=0.08089, over 4782.00 frames. ], tot_loss[loss=0.1758, simple_loss=0.2482, pruned_loss=0.05175, over 952623.08 frames. ], batch size: 29, lr: 3.18e-03, grad_scale: 32.0 2023-04-27 15:51:35,148 INFO [finetune.py:976] (1/7) Epoch 21, batch 3900, loss[loss=0.1476, simple_loss=0.2134, pruned_loss=0.04091, over 4914.00 frames. ], tot_loss[loss=0.1732, simple_loss=0.2452, pruned_loss=0.05065, over 952045.05 frames. ], batch size: 36, lr: 3.18e-03, grad_scale: 32.0 2023-04-27 15:51:40,430 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.017e+02 1.638e+02 1.891e+02 2.265e+02 7.777e+02, threshold=3.782e+02, percent-clipped=1.0 2023-04-27 15:51:49,047 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8466, 2.5482, 1.7581, 2.0643, 1.3701, 1.4075, 1.8190, 1.3479], device='cuda:1'), covar=tensor([0.1716, 0.1180, 0.1530, 0.1518, 0.2375, 0.2087, 0.1082, 0.2005], device='cuda:1'), in_proj_covar=tensor([0.0198, 0.0212, 0.0169, 0.0204, 0.0199, 0.0185, 0.0156, 0.0185], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-27 15:51:53,969 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9340, 1.9307, 1.9859, 2.2953, 2.2649, 2.3885, 1.8554, 4.7986], device='cuda:1'), covar=tensor([0.0486, 0.0730, 0.0732, 0.1097, 0.0570, 0.0435, 0.0697, 0.0122], device='cuda:1'), in_proj_covar=tensor([0.0037, 0.0038, 0.0039, 0.0043, 0.0039, 0.0038, 0.0037, 0.0055], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:1') 2023-04-27 15:52:07,427 INFO [finetune.py:976] (1/7) Epoch 21, batch 3950, loss[loss=0.1761, simple_loss=0.2435, pruned_loss=0.05439, over 4903.00 frames. ], tot_loss[loss=0.1701, simple_loss=0.2417, pruned_loss=0.04926, over 953532.83 frames. ], batch size: 32, lr: 3.18e-03, grad_scale: 32.0 2023-04-27 15:52:21,113 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=118524.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 15:52:24,860 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.7935, 2.0809, 1.8839, 2.5504, 2.7709, 2.3123, 2.2268, 1.9810], device='cuda:1'), covar=tensor([0.1591, 0.1591, 0.1842, 0.1726, 0.0952, 0.1640, 0.2171, 0.2176], device='cuda:1'), in_proj_covar=tensor([0.0309, 0.0309, 0.0349, 0.0287, 0.0324, 0.0306, 0.0300, 0.0368], device='cuda:1'), out_proj_covar=tensor([6.3332e-05, 6.3965e-05, 7.3671e-05, 5.7948e-05, 6.6930e-05, 6.4232e-05, 6.2816e-05, 7.8108e-05], device='cuda:1') 2023-04-27 15:52:35,281 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-04-27 15:52:40,872 INFO [finetune.py:976] (1/7) Epoch 21, batch 4000, loss[loss=0.1585, simple_loss=0.2212, pruned_loss=0.04793, over 4230.00 frames. ], tot_loss[loss=0.1692, simple_loss=0.2402, pruned_loss=0.04906, over 951684.26 frames. ], batch size: 18, lr: 3.18e-03, grad_scale: 32.0 2023-04-27 15:52:47,270 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.678e+01 1.545e+02 1.901e+02 2.367e+02 3.489e+02, threshold=3.803e+02, percent-clipped=0.0 2023-04-27 15:53:11,628 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([3.3695, 2.9251, 2.5468, 2.7503, 2.5487, 2.7963, 2.7146, 2.6892], device='cuda:1'), covar=tensor([0.3392, 0.4835, 0.4667, 0.4304, 0.4919, 0.5902, 0.5124, 0.4915], device='cuda:1'), in_proj_covar=tensor([0.0339, 0.0374, 0.0324, 0.0338, 0.0349, 0.0396, 0.0358, 0.0330], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 15:53:30,868 INFO [finetune.py:976] (1/7) Epoch 21, batch 4050, loss[loss=0.1622, simple_loss=0.2456, pruned_loss=0.03946, over 4765.00 frames. ], tot_loss[loss=0.1724, simple_loss=0.2437, pruned_loss=0.05055, over 949831.21 frames. ], batch size: 28, lr: 3.18e-03, grad_scale: 32.0 2023-04-27 15:53:56,771 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=118625.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:53:56,927 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.49 vs. limit=2.0 2023-04-27 15:54:06,606 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=118631.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:54:26,640 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=118645.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:54:33,133 INFO [finetune.py:976] (1/7) Epoch 21, batch 4100, loss[loss=0.1637, simple_loss=0.2295, pruned_loss=0.04892, over 4773.00 frames. ], tot_loss[loss=0.1734, simple_loss=0.2452, pruned_loss=0.05081, over 950413.05 frames. ], batch size: 26, lr: 3.18e-03, grad_scale: 32.0 2023-04-27 15:54:38,508 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.195e+02 1.657e+02 1.981e+02 2.296e+02 3.754e+02, threshold=3.963e+02, percent-clipped=0.0 2023-04-27 15:54:54,358 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=118685.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:54:58,636 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=118692.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:55:00,961 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=118696.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:55:06,701 INFO [finetune.py:976] (1/7) Epoch 21, batch 4150, loss[loss=0.1755, simple_loss=0.2479, pruned_loss=0.05158, over 4910.00 frames. ], tot_loss[loss=0.1761, simple_loss=0.248, pruned_loss=0.05215, over 951731.16 frames. ], batch size: 37, lr: 3.18e-03, grad_scale: 32.0 2023-04-27 15:55:56,017 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=118744.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:56:07,501 INFO [finetune.py:976] (1/7) Epoch 21, batch 4200, loss[loss=0.1681, simple_loss=0.2419, pruned_loss=0.04712, over 4817.00 frames. ], tot_loss[loss=0.1745, simple_loss=0.2471, pruned_loss=0.05094, over 953916.90 frames. ], batch size: 39, lr: 3.18e-03, grad_scale: 32.0 2023-04-27 15:56:19,457 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.047e+02 1.568e+02 1.859e+02 2.228e+02 3.642e+02, threshold=3.719e+02, percent-clipped=0.0 2023-04-27 15:56:58,125 INFO [finetune.py:976] (1/7) Epoch 21, batch 4250, loss[loss=0.1764, simple_loss=0.2498, pruned_loss=0.05149, over 4908.00 frames. ], tot_loss[loss=0.1726, simple_loss=0.2447, pruned_loss=0.05021, over 951579.25 frames. ], batch size: 43, lr: 3.18e-03, grad_scale: 32.0 2023-04-27 15:57:13,411 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=118824.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:57:32,153 INFO [finetune.py:976] (1/7) Epoch 21, batch 4300, loss[loss=0.1308, simple_loss=0.2036, pruned_loss=0.02898, over 4807.00 frames. ], tot_loss[loss=0.1707, simple_loss=0.2425, pruned_loss=0.0494, over 953060.97 frames. ], batch size: 51, lr: 3.18e-03, grad_scale: 32.0 2023-04-27 15:57:37,516 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.744e+01 1.518e+02 1.726e+02 2.140e+02 3.725e+02, threshold=3.451e+02, percent-clipped=1.0 2023-04-27 15:57:44,624 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=118872.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:57:51,478 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([4.8927, 4.9316, 3.3314, 5.5622, 4.9740, 4.9257, 2.4830, 4.8957], device='cuda:1'), covar=tensor([0.1681, 0.0921, 0.2738, 0.0967, 0.2110, 0.1564, 0.5014, 0.1819], device='cuda:1'), in_proj_covar=tensor([0.0246, 0.0216, 0.0253, 0.0307, 0.0297, 0.0247, 0.0276, 0.0272], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 15:57:56,967 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.2059, 1.3221, 3.8232, 3.5518, 3.3418, 3.6888, 3.7161, 3.3583], device='cuda:1'), covar=tensor([0.7167, 0.5575, 0.1180, 0.1803, 0.1286, 0.1597, 0.1213, 0.1443], device='cuda:1'), in_proj_covar=tensor([0.0310, 0.0306, 0.0405, 0.0404, 0.0347, 0.0408, 0.0310, 0.0366], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-27 15:58:06,089 INFO [finetune.py:976] (1/7) Epoch 21, batch 4350, loss[loss=0.1539, simple_loss=0.2265, pruned_loss=0.04066, over 4909.00 frames. ], tot_loss[loss=0.1683, simple_loss=0.2393, pruned_loss=0.04867, over 953955.57 frames. ], batch size: 35, lr: 3.18e-03, grad_scale: 32.0 2023-04-27 15:58:20,457 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=118925.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:58:39,672 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=118945.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:58:49,724 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=3.76 vs. limit=5.0 2023-04-27 15:58:51,404 INFO [finetune.py:976] (1/7) Epoch 21, batch 4400, loss[loss=0.1256, simple_loss=0.2024, pruned_loss=0.02444, over 4770.00 frames. ], tot_loss[loss=0.1704, simple_loss=0.2408, pruned_loss=0.04999, over 953832.06 frames. ], batch size: 26, lr: 3.18e-03, grad_scale: 32.0 2023-04-27 15:59:00,771 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.017e+02 1.550e+02 1.866e+02 2.301e+02 5.364e+02, threshold=3.732e+02, percent-clipped=5.0 2023-04-27 15:59:14,265 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=118973.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:59:35,148 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=118985.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:59:36,315 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=118987.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:59:45,377 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=118993.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:59:52,143 INFO [finetune.py:976] (1/7) Epoch 21, batch 4450, loss[loss=0.1789, simple_loss=0.2592, pruned_loss=0.04932, over 4867.00 frames. ], tot_loss[loss=0.173, simple_loss=0.2442, pruned_loss=0.05093, over 952726.36 frames. ], batch size: 34, lr: 3.18e-03, grad_scale: 32.0 2023-04-27 16:00:11,984 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=119033.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:00:25,718 INFO [finetune.py:976] (1/7) Epoch 21, batch 4500, loss[loss=0.1869, simple_loss=0.263, pruned_loss=0.05537, over 4852.00 frames. ], tot_loss[loss=0.1752, simple_loss=0.247, pruned_loss=0.0517, over 952321.35 frames. ], batch size: 44, lr: 3.18e-03, grad_scale: 32.0 2023-04-27 16:00:30,592 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.010e+02 1.788e+02 2.087e+02 2.542e+02 6.471e+02, threshold=4.174e+02, percent-clipped=4.0 2023-04-27 16:00:39,546 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=119075.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:00:58,654 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=119093.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:01:15,265 INFO [finetune.py:976] (1/7) Epoch 21, batch 4550, loss[loss=0.1687, simple_loss=0.2453, pruned_loss=0.04606, over 4785.00 frames. ], tot_loss[loss=0.1761, simple_loss=0.2484, pruned_loss=0.05195, over 952964.23 frames. ], batch size: 51, lr: 3.18e-03, grad_scale: 32.0 2023-04-27 16:01:53,827 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=119136.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:02:16,872 INFO [finetune.py:976] (1/7) Epoch 21, batch 4600, loss[loss=0.2142, simple_loss=0.273, pruned_loss=0.07774, over 4821.00 frames. ], tot_loss[loss=0.1745, simple_loss=0.2469, pruned_loss=0.05105, over 953210.09 frames. ], batch size: 30, lr: 3.18e-03, grad_scale: 32.0 2023-04-27 16:02:17,004 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=119154.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 16:02:20,089 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5650, 2.6830, 2.1369, 2.3000, 2.7658, 2.2775, 3.5602, 1.9874], device='cuda:1'), covar=tensor([0.3861, 0.2155, 0.4373, 0.3651, 0.1958, 0.2728, 0.1611, 0.4262], device='cuda:1'), in_proj_covar=tensor([0.0344, 0.0353, 0.0428, 0.0357, 0.0384, 0.0377, 0.0373, 0.0422], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 16:02:21,811 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.037e+02 1.477e+02 1.832e+02 2.213e+02 3.294e+02, threshold=3.663e+02, percent-clipped=0.0 2023-04-27 16:02:50,904 INFO [finetune.py:976] (1/7) Epoch 21, batch 4650, loss[loss=0.1488, simple_loss=0.2231, pruned_loss=0.03725, over 4868.00 frames. ], tot_loss[loss=0.1725, simple_loss=0.2442, pruned_loss=0.05037, over 954214.49 frames. ], batch size: 31, lr: 3.18e-03, grad_scale: 32.0 2023-04-27 16:03:00,205 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 2023-04-27 16:03:24,687 INFO [finetune.py:976] (1/7) Epoch 21, batch 4700, loss[loss=0.1431, simple_loss=0.2076, pruned_loss=0.03927, over 4773.00 frames. ], tot_loss[loss=0.1691, simple_loss=0.2403, pruned_loss=0.0489, over 955617.54 frames. ], batch size: 28, lr: 3.18e-03, grad_scale: 32.0 2023-04-27 16:03:29,613 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.111e+02 1.611e+02 1.939e+02 2.351e+02 3.791e+02, threshold=3.879e+02, percent-clipped=1.0 2023-04-27 16:03:45,990 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=119287.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:03:58,705 INFO [finetune.py:976] (1/7) Epoch 21, batch 4750, loss[loss=0.1626, simple_loss=0.2297, pruned_loss=0.04779, over 4872.00 frames. ], tot_loss[loss=0.169, simple_loss=0.2393, pruned_loss=0.04931, over 954785.02 frames. ], batch size: 31, lr: 3.17e-03, grad_scale: 32.0 2023-04-27 16:04:39,497 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=119335.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:04:54,474 INFO [finetune.py:976] (1/7) Epoch 21, batch 4800, loss[loss=0.1409, simple_loss=0.2244, pruned_loss=0.02873, over 4934.00 frames. ], tot_loss[loss=0.1702, simple_loss=0.2415, pruned_loss=0.04949, over 956342.75 frames. ], batch size: 33, lr: 3.17e-03, grad_scale: 32.0 2023-04-27 16:04:59,370 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.005e+02 1.524e+02 1.812e+02 2.153e+02 4.144e+02, threshold=3.624e+02, percent-clipped=1.0 2023-04-27 16:04:59,521 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.1980, 1.7118, 1.9670, 2.5488, 2.0323, 1.6698, 1.6616, 1.9275], device='cuda:1'), covar=tensor([0.2716, 0.3008, 0.1628, 0.1900, 0.2393, 0.2380, 0.3737, 0.1987], device='cuda:1'), in_proj_covar=tensor([0.0293, 0.0245, 0.0227, 0.0315, 0.0219, 0.0233, 0.0228, 0.0184], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-27 16:05:03,126 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.2345, 1.9651, 1.7017, 1.5904, 2.0479, 1.6990, 2.3699, 1.5245], device='cuda:1'), covar=tensor([0.2807, 0.1565, 0.3697, 0.2551, 0.1362, 0.1941, 0.1422, 0.3839], device='cuda:1'), in_proj_covar=tensor([0.0340, 0.0348, 0.0423, 0.0352, 0.0379, 0.0373, 0.0366, 0.0416], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 16:05:26,791 INFO [finetune.py:976] (1/7) Epoch 21, batch 4850, loss[loss=0.1311, simple_loss=0.2036, pruned_loss=0.02927, over 4786.00 frames. ], tot_loss[loss=0.1733, simple_loss=0.2451, pruned_loss=0.05074, over 952530.94 frames. ], batch size: 29, lr: 3.17e-03, grad_scale: 32.0 2023-04-27 16:05:28,036 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-04-27 16:05:30,291 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5109, 1.7881, 0.7609, 1.2530, 1.9727, 1.3769, 1.3184, 1.3514], device='cuda:1'), covar=tensor([0.0619, 0.0334, 0.0351, 0.0609, 0.0256, 0.0648, 0.0658, 0.0669], device='cuda:1'), in_proj_covar=tensor([0.0028, 0.0024, 0.0022, 0.0029, 0.0019, 0.0028, 0.0028, 0.0028], device='cuda:1'), out_proj_covar=tensor([0.0051, 0.0045, 0.0038, 0.0052, 0.0038, 0.0050, 0.0050, 0.0051], device='cuda:1') 2023-04-27 16:05:41,859 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.2024, 1.6925, 2.1110, 2.5946, 2.1258, 1.6758, 1.4661, 1.9040], device='cuda:1'), covar=tensor([0.3267, 0.3384, 0.1665, 0.2300, 0.2602, 0.2809, 0.4069, 0.2277], device='cuda:1'), in_proj_covar=tensor([0.0293, 0.0245, 0.0227, 0.0315, 0.0219, 0.0233, 0.0228, 0.0184], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-27 16:05:43,556 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=119431.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:05:55,512 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=119449.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 16:05:58,988 INFO [finetune.py:976] (1/7) Epoch 21, batch 4900, loss[loss=0.1944, simple_loss=0.2642, pruned_loss=0.06231, over 4851.00 frames. ], tot_loss[loss=0.1741, simple_loss=0.2463, pruned_loss=0.05095, over 952608.19 frames. ], batch size: 44, lr: 3.17e-03, grad_scale: 32.0 2023-04-27 16:06:04,797 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.059e+02 1.637e+02 1.876e+02 2.288e+02 4.482e+02, threshold=3.752e+02, percent-clipped=1.0 2023-04-27 16:06:35,841 INFO [finetune.py:976] (1/7) Epoch 21, batch 4950, loss[loss=0.1387, simple_loss=0.2116, pruned_loss=0.03284, over 4748.00 frames. ], tot_loss[loss=0.1739, simple_loss=0.2464, pruned_loss=0.05073, over 952196.53 frames. ], batch size: 23, lr: 3.17e-03, grad_scale: 16.0 2023-04-27 16:06:46,802 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.50 vs. limit=2.0 2023-04-27 16:06:47,778 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.0763, 0.8547, 0.8894, 0.8138, 1.2477, 0.9827, 0.9610, 0.9522], device='cuda:1'), covar=tensor([0.1769, 0.1726, 0.2246, 0.1800, 0.1232, 0.1761, 0.1734, 0.2771], device='cuda:1'), in_proj_covar=tensor([0.0311, 0.0310, 0.0351, 0.0288, 0.0326, 0.0307, 0.0300, 0.0370], device='cuda:1'), out_proj_covar=tensor([6.3840e-05, 6.4092e-05, 7.4140e-05, 5.8081e-05, 6.7302e-05, 6.4329e-05, 6.2916e-05, 7.8521e-05], device='cuda:1') 2023-04-27 16:06:48,360 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.2490, 3.0306, 0.8539, 1.6767, 1.6875, 2.2044, 1.7438, 0.9971], device='cuda:1'), covar=tensor([0.1511, 0.1045, 0.2001, 0.1263, 0.1145, 0.0971, 0.1559, 0.1849], device='cuda:1'), in_proj_covar=tensor([0.0117, 0.0238, 0.0136, 0.0119, 0.0132, 0.0151, 0.0115, 0.0118], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-27 16:06:49,603 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7186, 1.8436, 0.7244, 1.3783, 1.8745, 1.5715, 1.4567, 1.5827], device='cuda:1'), covar=tensor([0.0484, 0.0356, 0.0348, 0.0537, 0.0263, 0.0490, 0.0485, 0.0544], device='cuda:1'), in_proj_covar=tensor([0.0028, 0.0024, 0.0022, 0.0029, 0.0019, 0.0028, 0.0028, 0.0028], device='cuda:1'), out_proj_covar=tensor([0.0051, 0.0045, 0.0038, 0.0052, 0.0038, 0.0050, 0.0050, 0.0051], device='cuda:1') 2023-04-27 16:06:49,618 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=119514.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:07:42,888 INFO [finetune.py:976] (1/7) Epoch 21, batch 5000, loss[loss=0.1665, simple_loss=0.2365, pruned_loss=0.04824, over 4746.00 frames. ], tot_loss[loss=0.172, simple_loss=0.2442, pruned_loss=0.04992, over 953051.38 frames. ], batch size: 27, lr: 3.17e-03, grad_scale: 16.0 2023-04-27 16:07:56,240 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 8.721e+01 1.533e+02 1.932e+02 2.312e+02 5.244e+02, threshold=3.864e+02, percent-clipped=4.0 2023-04-27 16:08:15,223 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=119575.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:08:44,469 INFO [finetune.py:976] (1/7) Epoch 21, batch 5050, loss[loss=0.1588, simple_loss=0.23, pruned_loss=0.04374, over 4836.00 frames. ], tot_loss[loss=0.1692, simple_loss=0.2409, pruned_loss=0.04877, over 954477.82 frames. ], batch size: 33, lr: 3.17e-03, grad_scale: 16.0 2023-04-27 16:09:28,917 INFO [finetune.py:976] (1/7) Epoch 21, batch 5100, loss[loss=0.1382, simple_loss=0.2074, pruned_loss=0.03451, over 4820.00 frames. ], tot_loss[loss=0.1678, simple_loss=0.2389, pruned_loss=0.04833, over 955748.96 frames. ], batch size: 51, lr: 3.17e-03, grad_scale: 16.0 2023-04-27 16:09:41,673 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.943e+01 1.499e+02 1.766e+02 2.161e+02 3.760e+02, threshold=3.532e+02, percent-clipped=0.0 2023-04-27 16:09:43,048 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=119665.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:10:03,714 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-04-27 16:10:31,243 INFO [finetune.py:976] (1/7) Epoch 21, batch 5150, loss[loss=0.1791, simple_loss=0.2461, pruned_loss=0.05609, over 4765.00 frames. ], tot_loss[loss=0.1686, simple_loss=0.2396, pruned_loss=0.04881, over 955889.90 frames. ], batch size: 28, lr: 3.17e-03, grad_scale: 16.0 2023-04-27 16:11:05,064 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=119726.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:11:08,007 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=119731.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:11:28,051 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7910, 2.3121, 0.8975, 1.1991, 1.5098, 1.1142, 2.4611, 1.2449], device='cuda:1'), covar=tensor([0.0751, 0.0639, 0.0656, 0.1229, 0.0458, 0.1018, 0.0300, 0.0715], device='cuda:1'), in_proj_covar=tensor([0.0050, 0.0064, 0.0047, 0.0046, 0.0049, 0.0051, 0.0073, 0.0050], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0007], device='cuda:1') 2023-04-27 16:11:30,495 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=119749.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 16:11:33,427 INFO [finetune.py:976] (1/7) Epoch 21, batch 5200, loss[loss=0.2178, simple_loss=0.302, pruned_loss=0.06686, over 4849.00 frames. ], tot_loss[loss=0.172, simple_loss=0.2439, pruned_loss=0.05011, over 955036.62 frames. ], batch size: 44, lr: 3.17e-03, grad_scale: 16.0 2023-04-27 16:11:39,371 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 8.908e+01 1.576e+02 2.065e+02 2.326e+02 4.790e+02, threshold=4.130e+02, percent-clipped=1.0 2023-04-27 16:11:51,106 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=119779.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:12:01,650 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5391, 1.3654, 1.2424, 1.3577, 1.7971, 1.4918, 1.2591, 1.2092], device='cuda:1'), covar=tensor([0.1635, 0.1483, 0.2027, 0.1463, 0.0854, 0.1625, 0.1977, 0.2327], device='cuda:1'), in_proj_covar=tensor([0.0313, 0.0312, 0.0352, 0.0289, 0.0328, 0.0309, 0.0302, 0.0372], device='cuda:1'), out_proj_covar=tensor([6.4283e-05, 6.4473e-05, 7.4307e-05, 5.8269e-05, 6.7644e-05, 6.4680e-05, 6.3294e-05, 7.8964e-05], device='cuda:1') 2023-04-27 16:12:02,196 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=119797.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:12:02,846 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0852, 2.5932, 1.1126, 1.3784, 2.0817, 1.2775, 3.2588, 1.7214], device='cuda:1'), covar=tensor([0.0658, 0.0581, 0.0741, 0.1203, 0.0440, 0.0977, 0.0226, 0.0570], device='cuda:1'), in_proj_covar=tensor([0.0050, 0.0064, 0.0047, 0.0046, 0.0049, 0.0051, 0.0073, 0.0050], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0007], device='cuda:1') 2023-04-27 16:12:06,570 INFO [finetune.py:976] (1/7) Epoch 21, batch 5250, loss[loss=0.2013, simple_loss=0.2626, pruned_loss=0.07002, over 4897.00 frames. ], tot_loss[loss=0.1726, simple_loss=0.2451, pruned_loss=0.05008, over 955446.90 frames. ], batch size: 35, lr: 3.17e-03, grad_scale: 16.0 2023-04-27 16:12:40,884 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6374, 1.4976, 1.3375, 1.4836, 1.8952, 1.5722, 1.3262, 1.2781], device='cuda:1'), covar=tensor([0.1627, 0.1216, 0.1558, 0.1284, 0.0902, 0.1547, 0.1871, 0.2034], device='cuda:1'), in_proj_covar=tensor([0.0312, 0.0311, 0.0351, 0.0288, 0.0327, 0.0308, 0.0302, 0.0371], device='cuda:1'), out_proj_covar=tensor([6.4174e-05, 6.4355e-05, 7.4103e-05, 5.8117e-05, 6.7538e-05, 6.4518e-05, 6.3212e-05, 7.8852e-05], device='cuda:1') 2023-04-27 16:12:51,197 INFO [finetune.py:976] (1/7) Epoch 21, batch 5300, loss[loss=0.1743, simple_loss=0.2351, pruned_loss=0.05679, over 4185.00 frames. ], tot_loss[loss=0.1746, simple_loss=0.2473, pruned_loss=0.05096, over 955051.69 frames. ], batch size: 65, lr: 3.17e-03, grad_scale: 16.0 2023-04-27 16:13:02,329 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.110e+02 1.628e+02 1.910e+02 2.445e+02 5.972e+02, threshold=3.820e+02, percent-clipped=3.0 2023-04-27 16:13:04,756 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4234, 1.3526, 4.0996, 3.8476, 3.6693, 4.0086, 3.8821, 3.5963], device='cuda:1'), covar=tensor([0.7063, 0.6000, 0.1185, 0.1783, 0.1179, 0.1713, 0.1805, 0.1774], device='cuda:1'), in_proj_covar=tensor([0.0308, 0.0304, 0.0402, 0.0402, 0.0346, 0.0407, 0.0310, 0.0364], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-27 16:13:13,806 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=119870.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:13:37,219 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.1920, 1.5486, 2.0705, 2.3260, 2.0031, 1.5909, 1.3387, 1.7969], device='cuda:1'), covar=tensor([0.3167, 0.3179, 0.1556, 0.2234, 0.2456, 0.2589, 0.4041, 0.2058], device='cuda:1'), in_proj_covar=tensor([0.0292, 0.0244, 0.0226, 0.0314, 0.0219, 0.0233, 0.0228, 0.0183], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-27 16:13:53,932 INFO [finetune.py:976] (1/7) Epoch 21, batch 5350, loss[loss=0.1957, simple_loss=0.2628, pruned_loss=0.06426, over 4888.00 frames. ], tot_loss[loss=0.1749, simple_loss=0.2478, pruned_loss=0.05094, over 955993.67 frames. ], batch size: 43, lr: 3.17e-03, grad_scale: 16.0 2023-04-27 16:14:43,494 INFO [finetune.py:976] (1/7) Epoch 21, batch 5400, loss[loss=0.1505, simple_loss=0.2173, pruned_loss=0.0418, over 4117.00 frames. ], tot_loss[loss=0.173, simple_loss=0.2457, pruned_loss=0.05018, over 956598.33 frames. ], batch size: 65, lr: 3.17e-03, grad_scale: 16.0 2023-04-27 16:14:48,951 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.105e+02 1.397e+02 1.683e+02 2.017e+02 3.693e+02, threshold=3.366e+02, percent-clipped=0.0 2023-04-27 16:15:13,881 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=3.39 vs. limit=5.0 2023-04-27 16:15:18,156 INFO [finetune.py:976] (1/7) Epoch 21, batch 5450, loss[loss=0.1778, simple_loss=0.256, pruned_loss=0.04979, over 4902.00 frames. ], tot_loss[loss=0.1702, simple_loss=0.242, pruned_loss=0.04919, over 957460.31 frames. ], batch size: 37, lr: 3.17e-03, grad_scale: 16.0 2023-04-27 16:15:28,063 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.3291, 1.7591, 2.2138, 2.7007, 2.2542, 1.7588, 1.6285, 2.0979], device='cuda:1'), covar=tensor([0.3055, 0.2982, 0.1498, 0.2101, 0.2385, 0.2523, 0.4092, 0.1859], device='cuda:1'), in_proj_covar=tensor([0.0293, 0.0244, 0.0226, 0.0313, 0.0219, 0.0233, 0.0227, 0.0183], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-27 16:15:34,276 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=120021.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:16:03,187 INFO [finetune.py:976] (1/7) Epoch 21, batch 5500, loss[loss=0.1729, simple_loss=0.2475, pruned_loss=0.04919, over 4894.00 frames. ], tot_loss[loss=0.1683, simple_loss=0.2392, pruned_loss=0.04864, over 957503.83 frames. ], batch size: 32, lr: 3.17e-03, grad_scale: 16.0 2023-04-27 16:16:14,181 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.019e+02 1.665e+02 1.823e+02 2.184e+02 3.265e+02, threshold=3.646e+02, percent-clipped=0.0 2023-04-27 16:17:07,735 INFO [finetune.py:976] (1/7) Epoch 21, batch 5550, loss[loss=0.164, simple_loss=0.2284, pruned_loss=0.04974, over 4829.00 frames. ], tot_loss[loss=0.1681, simple_loss=0.2394, pruned_loss=0.04837, over 957424.44 frames. ], batch size: 25, lr: 3.17e-03, grad_scale: 16.0 2023-04-27 16:17:29,467 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5415, 3.3020, 1.1912, 1.8226, 1.8008, 2.4695, 1.9350, 0.9680], device='cuda:1'), covar=tensor([0.1376, 0.0984, 0.1741, 0.1196, 0.1097, 0.0928, 0.1581, 0.2085], device='cuda:1'), in_proj_covar=tensor([0.0118, 0.0242, 0.0138, 0.0120, 0.0134, 0.0153, 0.0117, 0.0119], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-27 16:17:37,968 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5475, 1.3601, 1.6844, 1.7057, 1.3915, 1.3206, 1.4580, 0.7453], device='cuda:1'), covar=tensor([0.0518, 0.0654, 0.0394, 0.0711, 0.0823, 0.1127, 0.0539, 0.0625], device='cuda:1'), in_proj_covar=tensor([0.0069, 0.0068, 0.0067, 0.0068, 0.0075, 0.0097, 0.0073, 0.0065], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-27 16:18:05,780 INFO [finetune.py:976] (1/7) Epoch 21, batch 5600, loss[loss=0.2116, simple_loss=0.2922, pruned_loss=0.06544, over 4837.00 frames. ], tot_loss[loss=0.1718, simple_loss=0.2445, pruned_loss=0.04961, over 955778.71 frames. ], batch size: 49, lr: 3.17e-03, grad_scale: 16.0 2023-04-27 16:18:08,925 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-04-27 16:18:11,019 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.101e+02 1.536e+02 1.871e+02 2.277e+02 4.229e+02, threshold=3.743e+02, percent-clipped=5.0 2023-04-27 16:18:13,454 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.0576, 1.2588, 1.1824, 1.5320, 1.4044, 1.4304, 1.2394, 2.4871], device='cuda:1'), covar=tensor([0.0624, 0.0856, 0.0860, 0.1229, 0.0659, 0.0543, 0.0793, 0.0213], device='cuda:1'), in_proj_covar=tensor([0.0037, 0.0038, 0.0039, 0.0043, 0.0040, 0.0038, 0.0038, 0.0055], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:1') 2023-04-27 16:18:15,215 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=120170.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:18:30,194 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-04-27 16:18:36,563 INFO [finetune.py:976] (1/7) Epoch 21, batch 5650, loss[loss=0.1563, simple_loss=0.2293, pruned_loss=0.04165, over 4838.00 frames. ], tot_loss[loss=0.1726, simple_loss=0.2459, pruned_loss=0.04962, over 954829.87 frames. ], batch size: 49, lr: 3.17e-03, grad_scale: 16.0 2023-04-27 16:18:45,334 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=120218.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:19:08,060 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9390, 1.5872, 2.0057, 2.1742, 1.7462, 1.6153, 1.8049, 1.2707], device='cuda:1'), covar=tensor([0.0388, 0.0625, 0.0386, 0.0407, 0.0592, 0.0996, 0.0527, 0.0555], device='cuda:1'), in_proj_covar=tensor([0.0069, 0.0068, 0.0067, 0.0068, 0.0075, 0.0097, 0.0073, 0.0065], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-27 16:19:09,211 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5614, 1.4508, 0.7716, 1.2887, 1.4456, 1.4152, 1.3609, 1.4032], device='cuda:1'), covar=tensor([0.0423, 0.0310, 0.0350, 0.0504, 0.0268, 0.0466, 0.0452, 0.0465], device='cuda:1'), in_proj_covar=tensor([0.0028, 0.0024, 0.0022, 0.0029, 0.0019, 0.0028, 0.0028, 0.0028], device='cuda:1'), out_proj_covar=tensor([0.0051, 0.0045, 0.0038, 0.0052, 0.0038, 0.0050, 0.0050, 0.0051], device='cuda:1') 2023-04-27 16:19:30,421 INFO [finetune.py:976] (1/7) Epoch 21, batch 5700, loss[loss=0.1583, simple_loss=0.2133, pruned_loss=0.05168, over 4266.00 frames. ], tot_loss[loss=0.1687, simple_loss=0.2406, pruned_loss=0.04837, over 934752.05 frames. ], batch size: 18, lr: 3.17e-03, grad_scale: 16.0 2023-04-27 16:19:39,340 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.3368, 2.7359, 1.2778, 1.6190, 2.2974, 1.5019, 3.6084, 2.2184], device='cuda:1'), covar=tensor([0.0645, 0.0608, 0.0709, 0.1227, 0.0437, 0.0946, 0.0317, 0.0524], device='cuda:1'), in_proj_covar=tensor([0.0051, 0.0064, 0.0047, 0.0046, 0.0049, 0.0052, 0.0073, 0.0050], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0007], device='cuda:1') 2023-04-27 16:19:39,350 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6662, 1.8252, 1.7322, 2.1049, 1.9937, 2.0775, 1.6871, 3.7002], device='cuda:1'), covar=tensor([0.0542, 0.0771, 0.0807, 0.1086, 0.0602, 0.0470, 0.0699, 0.0156], device='cuda:1'), in_proj_covar=tensor([0.0037, 0.0038, 0.0039, 0.0043, 0.0040, 0.0038, 0.0038, 0.0055], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:1') 2023-04-27 16:19:41,631 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.050e+02 1.444e+02 1.777e+02 2.267e+02 3.469e+02, threshold=3.555e+02, percent-clipped=0.0 2023-04-27 16:20:17,302 INFO [finetune.py:976] (1/7) Epoch 22, batch 0, loss[loss=0.1678, simple_loss=0.2403, pruned_loss=0.0476, over 4912.00 frames. ], tot_loss[loss=0.1678, simple_loss=0.2403, pruned_loss=0.0476, over 4912.00 frames. ], batch size: 38, lr: 3.16e-03, grad_scale: 16.0 2023-04-27 16:20:17,302 INFO [finetune.py:1001] (1/7) Computing validation loss 2023-04-27 16:20:33,672 INFO [finetune.py:1010] (1/7) Epoch 22, validation: loss=0.1546, simple_loss=0.2251, pruned_loss=0.04204, over 2265189.00 frames. 2023-04-27 16:20:33,673 INFO [finetune.py:1011] (1/7) Maximum memory allocated so far is 6391MB 2023-04-27 16:20:37,879 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9924, 1.8672, 2.0489, 2.4637, 2.4662, 1.9360, 1.6771, 2.2039], device='cuda:1'), covar=tensor([0.0882, 0.1070, 0.0761, 0.0572, 0.0617, 0.0882, 0.0814, 0.0578], device='cuda:1'), in_proj_covar=tensor([0.0187, 0.0201, 0.0183, 0.0174, 0.0177, 0.0180, 0.0152, 0.0178], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 16:20:59,174 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=120321.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:21:04,986 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=2.00 vs. limit=2.0 2023-04-27 16:21:06,359 INFO [finetune.py:976] (1/7) Epoch 22, batch 50, loss[loss=0.1495, simple_loss=0.231, pruned_loss=0.03399, over 4884.00 frames. ], tot_loss[loss=0.1702, simple_loss=0.245, pruned_loss=0.04765, over 216884.57 frames. ], batch size: 35, lr: 3.16e-03, grad_scale: 16.0 2023-04-27 16:21:21,923 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6515, 0.9728, 1.6784, 2.1232, 1.7369, 1.6131, 1.6425, 1.6216], device='cuda:1'), covar=tensor([0.4116, 0.6196, 0.5620, 0.5185, 0.5689, 0.6823, 0.7137, 0.8262], device='cuda:1'), in_proj_covar=tensor([0.0429, 0.0412, 0.0505, 0.0504, 0.0457, 0.0487, 0.0495, 0.0502], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 16:21:27,311 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 8.523e+01 1.622e+02 1.941e+02 2.362e+02 4.057e+02, threshold=3.882e+02, percent-clipped=2.0 2023-04-27 16:21:31,094 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=120369.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:21:51,383 INFO [finetune.py:976] (1/7) Epoch 22, batch 100, loss[loss=0.171, simple_loss=0.2377, pruned_loss=0.05216, over 4889.00 frames. ], tot_loss[loss=0.1667, simple_loss=0.2391, pruned_loss=0.04711, over 381101.55 frames. ], batch size: 35, lr: 3.16e-03, grad_scale: 16.0 2023-04-27 16:22:13,780 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([3.0584, 2.2861, 2.1661, 2.2140, 1.8933, 2.1872, 2.2982, 2.2400], device='cuda:1'), covar=tensor([0.3876, 0.6144, 0.5012, 0.5346, 0.6453, 0.7530, 0.5949, 0.5477], device='cuda:1'), in_proj_covar=tensor([0.0338, 0.0373, 0.0323, 0.0337, 0.0347, 0.0394, 0.0358, 0.0329], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 16:22:25,494 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.63 vs. limit=5.0 2023-04-27 16:22:55,499 INFO [finetune.py:976] (1/7) Epoch 22, batch 150, loss[loss=0.1796, simple_loss=0.2353, pruned_loss=0.06197, over 4307.00 frames. ], tot_loss[loss=0.165, simple_loss=0.2355, pruned_loss=0.04718, over 507428.31 frames. ], batch size: 65, lr: 3.16e-03, grad_scale: 16.0 2023-04-27 16:23:29,000 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.065e+02 1.459e+02 1.791e+02 2.121e+02 3.336e+02, threshold=3.582e+02, percent-clipped=0.0 2023-04-27 16:23:57,909 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-04-27 16:23:58,242 INFO [finetune.py:976] (1/7) Epoch 22, batch 200, loss[loss=0.1436, simple_loss=0.2165, pruned_loss=0.03531, over 4914.00 frames. ], tot_loss[loss=0.1639, simple_loss=0.2342, pruned_loss=0.04682, over 605829.24 frames. ], batch size: 36, lr: 3.16e-03, grad_scale: 16.0 2023-04-27 16:25:06,005 INFO [finetune.py:976] (1/7) Epoch 22, batch 250, loss[loss=0.2032, simple_loss=0.2634, pruned_loss=0.07147, over 4913.00 frames. ], tot_loss[loss=0.1647, simple_loss=0.2352, pruned_loss=0.04705, over 681887.89 frames. ], batch size: 35, lr: 3.16e-03, grad_scale: 16.0 2023-04-27 16:25:37,608 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-04-27 16:25:49,441 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.081e+02 1.595e+02 1.873e+02 2.274e+02 4.484e+02, threshold=3.746e+02, percent-clipped=2.0 2023-04-27 16:26:12,261 INFO [finetune.py:976] (1/7) Epoch 22, batch 300, loss[loss=0.1817, simple_loss=0.2463, pruned_loss=0.05853, over 4910.00 frames. ], tot_loss[loss=0.1691, simple_loss=0.2409, pruned_loss=0.0487, over 743927.32 frames. ], batch size: 43, lr: 3.16e-03, grad_scale: 16.0 2023-04-27 16:27:06,863 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([0.6760, 1.6688, 1.6314, 1.2351, 1.7785, 1.4332, 2.2301, 1.4222], device='cuda:1'), covar=tensor([0.3531, 0.1783, 0.4607, 0.2927, 0.1518, 0.2291, 0.1417, 0.4420], device='cuda:1'), in_proj_covar=tensor([0.0341, 0.0351, 0.0427, 0.0354, 0.0382, 0.0374, 0.0369, 0.0419], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 16:27:19,464 INFO [finetune.py:976] (1/7) Epoch 22, batch 350, loss[loss=0.2041, simple_loss=0.286, pruned_loss=0.06114, over 4923.00 frames. ], tot_loss[loss=0.173, simple_loss=0.2451, pruned_loss=0.05041, over 790822.01 frames. ], batch size: 38, lr: 3.16e-03, grad_scale: 16.0 2023-04-27 16:28:01,875 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.016e+02 1.528e+02 1.802e+02 2.377e+02 4.881e+02, threshold=3.603e+02, percent-clipped=3.0 2023-04-27 16:28:20,626 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=120676.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:28:24,207 INFO [finetune.py:976] (1/7) Epoch 22, batch 400, loss[loss=0.1533, simple_loss=0.2208, pruned_loss=0.04286, over 4216.00 frames. ], tot_loss[loss=0.1747, simple_loss=0.2471, pruned_loss=0.0511, over 828376.79 frames. ], batch size: 18, lr: 3.16e-03, grad_scale: 16.0 2023-04-27 16:28:44,119 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8369, 2.4416, 2.0096, 1.8228, 1.3324, 1.4098, 2.0600, 1.3082], device='cuda:1'), covar=tensor([0.1667, 0.1464, 0.1368, 0.1728, 0.2290, 0.1883, 0.0898, 0.2001], device='cuda:1'), in_proj_covar=tensor([0.0196, 0.0209, 0.0168, 0.0202, 0.0198, 0.0184, 0.0154, 0.0185], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-27 16:29:18,085 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.3276, 1.5607, 1.7450, 1.8207, 1.6975, 1.7826, 1.7781, 1.8238], device='cuda:1'), covar=tensor([0.3869, 0.4996, 0.4267, 0.4297, 0.5306, 0.6889, 0.4939, 0.4186], device='cuda:1'), in_proj_covar=tensor([0.0339, 0.0372, 0.0323, 0.0338, 0.0348, 0.0394, 0.0357, 0.0329], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 16:29:32,131 INFO [finetune.py:976] (1/7) Epoch 22, batch 450, loss[loss=0.2374, simple_loss=0.2892, pruned_loss=0.0928, over 4756.00 frames. ], tot_loss[loss=0.174, simple_loss=0.246, pruned_loss=0.05101, over 856554.08 frames. ], batch size: 54, lr: 3.16e-03, grad_scale: 16.0 2023-04-27 16:29:40,505 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=120737.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:29:48,355 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5904, 1.4545, 0.5822, 1.2921, 1.5052, 1.4576, 1.3540, 1.4084], device='cuda:1'), covar=tensor([0.0485, 0.0358, 0.0389, 0.0552, 0.0268, 0.0494, 0.0471, 0.0564], device='cuda:1'), in_proj_covar=tensor([0.0028, 0.0024, 0.0022, 0.0029, 0.0019, 0.0028, 0.0028, 0.0028], device='cuda:1'), out_proj_covar=tensor([0.0051, 0.0045, 0.0038, 0.0052, 0.0038, 0.0050, 0.0050, 0.0051], device='cuda:1') 2023-04-27 16:29:50,783 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.3381, 2.9871, 0.8287, 1.6311, 1.6066, 2.2684, 1.7598, 0.9432], device='cuda:1'), covar=tensor([0.1386, 0.0924, 0.1925, 0.1227, 0.1140, 0.0884, 0.1430, 0.1838], device='cuda:1'), in_proj_covar=tensor([0.0116, 0.0238, 0.0136, 0.0118, 0.0131, 0.0151, 0.0115, 0.0117], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-27 16:30:03,657 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.146e+02 1.511e+02 1.855e+02 2.296e+02 3.408e+02, threshold=3.711e+02, percent-clipped=0.0 2023-04-27 16:30:06,832 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([4.8095, 4.5787, 3.2092, 5.5487, 4.9470, 4.8344, 2.3570, 4.7502], device='cuda:1'), covar=tensor([0.1807, 0.0846, 0.2996, 0.1007, 0.2334, 0.1680, 0.5504, 0.2043], device='cuda:1'), in_proj_covar=tensor([0.0244, 0.0215, 0.0250, 0.0303, 0.0294, 0.0244, 0.0274, 0.0271], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 16:30:13,542 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.1365, 2.3312, 1.9879, 1.8901, 2.4190, 1.9489, 2.9949, 1.6895], device='cuda:1'), covar=tensor([0.3652, 0.1891, 0.4439, 0.2894, 0.1577, 0.2536, 0.1325, 0.4456], device='cuda:1'), in_proj_covar=tensor([0.0340, 0.0350, 0.0426, 0.0353, 0.0381, 0.0373, 0.0368, 0.0418], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 16:30:15,269 INFO [finetune.py:976] (1/7) Epoch 22, batch 500, loss[loss=0.1567, simple_loss=0.2224, pruned_loss=0.04544, over 4910.00 frames. ], tot_loss[loss=0.1714, simple_loss=0.2426, pruned_loss=0.0501, over 878611.04 frames. ], batch size: 37, lr: 3.16e-03, grad_scale: 16.0 2023-04-27 16:30:21,082 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2023-04-27 16:30:49,306 INFO [finetune.py:976] (1/7) Epoch 22, batch 550, loss[loss=0.1616, simple_loss=0.2296, pruned_loss=0.04684, over 4756.00 frames. ], tot_loss[loss=0.1687, simple_loss=0.2396, pruned_loss=0.0489, over 894702.52 frames. ], batch size: 27, lr: 3.16e-03, grad_scale: 16.0 2023-04-27 16:30:59,022 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4634, 1.3615, 4.1139, 3.8331, 3.5668, 3.9235, 3.8508, 3.6018], device='cuda:1'), covar=tensor([0.7015, 0.5703, 0.1086, 0.1697, 0.1198, 0.1833, 0.1593, 0.1505], device='cuda:1'), in_proj_covar=tensor([0.0314, 0.0307, 0.0409, 0.0408, 0.0350, 0.0412, 0.0314, 0.0369], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-27 16:31:15,724 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.143e+02 1.573e+02 1.876e+02 2.291e+02 4.467e+02, threshold=3.751e+02, percent-clipped=2.0 2023-04-27 16:31:38,950 INFO [finetune.py:976] (1/7) Epoch 22, batch 600, loss[loss=0.1746, simple_loss=0.2397, pruned_loss=0.05475, over 4837.00 frames. ], tot_loss[loss=0.1702, simple_loss=0.241, pruned_loss=0.0497, over 908921.03 frames. ], batch size: 30, lr: 3.16e-03, grad_scale: 16.0 2023-04-27 16:32:45,428 INFO [finetune.py:976] (1/7) Epoch 22, batch 650, loss[loss=0.187, simple_loss=0.2633, pruned_loss=0.05538, over 4808.00 frames. ], tot_loss[loss=0.1733, simple_loss=0.2448, pruned_loss=0.05085, over 920404.98 frames. ], batch size: 39, lr: 3.16e-03, grad_scale: 16.0 2023-04-27 16:33:26,708 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.051e+02 1.656e+02 1.853e+02 2.297e+02 3.999e+02, threshold=3.705e+02, percent-clipped=2.0 2023-04-27 16:33:52,001 INFO [finetune.py:976] (1/7) Epoch 22, batch 700, loss[loss=0.1615, simple_loss=0.2268, pruned_loss=0.04808, over 4829.00 frames. ], tot_loss[loss=0.1747, simple_loss=0.2469, pruned_loss=0.0513, over 929136.76 frames. ], batch size: 30, lr: 3.16e-03, grad_scale: 16.0 2023-04-27 16:34:21,029 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=121003.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:34:35,164 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.0499, 2.3986, 0.8307, 1.2446, 1.3902, 1.7845, 1.5922, 0.8414], device='cuda:1'), covar=tensor([0.1934, 0.1913, 0.2103, 0.1953, 0.1456, 0.1311, 0.1686, 0.1834], device='cuda:1'), in_proj_covar=tensor([0.0117, 0.0240, 0.0137, 0.0119, 0.0132, 0.0152, 0.0116, 0.0118], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-27 16:34:45,577 INFO [finetune.py:976] (1/7) Epoch 22, batch 750, loss[loss=0.1021, simple_loss=0.1642, pruned_loss=0.01994, over 3994.00 frames. ], tot_loss[loss=0.1754, simple_loss=0.2476, pruned_loss=0.05158, over 934121.52 frames. ], batch size: 17, lr: 3.16e-03, grad_scale: 16.0 2023-04-27 16:34:45,655 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=121032.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:34:56,537 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4893, 3.3744, 0.9952, 1.7810, 1.7841, 2.4936, 1.8733, 1.0520], device='cuda:1'), covar=tensor([0.1403, 0.0917, 0.2005, 0.1275, 0.1159, 0.0896, 0.1554, 0.2063], device='cuda:1'), in_proj_covar=tensor([0.0118, 0.0241, 0.0138, 0.0120, 0.0133, 0.0153, 0.0117, 0.0119], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-27 16:35:03,933 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.71 vs. limit=5.0 2023-04-27 16:35:04,851 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.944e+01 1.522e+02 1.865e+02 2.464e+02 7.582e+02, threshold=3.731e+02, percent-clipped=3.0 2023-04-27 16:35:05,576 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=121064.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:35:15,325 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-04-27 16:35:19,168 INFO [finetune.py:976] (1/7) Epoch 22, batch 800, loss[loss=0.1341, simple_loss=0.2177, pruned_loss=0.02524, over 4778.00 frames. ], tot_loss[loss=0.174, simple_loss=0.2465, pruned_loss=0.05074, over 937049.04 frames. ], batch size: 29, lr: 3.16e-03, grad_scale: 16.0 2023-04-27 16:35:52,468 INFO [finetune.py:976] (1/7) Epoch 22, batch 850, loss[loss=0.21, simple_loss=0.2691, pruned_loss=0.07545, over 4250.00 frames. ], tot_loss[loss=0.1721, simple_loss=0.2445, pruned_loss=0.04986, over 940022.98 frames. ], batch size: 66, lr: 3.16e-03, grad_scale: 16.0 2023-04-27 16:36:06,608 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.67 vs. limit=2.0 2023-04-27 16:36:08,869 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0807, 1.3932, 1.6184, 2.3824, 2.4415, 1.8710, 1.5672, 2.0415], device='cuda:1'), covar=tensor([0.0810, 0.1637, 0.1013, 0.0613, 0.0603, 0.1023, 0.0937, 0.0673], device='cuda:1'), in_proj_covar=tensor([0.0185, 0.0200, 0.0182, 0.0173, 0.0176, 0.0178, 0.0151, 0.0175], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 16:36:11,704 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.009e+02 1.493e+02 1.782e+02 2.235e+02 3.771e+02, threshold=3.565e+02, percent-clipped=1.0 2023-04-27 16:36:12,465 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=121164.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:36:14,913 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5607, 1.4808, 1.8199, 1.8137, 1.4151, 1.3053, 1.6333, 1.0287], device='cuda:1'), covar=tensor([0.0464, 0.0575, 0.0352, 0.0639, 0.0728, 0.1045, 0.0480, 0.0541], device='cuda:1'), in_proj_covar=tensor([0.0068, 0.0068, 0.0066, 0.0067, 0.0074, 0.0096, 0.0072, 0.0064], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-27 16:36:25,801 INFO [finetune.py:976] (1/7) Epoch 22, batch 900, loss[loss=0.1514, simple_loss=0.223, pruned_loss=0.03989, over 4822.00 frames. ], tot_loss[loss=0.171, simple_loss=0.2428, pruned_loss=0.04958, over 945219.88 frames. ], batch size: 30, lr: 3.16e-03, grad_scale: 16.0 2023-04-27 16:36:30,835 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-04-27 16:36:38,753 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=3.59 vs. limit=5.0 2023-04-27 16:36:52,990 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=121225.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:36:59,098 INFO [finetune.py:976] (1/7) Epoch 22, batch 950, loss[loss=0.1632, simple_loss=0.2315, pruned_loss=0.04742, over 4897.00 frames. ], tot_loss[loss=0.1695, simple_loss=0.2408, pruned_loss=0.04906, over 949140.27 frames. ], batch size: 46, lr: 3.16e-03, grad_scale: 16.0 2023-04-27 16:37:04,032 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8431, 2.4899, 1.7500, 1.9351, 1.5355, 1.4657, 1.8444, 1.4632], device='cuda:1'), covar=tensor([0.1334, 0.0997, 0.1367, 0.1346, 0.2034, 0.1795, 0.0795, 0.1724], device='cuda:1'), in_proj_covar=tensor([0.0195, 0.0208, 0.0167, 0.0201, 0.0198, 0.0183, 0.0154, 0.0184], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-27 16:37:35,638 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 8.216e+01 1.432e+02 1.893e+02 2.215e+02 5.796e+02, threshold=3.785e+02, percent-clipped=6.0 2023-04-27 16:37:37,510 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.2858, 2.1264, 2.3620, 2.6387, 2.7736, 2.1891, 1.8250, 2.3959], device='cuda:1'), covar=tensor([0.0707, 0.0863, 0.0580, 0.0556, 0.0500, 0.0779, 0.0804, 0.0480], device='cuda:1'), in_proj_covar=tensor([0.0184, 0.0199, 0.0181, 0.0172, 0.0175, 0.0177, 0.0150, 0.0175], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 16:38:01,690 INFO [finetune.py:976] (1/7) Epoch 22, batch 1000, loss[loss=0.2016, simple_loss=0.2653, pruned_loss=0.06897, over 4914.00 frames. ], tot_loss[loss=0.1718, simple_loss=0.2431, pruned_loss=0.05022, over 951141.94 frames. ], batch size: 35, lr: 3.16e-03, grad_scale: 16.0 2023-04-27 16:38:41,525 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=121311.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 16:39:01,159 INFO [finetune.py:976] (1/7) Epoch 22, batch 1050, loss[loss=0.1783, simple_loss=0.2539, pruned_loss=0.05133, over 4868.00 frames. ], tot_loss[loss=0.1745, simple_loss=0.2462, pruned_loss=0.05142, over 951525.30 frames. ], batch size: 31, lr: 3.16e-03, grad_scale: 16.0 2023-04-27 16:39:01,261 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=121332.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:39:01,449 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.43 vs. limit=5.0 2023-04-27 16:39:15,527 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.3787, 1.6249, 1.3992, 1.6192, 1.3385, 1.4523, 1.4948, 1.0706], device='cuda:1'), covar=tensor([0.1480, 0.1009, 0.0859, 0.0973, 0.3165, 0.0995, 0.1436, 0.2073], device='cuda:1'), in_proj_covar=tensor([0.0287, 0.0301, 0.0217, 0.0278, 0.0316, 0.0256, 0.0250, 0.0264], device='cuda:1'), out_proj_covar=tensor([1.1489e-04, 1.1901e-04, 8.5694e-05, 1.0992e-04, 1.2787e-04, 1.0127e-04, 1.0077e-04, 1.0434e-04], device='cuda:1') 2023-04-27 16:39:17,953 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.2445, 4.4657, 1.0185, 2.2383, 2.6187, 2.9113, 2.6563, 1.0423], device='cuda:1'), covar=tensor([0.1198, 0.0950, 0.2073, 0.1285, 0.0967, 0.1090, 0.1369, 0.2065], device='cuda:1'), in_proj_covar=tensor([0.0118, 0.0241, 0.0138, 0.0120, 0.0133, 0.0153, 0.0117, 0.0119], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-27 16:39:18,520 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=121359.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:39:20,858 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.274e+02 1.706e+02 2.101e+02 2.455e+02 4.347e+02, threshold=4.202e+02, percent-clipped=2.0 2023-04-27 16:39:22,165 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7486, 2.0142, 0.9905, 1.4448, 2.1789, 1.5648, 1.5121, 1.6284], device='cuda:1'), covar=tensor([0.0467, 0.0329, 0.0320, 0.0522, 0.0268, 0.0486, 0.0467, 0.0519], device='cuda:1'), in_proj_covar=tensor([0.0028, 0.0024, 0.0022, 0.0028, 0.0019, 0.0028, 0.0028, 0.0028], device='cuda:1'), out_proj_covar=tensor([0.0051, 0.0045, 0.0038, 0.0052, 0.0038, 0.0050, 0.0050, 0.0051], device='cuda:1') 2023-04-27 16:39:24,057 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7304, 1.2142, 1.8299, 2.1995, 1.8297, 1.6613, 1.7560, 1.7013], device='cuda:1'), covar=tensor([0.4483, 0.6988, 0.5892, 0.5248, 0.5643, 0.8062, 0.7883, 0.9441], device='cuda:1'), in_proj_covar=tensor([0.0430, 0.0413, 0.0506, 0.0506, 0.0459, 0.0489, 0.0495, 0.0504], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 16:39:26,936 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=121372.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 16:39:31,738 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=121380.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:39:33,384 INFO [finetune.py:976] (1/7) Epoch 22, batch 1100, loss[loss=0.1552, simple_loss=0.228, pruned_loss=0.04114, over 4753.00 frames. ], tot_loss[loss=0.176, simple_loss=0.2481, pruned_loss=0.052, over 951908.50 frames. ], batch size: 27, lr: 3.16e-03, grad_scale: 16.0 2023-04-27 16:39:42,215 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6659, 1.7871, 0.8533, 1.3432, 1.7568, 1.5182, 1.4092, 1.5161], device='cuda:1'), covar=tensor([0.0495, 0.0359, 0.0344, 0.0564, 0.0273, 0.0513, 0.0499, 0.0550], device='cuda:1'), in_proj_covar=tensor([0.0028, 0.0024, 0.0022, 0.0028, 0.0019, 0.0028, 0.0028, 0.0028], device='cuda:1'), out_proj_covar=tensor([0.0051, 0.0045, 0.0038, 0.0052, 0.0038, 0.0049, 0.0050, 0.0051], device='cuda:1') 2023-04-27 16:39:45,928 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5912, 0.9855, 1.6468, 2.0370, 1.6434, 1.5338, 1.6121, 1.5836], device='cuda:1'), covar=tensor([0.4346, 0.6515, 0.5862, 0.5610, 0.5892, 0.7508, 0.7357, 0.9042], device='cuda:1'), in_proj_covar=tensor([0.0430, 0.0414, 0.0506, 0.0507, 0.0459, 0.0489, 0.0495, 0.0504], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 16:39:47,851 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0602, 2.6027, 1.0921, 1.3683, 2.0841, 1.1013, 3.4631, 1.9215], device='cuda:1'), covar=tensor([0.0708, 0.0711, 0.0817, 0.1262, 0.0499, 0.1094, 0.0267, 0.0580], device='cuda:1'), in_proj_covar=tensor([0.0050, 0.0064, 0.0047, 0.0046, 0.0049, 0.0052, 0.0073, 0.0050], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0007], device='cuda:1') 2023-04-27 16:39:50,921 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=121407.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:39:52,699 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.2446, 1.3066, 1.3711, 1.5618, 1.5423, 1.2691, 0.9846, 1.4319], device='cuda:1'), covar=tensor([0.0900, 0.1255, 0.0990, 0.0700, 0.0783, 0.0940, 0.0928, 0.0732], device='cuda:1'), in_proj_covar=tensor([0.0185, 0.0199, 0.0182, 0.0172, 0.0176, 0.0178, 0.0151, 0.0175], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 16:40:06,342 INFO [finetune.py:976] (1/7) Epoch 22, batch 1150, loss[loss=0.1686, simple_loss=0.2478, pruned_loss=0.04472, over 4731.00 frames. ], tot_loss[loss=0.1764, simple_loss=0.2487, pruned_loss=0.05203, over 952628.73 frames. ], batch size: 54, lr: 3.16e-03, grad_scale: 16.0 2023-04-27 16:40:27,716 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.115e+02 1.622e+02 1.904e+02 2.379e+02 4.461e+02, threshold=3.808e+02, percent-clipped=1.0 2023-04-27 16:40:27,985 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.74 vs. limit=2.0 2023-04-27 16:40:30,881 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=121468.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:40:39,780 INFO [finetune.py:976] (1/7) Epoch 22, batch 1200, loss[loss=0.1655, simple_loss=0.2413, pruned_loss=0.04488, over 4926.00 frames. ], tot_loss[loss=0.1745, simple_loss=0.2468, pruned_loss=0.05112, over 953637.06 frames. ], batch size: 33, lr: 3.15e-03, grad_scale: 16.0 2023-04-27 16:41:05,061 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=121520.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:41:12,819 INFO [finetune.py:976] (1/7) Epoch 22, batch 1250, loss[loss=0.1291, simple_loss=0.1998, pruned_loss=0.02924, over 4907.00 frames. ], tot_loss[loss=0.1724, simple_loss=0.2444, pruned_loss=0.05022, over 953414.83 frames. ], batch size: 32, lr: 3.15e-03, grad_scale: 32.0 2023-04-27 16:41:34,045 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.134e+02 1.460e+02 1.755e+02 2.215e+02 4.561e+02, threshold=3.510e+02, percent-clipped=2.0 2023-04-27 16:41:46,167 INFO [finetune.py:976] (1/7) Epoch 22, batch 1300, loss[loss=0.1576, simple_loss=0.2262, pruned_loss=0.04447, over 4915.00 frames. ], tot_loss[loss=0.1696, simple_loss=0.2409, pruned_loss=0.04918, over 953837.69 frames. ], batch size: 43, lr: 3.15e-03, grad_scale: 32.0 2023-04-27 16:42:19,072 INFO [finetune.py:976] (1/7) Epoch 22, batch 1350, loss[loss=0.1723, simple_loss=0.2451, pruned_loss=0.04979, over 4835.00 frames. ], tot_loss[loss=0.1694, simple_loss=0.2407, pruned_loss=0.04902, over 955219.20 frames. ], batch size: 47, lr: 3.15e-03, grad_scale: 32.0 2023-04-27 16:42:32,456 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([3.8003, 3.7146, 2.7403, 4.3884, 3.8052, 3.7938, 1.5575, 3.7153], device='cuda:1'), covar=tensor([0.1691, 0.1346, 0.3166, 0.1697, 0.4406, 0.1823, 0.5908, 0.2242], device='cuda:1'), in_proj_covar=tensor([0.0242, 0.0214, 0.0249, 0.0303, 0.0292, 0.0244, 0.0272, 0.0270], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 16:42:55,952 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=121659.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:42:58,252 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.014e+02 1.587e+02 1.863e+02 2.182e+02 3.847e+02, threshold=3.726e+02, percent-clipped=2.0 2023-04-27 16:42:58,612 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-04-27 16:43:04,731 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=121667.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 16:43:19,351 INFO [finetune.py:976] (1/7) Epoch 22, batch 1400, loss[loss=0.1636, simple_loss=0.2518, pruned_loss=0.0377, over 4850.00 frames. ], tot_loss[loss=0.1706, simple_loss=0.2429, pruned_loss=0.04916, over 956335.29 frames. ], batch size: 44, lr: 3.15e-03, grad_scale: 32.0 2023-04-27 16:43:37,355 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.3226, 1.5612, 1.5845, 2.2064, 1.8145, 2.1450, 1.6395, 4.5168], device='cuda:1'), covar=tensor([0.0722, 0.1103, 0.1035, 0.1254, 0.0809, 0.0590, 0.0965, 0.0168], device='cuda:1'), in_proj_covar=tensor([0.0037, 0.0038, 0.0039, 0.0043, 0.0040, 0.0038, 0.0038, 0.0055], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:1') 2023-04-27 16:43:59,850 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=121707.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:44:03,632 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0322, 1.7839, 2.2510, 2.3965, 2.0657, 1.9628, 2.1389, 2.0733], device='cuda:1'), covar=tensor([0.4651, 0.6627, 0.6980, 0.5680, 0.5843, 0.8844, 0.8223, 0.9414], device='cuda:1'), in_proj_covar=tensor([0.0430, 0.0413, 0.0505, 0.0505, 0.0457, 0.0489, 0.0494, 0.0502], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 16:44:24,400 INFO [finetune.py:976] (1/7) Epoch 22, batch 1450, loss[loss=0.1612, simple_loss=0.2439, pruned_loss=0.03921, over 4768.00 frames. ], tot_loss[loss=0.1708, simple_loss=0.2439, pruned_loss=0.0488, over 956370.44 frames. ], batch size: 54, lr: 3.15e-03, grad_scale: 32.0 2023-04-27 16:44:44,670 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([0.9188, 1.8496, 1.6703, 1.4610, 1.9114, 1.6522, 2.3049, 1.4572], device='cuda:1'), covar=tensor([0.3275, 0.1601, 0.4144, 0.2639, 0.1409, 0.1984, 0.1423, 0.4430], device='cuda:1'), in_proj_covar=tensor([0.0339, 0.0348, 0.0423, 0.0350, 0.0379, 0.0373, 0.0365, 0.0417], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 16:45:04,293 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.075e+02 1.748e+02 1.956e+02 2.570e+02 3.972e+02, threshold=3.912e+02, percent-clipped=2.0 2023-04-27 16:45:04,381 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=121763.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:45:16,406 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4084, 1.1002, 3.8978, 3.3818, 3.4338, 3.5990, 3.5157, 3.3044], device='cuda:1'), covar=tensor([0.9414, 0.8487, 0.1808, 0.3089, 0.2224, 0.3245, 0.4294, 0.3377], device='cuda:1'), in_proj_covar=tensor([0.0309, 0.0306, 0.0405, 0.0405, 0.0346, 0.0408, 0.0310, 0.0364], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-27 16:45:27,548 INFO [finetune.py:976] (1/7) Epoch 22, batch 1500, loss[loss=0.1792, simple_loss=0.2524, pruned_loss=0.05298, over 4891.00 frames. ], tot_loss[loss=0.1724, simple_loss=0.2451, pruned_loss=0.04979, over 954345.92 frames. ], batch size: 35, lr: 3.15e-03, grad_scale: 32.0 2023-04-27 16:45:43,063 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8897, 2.4711, 1.8994, 1.7122, 1.3304, 1.3399, 2.0438, 1.3627], device='cuda:1'), covar=tensor([0.1595, 0.1196, 0.1320, 0.1631, 0.2262, 0.1841, 0.0880, 0.1906], device='cuda:1'), in_proj_covar=tensor([0.0197, 0.0211, 0.0169, 0.0203, 0.0200, 0.0185, 0.0155, 0.0186], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-27 16:45:58,962 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=121820.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:46:06,250 INFO [finetune.py:976] (1/7) Epoch 22, batch 1550, loss[loss=0.134, simple_loss=0.2025, pruned_loss=0.03272, over 4723.00 frames. ], tot_loss[loss=0.1729, simple_loss=0.2456, pruned_loss=0.05012, over 953159.50 frames. ], batch size: 23, lr: 3.15e-03, grad_scale: 32.0 2023-04-27 16:46:10,082 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.23 vs. limit=5.0 2023-04-27 16:46:28,048 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.267e+01 1.607e+02 1.890e+02 2.272e+02 5.935e+02, threshold=3.780e+02, percent-clipped=2.0 2023-04-27 16:46:31,191 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=121868.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:46:39,810 INFO [finetune.py:976] (1/7) Epoch 22, batch 1600, loss[loss=0.1821, simple_loss=0.2502, pruned_loss=0.05696, over 4675.00 frames. ], tot_loss[loss=0.1708, simple_loss=0.243, pruned_loss=0.04937, over 952852.95 frames. ], batch size: 23, lr: 3.15e-03, grad_scale: 32.0 2023-04-27 16:47:09,573 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5961, 1.8638, 1.8528, 2.2243, 2.1293, 2.1549, 1.7590, 4.6255], device='cuda:1'), covar=tensor([0.0507, 0.0768, 0.0725, 0.1097, 0.0569, 0.0488, 0.0691, 0.0108], device='cuda:1'), in_proj_covar=tensor([0.0038, 0.0038, 0.0039, 0.0043, 0.0040, 0.0038, 0.0038, 0.0055], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:1') 2023-04-27 16:47:13,760 INFO [finetune.py:976] (1/7) Epoch 22, batch 1650, loss[loss=0.122, simple_loss=0.2051, pruned_loss=0.01944, over 4791.00 frames. ], tot_loss[loss=0.1697, simple_loss=0.2413, pruned_loss=0.04907, over 953846.62 frames. ], batch size: 51, lr: 3.15e-03, grad_scale: 32.0 2023-04-27 16:47:24,845 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=121950.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:47:33,996 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.216e+02 1.538e+02 1.825e+02 2.150e+02 3.786e+02, threshold=3.649e+02, percent-clipped=1.0 2023-04-27 16:47:37,494 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=121967.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 16:47:47,186 INFO [finetune.py:976] (1/7) Epoch 22, batch 1700, loss[loss=0.1872, simple_loss=0.2613, pruned_loss=0.05651, over 4852.00 frames. ], tot_loss[loss=0.1674, simple_loss=0.2387, pruned_loss=0.048, over 955499.07 frames. ], batch size: 49, lr: 3.15e-03, grad_scale: 32.0 2023-04-27 16:47:47,324 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.6452, 2.0161, 2.5600, 3.1647, 2.4612, 1.9840, 1.9550, 2.4527], device='cuda:1'), covar=tensor([0.2952, 0.3067, 0.1448, 0.2196, 0.2522, 0.2500, 0.3621, 0.1786], device='cuda:1'), in_proj_covar=tensor([0.0292, 0.0243, 0.0226, 0.0313, 0.0220, 0.0232, 0.0227, 0.0183], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-27 16:48:06,898 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=122011.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:48:16,071 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=122015.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 16:48:37,959 INFO [finetune.py:976] (1/7) Epoch 22, batch 1750, loss[loss=0.1744, simple_loss=0.274, pruned_loss=0.03744, over 4814.00 frames. ], tot_loss[loss=0.1686, simple_loss=0.2404, pruned_loss=0.0484, over 955182.25 frames. ], batch size: 41, lr: 3.15e-03, grad_scale: 32.0 2023-04-27 16:48:52,474 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4284, 3.0447, 0.9274, 1.6012, 1.7883, 2.1364, 1.8171, 0.9796], device='cuda:1'), covar=tensor([0.1346, 0.0951, 0.1908, 0.1317, 0.1086, 0.1037, 0.1544, 0.1828], device='cuda:1'), in_proj_covar=tensor([0.0117, 0.0240, 0.0138, 0.0120, 0.0133, 0.0152, 0.0116, 0.0119], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-27 16:49:11,592 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8350, 2.2210, 0.9848, 1.2382, 1.6410, 1.2004, 2.4932, 1.3719], device='cuda:1'), covar=tensor([0.0687, 0.0570, 0.0609, 0.1206, 0.0418, 0.0974, 0.0308, 0.0680], device='cuda:1'), in_proj_covar=tensor([0.0050, 0.0064, 0.0047, 0.0046, 0.0049, 0.0052, 0.0073, 0.0050], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0007], device='cuda:1') 2023-04-27 16:49:13,878 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.158e+02 1.616e+02 1.845e+02 2.409e+02 4.396e+02, threshold=3.689e+02, percent-clipped=3.0 2023-04-27 16:49:13,976 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=122063.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:49:14,067 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=3.76 vs. limit=5.0 2023-04-27 16:49:43,438 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 2023-04-27 16:49:43,797 INFO [finetune.py:976] (1/7) Epoch 22, batch 1800, loss[loss=0.1888, simple_loss=0.2608, pruned_loss=0.05838, over 4883.00 frames. ], tot_loss[loss=0.1698, simple_loss=0.2419, pruned_loss=0.0488, over 951646.43 frames. ], batch size: 32, lr: 3.15e-03, grad_scale: 32.0 2023-04-27 16:49:57,706 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9707, 1.1660, 1.6584, 1.8006, 1.7817, 1.8206, 1.6682, 1.6661], device='cuda:1'), covar=tensor([0.3867, 0.5118, 0.4261, 0.4364, 0.5159, 0.7033, 0.4704, 0.4597], device='cuda:1'), in_proj_covar=tensor([0.0337, 0.0372, 0.0323, 0.0337, 0.0345, 0.0392, 0.0355, 0.0329], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 16:50:06,238 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.54 vs. limit=2.0 2023-04-27 16:50:18,106 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=122111.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:50:27,677 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=122117.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:50:48,288 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=2.05 vs. limit=2.0 2023-04-27 16:50:49,859 INFO [finetune.py:976] (1/7) Epoch 22, batch 1850, loss[loss=0.2259, simple_loss=0.2926, pruned_loss=0.07962, over 4903.00 frames. ], tot_loss[loss=0.1717, simple_loss=0.2442, pruned_loss=0.04958, over 954288.75 frames. ], batch size: 37, lr: 3.15e-03, grad_scale: 32.0 2023-04-27 16:51:14,578 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-04-27 16:51:22,417 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.1488, 1.3459, 1.2582, 1.6450, 1.4645, 1.5633, 1.2881, 2.4437], device='cuda:1'), covar=tensor([0.0601, 0.0888, 0.0848, 0.1294, 0.0703, 0.0509, 0.0802, 0.0226], device='cuda:1'), in_proj_covar=tensor([0.0038, 0.0039, 0.0039, 0.0043, 0.0040, 0.0038, 0.0038, 0.0055], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0013, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:1') 2023-04-27 16:51:27,102 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.233e+01 1.753e+02 1.978e+02 2.407e+02 4.460e+02, threshold=3.956e+02, percent-clipped=4.0 2023-04-27 16:51:48,054 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=122178.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 16:51:56,333 INFO [finetune.py:976] (1/7) Epoch 22, batch 1900, loss[loss=0.1589, simple_loss=0.2368, pruned_loss=0.04051, over 4735.00 frames. ], tot_loss[loss=0.172, simple_loss=0.245, pruned_loss=0.0495, over 954576.97 frames. ], batch size: 54, lr: 3.15e-03, grad_scale: 32.0 2023-04-27 16:52:10,111 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=122195.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:52:20,610 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.9191, 3.4363, 2.9296, 3.4652, 2.6123, 3.1039, 3.2154, 2.5837], device='cuda:1'), covar=tensor([0.1585, 0.1066, 0.0701, 0.0920, 0.2533, 0.0945, 0.1408, 0.2299], device='cuda:1'), in_proj_covar=tensor([0.0290, 0.0306, 0.0220, 0.0281, 0.0319, 0.0260, 0.0253, 0.0267], device='cuda:1'), out_proj_covar=tensor([1.1593e-04, 1.2117e-04, 8.6970e-05, 1.1113e-04, 1.2943e-04, 1.0274e-04, 1.0187e-04, 1.0564e-04], device='cuda:1') 2023-04-27 16:52:27,904 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-04-27 16:52:40,533 INFO [finetune.py:976] (1/7) Epoch 22, batch 1950, loss[loss=0.1819, simple_loss=0.2617, pruned_loss=0.05108, over 4920.00 frames. ], tot_loss[loss=0.1722, simple_loss=0.2453, pruned_loss=0.04957, over 955409.91 frames. ], batch size: 46, lr: 3.15e-03, grad_scale: 32.0 2023-04-27 16:52:49,305 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-04-27 16:52:55,258 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=122256.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:52:58,859 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.3988, 2.9523, 0.9182, 1.5097, 2.4932, 1.4600, 4.2541, 1.9156], device='cuda:1'), covar=tensor([0.0815, 0.0755, 0.1045, 0.1739, 0.0590, 0.1360, 0.0279, 0.0840], device='cuda:1'), in_proj_covar=tensor([0.0050, 0.0064, 0.0047, 0.0046, 0.0049, 0.0051, 0.0073, 0.0050], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0009, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0007], device='cuda:1') 2023-04-27 16:52:59,364 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.104e+02 1.614e+02 1.828e+02 2.204e+02 4.650e+02, threshold=3.656e+02, percent-clipped=1.0 2023-04-27 16:53:13,159 INFO [finetune.py:976] (1/7) Epoch 22, batch 2000, loss[loss=0.1629, simple_loss=0.2379, pruned_loss=0.04398, over 4919.00 frames. ], tot_loss[loss=0.1699, simple_loss=0.2425, pruned_loss=0.0486, over 957540.57 frames. ], batch size: 37, lr: 3.15e-03, grad_scale: 32.0 2023-04-27 16:53:29,024 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=122306.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:53:32,202 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.60 vs. limit=2.0 2023-04-27 16:53:37,056 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=122319.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:53:46,289 INFO [finetune.py:976] (1/7) Epoch 22, batch 2050, loss[loss=0.1365, simple_loss=0.2099, pruned_loss=0.03149, over 4931.00 frames. ], tot_loss[loss=0.1675, simple_loss=0.239, pruned_loss=0.04796, over 954291.99 frames. ], batch size: 33, lr: 3.15e-03, grad_scale: 32.0 2023-04-27 16:53:52,824 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.2543, 1.8657, 2.1437, 2.6723, 2.4398, 2.2289, 1.9242, 2.2396], device='cuda:1'), covar=tensor([0.0751, 0.1117, 0.0697, 0.0507, 0.0618, 0.0746, 0.0703, 0.0580], device='cuda:1'), in_proj_covar=tensor([0.0185, 0.0200, 0.0182, 0.0173, 0.0176, 0.0178, 0.0151, 0.0176], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 16:54:02,525 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.3906, 2.0642, 2.4513, 2.8467, 2.6775, 2.3820, 2.0384, 2.3939], device='cuda:1'), covar=tensor([0.0808, 0.1043, 0.0656, 0.0591, 0.0641, 0.0767, 0.0769, 0.0575], device='cuda:1'), in_proj_covar=tensor([0.0185, 0.0200, 0.0182, 0.0173, 0.0176, 0.0178, 0.0150, 0.0176], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 16:54:12,098 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.067e+02 1.559e+02 1.860e+02 2.262e+02 4.767e+02, threshold=3.720e+02, percent-clipped=3.0 2023-04-27 16:54:21,790 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.1252, 1.3186, 1.2438, 1.6032, 1.4221, 1.5799, 1.2507, 2.3316], device='cuda:1'), covar=tensor([0.0559, 0.0751, 0.0755, 0.1112, 0.0580, 0.0467, 0.0704, 0.0216], device='cuda:1'), in_proj_covar=tensor([0.0038, 0.0039, 0.0039, 0.0043, 0.0040, 0.0038, 0.0038, 0.0055], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0013, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:1') 2023-04-27 16:54:23,011 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4699, 1.8784, 1.7713, 2.0552, 2.0347, 2.0490, 1.7134, 3.6722], device='cuda:1'), covar=tensor([0.0524, 0.0647, 0.0661, 0.0977, 0.0499, 0.0779, 0.0663, 0.0182], device='cuda:1'), in_proj_covar=tensor([0.0038, 0.0039, 0.0039, 0.0043, 0.0040, 0.0038, 0.0038, 0.0055], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0013, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:1') 2023-04-27 16:54:35,040 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=122380.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:54:36,179 INFO [finetune.py:976] (1/7) Epoch 22, batch 2100, loss[loss=0.1938, simple_loss=0.2545, pruned_loss=0.06652, over 4915.00 frames. ], tot_loss[loss=0.1672, simple_loss=0.2387, pruned_loss=0.04785, over 953285.31 frames. ], batch size: 37, lr: 3.15e-03, grad_scale: 32.0 2023-04-27 16:55:40,951 INFO [finetune.py:976] (1/7) Epoch 22, batch 2150, loss[loss=0.2097, simple_loss=0.2819, pruned_loss=0.06879, over 4805.00 frames. ], tot_loss[loss=0.1693, simple_loss=0.2412, pruned_loss=0.04868, over 953870.03 frames. ], batch size: 41, lr: 3.15e-03, grad_scale: 32.0 2023-04-27 16:56:19,329 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.125e+02 1.667e+02 1.982e+02 2.424e+02 1.094e+03, threshold=3.964e+02, percent-clipped=3.0 2023-04-27 16:56:22,540 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=122468.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:56:30,954 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=122473.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 16:56:42,041 INFO [finetune.py:976] (1/7) Epoch 22, batch 2200, loss[loss=0.1912, simple_loss=0.2687, pruned_loss=0.05682, over 4905.00 frames. ], tot_loss[loss=0.171, simple_loss=0.2435, pruned_loss=0.04924, over 955633.33 frames. ], batch size: 43, lr: 3.15e-03, grad_scale: 32.0 2023-04-27 16:57:23,359 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.3617, 1.6778, 1.8469, 1.9501, 1.7541, 1.8490, 1.8710, 1.8530], device='cuda:1'), covar=tensor([0.3820, 0.5228, 0.4128, 0.4382, 0.5370, 0.7003, 0.4894, 0.4590], device='cuda:1'), in_proj_covar=tensor([0.0340, 0.0373, 0.0325, 0.0340, 0.0348, 0.0395, 0.0357, 0.0331], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 16:57:48,039 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=122529.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:57:49,741 INFO [finetune.py:976] (1/7) Epoch 22, batch 2250, loss[loss=0.1831, simple_loss=0.263, pruned_loss=0.05162, over 4898.00 frames. ], tot_loss[loss=0.1717, simple_loss=0.2443, pruned_loss=0.04958, over 955329.49 frames. ], batch size: 37, lr: 3.15e-03, grad_scale: 32.0 2023-04-27 16:58:19,792 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=122551.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:58:34,012 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.195e+02 1.620e+02 1.922e+02 2.285e+02 4.968e+02, threshold=3.845e+02, percent-clipped=1.0 2023-04-27 16:58:54,123 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8464, 1.4851, 1.5165, 1.5554, 2.0582, 1.6824, 1.3308, 1.3590], device='cuda:1'), covar=tensor([0.2014, 0.1540, 0.2369, 0.1901, 0.0996, 0.1855, 0.2969, 0.2740], device='cuda:1'), in_proj_covar=tensor([0.0313, 0.0311, 0.0351, 0.0290, 0.0326, 0.0309, 0.0301, 0.0372], device='cuda:1'), out_proj_covar=tensor([6.4234e-05, 6.4382e-05, 7.3854e-05, 5.8343e-05, 6.7039e-05, 6.4790e-05, 6.2916e-05, 7.9019e-05], device='cuda:1') 2023-04-27 16:58:54,201 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.62 vs. limit=5.0 2023-04-27 16:58:56,472 INFO [finetune.py:976] (1/7) Epoch 22, batch 2300, loss[loss=0.1294, simple_loss=0.2124, pruned_loss=0.02321, over 4743.00 frames. ], tot_loss[loss=0.1709, simple_loss=0.2443, pruned_loss=0.04876, over 955856.08 frames. ], batch size: 27, lr: 3.15e-03, grad_scale: 32.0 2023-04-27 16:59:36,002 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=122606.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:59:46,824 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-04-27 17:00:08,779 INFO [finetune.py:976] (1/7) Epoch 22, batch 2350, loss[loss=0.1553, simple_loss=0.2312, pruned_loss=0.03973, over 4815.00 frames. ], tot_loss[loss=0.1698, simple_loss=0.2429, pruned_loss=0.04832, over 956263.52 frames. ], batch size: 39, lr: 3.15e-03, grad_scale: 32.0 2023-04-27 17:00:40,760 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=122654.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:00:46,226 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.405e+01 1.589e+02 1.908e+02 2.329e+02 5.605e+02, threshold=3.816e+02, percent-clipped=4.0 2023-04-27 17:01:04,134 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=122675.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:01:14,124 INFO [finetune.py:976] (1/7) Epoch 22, batch 2400, loss[loss=0.1808, simple_loss=0.2382, pruned_loss=0.06163, over 4869.00 frames. ], tot_loss[loss=0.1683, simple_loss=0.2404, pruned_loss=0.04812, over 956104.88 frames. ], batch size: 34, lr: 3.15e-03, grad_scale: 32.0 2023-04-27 17:02:21,392 INFO [finetune.py:976] (1/7) Epoch 22, batch 2450, loss[loss=0.1626, simple_loss=0.2365, pruned_loss=0.04438, over 4830.00 frames. ], tot_loss[loss=0.1674, simple_loss=0.2385, pruned_loss=0.04814, over 955042.80 frames. ], batch size: 33, lr: 3.14e-03, grad_scale: 32.0 2023-04-27 17:02:42,828 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.071e+02 1.578e+02 1.755e+02 2.241e+02 3.458e+02, threshold=3.510e+02, percent-clipped=0.0 2023-04-27 17:02:48,952 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=122773.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 17:02:53,214 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([3.4311, 3.2604, 2.4761, 3.9026, 3.3897, 3.3687, 1.4658, 3.3377], device='cuda:1'), covar=tensor([0.1927, 0.1533, 0.3488, 0.2374, 0.3382, 0.2005, 0.5643, 0.2711], device='cuda:1'), in_proj_covar=tensor([0.0244, 0.0217, 0.0252, 0.0306, 0.0295, 0.0247, 0.0276, 0.0273], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 17:02:54,346 INFO [finetune.py:976] (1/7) Epoch 22, batch 2500, loss[loss=0.1684, simple_loss=0.2495, pruned_loss=0.04364, over 4840.00 frames. ], tot_loss[loss=0.1693, simple_loss=0.2404, pruned_loss=0.04908, over 956142.89 frames. ], batch size: 49, lr: 3.14e-03, grad_scale: 32.0 2023-04-27 17:03:21,716 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=122821.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:03:23,546 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=122824.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:03:28,387 INFO [finetune.py:976] (1/7) Epoch 22, batch 2550, loss[loss=0.1454, simple_loss=0.2301, pruned_loss=0.03031, over 4933.00 frames. ], tot_loss[loss=0.1723, simple_loss=0.2445, pruned_loss=0.05, over 957909.64 frames. ], batch size: 38, lr: 3.14e-03, grad_scale: 32.0 2023-04-27 17:03:40,086 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=122851.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:03:48,775 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.124e+02 1.572e+02 1.988e+02 2.306e+02 4.105e+02, threshold=3.976e+02, percent-clipped=1.0 2023-04-27 17:04:01,383 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=122872.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:04:01,993 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0002, 2.3890, 1.0820, 1.3146, 2.0222, 1.2281, 2.9486, 1.5378], device='cuda:1'), covar=tensor([0.0718, 0.0573, 0.0710, 0.1251, 0.0427, 0.0986, 0.0299, 0.0667], device='cuda:1'), in_proj_covar=tensor([0.0050, 0.0064, 0.0047, 0.0046, 0.0049, 0.0051, 0.0073, 0.0050], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0009, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:1') 2023-04-27 17:04:13,224 INFO [finetune.py:976] (1/7) Epoch 22, batch 2600, loss[loss=0.1304, simple_loss=0.2076, pruned_loss=0.02664, over 4767.00 frames. ], tot_loss[loss=0.1727, simple_loss=0.2454, pruned_loss=0.05001, over 957118.98 frames. ], batch size: 28, lr: 3.14e-03, grad_scale: 32.0 2023-04-27 17:04:34,928 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=122899.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:05:08,452 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.2119, 1.5719, 2.0316, 2.4859, 2.1001, 1.6187, 1.3951, 1.9307], device='cuda:1'), covar=tensor([0.3336, 0.3484, 0.1891, 0.2335, 0.2606, 0.2935, 0.4180, 0.1989], device='cuda:1'), in_proj_covar=tensor([0.0292, 0.0244, 0.0226, 0.0313, 0.0220, 0.0232, 0.0227, 0.0183], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-27 17:05:19,783 INFO [finetune.py:976] (1/7) Epoch 22, batch 2650, loss[loss=0.1541, simple_loss=0.2205, pruned_loss=0.04387, over 4712.00 frames. ], tot_loss[loss=0.1738, simple_loss=0.247, pruned_loss=0.05031, over 956975.86 frames. ], batch size: 23, lr: 3.14e-03, grad_scale: 32.0 2023-04-27 17:05:20,513 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=122933.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:05:30,525 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.79 vs. limit=2.0 2023-04-27 17:06:00,531 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.091e+02 1.522e+02 1.838e+02 2.219e+02 4.134e+02, threshold=3.677e+02, percent-clipped=1.0 2023-04-27 17:06:15,879 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=122975.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:06:16,628 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-04-27 17:06:26,129 INFO [finetune.py:976] (1/7) Epoch 22, batch 2700, loss[loss=0.1553, simple_loss=0.2388, pruned_loss=0.03594, over 4899.00 frames. ], tot_loss[loss=0.1719, simple_loss=0.2457, pruned_loss=0.0491, over 957327.99 frames. ], batch size: 35, lr: 3.14e-03, grad_scale: 32.0 2023-04-27 17:06:28,217 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.63 vs. limit=2.0 2023-04-27 17:06:42,440 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([0.4308, 1.2754, 1.3771, 0.9894, 1.3030, 1.1508, 1.6875, 1.3762], device='cuda:1'), covar=tensor([0.3656, 0.2060, 0.5263, 0.2858, 0.1746, 0.2325, 0.1754, 0.4832], device='cuda:1'), in_proj_covar=tensor([0.0339, 0.0350, 0.0423, 0.0354, 0.0382, 0.0376, 0.0368, 0.0419], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 17:06:56,059 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=123023.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:06:56,133 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6624, 1.2994, 1.3596, 1.4077, 1.8200, 1.5041, 1.1716, 1.2524], device='cuda:1'), covar=tensor([0.1761, 0.1547, 0.1894, 0.1534, 0.0934, 0.1769, 0.2320, 0.2756], device='cuda:1'), in_proj_covar=tensor([0.0313, 0.0312, 0.0351, 0.0289, 0.0326, 0.0310, 0.0301, 0.0373], device='cuda:1'), out_proj_covar=tensor([6.4200e-05, 6.4572e-05, 7.3989e-05, 5.8243e-05, 6.7128e-05, 6.4967e-05, 6.2876e-05, 7.9191e-05], device='cuda:1') 2023-04-27 17:07:02,522 INFO [finetune.py:976] (1/7) Epoch 22, batch 2750, loss[loss=0.1552, simple_loss=0.2259, pruned_loss=0.04223, over 4925.00 frames. ], tot_loss[loss=0.1704, simple_loss=0.2433, pruned_loss=0.04877, over 956679.58 frames. ], batch size: 38, lr: 3.14e-03, grad_scale: 32.0 2023-04-27 17:07:28,110 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.026e+01 1.558e+02 1.841e+02 2.200e+02 4.189e+02, threshold=3.681e+02, percent-clipped=2.0 2023-04-27 17:07:53,155 INFO [finetune.py:976] (1/7) Epoch 22, batch 2800, loss[loss=0.174, simple_loss=0.2372, pruned_loss=0.05539, over 4847.00 frames. ], tot_loss[loss=0.1684, simple_loss=0.2404, pruned_loss=0.04815, over 956604.36 frames. ], batch size: 47, lr: 3.14e-03, grad_scale: 32.0 2023-04-27 17:08:48,176 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=123124.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:08:53,533 INFO [finetune.py:976] (1/7) Epoch 22, batch 2850, loss[loss=0.1393, simple_loss=0.2133, pruned_loss=0.03263, over 4790.00 frames. ], tot_loss[loss=0.1659, simple_loss=0.2379, pruned_loss=0.04695, over 958316.90 frames. ], batch size: 29, lr: 3.14e-03, grad_scale: 32.0 2023-04-27 17:09:12,896 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.070e+02 1.588e+02 1.846e+02 2.327e+02 3.843e+02, threshold=3.691e+02, percent-clipped=1.0 2023-04-27 17:09:19,395 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=123172.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:09:27,496 INFO [finetune.py:976] (1/7) Epoch 22, batch 2900, loss[loss=0.1758, simple_loss=0.2362, pruned_loss=0.05774, over 4762.00 frames. ], tot_loss[loss=0.1688, simple_loss=0.2412, pruned_loss=0.0482, over 957971.09 frames. ], batch size: 27, lr: 3.14e-03, grad_scale: 32.0 2023-04-27 17:09:42,974 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.2754, 2.9188, 0.8236, 1.7252, 1.7559, 2.1447, 1.7685, 0.9346], device='cuda:1'), covar=tensor([0.1327, 0.1039, 0.1885, 0.1142, 0.1059, 0.0917, 0.1360, 0.1832], device='cuda:1'), in_proj_covar=tensor([0.0116, 0.0240, 0.0137, 0.0119, 0.0132, 0.0152, 0.0117, 0.0119], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-27 17:09:48,564 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9713, 2.3566, 1.0432, 1.3187, 2.2174, 1.1417, 3.2311, 1.7007], device='cuda:1'), covar=tensor([0.0721, 0.0685, 0.0823, 0.1267, 0.0458, 0.1089, 0.0215, 0.0617], device='cuda:1'), in_proj_covar=tensor([0.0050, 0.0064, 0.0048, 0.0046, 0.0049, 0.0052, 0.0074, 0.0051], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:1') 2023-04-27 17:10:09,342 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=123228.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:10:11,740 INFO [finetune.py:976] (1/7) Epoch 22, batch 2950, loss[loss=0.1437, simple_loss=0.2285, pruned_loss=0.02943, over 4787.00 frames. ], tot_loss[loss=0.1726, simple_loss=0.2458, pruned_loss=0.04971, over 957199.44 frames. ], batch size: 29, lr: 3.14e-03, grad_scale: 32.0 2023-04-27 17:10:20,195 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-04-27 17:10:52,159 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.3375, 3.0634, 0.9573, 1.7019, 1.8590, 2.2352, 1.8019, 0.9526], device='cuda:1'), covar=tensor([0.1373, 0.0981, 0.1750, 0.1289, 0.1018, 0.0952, 0.1658, 0.1750], device='cuda:1'), in_proj_covar=tensor([0.0117, 0.0240, 0.0138, 0.0120, 0.0133, 0.0153, 0.0117, 0.0119], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-27 17:10:53,371 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5893, 3.3610, 1.0054, 1.9338, 2.1358, 2.5066, 2.0668, 1.2237], device='cuda:1'), covar=tensor([0.1262, 0.1001, 0.1817, 0.1168, 0.0876, 0.0914, 0.1314, 0.1594], device='cuda:1'), in_proj_covar=tensor([0.0117, 0.0240, 0.0138, 0.0120, 0.0133, 0.0153, 0.0117, 0.0119], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-27 17:10:55,130 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.161e+02 1.651e+02 2.036e+02 2.473e+02 7.289e+02, threshold=4.071e+02, percent-clipped=4.0 2023-04-27 17:11:18,375 INFO [finetune.py:976] (1/7) Epoch 22, batch 3000, loss[loss=0.2026, simple_loss=0.2674, pruned_loss=0.06889, over 4877.00 frames. ], tot_loss[loss=0.1729, simple_loss=0.2461, pruned_loss=0.04988, over 957024.80 frames. ], batch size: 43, lr: 3.14e-03, grad_scale: 32.0 2023-04-27 17:11:18,375 INFO [finetune.py:1001] (1/7) Computing validation loss 2023-04-27 17:11:24,734 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9932, 2.2343, 1.9467, 1.6244, 1.4621, 1.4979, 1.8913, 1.4433], device='cuda:1'), covar=tensor([0.1688, 0.1504, 0.1534, 0.1812, 0.2377, 0.2067, 0.1092, 0.2162], device='cuda:1'), in_proj_covar=tensor([0.0197, 0.0211, 0.0169, 0.0204, 0.0200, 0.0186, 0.0156, 0.0186], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-27 17:11:29,116 INFO [finetune.py:1010] (1/7) Epoch 22, validation: loss=0.1537, simple_loss=0.2227, pruned_loss=0.04237, over 2265189.00 frames. 2023-04-27 17:11:29,116 INFO [finetune.py:1011] (1/7) Maximum memory allocated so far is 6391MB 2023-04-27 17:12:11,352 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=123314.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:12:11,515 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 2023-04-27 17:12:33,515 INFO [finetune.py:976] (1/7) Epoch 22, batch 3050, loss[loss=0.1621, simple_loss=0.2405, pruned_loss=0.04187, over 4788.00 frames. ], tot_loss[loss=0.174, simple_loss=0.2474, pruned_loss=0.05028, over 956875.17 frames. ], batch size: 29, lr: 3.14e-03, grad_scale: 32.0 2023-04-27 17:12:42,339 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-04-27 17:12:59,243 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.5568, 1.7936, 1.7124, 2.3088, 2.5127, 1.9764, 1.9374, 1.7271], device='cuda:1'), covar=tensor([0.1306, 0.1658, 0.1734, 0.1245, 0.0911, 0.1855, 0.2017, 0.2262], device='cuda:1'), in_proj_covar=tensor([0.0316, 0.0315, 0.0355, 0.0291, 0.0329, 0.0313, 0.0304, 0.0377], device='cuda:1'), out_proj_covar=tensor([6.4911e-05, 6.5066e-05, 7.4855e-05, 5.8730e-05, 6.7857e-05, 6.5524e-05, 6.3455e-05, 8.0009e-05], device='cuda:1') 2023-04-27 17:13:01,590 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 8.350e+01 1.609e+02 1.855e+02 2.357e+02 3.763e+02, threshold=3.710e+02, percent-clipped=0.0 2023-04-27 17:13:14,330 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=123375.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:13:24,030 INFO [finetune.py:976] (1/7) Epoch 22, batch 3100, loss[loss=0.2023, simple_loss=0.266, pruned_loss=0.06931, over 4931.00 frames. ], tot_loss[loss=0.1728, simple_loss=0.2457, pruned_loss=0.04992, over 956687.39 frames. ], batch size: 33, lr: 3.14e-03, grad_scale: 32.0 2023-04-27 17:14:29,415 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=123428.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:14:31,786 INFO [finetune.py:976] (1/7) Epoch 22, batch 3150, loss[loss=0.1623, simple_loss=0.2341, pruned_loss=0.04524, over 4862.00 frames. ], tot_loss[loss=0.1721, simple_loss=0.2441, pruned_loss=0.04999, over 957207.19 frames. ], batch size: 34, lr: 3.14e-03, grad_scale: 32.0 2023-04-27 17:15:14,443 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.79 vs. limit=2.0 2023-04-27 17:15:16,033 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.001e+02 1.572e+02 1.779e+02 2.164e+02 3.617e+02, threshold=3.558e+02, percent-clipped=0.0 2023-04-27 17:15:38,621 INFO [finetune.py:976] (1/7) Epoch 22, batch 3200, loss[loss=0.1545, simple_loss=0.2223, pruned_loss=0.04333, over 4871.00 frames. ], tot_loss[loss=0.1693, simple_loss=0.2409, pruned_loss=0.04889, over 957595.53 frames. ], batch size: 34, lr: 3.14e-03, grad_scale: 32.0 2023-04-27 17:15:48,872 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=123489.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:16:43,134 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=123528.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:16:51,253 INFO [finetune.py:976] (1/7) Epoch 22, batch 3250, loss[loss=0.1488, simple_loss=0.2337, pruned_loss=0.03192, over 4784.00 frames. ], tot_loss[loss=0.1688, simple_loss=0.2403, pruned_loss=0.04866, over 956072.64 frames. ], batch size: 26, lr: 3.14e-03, grad_scale: 64.0 2023-04-27 17:17:26,394 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.204e+02 1.572e+02 1.866e+02 2.251e+02 4.300e+02, threshold=3.733e+02, percent-clipped=4.0 2023-04-27 17:17:40,417 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=123576.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:17:44,044 INFO [finetune.py:976] (1/7) Epoch 22, batch 3300, loss[loss=0.2019, simple_loss=0.2798, pruned_loss=0.06198, over 4817.00 frames. ], tot_loss[loss=0.1711, simple_loss=0.2434, pruned_loss=0.04938, over 955384.73 frames. ], batch size: 33, lr: 3.14e-03, grad_scale: 64.0 2023-04-27 17:17:45,389 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6286, 1.0887, 1.7108, 2.0549, 1.7030, 1.6157, 1.6663, 1.6525], device='cuda:1'), covar=tensor([0.4395, 0.7039, 0.6391, 0.5990, 0.5786, 0.7786, 0.7679, 0.9061], device='cuda:1'), in_proj_covar=tensor([0.0431, 0.0415, 0.0508, 0.0507, 0.0460, 0.0489, 0.0498, 0.0506], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 17:18:05,174 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.5375, 1.7916, 1.8240, 1.9605, 1.7546, 1.8062, 1.8948, 1.8531], device='cuda:1'), covar=tensor([0.4150, 0.5381, 0.4331, 0.4025, 0.5189, 0.7138, 0.5098, 0.4785], device='cuda:1'), in_proj_covar=tensor([0.0339, 0.0374, 0.0326, 0.0339, 0.0348, 0.0395, 0.0357, 0.0331], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 17:18:17,760 INFO [finetune.py:976] (1/7) Epoch 22, batch 3350, loss[loss=0.1965, simple_loss=0.2682, pruned_loss=0.06242, over 4851.00 frames. ], tot_loss[loss=0.1725, simple_loss=0.2455, pruned_loss=0.04972, over 957362.64 frames. ], batch size: 44, lr: 3.14e-03, grad_scale: 64.0 2023-04-27 17:18:23,546 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-04-27 17:18:40,219 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.125e+02 1.630e+02 1.835e+02 2.174e+02 3.522e+02, threshold=3.670e+02, percent-clipped=0.0 2023-04-27 17:18:43,918 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=123670.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:18:48,062 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4119, 1.7333, 1.6430, 1.9364, 1.9125, 1.9590, 1.6304, 3.2255], device='cuda:1'), covar=tensor([0.0563, 0.0657, 0.0668, 0.0980, 0.0488, 0.0704, 0.0652, 0.0177], device='cuda:1'), in_proj_covar=tensor([0.0037, 0.0038, 0.0039, 0.0043, 0.0040, 0.0038, 0.0038, 0.0055], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:1') 2023-04-27 17:18:51,586 INFO [finetune.py:976] (1/7) Epoch 22, batch 3400, loss[loss=0.2131, simple_loss=0.2772, pruned_loss=0.07445, over 4234.00 frames. ], tot_loss[loss=0.1733, simple_loss=0.2464, pruned_loss=0.05014, over 956698.78 frames. ], batch size: 65, lr: 3.14e-03, grad_scale: 32.0 2023-04-27 17:18:59,630 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7804, 1.8748, 0.8595, 1.4619, 2.0297, 1.6003, 1.5473, 1.6496], device='cuda:1'), covar=tensor([0.0475, 0.0369, 0.0325, 0.0550, 0.0251, 0.0512, 0.0501, 0.0539], device='cuda:1'), in_proj_covar=tensor([0.0028, 0.0024, 0.0022, 0.0028, 0.0019, 0.0028, 0.0028, 0.0028], device='cuda:1'), out_proj_covar=tensor([0.0051, 0.0045, 0.0038, 0.0052, 0.0038, 0.0050, 0.0050, 0.0051], device='cuda:1') 2023-04-27 17:19:25,130 INFO [finetune.py:976] (1/7) Epoch 22, batch 3450, loss[loss=0.1487, simple_loss=0.2218, pruned_loss=0.03783, over 4793.00 frames. ], tot_loss[loss=0.1725, simple_loss=0.2459, pruned_loss=0.0496, over 955983.59 frames. ], batch size: 29, lr: 3.14e-03, grad_scale: 32.0 2023-04-27 17:19:57,039 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.012e+02 1.610e+02 1.907e+02 2.404e+02 3.474e+02, threshold=3.815e+02, percent-clipped=0.0 2023-04-27 17:20:20,667 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5107, 3.5332, 0.8184, 1.8520, 2.0300, 2.4194, 2.0602, 1.0717], device='cuda:1'), covar=tensor([0.1444, 0.1011, 0.2131, 0.1354, 0.1024, 0.1153, 0.1524, 0.2043], device='cuda:1'), in_proj_covar=tensor([0.0117, 0.0241, 0.0138, 0.0120, 0.0133, 0.0153, 0.0117, 0.0119], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-27 17:20:21,195 INFO [finetune.py:976] (1/7) Epoch 22, batch 3500, loss[loss=0.1733, simple_loss=0.2232, pruned_loss=0.06173, over 4904.00 frames. ], tot_loss[loss=0.1715, simple_loss=0.2439, pruned_loss=0.04954, over 957723.72 frames. ], batch size: 35, lr: 3.14e-03, grad_scale: 32.0 2023-04-27 17:20:28,341 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=123784.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:20:44,250 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([3.2853, 3.2637, 2.4942, 3.7792, 3.3355, 3.2854, 1.5039, 3.1943], device='cuda:1'), covar=tensor([0.2070, 0.1484, 0.3674, 0.2486, 0.3496, 0.1909, 0.5873, 0.2807], device='cuda:1'), in_proj_covar=tensor([0.0246, 0.0219, 0.0255, 0.0308, 0.0299, 0.0247, 0.0279, 0.0273], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 17:21:22,867 INFO [finetune.py:976] (1/7) Epoch 22, batch 3550, loss[loss=0.1365, simple_loss=0.2043, pruned_loss=0.03435, over 4807.00 frames. ], tot_loss[loss=0.1699, simple_loss=0.2409, pruned_loss=0.04939, over 955955.17 frames. ], batch size: 25, lr: 3.14e-03, grad_scale: 32.0 2023-04-27 17:21:43,412 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.018e+02 1.536e+02 1.822e+02 2.195e+02 3.918e+02, threshold=3.645e+02, percent-clipped=1.0 2023-04-27 17:21:49,382 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=123872.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:21:56,772 INFO [finetune.py:976] (1/7) Epoch 22, batch 3600, loss[loss=0.1922, simple_loss=0.2572, pruned_loss=0.06364, over 4733.00 frames. ], tot_loss[loss=0.1687, simple_loss=0.2396, pruned_loss=0.04887, over 957296.59 frames. ], batch size: 59, lr: 3.14e-03, grad_scale: 32.0 2023-04-27 17:22:30,655 INFO [finetune.py:976] (1/7) Epoch 22, batch 3650, loss[loss=0.1909, simple_loss=0.2775, pruned_loss=0.05214, over 4834.00 frames. ], tot_loss[loss=0.1713, simple_loss=0.2421, pruned_loss=0.05026, over 955401.18 frames. ], batch size: 40, lr: 3.14e-03, grad_scale: 32.0 2023-04-27 17:22:31,402 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=123933.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:23:14,431 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.182e+02 1.652e+02 1.968e+02 2.235e+02 6.540e+02, threshold=3.937e+02, percent-clipped=5.0 2023-04-27 17:23:23,355 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=123970.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:23:25,805 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-04-27 17:23:36,555 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.14 vs. limit=2.0 2023-04-27 17:23:37,657 INFO [finetune.py:976] (1/7) Epoch 22, batch 3700, loss[loss=0.1259, simple_loss=0.2027, pruned_loss=0.02451, over 4794.00 frames. ], tot_loss[loss=0.1723, simple_loss=0.2443, pruned_loss=0.05018, over 955311.29 frames. ], batch size: 29, lr: 3.14e-03, grad_scale: 32.0 2023-04-27 17:24:11,408 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.3554, 1.2316, 1.5891, 1.5605, 1.2775, 1.1843, 1.2779, 0.7865], device='cuda:1'), covar=tensor([0.0602, 0.0795, 0.0441, 0.0741, 0.0888, 0.1246, 0.0595, 0.0638], device='cuda:1'), in_proj_covar=tensor([0.0069, 0.0069, 0.0067, 0.0068, 0.0075, 0.0096, 0.0073, 0.0065], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-27 17:24:22,550 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=124018.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:24:44,186 INFO [finetune.py:976] (1/7) Epoch 22, batch 3750, loss[loss=0.1413, simple_loss=0.2217, pruned_loss=0.03048, over 4907.00 frames. ], tot_loss[loss=0.173, simple_loss=0.2455, pruned_loss=0.05031, over 953493.52 frames. ], batch size: 37, lr: 3.13e-03, grad_scale: 32.0 2023-04-27 17:24:52,246 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=124036.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:25:08,747 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=124062.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:25:09,866 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.228e+02 1.543e+02 1.791e+02 2.127e+02 4.626e+02, threshold=3.581e+02, percent-clipped=2.0 2023-04-27 17:25:22,231 INFO [finetune.py:976] (1/7) Epoch 22, batch 3800, loss[loss=0.1835, simple_loss=0.2444, pruned_loss=0.0613, over 4693.00 frames. ], tot_loss[loss=0.1732, simple_loss=0.2459, pruned_loss=0.05024, over 951452.80 frames. ], batch size: 23, lr: 3.13e-03, grad_scale: 32.0 2023-04-27 17:25:23,547 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=124084.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:25:32,952 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=124097.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:25:49,432 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=124123.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:25:56,294 INFO [finetune.py:976] (1/7) Epoch 22, batch 3850, loss[loss=0.1348, simple_loss=0.2118, pruned_loss=0.02891, over 4757.00 frames. ], tot_loss[loss=0.1716, simple_loss=0.2442, pruned_loss=0.04947, over 952178.80 frames. ], batch size: 27, lr: 3.13e-03, grad_scale: 32.0 2023-04-27 17:25:56,353 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=124132.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:26:06,440 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6779, 1.8240, 0.9787, 1.3744, 1.7774, 1.5403, 1.4415, 1.5729], device='cuda:1'), covar=tensor([0.0464, 0.0317, 0.0350, 0.0503, 0.0283, 0.0469, 0.0454, 0.0518], device='cuda:1'), in_proj_covar=tensor([0.0028, 0.0024, 0.0022, 0.0028, 0.0019, 0.0028, 0.0027, 0.0028], device='cuda:1'), out_proj_covar=tensor([0.0051, 0.0045, 0.0038, 0.0052, 0.0038, 0.0050, 0.0050, 0.0051], device='cuda:1') 2023-04-27 17:26:17,750 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.098e+02 1.596e+02 1.902e+02 2.251e+02 4.955e+02, threshold=3.804e+02, percent-clipped=1.0 2023-04-27 17:26:20,299 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.4543, 2.9575, 2.4056, 2.9177, 2.2621, 2.7778, 2.7819, 2.0687], device='cuda:1'), covar=tensor([0.1909, 0.1209, 0.0821, 0.1092, 0.2814, 0.0988, 0.1789, 0.2440], device='cuda:1'), in_proj_covar=tensor([0.0285, 0.0302, 0.0218, 0.0278, 0.0315, 0.0256, 0.0249, 0.0263], device='cuda:1'), out_proj_covar=tensor([1.1413e-04, 1.1937e-04, 8.5972e-05, 1.0984e-04, 1.2755e-04, 1.0136e-04, 1.0069e-04, 1.0384e-04], device='cuda:1') 2023-04-27 17:26:21,485 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.1206, 1.3849, 1.3711, 1.7160, 1.5580, 1.5214, 1.3388, 2.4776], device='cuda:1'), covar=tensor([0.0600, 0.0806, 0.0780, 0.1156, 0.0624, 0.0446, 0.0733, 0.0207], device='cuda:1'), in_proj_covar=tensor([0.0037, 0.0038, 0.0039, 0.0043, 0.0040, 0.0037, 0.0038, 0.0055], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:1') 2023-04-27 17:26:35,209 INFO [finetune.py:976] (1/7) Epoch 22, batch 3900, loss[loss=0.1784, simple_loss=0.2464, pruned_loss=0.05514, over 4781.00 frames. ], tot_loss[loss=0.1699, simple_loss=0.2423, pruned_loss=0.04881, over 951556.78 frames. ], batch size: 26, lr: 3.13e-03, grad_scale: 32.0 2023-04-27 17:27:38,094 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=124228.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:27:40,956 INFO [finetune.py:976] (1/7) Epoch 22, batch 3950, loss[loss=0.1643, simple_loss=0.2373, pruned_loss=0.04568, over 4751.00 frames. ], tot_loss[loss=0.1685, simple_loss=0.2405, pruned_loss=0.0482, over 953714.06 frames. ], batch size: 27, lr: 3.13e-03, grad_scale: 32.0 2023-04-27 17:27:54,334 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-04-27 17:28:07,730 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 8.810e+01 1.371e+02 1.651e+02 2.165e+02 4.664e+02, threshold=3.302e+02, percent-clipped=1.0 2023-04-27 17:28:13,203 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=124272.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:28:19,116 INFO [finetune.py:976] (1/7) Epoch 22, batch 4000, loss[loss=0.1717, simple_loss=0.2522, pruned_loss=0.04554, over 4907.00 frames. ], tot_loss[loss=0.1663, simple_loss=0.2379, pruned_loss=0.04735, over 953090.90 frames. ], batch size: 37, lr: 3.13e-03, grad_scale: 32.0 2023-04-27 17:29:09,167 INFO [finetune.py:976] (1/7) Epoch 22, batch 4050, loss[loss=0.2189, simple_loss=0.2927, pruned_loss=0.07249, over 4911.00 frames. ], tot_loss[loss=0.1687, simple_loss=0.2408, pruned_loss=0.04833, over 954598.68 frames. ], batch size: 36, lr: 3.13e-03, grad_scale: 32.0 2023-04-27 17:29:09,891 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=124333.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:29:33,238 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=124348.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 17:29:54,262 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.048e+02 1.579e+02 1.902e+02 2.335e+02 4.180e+02, threshold=3.804e+02, percent-clipped=3.0 2023-04-27 17:30:17,167 INFO [finetune.py:976] (1/7) Epoch 22, batch 4100, loss[loss=0.183, simple_loss=0.2649, pruned_loss=0.05059, over 4802.00 frames. ], tot_loss[loss=0.1707, simple_loss=0.243, pruned_loss=0.04924, over 952092.96 frames. ], batch size: 45, lr: 3.13e-03, grad_scale: 32.0 2023-04-27 17:30:27,280 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.3733, 1.5262, 1.2785, 1.5441, 1.2882, 1.2822, 1.4294, 1.0284], device='cuda:1'), covar=tensor([0.1421, 0.1150, 0.0909, 0.1083, 0.3253, 0.1198, 0.1542, 0.1912], device='cuda:1'), in_proj_covar=tensor([0.0288, 0.0304, 0.0220, 0.0280, 0.0319, 0.0259, 0.0252, 0.0265], device='cuda:1'), out_proj_covar=tensor([1.1511e-04, 1.2032e-04, 8.6744e-05, 1.1060e-04, 1.2888e-04, 1.0256e-04, 1.0171e-04, 1.0485e-04], device='cuda:1') 2023-04-27 17:30:34,102 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=124392.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:30:46,066 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 2023-04-27 17:30:46,902 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7617, 2.0520, 1.7381, 2.0140, 1.6828, 1.7149, 1.7071, 1.4047], device='cuda:1'), covar=tensor([0.1466, 0.1139, 0.0855, 0.1096, 0.3068, 0.1249, 0.1891, 0.2170], device='cuda:1'), in_proj_covar=tensor([0.0287, 0.0304, 0.0220, 0.0280, 0.0318, 0.0259, 0.0252, 0.0265], device='cuda:1'), out_proj_covar=tensor([1.1505e-04, 1.2024e-04, 8.6720e-05, 1.1052e-04, 1.2878e-04, 1.0258e-04, 1.0174e-04, 1.0482e-04], device='cuda:1') 2023-04-27 17:30:58,093 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=124409.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 17:30:58,713 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.0883, 2.0660, 1.8260, 1.6911, 2.1597, 1.7587, 2.6997, 1.6459], device='cuda:1'), covar=tensor([0.3647, 0.1853, 0.4736, 0.3013, 0.1645, 0.2389, 0.1216, 0.4638], device='cuda:1'), in_proj_covar=tensor([0.0338, 0.0346, 0.0423, 0.0353, 0.0378, 0.0372, 0.0368, 0.0417], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 17:31:00,638 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.54 vs. limit=2.0 2023-04-27 17:31:08,994 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=124418.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:31:27,631 INFO [finetune.py:976] (1/7) Epoch 22, batch 4150, loss[loss=0.1442, simple_loss=0.2133, pruned_loss=0.03753, over 4815.00 frames. ], tot_loss[loss=0.1712, simple_loss=0.2436, pruned_loss=0.0494, over 950143.87 frames. ], batch size: 25, lr: 3.13e-03, grad_scale: 32.0 2023-04-27 17:31:27,752 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=124432.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:31:31,546 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-04-27 17:32:10,278 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.606e+01 1.564e+02 1.862e+02 2.379e+02 7.363e+02, threshold=3.724e+02, percent-clipped=1.0 2023-04-27 17:32:33,362 INFO [finetune.py:976] (1/7) Epoch 22, batch 4200, loss[loss=0.1524, simple_loss=0.2313, pruned_loss=0.0368, over 4808.00 frames. ], tot_loss[loss=0.1705, simple_loss=0.2437, pruned_loss=0.04866, over 950884.70 frames. ], batch size: 41, lr: 3.13e-03, grad_scale: 32.0 2023-04-27 17:32:36,549 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5810, 1.5212, 0.5254, 1.2708, 1.4508, 1.4586, 1.3603, 1.3721], device='cuda:1'), covar=tensor([0.0457, 0.0351, 0.0392, 0.0535, 0.0278, 0.0487, 0.0468, 0.0538], device='cuda:1'), in_proj_covar=tensor([0.0028, 0.0024, 0.0022, 0.0028, 0.0019, 0.0028, 0.0027, 0.0028], device='cuda:1'), out_proj_covar=tensor([0.0051, 0.0045, 0.0038, 0.0052, 0.0038, 0.0050, 0.0050, 0.0051], device='cuda:1') 2023-04-27 17:32:36,699 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.73 vs. limit=2.0 2023-04-27 17:32:45,642 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=124493.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:32:53,306 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8888, 4.2528, 0.9723, 2.1398, 2.4765, 2.9136, 2.4991, 0.9650], device='cuda:1'), covar=tensor([0.1358, 0.0960, 0.2063, 0.1261, 0.0929, 0.1030, 0.1404, 0.2185], device='cuda:1'), in_proj_covar=tensor([0.0118, 0.0242, 0.0139, 0.0121, 0.0134, 0.0153, 0.0118, 0.0120], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-27 17:32:59,123 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5217, 1.7162, 1.7163, 2.0096, 1.9693, 2.0514, 1.5997, 3.8151], device='cuda:1'), covar=tensor([0.0512, 0.0760, 0.0747, 0.1177, 0.0572, 0.0480, 0.0758, 0.0168], device='cuda:1'), in_proj_covar=tensor([0.0037, 0.0038, 0.0039, 0.0043, 0.0040, 0.0038, 0.0038, 0.0055], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:1') 2023-04-27 17:33:09,234 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=124528.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:33:11,581 INFO [finetune.py:976] (1/7) Epoch 22, batch 4250, loss[loss=0.1346, simple_loss=0.2067, pruned_loss=0.03128, over 4722.00 frames. ], tot_loss[loss=0.1691, simple_loss=0.2418, pruned_loss=0.04819, over 951636.51 frames. ], batch size: 23, lr: 3.13e-03, grad_scale: 32.0 2023-04-27 17:33:33,183 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.121e+02 1.562e+02 1.869e+02 2.245e+02 4.302e+02, threshold=3.738e+02, percent-clipped=2.0 2023-04-27 17:33:40,655 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=124576.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:33:44,720 INFO [finetune.py:976] (1/7) Epoch 22, batch 4300, loss[loss=0.1773, simple_loss=0.2442, pruned_loss=0.05525, over 4787.00 frames. ], tot_loss[loss=0.169, simple_loss=0.2407, pruned_loss=0.04861, over 954133.01 frames. ], batch size: 29, lr: 3.13e-03, grad_scale: 32.0 2023-04-27 17:33:52,888 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 2023-04-27 17:33:57,625 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8081, 2.3130, 1.8197, 2.2627, 1.5390, 1.8154, 1.9444, 1.4680], device='cuda:1'), covar=tensor([0.2188, 0.1409, 0.1062, 0.1436, 0.3663, 0.1429, 0.1990, 0.2447], device='cuda:1'), in_proj_covar=tensor([0.0285, 0.0301, 0.0218, 0.0277, 0.0316, 0.0257, 0.0250, 0.0263], device='cuda:1'), out_proj_covar=tensor([1.1417e-04, 1.1914e-04, 8.5941e-05, 1.0938e-04, 1.2782e-04, 1.0167e-04, 1.0086e-04, 1.0384e-04], device='cuda:1') 2023-04-27 17:34:15,444 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=124628.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:34:17,829 INFO [finetune.py:976] (1/7) Epoch 22, batch 4350, loss[loss=0.1801, simple_loss=0.2597, pruned_loss=0.05022, over 4857.00 frames. ], tot_loss[loss=0.1666, simple_loss=0.2377, pruned_loss=0.0477, over 952940.32 frames. ], batch size: 49, lr: 3.13e-03, grad_scale: 32.0 2023-04-27 17:34:22,038 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=124638.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:34:38,884 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=124663.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:34:39,361 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.058e+02 1.483e+02 1.838e+02 2.190e+02 4.498e+02, threshold=3.677e+02, percent-clipped=2.0 2023-04-27 17:34:51,252 INFO [finetune.py:976] (1/7) Epoch 22, batch 4400, loss[loss=0.1583, simple_loss=0.2344, pruned_loss=0.04115, over 4779.00 frames. ], tot_loss[loss=0.1664, simple_loss=0.2376, pruned_loss=0.04763, over 952433.38 frames. ], batch size: 26, lr: 3.13e-03, grad_scale: 32.0 2023-04-27 17:34:58,033 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=124692.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:35:02,328 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=124699.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:35:02,351 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7338, 2.1856, 1.8745, 1.6279, 1.3071, 1.3155, 1.9651, 1.2701], device='cuda:1'), covar=tensor([0.1657, 0.1400, 0.1497, 0.1786, 0.2384, 0.2060, 0.0945, 0.2175], device='cuda:1'), in_proj_covar=tensor([0.0197, 0.0212, 0.0169, 0.0204, 0.0200, 0.0186, 0.0156, 0.0187], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-27 17:35:11,385 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=124704.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 17:35:13,847 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-04-27 17:35:32,550 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=124718.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:35:42,963 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=124724.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:35:44,220 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.4955, 1.7788, 1.9394, 2.0347, 1.8870, 1.9136, 1.9692, 1.8985], device='cuda:1'), covar=tensor([0.3968, 0.5728, 0.4520, 0.4517, 0.5592, 0.7049, 0.5405, 0.5057], device='cuda:1'), in_proj_covar=tensor([0.0340, 0.0373, 0.0324, 0.0340, 0.0348, 0.0394, 0.0358, 0.0330], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 17:35:53,107 INFO [finetune.py:976] (1/7) Epoch 22, batch 4450, loss[loss=0.1742, simple_loss=0.2577, pruned_loss=0.0453, over 4747.00 frames. ], tot_loss[loss=0.1699, simple_loss=0.2418, pruned_loss=0.04895, over 953551.35 frames. ], batch size: 28, lr: 3.13e-03, grad_scale: 32.0 2023-04-27 17:36:03,252 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9889, 2.6642, 1.1718, 1.4579, 2.2366, 1.3422, 3.4301, 1.6646], device='cuda:1'), covar=tensor([0.0734, 0.0816, 0.0766, 0.1242, 0.0466, 0.0975, 0.0230, 0.0667], device='cuda:1'), in_proj_covar=tensor([0.0050, 0.0064, 0.0047, 0.0046, 0.0049, 0.0051, 0.0073, 0.0050], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:1') 2023-04-27 17:36:03,818 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=124740.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:36:30,396 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.087e+02 1.602e+02 1.930e+02 2.289e+02 4.578e+02, threshold=3.860e+02, percent-clipped=3.0 2023-04-27 17:36:38,227 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=124766.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:37:00,152 INFO [finetune.py:976] (1/7) Epoch 22, batch 4500, loss[loss=0.2147, simple_loss=0.2865, pruned_loss=0.07148, over 4756.00 frames. ], tot_loss[loss=0.172, simple_loss=0.2439, pruned_loss=0.05003, over 952489.99 frames. ], batch size: 59, lr: 3.13e-03, grad_scale: 32.0 2023-04-27 17:37:03,928 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=124788.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:37:47,362 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=124820.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:38:07,812 INFO [finetune.py:976] (1/7) Epoch 22, batch 4550, loss[loss=0.1769, simple_loss=0.2566, pruned_loss=0.04861, over 4824.00 frames. ], tot_loss[loss=0.1735, simple_loss=0.2456, pruned_loss=0.05066, over 951906.40 frames. ], batch size: 41, lr: 3.13e-03, grad_scale: 32.0 2023-04-27 17:38:49,604 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.035e+02 1.528e+02 1.862e+02 2.235e+02 3.504e+02, threshold=3.725e+02, percent-clipped=0.0 2023-04-27 17:39:13,984 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=124881.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:39:14,477 INFO [finetune.py:976] (1/7) Epoch 22, batch 4600, loss[loss=0.183, simple_loss=0.2592, pruned_loss=0.05336, over 4899.00 frames. ], tot_loss[loss=0.1727, simple_loss=0.245, pruned_loss=0.05022, over 951045.98 frames. ], batch size: 35, lr: 3.13e-03, grad_scale: 32.0 2023-04-27 17:40:00,413 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=124928.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:40:00,516 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.76 vs. limit=5.0 2023-04-27 17:40:03,269 INFO [finetune.py:976] (1/7) Epoch 22, batch 4650, loss[loss=0.1897, simple_loss=0.2447, pruned_loss=0.06735, over 4242.00 frames. ], tot_loss[loss=0.1705, simple_loss=0.2422, pruned_loss=0.04942, over 951890.10 frames. ], batch size: 65, lr: 3.13e-03, grad_scale: 32.0 2023-04-27 17:40:23,402 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.522e+01 1.501e+02 1.822e+02 2.156e+02 6.565e+02, threshold=3.645e+02, percent-clipped=1.0 2023-04-27 17:40:31,674 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=124976.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:40:36,852 INFO [finetune.py:976] (1/7) Epoch 22, batch 4700, loss[loss=0.1659, simple_loss=0.2287, pruned_loss=0.05154, over 4822.00 frames. ], tot_loss[loss=0.1679, simple_loss=0.2392, pruned_loss=0.04827, over 951978.66 frames. ], batch size: 38, lr: 3.13e-03, grad_scale: 32.0 2023-04-27 17:40:37,578 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.1100, 1.3365, 1.2863, 1.6468, 1.5020, 1.4571, 1.3406, 2.4248], device='cuda:1'), covar=tensor([0.0610, 0.0810, 0.0738, 0.1144, 0.0623, 0.0526, 0.0718, 0.0237], device='cuda:1'), in_proj_covar=tensor([0.0037, 0.0038, 0.0039, 0.0042, 0.0039, 0.0037, 0.0037, 0.0055], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:1') 2023-04-27 17:40:44,818 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=124994.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:40:51,173 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=125004.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 17:41:00,854 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=125019.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:41:05,410 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6153, 1.4727, 1.9181, 1.9329, 1.4708, 1.3593, 1.5618, 0.9172], device='cuda:1'), covar=tensor([0.0531, 0.0672, 0.0398, 0.0616, 0.0824, 0.1111, 0.0579, 0.0731], device='cuda:1'), in_proj_covar=tensor([0.0069, 0.0068, 0.0066, 0.0068, 0.0075, 0.0095, 0.0072, 0.0065], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-27 17:41:10,030 INFO [finetune.py:976] (1/7) Epoch 22, batch 4750, loss[loss=0.2096, simple_loss=0.2758, pruned_loss=0.07165, over 4814.00 frames. ], tot_loss[loss=0.1672, simple_loss=0.2382, pruned_loss=0.04811, over 953596.78 frames. ], batch size: 39, lr: 3.13e-03, grad_scale: 32.0 2023-04-27 17:41:12,444 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8352, 1.4673, 1.9781, 2.2717, 1.9504, 1.8164, 1.9251, 1.8364], device='cuda:1'), covar=tensor([0.3933, 0.6523, 0.5970, 0.5223, 0.5416, 0.7379, 0.7262, 0.8715], device='cuda:1'), in_proj_covar=tensor([0.0433, 0.0417, 0.0511, 0.0508, 0.0463, 0.0493, 0.0499, 0.0510], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 17:41:23,893 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=125052.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 17:41:31,617 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 8.684e+01 1.607e+02 1.875e+02 2.293e+02 4.178e+02, threshold=3.749e+02, percent-clipped=1.0 2023-04-27 17:41:43,190 INFO [finetune.py:976] (1/7) Epoch 22, batch 4800, loss[loss=0.1938, simple_loss=0.2728, pruned_loss=0.05736, over 4836.00 frames. ], tot_loss[loss=0.1683, simple_loss=0.24, pruned_loss=0.04835, over 952866.75 frames. ], batch size: 49, lr: 3.13e-03, grad_scale: 32.0 2023-04-27 17:41:47,145 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([4.4503, 4.2820, 2.9897, 5.0338, 4.4316, 4.3700, 1.6722, 4.2739], device='cuda:1'), covar=tensor([0.1515, 0.0878, 0.3818, 0.0978, 0.2846, 0.1603, 0.5789, 0.2138], device='cuda:1'), in_proj_covar=tensor([0.0243, 0.0216, 0.0250, 0.0304, 0.0294, 0.0245, 0.0274, 0.0271], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 17:41:48,389 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=125088.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:42:14,808 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=125128.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:42:17,142 INFO [finetune.py:976] (1/7) Epoch 22, batch 4850, loss[loss=0.1686, simple_loss=0.2512, pruned_loss=0.04301, over 4743.00 frames. ], tot_loss[loss=0.1695, simple_loss=0.2425, pruned_loss=0.04823, over 954867.08 frames. ], batch size: 54, lr: 3.13e-03, grad_scale: 32.0 2023-04-27 17:42:20,150 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=125136.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:42:28,243 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([3.6079, 3.4694, 2.8890, 4.1348, 3.4638, 3.5525, 1.8988, 3.6269], device='cuda:1'), covar=tensor([0.1581, 0.1331, 0.4043, 0.1264, 0.2543, 0.1754, 0.4626, 0.2113], device='cuda:1'), in_proj_covar=tensor([0.0241, 0.0215, 0.0249, 0.0302, 0.0293, 0.0243, 0.0272, 0.0269], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 17:42:44,288 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.162e+02 1.525e+02 1.734e+02 2.236e+02 4.357e+02, threshold=3.469e+02, percent-clipped=1.0 2023-04-27 17:42:50,789 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7976, 2.0641, 1.4175, 1.6776, 2.1549, 1.6851, 1.6798, 1.7256], device='cuda:1'), covar=tensor([0.0414, 0.0293, 0.0270, 0.0430, 0.0248, 0.0434, 0.0403, 0.0473], device='cuda:1'), in_proj_covar=tensor([0.0028, 0.0024, 0.0022, 0.0028, 0.0019, 0.0028, 0.0028, 0.0028], device='cuda:1'), out_proj_covar=tensor([0.0051, 0.0046, 0.0038, 0.0052, 0.0038, 0.0050, 0.0050, 0.0052], device='cuda:1') 2023-04-27 17:43:02,105 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.7309, 3.3924, 2.9371, 3.2546, 2.3980, 2.8962, 3.1131, 2.3138], device='cuda:1'), covar=tensor([0.1866, 0.1069, 0.0653, 0.1085, 0.3222, 0.1256, 0.1648, 0.2571], device='cuda:1'), in_proj_covar=tensor([0.0282, 0.0298, 0.0216, 0.0274, 0.0314, 0.0254, 0.0248, 0.0260], device='cuda:1'), out_proj_covar=tensor([1.1287e-04, 1.1786e-04, 8.5211e-05, 1.0815e-04, 1.2708e-04, 1.0061e-04, 9.9906e-05, 1.0291e-04], device='cuda:1') 2023-04-27 17:43:02,726 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.2065, 1.3904, 1.3959, 1.7305, 1.5500, 1.7205, 1.4152, 3.1048], device='cuda:1'), covar=tensor([0.0622, 0.0795, 0.0792, 0.1202, 0.0632, 0.0483, 0.0710, 0.0160], device='cuda:1'), in_proj_covar=tensor([0.0037, 0.0038, 0.0039, 0.0042, 0.0039, 0.0037, 0.0037, 0.0054], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:1') 2023-04-27 17:43:03,303 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=125176.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:43:13,303 INFO [finetune.py:976] (1/7) Epoch 22, batch 4900, loss[loss=0.1368, simple_loss=0.2175, pruned_loss=0.02806, over 4837.00 frames. ], tot_loss[loss=0.1706, simple_loss=0.2436, pruned_loss=0.04878, over 954041.62 frames. ], batch size: 49, lr: 3.13e-03, grad_scale: 32.0 2023-04-27 17:43:18,228 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=125189.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:43:24,046 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=3.85 vs. limit=5.0 2023-04-27 17:44:14,015 INFO [finetune.py:976] (1/7) Epoch 22, batch 4950, loss[loss=0.1994, simple_loss=0.2691, pruned_loss=0.06481, over 4812.00 frames. ], tot_loss[loss=0.1717, simple_loss=0.2451, pruned_loss=0.04919, over 955570.68 frames. ], batch size: 30, lr: 3.13e-03, grad_scale: 32.0 2023-04-27 17:44:50,834 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.101e+02 1.480e+02 1.752e+02 2.103e+02 4.760e+02, threshold=3.505e+02, percent-clipped=3.0 2023-04-27 17:45:13,456 INFO [finetune.py:976] (1/7) Epoch 22, batch 5000, loss[loss=0.1424, simple_loss=0.2042, pruned_loss=0.04028, over 4895.00 frames. ], tot_loss[loss=0.1708, simple_loss=0.2437, pruned_loss=0.0489, over 952818.00 frames. ], batch size: 32, lr: 3.13e-03, grad_scale: 16.0 2023-04-27 17:45:32,809 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=125294.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:46:05,804 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=125319.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:46:18,221 INFO [finetune.py:976] (1/7) Epoch 22, batch 5050, loss[loss=0.1098, simple_loss=0.1867, pruned_loss=0.01644, over 4693.00 frames. ], tot_loss[loss=0.1693, simple_loss=0.2417, pruned_loss=0.04849, over 953527.72 frames. ], batch size: 23, lr: 3.13e-03, grad_scale: 16.0 2023-04-27 17:46:32,844 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=125342.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:46:43,449 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=125356.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:46:48,792 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.011e+02 1.666e+02 1.966e+02 2.426e+02 4.294e+02, threshold=3.932e+02, percent-clipped=5.0 2023-04-27 17:46:50,092 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=125367.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:46:55,674 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=125376.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:46:59,787 INFO [finetune.py:976] (1/7) Epoch 22, batch 5100, loss[loss=0.1727, simple_loss=0.2508, pruned_loss=0.04729, over 4925.00 frames. ], tot_loss[loss=0.1672, simple_loss=0.239, pruned_loss=0.04772, over 954006.41 frames. ], batch size: 46, lr: 3.12e-03, grad_scale: 16.0 2023-04-27 17:47:03,027 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.1822, 0.5481, 0.9854, 0.9216, 1.2641, 1.0721, 0.9076, 0.9767], device='cuda:1'), covar=tensor([0.1347, 0.1320, 0.1485, 0.1158, 0.0781, 0.1005, 0.1503, 0.1672], device='cuda:1'), in_proj_covar=tensor([0.0310, 0.0307, 0.0347, 0.0284, 0.0323, 0.0304, 0.0297, 0.0369], device='cuda:1'), out_proj_covar=tensor([6.3576e-05, 6.3571e-05, 7.3244e-05, 5.7166e-05, 6.6547e-05, 6.3679e-05, 6.1918e-05, 7.8337e-05], device='cuda:1') 2023-04-27 17:47:04,347 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.71 vs. limit=2.0 2023-04-27 17:47:24,361 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=125417.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 17:47:26,786 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=125421.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:47:33,405 INFO [finetune.py:976] (1/7) Epoch 22, batch 5150, loss[loss=0.1905, simple_loss=0.2772, pruned_loss=0.05194, over 4830.00 frames. ], tot_loss[loss=0.1689, simple_loss=0.2403, pruned_loss=0.04872, over 955216.87 frames. ], batch size: 39, lr: 3.12e-03, grad_scale: 16.0 2023-04-27 17:47:37,083 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=125437.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:47:47,113 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([3.8507, 3.8127, 2.7415, 4.4491, 3.8302, 3.8665, 1.6980, 3.7777], device='cuda:1'), covar=tensor([0.1996, 0.1202, 0.3321, 0.1728, 0.3952, 0.2141, 0.5888, 0.2750], device='cuda:1'), in_proj_covar=tensor([0.0244, 0.0216, 0.0251, 0.0305, 0.0296, 0.0246, 0.0274, 0.0271], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 17:48:02,911 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.720e+01 1.588e+02 1.846e+02 2.174e+02 3.765e+02, threshold=3.692e+02, percent-clipped=0.0 2023-04-27 17:48:09,689 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=125476.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:48:11,002 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.13 vs. limit=5.0 2023-04-27 17:48:13,157 INFO [finetune.py:976] (1/7) Epoch 22, batch 5200, loss[loss=0.1531, simple_loss=0.2413, pruned_loss=0.03245, over 4763.00 frames. ], tot_loss[loss=0.1712, simple_loss=0.2438, pruned_loss=0.04934, over 956874.87 frames. ], batch size: 54, lr: 3.12e-03, grad_scale: 16.0 2023-04-27 17:48:13,302 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=125482.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:48:14,968 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=125484.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:48:15,618 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5353, 1.3366, 1.1905, 1.4429, 1.7732, 1.4695, 1.2680, 1.1973], device='cuda:1'), covar=tensor([0.1598, 0.1141, 0.1629, 0.1141, 0.0720, 0.1333, 0.1978, 0.2030], device='cuda:1'), in_proj_covar=tensor([0.0309, 0.0307, 0.0348, 0.0284, 0.0323, 0.0304, 0.0297, 0.0370], device='cuda:1'), out_proj_covar=tensor([6.3453e-05, 6.3595e-05, 7.3296e-05, 5.7019e-05, 6.6475e-05, 6.3649e-05, 6.1958e-05, 7.8508e-05], device='cuda:1') 2023-04-27 17:48:35,250 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4850, 1.3903, 1.7242, 1.7136, 1.4370, 1.2632, 1.5541, 0.9937], device='cuda:1'), covar=tensor([0.0584, 0.0602, 0.0453, 0.0649, 0.0654, 0.0954, 0.0602, 0.0645], device='cuda:1'), in_proj_covar=tensor([0.0069, 0.0068, 0.0067, 0.0068, 0.0075, 0.0096, 0.0073, 0.0065], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-27 17:48:42,432 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=125524.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:48:47,213 INFO [finetune.py:976] (1/7) Epoch 22, batch 5250, loss[loss=0.1811, simple_loss=0.2538, pruned_loss=0.0542, over 4763.00 frames. ], tot_loss[loss=0.1734, simple_loss=0.2464, pruned_loss=0.0502, over 956029.35 frames. ], batch size: 54, lr: 3.12e-03, grad_scale: 16.0 2023-04-27 17:49:08,101 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([0.9635, 1.9227, 1.7906, 1.6252, 2.0117, 1.6915, 2.5480, 1.5935], device='cuda:1'), covar=tensor([0.3353, 0.1894, 0.4543, 0.2755, 0.1608, 0.2220, 0.1352, 0.4077], device='cuda:1'), in_proj_covar=tensor([0.0336, 0.0347, 0.0424, 0.0352, 0.0378, 0.0372, 0.0366, 0.0417], device='cuda:1'), out_proj_covar=tensor([9.9629e-05, 1.0384e-04, 1.2857e-04, 1.0580e-04, 1.1243e-04, 1.1101e-04, 1.0757e-04, 1.2571e-04], device='cuda:1') 2023-04-27 17:49:09,789 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.141e+02 1.600e+02 1.860e+02 2.252e+02 4.468e+02, threshold=3.721e+02, percent-clipped=2.0 2023-04-27 17:49:13,380 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4081, 1.0576, 0.4105, 1.1284, 1.0134, 1.2779, 1.2086, 1.2035], device='cuda:1'), covar=tensor([0.0510, 0.0404, 0.0412, 0.0563, 0.0323, 0.0527, 0.0500, 0.0582], device='cuda:1'), in_proj_covar=tensor([0.0028, 0.0024, 0.0022, 0.0028, 0.0019, 0.0028, 0.0028, 0.0028], device='cuda:1'), out_proj_covar=tensor([0.0051, 0.0046, 0.0038, 0.0052, 0.0039, 0.0050, 0.0050, 0.0051], device='cuda:1') 2023-04-27 17:49:20,594 INFO [finetune.py:976] (1/7) Epoch 22, batch 5300, loss[loss=0.2321, simple_loss=0.2837, pruned_loss=0.09021, over 4903.00 frames. ], tot_loss[loss=0.1758, simple_loss=0.2487, pruned_loss=0.05147, over 955673.68 frames. ], batch size: 37, lr: 3.12e-03, grad_scale: 16.0 2023-04-27 17:49:34,598 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=125603.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:49:54,072 INFO [finetune.py:976] (1/7) Epoch 22, batch 5350, loss[loss=0.1562, simple_loss=0.2281, pruned_loss=0.04212, over 4896.00 frames. ], tot_loss[loss=0.1744, simple_loss=0.2474, pruned_loss=0.05071, over 955356.21 frames. ], batch size: 36, lr: 3.12e-03, grad_scale: 16.0 2023-04-27 17:50:02,635 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.4356, 1.9919, 2.3381, 2.6322, 2.5904, 2.1125, 1.9941, 2.4164], device='cuda:1'), covar=tensor([0.0673, 0.0945, 0.0573, 0.0567, 0.0550, 0.0810, 0.0658, 0.0486], device='cuda:1'), in_proj_covar=tensor([0.0188, 0.0202, 0.0183, 0.0177, 0.0178, 0.0182, 0.0153, 0.0179], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 17:50:12,895 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.65 vs. limit=2.0 2023-04-27 17:50:15,137 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=125664.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:50:16,102 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.320e+01 1.581e+02 1.868e+02 2.376e+02 4.466e+02, threshold=3.736e+02, percent-clipped=2.0 2023-04-27 17:50:38,573 INFO [finetune.py:976] (1/7) Epoch 22, batch 5400, loss[loss=0.1919, simple_loss=0.2651, pruned_loss=0.05937, over 4721.00 frames. ], tot_loss[loss=0.172, simple_loss=0.2446, pruned_loss=0.04971, over 955247.59 frames. ], batch size: 59, lr: 3.12e-03, grad_scale: 16.0 2023-04-27 17:50:48,905 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0772, 2.5187, 1.1214, 1.3766, 1.9433, 1.2472, 3.1609, 1.6027], device='cuda:1'), covar=tensor([0.0676, 0.0629, 0.0800, 0.1128, 0.0464, 0.0929, 0.0288, 0.0608], device='cuda:1'), in_proj_covar=tensor([0.0051, 0.0064, 0.0048, 0.0046, 0.0050, 0.0052, 0.0074, 0.0051], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:1') 2023-04-27 17:50:57,523 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.23 vs. limit=5.0 2023-04-27 17:51:19,930 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=125712.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 17:51:40,594 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=125725.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:51:42,264 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.2071, 2.5700, 0.8203, 1.5597, 1.6063, 1.8207, 1.6824, 0.8752], device='cuda:1'), covar=tensor([0.1404, 0.1120, 0.1760, 0.1217, 0.1065, 0.0939, 0.1517, 0.1625], device='cuda:1'), in_proj_covar=tensor([0.0119, 0.0243, 0.0139, 0.0121, 0.0134, 0.0153, 0.0119, 0.0120], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-27 17:51:45,247 INFO [finetune.py:976] (1/7) Epoch 22, batch 5450, loss[loss=0.1531, simple_loss=0.2307, pruned_loss=0.03774, over 4897.00 frames. ], tot_loss[loss=0.1696, simple_loss=0.2415, pruned_loss=0.04881, over 956195.52 frames. ], batch size: 35, lr: 3.12e-03, grad_scale: 16.0 2023-04-27 17:51:45,328 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=125732.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:52:28,292 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.173e+01 1.432e+02 1.653e+02 1.978e+02 3.483e+02, threshold=3.306e+02, percent-clipped=0.0 2023-04-27 17:52:36,389 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=3.23 vs. limit=5.0 2023-04-27 17:52:47,945 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=125777.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:52:51,421 INFO [finetune.py:976] (1/7) Epoch 22, batch 5500, loss[loss=0.1822, simple_loss=0.2483, pruned_loss=0.05804, over 4907.00 frames. ], tot_loss[loss=0.1681, simple_loss=0.2392, pruned_loss=0.04849, over 955195.55 frames. ], batch size: 46, lr: 3.12e-03, grad_scale: 16.0 2023-04-27 17:52:58,404 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=125784.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:52:59,621 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=125786.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:53:42,124 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-04-27 17:53:57,322 INFO [finetune.py:976] (1/7) Epoch 22, batch 5550, loss[loss=0.182, simple_loss=0.2627, pruned_loss=0.05068, over 4850.00 frames. ], tot_loss[loss=0.1683, simple_loss=0.24, pruned_loss=0.04832, over 955318.02 frames. ], batch size: 44, lr: 3.12e-03, grad_scale: 16.0 2023-04-27 17:54:03,349 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=125832.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:54:41,115 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.131e+02 1.577e+02 1.939e+02 2.274e+02 3.783e+02, threshold=3.878e+02, percent-clipped=3.0 2023-04-27 17:54:49,411 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.35 vs. limit=5.0 2023-04-27 17:54:51,075 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.2994, 1.2567, 1.3376, 1.5606, 1.5601, 1.1479, 0.9792, 1.3607], device='cuda:1'), covar=tensor([0.0864, 0.1260, 0.0899, 0.0727, 0.0701, 0.0907, 0.0940, 0.0674], device='cuda:1'), in_proj_covar=tensor([0.0187, 0.0202, 0.0184, 0.0176, 0.0177, 0.0181, 0.0152, 0.0179], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 17:55:02,079 INFO [finetune.py:976] (1/7) Epoch 22, batch 5600, loss[loss=0.1707, simple_loss=0.2447, pruned_loss=0.04835, over 4811.00 frames. ], tot_loss[loss=0.1719, simple_loss=0.2447, pruned_loss=0.04957, over 953652.27 frames. ], batch size: 41, lr: 3.12e-03, grad_scale: 16.0 2023-04-27 17:55:32,970 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6793, 2.1227, 1.6759, 1.4066, 1.2787, 1.2672, 1.6820, 1.2104], device='cuda:1'), covar=tensor([0.1758, 0.1307, 0.1520, 0.1746, 0.2346, 0.2053, 0.1025, 0.2120], device='cuda:1'), in_proj_covar=tensor([0.0197, 0.0212, 0.0170, 0.0204, 0.0200, 0.0186, 0.0156, 0.0187], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-27 17:55:54,639 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6866, 1.2052, 1.7204, 2.2104, 1.7751, 1.6015, 1.6772, 1.6414], device='cuda:1'), covar=tensor([0.4625, 0.6754, 0.6558, 0.5658, 0.5897, 0.7853, 0.7959, 0.9100], device='cuda:1'), in_proj_covar=tensor([0.0432, 0.0414, 0.0507, 0.0505, 0.0462, 0.0490, 0.0497, 0.0508], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 17:56:05,132 INFO [finetune.py:976] (1/7) Epoch 22, batch 5650, loss[loss=0.1849, simple_loss=0.2628, pruned_loss=0.05353, over 4842.00 frames. ], tot_loss[loss=0.1736, simple_loss=0.2473, pruned_loss=0.04996, over 954289.49 frames. ], batch size: 47, lr: 3.12e-03, grad_scale: 16.0 2023-04-27 17:56:13,519 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([3.3657, 3.3870, 2.4385, 3.8375, 3.3064, 3.3350, 1.4081, 3.3063], device='cuda:1'), covar=tensor([0.1916, 0.1296, 0.3437, 0.2423, 0.3402, 0.1936, 0.6206, 0.2635], device='cuda:1'), in_proj_covar=tensor([0.0244, 0.0216, 0.0252, 0.0305, 0.0296, 0.0246, 0.0274, 0.0271], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 17:56:31,892 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=125959.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:56:35,408 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.900e+01 1.470e+02 1.804e+02 2.167e+02 3.376e+02, threshold=3.608e+02, percent-clipped=0.0 2023-04-27 17:56:45,404 INFO [finetune.py:976] (1/7) Epoch 22, batch 5700, loss[loss=0.1568, simple_loss=0.2215, pruned_loss=0.04611, over 4308.00 frames. ], tot_loss[loss=0.1706, simple_loss=0.2429, pruned_loss=0.04918, over 936366.05 frames. ], batch size: 18, lr: 3.12e-03, grad_scale: 16.0 2023-04-27 17:56:54,942 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=125998.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:57:25,281 INFO [finetune.py:976] (1/7) Epoch 23, batch 0, loss[loss=0.2157, simple_loss=0.2806, pruned_loss=0.07545, over 4841.00 frames. ], tot_loss[loss=0.2157, simple_loss=0.2806, pruned_loss=0.07545, over 4841.00 frames. ], batch size: 49, lr: 3.12e-03, grad_scale: 16.0 2023-04-27 17:57:25,281 INFO [finetune.py:1001] (1/7) Computing validation loss 2023-04-27 17:57:31,438 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6422, 2.8852, 1.1734, 1.9157, 1.9377, 2.2372, 1.9393, 1.2340], device='cuda:1'), covar=tensor([0.1031, 0.0977, 0.1551, 0.0989, 0.0841, 0.0798, 0.1217, 0.1482], device='cuda:1'), in_proj_covar=tensor([0.0118, 0.0240, 0.0138, 0.0120, 0.0132, 0.0151, 0.0118, 0.0119], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-27 17:57:32,454 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8105, 1.6280, 1.7946, 2.1546, 2.1102, 1.7152, 1.5609, 1.9477], device='cuda:1'), covar=tensor([0.0841, 0.1131, 0.0751, 0.0601, 0.0597, 0.0838, 0.0740, 0.0514], device='cuda:1'), in_proj_covar=tensor([0.0187, 0.0202, 0.0184, 0.0175, 0.0177, 0.0181, 0.0152, 0.0178], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 17:57:41,119 INFO [finetune.py:1010] (1/7) Epoch 23, validation: loss=0.1552, simple_loss=0.2246, pruned_loss=0.04292, over 2265189.00 frames. 2023-04-27 17:57:41,120 INFO [finetune.py:1011] (1/7) Maximum memory allocated so far is 6391MB 2023-04-27 17:57:48,906 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=126012.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 17:58:05,546 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=126032.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:58:11,596 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.3590, 1.5748, 1.4227, 1.6003, 1.3498, 1.3519, 1.4209, 1.0588], device='cuda:1'), covar=tensor([0.1678, 0.1265, 0.0948, 0.1166, 0.3714, 0.1236, 0.1553, 0.2231], device='cuda:1'), in_proj_covar=tensor([0.0286, 0.0301, 0.0217, 0.0278, 0.0316, 0.0256, 0.0250, 0.0263], device='cuda:1'), out_proj_covar=tensor([1.1424e-04, 1.1925e-04, 8.5693e-05, 1.0962e-04, 1.2800e-04, 1.0141e-04, 1.0075e-04, 1.0390e-04], device='cuda:1') 2023-04-27 17:58:23,638 INFO [finetune.py:976] (1/7) Epoch 23, batch 50, loss[loss=0.207, simple_loss=0.2689, pruned_loss=0.07254, over 4744.00 frames. ], tot_loss[loss=0.1757, simple_loss=0.2469, pruned_loss=0.05222, over 214380.02 frames. ], batch size: 59, lr: 3.12e-03, grad_scale: 16.0 2023-04-27 17:58:23,756 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=126059.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:58:24,776 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=126060.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:58:28,244 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.078e+02 1.502e+02 1.744e+02 2.087e+02 3.495e+02, threshold=3.488e+02, percent-clipped=0.0 2023-04-27 17:58:30,835 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-04-27 17:58:41,181 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=126077.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:58:42,972 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=126080.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:58:48,842 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=126081.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:58:51,314 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5574, 3.4371, 0.9113, 1.8284, 1.8710, 2.2388, 1.8730, 1.0198], device='cuda:1'), covar=tensor([0.1291, 0.0914, 0.1992, 0.1246, 0.1101, 0.1131, 0.1530, 0.2048], device='cuda:1'), in_proj_covar=tensor([0.0117, 0.0239, 0.0138, 0.0120, 0.0132, 0.0151, 0.0117, 0.0119], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-27 17:59:02,837 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6272, 1.7876, 0.8104, 1.3799, 1.7552, 1.4875, 1.4445, 1.5151], device='cuda:1'), covar=tensor([0.0537, 0.0387, 0.0377, 0.0583, 0.0298, 0.0554, 0.0552, 0.0617], device='cuda:1'), in_proj_covar=tensor([0.0028, 0.0024, 0.0022, 0.0028, 0.0019, 0.0028, 0.0028, 0.0028], device='cuda:1'), out_proj_covar=tensor([0.0051, 0.0046, 0.0038, 0.0052, 0.0039, 0.0050, 0.0050, 0.0052], device='cuda:1') 2023-04-27 17:59:21,047 INFO [finetune.py:976] (1/7) Epoch 23, batch 100, loss[loss=0.1294, simple_loss=0.2081, pruned_loss=0.0254, over 4782.00 frames. ], tot_loss[loss=0.1683, simple_loss=0.2393, pruned_loss=0.0486, over 380403.50 frames. ], batch size: 29, lr: 3.12e-03, grad_scale: 16.0 2023-04-27 17:59:42,926 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=126125.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:00:04,633 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.1941, 2.1400, 1.8650, 1.7612, 2.2671, 1.7791, 2.7215, 1.7070], device='cuda:1'), covar=tensor([0.3318, 0.1838, 0.4251, 0.3055, 0.1516, 0.2315, 0.1077, 0.3917], device='cuda:1'), in_proj_covar=tensor([0.0338, 0.0350, 0.0426, 0.0353, 0.0380, 0.0374, 0.0369, 0.0420], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 18:00:05,873 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([0.4165, 1.3261, 1.4384, 0.9481, 1.3002, 1.1278, 1.6933, 1.3729], device='cuda:1'), covar=tensor([0.3721, 0.1890, 0.5119, 0.2842, 0.1670, 0.2206, 0.1586, 0.4548], device='cuda:1'), in_proj_covar=tensor([0.0338, 0.0350, 0.0426, 0.0353, 0.0380, 0.0374, 0.0369, 0.0420], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 18:00:28,538 INFO [finetune.py:976] (1/7) Epoch 23, batch 150, loss[loss=0.1739, simple_loss=0.2355, pruned_loss=0.05616, over 4817.00 frames. ], tot_loss[loss=0.1648, simple_loss=0.2352, pruned_loss=0.04725, over 508766.36 frames. ], batch size: 33, lr: 3.12e-03, grad_scale: 16.0 2023-04-27 18:00:37,393 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.2270, 1.9067, 2.4064, 2.6830, 2.3405, 2.2061, 2.2861, 2.0973], device='cuda:1'), covar=tensor([0.3897, 0.5738, 0.5707, 0.5011, 0.5246, 0.6883, 0.6720, 0.7330], device='cuda:1'), in_proj_covar=tensor([0.0433, 0.0416, 0.0508, 0.0506, 0.0463, 0.0491, 0.0498, 0.0508], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 18:00:38,440 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 8.878e+01 1.494e+02 1.914e+02 2.300e+02 4.167e+02, threshold=3.828e+02, percent-clipped=5.0 2023-04-27 18:01:10,255 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.2243, 1.5974, 1.6029, 1.9146, 1.9695, 2.1447, 1.6073, 4.1457], device='cuda:1'), covar=tensor([0.0577, 0.0804, 0.0756, 0.1193, 0.0585, 0.0517, 0.0699, 0.0098], device='cuda:1'), in_proj_covar=tensor([0.0037, 0.0038, 0.0039, 0.0043, 0.0040, 0.0037, 0.0037, 0.0055], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:1') 2023-04-27 18:01:15,048 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7335, 2.7472, 2.2951, 2.4210, 2.8620, 2.4725, 3.7120, 2.1686], device='cuda:1'), covar=tensor([0.3778, 0.2369, 0.3916, 0.3676, 0.2007, 0.2525, 0.1675, 0.4393], device='cuda:1'), in_proj_covar=tensor([0.0336, 0.0348, 0.0424, 0.0351, 0.0378, 0.0372, 0.0367, 0.0419], device='cuda:1'), out_proj_covar=tensor([9.9704e-05, 1.0424e-04, 1.2856e-04, 1.0561e-04, 1.1260e-04, 1.1091e-04, 1.0804e-04, 1.2615e-04], device='cuda:1') 2023-04-27 18:01:34,970 INFO [finetune.py:976] (1/7) Epoch 23, batch 200, loss[loss=0.1491, simple_loss=0.23, pruned_loss=0.0341, over 4908.00 frames. ], tot_loss[loss=0.1644, simple_loss=0.2343, pruned_loss=0.04725, over 607039.15 frames. ], batch size: 35, lr: 3.12e-03, grad_scale: 16.0 2023-04-27 18:02:03,931 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.5319, 2.1405, 2.4658, 2.7321, 2.7425, 2.2534, 2.0721, 2.5237], device='cuda:1'), covar=tensor([0.0782, 0.1002, 0.0644, 0.0629, 0.0660, 0.0855, 0.0701, 0.0546], device='cuda:1'), in_proj_covar=tensor([0.0186, 0.0201, 0.0183, 0.0174, 0.0176, 0.0180, 0.0151, 0.0177], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 18:02:19,986 INFO [finetune.py:976] (1/7) Epoch 23, batch 250, loss[loss=0.2154, simple_loss=0.2916, pruned_loss=0.06965, over 4809.00 frames. ], tot_loss[loss=0.1672, simple_loss=0.238, pruned_loss=0.04818, over 684819.92 frames. ], batch size: 51, lr: 3.11e-03, grad_scale: 16.0 2023-04-27 18:02:20,594 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=126259.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:02:24,185 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.438e+01 1.622e+02 1.975e+02 2.331e+02 7.246e+02, threshold=3.950e+02, percent-clipped=3.0 2023-04-27 18:02:45,438 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6593, 2.7349, 2.2911, 2.4022, 2.8147, 2.3415, 3.7541, 2.2032], device='cuda:1'), covar=tensor([0.3923, 0.2230, 0.3896, 0.3243, 0.1812, 0.2552, 0.1130, 0.4070], device='cuda:1'), in_proj_covar=tensor([0.0339, 0.0352, 0.0429, 0.0355, 0.0383, 0.0376, 0.0371, 0.0423], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 18:02:51,715 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=126307.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:02:53,383 INFO [finetune.py:976] (1/7) Epoch 23, batch 300, loss[loss=0.122, simple_loss=0.1988, pruned_loss=0.0226, over 4800.00 frames. ], tot_loss[loss=0.169, simple_loss=0.2406, pruned_loss=0.04867, over 743067.11 frames. ], batch size: 25, lr: 3.11e-03, grad_scale: 16.0 2023-04-27 18:03:22,894 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=126354.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:03:26,334 INFO [finetune.py:976] (1/7) Epoch 23, batch 350, loss[loss=0.1297, simple_loss=0.2134, pruned_loss=0.023, over 4811.00 frames. ], tot_loss[loss=0.1705, simple_loss=0.243, pruned_loss=0.04897, over 790619.81 frames. ], batch size: 25, lr: 3.11e-03, grad_scale: 16.0 2023-04-27 18:03:30,388 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.435e+01 1.518e+02 1.849e+02 2.150e+02 3.695e+02, threshold=3.697e+02, percent-clipped=0.0 2023-04-27 18:03:39,078 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8286, 4.1567, 1.0335, 2.0844, 2.2446, 2.6257, 2.2947, 0.9631], device='cuda:1'), covar=tensor([0.1324, 0.0841, 0.1901, 0.1295, 0.1047, 0.1102, 0.1539, 0.2184], device='cuda:1'), in_proj_covar=tensor([0.0117, 0.0238, 0.0137, 0.0119, 0.0131, 0.0150, 0.0116, 0.0118], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-27 18:03:41,552 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=126380.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:03:42,119 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=126381.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:03:48,269 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6823, 2.8721, 2.3940, 2.5342, 3.0212, 2.4600, 3.9360, 2.3039], device='cuda:1'), covar=tensor([0.3873, 0.2023, 0.3746, 0.3032, 0.1792, 0.2587, 0.1084, 0.3907], device='cuda:1'), in_proj_covar=tensor([0.0340, 0.0352, 0.0429, 0.0356, 0.0383, 0.0376, 0.0372, 0.0423], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 18:03:48,907 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6538, 1.0129, 1.6243, 2.0954, 1.7299, 1.5363, 1.5821, 1.5952], device='cuda:1'), covar=tensor([0.4152, 0.6366, 0.5836, 0.5464, 0.5362, 0.7449, 0.7514, 0.8807], device='cuda:1'), in_proj_covar=tensor([0.0434, 0.0416, 0.0509, 0.0507, 0.0463, 0.0492, 0.0499, 0.0509], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 18:03:50,113 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8654, 1.7534, 2.2490, 2.3142, 1.6722, 1.5486, 1.8242, 1.0165], device='cuda:1'), covar=tensor([0.0566, 0.0722, 0.0362, 0.0705, 0.0753, 0.1018, 0.0714, 0.0703], device='cuda:1'), in_proj_covar=tensor([0.0069, 0.0068, 0.0066, 0.0068, 0.0075, 0.0095, 0.0073, 0.0064], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-27 18:03:58,001 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.9750, 2.4708, 2.8562, 3.1065, 3.0088, 2.7150, 2.3826, 2.8837], device='cuda:1'), covar=tensor([0.0733, 0.1013, 0.0600, 0.0672, 0.0697, 0.0769, 0.0690, 0.0507], device='cuda:1'), in_proj_covar=tensor([0.0186, 0.0202, 0.0183, 0.0175, 0.0176, 0.0180, 0.0151, 0.0178], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 18:03:59,739 INFO [finetune.py:976] (1/7) Epoch 23, batch 400, loss[loss=0.1573, simple_loss=0.2274, pruned_loss=0.04362, over 4026.00 frames. ], tot_loss[loss=0.17, simple_loss=0.2433, pruned_loss=0.04841, over 824061.48 frames. ], batch size: 17, lr: 3.11e-03, grad_scale: 16.0 2023-04-27 18:04:30,797 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=126429.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:04:42,873 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=126441.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:04:48,372 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=126450.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:04:52,733 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-04-27 18:04:54,270 INFO [finetune.py:976] (1/7) Epoch 23, batch 450, loss[loss=0.1741, simple_loss=0.2384, pruned_loss=0.05487, over 4666.00 frames. ], tot_loss[loss=0.17, simple_loss=0.2431, pruned_loss=0.04849, over 854437.98 frames. ], batch size: 23, lr: 3.11e-03, grad_scale: 16.0 2023-04-27 18:04:58,394 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.037e+02 1.511e+02 1.837e+02 2.263e+02 5.457e+02, threshold=3.673e+02, percent-clipped=4.0 2023-04-27 18:05:12,081 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.54 vs. limit=5.0 2023-04-27 18:05:27,628 INFO [finetune.py:976] (1/7) Epoch 23, batch 500, loss[loss=0.1657, simple_loss=0.2375, pruned_loss=0.04689, over 4935.00 frames. ], tot_loss[loss=0.1706, simple_loss=0.2425, pruned_loss=0.04936, over 877423.51 frames. ], batch size: 38, lr: 3.11e-03, grad_scale: 16.0 2023-04-27 18:05:29,450 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=126511.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:06:06,864 INFO [finetune.py:976] (1/7) Epoch 23, batch 550, loss[loss=0.1527, simple_loss=0.2077, pruned_loss=0.04884, over 4137.00 frames. ], tot_loss[loss=0.1694, simple_loss=0.2404, pruned_loss=0.04915, over 895766.42 frames. ], batch size: 18, lr: 3.11e-03, grad_scale: 16.0 2023-04-27 18:06:16,306 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.004e+02 1.502e+02 1.815e+02 2.160e+02 5.481e+02, threshold=3.630e+02, percent-clipped=1.0 2023-04-27 18:06:18,771 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5303, 2.6352, 2.2401, 2.3712, 2.7834, 2.2356, 3.7136, 2.0160], device='cuda:1'), covar=tensor([0.3570, 0.2258, 0.4305, 0.3291, 0.1909, 0.2701, 0.1215, 0.4254], device='cuda:1'), in_proj_covar=tensor([0.0337, 0.0348, 0.0425, 0.0351, 0.0378, 0.0372, 0.0369, 0.0419], device='cuda:1'), out_proj_covar=tensor([9.9877e-05, 1.0426e-04, 1.2878e-04, 1.0571e-04, 1.1238e-04, 1.1095e-04, 1.0846e-04, 1.2620e-04], device='cuda:1') 2023-04-27 18:06:30,306 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4735, 1.1039, 4.1632, 3.8942, 3.5959, 3.9369, 3.8735, 3.6346], device='cuda:1'), covar=tensor([0.7329, 0.6414, 0.1101, 0.1932, 0.1298, 0.1896, 0.1610, 0.1552], device='cuda:1'), in_proj_covar=tensor([0.0306, 0.0304, 0.0402, 0.0403, 0.0345, 0.0406, 0.0311, 0.0361], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-27 18:07:12,074 INFO [finetune.py:976] (1/7) Epoch 23, batch 600, loss[loss=0.1405, simple_loss=0.2176, pruned_loss=0.03174, over 4825.00 frames. ], tot_loss[loss=0.1704, simple_loss=0.2416, pruned_loss=0.04956, over 906101.12 frames. ], batch size: 33, lr: 3.11e-03, grad_scale: 16.0 2023-04-27 18:08:07,997 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5103, 1.5105, 0.5810, 1.2559, 1.6354, 1.3512, 1.2869, 1.3556], device='cuda:1'), covar=tensor([0.0518, 0.0385, 0.0384, 0.0559, 0.0298, 0.0519, 0.0513, 0.0601], device='cuda:1'), in_proj_covar=tensor([0.0028, 0.0024, 0.0022, 0.0028, 0.0019, 0.0028, 0.0027, 0.0028], device='cuda:1'), out_proj_covar=tensor([0.0051, 0.0045, 0.0038, 0.0052, 0.0039, 0.0050, 0.0050, 0.0051], device='cuda:1') 2023-04-27 18:08:14,765 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=126654.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:08:17,790 INFO [finetune.py:976] (1/7) Epoch 23, batch 650, loss[loss=0.1656, simple_loss=0.2394, pruned_loss=0.04586, over 4858.00 frames. ], tot_loss[loss=0.1725, simple_loss=0.2448, pruned_loss=0.05016, over 918011.95 frames. ], batch size: 31, lr: 3.11e-03, grad_scale: 16.0 2023-04-27 18:08:26,640 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.220e+02 1.615e+02 1.965e+02 2.363e+02 5.710e+02, threshold=3.929e+02, percent-clipped=5.0 2023-04-27 18:09:20,361 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=126702.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:09:24,629 INFO [finetune.py:976] (1/7) Epoch 23, batch 700, loss[loss=0.1636, simple_loss=0.2303, pruned_loss=0.0484, over 4787.00 frames. ], tot_loss[loss=0.1741, simple_loss=0.2465, pruned_loss=0.05083, over 927437.74 frames. ], batch size: 26, lr: 3.11e-03, grad_scale: 16.0 2023-04-27 18:09:31,801 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8786, 1.3149, 1.9497, 2.3669, 1.9426, 1.8150, 1.9203, 1.8809], device='cuda:1'), covar=tensor([0.5015, 0.7122, 0.7116, 0.6195, 0.6478, 0.8825, 0.8603, 0.8478], device='cuda:1'), in_proj_covar=tensor([0.0434, 0.0417, 0.0509, 0.0508, 0.0463, 0.0493, 0.0500, 0.0509], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 18:10:03,678 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=126736.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:10:19,571 INFO [finetune.py:976] (1/7) Epoch 23, batch 750, loss[loss=0.1555, simple_loss=0.2376, pruned_loss=0.03671, over 4769.00 frames. ], tot_loss[loss=0.1732, simple_loss=0.2464, pruned_loss=0.05004, over 934724.04 frames. ], batch size: 28, lr: 3.11e-03, grad_scale: 16.0 2023-04-27 18:10:23,175 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.166e+01 1.471e+02 1.836e+02 2.076e+02 2.754e+02, threshold=3.672e+02, percent-clipped=0.0 2023-04-27 18:10:51,753 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=126806.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:10:53,513 INFO [finetune.py:976] (1/7) Epoch 23, batch 800, loss[loss=0.1299, simple_loss=0.215, pruned_loss=0.02234, over 4886.00 frames. ], tot_loss[loss=0.172, simple_loss=0.2453, pruned_loss=0.04936, over 939761.70 frames. ], batch size: 32, lr: 3.11e-03, grad_scale: 16.0 2023-04-27 18:10:57,928 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=126816.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:10:58,551 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.0924, 1.5531, 1.5104, 1.7626, 1.7035, 1.8439, 1.4133, 3.3645], device='cuda:1'), covar=tensor([0.0649, 0.0790, 0.0780, 0.1165, 0.0629, 0.0480, 0.0692, 0.0144], device='cuda:1'), in_proj_covar=tensor([0.0037, 0.0038, 0.0039, 0.0042, 0.0040, 0.0037, 0.0037, 0.0054], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:1') 2023-04-27 18:11:27,334 INFO [finetune.py:976] (1/7) Epoch 23, batch 850, loss[loss=0.1653, simple_loss=0.2391, pruned_loss=0.04574, over 4714.00 frames. ], tot_loss[loss=0.1695, simple_loss=0.2425, pruned_loss=0.04821, over 943733.31 frames. ], batch size: 23, lr: 3.11e-03, grad_scale: 16.0 2023-04-27 18:11:29,900 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.6823, 2.0977, 2.1094, 2.2051, 2.1085, 2.1975, 2.2123, 2.2057], device='cuda:1'), covar=tensor([0.3758, 0.5078, 0.4669, 0.4827, 0.5295, 0.6505, 0.5401, 0.4551], device='cuda:1'), in_proj_covar=tensor([0.0338, 0.0373, 0.0323, 0.0338, 0.0345, 0.0394, 0.0356, 0.0330], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 18:11:30,954 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.555e+01 1.428e+02 1.632e+02 2.009e+02 4.490e+02, threshold=3.264e+02, percent-clipped=1.0 2023-04-27 18:11:44,256 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=126877.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:12:28,952 INFO [finetune.py:976] (1/7) Epoch 23, batch 900, loss[loss=0.1878, simple_loss=0.2497, pruned_loss=0.06295, over 4820.00 frames. ], tot_loss[loss=0.1674, simple_loss=0.2397, pruned_loss=0.04755, over 945810.16 frames. ], batch size: 38, lr: 3.11e-03, grad_scale: 16.0 2023-04-27 18:12:30,274 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.3237, 1.2802, 4.1299, 3.8856, 3.6885, 3.9639, 3.9169, 3.6784], device='cuda:1'), covar=tensor([0.7060, 0.5899, 0.1042, 0.1526, 0.1071, 0.1491, 0.1477, 0.1397], device='cuda:1'), in_proj_covar=tensor([0.0310, 0.0308, 0.0406, 0.0408, 0.0349, 0.0411, 0.0314, 0.0364], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-27 18:13:36,983 INFO [finetune.py:976] (1/7) Epoch 23, batch 950, loss[loss=0.1589, simple_loss=0.2388, pruned_loss=0.0395, over 4808.00 frames. ], tot_loss[loss=0.1672, simple_loss=0.2388, pruned_loss=0.04777, over 947162.66 frames. ], batch size: 38, lr: 3.11e-03, grad_scale: 16.0 2023-04-27 18:13:40,660 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.057e+02 1.507e+02 1.824e+02 2.265e+02 3.979e+02, threshold=3.647e+02, percent-clipped=2.0 2023-04-27 18:14:11,163 INFO [finetune.py:976] (1/7) Epoch 23, batch 1000, loss[loss=0.1748, simple_loss=0.2572, pruned_loss=0.04625, over 4820.00 frames. ], tot_loss[loss=0.1692, simple_loss=0.241, pruned_loss=0.04866, over 949323.60 frames. ], batch size: 38, lr: 3.11e-03, grad_scale: 16.0 2023-04-27 18:14:28,600 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=127036.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:14:42,828 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7058, 1.7467, 0.9161, 1.4016, 1.7152, 1.5520, 1.4810, 1.5472], device='cuda:1'), covar=tensor([0.0490, 0.0367, 0.0320, 0.0565, 0.0266, 0.0511, 0.0485, 0.0573], device='cuda:1'), in_proj_covar=tensor([0.0028, 0.0024, 0.0022, 0.0029, 0.0019, 0.0028, 0.0027, 0.0028], device='cuda:1'), out_proj_covar=tensor([0.0051, 0.0046, 0.0038, 0.0052, 0.0039, 0.0050, 0.0050, 0.0052], device='cuda:1') 2023-04-27 18:14:45,073 INFO [finetune.py:976] (1/7) Epoch 23, batch 1050, loss[loss=0.1781, simple_loss=0.2727, pruned_loss=0.0418, over 4837.00 frames. ], tot_loss[loss=0.1721, simple_loss=0.245, pruned_loss=0.04958, over 950786.27 frames. ], batch size: 47, lr: 3.11e-03, grad_scale: 16.0 2023-04-27 18:14:48,716 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.212e+02 1.574e+02 1.858e+02 2.309e+02 5.197e+02, threshold=3.716e+02, percent-clipped=5.0 2023-04-27 18:15:00,297 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=127084.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:15:16,206 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=127106.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:15:18,461 INFO [finetune.py:976] (1/7) Epoch 23, batch 1100, loss[loss=0.1181, simple_loss=0.1892, pruned_loss=0.02352, over 4682.00 frames. ], tot_loss[loss=0.1713, simple_loss=0.2447, pruned_loss=0.04893, over 952410.66 frames. ], batch size: 23, lr: 3.11e-03, grad_scale: 16.0 2023-04-27 18:15:47,875 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=127154.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:15:51,853 INFO [finetune.py:976] (1/7) Epoch 23, batch 1150, loss[loss=0.1479, simple_loss=0.2178, pruned_loss=0.03897, over 4824.00 frames. ], tot_loss[loss=0.1727, simple_loss=0.2462, pruned_loss=0.04965, over 953280.53 frames. ], batch size: 30, lr: 3.11e-03, grad_scale: 16.0 2023-04-27 18:15:52,079 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 2023-04-27 18:15:56,439 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.040e+02 1.449e+02 1.782e+02 2.216e+02 4.823e+02, threshold=3.563e+02, percent-clipped=1.0 2023-04-27 18:16:00,785 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=127172.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:16:25,333 INFO [finetune.py:976] (1/7) Epoch 23, batch 1200, loss[loss=0.169, simple_loss=0.2413, pruned_loss=0.04836, over 4798.00 frames. ], tot_loss[loss=0.1717, simple_loss=0.2446, pruned_loss=0.04936, over 953033.67 frames. ], batch size: 51, lr: 3.11e-03, grad_scale: 16.0 2023-04-27 18:16:31,797 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9182, 1.5823, 1.4915, 1.9280, 2.2055, 1.8329, 1.5858, 1.4555], device='cuda:1'), covar=tensor([0.1649, 0.1645, 0.2099, 0.1094, 0.0885, 0.1517, 0.1997, 0.2400], device='cuda:1'), in_proj_covar=tensor([0.0315, 0.0311, 0.0352, 0.0288, 0.0327, 0.0309, 0.0299, 0.0372], device='cuda:1'), out_proj_covar=tensor([6.4577e-05, 6.4366e-05, 7.4265e-05, 5.8052e-05, 6.7436e-05, 6.4867e-05, 6.2444e-05, 7.8954e-05], device='cuda:1') 2023-04-27 18:16:52,076 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=127249.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:16:53,185 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.4083, 1.9946, 2.2787, 2.8456, 2.8098, 2.1693, 2.0638, 2.3463], device='cuda:1'), covar=tensor([0.0885, 0.1198, 0.0726, 0.0538, 0.0560, 0.0980, 0.0754, 0.0660], device='cuda:1'), in_proj_covar=tensor([0.0187, 0.0202, 0.0184, 0.0174, 0.0176, 0.0180, 0.0151, 0.0178], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 18:16:55,710 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-04-27 18:17:03,695 INFO [finetune.py:976] (1/7) Epoch 23, batch 1250, loss[loss=0.1519, simple_loss=0.2169, pruned_loss=0.0435, over 4722.00 frames. ], tot_loss[loss=0.1689, simple_loss=0.2412, pruned_loss=0.0483, over 953016.37 frames. ], batch size: 23, lr: 3.11e-03, grad_scale: 16.0 2023-04-27 18:17:14,004 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.661e+01 1.481e+02 1.752e+02 2.149e+02 4.730e+02, threshold=3.504e+02, percent-clipped=1.0 2023-04-27 18:18:08,738 INFO [finetune.py:976] (1/7) Epoch 23, batch 1300, loss[loss=0.1339, simple_loss=0.2104, pruned_loss=0.02873, over 4825.00 frames. ], tot_loss[loss=0.1667, simple_loss=0.2384, pruned_loss=0.0475, over 955201.98 frames. ], batch size: 40, lr: 3.11e-03, grad_scale: 32.0 2023-04-27 18:18:09,509 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=127310.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:19:13,548 INFO [finetune.py:976] (1/7) Epoch 23, batch 1350, loss[loss=0.2072, simple_loss=0.2712, pruned_loss=0.07161, over 4910.00 frames. ], tot_loss[loss=0.167, simple_loss=0.2386, pruned_loss=0.04767, over 954537.07 frames. ], batch size: 36, lr: 3.11e-03, grad_scale: 32.0 2023-04-27 18:19:23,394 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.776e+01 1.488e+02 1.772e+02 2.253e+02 3.450e+02, threshold=3.544e+02, percent-clipped=0.0 2023-04-27 18:20:19,976 INFO [finetune.py:976] (1/7) Epoch 23, batch 1400, loss[loss=0.1588, simple_loss=0.2417, pruned_loss=0.03791, over 4904.00 frames. ], tot_loss[loss=0.1685, simple_loss=0.2412, pruned_loss=0.04791, over 954301.58 frames. ], batch size: 43, lr: 3.11e-03, grad_scale: 32.0 2023-04-27 18:20:43,132 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=127427.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 18:21:02,767 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6155, 1.8948, 1.8401, 2.1414, 2.1045, 2.1488, 1.8071, 3.6676], device='cuda:1'), covar=tensor([0.0510, 0.0642, 0.0617, 0.0949, 0.0484, 0.0490, 0.0594, 0.0155], device='cuda:1'), in_proj_covar=tensor([0.0037, 0.0038, 0.0039, 0.0042, 0.0040, 0.0037, 0.0037, 0.0054], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:1') 2023-04-27 18:21:15,607 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7675, 2.3681, 1.8780, 1.7654, 1.2802, 1.3034, 1.9050, 1.2563], device='cuda:1'), covar=tensor([0.1602, 0.1414, 0.1400, 0.1672, 0.2312, 0.1903, 0.0948, 0.1985], device='cuda:1'), in_proj_covar=tensor([0.0195, 0.0209, 0.0167, 0.0202, 0.0198, 0.0184, 0.0154, 0.0185], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-27 18:21:24,257 INFO [finetune.py:976] (1/7) Epoch 23, batch 1450, loss[loss=0.1565, simple_loss=0.2465, pruned_loss=0.03322, over 4875.00 frames. ], tot_loss[loss=0.1714, simple_loss=0.2441, pruned_loss=0.04937, over 952954.18 frames. ], batch size: 34, lr: 3.11e-03, grad_scale: 32.0 2023-04-27 18:21:34,074 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.128e+02 1.543e+02 1.916e+02 2.299e+02 4.494e+02, threshold=3.833e+02, percent-clipped=8.0 2023-04-27 18:21:43,899 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=127472.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:22:06,036 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=127488.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 18:22:29,896 INFO [finetune.py:976] (1/7) Epoch 23, batch 1500, loss[loss=0.1351, simple_loss=0.209, pruned_loss=0.03057, over 4692.00 frames. ], tot_loss[loss=0.1719, simple_loss=0.245, pruned_loss=0.04937, over 955234.40 frames. ], batch size: 23, lr: 3.11e-03, grad_scale: 32.0 2023-04-27 18:22:36,465 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8118, 2.2971, 1.8978, 2.2124, 1.6873, 1.8588, 1.8975, 1.4436], device='cuda:1'), covar=tensor([0.2125, 0.1336, 0.0883, 0.1243, 0.3208, 0.1303, 0.1919, 0.2554], device='cuda:1'), in_proj_covar=tensor([0.0281, 0.0295, 0.0212, 0.0273, 0.0309, 0.0252, 0.0245, 0.0259], device='cuda:1'), out_proj_covar=tensor([1.1212e-04, 1.1697e-04, 8.3515e-05, 1.0760e-04, 1.2488e-04, 9.9605e-05, 9.8769e-05, 1.0214e-04], device='cuda:1') 2023-04-27 18:22:37,636 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=127520.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:22:39,511 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8647, 2.3458, 0.9185, 1.2583, 1.6005, 1.2193, 2.4847, 1.3904], device='cuda:1'), covar=tensor([0.0669, 0.0539, 0.0646, 0.1261, 0.0456, 0.1010, 0.0307, 0.0685], device='cuda:1'), in_proj_covar=tensor([0.0050, 0.0065, 0.0047, 0.0046, 0.0050, 0.0051, 0.0073, 0.0051], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:1') 2023-04-27 18:22:43,214 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=127529.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:23:19,694 INFO [finetune.py:976] (1/7) Epoch 23, batch 1550, loss[loss=0.159, simple_loss=0.242, pruned_loss=0.03802, over 4892.00 frames. ], tot_loss[loss=0.1737, simple_loss=0.2469, pruned_loss=0.05024, over 954716.23 frames. ], batch size: 43, lr: 3.10e-03, grad_scale: 32.0 2023-04-27 18:23:23,843 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.013e+02 1.538e+02 1.819e+02 2.091e+02 4.341e+02, threshold=3.638e+02, percent-clipped=1.0 2023-04-27 18:23:41,775 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=127590.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:23:50,928 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=127605.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:23:53,256 INFO [finetune.py:976] (1/7) Epoch 23, batch 1600, loss[loss=0.1727, simple_loss=0.2388, pruned_loss=0.05326, over 4815.00 frames. ], tot_loss[loss=0.1714, simple_loss=0.2438, pruned_loss=0.04952, over 955542.45 frames. ], batch size: 41, lr: 3.10e-03, grad_scale: 32.0 2023-04-27 18:24:20,518 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.54 vs. limit=2.0 2023-04-27 18:24:26,892 INFO [finetune.py:976] (1/7) Epoch 23, batch 1650, loss[loss=0.1639, simple_loss=0.2328, pruned_loss=0.04748, over 4936.00 frames. ], tot_loss[loss=0.1685, simple_loss=0.2403, pruned_loss=0.04833, over 955978.01 frames. ], batch size: 38, lr: 3.10e-03, grad_scale: 32.0 2023-04-27 18:24:28,866 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=127662.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:24:31,045 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.001e+02 1.420e+02 1.883e+02 2.201e+02 3.441e+02, threshold=3.766e+02, percent-clipped=0.0 2023-04-27 18:25:00,765 INFO [finetune.py:976] (1/7) Epoch 23, batch 1700, loss[loss=0.2125, simple_loss=0.2687, pruned_loss=0.07813, over 4821.00 frames. ], tot_loss[loss=0.167, simple_loss=0.2386, pruned_loss=0.04768, over 956299.50 frames. ], batch size: 38, lr: 3.10e-03, grad_scale: 32.0 2023-04-27 18:25:10,334 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=127723.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:25:34,589 INFO [finetune.py:976] (1/7) Epoch 23, batch 1750, loss[loss=0.2035, simple_loss=0.2683, pruned_loss=0.06934, over 4926.00 frames. ], tot_loss[loss=0.1685, simple_loss=0.2402, pruned_loss=0.04843, over 955704.36 frames. ], batch size: 38, lr: 3.10e-03, grad_scale: 32.0 2023-04-27 18:25:38,229 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.809e+01 1.624e+02 1.849e+02 2.197e+02 5.063e+02, threshold=3.698e+02, percent-clipped=3.0 2023-04-27 18:25:40,852 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.3731, 3.0305, 2.3801, 2.3538, 1.7277, 1.6662, 2.4549, 1.6623], device='cuda:1'), covar=tensor([0.1556, 0.1267, 0.1237, 0.1535, 0.2087, 0.1846, 0.0872, 0.1896], device='cuda:1'), in_proj_covar=tensor([0.0196, 0.0209, 0.0167, 0.0202, 0.0197, 0.0184, 0.0154, 0.0185], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-27 18:25:46,619 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.0752, 1.2316, 3.0423, 2.8406, 2.7628, 2.9022, 2.8672, 2.6887], device='cuda:1'), covar=tensor([0.7048, 0.4815, 0.1326, 0.1733, 0.1236, 0.1929, 0.3587, 0.1671], device='cuda:1'), in_proj_covar=tensor([0.0309, 0.0307, 0.0405, 0.0408, 0.0348, 0.0410, 0.0314, 0.0365], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-27 18:25:48,413 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=127779.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:25:50,842 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=127783.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 18:25:56,112 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8913, 1.7147, 2.0001, 2.2287, 1.6711, 1.4839, 1.7149, 1.1182], device='cuda:1'), covar=tensor([0.0442, 0.0698, 0.0411, 0.0490, 0.0659, 0.1131, 0.0601, 0.0656], device='cuda:1'), in_proj_covar=tensor([0.0069, 0.0068, 0.0067, 0.0068, 0.0075, 0.0096, 0.0073, 0.0065], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-27 18:26:14,171 INFO [finetune.py:976] (1/7) Epoch 23, batch 1800, loss[loss=0.134, simple_loss=0.2086, pruned_loss=0.02973, over 4749.00 frames. ], tot_loss[loss=0.1707, simple_loss=0.2428, pruned_loss=0.04931, over 956018.28 frames. ], batch size: 27, lr: 3.10e-03, grad_scale: 32.0 2023-04-27 18:26:57,475 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=127840.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:27:09,550 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=127850.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:27:21,426 INFO [finetune.py:976] (1/7) Epoch 23, batch 1850, loss[loss=0.177, simple_loss=0.2667, pruned_loss=0.04363, over 4809.00 frames. ], tot_loss[loss=0.1719, simple_loss=0.2445, pruned_loss=0.0497, over 955709.11 frames. ], batch size: 45, lr: 3.10e-03, grad_scale: 32.0 2023-04-27 18:27:31,370 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.092e+02 1.632e+02 1.910e+02 2.193e+02 3.676e+02, threshold=3.820e+02, percent-clipped=0.0 2023-04-27 18:27:32,671 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.2872, 1.2224, 3.8337, 3.5574, 3.3910, 3.7251, 3.7357, 3.3264], device='cuda:1'), covar=tensor([0.7422, 0.5969, 0.1220, 0.2016, 0.1275, 0.1647, 0.1387, 0.1878], device='cuda:1'), in_proj_covar=tensor([0.0310, 0.0307, 0.0406, 0.0408, 0.0350, 0.0411, 0.0316, 0.0366], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-27 18:27:40,523 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=127872.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:27:44,162 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2023-04-27 18:27:49,850 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=127885.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:28:04,011 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=127905.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:28:06,378 INFO [finetune.py:976] (1/7) Epoch 23, batch 1900, loss[loss=0.1479, simple_loss=0.2296, pruned_loss=0.03315, over 4841.00 frames. ], tot_loss[loss=0.1725, simple_loss=0.2453, pruned_loss=0.04987, over 955290.53 frames. ], batch size: 49, lr: 3.10e-03, grad_scale: 32.0 2023-04-27 18:28:07,698 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=127911.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:28:32,853 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=127933.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:28:57,590 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=127953.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:29:07,612 INFO [finetune.py:976] (1/7) Epoch 23, batch 1950, loss[loss=0.1378, simple_loss=0.1956, pruned_loss=0.04001, over 4011.00 frames. ], tot_loss[loss=0.1714, simple_loss=0.2441, pruned_loss=0.04934, over 953154.90 frames. ], batch size: 17, lr: 3.10e-03, grad_scale: 32.0 2023-04-27 18:29:16,664 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.063e+02 1.493e+02 1.855e+02 2.257e+02 4.055e+02, threshold=3.710e+02, percent-clipped=1.0 2023-04-27 18:30:00,659 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([3.8779, 3.8519, 3.0712, 4.4923, 3.6780, 3.8561, 1.8838, 3.9467], device='cuda:1'), covar=tensor([0.1655, 0.1103, 0.3830, 0.1072, 0.3561, 0.1687, 0.5232, 0.1898], device='cuda:1'), in_proj_covar=tensor([0.0247, 0.0219, 0.0255, 0.0307, 0.0300, 0.0248, 0.0276, 0.0274], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 18:30:02,324 INFO [finetune.py:976] (1/7) Epoch 23, batch 2000, loss[loss=0.1383, simple_loss=0.205, pruned_loss=0.03576, over 4709.00 frames. ], tot_loss[loss=0.1703, simple_loss=0.2424, pruned_loss=0.04914, over 954355.02 frames. ], batch size: 23, lr: 3.10e-03, grad_scale: 32.0 2023-04-27 18:30:07,960 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=128018.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:30:15,989 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.1882, 2.4814, 1.0796, 1.4907, 2.0255, 1.2361, 3.4863, 1.7637], device='cuda:1'), covar=tensor([0.0632, 0.0647, 0.0816, 0.1264, 0.0511, 0.1016, 0.0245, 0.0638], device='cuda:1'), in_proj_covar=tensor([0.0050, 0.0064, 0.0047, 0.0046, 0.0050, 0.0051, 0.0073, 0.0050], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:1') 2023-04-27 18:30:35,022 INFO [finetune.py:976] (1/7) Epoch 23, batch 2050, loss[loss=0.1637, simple_loss=0.2378, pruned_loss=0.04481, over 4898.00 frames. ], tot_loss[loss=0.1678, simple_loss=0.2395, pruned_loss=0.04808, over 954656.28 frames. ], batch size: 32, lr: 3.10e-03, grad_scale: 32.0 2023-04-27 18:30:39,640 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.902e+01 1.493e+02 1.754e+02 2.070e+02 4.110e+02, threshold=3.508e+02, percent-clipped=2.0 2023-04-27 18:30:50,717 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=128083.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 18:31:08,824 INFO [finetune.py:976] (1/7) Epoch 23, batch 2100, loss[loss=0.2007, simple_loss=0.272, pruned_loss=0.06469, over 4871.00 frames. ], tot_loss[loss=0.1686, simple_loss=0.2396, pruned_loss=0.04878, over 955054.11 frames. ], batch size: 34, lr: 3.10e-03, grad_scale: 32.0 2023-04-27 18:31:22,850 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=128131.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 18:31:25,335 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=128135.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:31:42,238 INFO [finetune.py:976] (1/7) Epoch 23, batch 2150, loss[loss=0.1963, simple_loss=0.2667, pruned_loss=0.06289, over 4933.00 frames. ], tot_loss[loss=0.1703, simple_loss=0.242, pruned_loss=0.04933, over 955736.06 frames. ], batch size: 33, lr: 3.10e-03, grad_scale: 32.0 2023-04-27 18:31:46,840 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.015e+02 1.634e+02 1.958e+02 2.421e+02 4.804e+02, threshold=3.916e+02, percent-clipped=1.0 2023-04-27 18:31:59,141 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=128185.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:32:13,546 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=128206.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:32:15,346 INFO [finetune.py:976] (1/7) Epoch 23, batch 2200, loss[loss=0.1382, simple_loss=0.2224, pruned_loss=0.02696, over 4827.00 frames. ], tot_loss[loss=0.171, simple_loss=0.2436, pruned_loss=0.04917, over 955614.41 frames. ], batch size: 47, lr: 3.10e-03, grad_scale: 32.0 2023-04-27 18:32:39,396 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=128228.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:32:40,696 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.2210, 1.4367, 1.7702, 1.8508, 1.7811, 1.8557, 1.7596, 1.7760], device='cuda:1'), covar=tensor([0.3576, 0.5286, 0.4224, 0.4399, 0.5154, 0.6949, 0.4916, 0.4794], device='cuda:1'), in_proj_covar=tensor([0.0336, 0.0371, 0.0323, 0.0336, 0.0345, 0.0392, 0.0356, 0.0328], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 18:32:42,391 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=128233.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:33:12,840 INFO [finetune.py:976] (1/7) Epoch 23, batch 2250, loss[loss=0.1826, simple_loss=0.2594, pruned_loss=0.05285, over 4888.00 frames. ], tot_loss[loss=0.1716, simple_loss=0.2448, pruned_loss=0.04917, over 956684.79 frames. ], batch size: 35, lr: 3.10e-03, grad_scale: 32.0 2023-04-27 18:33:22,484 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.137e+02 1.532e+02 1.818e+02 2.293e+02 4.621e+02, threshold=3.635e+02, percent-clipped=2.0 2023-04-27 18:33:57,139 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7382, 2.1705, 1.1136, 1.4967, 2.1906, 1.5775, 1.5518, 1.7497], device='cuda:1'), covar=tensor([0.0502, 0.0341, 0.0282, 0.0554, 0.0225, 0.0500, 0.0495, 0.0546], device='cuda:1'), in_proj_covar=tensor([0.0028, 0.0024, 0.0022, 0.0028, 0.0019, 0.0028, 0.0027, 0.0028], device='cuda:1'), out_proj_covar=tensor([0.0051, 0.0046, 0.0038, 0.0052, 0.0038, 0.0050, 0.0050, 0.0051], device='cuda:1') 2023-04-27 18:34:26,340 INFO [finetune.py:976] (1/7) Epoch 23, batch 2300, loss[loss=0.1925, simple_loss=0.265, pruned_loss=0.06003, over 4896.00 frames. ], tot_loss[loss=0.1717, simple_loss=0.2456, pruned_loss=0.0489, over 954591.06 frames. ], batch size: 36, lr: 3.10e-03, grad_scale: 32.0 2023-04-27 18:34:37,303 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=128318.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:34:37,949 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8974, 1.7564, 2.2363, 2.3757, 1.7104, 1.5282, 1.8992, 1.0218], device='cuda:1'), covar=tensor([0.0604, 0.0767, 0.0457, 0.0696, 0.0821, 0.1151, 0.0633, 0.0782], device='cuda:1'), in_proj_covar=tensor([0.0069, 0.0068, 0.0067, 0.0068, 0.0075, 0.0095, 0.0073, 0.0064], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-27 18:34:39,025 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=128320.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:35:34,258 INFO [finetune.py:976] (1/7) Epoch 23, batch 2350, loss[loss=0.2166, simple_loss=0.2794, pruned_loss=0.0769, over 4814.00 frames. ], tot_loss[loss=0.1706, simple_loss=0.2435, pruned_loss=0.04883, over 954296.76 frames. ], batch size: 39, lr: 3.10e-03, grad_scale: 32.0 2023-04-27 18:35:37,973 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.014e+02 1.560e+02 1.785e+02 2.201e+02 3.854e+02, threshold=3.569e+02, percent-clipped=2.0 2023-04-27 18:35:38,685 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=128366.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:36:01,146 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=128381.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:36:31,442 INFO [finetune.py:976] (1/7) Epoch 23, batch 2400, loss[loss=0.1725, simple_loss=0.2402, pruned_loss=0.05238, over 4914.00 frames. ], tot_loss[loss=0.17, simple_loss=0.242, pruned_loss=0.04903, over 956213.17 frames. ], batch size: 43, lr: 3.10e-03, grad_scale: 32.0 2023-04-27 18:36:34,508 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=128413.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:36:58,489 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.2056, 3.0868, 0.9039, 1.6895, 1.7819, 2.3294, 1.7679, 1.0262], device='cuda:1'), covar=tensor([0.1695, 0.1392, 0.2187, 0.1473, 0.1230, 0.1121, 0.1697, 0.2149], device='cuda:1'), in_proj_covar=tensor([0.0116, 0.0238, 0.0137, 0.0119, 0.0131, 0.0150, 0.0117, 0.0118], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-27 18:37:01,394 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=128435.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:37:29,230 INFO [finetune.py:976] (1/7) Epoch 23, batch 2450, loss[loss=0.1397, simple_loss=0.2163, pruned_loss=0.03156, over 4856.00 frames. ], tot_loss[loss=0.1684, simple_loss=0.2399, pruned_loss=0.0485, over 953719.72 frames. ], batch size: 49, lr: 3.10e-03, grad_scale: 32.0 2023-04-27 18:37:35,177 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.697e+01 1.581e+02 1.977e+02 2.297e+02 4.168e+02, threshold=3.955e+02, percent-clipped=2.0 2023-04-27 18:37:46,742 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=128474.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:37:56,933 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=128483.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:38:27,889 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=128506.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:38:29,612 INFO [finetune.py:976] (1/7) Epoch 23, batch 2500, loss[loss=0.1642, simple_loss=0.2405, pruned_loss=0.044, over 4830.00 frames. ], tot_loss[loss=0.1698, simple_loss=0.241, pruned_loss=0.04925, over 953943.75 frames. ], batch size: 33, lr: 3.10e-03, grad_scale: 32.0 2023-04-27 18:38:53,265 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=128528.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:39:29,765 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=128554.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:39:32,694 INFO [finetune.py:976] (1/7) Epoch 23, batch 2550, loss[loss=0.208, simple_loss=0.2816, pruned_loss=0.06717, over 4830.00 frames. ], tot_loss[loss=0.1731, simple_loss=0.2452, pruned_loss=0.05053, over 954667.74 frames. ], batch size: 33, lr: 3.10e-03, grad_scale: 32.0 2023-04-27 18:39:42,963 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.076e+02 1.730e+02 1.875e+02 2.334e+02 4.772e+02, threshold=3.751e+02, percent-clipped=1.0 2023-04-27 18:39:55,794 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=128576.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:40:40,367 INFO [finetune.py:976] (1/7) Epoch 23, batch 2600, loss[loss=0.1618, simple_loss=0.2394, pruned_loss=0.04216, over 4905.00 frames. ], tot_loss[loss=0.1733, simple_loss=0.2462, pruned_loss=0.05018, over 954495.66 frames. ], batch size: 43, lr: 3.10e-03, grad_scale: 32.0 2023-04-27 18:41:01,857 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.3657, 2.7005, 1.1940, 1.6112, 2.1224, 1.5174, 3.6621, 2.1606], device='cuda:1'), covar=tensor([0.0608, 0.0617, 0.0750, 0.1194, 0.0481, 0.0919, 0.0216, 0.0544], device='cuda:1'), in_proj_covar=tensor([0.0051, 0.0065, 0.0048, 0.0047, 0.0050, 0.0052, 0.0073, 0.0051], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:1') 2023-04-27 18:41:44,290 INFO [finetune.py:976] (1/7) Epoch 23, batch 2650, loss[loss=0.1726, simple_loss=0.2473, pruned_loss=0.04894, over 4751.00 frames. ], tot_loss[loss=0.174, simple_loss=0.2475, pruned_loss=0.0503, over 954849.02 frames. ], batch size: 26, lr: 3.10e-03, grad_scale: 32.0 2023-04-27 18:41:44,367 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([3.6970, 3.6379, 2.8045, 4.2887, 3.6366, 3.6718, 1.6995, 3.6647], device='cuda:1'), covar=tensor([0.1663, 0.1339, 0.3579, 0.1624, 0.3067, 0.1728, 0.5371, 0.2328], device='cuda:1'), in_proj_covar=tensor([0.0247, 0.0220, 0.0255, 0.0308, 0.0299, 0.0248, 0.0277, 0.0276], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 18:41:53,484 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.191e+02 1.651e+02 1.881e+02 2.286e+02 5.160e+02, threshold=3.761e+02, percent-clipped=1.0 2023-04-27 18:42:11,798 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=128676.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:42:55,231 INFO [finetune.py:976] (1/7) Epoch 23, batch 2700, loss[loss=0.1616, simple_loss=0.2317, pruned_loss=0.04576, over 4932.00 frames. ], tot_loss[loss=0.1719, simple_loss=0.2453, pruned_loss=0.04928, over 953360.26 frames. ], batch size: 33, lr: 3.10e-03, grad_scale: 32.0 2023-04-27 18:43:52,305 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6819, 2.3409, 1.6116, 1.6029, 1.2906, 1.3253, 1.6169, 1.2467], device='cuda:1'), covar=tensor([0.1816, 0.1247, 0.1662, 0.1837, 0.2470, 0.2253, 0.1056, 0.2177], device='cuda:1'), in_proj_covar=tensor([0.0197, 0.0211, 0.0169, 0.0205, 0.0199, 0.0186, 0.0156, 0.0187], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-27 18:44:00,967 INFO [finetune.py:976] (1/7) Epoch 23, batch 2750, loss[loss=0.154, simple_loss=0.2336, pruned_loss=0.03715, over 4778.00 frames. ], tot_loss[loss=0.17, simple_loss=0.2429, pruned_loss=0.04858, over 952902.94 frames. ], batch size: 26, lr: 3.10e-03, grad_scale: 32.0 2023-04-27 18:44:04,649 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 8.255e+01 1.547e+02 1.823e+02 2.138e+02 3.437e+02, threshold=3.646e+02, percent-clipped=0.0 2023-04-27 18:44:12,834 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=128769.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:44:27,138 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=128781.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:44:45,355 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4924, 0.9169, 0.4813, 1.1896, 1.0778, 1.3522, 1.2502, 1.2442], device='cuda:1'), covar=tensor([0.0504, 0.0414, 0.0384, 0.0576, 0.0296, 0.0518, 0.0490, 0.0583], device='cuda:1'), in_proj_covar=tensor([0.0028, 0.0024, 0.0022, 0.0028, 0.0019, 0.0028, 0.0027, 0.0028], device='cuda:1'), out_proj_covar=tensor([0.0051, 0.0046, 0.0038, 0.0052, 0.0039, 0.0050, 0.0050, 0.0052], device='cuda:1') 2023-04-27 18:45:09,454 INFO [finetune.py:976] (1/7) Epoch 23, batch 2800, loss[loss=0.1308, simple_loss=0.2053, pruned_loss=0.0282, over 4868.00 frames. ], tot_loss[loss=0.1671, simple_loss=0.2395, pruned_loss=0.04735, over 954331.55 frames. ], batch size: 31, lr: 3.10e-03, grad_scale: 32.0 2023-04-27 18:45:55,110 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=128842.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:46:15,066 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-04-27 18:46:17,709 INFO [finetune.py:976] (1/7) Epoch 23, batch 2850, loss[loss=0.1954, simple_loss=0.2741, pruned_loss=0.05831, over 4825.00 frames. ], tot_loss[loss=0.1653, simple_loss=0.237, pruned_loss=0.04684, over 953077.65 frames. ], batch size: 39, lr: 3.10e-03, grad_scale: 32.0 2023-04-27 18:46:22,646 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.077e+02 1.467e+02 1.826e+02 2.202e+02 4.194e+02, threshold=3.652e+02, percent-clipped=1.0 2023-04-27 18:46:32,685 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6625, 2.1174, 1.6173, 1.4208, 1.2871, 1.2775, 1.6271, 1.1729], device='cuda:1'), covar=tensor([0.1598, 0.1265, 0.1363, 0.1637, 0.2158, 0.1870, 0.0944, 0.1990], device='cuda:1'), in_proj_covar=tensor([0.0197, 0.0210, 0.0169, 0.0205, 0.0199, 0.0186, 0.0156, 0.0186], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-27 18:46:43,958 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([3.4320, 3.3557, 2.5528, 3.9187, 3.4162, 3.3864, 1.2832, 3.3817], device='cuda:1'), covar=tensor([0.1685, 0.1485, 0.3086, 0.2027, 0.3932, 0.1719, 0.5974, 0.2526], device='cuda:1'), in_proj_covar=tensor([0.0243, 0.0218, 0.0251, 0.0304, 0.0295, 0.0245, 0.0273, 0.0271], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 18:46:50,215 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=128890.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 18:47:11,862 INFO [finetune.py:976] (1/7) Epoch 23, batch 2900, loss[loss=0.1832, simple_loss=0.2569, pruned_loss=0.05477, over 4927.00 frames. ], tot_loss[loss=0.1679, simple_loss=0.2401, pruned_loss=0.04785, over 954349.49 frames. ], batch size: 33, lr: 3.09e-03, grad_scale: 32.0 2023-04-27 18:47:17,831 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9925, 3.1398, 2.6545, 2.8903, 3.2641, 2.8162, 4.1034, 2.4279], device='cuda:1'), covar=tensor([0.3071, 0.1570, 0.2937, 0.2281, 0.1391, 0.2111, 0.0930, 0.3145], device='cuda:1'), in_proj_covar=tensor([0.0338, 0.0348, 0.0426, 0.0351, 0.0377, 0.0375, 0.0368, 0.0419], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 18:47:36,711 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=3.71 vs. limit=5.0 2023-04-27 18:47:56,530 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=128951.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 18:48:04,747 INFO [finetune.py:976] (1/7) Epoch 23, batch 2950, loss[loss=0.1745, simple_loss=0.2503, pruned_loss=0.0493, over 4886.00 frames. ], tot_loss[loss=0.1709, simple_loss=0.2435, pruned_loss=0.04913, over 950975.98 frames. ], batch size: 35, lr: 3.09e-03, grad_scale: 32.0 2023-04-27 18:48:10,254 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.058e+02 1.560e+02 1.806e+02 2.158e+02 5.289e+02, threshold=3.612e+02, percent-clipped=5.0 2023-04-27 18:48:22,821 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=128976.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:48:48,718 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.3115, 1.5384, 1.8769, 1.9563, 1.9119, 1.9377, 1.8852, 1.9033], device='cuda:1'), covar=tensor([0.3930, 0.4605, 0.4091, 0.4305, 0.4898, 0.6392, 0.4465, 0.4226], device='cuda:1'), in_proj_covar=tensor([0.0337, 0.0372, 0.0324, 0.0336, 0.0346, 0.0392, 0.0355, 0.0328], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 18:49:01,104 INFO [finetune.py:976] (1/7) Epoch 23, batch 3000, loss[loss=0.1559, simple_loss=0.2225, pruned_loss=0.0446, over 4409.00 frames. ], tot_loss[loss=0.1722, simple_loss=0.2449, pruned_loss=0.0498, over 953024.67 frames. ], batch size: 19, lr: 3.09e-03, grad_scale: 32.0 2023-04-27 18:49:01,104 INFO [finetune.py:1001] (1/7) Computing validation loss 2023-04-27 18:49:17,637 INFO [finetune.py:1010] (1/7) Epoch 23, validation: loss=0.1527, simple_loss=0.2222, pruned_loss=0.04158, over 2265189.00 frames. 2023-04-27 18:49:17,638 INFO [finetune.py:1011] (1/7) Maximum memory allocated so far is 6391MB 2023-04-27 18:49:31,984 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=129024.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:49:51,611 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5433, 3.4779, 0.8475, 1.8120, 1.8348, 2.4352, 1.8897, 0.9428], device='cuda:1'), covar=tensor([0.1382, 0.0933, 0.2125, 0.1219, 0.1044, 0.1020, 0.1478, 0.2051], device='cuda:1'), in_proj_covar=tensor([0.0116, 0.0236, 0.0136, 0.0119, 0.0130, 0.0150, 0.0116, 0.0117], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-27 18:50:16,853 INFO [finetune.py:976] (1/7) Epoch 23, batch 3050, loss[loss=0.1297, simple_loss=0.1991, pruned_loss=0.03014, over 3892.00 frames. ], tot_loss[loss=0.1722, simple_loss=0.2456, pruned_loss=0.04935, over 954168.26 frames. ], batch size: 17, lr: 3.09e-03, grad_scale: 32.0 2023-04-27 18:50:25,142 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.041e+02 1.682e+02 1.957e+02 2.360e+02 4.958e+02, threshold=3.913e+02, percent-clipped=5.0 2023-04-27 18:50:34,465 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=129069.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:51:09,073 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([0.5538, 1.4299, 1.4092, 1.1101, 1.3104, 1.2881, 1.6795, 1.2304], device='cuda:1'), covar=tensor([0.3095, 0.1731, 0.4871, 0.2398, 0.1578, 0.1955, 0.1490, 0.5018], device='cuda:1'), in_proj_covar=tensor([0.0339, 0.0348, 0.0426, 0.0351, 0.0378, 0.0374, 0.0367, 0.0420], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 18:51:30,728 INFO [finetune.py:976] (1/7) Epoch 23, batch 3100, loss[loss=0.1617, simple_loss=0.2455, pruned_loss=0.03891, over 4821.00 frames. ], tot_loss[loss=0.1711, simple_loss=0.2443, pruned_loss=0.04899, over 955625.57 frames. ], batch size: 38, lr: 3.09e-03, grad_scale: 16.0 2023-04-27 18:51:41,686 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=129117.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:52:07,895 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=129137.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:52:39,821 INFO [finetune.py:976] (1/7) Epoch 23, batch 3150, loss[loss=0.1499, simple_loss=0.2235, pruned_loss=0.03812, over 4906.00 frames. ], tot_loss[loss=0.1704, simple_loss=0.2426, pruned_loss=0.0491, over 955491.59 frames. ], batch size: 37, lr: 3.09e-03, grad_scale: 16.0 2023-04-27 18:52:50,086 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.045e+02 1.492e+02 1.784e+02 2.330e+02 3.707e+02, threshold=3.568e+02, percent-clipped=0.0 2023-04-27 18:53:03,022 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.3653, 1.5392, 1.7947, 1.9270, 1.8552, 1.9376, 1.8293, 1.8891], device='cuda:1'), covar=tensor([0.3587, 0.5169, 0.3974, 0.4107, 0.4963, 0.6697, 0.4847, 0.4366], device='cuda:1'), in_proj_covar=tensor([0.0337, 0.0372, 0.0324, 0.0336, 0.0345, 0.0391, 0.0355, 0.0327], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 18:53:22,804 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.2003, 1.4969, 1.4180, 1.8656, 1.6381, 1.7667, 1.3459, 3.0989], device='cuda:1'), covar=tensor([0.0666, 0.0832, 0.0820, 0.1206, 0.0624, 0.0459, 0.0797, 0.0197], device='cuda:1'), in_proj_covar=tensor([0.0037, 0.0039, 0.0039, 0.0043, 0.0040, 0.0038, 0.0038, 0.0055], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0013, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:1') 2023-04-27 18:53:46,920 INFO [finetune.py:976] (1/7) Epoch 23, batch 3200, loss[loss=0.1146, simple_loss=0.1942, pruned_loss=0.01753, over 4753.00 frames. ], tot_loss[loss=0.1663, simple_loss=0.2382, pruned_loss=0.0472, over 955879.68 frames. ], batch size: 27, lr: 3.09e-03, grad_scale: 16.0 2023-04-27 18:54:42,102 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=129246.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 18:54:55,919 INFO [finetune.py:976] (1/7) Epoch 23, batch 3250, loss[loss=0.2036, simple_loss=0.2642, pruned_loss=0.07145, over 4872.00 frames. ], tot_loss[loss=0.1677, simple_loss=0.2391, pruned_loss=0.04819, over 955386.96 frames. ], batch size: 34, lr: 3.09e-03, grad_scale: 16.0 2023-04-27 18:54:58,815 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4237, 1.2417, 4.1025, 3.8292, 3.6195, 3.8995, 3.9179, 3.5744], device='cuda:1'), covar=tensor([0.7265, 0.6083, 0.1134, 0.2120, 0.1110, 0.1701, 0.1435, 0.1752], device='cuda:1'), in_proj_covar=tensor([0.0311, 0.0308, 0.0408, 0.0410, 0.0350, 0.0413, 0.0317, 0.0369], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-27 18:55:06,949 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 8.754e+01 1.533e+02 1.786e+02 2.239e+02 3.821e+02, threshold=3.572e+02, percent-clipped=2.0 2023-04-27 18:55:42,376 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=2.07 vs. limit=2.0 2023-04-27 18:55:51,326 INFO [finetune.py:976] (1/7) Epoch 23, batch 3300, loss[loss=0.1741, simple_loss=0.2601, pruned_loss=0.04412, over 4922.00 frames. ], tot_loss[loss=0.1696, simple_loss=0.2414, pruned_loss=0.04885, over 953105.72 frames. ], batch size: 42, lr: 3.09e-03, grad_scale: 16.0 2023-04-27 18:56:51,618 INFO [finetune.py:976] (1/7) Epoch 23, batch 3350, loss[loss=0.1935, simple_loss=0.2651, pruned_loss=0.06091, over 4789.00 frames. ], tot_loss[loss=0.1726, simple_loss=0.2448, pruned_loss=0.05018, over 954401.78 frames. ], batch size: 51, lr: 3.09e-03, grad_scale: 16.0 2023-04-27 18:57:02,542 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.057e+02 1.708e+02 1.958e+02 2.263e+02 4.173e+02, threshold=3.917e+02, percent-clipped=1.0 2023-04-27 18:57:58,595 INFO [finetune.py:976] (1/7) Epoch 23, batch 3400, loss[loss=0.2217, simple_loss=0.2798, pruned_loss=0.08181, over 4872.00 frames. ], tot_loss[loss=0.1734, simple_loss=0.246, pruned_loss=0.05036, over 954242.92 frames. ], batch size: 34, lr: 3.09e-03, grad_scale: 16.0 2023-04-27 18:58:06,854 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=129413.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 18:58:40,739 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=129437.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:59:06,106 INFO [finetune.py:976] (1/7) Epoch 23, batch 3450, loss[loss=0.1597, simple_loss=0.2418, pruned_loss=0.03884, over 4850.00 frames. ], tot_loss[loss=0.1724, simple_loss=0.2455, pruned_loss=0.04971, over 953818.71 frames. ], batch size: 49, lr: 3.09e-03, grad_scale: 16.0 2023-04-27 18:59:15,877 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 8.992e+01 1.559e+02 1.867e+02 2.155e+02 3.707e+02, threshold=3.734e+02, percent-clipped=0.0 2023-04-27 18:59:26,957 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=129474.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 18:59:39,676 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=129485.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:00:07,510 INFO [finetune.py:976] (1/7) Epoch 23, batch 3500, loss[loss=0.1595, simple_loss=0.2451, pruned_loss=0.0369, over 4812.00 frames. ], tot_loss[loss=0.1707, simple_loss=0.2437, pruned_loss=0.04887, over 954122.21 frames. ], batch size: 40, lr: 3.09e-03, grad_scale: 16.0 2023-04-27 19:00:46,494 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=129546.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 19:01:00,754 INFO [finetune.py:976] (1/7) Epoch 23, batch 3550, loss[loss=0.1622, simple_loss=0.2348, pruned_loss=0.0448, over 4825.00 frames. ], tot_loss[loss=0.1684, simple_loss=0.2406, pruned_loss=0.0481, over 954602.30 frames. ], batch size: 38, lr: 3.09e-03, grad_scale: 16.0 2023-04-27 19:01:08,387 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.192e+01 1.610e+02 1.893e+02 2.308e+02 5.470e+02, threshold=3.785e+02, percent-clipped=3.0 2023-04-27 19:01:14,605 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.3265, 3.0260, 0.8780, 1.6049, 1.7097, 2.0940, 1.7993, 1.0388], device='cuda:1'), covar=tensor([0.1347, 0.0962, 0.1882, 0.1199, 0.1040, 0.0980, 0.1391, 0.1703], device='cuda:1'), in_proj_covar=tensor([0.0117, 0.0239, 0.0136, 0.0120, 0.0132, 0.0151, 0.0117, 0.0119], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-27 19:01:38,374 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=129594.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 19:01:38,402 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.3890, 3.0943, 0.8790, 1.5517, 2.3756, 1.5745, 4.3621, 1.9913], device='cuda:1'), covar=tensor([0.0750, 0.0866, 0.1003, 0.1473, 0.0575, 0.1108, 0.0230, 0.0702], device='cuda:1'), in_proj_covar=tensor([0.0050, 0.0064, 0.0047, 0.0046, 0.0049, 0.0051, 0.0072, 0.0050], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0009, 0.0007, 0.0007, 0.0008, 0.0008, 0.0010, 0.0007], device='cuda:1') 2023-04-27 19:01:57,767 INFO [finetune.py:976] (1/7) Epoch 23, batch 3600, loss[loss=0.1497, simple_loss=0.2116, pruned_loss=0.04388, over 4889.00 frames. ], tot_loss[loss=0.1673, simple_loss=0.2385, pruned_loss=0.04805, over 955080.86 frames. ], batch size: 32, lr: 3.09e-03, grad_scale: 16.0 2023-04-27 19:01:59,687 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=129612.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:02:42,840 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-04-27 19:02:47,465 INFO [finetune.py:976] (1/7) Epoch 23, batch 3650, loss[loss=0.1523, simple_loss=0.2356, pruned_loss=0.03449, over 4716.00 frames. ], tot_loss[loss=0.169, simple_loss=0.2401, pruned_loss=0.04894, over 954190.98 frames. ], batch size: 23, lr: 3.09e-03, grad_scale: 16.0 2023-04-27 19:02:51,895 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.776e+01 1.523e+02 1.795e+02 2.282e+02 4.473e+02, threshold=3.590e+02, percent-clipped=4.0 2023-04-27 19:03:02,051 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=129673.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:03:26,419 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.1170, 0.7884, 0.9362, 0.7883, 1.2255, 0.9815, 0.8835, 0.9857], device='cuda:1'), covar=tensor([0.1927, 0.1739, 0.2359, 0.1947, 0.1194, 0.1860, 0.2146, 0.2912], device='cuda:1'), in_proj_covar=tensor([0.0314, 0.0308, 0.0350, 0.0286, 0.0324, 0.0309, 0.0299, 0.0371], device='cuda:1'), out_proj_covar=tensor([6.4415e-05, 6.3703e-05, 7.3705e-05, 5.7444e-05, 6.6646e-05, 6.4713e-05, 6.2408e-05, 7.8755e-05], device='cuda:1') 2023-04-27 19:03:55,979 INFO [finetune.py:976] (1/7) Epoch 23, batch 3700, loss[loss=0.2209, simple_loss=0.2974, pruned_loss=0.07226, over 4862.00 frames. ], tot_loss[loss=0.1717, simple_loss=0.2439, pruned_loss=0.04977, over 954258.36 frames. ], batch size: 34, lr: 3.09e-03, grad_scale: 16.0 2023-04-27 19:04:39,371 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9034, 1.5454, 1.4550, 1.6994, 2.1437, 1.7400, 1.4571, 1.3887], device='cuda:1'), covar=tensor([0.1585, 0.1587, 0.1909, 0.1313, 0.0882, 0.1626, 0.2371, 0.2330], device='cuda:1'), in_proj_covar=tensor([0.0315, 0.0309, 0.0351, 0.0286, 0.0324, 0.0309, 0.0300, 0.0372], device='cuda:1'), out_proj_covar=tensor([6.4563e-05, 6.3854e-05, 7.3962e-05, 5.7529e-05, 6.6758e-05, 6.4824e-05, 6.2566e-05, 7.9025e-05], device='cuda:1') 2023-04-27 19:04:52,813 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5329, 1.3362, 0.5901, 1.2378, 1.4122, 1.3775, 1.3168, 1.3510], device='cuda:1'), covar=tensor([0.0516, 0.0400, 0.0381, 0.0561, 0.0296, 0.0508, 0.0503, 0.0604], device='cuda:1'), in_proj_covar=tensor([0.0028, 0.0023, 0.0022, 0.0028, 0.0019, 0.0027, 0.0027, 0.0028], device='cuda:1'), out_proj_covar=tensor([0.0051, 0.0045, 0.0038, 0.0052, 0.0038, 0.0050, 0.0050, 0.0051], device='cuda:1') 2023-04-27 19:05:03,363 INFO [finetune.py:976] (1/7) Epoch 23, batch 3750, loss[loss=0.1491, simple_loss=0.2231, pruned_loss=0.03751, over 4768.00 frames. ], tot_loss[loss=0.1723, simple_loss=0.2452, pruned_loss=0.04967, over 954300.21 frames. ], batch size: 26, lr: 3.09e-03, grad_scale: 16.0 2023-04-27 19:05:12,865 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.105e+02 1.590e+02 1.958e+02 2.395e+02 5.557e+02, threshold=3.915e+02, percent-clipped=3.0 2023-04-27 19:05:15,279 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=129769.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 19:05:45,709 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=129791.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:05:46,351 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7932, 2.0021, 0.8102, 1.1966, 1.5277, 1.1508, 2.2431, 1.3698], device='cuda:1'), covar=tensor([0.0588, 0.0718, 0.0684, 0.1110, 0.0390, 0.0841, 0.0286, 0.0555], device='cuda:1'), in_proj_covar=tensor([0.0050, 0.0064, 0.0047, 0.0046, 0.0049, 0.0051, 0.0072, 0.0050], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0009, 0.0007, 0.0007, 0.0008, 0.0008, 0.0010, 0.0007], device='cuda:1') 2023-04-27 19:06:08,151 INFO [finetune.py:976] (1/7) Epoch 23, batch 3800, loss[loss=0.146, simple_loss=0.2119, pruned_loss=0.04001, over 4690.00 frames. ], tot_loss[loss=0.1729, simple_loss=0.2458, pruned_loss=0.05, over 953308.85 frames. ], batch size: 23, lr: 3.09e-03, grad_scale: 16.0 2023-04-27 19:07:04,487 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=129852.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:07:13,928 INFO [finetune.py:976] (1/7) Epoch 23, batch 3850, loss[loss=0.1826, simple_loss=0.2476, pruned_loss=0.05886, over 4832.00 frames. ], tot_loss[loss=0.1708, simple_loss=0.2439, pruned_loss=0.04888, over 953219.31 frames. ], batch size: 30, lr: 3.09e-03, grad_scale: 16.0 2023-04-27 19:07:23,061 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9111, 2.5405, 0.9837, 1.2733, 1.9014, 1.0572, 3.3748, 1.5763], device='cuda:1'), covar=tensor([0.0940, 0.0887, 0.1028, 0.1598, 0.0636, 0.1388, 0.0293, 0.0870], device='cuda:1'), in_proj_covar=tensor([0.0050, 0.0064, 0.0047, 0.0046, 0.0049, 0.0051, 0.0072, 0.0050], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0009, 0.0007, 0.0007, 0.0008, 0.0008, 0.0010, 0.0007], device='cuda:1') 2023-04-27 19:07:24,864 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.155e+02 1.706e+02 1.886e+02 2.157e+02 5.437e+02, threshold=3.771e+02, percent-clipped=4.0 2023-04-27 19:08:11,818 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4059, 1.1660, 0.4826, 1.1605, 1.1586, 1.2880, 1.2385, 1.2301], device='cuda:1'), covar=tensor([0.0421, 0.0308, 0.0394, 0.0462, 0.0324, 0.0415, 0.0385, 0.0467], device='cuda:1'), in_proj_covar=tensor([0.0028, 0.0024, 0.0022, 0.0028, 0.0019, 0.0028, 0.0027, 0.0028], device='cuda:1'), out_proj_covar=tensor([0.0051, 0.0046, 0.0038, 0.0052, 0.0039, 0.0050, 0.0050, 0.0051], device='cuda:1') 2023-04-27 19:08:15,871 INFO [finetune.py:976] (1/7) Epoch 23, batch 3900, loss[loss=0.1306, simple_loss=0.2105, pruned_loss=0.02532, over 4776.00 frames. ], tot_loss[loss=0.1687, simple_loss=0.2409, pruned_loss=0.04825, over 953054.76 frames. ], batch size: 29, lr: 3.09e-03, grad_scale: 16.0 2023-04-27 19:08:51,741 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6846, 1.5905, 0.7270, 1.3596, 1.6780, 1.5088, 1.4208, 1.4924], device='cuda:1'), covar=tensor([0.0487, 0.0369, 0.0353, 0.0569, 0.0266, 0.0499, 0.0498, 0.0572], device='cuda:1'), in_proj_covar=tensor([0.0028, 0.0023, 0.0022, 0.0028, 0.0019, 0.0028, 0.0027, 0.0028], device='cuda:1'), out_proj_covar=tensor([0.0051, 0.0045, 0.0038, 0.0052, 0.0038, 0.0050, 0.0050, 0.0051], device='cuda:1') 2023-04-27 19:09:17,305 INFO [finetune.py:976] (1/7) Epoch 23, batch 3950, loss[loss=0.1363, simple_loss=0.2142, pruned_loss=0.02917, over 4739.00 frames. ], tot_loss[loss=0.1657, simple_loss=0.2375, pruned_loss=0.04696, over 956155.60 frames. ], batch size: 27, lr: 3.09e-03, grad_scale: 16.0 2023-04-27 19:09:27,940 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.540e+01 1.458e+02 1.749e+02 1.978e+02 4.242e+02, threshold=3.497e+02, percent-clipped=1.0 2023-04-27 19:09:35,335 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=129968.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:10:16,049 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=130000.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:10:22,144 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=130002.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:10:33,621 INFO [finetune.py:976] (1/7) Epoch 23, batch 4000, loss[loss=0.186, simple_loss=0.264, pruned_loss=0.05399, over 4927.00 frames. ], tot_loss[loss=0.1675, simple_loss=0.2385, pruned_loss=0.04823, over 956238.82 frames. ], batch size: 33, lr: 3.09e-03, grad_scale: 16.0 2023-04-27 19:11:22,103 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9149, 2.2752, 2.1018, 2.3295, 1.7097, 2.0302, 1.9831, 1.5679], device='cuda:1'), covar=tensor([0.2013, 0.1398, 0.0697, 0.1140, 0.3019, 0.1040, 0.1921, 0.2508], device='cuda:1'), in_proj_covar=tensor([0.0281, 0.0299, 0.0212, 0.0273, 0.0311, 0.0253, 0.0244, 0.0259], device='cuda:1'), out_proj_covar=tensor([1.1237e-04, 1.1825e-04, 8.3443e-05, 1.0760e-04, 1.2543e-04, 9.9958e-05, 9.8535e-05, 1.0238e-04], device='cuda:1') 2023-04-27 19:11:43,267 INFO [finetune.py:976] (1/7) Epoch 23, batch 4050, loss[loss=0.1699, simple_loss=0.2544, pruned_loss=0.04269, over 4759.00 frames. ], tot_loss[loss=0.1704, simple_loss=0.2423, pruned_loss=0.04932, over 956569.71 frames. ], batch size: 54, lr: 3.09e-03, grad_scale: 16.0 2023-04-27 19:11:44,627 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=130061.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:11:45,892 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=130063.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 19:11:53,610 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.121e+02 1.688e+02 1.937e+02 2.332e+02 4.574e+02, threshold=3.874e+02, percent-clipped=1.0 2023-04-27 19:11:56,081 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=130069.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 19:12:52,034 INFO [finetune.py:976] (1/7) Epoch 23, batch 4100, loss[loss=0.1672, simple_loss=0.2353, pruned_loss=0.04961, over 4711.00 frames. ], tot_loss[loss=0.1717, simple_loss=0.2441, pruned_loss=0.0497, over 956624.03 frames. ], batch size: 23, lr: 3.09e-03, grad_scale: 16.0 2023-04-27 19:13:01,825 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=130117.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 19:13:06,863 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-04-27 19:13:21,674 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7073, 3.4781, 1.0284, 1.8894, 2.1015, 2.4156, 1.9899, 1.0222], device='cuda:1'), covar=tensor([0.1151, 0.0842, 0.1903, 0.1136, 0.0898, 0.1018, 0.1439, 0.1897], device='cuda:1'), in_proj_covar=tensor([0.0116, 0.0237, 0.0136, 0.0119, 0.0131, 0.0151, 0.0116, 0.0118], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-27 19:13:32,090 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=130147.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:13:51,930 INFO [finetune.py:976] (1/7) Epoch 23, batch 4150, loss[loss=0.1322, simple_loss=0.2133, pruned_loss=0.02558, over 4891.00 frames. ], tot_loss[loss=0.1738, simple_loss=0.2464, pruned_loss=0.05058, over 956007.72 frames. ], batch size: 32, lr: 3.09e-03, grad_scale: 16.0 2023-04-27 19:14:02,699 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.039e+02 1.725e+02 1.972e+02 2.342e+02 5.183e+02, threshold=3.944e+02, percent-clipped=4.0 2023-04-27 19:14:12,071 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4369, 2.8583, 1.1682, 1.7817, 1.7824, 2.3413, 1.8443, 1.2220], device='cuda:1'), covar=tensor([0.1112, 0.0782, 0.1497, 0.1028, 0.0862, 0.0726, 0.1233, 0.1651], device='cuda:1'), in_proj_covar=tensor([0.0116, 0.0237, 0.0136, 0.0119, 0.0131, 0.0150, 0.0116, 0.0118], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-27 19:14:13,479 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-04-27 19:14:38,930 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8995, 4.3397, 0.9116, 2.1707, 2.5167, 2.9743, 2.4943, 1.0637], device='cuda:1'), covar=tensor([0.1387, 0.0724, 0.2176, 0.1195, 0.0929, 0.1011, 0.1412, 0.2070], device='cuda:1'), in_proj_covar=tensor([0.0116, 0.0237, 0.0136, 0.0119, 0.0131, 0.0150, 0.0116, 0.0118], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-27 19:14:48,093 INFO [finetune.py:976] (1/7) Epoch 23, batch 4200, loss[loss=0.1433, simple_loss=0.2291, pruned_loss=0.02875, over 4792.00 frames. ], tot_loss[loss=0.173, simple_loss=0.2465, pruned_loss=0.04972, over 954958.98 frames. ], batch size: 51, lr: 3.09e-03, grad_scale: 16.0 2023-04-27 19:15:51,651 INFO [finetune.py:976] (1/7) Epoch 23, batch 4250, loss[loss=0.1583, simple_loss=0.2303, pruned_loss=0.0432, over 4904.00 frames. ], tot_loss[loss=0.1712, simple_loss=0.2443, pruned_loss=0.04903, over 955698.32 frames. ], batch size: 36, lr: 3.08e-03, grad_scale: 16.0 2023-04-27 19:16:01,964 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.003e+02 1.509e+02 1.751e+02 2.292e+02 3.450e+02, threshold=3.503e+02, percent-clipped=0.0 2023-04-27 19:16:03,352 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=130268.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:16:04,681 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.73 vs. limit=2.0 2023-04-27 19:16:11,026 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.73 vs. limit=2.0 2023-04-27 19:16:55,315 INFO [finetune.py:976] (1/7) Epoch 23, batch 4300, loss[loss=0.1447, simple_loss=0.2179, pruned_loss=0.0357, over 4892.00 frames. ], tot_loss[loss=0.169, simple_loss=0.2414, pruned_loss=0.04827, over 957237.75 frames. ], batch size: 32, lr: 3.08e-03, grad_scale: 16.0 2023-04-27 19:17:05,110 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=130316.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:17:58,666 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=130356.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:17:59,934 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=130358.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 19:18:05,660 INFO [finetune.py:976] (1/7) Epoch 23, batch 4350, loss[loss=0.1776, simple_loss=0.2413, pruned_loss=0.05697, over 4743.00 frames. ], tot_loss[loss=0.1669, simple_loss=0.2388, pruned_loss=0.04749, over 956751.37 frames. ], batch size: 54, lr: 3.08e-03, grad_scale: 16.0 2023-04-27 19:18:05,749 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5073, 1.3918, 4.1880, 3.9130, 3.6851, 3.9686, 4.0043, 3.6560], device='cuda:1'), covar=tensor([0.7058, 0.5937, 0.1177, 0.1885, 0.1291, 0.1988, 0.1175, 0.1784], device='cuda:1'), in_proj_covar=tensor([0.0311, 0.0308, 0.0408, 0.0408, 0.0348, 0.0411, 0.0316, 0.0368], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-27 19:18:10,118 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.083e+02 1.587e+02 1.809e+02 2.203e+02 4.398e+02, threshold=3.619e+02, percent-clipped=3.0 2023-04-27 19:18:10,515 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2023-04-27 19:18:29,070 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=130384.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:18:40,615 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4376, 1.3396, 3.9181, 3.6505, 3.4756, 3.6828, 3.6195, 3.4139], device='cuda:1'), covar=tensor([0.7697, 0.5615, 0.1084, 0.1756, 0.1181, 0.1843, 0.2700, 0.1771], device='cuda:1'), in_proj_covar=tensor([0.0312, 0.0308, 0.0408, 0.0409, 0.0348, 0.0411, 0.0317, 0.0369], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-27 19:19:02,657 INFO [finetune.py:976] (1/7) Epoch 23, batch 4400, loss[loss=0.1574, simple_loss=0.2407, pruned_loss=0.03707, over 4806.00 frames. ], tot_loss[loss=0.169, simple_loss=0.2406, pruned_loss=0.04871, over 955358.63 frames. ], batch size: 45, lr: 3.08e-03, grad_scale: 16.0 2023-04-27 19:19:11,773 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-04-27 19:19:34,764 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=130432.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:19:45,122 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=130438.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:19:55,329 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=130445.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:19:56,583 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=130447.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:20:05,833 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6424, 1.1420, 4.4360, 4.1762, 3.9083, 4.2763, 4.1026, 3.9454], device='cuda:1'), covar=tensor([0.7163, 0.6533, 0.1012, 0.1616, 0.1148, 0.1824, 0.1829, 0.1536], device='cuda:1'), in_proj_covar=tensor([0.0313, 0.0310, 0.0410, 0.0410, 0.0350, 0.0414, 0.0319, 0.0370], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-27 19:20:16,251 INFO [finetune.py:976] (1/7) Epoch 23, batch 4450, loss[loss=0.236, simple_loss=0.3162, pruned_loss=0.07795, over 4803.00 frames. ], tot_loss[loss=0.1701, simple_loss=0.2425, pruned_loss=0.04883, over 953789.45 frames. ], batch size: 45, lr: 3.08e-03, grad_scale: 16.0 2023-04-27 19:20:25,828 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.347e+01 1.587e+02 1.803e+02 2.328e+02 4.838e+02, threshold=3.606e+02, percent-clipped=2.0 2023-04-27 19:20:38,839 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.1964, 1.4914, 1.3432, 1.6221, 1.5972, 1.8341, 1.3087, 3.3371], device='cuda:1'), covar=tensor([0.0609, 0.0816, 0.0782, 0.1196, 0.0626, 0.0548, 0.0760, 0.0141], device='cuda:1'), in_proj_covar=tensor([0.0037, 0.0038, 0.0039, 0.0042, 0.0040, 0.0038, 0.0037, 0.0055], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0013, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:1') 2023-04-27 19:20:41,710 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([3.9779, 3.9781, 3.0442, 4.6153, 4.0741, 4.0358, 2.0669, 3.9285], device='cuda:1'), covar=tensor([0.2080, 0.1221, 0.3307, 0.1799, 0.3670, 0.2162, 0.6177, 0.2745], device='cuda:1'), in_proj_covar=tensor([0.0250, 0.0222, 0.0257, 0.0310, 0.0302, 0.0251, 0.0280, 0.0278], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 19:21:01,287 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=130493.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:21:02,961 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=130495.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:21:11,828 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=130499.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:21:23,654 INFO [finetune.py:976] (1/7) Epoch 23, batch 4500, loss[loss=0.228, simple_loss=0.2939, pruned_loss=0.08103, over 4841.00 frames. ], tot_loss[loss=0.171, simple_loss=0.244, pruned_loss=0.04902, over 954342.44 frames. ], batch size: 44, lr: 3.08e-03, grad_scale: 16.0 2023-04-27 19:21:46,476 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.53 vs. limit=2.0 2023-04-27 19:22:17,944 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=130548.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:22:31,479 INFO [finetune.py:976] (1/7) Epoch 23, batch 4550, loss[loss=0.2211, simple_loss=0.2976, pruned_loss=0.07233, over 4727.00 frames. ], tot_loss[loss=0.1725, simple_loss=0.2458, pruned_loss=0.04957, over 954875.90 frames. ], batch size: 59, lr: 3.08e-03, grad_scale: 16.0 2023-04-27 19:22:42,261 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=130565.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:22:42,754 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.159e+02 1.614e+02 1.836e+02 2.049e+02 5.292e+02, threshold=3.672e+02, percent-clipped=1.0 2023-04-27 19:23:36,953 INFO [finetune.py:976] (1/7) Epoch 23, batch 4600, loss[loss=0.1636, simple_loss=0.2214, pruned_loss=0.05292, over 4888.00 frames. ], tot_loss[loss=0.1711, simple_loss=0.2444, pruned_loss=0.04893, over 954530.48 frames. ], batch size: 35, lr: 3.08e-03, grad_scale: 16.0 2023-04-27 19:23:37,073 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=130609.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 19:23:56,078 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=130626.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:24:06,302 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.5403, 1.9159, 1.9903, 2.0606, 1.9104, 1.9222, 1.9704, 2.0634], device='cuda:1'), covar=tensor([0.4134, 0.5448, 0.4588, 0.4580, 0.5570, 0.7198, 0.5892, 0.4999], device='cuda:1'), in_proj_covar=tensor([0.0342, 0.0376, 0.0328, 0.0341, 0.0350, 0.0396, 0.0359, 0.0332], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 19:24:36,991 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=130656.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:24:38,204 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=130658.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 19:24:39,269 INFO [finetune.py:976] (1/7) Epoch 23, batch 4650, loss[loss=0.1475, simple_loss=0.217, pruned_loss=0.03899, over 4737.00 frames. ], tot_loss[loss=0.1691, simple_loss=0.2416, pruned_loss=0.04828, over 955068.52 frames. ], batch size: 28, lr: 3.08e-03, grad_scale: 16.0 2023-04-27 19:24:48,332 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.298e+01 1.478e+02 1.669e+02 2.060e+02 3.844e+02, threshold=3.337e+02, percent-clipped=2.0 2023-04-27 19:25:05,618 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.2956, 1.4915, 1.3330, 1.7507, 1.5748, 1.8608, 1.2778, 3.5828], device='cuda:1'), covar=tensor([0.0683, 0.1072, 0.1025, 0.1321, 0.0824, 0.0594, 0.0981, 0.0197], device='cuda:1'), in_proj_covar=tensor([0.0037, 0.0039, 0.0039, 0.0043, 0.0040, 0.0038, 0.0038, 0.0055], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0013, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:1') 2023-04-27 19:25:39,306 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=130704.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:25:40,550 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=130706.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:25:47,204 INFO [finetune.py:976] (1/7) Epoch 23, batch 4700, loss[loss=0.1355, simple_loss=0.2084, pruned_loss=0.03128, over 4906.00 frames. ], tot_loss[loss=0.1674, simple_loss=0.2388, pruned_loss=0.04799, over 956191.79 frames. ], batch size: 43, lr: 3.08e-03, grad_scale: 16.0 2023-04-27 19:26:22,719 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=130740.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:26:41,806 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-04-27 19:26:50,254 INFO [finetune.py:976] (1/7) Epoch 23, batch 4750, loss[loss=0.1715, simple_loss=0.2429, pruned_loss=0.05005, over 4752.00 frames. ], tot_loss[loss=0.166, simple_loss=0.2368, pruned_loss=0.04761, over 954139.55 frames. ], batch size: 54, lr: 3.08e-03, grad_scale: 16.0 2023-04-27 19:27:00,092 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.073e+02 1.517e+02 1.838e+02 2.137e+02 3.767e+02, threshold=3.677e+02, percent-clipped=3.0 2023-04-27 19:27:24,061 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=130788.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:27:33,678 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=130794.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:27:54,320 INFO [finetune.py:976] (1/7) Epoch 23, batch 4800, loss[loss=0.1804, simple_loss=0.2575, pruned_loss=0.05162, over 4909.00 frames. ], tot_loss[loss=0.1696, simple_loss=0.2409, pruned_loss=0.04916, over 953449.34 frames. ], batch size: 43, lr: 3.08e-03, grad_scale: 16.0 2023-04-27 19:28:54,617 INFO [finetune.py:976] (1/7) Epoch 23, batch 4850, loss[loss=0.169, simple_loss=0.2275, pruned_loss=0.05523, over 4167.00 frames. ], tot_loss[loss=0.1716, simple_loss=0.2436, pruned_loss=0.04976, over 952066.41 frames. ], batch size: 18, lr: 3.08e-03, grad_scale: 16.0 2023-04-27 19:28:55,520 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-04-27 19:29:05,296 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.174e+02 1.564e+02 1.952e+02 2.266e+02 5.605e+02, threshold=3.905e+02, percent-clipped=3.0 2023-04-27 19:30:00,417 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=130904.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 19:30:03,972 INFO [finetune.py:976] (1/7) Epoch 23, batch 4900, loss[loss=0.1908, simple_loss=0.2618, pruned_loss=0.05993, over 4835.00 frames. ], tot_loss[loss=0.1721, simple_loss=0.2445, pruned_loss=0.04986, over 952876.32 frames. ], batch size: 47, lr: 3.08e-03, grad_scale: 16.0 2023-04-27 19:30:23,218 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=130921.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:31:11,883 INFO [finetune.py:976] (1/7) Epoch 23, batch 4950, loss[loss=0.1651, simple_loss=0.2522, pruned_loss=0.03902, over 4820.00 frames. ], tot_loss[loss=0.1725, simple_loss=0.2455, pruned_loss=0.0497, over 953993.77 frames. ], batch size: 38, lr: 3.08e-03, grad_scale: 16.0 2023-04-27 19:31:21,964 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.168e+02 1.568e+02 1.814e+02 2.171e+02 6.365e+02, threshold=3.628e+02, percent-clipped=1.0 2023-04-27 19:31:33,179 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9426, 1.4800, 1.5364, 1.6370, 2.0695, 1.6928, 1.4233, 1.4828], device='cuda:1'), covar=tensor([0.1568, 0.1412, 0.2181, 0.1407, 0.0853, 0.1506, 0.2107, 0.2138], device='cuda:1'), in_proj_covar=tensor([0.0314, 0.0311, 0.0353, 0.0287, 0.0328, 0.0308, 0.0301, 0.0375], device='cuda:1'), out_proj_covar=tensor([6.4423e-05, 6.4193e-05, 7.4336e-05, 5.7651e-05, 6.7565e-05, 6.4625e-05, 6.2810e-05, 7.9532e-05], device='cuda:1') 2023-04-27 19:32:26,076 INFO [finetune.py:976] (1/7) Epoch 23, batch 5000, loss[loss=0.1687, simple_loss=0.2393, pruned_loss=0.04901, over 4825.00 frames. ], tot_loss[loss=0.1709, simple_loss=0.244, pruned_loss=0.0489, over 956633.50 frames. ], batch size: 38, lr: 3.08e-03, grad_scale: 16.0 2023-04-27 19:33:10,246 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=131040.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:33:10,968 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.67 vs. limit=2.0 2023-04-27 19:33:35,487 INFO [finetune.py:976] (1/7) Epoch 23, batch 5050, loss[loss=0.1535, simple_loss=0.2265, pruned_loss=0.04025, over 4828.00 frames. ], tot_loss[loss=0.1689, simple_loss=0.2415, pruned_loss=0.04814, over 956872.55 frames. ], batch size: 33, lr: 3.08e-03, grad_scale: 16.0 2023-04-27 19:33:41,159 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.129e+02 1.462e+02 1.838e+02 2.219e+02 3.974e+02, threshold=3.675e+02, percent-clipped=2.0 2023-04-27 19:34:08,386 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=131088.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:34:08,417 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=131088.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:34:18,270 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=131094.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:34:38,867 INFO [finetune.py:976] (1/7) Epoch 23, batch 5100, loss[loss=0.176, simple_loss=0.2428, pruned_loss=0.0546, over 4887.00 frames. ], tot_loss[loss=0.1666, simple_loss=0.2382, pruned_loss=0.0475, over 956072.72 frames. ], batch size: 32, lr: 3.08e-03, grad_scale: 32.0 2023-04-27 19:35:10,695 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.99 vs. limit=2.0 2023-04-27 19:35:11,740 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=131136.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:35:21,307 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=131142.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:35:41,499 INFO [finetune.py:976] (1/7) Epoch 23, batch 5150, loss[loss=0.1783, simple_loss=0.2529, pruned_loss=0.05182, over 4809.00 frames. ], tot_loss[loss=0.1674, simple_loss=0.239, pruned_loss=0.04792, over 955202.49 frames. ], batch size: 41, lr: 3.08e-03, grad_scale: 32.0 2023-04-27 19:35:51,236 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.008e+02 1.590e+02 1.900e+02 2.291e+02 6.669e+02, threshold=3.800e+02, percent-clipped=5.0 2023-04-27 19:36:02,280 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=131174.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:36:42,353 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=131204.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 19:36:45,362 INFO [finetune.py:976] (1/7) Epoch 23, batch 5200, loss[loss=0.2098, simple_loss=0.2811, pruned_loss=0.0692, over 4760.00 frames. ], tot_loss[loss=0.1691, simple_loss=0.2413, pruned_loss=0.04846, over 954957.31 frames. ], batch size: 54, lr: 3.08e-03, grad_scale: 32.0 2023-04-27 19:36:51,013 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6531, 1.9731, 1.7068, 1.9306, 1.4915, 1.7349, 1.6411, 1.3392], device='cuda:1'), covar=tensor([0.1817, 0.1309, 0.0898, 0.1164, 0.3564, 0.1231, 0.1806, 0.2493], device='cuda:1'), in_proj_covar=tensor([0.0283, 0.0300, 0.0214, 0.0274, 0.0314, 0.0255, 0.0247, 0.0261], device='cuda:1'), out_proj_covar=tensor([1.1313e-04, 1.1870e-04, 8.4265e-05, 1.0810e-04, 1.2692e-04, 1.0055e-04, 9.9608e-05, 1.0317e-04], device='cuda:1') 2023-04-27 19:37:04,427 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=131221.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:37:22,658 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=131235.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 19:37:43,643 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=131252.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:37:52,693 INFO [finetune.py:976] (1/7) Epoch 23, batch 5250, loss[loss=0.1591, simple_loss=0.2305, pruned_loss=0.04385, over 4216.00 frames. ], tot_loss[loss=0.1711, simple_loss=0.2439, pruned_loss=0.04915, over 953580.48 frames. ], batch size: 65, lr: 3.08e-03, grad_scale: 32.0 2023-04-27 19:37:57,553 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.170e+02 1.647e+02 1.953e+02 2.357e+02 4.289e+02, threshold=3.906e+02, percent-clipped=1.0 2023-04-27 19:38:04,766 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=131269.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:38:50,170 INFO [finetune.py:976] (1/7) Epoch 23, batch 5300, loss[loss=0.1295, simple_loss=0.2051, pruned_loss=0.027, over 4775.00 frames. ], tot_loss[loss=0.1725, simple_loss=0.2456, pruned_loss=0.04969, over 954743.10 frames. ], batch size: 26, lr: 3.08e-03, grad_scale: 32.0 2023-04-27 19:39:21,827 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-04-27 19:39:56,069 INFO [finetune.py:976] (1/7) Epoch 23, batch 5350, loss[loss=0.1696, simple_loss=0.2376, pruned_loss=0.05084, over 4729.00 frames. ], tot_loss[loss=0.1721, simple_loss=0.2454, pruned_loss=0.04944, over 956583.06 frames. ], batch size: 23, lr: 3.08e-03, grad_scale: 32.0 2023-04-27 19:40:07,007 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 8.812e+01 1.537e+02 1.790e+02 2.239e+02 4.084e+02, threshold=3.580e+02, percent-clipped=2.0 2023-04-27 19:40:38,635 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.2252, 2.7743, 2.1245, 2.1282, 1.5506, 1.5712, 2.2752, 1.4999], device='cuda:1'), covar=tensor([0.1468, 0.1324, 0.1272, 0.1624, 0.2119, 0.1842, 0.0857, 0.1903], device='cuda:1'), in_proj_covar=tensor([0.0195, 0.0208, 0.0167, 0.0203, 0.0197, 0.0183, 0.0154, 0.0183], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-27 19:41:03,583 INFO [finetune.py:976] (1/7) Epoch 23, batch 5400, loss[loss=0.1642, simple_loss=0.2354, pruned_loss=0.04644, over 4937.00 frames. ], tot_loss[loss=0.1705, simple_loss=0.243, pruned_loss=0.04906, over 953128.04 frames. ], batch size: 38, lr: 3.08e-03, grad_scale: 32.0 2023-04-27 19:41:03,699 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=131409.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:41:10,426 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.66 vs. limit=5.0 2023-04-27 19:41:12,779 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=131416.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:42:12,941 INFO [finetune.py:976] (1/7) Epoch 23, batch 5450, loss[loss=0.1866, simple_loss=0.2558, pruned_loss=0.05866, over 4808.00 frames. ], tot_loss[loss=0.1689, simple_loss=0.2405, pruned_loss=0.04864, over 953289.37 frames. ], batch size: 38, lr: 3.08e-03, grad_scale: 32.0 2023-04-27 19:42:22,593 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.036e+02 1.575e+02 1.987e+02 2.572e+02 5.206e+02, threshold=3.973e+02, percent-clipped=5.0 2023-04-27 19:42:31,945 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=131470.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:42:43,226 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=131477.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:43:05,095 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=131492.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:43:08,869 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=3.78 vs. limit=5.0 2023-04-27 19:43:20,923 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.5819, 3.3071, 2.6140, 2.7733, 1.9787, 1.9798, 2.8528, 1.9778], device='cuda:1'), covar=tensor([0.1607, 0.1308, 0.1373, 0.1574, 0.2312, 0.1816, 0.0853, 0.1903], device='cuda:1'), in_proj_covar=tensor([0.0196, 0.0208, 0.0167, 0.0203, 0.0198, 0.0183, 0.0154, 0.0184], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-27 19:43:22,050 INFO [finetune.py:976] (1/7) Epoch 23, batch 5500, loss[loss=0.1395, simple_loss=0.2141, pruned_loss=0.03245, over 4825.00 frames. ], tot_loss[loss=0.1667, simple_loss=0.238, pruned_loss=0.04771, over 954440.50 frames. ], batch size: 38, lr: 3.08e-03, grad_scale: 32.0 2023-04-27 19:43:52,912 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=131530.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 19:44:16,328 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=131553.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:44:21,722 INFO [finetune.py:976] (1/7) Epoch 23, batch 5550, loss[loss=0.2248, simple_loss=0.289, pruned_loss=0.08029, over 4828.00 frames. ], tot_loss[loss=0.1684, simple_loss=0.24, pruned_loss=0.04837, over 952533.20 frames. ], batch size: 33, lr: 3.08e-03, grad_scale: 32.0 2023-04-27 19:44:32,379 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.039e+02 1.589e+02 1.849e+02 2.302e+02 7.246e+02, threshold=3.698e+02, percent-clipped=1.0 2023-04-27 19:44:50,770 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.3538, 2.7297, 1.0124, 1.6410, 2.1467, 1.3522, 3.9988, 1.8319], device='cuda:1'), covar=tensor([0.0653, 0.0850, 0.0856, 0.1183, 0.0533, 0.0996, 0.0238, 0.0617], device='cuda:1'), in_proj_covar=tensor([0.0050, 0.0064, 0.0047, 0.0046, 0.0049, 0.0051, 0.0072, 0.0050], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0009, 0.0007, 0.0007, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:1') 2023-04-27 19:45:12,495 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9037, 2.4176, 1.7653, 1.6585, 1.4104, 1.3907, 1.8120, 1.3195], device='cuda:1'), covar=tensor([0.1574, 0.1193, 0.1386, 0.1684, 0.2164, 0.1856, 0.0958, 0.1919], device='cuda:1'), in_proj_covar=tensor([0.0196, 0.0208, 0.0167, 0.0203, 0.0198, 0.0183, 0.0154, 0.0183], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-27 19:45:32,464 INFO [finetune.py:976] (1/7) Epoch 23, batch 5600, loss[loss=0.2007, simple_loss=0.2776, pruned_loss=0.06191, over 4864.00 frames. ], tot_loss[loss=0.1709, simple_loss=0.2435, pruned_loss=0.04911, over 953200.46 frames. ], batch size: 31, lr: 3.07e-03, grad_scale: 32.0 2023-04-27 19:46:28,599 INFO [finetune.py:976] (1/7) Epoch 23, batch 5650, loss[loss=0.127, simple_loss=0.2007, pruned_loss=0.02668, over 4772.00 frames. ], tot_loss[loss=0.1712, simple_loss=0.245, pruned_loss=0.04871, over 953348.31 frames. ], batch size: 28, lr: 3.07e-03, grad_scale: 32.0 2023-04-27 19:46:38,007 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.000e+01 1.549e+02 1.818e+02 2.173e+02 4.178e+02, threshold=3.635e+02, percent-clipped=2.0 2023-04-27 19:47:00,952 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([3.3944, 3.4170, 2.6070, 3.9457, 3.3821, 3.3658, 1.5931, 3.3012], device='cuda:1'), covar=tensor([0.1807, 0.1286, 0.3014, 0.1999, 0.4444, 0.1958, 0.5388, 0.2453], device='cuda:1'), in_proj_covar=tensor([0.0244, 0.0218, 0.0252, 0.0303, 0.0293, 0.0245, 0.0272, 0.0271], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 19:47:29,408 INFO [finetune.py:976] (1/7) Epoch 23, batch 5700, loss[loss=0.1653, simple_loss=0.2331, pruned_loss=0.04871, over 4275.00 frames. ], tot_loss[loss=0.1677, simple_loss=0.2405, pruned_loss=0.04748, over 937639.68 frames. ], batch size: 18, lr: 3.07e-03, grad_scale: 32.0 2023-04-27 19:48:21,866 INFO [finetune.py:976] (1/7) Epoch 24, batch 0, loss[loss=0.1475, simple_loss=0.2308, pruned_loss=0.0321, over 4781.00 frames. ], tot_loss[loss=0.1475, simple_loss=0.2308, pruned_loss=0.0321, over 4781.00 frames. ], batch size: 28, lr: 3.07e-03, grad_scale: 32.0 2023-04-27 19:48:21,866 INFO [finetune.py:1001] (1/7) Computing validation loss 2023-04-27 19:48:25,600 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.1910, 1.4490, 1.7458, 1.8751, 1.8008, 1.8544, 1.7714, 1.7848], device='cuda:1'), covar=tensor([0.4093, 0.5780, 0.4766, 0.4805, 0.5696, 0.7620, 0.5092, 0.5122], device='cuda:1'), in_proj_covar=tensor([0.0340, 0.0375, 0.0328, 0.0339, 0.0349, 0.0394, 0.0357, 0.0329], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 19:48:32,798 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.2618, 1.5316, 1.8528, 1.9728, 1.9520, 2.0278, 1.8778, 1.9313], device='cuda:1'), covar=tensor([0.3895, 0.5574, 0.4853, 0.4738, 0.5540, 0.6933, 0.4963, 0.4920], device='cuda:1'), in_proj_covar=tensor([0.0340, 0.0375, 0.0328, 0.0339, 0.0349, 0.0394, 0.0357, 0.0329], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 19:48:37,425 INFO [finetune.py:1010] (1/7) Epoch 24, validation: loss=0.1552, simple_loss=0.2243, pruned_loss=0.04308, over 2265189.00 frames. 2023-04-27 19:48:37,425 INFO [finetune.py:1011] (1/7) Maximum memory allocated so far is 6391MB 2023-04-27 19:48:45,625 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.1369, 1.5630, 1.4015, 1.7300, 1.7099, 1.9374, 1.3928, 3.4978], device='cuda:1'), covar=tensor([0.0608, 0.0786, 0.0776, 0.1234, 0.0620, 0.0567, 0.0755, 0.0140], device='cuda:1'), in_proj_covar=tensor([0.0037, 0.0038, 0.0039, 0.0043, 0.0040, 0.0038, 0.0038, 0.0055], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0013, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:1') 2023-04-27 19:49:12,805 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=131765.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:49:13,311 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.101e+01 1.500e+02 1.790e+02 2.328e+02 5.462e+02, threshold=3.579e+02, percent-clipped=4.0 2023-04-27 19:49:21,988 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=131772.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:49:37,727 INFO [finetune.py:976] (1/7) Epoch 24, batch 50, loss[loss=0.1468, simple_loss=0.2166, pruned_loss=0.03851, over 4862.00 frames. ], tot_loss[loss=0.1704, simple_loss=0.243, pruned_loss=0.04888, over 216920.39 frames. ], batch size: 34, lr: 3.07e-03, grad_scale: 32.0 2023-04-27 19:50:39,505 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=131830.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:50:48,615 INFO [finetune.py:976] (1/7) Epoch 24, batch 100, loss[loss=0.1587, simple_loss=0.2346, pruned_loss=0.04142, over 4908.00 frames. ], tot_loss[loss=0.1668, simple_loss=0.239, pruned_loss=0.0473, over 381831.34 frames. ], batch size: 37, lr: 3.07e-03, grad_scale: 32.0 2023-04-27 19:50:56,378 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=131848.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:50:58,847 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9802, 1.7200, 1.9812, 2.3940, 2.4397, 1.9427, 1.7125, 2.1550], device='cuda:1'), covar=tensor([0.0750, 0.1262, 0.0747, 0.0531, 0.0520, 0.0710, 0.0710, 0.0538], device='cuda:1'), in_proj_covar=tensor([0.0185, 0.0201, 0.0182, 0.0171, 0.0174, 0.0177, 0.0147, 0.0176], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 19:51:01,829 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=131857.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:51:07,130 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.089e+01 1.524e+02 1.795e+02 2.090e+02 4.072e+02, threshold=3.590e+02, percent-clipped=1.0 2023-04-27 19:51:23,454 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=131878.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:51:35,579 INFO [finetune.py:976] (1/7) Epoch 24, batch 150, loss[loss=0.1555, simple_loss=0.2232, pruned_loss=0.04393, over 4754.00 frames. ], tot_loss[loss=0.1645, simple_loss=0.2357, pruned_loss=0.04668, over 508449.05 frames. ], batch size: 26, lr: 3.07e-03, grad_scale: 32.0 2023-04-27 19:52:18,775 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=131918.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:52:44,207 INFO [finetune.py:976] (1/7) Epoch 24, batch 200, loss[loss=0.1689, simple_loss=0.2454, pruned_loss=0.04625, over 4836.00 frames. ], tot_loss[loss=0.1636, simple_loss=0.2343, pruned_loss=0.0465, over 608936.55 frames. ], batch size: 30, lr: 3.07e-03, grad_scale: 32.0 2023-04-27 19:53:08,194 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8493, 2.1393, 1.8655, 1.5665, 1.4322, 1.5199, 1.8972, 1.4072], device='cuda:1'), covar=tensor([0.1477, 0.1437, 0.1338, 0.1581, 0.2173, 0.1782, 0.0906, 0.1834], device='cuda:1'), in_proj_covar=tensor([0.0198, 0.0210, 0.0169, 0.0205, 0.0200, 0.0186, 0.0156, 0.0185], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-27 19:53:12,914 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.093e+02 1.534e+02 1.797e+02 2.198e+02 4.905e+02, threshold=3.594e+02, percent-clipped=2.0 2023-04-27 19:53:27,656 INFO [finetune.py:976] (1/7) Epoch 24, batch 250, loss[loss=0.2035, simple_loss=0.2891, pruned_loss=0.05891, over 4834.00 frames. ], tot_loss[loss=0.167, simple_loss=0.2385, pruned_loss=0.04777, over 684855.45 frames. ], batch size: 49, lr: 3.07e-03, grad_scale: 32.0 2023-04-27 19:53:48,671 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.3938, 1.3755, 1.6662, 1.6897, 1.3165, 1.1521, 1.4562, 0.9151], device='cuda:1'), covar=tensor([0.0582, 0.0502, 0.0369, 0.0588, 0.0698, 0.1045, 0.0545, 0.0595], device='cuda:1'), in_proj_covar=tensor([0.0069, 0.0067, 0.0066, 0.0068, 0.0074, 0.0095, 0.0072, 0.0064], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-27 19:53:57,448 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=132031.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:54:01,445 INFO [finetune.py:976] (1/7) Epoch 24, batch 300, loss[loss=0.1716, simple_loss=0.2388, pruned_loss=0.05221, over 4926.00 frames. ], tot_loss[loss=0.1696, simple_loss=0.2416, pruned_loss=0.04877, over 744134.50 frames. ], batch size: 33, lr: 3.07e-03, grad_scale: 32.0 2023-04-27 19:54:24,647 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=132054.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:54:36,475 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=132065.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:54:36,978 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.215e+02 1.704e+02 1.925e+02 2.353e+02 6.924e+02, threshold=3.849e+02, percent-clipped=2.0 2023-04-27 19:54:46,358 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=132072.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:55:00,436 INFO [finetune.py:976] (1/7) Epoch 24, batch 350, loss[loss=0.1696, simple_loss=0.2508, pruned_loss=0.0442, over 4784.00 frames. ], tot_loss[loss=0.172, simple_loss=0.2443, pruned_loss=0.04983, over 791559.85 frames. ], batch size: 51, lr: 3.07e-03, grad_scale: 32.0 2023-04-27 19:55:10,370 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=132092.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:55:40,630 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=132113.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:55:41,944 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=132115.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 19:55:44,961 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=132120.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:56:06,855 INFO [finetune.py:976] (1/7) Epoch 24, batch 400, loss[loss=0.2058, simple_loss=0.2725, pruned_loss=0.0696, over 4235.00 frames. ], tot_loss[loss=0.173, simple_loss=0.2458, pruned_loss=0.05014, over 828263.57 frames. ], batch size: 65, lr: 3.07e-03, grad_scale: 32.0 2023-04-27 19:56:24,609 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=3.92 vs. limit=5.0 2023-04-27 19:56:25,259 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 2023-04-27 19:56:25,706 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=132148.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:56:25,821 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.55 vs. limit=2.0 2023-04-27 19:56:49,859 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.121e+02 1.575e+02 1.808e+02 2.256e+02 3.614e+02, threshold=3.616e+02, percent-clipped=0.0 2023-04-27 19:57:08,719 INFO [finetune.py:976] (1/7) Epoch 24, batch 450, loss[loss=0.1699, simple_loss=0.2391, pruned_loss=0.05031, over 4791.00 frames. ], tot_loss[loss=0.1716, simple_loss=0.2446, pruned_loss=0.04935, over 857007.39 frames. ], batch size: 51, lr: 3.07e-03, grad_scale: 32.0 2023-04-27 19:57:08,926 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2023-04-27 19:57:15,195 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=132196.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:57:27,639 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=132213.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:57:42,205 INFO [finetune.py:976] (1/7) Epoch 24, batch 500, loss[loss=0.143, simple_loss=0.2076, pruned_loss=0.03923, over 4752.00 frames. ], tot_loss[loss=0.1699, simple_loss=0.2421, pruned_loss=0.04891, over 880062.11 frames. ], batch size: 54, lr: 3.07e-03, grad_scale: 32.0 2023-04-27 19:57:42,819 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([0.5590, 1.4768, 1.4316, 1.1654, 1.3650, 1.2746, 1.6603, 1.4203], device='cuda:1'), covar=tensor([0.3344, 0.1719, 0.5086, 0.2664, 0.1605, 0.2160, 0.1589, 0.4793], device='cuda:1'), in_proj_covar=tensor([0.0338, 0.0352, 0.0426, 0.0354, 0.0379, 0.0376, 0.0369, 0.0421], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 19:58:03,235 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.098e+02 1.544e+02 1.768e+02 2.170e+02 4.737e+02, threshold=3.537e+02, percent-clipped=2.0 2023-04-27 19:58:16,036 INFO [finetune.py:976] (1/7) Epoch 24, batch 550, loss[loss=0.1795, simple_loss=0.251, pruned_loss=0.054, over 4771.00 frames. ], tot_loss[loss=0.1684, simple_loss=0.2398, pruned_loss=0.04846, over 896810.16 frames. ], batch size: 54, lr: 3.07e-03, grad_scale: 32.0 2023-04-27 19:58:49,838 INFO [finetune.py:976] (1/7) Epoch 24, batch 600, loss[loss=0.2108, simple_loss=0.2928, pruned_loss=0.06439, over 4863.00 frames. ], tot_loss[loss=0.1695, simple_loss=0.2408, pruned_loss=0.04913, over 909876.70 frames. ], batch size: 44, lr: 3.07e-03, grad_scale: 32.0 2023-04-27 19:59:10,243 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.026e+02 1.689e+02 1.942e+02 2.477e+02 4.504e+02, threshold=3.885e+02, percent-clipped=3.0 2023-04-27 19:59:22,996 INFO [finetune.py:976] (1/7) Epoch 24, batch 650, loss[loss=0.206, simple_loss=0.2736, pruned_loss=0.06918, over 4817.00 frames. ], tot_loss[loss=0.1725, simple_loss=0.2441, pruned_loss=0.05039, over 920130.16 frames. ], batch size: 40, lr: 3.07e-03, grad_scale: 32.0 2023-04-27 19:59:23,067 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=132387.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:59:39,120 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=132410.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 19:59:59,422 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-04-27 20:00:08,100 INFO [finetune.py:976] (1/7) Epoch 24, batch 700, loss[loss=0.1487, simple_loss=0.2105, pruned_loss=0.04347, over 4822.00 frames. ], tot_loss[loss=0.1721, simple_loss=0.2444, pruned_loss=0.0499, over 928735.16 frames. ], batch size: 25, lr: 3.07e-03, grad_scale: 32.0 2023-04-27 20:00:50,349 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.129e+02 1.633e+02 1.958e+02 2.360e+02 4.192e+02, threshold=3.915e+02, percent-clipped=4.0 2023-04-27 20:01:15,827 INFO [finetune.py:976] (1/7) Epoch 24, batch 750, loss[loss=0.2178, simple_loss=0.2986, pruned_loss=0.06845, over 4757.00 frames. ], tot_loss[loss=0.1732, simple_loss=0.2458, pruned_loss=0.05029, over 934983.98 frames. ], batch size: 54, lr: 3.06e-03, grad_scale: 32.0 2023-04-27 20:01:48,572 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=132513.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:02:21,498 INFO [finetune.py:976] (1/7) Epoch 24, batch 800, loss[loss=0.1758, simple_loss=0.2424, pruned_loss=0.05463, over 4782.00 frames. ], tot_loss[loss=0.1721, simple_loss=0.2452, pruned_loss=0.04951, over 939739.35 frames. ], batch size: 29, lr: 3.06e-03, grad_scale: 32.0 2023-04-27 20:02:40,448 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.2867, 1.6333, 1.5285, 2.0750, 2.0097, 2.0895, 1.6030, 4.4818], device='cuda:1'), covar=tensor([0.0545, 0.0823, 0.0789, 0.1147, 0.0613, 0.0550, 0.0722, 0.0106], device='cuda:1'), in_proj_covar=tensor([0.0037, 0.0039, 0.0039, 0.0043, 0.0040, 0.0038, 0.0038, 0.0055], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0013, 0.0013, 0.0015, 0.0014, 0.0013, 0.0013, 0.0016], device='cuda:1') 2023-04-27 20:02:53,590 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=132561.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:03:01,810 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 8.951e+01 1.478e+02 1.835e+02 2.240e+02 6.510e+02, threshold=3.671e+02, percent-clipped=1.0 2023-04-27 20:03:17,393 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5660, 3.2219, 1.1895, 1.8882, 1.9727, 2.5661, 2.0240, 1.3262], device='cuda:1'), covar=tensor([0.1146, 0.0780, 0.1769, 0.1157, 0.0917, 0.0791, 0.1294, 0.1900], device='cuda:1'), in_proj_covar=tensor([0.0115, 0.0236, 0.0135, 0.0119, 0.0131, 0.0150, 0.0115, 0.0117], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-27 20:03:27,812 INFO [finetune.py:976] (1/7) Epoch 24, batch 850, loss[loss=0.1301, simple_loss=0.2025, pruned_loss=0.02884, over 3995.00 frames. ], tot_loss[loss=0.1705, simple_loss=0.243, pruned_loss=0.04893, over 943681.39 frames. ], batch size: 17, lr: 3.06e-03, grad_scale: 16.0 2023-04-27 20:04:08,621 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4900, 2.5792, 2.3282, 2.2544, 2.6525, 2.1717, 3.1432, 2.0363], device='cuda:1'), covar=tensor([0.3030, 0.1436, 0.3412, 0.2302, 0.1228, 0.2100, 0.1233, 0.3740], device='cuda:1'), in_proj_covar=tensor([0.0338, 0.0351, 0.0424, 0.0351, 0.0377, 0.0375, 0.0367, 0.0420], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 20:04:34,502 INFO [finetune.py:976] (1/7) Epoch 24, batch 900, loss[loss=0.1618, simple_loss=0.2363, pruned_loss=0.04366, over 4908.00 frames. ], tot_loss[loss=0.1688, simple_loss=0.2408, pruned_loss=0.04837, over 947025.83 frames. ], batch size: 37, lr: 3.06e-03, grad_scale: 16.0 2023-04-27 20:05:01,488 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.0659, 2.1690, 1.8668, 1.8720, 2.2606, 1.8091, 2.7629, 1.6839], device='cuda:1'), covar=tensor([0.3885, 0.1789, 0.4637, 0.2812, 0.1675, 0.2583, 0.1239, 0.4468], device='cuda:1'), in_proj_covar=tensor([0.0338, 0.0350, 0.0424, 0.0351, 0.0377, 0.0374, 0.0367, 0.0419], device='cuda:1'), out_proj_covar=tensor([9.9948e-05, 1.0460e-04, 1.2852e-04, 1.0544e-04, 1.1195e-04, 1.1158e-04, 1.0777e-04, 1.2605e-04], device='cuda:1') 2023-04-27 20:05:02,101 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=132660.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:05:12,904 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.989e+01 1.487e+02 1.792e+02 2.042e+02 3.425e+02, threshold=3.585e+02, percent-clipped=0.0 2023-04-27 20:05:44,310 INFO [finetune.py:976] (1/7) Epoch 24, batch 950, loss[loss=0.1747, simple_loss=0.2459, pruned_loss=0.05177, over 4931.00 frames. ], tot_loss[loss=0.1664, simple_loss=0.2383, pruned_loss=0.04728, over 949147.48 frames. ], batch size: 33, lr: 3.06e-03, grad_scale: 16.0 2023-04-27 20:05:44,394 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=132687.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:06:08,276 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=132710.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 20:06:22,014 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-04-27 20:06:27,930 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=132721.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:06:29,148 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9566, 2.5040, 2.0915, 2.3995, 1.6447, 2.1411, 2.0293, 1.7148], device='cuda:1'), covar=tensor([0.2132, 0.1088, 0.0865, 0.1161, 0.3405, 0.1116, 0.1829, 0.2369], device='cuda:1'), in_proj_covar=tensor([0.0287, 0.0303, 0.0216, 0.0275, 0.0316, 0.0256, 0.0248, 0.0265], device='cuda:1'), out_proj_covar=tensor([1.1478e-04, 1.1992e-04, 8.5110e-05, 1.0858e-04, 1.2732e-04, 1.0104e-04, 1.0026e-04, 1.0445e-04], device='cuda:1') 2023-04-27 20:06:43,233 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=132735.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:06:49,487 INFO [finetune.py:976] (1/7) Epoch 24, batch 1000, loss[loss=0.2204, simple_loss=0.2925, pruned_loss=0.07418, over 4819.00 frames. ], tot_loss[loss=0.1684, simple_loss=0.2407, pruned_loss=0.04805, over 951419.80 frames. ], batch size: 39, lr: 3.06e-03, grad_scale: 16.0 2023-04-27 20:06:55,150 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=132746.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:07:02,346 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=132758.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:07:07,765 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.158e+02 1.549e+02 1.832e+02 2.153e+02 4.117e+02, threshold=3.664e+02, percent-clipped=1.0 2023-04-27 20:07:22,250 INFO [finetune.py:976] (1/7) Epoch 24, batch 1050, loss[loss=0.2015, simple_loss=0.2606, pruned_loss=0.07121, over 4749.00 frames. ], tot_loss[loss=0.1697, simple_loss=0.2423, pruned_loss=0.04856, over 950850.87 frames. ], batch size: 26, lr: 3.06e-03, grad_scale: 16.0 2023-04-27 20:07:35,151 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=132807.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:07:38,791 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.8424, 2.1099, 2.0956, 2.2692, 2.0294, 2.1552, 2.1668, 2.1395], device='cuda:1'), covar=tensor([0.3821, 0.6170, 0.5167, 0.4526, 0.5942, 0.6944, 0.6270, 0.5210], device='cuda:1'), in_proj_covar=tensor([0.0342, 0.0378, 0.0329, 0.0341, 0.0351, 0.0396, 0.0359, 0.0332], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 20:07:56,051 INFO [finetune.py:976] (1/7) Epoch 24, batch 1100, loss[loss=0.1875, simple_loss=0.2708, pruned_loss=0.05213, over 4816.00 frames. ], tot_loss[loss=0.1706, simple_loss=0.2438, pruned_loss=0.04867, over 953404.03 frames. ], batch size: 39, lr: 3.06e-03, grad_scale: 16.0 2023-04-27 20:08:14,645 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.617e+01 1.589e+02 1.846e+02 2.225e+02 4.028e+02, threshold=3.692e+02, percent-clipped=2.0 2023-04-27 20:08:28,712 INFO [finetune.py:976] (1/7) Epoch 24, batch 1150, loss[loss=0.1656, simple_loss=0.2449, pruned_loss=0.04309, over 4891.00 frames. ], tot_loss[loss=0.1721, simple_loss=0.2454, pruned_loss=0.04941, over 954882.66 frames. ], batch size: 43, lr: 3.06e-03, grad_scale: 16.0 2023-04-27 20:09:11,141 INFO [finetune.py:976] (1/7) Epoch 24, batch 1200, loss[loss=0.1703, simple_loss=0.2429, pruned_loss=0.04881, over 4905.00 frames. ], tot_loss[loss=0.1703, simple_loss=0.2435, pruned_loss=0.04853, over 956032.43 frames. ], batch size: 37, lr: 3.06e-03, grad_scale: 16.0 2023-04-27 20:09:48,368 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.050e+02 1.568e+02 1.953e+02 2.291e+02 3.731e+02, threshold=3.906e+02, percent-clipped=1.0 2023-04-27 20:10:09,523 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([4.5995, 4.5212, 3.2113, 5.2601, 4.6034, 4.5685, 1.7576, 4.4526], device='cuda:1'), covar=tensor([0.1554, 0.1045, 0.2932, 0.0866, 0.2301, 0.1633, 0.5825, 0.1983], device='cuda:1'), in_proj_covar=tensor([0.0244, 0.0219, 0.0252, 0.0304, 0.0293, 0.0246, 0.0273, 0.0270], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 20:10:17,850 INFO [finetune.py:976] (1/7) Epoch 24, batch 1250, loss[loss=0.215, simple_loss=0.2773, pruned_loss=0.07635, over 4263.00 frames. ], tot_loss[loss=0.1691, simple_loss=0.2414, pruned_loss=0.04838, over 956582.73 frames. ], batch size: 65, lr: 3.06e-03, grad_scale: 16.0 2023-04-27 20:10:30,024 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=3.52 vs. limit=5.0 2023-04-27 20:10:53,170 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=133016.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:11:17,185 INFO [finetune.py:976] (1/7) Epoch 24, batch 1300, loss[loss=0.2091, simple_loss=0.2691, pruned_loss=0.07456, over 4233.00 frames. ], tot_loss[loss=0.1655, simple_loss=0.2372, pruned_loss=0.04692, over 955128.53 frames. ], batch size: 65, lr: 3.06e-03, grad_scale: 16.0 2023-04-27 20:11:22,042 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.61 vs. limit=5.0 2023-04-27 20:11:37,887 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.023e+02 1.644e+02 1.911e+02 2.318e+02 6.826e+02, threshold=3.821e+02, percent-clipped=2.0 2023-04-27 20:11:41,703 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.4520, 2.2084, 2.4084, 2.9400, 2.8897, 2.3425, 2.0534, 2.5180], device='cuda:1'), covar=tensor([0.0728, 0.1022, 0.0701, 0.0543, 0.0546, 0.0772, 0.0683, 0.0530], device='cuda:1'), in_proj_covar=tensor([0.0186, 0.0204, 0.0185, 0.0172, 0.0177, 0.0179, 0.0149, 0.0178], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 20:11:50,600 INFO [finetune.py:976] (1/7) Epoch 24, batch 1350, loss[loss=0.1686, simple_loss=0.2477, pruned_loss=0.04479, over 4827.00 frames. ], tot_loss[loss=0.1664, simple_loss=0.2376, pruned_loss=0.04757, over 953262.62 frames. ], batch size: 33, lr: 3.06e-03, grad_scale: 16.0 2023-04-27 20:12:07,617 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=133102.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:12:21,174 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([0.5333, 1.4394, 1.4310, 1.1807, 1.4328, 1.2892, 1.7293, 1.3403], device='cuda:1'), covar=tensor([0.3106, 0.1666, 0.4088, 0.2195, 0.1421, 0.2013, 0.1369, 0.4367], device='cuda:1'), in_proj_covar=tensor([0.0340, 0.0354, 0.0427, 0.0352, 0.0379, 0.0376, 0.0369, 0.0421], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 20:12:52,452 INFO [finetune.py:976] (1/7) Epoch 24, batch 1400, loss[loss=0.2831, simple_loss=0.3383, pruned_loss=0.1139, over 4262.00 frames. ], tot_loss[loss=0.1688, simple_loss=0.2404, pruned_loss=0.04865, over 952128.87 frames. ], batch size: 65, lr: 3.06e-03, grad_scale: 16.0 2023-04-27 20:12:54,998 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8453, 2.2145, 1.1226, 1.2092, 1.8949, 1.2404, 2.9246, 1.4056], device='cuda:1'), covar=tensor([0.0747, 0.0690, 0.0758, 0.1443, 0.0509, 0.1055, 0.0350, 0.0754], device='cuda:1'), in_proj_covar=tensor([0.0050, 0.0063, 0.0047, 0.0045, 0.0049, 0.0050, 0.0072, 0.0050], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0009, 0.0007, 0.0007, 0.0008, 0.0008, 0.0010, 0.0007], device='cuda:1') 2023-04-27 20:13:29,246 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.160e+02 1.717e+02 2.018e+02 2.257e+02 3.714e+02, threshold=4.037e+02, percent-clipped=0.0 2023-04-27 20:13:41,457 INFO [finetune.py:976] (1/7) Epoch 24, batch 1450, loss[loss=0.153, simple_loss=0.2211, pruned_loss=0.04244, over 4708.00 frames. ], tot_loss[loss=0.1711, simple_loss=0.2435, pruned_loss=0.04936, over 954398.34 frames. ], batch size: 23, lr: 3.06e-03, grad_scale: 16.0 2023-04-27 20:14:31,507 INFO [finetune.py:976] (1/7) Epoch 24, batch 1500, loss[loss=0.1916, simple_loss=0.2496, pruned_loss=0.0668, over 4134.00 frames. ], tot_loss[loss=0.1716, simple_loss=0.2441, pruned_loss=0.04952, over 952928.91 frames. ], batch size: 18, lr: 3.06e-03, grad_scale: 16.0 2023-04-27 20:15:14,559 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.127e+02 1.609e+02 1.860e+02 2.213e+02 4.460e+02, threshold=3.721e+02, percent-clipped=1.0 2023-04-27 20:15:44,450 INFO [finetune.py:976] (1/7) Epoch 24, batch 1550, loss[loss=0.1677, simple_loss=0.2423, pruned_loss=0.0465, over 4836.00 frames. ], tot_loss[loss=0.1717, simple_loss=0.2445, pruned_loss=0.04946, over 953783.01 frames. ], batch size: 30, lr: 3.06e-03, grad_scale: 16.0 2023-04-27 20:16:10,962 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=133316.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:16:34,614 INFO [finetune.py:976] (1/7) Epoch 24, batch 1600, loss[loss=0.1556, simple_loss=0.2277, pruned_loss=0.04176, over 4912.00 frames. ], tot_loss[loss=0.1716, simple_loss=0.2437, pruned_loss=0.04975, over 953656.42 frames. ], batch size: 37, lr: 3.06e-03, grad_scale: 16.0 2023-04-27 20:16:43,858 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0101, 1.8023, 2.0046, 2.3950, 2.4241, 1.8636, 1.6643, 2.1440], device='cuda:1'), covar=tensor([0.0702, 0.1083, 0.0710, 0.0485, 0.0483, 0.0820, 0.0687, 0.0488], device='cuda:1'), in_proj_covar=tensor([0.0188, 0.0205, 0.0187, 0.0174, 0.0179, 0.0181, 0.0151, 0.0180], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 20:16:54,207 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([3.9055, 3.8911, 2.8672, 4.4751, 3.9943, 3.8883, 1.6770, 3.7999], device='cuda:1'), covar=tensor([0.1817, 0.1241, 0.3338, 0.1604, 0.3218, 0.1932, 0.6123, 0.2530], device='cuda:1'), in_proj_covar=tensor([0.0243, 0.0217, 0.0250, 0.0303, 0.0293, 0.0245, 0.0272, 0.0270], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 20:17:16,920 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=133364.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:17:19,224 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.014e+02 1.451e+02 1.794e+02 2.296e+02 6.131e+02, threshold=3.588e+02, percent-clipped=2.0 2023-04-27 20:17:29,313 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=133374.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 20:17:30,580 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=133376.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:17:42,835 INFO [finetune.py:976] (1/7) Epoch 24, batch 1650, loss[loss=0.1587, simple_loss=0.2312, pruned_loss=0.04309, over 4824.00 frames. ], tot_loss[loss=0.1696, simple_loss=0.2415, pruned_loss=0.04886, over 954755.09 frames. ], batch size: 39, lr: 3.06e-03, grad_scale: 16.0 2023-04-27 20:18:01,036 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.2780, 1.4209, 1.7802, 1.8777, 1.8276, 1.8530, 1.7954, 1.8354], device='cuda:1'), covar=tensor([0.3754, 0.4826, 0.4022, 0.4072, 0.5163, 0.6668, 0.4250, 0.4381], device='cuda:1'), in_proj_covar=tensor([0.0340, 0.0376, 0.0327, 0.0339, 0.0347, 0.0393, 0.0357, 0.0331], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 20:18:02,172 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=133402.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:18:23,685 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.1335, 2.6806, 2.2334, 2.6067, 1.8349, 2.3336, 2.5956, 1.8844], device='cuda:1'), covar=tensor([0.2423, 0.1684, 0.1086, 0.1445, 0.3604, 0.1310, 0.1885, 0.2620], device='cuda:1'), in_proj_covar=tensor([0.0287, 0.0303, 0.0217, 0.0277, 0.0316, 0.0257, 0.0249, 0.0265], device='cuda:1'), out_proj_covar=tensor([1.1483e-04, 1.1978e-04, 8.5416e-05, 1.0917e-04, 1.2732e-04, 1.0132e-04, 1.0069e-04, 1.0485e-04], device='cuda:1') 2023-04-27 20:18:48,095 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=133435.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 20:18:49,169 INFO [finetune.py:976] (1/7) Epoch 24, batch 1700, loss[loss=0.1926, simple_loss=0.271, pruned_loss=0.05712, over 4724.00 frames. ], tot_loss[loss=0.1675, simple_loss=0.2386, pruned_loss=0.04818, over 955699.87 frames. ], batch size: 59, lr: 3.06e-03, grad_scale: 16.0 2023-04-27 20:18:49,283 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=133437.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:19:07,547 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 2023-04-27 20:19:08,005 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=133450.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:19:19,707 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7054, 1.4178, 1.3529, 1.4954, 1.9176, 1.5195, 1.2810, 1.3504], device='cuda:1'), covar=tensor([0.1552, 0.1382, 0.1843, 0.1265, 0.0841, 0.1444, 0.2038, 0.2127], device='cuda:1'), in_proj_covar=tensor([0.0312, 0.0307, 0.0350, 0.0284, 0.0327, 0.0306, 0.0299, 0.0372], device='cuda:1'), out_proj_covar=tensor([6.3920e-05, 6.3234e-05, 7.3539e-05, 5.6960e-05, 6.7306e-05, 6.4024e-05, 6.2226e-05, 7.8938e-05], device='cuda:1') 2023-04-27 20:19:31,793 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.276e+01 1.467e+02 1.697e+02 2.216e+02 4.193e+02, threshold=3.394e+02, percent-clipped=4.0 2023-04-27 20:19:41,291 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=133474.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:19:55,042 INFO [finetune.py:976] (1/7) Epoch 24, batch 1750, loss[loss=0.1409, simple_loss=0.2134, pruned_loss=0.03418, over 4906.00 frames. ], tot_loss[loss=0.1701, simple_loss=0.2419, pruned_loss=0.04922, over 955442.19 frames. ], batch size: 36, lr: 3.06e-03, grad_scale: 16.0 2023-04-27 20:20:47,042 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7669, 2.4149, 1.8536, 1.9341, 1.3080, 1.3539, 1.9026, 1.2539], device='cuda:1'), covar=tensor([0.1593, 0.1383, 0.1445, 0.1587, 0.2350, 0.2000, 0.1002, 0.2098], device='cuda:1'), in_proj_covar=tensor([0.0198, 0.0212, 0.0170, 0.0206, 0.0201, 0.0187, 0.0157, 0.0186], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-27 20:21:01,609 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=133535.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:21:08,174 INFO [finetune.py:976] (1/7) Epoch 24, batch 1800, loss[loss=0.1382, simple_loss=0.2074, pruned_loss=0.03449, over 4723.00 frames. ], tot_loss[loss=0.1717, simple_loss=0.2445, pruned_loss=0.04945, over 955767.30 frames. ], batch size: 23, lr: 3.06e-03, grad_scale: 16.0 2023-04-27 20:21:22,929 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.3420, 1.4866, 1.3633, 1.7029, 1.6632, 1.8648, 1.3831, 3.3092], device='cuda:1'), covar=tensor([0.0587, 0.0805, 0.0804, 0.1198, 0.0642, 0.0516, 0.0702, 0.0167], device='cuda:1'), in_proj_covar=tensor([0.0038, 0.0039, 0.0040, 0.0043, 0.0040, 0.0038, 0.0038, 0.0055], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0013, 0.0013, 0.0015, 0.0014, 0.0013, 0.0013, 0.0016], device='cuda:1') 2023-04-27 20:21:44,188 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.423e+01 1.601e+02 1.910e+02 2.304e+02 3.946e+02, threshold=3.820e+02, percent-clipped=1.0 2023-04-27 20:22:08,895 INFO [finetune.py:976] (1/7) Epoch 24, batch 1850, loss[loss=0.1577, simple_loss=0.2356, pruned_loss=0.03984, over 4746.00 frames. ], tot_loss[loss=0.173, simple_loss=0.2456, pruned_loss=0.05019, over 955171.66 frames. ], batch size: 28, lr: 3.06e-03, grad_scale: 16.0 2023-04-27 20:22:42,122 INFO [finetune.py:976] (1/7) Epoch 24, batch 1900, loss[loss=0.1628, simple_loss=0.2444, pruned_loss=0.04059, over 4784.00 frames. ], tot_loss[loss=0.1728, simple_loss=0.246, pruned_loss=0.04983, over 954893.54 frames. ], batch size: 51, lr: 3.06e-03, grad_scale: 16.0 2023-04-27 20:22:51,303 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=133652.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:23:12,126 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.071e+02 1.575e+02 1.806e+02 2.357e+02 5.107e+02, threshold=3.612e+02, percent-clipped=3.0 2023-04-27 20:23:26,868 INFO [finetune.py:976] (1/7) Epoch 24, batch 1950, loss[loss=0.1443, simple_loss=0.2305, pruned_loss=0.0291, over 4791.00 frames. ], tot_loss[loss=0.1706, simple_loss=0.2441, pruned_loss=0.04852, over 954922.46 frames. ], batch size: 29, lr: 3.06e-03, grad_scale: 16.0 2023-04-27 20:23:43,302 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=133713.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:23:56,164 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=133730.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 20:23:57,384 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=133732.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:24:00,321 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5734, 1.8522, 1.7756, 2.2151, 2.1874, 2.0853, 1.6898, 4.4870], device='cuda:1'), covar=tensor([0.0495, 0.0767, 0.0727, 0.1049, 0.0573, 0.0520, 0.0674, 0.0100], device='cuda:1'), in_proj_covar=tensor([0.0037, 0.0038, 0.0039, 0.0042, 0.0040, 0.0038, 0.0038, 0.0055], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0013, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:1') 2023-04-27 20:24:00,817 INFO [finetune.py:976] (1/7) Epoch 24, batch 2000, loss[loss=0.1353, simple_loss=0.2061, pruned_loss=0.03227, over 4789.00 frames. ], tot_loss[loss=0.1668, simple_loss=0.2395, pruned_loss=0.0471, over 953010.32 frames. ], batch size: 29, lr: 3.06e-03, grad_scale: 16.0 2023-04-27 20:24:25,086 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.642e+01 1.427e+02 1.749e+02 2.000e+02 3.515e+02, threshold=3.497e+02, percent-clipped=0.0 2023-04-27 20:24:25,844 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9207, 2.4717, 1.9313, 1.8815, 1.4276, 1.4515, 2.0643, 1.3932], device='cuda:1'), covar=tensor([0.1674, 0.1397, 0.1353, 0.1720, 0.2274, 0.1892, 0.0932, 0.2022], device='cuda:1'), in_proj_covar=tensor([0.0197, 0.0210, 0.0169, 0.0205, 0.0200, 0.0185, 0.0156, 0.0185], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-27 20:24:53,611 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.3247, 2.9744, 2.4215, 2.7997, 2.1818, 2.5435, 2.6378, 2.0226], device='cuda:1'), covar=tensor([0.2079, 0.1198, 0.0837, 0.1254, 0.2922, 0.1117, 0.2049, 0.2522], device='cuda:1'), in_proj_covar=tensor([0.0287, 0.0302, 0.0216, 0.0277, 0.0315, 0.0256, 0.0250, 0.0265], device='cuda:1'), out_proj_covar=tensor([1.1454e-04, 1.1957e-04, 8.5092e-05, 1.0926e-04, 1.2691e-04, 1.0115e-04, 1.0083e-04, 1.0456e-04], device='cuda:1') 2023-04-27 20:24:55,311 INFO [finetune.py:976] (1/7) Epoch 24, batch 2050, loss[loss=0.1716, simple_loss=0.2488, pruned_loss=0.04718, over 4928.00 frames. ], tot_loss[loss=0.1647, simple_loss=0.2367, pruned_loss=0.04635, over 954094.33 frames. ], batch size: 38, lr: 3.06e-03, grad_scale: 16.0 2023-04-27 20:25:06,946 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([3.4008, 2.7447, 2.5694, 2.7926, 2.5782, 2.7165, 2.6810, 2.5978], device='cuda:1'), covar=tensor([0.3633, 0.4934, 0.4216, 0.4107, 0.4959, 0.5884, 0.5137, 0.4682], device='cuda:1'), in_proj_covar=tensor([0.0341, 0.0376, 0.0328, 0.0339, 0.0348, 0.0395, 0.0358, 0.0331], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 20:25:17,215 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-04-27 20:25:48,584 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=133830.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:25:57,779 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.3773, 3.3621, 0.8824, 1.8537, 1.7514, 2.4784, 1.9302, 1.0175], device='cuda:1'), covar=tensor([0.1302, 0.0950, 0.1971, 0.1170, 0.1125, 0.0962, 0.1332, 0.2051], device='cuda:1'), in_proj_covar=tensor([0.0117, 0.0240, 0.0137, 0.0121, 0.0133, 0.0152, 0.0117, 0.0119], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-27 20:25:58,914 INFO [finetune.py:976] (1/7) Epoch 24, batch 2100, loss[loss=0.2457, simple_loss=0.3057, pruned_loss=0.09287, over 4185.00 frames. ], tot_loss[loss=0.1654, simple_loss=0.2367, pruned_loss=0.04707, over 955030.37 frames. ], batch size: 65, lr: 3.06e-03, grad_scale: 16.0 2023-04-27 20:26:07,182 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([3.7483, 3.6688, 2.7719, 4.3752, 3.7606, 3.7680, 1.7038, 3.7092], device='cuda:1'), covar=tensor([0.1709, 0.1408, 0.3615, 0.1537, 0.3787, 0.1830, 0.5657, 0.2473], device='cuda:1'), in_proj_covar=tensor([0.0246, 0.0220, 0.0253, 0.0306, 0.0296, 0.0247, 0.0275, 0.0274], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 20:26:22,908 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.079e+02 1.658e+02 2.018e+02 2.431e+02 5.100e+02, threshold=4.036e+02, percent-clipped=6.0 2023-04-27 20:26:37,512 INFO [finetune.py:976] (1/7) Epoch 24, batch 2150, loss[loss=0.1796, simple_loss=0.2586, pruned_loss=0.05032, over 4853.00 frames. ], tot_loss[loss=0.1688, simple_loss=0.2405, pruned_loss=0.04854, over 955938.07 frames. ], batch size: 49, lr: 3.05e-03, grad_scale: 16.0 2023-04-27 20:26:50,716 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.9345, 2.8702, 2.2768, 3.3239, 2.9390, 2.9102, 1.1540, 2.7989], device='cuda:1'), covar=tensor([0.2160, 0.1728, 0.3105, 0.2687, 0.3813, 0.2199, 0.6186, 0.2848], device='cuda:1'), in_proj_covar=tensor([0.0246, 0.0219, 0.0253, 0.0306, 0.0296, 0.0247, 0.0275, 0.0273], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 20:27:09,767 INFO [finetune.py:976] (1/7) Epoch 24, batch 2200, loss[loss=0.2068, simple_loss=0.2695, pruned_loss=0.07205, over 4888.00 frames. ], tot_loss[loss=0.169, simple_loss=0.2419, pruned_loss=0.04805, over 954502.45 frames. ], batch size: 32, lr: 3.05e-03, grad_scale: 16.0 2023-04-27 20:27:48,467 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.369e+01 1.555e+02 1.795e+02 2.184e+02 7.045e+02, threshold=3.590e+02, percent-clipped=2.0 2023-04-27 20:28:06,398 INFO [finetune.py:976] (1/7) Epoch 24, batch 2250, loss[loss=0.1748, simple_loss=0.2457, pruned_loss=0.05195, over 4894.00 frames. ], tot_loss[loss=0.1709, simple_loss=0.2441, pruned_loss=0.04886, over 954287.68 frames. ], batch size: 32, lr: 3.05e-03, grad_scale: 16.0 2023-04-27 20:28:06,477 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([3.8271, 3.8518, 2.8153, 4.4243, 3.9789, 3.7995, 1.6886, 3.7607], device='cuda:1'), covar=tensor([0.1803, 0.1215, 0.3000, 0.1530, 0.2464, 0.1737, 0.5750, 0.2527], device='cuda:1'), in_proj_covar=tensor([0.0246, 0.0218, 0.0252, 0.0306, 0.0295, 0.0247, 0.0274, 0.0273], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 20:28:06,522 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7464, 1.9983, 0.8889, 1.4748, 1.6947, 1.6286, 1.5132, 1.6771], device='cuda:1'), covar=tensor([0.0491, 0.0338, 0.0329, 0.0537, 0.0277, 0.0506, 0.0479, 0.0551], device='cuda:1'), in_proj_covar=tensor([0.0028, 0.0023, 0.0022, 0.0028, 0.0019, 0.0028, 0.0027, 0.0028], device='cuda:1'), out_proj_covar=tensor([0.0051, 0.0046, 0.0038, 0.0052, 0.0039, 0.0050, 0.0050, 0.0052], device='cuda:1') 2023-04-27 20:28:14,489 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=133997.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:28:23,002 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=134008.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:28:36,971 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=134030.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 20:28:38,190 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=134032.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:28:41,583 INFO [finetune.py:976] (1/7) Epoch 24, batch 2300, loss[loss=0.1777, simple_loss=0.2454, pruned_loss=0.055, over 4904.00 frames. ], tot_loss[loss=0.1709, simple_loss=0.2446, pruned_loss=0.04864, over 953024.88 frames. ], batch size: 43, lr: 3.05e-03, grad_scale: 16.0 2023-04-27 20:28:55,872 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=134058.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:29:01,260 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.437e+01 1.459e+02 1.887e+02 2.167e+02 3.394e+02, threshold=3.773e+02, percent-clipped=1.0 2023-04-27 20:29:08,084 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=134078.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 20:29:09,323 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=134080.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:29:14,452 INFO [finetune.py:976] (1/7) Epoch 24, batch 2350, loss[loss=0.1486, simple_loss=0.2238, pruned_loss=0.03674, over 4897.00 frames. ], tot_loss[loss=0.1705, simple_loss=0.2434, pruned_loss=0.04879, over 954302.69 frames. ], batch size: 36, lr: 3.05e-03, grad_scale: 16.0 2023-04-27 20:29:42,025 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=134130.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:29:46,678 INFO [finetune.py:976] (1/7) Epoch 24, batch 2400, loss[loss=0.1454, simple_loss=0.2236, pruned_loss=0.03361, over 4923.00 frames. ], tot_loss[loss=0.169, simple_loss=0.2409, pruned_loss=0.04851, over 954390.90 frames. ], batch size: 46, lr: 3.05e-03, grad_scale: 16.0 2023-04-27 20:30:06,804 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.831e+01 1.496e+02 1.764e+02 2.072e+02 3.760e+02, threshold=3.528e+02, percent-clipped=0.0 2023-04-27 20:30:13,577 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=134178.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:30:19,538 INFO [finetune.py:976] (1/7) Epoch 24, batch 2450, loss[loss=0.2055, simple_loss=0.2712, pruned_loss=0.0699, over 4794.00 frames. ], tot_loss[loss=0.166, simple_loss=0.2376, pruned_loss=0.04722, over 954649.58 frames. ], batch size: 51, lr: 3.05e-03, grad_scale: 16.0 2023-04-27 20:30:21,494 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=5.00 vs. limit=5.0 2023-04-27 20:30:56,615 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8162, 1.2457, 4.9367, 4.6280, 4.2539, 4.6548, 4.3686, 4.3150], device='cuda:1'), covar=tensor([0.7180, 0.6176, 0.0953, 0.1613, 0.1244, 0.1585, 0.1800, 0.1746], device='cuda:1'), in_proj_covar=tensor([0.0308, 0.0305, 0.0405, 0.0405, 0.0348, 0.0408, 0.0316, 0.0366], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-27 20:31:07,207 INFO [finetune.py:976] (1/7) Epoch 24, batch 2500, loss[loss=0.1704, simple_loss=0.2507, pruned_loss=0.04501, over 4859.00 frames. ], tot_loss[loss=0.1671, simple_loss=0.2389, pruned_loss=0.04766, over 955395.60 frames. ], batch size: 44, lr: 3.05e-03, grad_scale: 16.0 2023-04-27 20:31:39,322 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.116e+02 1.581e+02 1.827e+02 2.358e+02 4.176e+02, threshold=3.655e+02, percent-clipped=4.0 2023-04-27 20:32:02,532 INFO [finetune.py:976] (1/7) Epoch 24, batch 2550, loss[loss=0.1452, simple_loss=0.2332, pruned_loss=0.0286, over 4755.00 frames. ], tot_loss[loss=0.1699, simple_loss=0.2426, pruned_loss=0.04857, over 953912.80 frames. ], batch size: 54, lr: 3.05e-03, grad_scale: 16.0 2023-04-27 20:32:26,932 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=134304.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:32:34,603 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=134308.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:32:35,207 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=134309.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:33:07,652 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.2281, 2.9654, 1.0853, 1.7118, 1.6720, 2.3139, 1.7425, 1.0079], device='cuda:1'), covar=tensor([0.1395, 0.0897, 0.1640, 0.1224, 0.1123, 0.0855, 0.1464, 0.1806], device='cuda:1'), in_proj_covar=tensor([0.0117, 0.0240, 0.0137, 0.0121, 0.0133, 0.0152, 0.0117, 0.0120], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-27 20:33:08,333 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.4544, 1.6971, 1.9004, 2.0113, 1.8580, 1.8178, 1.9691, 1.8655], device='cuda:1'), covar=tensor([0.4365, 0.5818, 0.4534, 0.4262, 0.5802, 0.7470, 0.5232, 0.5065], device='cuda:1'), in_proj_covar=tensor([0.0340, 0.0374, 0.0326, 0.0337, 0.0347, 0.0392, 0.0357, 0.0329], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 20:33:08,806 INFO [finetune.py:976] (1/7) Epoch 24, batch 2600, loss[loss=0.1975, simple_loss=0.2697, pruned_loss=0.06269, over 4810.00 frames. ], tot_loss[loss=0.1703, simple_loss=0.2432, pruned_loss=0.04867, over 953108.01 frames. ], batch size: 40, lr: 3.05e-03, grad_scale: 16.0 2023-04-27 20:33:21,905 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.5077, 1.7648, 1.9106, 2.0088, 1.8071, 1.8422, 2.0039, 1.8835], device='cuda:1'), covar=tensor([0.3495, 0.5543, 0.4212, 0.4030, 0.5486, 0.6983, 0.5216, 0.4970], device='cuda:1'), in_proj_covar=tensor([0.0340, 0.0374, 0.0326, 0.0337, 0.0346, 0.0392, 0.0357, 0.0329], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 20:33:27,027 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=134353.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:33:29,506 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=134356.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:33:42,151 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=134365.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:33:43,228 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.125e+02 1.588e+02 1.844e+02 2.194e+02 4.309e+02, threshold=3.689e+02, percent-clipped=2.0 2023-04-27 20:33:45,765 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=134370.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:34:07,321 INFO [finetune.py:976] (1/7) Epoch 24, batch 2650, loss[loss=0.1412, simple_loss=0.2258, pruned_loss=0.0283, over 4771.00 frames. ], tot_loss[loss=0.1709, simple_loss=0.2443, pruned_loss=0.04869, over 953586.86 frames. ], batch size: 26, lr: 3.05e-03, grad_scale: 16.0 2023-04-27 20:34:16,587 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.57 vs. limit=2.0 2023-04-27 20:35:02,815 INFO [finetune.py:976] (1/7) Epoch 24, batch 2700, loss[loss=0.1777, simple_loss=0.2402, pruned_loss=0.0576, over 4773.00 frames. ], tot_loss[loss=0.1689, simple_loss=0.2424, pruned_loss=0.04774, over 952358.03 frames. ], batch size: 26, lr: 3.05e-03, grad_scale: 16.0 2023-04-27 20:35:23,806 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.513e+01 1.459e+02 1.776e+02 2.011e+02 3.451e+02, threshold=3.553e+02, percent-clipped=0.0 2023-04-27 20:35:36,506 INFO [finetune.py:976] (1/7) Epoch 24, batch 2750, loss[loss=0.1458, simple_loss=0.2285, pruned_loss=0.03157, over 4910.00 frames. ], tot_loss[loss=0.1681, simple_loss=0.2408, pruned_loss=0.04771, over 951470.84 frames. ], batch size: 36, lr: 3.05e-03, grad_scale: 16.0 2023-04-27 20:35:54,521 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8763, 1.5035, 1.5099, 1.6221, 2.0402, 1.6966, 1.3768, 1.4258], device='cuda:1'), covar=tensor([0.1572, 0.1457, 0.1764, 0.1327, 0.0883, 0.1527, 0.1889, 0.2202], device='cuda:1'), in_proj_covar=tensor([0.0309, 0.0304, 0.0347, 0.0281, 0.0325, 0.0303, 0.0296, 0.0369], device='cuda:1'), out_proj_covar=tensor([6.3156e-05, 6.2699e-05, 7.2919e-05, 5.6413e-05, 6.6977e-05, 6.3568e-05, 6.1595e-05, 7.8228e-05], device='cuda:1') 2023-04-27 20:36:16,408 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-04-27 20:36:43,196 INFO [finetune.py:976] (1/7) Epoch 24, batch 2800, loss[loss=0.1822, simple_loss=0.2449, pruned_loss=0.0598, over 4912.00 frames. ], tot_loss[loss=0.1656, simple_loss=0.2378, pruned_loss=0.04672, over 952113.49 frames. ], batch size: 32, lr: 3.05e-03, grad_scale: 16.0 2023-04-27 20:37:23,286 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.232e+01 1.527e+02 1.797e+02 2.170e+02 3.863e+02, threshold=3.594e+02, percent-clipped=3.0 2023-04-27 20:37:47,790 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5877, 3.5313, 0.7651, 1.9416, 2.0901, 2.5476, 1.9599, 1.0287], device='cuda:1'), covar=tensor([0.1418, 0.1047, 0.2386, 0.1284, 0.1123, 0.1021, 0.1627, 0.2070], device='cuda:1'), in_proj_covar=tensor([0.0118, 0.0241, 0.0137, 0.0122, 0.0134, 0.0153, 0.0118, 0.0120], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-27 20:37:48,299 INFO [finetune.py:976] (1/7) Epoch 24, batch 2850, loss[loss=0.169, simple_loss=0.2334, pruned_loss=0.0523, over 4829.00 frames. ], tot_loss[loss=0.1654, simple_loss=0.2369, pruned_loss=0.04699, over 948934.71 frames. ], batch size: 33, lr: 3.05e-03, grad_scale: 32.0 2023-04-27 20:38:59,968 INFO [finetune.py:976] (1/7) Epoch 24, batch 2900, loss[loss=0.1633, simple_loss=0.2383, pruned_loss=0.04418, over 4911.00 frames. ], tot_loss[loss=0.1687, simple_loss=0.2409, pruned_loss=0.04824, over 951350.15 frames. ], batch size: 37, lr: 3.05e-03, grad_scale: 32.0 2023-04-27 20:39:15,051 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=134653.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:39:24,533 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=134660.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:39:33,040 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=134665.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:39:34,174 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.095e+02 1.576e+02 1.923e+02 2.277e+02 4.210e+02, threshold=3.845e+02, percent-clipped=2.0 2023-04-27 20:40:04,275 INFO [finetune.py:976] (1/7) Epoch 24, batch 2950, loss[loss=0.156, simple_loss=0.2419, pruned_loss=0.0351, over 4854.00 frames. ], tot_loss[loss=0.1695, simple_loss=0.2427, pruned_loss=0.0482, over 952597.47 frames. ], batch size: 44, lr: 3.05e-03, grad_scale: 32.0 2023-04-27 20:40:14,283 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.0083, 0.9346, 1.1522, 1.1341, 0.9866, 0.8684, 1.0048, 0.5836], device='cuda:1'), covar=tensor([0.0521, 0.0483, 0.0448, 0.0492, 0.0637, 0.1092, 0.0450, 0.0591], device='cuda:1'), in_proj_covar=tensor([0.0069, 0.0067, 0.0066, 0.0067, 0.0074, 0.0094, 0.0072, 0.0063], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-27 20:40:18,487 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=134701.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:40:28,853 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8113, 2.5420, 1.8910, 1.8834, 1.2967, 1.3585, 1.8821, 1.2829], device='cuda:1'), covar=tensor([0.1692, 0.1299, 0.1348, 0.1629, 0.2253, 0.1890, 0.0953, 0.2003], device='cuda:1'), in_proj_covar=tensor([0.0196, 0.0208, 0.0168, 0.0204, 0.0198, 0.0183, 0.0155, 0.0185], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-27 20:40:47,452 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5681, 1.4863, 1.8271, 1.8648, 1.4131, 1.2640, 1.5495, 0.8856], device='cuda:1'), covar=tensor([0.0494, 0.0546, 0.0367, 0.0524, 0.0690, 0.1122, 0.0576, 0.0608], device='cuda:1'), in_proj_covar=tensor([0.0069, 0.0067, 0.0066, 0.0067, 0.0074, 0.0095, 0.0073, 0.0064], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-27 20:41:09,548 INFO [finetune.py:976] (1/7) Epoch 24, batch 3000, loss[loss=0.1434, simple_loss=0.2142, pruned_loss=0.03631, over 4766.00 frames. ], tot_loss[loss=0.1706, simple_loss=0.2436, pruned_loss=0.04882, over 951653.62 frames. ], batch size: 26, lr: 3.05e-03, grad_scale: 32.0 2023-04-27 20:41:09,548 INFO [finetune.py:1001] (1/7) Computing validation loss 2023-04-27 20:41:25,508 INFO [finetune.py:1010] (1/7) Epoch 24, validation: loss=0.1526, simple_loss=0.2221, pruned_loss=0.04154, over 2265189.00 frames. 2023-04-27 20:41:25,508 INFO [finetune.py:1011] (1/7) Maximum memory allocated so far is 6391MB 2023-04-27 20:41:44,218 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.083e+02 1.565e+02 1.917e+02 2.257e+02 3.857e+02, threshold=3.833e+02, percent-clipped=1.0 2023-04-27 20:41:57,445 INFO [finetune.py:976] (1/7) Epoch 24, batch 3050, loss[loss=0.176, simple_loss=0.2488, pruned_loss=0.05159, over 4779.00 frames. ], tot_loss[loss=0.172, simple_loss=0.2455, pruned_loss=0.04927, over 953795.86 frames. ], batch size: 51, lr: 3.05e-03, grad_scale: 32.0 2023-04-27 20:42:18,195 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5201, 1.4335, 1.8639, 1.8202, 1.3404, 1.2220, 1.5389, 0.9733], device='cuda:1'), covar=tensor([0.0494, 0.0528, 0.0392, 0.0502, 0.0684, 0.1055, 0.0549, 0.0577], device='cuda:1'), in_proj_covar=tensor([0.0070, 0.0067, 0.0066, 0.0068, 0.0074, 0.0095, 0.0073, 0.0064], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-27 20:42:30,089 INFO [finetune.py:976] (1/7) Epoch 24, batch 3100, loss[loss=0.1661, simple_loss=0.2355, pruned_loss=0.04832, over 4732.00 frames. ], tot_loss[loss=0.1704, simple_loss=0.2438, pruned_loss=0.04849, over 955125.13 frames. ], batch size: 59, lr: 3.05e-03, grad_scale: 32.0 2023-04-27 20:42:47,417 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=134862.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:42:50,371 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.815e+01 1.536e+02 1.798e+02 2.119e+02 3.796e+02, threshold=3.595e+02, percent-clipped=0.0 2023-04-27 20:43:02,614 INFO [finetune.py:976] (1/7) Epoch 24, batch 3150, loss[loss=0.1579, simple_loss=0.2275, pruned_loss=0.04411, over 4920.00 frames. ], tot_loss[loss=0.1686, simple_loss=0.2414, pruned_loss=0.0479, over 953058.78 frames. ], batch size: 37, lr: 3.05e-03, grad_scale: 32.0 2023-04-27 20:43:28,107 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=134923.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:43:36,613 INFO [finetune.py:976] (1/7) Epoch 24, batch 3200, loss[loss=0.1637, simple_loss=0.2226, pruned_loss=0.05235, over 4911.00 frames. ], tot_loss[loss=0.1664, simple_loss=0.2387, pruned_loss=0.04708, over 953838.33 frames. ], batch size: 43, lr: 3.05e-03, grad_scale: 32.0 2023-04-27 20:43:52,066 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.50 vs. limit=5.0 2023-04-27 20:43:53,563 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=134960.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:43:56,582 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=134965.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:43:57,659 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.048e+02 1.515e+02 1.832e+02 2.291e+02 6.391e+02, threshold=3.663e+02, percent-clipped=4.0 2023-04-27 20:44:10,013 INFO [finetune.py:976] (1/7) Epoch 24, batch 3250, loss[loss=0.1407, simple_loss=0.2069, pruned_loss=0.03726, over 3919.00 frames. ], tot_loss[loss=0.1674, simple_loss=0.2392, pruned_loss=0.04778, over 952171.72 frames. ], batch size: 17, lr: 3.05e-03, grad_scale: 32.0 2023-04-27 20:44:11,980 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.14 vs. limit=2.0 2023-04-27 20:44:25,316 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=135008.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:44:28,803 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=135013.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:44:43,397 INFO [finetune.py:976] (1/7) Epoch 24, batch 3300, loss[loss=0.1868, simple_loss=0.2668, pruned_loss=0.05338, over 4928.00 frames. ], tot_loss[loss=0.1697, simple_loss=0.2426, pruned_loss=0.04843, over 952510.67 frames. ], batch size: 38, lr: 3.05e-03, grad_scale: 32.0 2023-04-27 20:45:15,629 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.097e+02 1.494e+02 1.834e+02 2.222e+02 6.565e+02, threshold=3.668e+02, percent-clipped=2.0 2023-04-27 20:45:44,560 INFO [finetune.py:976] (1/7) Epoch 24, batch 3350, loss[loss=0.1839, simple_loss=0.2503, pruned_loss=0.05878, over 4281.00 frames. ], tot_loss[loss=0.1714, simple_loss=0.245, pruned_loss=0.04886, over 954502.63 frames. ], batch size: 65, lr: 3.05e-03, grad_scale: 32.0 2023-04-27 20:46:15,698 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6515, 1.7016, 0.7372, 1.3855, 1.6630, 1.5112, 1.4193, 1.5177], device='cuda:1'), covar=tensor([0.0446, 0.0340, 0.0342, 0.0529, 0.0268, 0.0467, 0.0470, 0.0495], device='cuda:1'), in_proj_covar=tensor([0.0028, 0.0023, 0.0022, 0.0028, 0.0019, 0.0028, 0.0027, 0.0028], device='cuda:1'), out_proj_covar=tensor([0.0051, 0.0046, 0.0038, 0.0052, 0.0039, 0.0050, 0.0050, 0.0052], device='cuda:1') 2023-04-27 20:46:40,850 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5728, 0.6240, 1.4787, 1.9290, 1.6465, 1.5188, 1.5283, 1.5268], device='cuda:1'), covar=tensor([0.4389, 0.6482, 0.5816, 0.6092, 0.5804, 0.7073, 0.7464, 0.8206], device='cuda:1'), in_proj_covar=tensor([0.0435, 0.0418, 0.0510, 0.0506, 0.0465, 0.0498, 0.0502, 0.0513], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 20:46:49,145 INFO [finetune.py:976] (1/7) Epoch 24, batch 3400, loss[loss=0.169, simple_loss=0.2324, pruned_loss=0.05281, over 4746.00 frames. ], tot_loss[loss=0.1729, simple_loss=0.2466, pruned_loss=0.04965, over 956198.60 frames. ], batch size: 27, lr: 3.05e-03, grad_scale: 32.0 2023-04-27 20:46:51,742 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.3309, 1.5366, 1.4260, 1.8279, 1.6954, 1.7400, 1.4489, 3.0068], device='cuda:1'), covar=tensor([0.0628, 0.0765, 0.0814, 0.1067, 0.0623, 0.0509, 0.0688, 0.0237], device='cuda:1'), in_proj_covar=tensor([0.0038, 0.0039, 0.0040, 0.0043, 0.0040, 0.0038, 0.0038, 0.0055], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0013, 0.0014, 0.0015, 0.0014, 0.0013, 0.0013, 0.0016], device='cuda:1') 2023-04-27 20:46:54,294 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.61 vs. limit=2.0 2023-04-27 20:47:25,516 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.127e+02 1.583e+02 1.888e+02 2.345e+02 4.447e+02, threshold=3.777e+02, percent-clipped=3.0 2023-04-27 20:47:33,952 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=135171.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:47:35,754 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.76 vs. limit=2.0 2023-04-27 20:47:42,924 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-04-27 20:47:54,152 INFO [finetune.py:976] (1/7) Epoch 24, batch 3450, loss[loss=0.1628, simple_loss=0.2372, pruned_loss=0.04425, over 4834.00 frames. ], tot_loss[loss=0.1725, simple_loss=0.2465, pruned_loss=0.0492, over 956556.21 frames. ], batch size: 47, lr: 3.05e-03, grad_scale: 32.0 2023-04-27 20:48:27,447 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=135218.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:48:36,540 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=135232.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:48:39,438 INFO [finetune.py:976] (1/7) Epoch 24, batch 3500, loss[loss=0.1634, simple_loss=0.2302, pruned_loss=0.04827, over 4913.00 frames. ], tot_loss[loss=0.1701, simple_loss=0.2436, pruned_loss=0.04827, over 957264.58 frames. ], batch size: 43, lr: 3.04e-03, grad_scale: 32.0 2023-04-27 20:48:59,253 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.073e+02 1.566e+02 1.818e+02 2.142e+02 4.699e+02, threshold=3.637e+02, percent-clipped=2.0 2023-04-27 20:49:13,318 INFO [finetune.py:976] (1/7) Epoch 24, batch 3550, loss[loss=0.1274, simple_loss=0.2069, pruned_loss=0.02393, over 4790.00 frames. ], tot_loss[loss=0.1674, simple_loss=0.2403, pruned_loss=0.04722, over 955285.07 frames. ], batch size: 29, lr: 3.04e-03, grad_scale: 32.0 2023-04-27 20:49:17,142 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.2139, 1.6678, 2.0366, 2.2406, 2.0418, 1.6315, 1.1117, 1.7185], device='cuda:1'), covar=tensor([0.3096, 0.3148, 0.1727, 0.2249, 0.2677, 0.2598, 0.4294, 0.1989], device='cuda:1'), in_proj_covar=tensor([0.0296, 0.0246, 0.0229, 0.0316, 0.0222, 0.0236, 0.0230, 0.0186], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-27 20:49:47,282 INFO [finetune.py:976] (1/7) Epoch 24, batch 3600, loss[loss=0.1445, simple_loss=0.2299, pruned_loss=0.02956, over 4751.00 frames. ], tot_loss[loss=0.1646, simple_loss=0.2376, pruned_loss=0.04578, over 956452.10 frames. ], batch size: 27, lr: 3.04e-03, grad_scale: 32.0 2023-04-27 20:50:01,430 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 2023-04-27 20:50:05,962 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.034e+02 1.602e+02 1.884e+02 2.235e+02 3.277e+02, threshold=3.769e+02, percent-clipped=0.0 2023-04-27 20:50:20,184 INFO [finetune.py:976] (1/7) Epoch 24, batch 3650, loss[loss=0.1828, simple_loss=0.2525, pruned_loss=0.05659, over 4750.00 frames. ], tot_loss[loss=0.1667, simple_loss=0.2396, pruned_loss=0.04694, over 957439.15 frames. ], batch size: 27, lr: 3.04e-03, grad_scale: 32.0 2023-04-27 20:50:28,273 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-04-27 20:50:53,969 INFO [finetune.py:976] (1/7) Epoch 24, batch 3700, loss[loss=0.212, simple_loss=0.2899, pruned_loss=0.06706, over 4843.00 frames. ], tot_loss[loss=0.1707, simple_loss=0.2445, pruned_loss=0.04848, over 957182.39 frames. ], batch size: 44, lr: 3.04e-03, grad_scale: 32.0 2023-04-27 20:50:57,585 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5173, 2.5307, 2.0665, 2.1895, 2.6222, 2.2662, 3.4892, 1.9973], device='cuda:1'), covar=tensor([0.3681, 0.2341, 0.4760, 0.3176, 0.1797, 0.2526, 0.1477, 0.4053], device='cuda:1'), in_proj_covar=tensor([0.0339, 0.0353, 0.0425, 0.0351, 0.0379, 0.0377, 0.0369, 0.0421], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 20:51:12,482 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.055e+02 1.617e+02 1.866e+02 2.198e+02 4.179e+02, threshold=3.733e+02, percent-clipped=1.0 2023-04-27 20:51:27,090 INFO [finetune.py:976] (1/7) Epoch 24, batch 3750, loss[loss=0.1587, simple_loss=0.2343, pruned_loss=0.04158, over 4775.00 frames. ], tot_loss[loss=0.1721, simple_loss=0.2458, pruned_loss=0.04919, over 955943.31 frames. ], batch size: 28, lr: 3.04e-03, grad_scale: 32.0 2023-04-27 20:52:05,332 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=135518.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:52:16,514 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=135527.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:52:16,524 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.1020, 2.5368, 0.7748, 1.4032, 1.5090, 1.8800, 1.5918, 0.8502], device='cuda:1'), covar=tensor([0.1471, 0.0960, 0.1787, 0.1341, 0.1129, 0.0973, 0.1548, 0.1723], device='cuda:1'), in_proj_covar=tensor([0.0116, 0.0237, 0.0135, 0.0120, 0.0131, 0.0151, 0.0116, 0.0118], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-27 20:52:29,077 INFO [finetune.py:976] (1/7) Epoch 24, batch 3800, loss[loss=0.1917, simple_loss=0.2559, pruned_loss=0.06374, over 4832.00 frames. ], tot_loss[loss=0.1727, simple_loss=0.2465, pruned_loss=0.04949, over 956776.56 frames. ], batch size: 30, lr: 3.04e-03, grad_scale: 32.0 2023-04-27 20:53:08,358 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=135566.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:53:08,430 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7264, 1.3582, 1.4140, 1.4035, 1.8632, 1.5206, 1.2781, 1.3257], device='cuda:1'), covar=tensor([0.1511, 0.1223, 0.1749, 0.1201, 0.0729, 0.1389, 0.1658, 0.2028], device='cuda:1'), in_proj_covar=tensor([0.0311, 0.0306, 0.0349, 0.0283, 0.0326, 0.0305, 0.0297, 0.0369], device='cuda:1'), out_proj_covar=tensor([6.3589e-05, 6.3084e-05, 7.3463e-05, 5.6749e-05, 6.7010e-05, 6.3833e-05, 6.1753e-05, 7.8171e-05], device='cuda:1') 2023-04-27 20:53:08,920 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.795e+01 1.517e+02 1.799e+02 2.182e+02 4.922e+02, threshold=3.597e+02, percent-clipped=3.0 2023-04-27 20:53:32,877 INFO [finetune.py:976] (1/7) Epoch 24, batch 3850, loss[loss=0.1508, simple_loss=0.2305, pruned_loss=0.03558, over 4856.00 frames. ], tot_loss[loss=0.1718, simple_loss=0.245, pruned_loss=0.04925, over 956326.80 frames. ], batch size: 44, lr: 3.04e-03, grad_scale: 32.0 2023-04-27 20:53:40,269 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5812, 1.5055, 4.2976, 4.0557, 3.7428, 4.0524, 4.0349, 3.8263], device='cuda:1'), covar=tensor([0.6902, 0.5401, 0.1123, 0.1629, 0.1142, 0.1521, 0.1512, 0.1731], device='cuda:1'), in_proj_covar=tensor([0.0310, 0.0306, 0.0405, 0.0407, 0.0349, 0.0408, 0.0318, 0.0368], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-27 20:53:59,381 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.78 vs. limit=2.0 2023-04-27 20:54:00,565 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9782, 1.5736, 2.1000, 2.4306, 2.0783, 1.9312, 2.0321, 1.9610], device='cuda:1'), covar=tensor([0.4364, 0.6404, 0.6010, 0.5431, 0.5640, 0.8180, 0.7816, 0.8116], device='cuda:1'), in_proj_covar=tensor([0.0434, 0.0418, 0.0509, 0.0505, 0.0464, 0.0496, 0.0501, 0.0512], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 20:54:23,659 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=135630.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:54:33,840 INFO [finetune.py:976] (1/7) Epoch 24, batch 3900, loss[loss=0.2001, simple_loss=0.2563, pruned_loss=0.07201, over 4847.00 frames. ], tot_loss[loss=0.1687, simple_loss=0.2414, pruned_loss=0.04804, over 956126.62 frames. ], batch size: 49, lr: 3.04e-03, grad_scale: 32.0 2023-04-27 20:55:06,842 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7262, 0.7696, 1.6376, 2.0359, 1.7835, 1.6240, 1.6237, 1.6623], device='cuda:1'), covar=tensor([0.4454, 0.6583, 0.6071, 0.6223, 0.6211, 0.7294, 0.7310, 0.8025], device='cuda:1'), in_proj_covar=tensor([0.0435, 0.0419, 0.0510, 0.0506, 0.0464, 0.0497, 0.0502, 0.0514], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 20:55:15,855 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.029e+02 1.536e+02 1.810e+02 2.224e+02 3.785e+02, threshold=3.619e+02, percent-clipped=2.0 2023-04-27 20:55:44,887 INFO [finetune.py:976] (1/7) Epoch 24, batch 3950, loss[loss=0.169, simple_loss=0.2447, pruned_loss=0.04665, over 4778.00 frames. ], tot_loss[loss=0.1658, simple_loss=0.2381, pruned_loss=0.04676, over 954843.02 frames. ], batch size: 29, lr: 3.04e-03, grad_scale: 32.0 2023-04-27 20:55:47,447 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=135691.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:56:44,335 INFO [finetune.py:976] (1/7) Epoch 24, batch 4000, loss[loss=0.1881, simple_loss=0.2625, pruned_loss=0.0569, over 4823.00 frames. ], tot_loss[loss=0.1664, simple_loss=0.2384, pruned_loss=0.04716, over 957111.62 frames. ], batch size: 39, lr: 3.04e-03, grad_scale: 32.0 2023-04-27 20:57:19,946 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0006, 2.6959, 1.1388, 1.5736, 1.9739, 1.2311, 3.4188, 1.7983], device='cuda:1'), covar=tensor([0.0661, 0.0634, 0.0816, 0.1193, 0.0518, 0.0962, 0.0201, 0.0611], device='cuda:1'), in_proj_covar=tensor([0.0050, 0.0064, 0.0047, 0.0046, 0.0049, 0.0051, 0.0072, 0.0050], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0007, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:1') 2023-04-27 20:57:22,843 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.079e+02 1.579e+02 1.856e+02 2.321e+02 4.338e+02, threshold=3.712e+02, percent-clipped=1.0 2023-04-27 20:57:52,044 INFO [finetune.py:976] (1/7) Epoch 24, batch 4050, loss[loss=0.1397, simple_loss=0.2139, pruned_loss=0.03275, over 4679.00 frames. ], tot_loss[loss=0.1701, simple_loss=0.2421, pruned_loss=0.04906, over 956504.58 frames. ], batch size: 23, lr: 3.04e-03, grad_scale: 32.0 2023-04-27 20:58:39,876 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=135827.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:58:52,538 INFO [finetune.py:976] (1/7) Epoch 24, batch 4100, loss[loss=0.1419, simple_loss=0.2134, pruned_loss=0.03524, over 4676.00 frames. ], tot_loss[loss=0.1699, simple_loss=0.2427, pruned_loss=0.04851, over 954145.91 frames. ], batch size: 23, lr: 3.04e-03, grad_scale: 32.0 2023-04-27 20:59:28,241 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.032e+02 1.687e+02 1.984e+02 2.360e+02 5.004e+02, threshold=3.968e+02, percent-clipped=3.0 2023-04-27 20:59:29,552 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4770, 1.0971, 4.2360, 3.9983, 3.7171, 3.9906, 3.9180, 3.7702], device='cuda:1'), covar=tensor([0.7133, 0.6552, 0.1120, 0.1630, 0.1128, 0.1610, 0.1818, 0.1689], device='cuda:1'), in_proj_covar=tensor([0.0310, 0.0306, 0.0405, 0.0407, 0.0349, 0.0409, 0.0319, 0.0368], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-27 20:59:33,106 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=135875.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:59:34,547 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.83 vs. limit=5.0 2023-04-27 20:59:40,397 INFO [finetune.py:976] (1/7) Epoch 24, batch 4150, loss[loss=0.1763, simple_loss=0.2546, pruned_loss=0.04904, over 4807.00 frames. ], tot_loss[loss=0.1701, simple_loss=0.2433, pruned_loss=0.04844, over 951861.62 frames. ], batch size: 51, lr: 3.04e-03, grad_scale: 32.0 2023-04-27 21:00:14,126 INFO [finetune.py:976] (1/7) Epoch 24, batch 4200, loss[loss=0.1418, simple_loss=0.2259, pruned_loss=0.02885, over 4877.00 frames. ], tot_loss[loss=0.1702, simple_loss=0.244, pruned_loss=0.04818, over 952188.36 frames. ], batch size: 43, lr: 3.04e-03, grad_scale: 32.0 2023-04-27 21:00:23,667 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-04-27 21:00:32,453 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-04-27 21:00:35,175 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.056e+02 1.525e+02 1.774e+02 2.091e+02 5.050e+02, threshold=3.548e+02, percent-clipped=3.0 2023-04-27 21:00:42,424 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 2023-04-27 21:00:44,758 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=135982.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:00:47,164 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=135986.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:00:47,700 INFO [finetune.py:976] (1/7) Epoch 24, batch 4250, loss[loss=0.2103, simple_loss=0.2673, pruned_loss=0.0766, over 4928.00 frames. ], tot_loss[loss=0.169, simple_loss=0.242, pruned_loss=0.04796, over 953751.00 frames. ], batch size: 38, lr: 3.04e-03, grad_scale: 32.0 2023-04-27 21:00:52,089 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=135994.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:01:22,183 INFO [finetune.py:976] (1/7) Epoch 24, batch 4300, loss[loss=0.1388, simple_loss=0.2237, pruned_loss=0.02692, over 4837.00 frames. ], tot_loss[loss=0.1672, simple_loss=0.2397, pruned_loss=0.04739, over 953000.08 frames. ], batch size: 30, lr: 3.04e-03, grad_scale: 32.0 2023-04-27 21:01:26,004 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=136043.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:01:34,328 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=136055.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:01:42,941 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.901e+01 1.479e+02 1.753e+02 2.087e+02 3.589e+02, threshold=3.506e+02, percent-clipped=1.0 2023-04-27 21:01:55,531 INFO [finetune.py:976] (1/7) Epoch 24, batch 4350, loss[loss=0.147, simple_loss=0.2268, pruned_loss=0.03364, over 4758.00 frames. ], tot_loss[loss=0.1655, simple_loss=0.2372, pruned_loss=0.04693, over 951475.08 frames. ], batch size: 27, lr: 3.04e-03, grad_scale: 32.0 2023-04-27 21:01:55,741 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-04-27 21:02:33,981 INFO [finetune.py:976] (1/7) Epoch 24, batch 4400, loss[loss=0.1758, simple_loss=0.2542, pruned_loss=0.04871, over 4784.00 frames. ], tot_loss[loss=0.1665, simple_loss=0.2384, pruned_loss=0.04734, over 951556.59 frames. ], batch size: 29, lr: 3.04e-03, grad_scale: 32.0 2023-04-27 21:02:41,368 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.3175, 1.2527, 1.3524, 1.5058, 1.6082, 1.2705, 1.0000, 1.4798], device='cuda:1'), covar=tensor([0.0763, 0.1203, 0.0772, 0.0589, 0.0639, 0.0732, 0.0832, 0.0527], device='cuda:1'), in_proj_covar=tensor([0.0189, 0.0205, 0.0187, 0.0176, 0.0180, 0.0181, 0.0152, 0.0180], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 21:02:42,582 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=136142.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:02:52,609 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 2023-04-27 21:03:16,196 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.001e+02 1.622e+02 1.920e+02 2.374e+02 4.375e+02, threshold=3.840e+02, percent-clipped=3.0 2023-04-27 21:03:40,889 INFO [finetune.py:976] (1/7) Epoch 24, batch 4450, loss[loss=0.1783, simple_loss=0.2397, pruned_loss=0.05845, over 4218.00 frames. ], tot_loss[loss=0.1689, simple_loss=0.2418, pruned_loss=0.04799, over 953243.89 frames. ], batch size: 65, lr: 3.04e-03, grad_scale: 32.0 2023-04-27 21:04:02,262 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=136203.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:04:53,816 INFO [finetune.py:976] (1/7) Epoch 24, batch 4500, loss[loss=0.2059, simple_loss=0.2718, pruned_loss=0.07004, over 4835.00 frames. ], tot_loss[loss=0.1712, simple_loss=0.2441, pruned_loss=0.04917, over 951897.14 frames. ], batch size: 44, lr: 3.04e-03, grad_scale: 32.0 2023-04-27 21:05:29,627 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.760e+01 1.501e+02 1.833e+02 2.147e+02 4.216e+02, threshold=3.666e+02, percent-clipped=1.0 2023-04-27 21:05:58,042 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=136286.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:05:58,524 INFO [finetune.py:976] (1/7) Epoch 24, batch 4550, loss[loss=0.1403, simple_loss=0.2002, pruned_loss=0.04015, over 4239.00 frames. ], tot_loss[loss=0.1697, simple_loss=0.2429, pruned_loss=0.04828, over 952365.44 frames. ], batch size: 18, lr: 3.04e-03, grad_scale: 32.0 2023-04-27 21:06:56,730 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=136334.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:06:57,404 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=136335.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:07:04,271 INFO [finetune.py:976] (1/7) Epoch 24, batch 4600, loss[loss=0.203, simple_loss=0.2642, pruned_loss=0.07087, over 4872.00 frames. ], tot_loss[loss=0.17, simple_loss=0.2431, pruned_loss=0.04847, over 950845.75 frames. ], batch size: 34, lr: 3.04e-03, grad_scale: 32.0 2023-04-27 21:07:04,977 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=136338.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:07:18,411 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=136350.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:07:18,470 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5216, 1.4932, 1.9811, 2.0000, 1.4058, 1.2730, 1.6486, 1.1151], device='cuda:1'), covar=tensor([0.0546, 0.0604, 0.0329, 0.0500, 0.0751, 0.1191, 0.0588, 0.0591], device='cuda:1'), in_proj_covar=tensor([0.0070, 0.0067, 0.0066, 0.0067, 0.0074, 0.0094, 0.0072, 0.0063], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-27 21:07:40,423 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.023e+02 1.579e+02 1.858e+02 2.127e+02 3.942e+02, threshold=3.716e+02, percent-clipped=1.0 2023-04-27 21:08:03,894 INFO [finetune.py:976] (1/7) Epoch 24, batch 4650, loss[loss=0.1846, simple_loss=0.255, pruned_loss=0.05707, over 4837.00 frames. ], tot_loss[loss=0.1688, simple_loss=0.2413, pruned_loss=0.04811, over 952174.36 frames. ], batch size: 33, lr: 3.04e-03, grad_scale: 32.0 2023-04-27 21:08:20,813 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=136396.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:08:33,856 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.4947, 1.8101, 1.9053, 2.0577, 1.8703, 1.9277, 1.9182, 1.9460], device='cuda:1'), covar=tensor([0.4623, 0.5416, 0.4465, 0.4080, 0.5886, 0.7490, 0.5294, 0.4840], device='cuda:1'), in_proj_covar=tensor([0.0343, 0.0376, 0.0327, 0.0342, 0.0350, 0.0397, 0.0360, 0.0333], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 21:08:54,726 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-04-27 21:09:13,811 INFO [finetune.py:976] (1/7) Epoch 24, batch 4700, loss[loss=0.1428, simple_loss=0.2164, pruned_loss=0.03461, over 4824.00 frames. ], tot_loss[loss=0.1658, simple_loss=0.238, pruned_loss=0.04678, over 953656.17 frames. ], batch size: 39, lr: 3.04e-03, grad_scale: 32.0 2023-04-27 21:09:18,247 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.7311, 2.3194, 2.7493, 3.1968, 2.6049, 2.1862, 2.2326, 2.5747], device='cuda:1'), covar=tensor([0.3080, 0.2645, 0.1336, 0.2127, 0.2630, 0.2425, 0.3252, 0.1677], device='cuda:1'), in_proj_covar=tensor([0.0293, 0.0244, 0.0228, 0.0313, 0.0221, 0.0234, 0.0227, 0.0186], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-27 21:09:45,098 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.039e+02 1.527e+02 1.865e+02 2.204e+02 5.324e+02, threshold=3.729e+02, percent-clipped=2.0 2023-04-27 21:09:58,843 INFO [finetune.py:976] (1/7) Epoch 24, batch 4750, loss[loss=0.1492, simple_loss=0.2189, pruned_loss=0.03978, over 4907.00 frames. ], tot_loss[loss=0.1628, simple_loss=0.2351, pruned_loss=0.04532, over 954396.17 frames. ], batch size: 35, lr: 3.04e-03, grad_scale: 32.0 2023-04-27 21:10:07,140 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=136498.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:10:23,725 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([3.3218, 3.2536, 2.4548, 3.8240, 3.3391, 3.2756, 1.4703, 3.2270], device='cuda:1'), covar=tensor([0.1858, 0.1543, 0.3555, 0.2477, 0.3673, 0.2102, 0.5873, 0.2936], device='cuda:1'), in_proj_covar=tensor([0.0246, 0.0222, 0.0254, 0.0310, 0.0299, 0.0251, 0.0277, 0.0274], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 21:10:23,844 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.65 vs. limit=2.0 2023-04-27 21:10:24,982 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.2834, 2.1869, 1.8152, 1.9115, 2.4442, 1.9558, 2.7874, 1.6050], device='cuda:1'), covar=tensor([0.3640, 0.1941, 0.4594, 0.2929, 0.1423, 0.2310, 0.1475, 0.4192], device='cuda:1'), in_proj_covar=tensor([0.0337, 0.0351, 0.0423, 0.0349, 0.0377, 0.0374, 0.0367, 0.0420], device='cuda:1'), out_proj_covar=tensor([9.9784e-05, 1.0481e-04, 1.2827e-04, 1.0485e-04, 1.1177e-04, 1.1128e-04, 1.0770e-04, 1.2635e-04], device='cuda:1') 2023-04-27 21:10:47,890 INFO [finetune.py:976] (1/7) Epoch 24, batch 4800, loss[loss=0.2025, simple_loss=0.2817, pruned_loss=0.06165, over 4815.00 frames. ], tot_loss[loss=0.1663, simple_loss=0.239, pruned_loss=0.04677, over 956034.62 frames. ], batch size: 39, lr: 3.04e-03, grad_scale: 32.0 2023-04-27 21:11:29,602 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.193e+02 1.561e+02 1.788e+02 2.083e+02 3.546e+02, threshold=3.576e+02, percent-clipped=0.0 2023-04-27 21:11:44,096 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7831, 2.0311, 1.0721, 1.5122, 2.1483, 1.6033, 1.5915, 1.7471], device='cuda:1'), covar=tensor([0.0483, 0.0354, 0.0306, 0.0569, 0.0256, 0.0509, 0.0527, 0.0546], device='cuda:1'), in_proj_covar=tensor([0.0028, 0.0023, 0.0022, 0.0029, 0.0019, 0.0028, 0.0027, 0.0028], device='cuda:1'), out_proj_covar=tensor([0.0051, 0.0046, 0.0038, 0.0053, 0.0039, 0.0050, 0.0050, 0.0052], device='cuda:1') 2023-04-27 21:11:53,506 INFO [finetune.py:976] (1/7) Epoch 24, batch 4850, loss[loss=0.2034, simple_loss=0.2763, pruned_loss=0.06524, over 4836.00 frames. ], tot_loss[loss=0.1679, simple_loss=0.2415, pruned_loss=0.0471, over 954332.42 frames. ], batch size: 47, lr: 3.04e-03, grad_scale: 64.0 2023-04-27 21:12:08,958 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-04-27 21:12:26,524 INFO [finetune.py:976] (1/7) Epoch 24, batch 4900, loss[loss=0.1678, simple_loss=0.2515, pruned_loss=0.04203, over 4915.00 frames. ], tot_loss[loss=0.1695, simple_loss=0.2433, pruned_loss=0.04785, over 954399.82 frames. ], batch size: 37, lr: 3.03e-03, grad_scale: 32.0 2023-04-27 21:12:27,224 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=136638.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:12:30,287 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=136643.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:12:35,616 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=136650.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:12:46,886 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.047e+02 1.604e+02 1.891e+02 2.229e+02 4.573e+02, threshold=3.781e+02, percent-clipped=1.0 2023-04-27 21:12:54,895 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.5502, 2.0898, 2.5142, 2.9985, 2.4439, 1.9649, 2.0034, 2.4660], device='cuda:1'), covar=tensor([0.3343, 0.3228, 0.1755, 0.2537, 0.2712, 0.2628, 0.3763, 0.2094], device='cuda:1'), in_proj_covar=tensor([0.0292, 0.0244, 0.0227, 0.0312, 0.0220, 0.0234, 0.0226, 0.0185], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-27 21:12:58,275 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=136686.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:12:58,835 INFO [finetune.py:976] (1/7) Epoch 24, batch 4950, loss[loss=0.2119, simple_loss=0.2722, pruned_loss=0.07585, over 4877.00 frames. ], tot_loss[loss=0.1701, simple_loss=0.2439, pruned_loss=0.0481, over 955237.64 frames. ], batch size: 35, lr: 3.03e-03, grad_scale: 32.0 2023-04-27 21:13:02,154 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=136691.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:13:07,457 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=136698.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:13:11,649 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=136704.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:13:32,451 INFO [finetune.py:976] (1/7) Epoch 24, batch 5000, loss[loss=0.1494, simple_loss=0.2194, pruned_loss=0.03963, over 4776.00 frames. ], tot_loss[loss=0.1686, simple_loss=0.2424, pruned_loss=0.04739, over 956331.39 frames. ], batch size: 25, lr: 3.03e-03, grad_scale: 32.0 2023-04-27 21:13:47,073 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4460, 1.2458, 4.1448, 3.8950, 3.6773, 3.9252, 3.9433, 3.5546], device='cuda:1'), covar=tensor([0.7073, 0.6059, 0.1022, 0.1637, 0.1081, 0.1606, 0.1347, 0.1804], device='cuda:1'), in_proj_covar=tensor([0.0312, 0.0307, 0.0406, 0.0409, 0.0350, 0.0410, 0.0320, 0.0369], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-27 21:13:53,647 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.026e+02 1.571e+02 1.815e+02 2.179e+02 3.304e+02, threshold=3.630e+02, percent-clipped=0.0 2023-04-27 21:14:05,878 INFO [finetune.py:976] (1/7) Epoch 24, batch 5050, loss[loss=0.1604, simple_loss=0.2287, pruned_loss=0.04601, over 4912.00 frames. ], tot_loss[loss=0.1674, simple_loss=0.24, pruned_loss=0.04746, over 955205.07 frames. ], batch size: 36, lr: 3.03e-03, grad_scale: 32.0 2023-04-27 21:14:18,815 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=136798.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:14:29,844 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=136806.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:15:01,715 INFO [finetune.py:976] (1/7) Epoch 24, batch 5100, loss[loss=0.1704, simple_loss=0.2442, pruned_loss=0.04826, over 4933.00 frames. ], tot_loss[loss=0.1656, simple_loss=0.2376, pruned_loss=0.04676, over 953743.62 frames. ], batch size: 38, lr: 3.03e-03, grad_scale: 32.0 2023-04-27 21:15:07,733 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=136846.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:15:22,613 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=136867.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:15:23,110 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.056e+02 1.537e+02 1.873e+02 2.196e+02 3.846e+02, threshold=3.746e+02, percent-clipped=1.0 2023-04-27 21:15:33,518 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.1285, 1.8693, 2.2936, 2.5864, 2.1195, 2.0147, 2.1480, 2.1040], device='cuda:1'), covar=tensor([0.5328, 0.8206, 0.7864, 0.6059, 0.6848, 0.9295, 0.9429, 1.1150], device='cuda:1'), in_proj_covar=tensor([0.0436, 0.0420, 0.0511, 0.0507, 0.0465, 0.0498, 0.0501, 0.0515], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 21:15:35,186 INFO [finetune.py:976] (1/7) Epoch 24, batch 5150, loss[loss=0.1678, simple_loss=0.236, pruned_loss=0.04975, over 4901.00 frames. ], tot_loss[loss=0.1659, simple_loss=0.2373, pruned_loss=0.04725, over 955090.58 frames. ], batch size: 32, lr: 3.03e-03, grad_scale: 32.0 2023-04-27 21:16:16,512 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2023-04-27 21:16:26,994 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6597, 1.4571, 5.0476, 4.7448, 4.4635, 4.8392, 4.5523, 4.4459], device='cuda:1'), covar=tensor([0.7600, 0.6011, 0.1156, 0.1897, 0.1082, 0.1556, 0.1450, 0.1894], device='cuda:1'), in_proj_covar=tensor([0.0313, 0.0307, 0.0407, 0.0410, 0.0351, 0.0411, 0.0321, 0.0370], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-27 21:16:29,340 INFO [finetune.py:976] (1/7) Epoch 24, batch 5200, loss[loss=0.181, simple_loss=0.2644, pruned_loss=0.04881, over 4800.00 frames. ], tot_loss[loss=0.1684, simple_loss=0.2405, pruned_loss=0.04812, over 951278.56 frames. ], batch size: 51, lr: 3.03e-03, grad_scale: 32.0 2023-04-27 21:16:39,429 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5023, 1.5690, 0.8304, 1.2377, 1.5372, 1.3430, 1.2468, 1.3478], device='cuda:1'), covar=tensor([0.0488, 0.0381, 0.0367, 0.0556, 0.0311, 0.0530, 0.0519, 0.0580], device='cuda:1'), in_proj_covar=tensor([0.0028, 0.0023, 0.0022, 0.0028, 0.0019, 0.0028, 0.0027, 0.0028], device='cuda:1'), out_proj_covar=tensor([0.0051, 0.0046, 0.0039, 0.0052, 0.0039, 0.0050, 0.0050, 0.0052], device='cuda:1') 2023-04-27 21:16:59,706 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.230e+02 1.748e+02 2.026e+02 2.510e+02 4.483e+02, threshold=4.051e+02, percent-clipped=3.0 2023-04-27 21:17:22,716 INFO [finetune.py:976] (1/7) Epoch 24, batch 5250, loss[loss=0.1669, simple_loss=0.2513, pruned_loss=0.04119, over 4801.00 frames. ], tot_loss[loss=0.172, simple_loss=0.2445, pruned_loss=0.04971, over 952995.43 frames. ], batch size: 45, lr: 3.03e-03, grad_scale: 32.0 2023-04-27 21:17:25,240 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=136991.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:17:36,061 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=136999.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:18:10,122 INFO [finetune.py:976] (1/7) Epoch 24, batch 5300, loss[loss=0.1881, simple_loss=0.2703, pruned_loss=0.05295, over 4852.00 frames. ], tot_loss[loss=0.1721, simple_loss=0.2452, pruned_loss=0.04952, over 952027.42 frames. ], batch size: 44, lr: 3.03e-03, grad_scale: 32.0 2023-04-27 21:18:11,867 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=137039.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:18:25,208 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.0881, 0.7532, 0.8959, 0.8091, 1.1802, 0.9906, 0.8840, 0.9410], device='cuda:1'), covar=tensor([0.1789, 0.1417, 0.1937, 0.1509, 0.0901, 0.1300, 0.1553, 0.2135], device='cuda:1'), in_proj_covar=tensor([0.0313, 0.0307, 0.0350, 0.0283, 0.0326, 0.0303, 0.0298, 0.0371], device='cuda:1'), out_proj_covar=tensor([6.4093e-05, 6.3210e-05, 7.3598e-05, 5.6769e-05, 6.6994e-05, 6.3496e-05, 6.1889e-05, 7.8728e-05], device='cuda:1') 2023-04-27 21:18:30,932 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.102e+02 1.584e+02 1.901e+02 2.207e+02 4.060e+02, threshold=3.801e+02, percent-clipped=1.0 2023-04-27 21:18:44,060 INFO [finetune.py:976] (1/7) Epoch 24, batch 5350, loss[loss=0.1516, simple_loss=0.2289, pruned_loss=0.03719, over 4904.00 frames. ], tot_loss[loss=0.1716, simple_loss=0.2451, pruned_loss=0.04907, over 953381.05 frames. ], batch size: 36, lr: 3.03e-03, grad_scale: 32.0 2023-04-27 21:19:16,914 INFO [finetune.py:976] (1/7) Epoch 24, batch 5400, loss[loss=0.2021, simple_loss=0.2698, pruned_loss=0.06723, over 4792.00 frames. ], tot_loss[loss=0.17, simple_loss=0.2426, pruned_loss=0.04866, over 952011.41 frames. ], batch size: 45, lr: 3.03e-03, grad_scale: 32.0 2023-04-27 21:19:20,189 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2023-04-27 21:19:26,631 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6985, 1.2340, 1.3720, 1.4082, 1.8266, 1.4912, 1.2278, 1.2975], device='cuda:1'), covar=tensor([0.1521, 0.1444, 0.1702, 0.1301, 0.0778, 0.1535, 0.2060, 0.2243], device='cuda:1'), in_proj_covar=tensor([0.0313, 0.0308, 0.0351, 0.0284, 0.0326, 0.0305, 0.0299, 0.0373], device='cuda:1'), out_proj_covar=tensor([6.4151e-05, 6.3466e-05, 7.4039e-05, 5.6923e-05, 6.7069e-05, 6.3758e-05, 6.2193e-05, 7.9080e-05], device='cuda:1') 2023-04-27 21:19:33,101 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=137162.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:19:37,581 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.035e+02 1.430e+02 1.764e+02 2.146e+02 4.768e+02, threshold=3.527e+02, percent-clipped=1.0 2023-04-27 21:19:50,715 INFO [finetune.py:976] (1/7) Epoch 24, batch 5450, loss[loss=0.1609, simple_loss=0.2239, pruned_loss=0.04894, over 4827.00 frames. ], tot_loss[loss=0.1679, simple_loss=0.24, pruned_loss=0.04795, over 952006.11 frames. ], batch size: 51, lr: 3.03e-03, grad_scale: 32.0 2023-04-27 21:20:00,001 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.1020, 2.8143, 2.2624, 2.2996, 1.6044, 1.5914, 2.3078, 1.5546], device='cuda:1'), covar=tensor([0.1289, 0.1248, 0.1065, 0.1415, 0.1783, 0.1567, 0.0739, 0.1643], device='cuda:1'), in_proj_covar=tensor([0.0196, 0.0210, 0.0167, 0.0203, 0.0198, 0.0184, 0.0155, 0.0185], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-27 21:20:05,494 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.5512, 3.2112, 2.8269, 3.0635, 2.3890, 2.6626, 2.7961, 2.0539], device='cuda:1'), covar=tensor([0.1869, 0.1173, 0.0803, 0.1084, 0.2710, 0.1207, 0.1883, 0.2911], device='cuda:1'), in_proj_covar=tensor([0.0282, 0.0297, 0.0214, 0.0274, 0.0310, 0.0254, 0.0248, 0.0263], device='cuda:1'), out_proj_covar=tensor([1.1265e-04, 1.1735e-04, 8.4302e-05, 1.0826e-04, 1.2520e-04, 9.9931e-05, 1.0021e-04, 1.0387e-04], device='cuda:1') 2023-04-27 21:20:24,727 INFO [finetune.py:976] (1/7) Epoch 24, batch 5500, loss[loss=0.1726, simple_loss=0.2433, pruned_loss=0.05102, over 4750.00 frames. ], tot_loss[loss=0.1652, simple_loss=0.2364, pruned_loss=0.04703, over 953334.87 frames. ], batch size: 27, lr: 3.03e-03, grad_scale: 32.0 2023-04-27 21:20:28,073 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.79 vs. limit=2.0 2023-04-27 21:20:32,762 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.4114, 2.9670, 1.0164, 1.7440, 2.2491, 1.3106, 3.9794, 1.8410], device='cuda:1'), covar=tensor([0.0635, 0.0747, 0.0916, 0.1256, 0.0531, 0.1006, 0.0214, 0.0639], device='cuda:1'), in_proj_covar=tensor([0.0050, 0.0063, 0.0047, 0.0046, 0.0048, 0.0051, 0.0072, 0.0050], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0009, 0.0007, 0.0007, 0.0007, 0.0008, 0.0010, 0.0007], device='cuda:1') 2023-04-27 21:20:44,122 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 8.686e+01 1.504e+02 1.776e+02 2.176e+02 3.969e+02, threshold=3.551e+02, percent-clipped=4.0 2023-04-27 21:20:51,858 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.2543, 1.6831, 2.0745, 2.2864, 2.0572, 1.6947, 1.1360, 1.7394], device='cuda:1'), covar=tensor([0.3434, 0.3265, 0.1810, 0.2235, 0.2568, 0.2812, 0.4246, 0.2083], device='cuda:1'), in_proj_covar=tensor([0.0293, 0.0245, 0.0227, 0.0312, 0.0221, 0.0234, 0.0227, 0.0185], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-27 21:20:57,634 INFO [finetune.py:976] (1/7) Epoch 24, batch 5550, loss[loss=0.1483, simple_loss=0.2226, pruned_loss=0.03701, over 4779.00 frames. ], tot_loss[loss=0.1665, simple_loss=0.2381, pruned_loss=0.04748, over 952975.11 frames. ], batch size: 29, lr: 3.03e-03, grad_scale: 32.0 2023-04-27 21:21:02,607 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4526, 1.1967, 0.4877, 1.1271, 1.0901, 1.3023, 1.2118, 1.1993], device='cuda:1'), covar=tensor([0.0526, 0.0392, 0.0394, 0.0588, 0.0306, 0.0535, 0.0533, 0.0593], device='cuda:1'), in_proj_covar=tensor([0.0028, 0.0023, 0.0022, 0.0028, 0.0019, 0.0028, 0.0027, 0.0028], device='cuda:1'), out_proj_covar=tensor([0.0051, 0.0046, 0.0038, 0.0052, 0.0039, 0.0050, 0.0050, 0.0052], device='cuda:1') 2023-04-27 21:21:05,049 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=137299.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:21:23,436 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7125, 1.2227, 1.3730, 1.3757, 1.8062, 1.4624, 1.2002, 1.2769], device='cuda:1'), covar=tensor([0.1632, 0.1484, 0.2006, 0.1476, 0.0895, 0.1655, 0.2058, 0.2454], device='cuda:1'), in_proj_covar=tensor([0.0315, 0.0309, 0.0353, 0.0285, 0.0328, 0.0306, 0.0300, 0.0374], device='cuda:1'), out_proj_covar=tensor([6.4426e-05, 6.3691e-05, 7.4296e-05, 5.7206e-05, 6.7507e-05, 6.4012e-05, 6.2475e-05, 7.9334e-05], device='cuda:1') 2023-04-27 21:21:34,762 INFO [finetune.py:976] (1/7) Epoch 24, batch 5600, loss[loss=0.1684, simple_loss=0.2498, pruned_loss=0.04347, over 4839.00 frames. ], tot_loss[loss=0.1685, simple_loss=0.241, pruned_loss=0.04801, over 953716.10 frames. ], batch size: 47, lr: 3.03e-03, grad_scale: 32.0 2023-04-27 21:21:46,034 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=137347.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:21:57,103 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.64 vs. limit=2.0 2023-04-27 21:22:14,767 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.035e+02 1.541e+02 1.789e+02 2.150e+02 5.679e+02, threshold=3.578e+02, percent-clipped=2.0 2023-04-27 21:22:37,496 INFO [finetune.py:976] (1/7) Epoch 24, batch 5650, loss[loss=0.2259, simple_loss=0.2925, pruned_loss=0.07964, over 4857.00 frames. ], tot_loss[loss=0.17, simple_loss=0.2433, pruned_loss=0.04832, over 952860.40 frames. ], batch size: 49, lr: 3.03e-03, grad_scale: 32.0 2023-04-27 21:22:47,382 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-04-27 21:23:28,791 INFO [finetune.py:976] (1/7) Epoch 24, batch 5700, loss[loss=0.1681, simple_loss=0.225, pruned_loss=0.05557, over 4090.00 frames. ], tot_loss[loss=0.1678, simple_loss=0.2402, pruned_loss=0.04774, over 938737.08 frames. ], batch size: 17, lr: 3.03e-03, grad_scale: 32.0 2023-04-27 21:23:38,998 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8859, 2.1166, 1.8254, 1.6160, 1.4603, 1.4726, 1.8859, 1.4139], device='cuda:1'), covar=tensor([0.1632, 0.1306, 0.1408, 0.1514, 0.2051, 0.1787, 0.0896, 0.1863], device='cuda:1'), in_proj_covar=tensor([0.0196, 0.0209, 0.0167, 0.0201, 0.0197, 0.0183, 0.0155, 0.0184], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-27 21:23:43,837 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=137462.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:23:57,746 INFO [finetune.py:976] (1/7) Epoch 25, batch 0, loss[loss=0.1951, simple_loss=0.2747, pruned_loss=0.05779, over 4817.00 frames. ], tot_loss[loss=0.1951, simple_loss=0.2747, pruned_loss=0.05779, over 4817.00 frames. ], batch size: 38, lr: 3.03e-03, grad_scale: 32.0 2023-04-27 21:23:57,746 INFO [finetune.py:1001] (1/7) Computing validation loss 2023-04-27 21:24:03,503 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([3.4053, 3.4712, 2.5385, 3.8915, 3.5734, 3.4441, 1.5429, 3.4470], device='cuda:1'), covar=tensor([0.1874, 0.1562, 0.3071, 0.2142, 0.3485, 0.2031, 0.5698, 0.2436], device='cuda:1'), in_proj_covar=tensor([0.0245, 0.0220, 0.0253, 0.0308, 0.0299, 0.0248, 0.0277, 0.0273], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 21:24:08,111 INFO [finetune.py:1010] (1/7) Epoch 25, validation: loss=0.155, simple_loss=0.224, pruned_loss=0.04295, over 2265189.00 frames. 2023-04-27 21:24:08,111 INFO [finetune.py:1011] (1/7) Maximum memory allocated so far is 6391MB 2023-04-27 21:24:09,923 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.060e+01 1.467e+02 1.816e+02 2.337e+02 4.208e+02, threshold=3.632e+02, percent-clipped=2.0 2023-04-27 21:24:37,393 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=137510.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:24:40,446 INFO [finetune.py:976] (1/7) Epoch 25, batch 50, loss[loss=0.1776, simple_loss=0.2551, pruned_loss=0.05008, over 4898.00 frames. ], tot_loss[loss=0.1768, simple_loss=0.2495, pruned_loss=0.05202, over 216004.29 frames. ], batch size: 37, lr: 3.02e-03, grad_scale: 32.0 2023-04-27 21:24:41,792 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7086, 1.7795, 0.9302, 1.4375, 1.8206, 1.5407, 1.4529, 1.5651], device='cuda:1'), covar=tensor([0.0483, 0.0343, 0.0322, 0.0546, 0.0253, 0.0494, 0.0485, 0.0525], device='cuda:1'), in_proj_covar=tensor([0.0028, 0.0023, 0.0022, 0.0028, 0.0019, 0.0028, 0.0027, 0.0028], device='cuda:1'), out_proj_covar=tensor([0.0051, 0.0046, 0.0038, 0.0052, 0.0038, 0.0050, 0.0050, 0.0052], device='cuda:1') 2023-04-27 21:24:44,162 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.44 vs. limit=5.0 2023-04-27 21:24:55,460 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0371, 2.4759, 1.0837, 1.4295, 2.0768, 1.1573, 3.1555, 1.5791], device='cuda:1'), covar=tensor([0.0637, 0.0588, 0.0781, 0.1147, 0.0430, 0.0994, 0.0249, 0.0612], device='cuda:1'), in_proj_covar=tensor([0.0050, 0.0063, 0.0047, 0.0046, 0.0048, 0.0051, 0.0072, 0.0050], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0009, 0.0007, 0.0007, 0.0007, 0.0008, 0.0010, 0.0007], device='cuda:1') 2023-04-27 21:24:57,772 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.0202, 2.3578, 0.7441, 1.3185, 1.3656, 1.7837, 1.5688, 0.8727], device='cuda:1'), covar=tensor([0.1826, 0.1587, 0.2258, 0.1826, 0.1399, 0.1252, 0.1750, 0.1953], device='cuda:1'), in_proj_covar=tensor([0.0116, 0.0236, 0.0135, 0.0119, 0.0131, 0.0151, 0.0117, 0.0117], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-27 21:24:59,620 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5409, 1.0874, 1.3151, 1.2546, 1.6230, 1.3512, 1.1211, 1.2441], device='cuda:1'), covar=tensor([0.1533, 0.1363, 0.1739, 0.1339, 0.0944, 0.1337, 0.1705, 0.2415], device='cuda:1'), in_proj_covar=tensor([0.0310, 0.0304, 0.0348, 0.0282, 0.0324, 0.0302, 0.0296, 0.0369], device='cuda:1'), out_proj_covar=tensor([6.3359e-05, 6.2663e-05, 7.3322e-05, 5.6520e-05, 6.6681e-05, 6.3163e-05, 6.1477e-05, 7.8147e-05], device='cuda:1') 2023-04-27 21:25:13,440 INFO [finetune.py:976] (1/7) Epoch 25, batch 100, loss[loss=0.1377, simple_loss=0.2179, pruned_loss=0.02874, over 4906.00 frames. ], tot_loss[loss=0.1711, simple_loss=0.2428, pruned_loss=0.04971, over 381837.69 frames. ], batch size: 32, lr: 3.02e-03, grad_scale: 32.0 2023-04-27 21:25:15,237 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.734e+01 1.567e+02 1.856e+02 2.214e+02 3.687e+02, threshold=3.711e+02, percent-clipped=3.0 2023-04-27 21:25:46,408 INFO [finetune.py:976] (1/7) Epoch 25, batch 150, loss[loss=0.189, simple_loss=0.2449, pruned_loss=0.06659, over 4895.00 frames. ], tot_loss[loss=0.1673, simple_loss=0.2379, pruned_loss=0.04837, over 510796.99 frames. ], batch size: 32, lr: 3.02e-03, grad_scale: 32.0 2023-04-27 21:25:55,291 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-04-27 21:26:20,277 INFO [finetune.py:976] (1/7) Epoch 25, batch 200, loss[loss=0.1855, simple_loss=0.263, pruned_loss=0.054, over 4919.00 frames. ], tot_loss[loss=0.1653, simple_loss=0.2353, pruned_loss=0.04762, over 610378.61 frames. ], batch size: 37, lr: 3.02e-03, grad_scale: 32.0 2023-04-27 21:26:22,073 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.158e+02 1.527e+02 1.776e+02 2.195e+02 3.612e+02, threshold=3.551e+02, percent-clipped=0.0 2023-04-27 21:26:33,030 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.12 vs. limit=2.0 2023-04-27 21:27:09,388 INFO [finetune.py:976] (1/7) Epoch 25, batch 250, loss[loss=0.2028, simple_loss=0.2821, pruned_loss=0.06174, over 4805.00 frames. ], tot_loss[loss=0.1668, simple_loss=0.238, pruned_loss=0.04783, over 686982.23 frames. ], batch size: 41, lr: 3.02e-03, grad_scale: 32.0 2023-04-27 21:27:18,679 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.4701, 1.8116, 1.8666, 1.9644, 1.8170, 1.8914, 1.9145, 1.8652], device='cuda:1'), covar=tensor([0.3732, 0.5252, 0.4386, 0.4367, 0.5292, 0.6590, 0.5308, 0.4867], device='cuda:1'), in_proj_covar=tensor([0.0339, 0.0372, 0.0324, 0.0338, 0.0348, 0.0392, 0.0356, 0.0329], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 21:28:04,796 INFO [finetune.py:976] (1/7) Epoch 25, batch 300, loss[loss=0.1521, simple_loss=0.2301, pruned_loss=0.03709, over 4791.00 frames. ], tot_loss[loss=0.1678, simple_loss=0.2405, pruned_loss=0.04754, over 746952.44 frames. ], batch size: 29, lr: 3.02e-03, grad_scale: 32.0 2023-04-27 21:28:05,580 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6429, 1.0370, 1.6720, 2.1698, 1.7452, 1.5811, 1.6563, 1.5773], device='cuda:1'), covar=tensor([0.3693, 0.5696, 0.4848, 0.4695, 0.4566, 0.6139, 0.6082, 0.7803], device='cuda:1'), in_proj_covar=tensor([0.0436, 0.0420, 0.0512, 0.0508, 0.0466, 0.0500, 0.0503, 0.0516], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 21:28:06,624 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.160e+02 1.558e+02 1.872e+02 2.368e+02 4.247e+02, threshold=3.743e+02, percent-clipped=2.0 2023-04-27 21:28:43,570 INFO [finetune.py:976] (1/7) Epoch 25, batch 350, loss[loss=0.173, simple_loss=0.2582, pruned_loss=0.04393, over 4818.00 frames. ], tot_loss[loss=0.1697, simple_loss=0.243, pruned_loss=0.04816, over 794157.21 frames. ], batch size: 39, lr: 3.02e-03, grad_scale: 32.0 2023-04-27 21:29:09,315 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.97 vs. limit=2.0 2023-04-27 21:29:21,925 INFO [finetune.py:976] (1/7) Epoch 25, batch 400, loss[loss=0.1606, simple_loss=0.2396, pruned_loss=0.04082, over 4827.00 frames. ], tot_loss[loss=0.1717, simple_loss=0.2454, pruned_loss=0.04897, over 829686.32 frames. ], batch size: 30, lr: 3.02e-03, grad_scale: 32.0 2023-04-27 21:29:29,042 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.177e+02 1.617e+02 1.875e+02 2.197e+02 4.707e+02, threshold=3.750e+02, percent-clipped=1.0 2023-04-27 21:30:01,120 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=137892.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:30:02,926 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=137895.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:30:27,175 INFO [finetune.py:976] (1/7) Epoch 25, batch 450, loss[loss=0.1248, simple_loss=0.1895, pruned_loss=0.03005, over 4240.00 frames. ], tot_loss[loss=0.1698, simple_loss=0.2436, pruned_loss=0.04802, over 859504.34 frames. ], batch size: 18, lr: 3.02e-03, grad_scale: 32.0 2023-04-27 21:30:28,487 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.5311, 2.7382, 1.4725, 1.8992, 2.4906, 1.5836, 3.8111, 2.0830], device='cuda:1'), covar=tensor([0.0543, 0.0583, 0.0618, 0.1116, 0.0386, 0.0909, 0.0248, 0.0548], device='cuda:1'), in_proj_covar=tensor([0.0050, 0.0064, 0.0047, 0.0046, 0.0048, 0.0051, 0.0072, 0.0050], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0007, 0.0007, 0.0008, 0.0010, 0.0008], device='cuda:1') 2023-04-27 21:31:27,309 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=137953.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:31:33,956 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=137956.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:31:44,823 INFO [finetune.py:976] (1/7) Epoch 25, batch 500, loss[loss=0.1691, simple_loss=0.2394, pruned_loss=0.04941, over 4816.00 frames. ], tot_loss[loss=0.1685, simple_loss=0.2414, pruned_loss=0.04776, over 879418.70 frames. ], batch size: 41, lr: 3.02e-03, grad_scale: 32.0 2023-04-27 21:31:46,603 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 8.991e+01 1.587e+02 1.886e+02 2.282e+02 3.867e+02, threshold=3.771e+02, percent-clipped=1.0 2023-04-27 21:32:25,646 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-04-27 21:32:27,220 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8373, 1.6411, 2.0918, 2.2774, 1.7290, 1.5258, 1.7510, 1.2108], device='cuda:1'), covar=tensor([0.0459, 0.0748, 0.0401, 0.0461, 0.0702, 0.1056, 0.0601, 0.0619], device='cuda:1'), in_proj_covar=tensor([0.0070, 0.0068, 0.0066, 0.0068, 0.0075, 0.0096, 0.0073, 0.0064], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-27 21:32:30,094 INFO [finetune.py:976] (1/7) Epoch 25, batch 550, loss[loss=0.1564, simple_loss=0.2277, pruned_loss=0.04252, over 4866.00 frames. ], tot_loss[loss=0.1657, simple_loss=0.238, pruned_loss=0.04673, over 896933.21 frames. ], batch size: 31, lr: 3.02e-03, grad_scale: 32.0 2023-04-27 21:33:20,385 INFO [finetune.py:976] (1/7) Epoch 25, batch 600, loss[loss=0.1748, simple_loss=0.2448, pruned_loss=0.05237, over 4917.00 frames. ], tot_loss[loss=0.1662, simple_loss=0.2382, pruned_loss=0.0471, over 910900.23 frames. ], batch size: 36, lr: 3.02e-03, grad_scale: 32.0 2023-04-27 21:33:22,212 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.059e+02 1.469e+02 1.786e+02 2.001e+02 2.950e+02, threshold=3.571e+02, percent-clipped=0.0 2023-04-27 21:33:53,151 INFO [finetune.py:976] (1/7) Epoch 25, batch 650, loss[loss=0.1794, simple_loss=0.2529, pruned_loss=0.05293, over 4753.00 frames. ], tot_loss[loss=0.1683, simple_loss=0.2408, pruned_loss=0.04786, over 921330.42 frames. ], batch size: 27, lr: 3.02e-03, grad_scale: 32.0 2023-04-27 21:34:10,466 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.1525, 0.8306, 0.9020, 0.8005, 1.2396, 1.0002, 0.9627, 0.9430], device='cuda:1'), covar=tensor([0.1877, 0.1658, 0.2293, 0.1725, 0.1151, 0.1531, 0.1779, 0.2827], device='cuda:1'), in_proj_covar=tensor([0.0311, 0.0306, 0.0349, 0.0283, 0.0324, 0.0303, 0.0297, 0.0370], device='cuda:1'), out_proj_covar=tensor([6.3625e-05, 6.3003e-05, 7.3522e-05, 5.6712e-05, 6.6559e-05, 6.3385e-05, 6.1639e-05, 7.8331e-05], device='cuda:1') 2023-04-27 21:34:26,502 INFO [finetune.py:976] (1/7) Epoch 25, batch 700, loss[loss=0.1383, simple_loss=0.1972, pruned_loss=0.03972, over 4702.00 frames. ], tot_loss[loss=0.1707, simple_loss=0.2438, pruned_loss=0.04882, over 928306.74 frames. ], batch size: 23, lr: 3.02e-03, grad_scale: 32.0 2023-04-27 21:34:28,315 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.080e+02 1.605e+02 1.838e+02 2.214e+02 4.494e+02, threshold=3.677e+02, percent-clipped=3.0 2023-04-27 21:35:14,707 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.79 vs. limit=2.0 2023-04-27 21:35:26,623 INFO [finetune.py:976] (1/7) Epoch 25, batch 750, loss[loss=0.1487, simple_loss=0.229, pruned_loss=0.03417, over 4759.00 frames. ], tot_loss[loss=0.1716, simple_loss=0.2448, pruned_loss=0.04922, over 935854.30 frames. ], batch size: 59, lr: 3.02e-03, grad_scale: 32.0 2023-04-27 21:36:09,233 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=138248.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:36:16,834 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=138251.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:36:25,762 INFO [finetune.py:976] (1/7) Epoch 25, batch 800, loss[loss=0.1659, simple_loss=0.2482, pruned_loss=0.04182, over 4906.00 frames. ], tot_loss[loss=0.1713, simple_loss=0.2449, pruned_loss=0.04888, over 937894.63 frames. ], batch size: 43, lr: 3.02e-03, grad_scale: 32.0 2023-04-27 21:36:27,577 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.037e+01 1.577e+02 1.890e+02 2.276e+02 6.092e+02, threshold=3.780e+02, percent-clipped=1.0 2023-04-27 21:36:28,893 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=138270.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:36:41,912 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.55 vs. limit=2.0 2023-04-27 21:37:09,964 INFO [finetune.py:976] (1/7) Epoch 25, batch 850, loss[loss=0.1391, simple_loss=0.2107, pruned_loss=0.03381, over 4693.00 frames. ], tot_loss[loss=0.1691, simple_loss=0.2427, pruned_loss=0.04775, over 942310.65 frames. ], batch size: 23, lr: 3.02e-03, grad_scale: 32.0 2023-04-27 21:37:14,317 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=138322.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:37:15,577 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.3283, 1.7563, 1.7964, 1.8909, 1.8292, 1.9315, 1.8319, 1.8366], device='cuda:1'), covar=tensor([0.4036, 0.4308, 0.4198, 0.4088, 0.4998, 0.6255, 0.4704, 0.4638], device='cuda:1'), in_proj_covar=tensor([0.0342, 0.0375, 0.0327, 0.0341, 0.0350, 0.0394, 0.0358, 0.0332], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 21:37:24,961 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=138331.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:38:00,618 INFO [finetune.py:976] (1/7) Epoch 25, batch 900, loss[loss=0.1265, simple_loss=0.1914, pruned_loss=0.03079, over 4104.00 frames. ], tot_loss[loss=0.1668, simple_loss=0.2398, pruned_loss=0.04691, over 944805.69 frames. ], batch size: 18, lr: 3.02e-03, grad_scale: 32.0 2023-04-27 21:38:02,478 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.023e+01 1.492e+02 1.759e+02 2.095e+02 3.711e+02, threshold=3.518e+02, percent-clipped=0.0 2023-04-27 21:38:20,779 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7561, 1.7820, 0.9317, 1.4981, 1.8930, 1.5979, 1.5248, 1.6181], device='cuda:1'), covar=tensor([0.0475, 0.0372, 0.0325, 0.0534, 0.0252, 0.0521, 0.0517, 0.0550], device='cuda:1'), in_proj_covar=tensor([0.0028, 0.0023, 0.0022, 0.0028, 0.0019, 0.0028, 0.0027, 0.0028], device='cuda:1'), out_proj_covar=tensor([0.0052, 0.0046, 0.0039, 0.0053, 0.0039, 0.0050, 0.0050, 0.0052], device='cuda:1') 2023-04-27 21:38:22,002 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=138383.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:38:54,305 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7638, 2.2339, 2.0495, 1.8344, 1.2900, 1.3724, 2.1446, 1.3460], device='cuda:1'), covar=tensor([0.1721, 0.1516, 0.1273, 0.1681, 0.2280, 0.1940, 0.0838, 0.1992], device='cuda:1'), in_proj_covar=tensor([0.0197, 0.0209, 0.0168, 0.0203, 0.0199, 0.0185, 0.0156, 0.0185], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-27 21:39:01,961 INFO [finetune.py:976] (1/7) Epoch 25, batch 950, loss[loss=0.09991, simple_loss=0.1658, pruned_loss=0.01701, over 4238.00 frames. ], tot_loss[loss=0.1657, simple_loss=0.2379, pruned_loss=0.04676, over 946649.50 frames. ], batch size: 18, lr: 3.02e-03, grad_scale: 32.0 2023-04-27 21:39:13,734 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8185, 1.8139, 2.1873, 2.2249, 1.6859, 1.4281, 1.7573, 1.0009], device='cuda:1'), covar=tensor([0.0812, 0.0667, 0.0513, 0.0882, 0.0799, 0.1165, 0.0755, 0.0784], device='cuda:1'), in_proj_covar=tensor([0.0070, 0.0068, 0.0066, 0.0068, 0.0075, 0.0096, 0.0073, 0.0064], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-27 21:40:06,146 INFO [finetune.py:976] (1/7) Epoch 25, batch 1000, loss[loss=0.1518, simple_loss=0.2304, pruned_loss=0.03661, over 4815.00 frames. ], tot_loss[loss=0.1656, simple_loss=0.238, pruned_loss=0.04658, over 948386.29 frames. ], batch size: 51, lr: 3.02e-03, grad_scale: 32.0 2023-04-27 21:40:07,971 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.116e+02 1.516e+02 1.776e+02 2.068e+02 3.891e+02, threshold=3.551e+02, percent-clipped=2.0 2023-04-27 21:40:15,853 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.3508, 1.5965, 1.7809, 1.9035, 1.7680, 1.7475, 1.8456, 1.7922], device='cuda:1'), covar=tensor([0.3804, 0.5474, 0.4175, 0.4282, 0.5313, 0.6895, 0.4823, 0.4573], device='cuda:1'), in_proj_covar=tensor([0.0340, 0.0375, 0.0327, 0.0341, 0.0350, 0.0394, 0.0358, 0.0331], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 21:41:09,819 INFO [finetune.py:976] (1/7) Epoch 25, batch 1050, loss[loss=0.1462, simple_loss=0.2231, pruned_loss=0.03465, over 4756.00 frames. ], tot_loss[loss=0.1683, simple_loss=0.2414, pruned_loss=0.04763, over 949135.22 frames. ], batch size: 27, lr: 3.02e-03, grad_scale: 32.0 2023-04-27 21:41:21,495 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0253, 1.7747, 1.9861, 2.4087, 2.3234, 1.9283, 1.7262, 2.2086], device='cuda:1'), covar=tensor([0.0813, 0.1198, 0.0768, 0.0551, 0.0694, 0.0877, 0.0753, 0.0510], device='cuda:1'), in_proj_covar=tensor([0.0187, 0.0203, 0.0187, 0.0174, 0.0179, 0.0179, 0.0152, 0.0179], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 21:41:52,282 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=138548.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:41:53,397 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9902, 2.3894, 2.0800, 2.3146, 1.7170, 1.9666, 2.0070, 1.5906], device='cuda:1'), covar=tensor([0.1504, 0.1083, 0.0723, 0.0919, 0.3062, 0.1021, 0.1537, 0.2287], device='cuda:1'), in_proj_covar=tensor([0.0284, 0.0299, 0.0213, 0.0276, 0.0311, 0.0255, 0.0248, 0.0263], device='cuda:1'), out_proj_covar=tensor([1.1338e-04, 1.1795e-04, 8.3949e-05, 1.0884e-04, 1.2539e-04, 1.0041e-04, 9.9919e-05, 1.0392e-04], device='cuda:1') 2023-04-27 21:41:54,998 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=138551.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:42:14,665 INFO [finetune.py:976] (1/7) Epoch 25, batch 1100, loss[loss=0.1703, simple_loss=0.2476, pruned_loss=0.04653, over 4856.00 frames. ], tot_loss[loss=0.171, simple_loss=0.2442, pruned_loss=0.04887, over 949622.96 frames. ], batch size: 44, lr: 3.02e-03, grad_scale: 32.0 2023-04-27 21:42:16,463 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.782e+01 1.591e+02 1.867e+02 2.331e+02 5.511e+02, threshold=3.734e+02, percent-clipped=3.0 2023-04-27 21:42:55,899 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=138596.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:42:57,699 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=138599.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:43:18,592 INFO [finetune.py:976] (1/7) Epoch 25, batch 1150, loss[loss=0.1552, simple_loss=0.2362, pruned_loss=0.03712, over 4815.00 frames. ], tot_loss[loss=0.1721, simple_loss=0.2454, pruned_loss=0.04941, over 951642.38 frames. ], batch size: 38, lr: 3.02e-03, grad_scale: 32.0 2023-04-27 21:43:38,273 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=138626.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:44:23,366 INFO [finetune.py:976] (1/7) Epoch 25, batch 1200, loss[loss=0.1454, simple_loss=0.2252, pruned_loss=0.03279, over 4896.00 frames. ], tot_loss[loss=0.1708, simple_loss=0.2441, pruned_loss=0.04875, over 953420.19 frames. ], batch size: 43, lr: 3.02e-03, grad_scale: 64.0 2023-04-27 21:44:26,076 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.080e+02 1.537e+02 1.836e+02 2.355e+02 3.778e+02, threshold=3.672e+02, percent-clipped=1.0 2023-04-27 21:44:43,175 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=138678.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:44:44,973 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=138681.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:44:53,094 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=138686.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:45:04,529 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0036, 1.5390, 1.6324, 1.6930, 2.1167, 1.7807, 1.5309, 1.5001], device='cuda:1'), covar=tensor([0.1613, 0.1256, 0.1966, 0.1369, 0.0845, 0.1493, 0.2011, 0.2430], device='cuda:1'), in_proj_covar=tensor([0.0312, 0.0307, 0.0350, 0.0284, 0.0325, 0.0305, 0.0299, 0.0371], device='cuda:1'), out_proj_covar=tensor([6.3804e-05, 6.3071e-05, 7.3778e-05, 5.6973e-05, 6.6833e-05, 6.3873e-05, 6.2012e-05, 7.8693e-05], device='cuda:1') 2023-04-27 21:45:05,116 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=138697.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:45:29,240 INFO [finetune.py:976] (1/7) Epoch 25, batch 1250, loss[loss=0.1769, simple_loss=0.2373, pruned_loss=0.05827, over 4257.00 frames. ], tot_loss[loss=0.1683, simple_loss=0.2409, pruned_loss=0.04788, over 954514.07 frames. ], batch size: 65, lr: 3.02e-03, grad_scale: 64.0 2023-04-27 21:46:08,004 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=138742.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:46:11,070 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=138747.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:46:23,975 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=138758.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:46:33,339 INFO [finetune.py:976] (1/7) Epoch 25, batch 1300, loss[loss=0.1667, simple_loss=0.2324, pruned_loss=0.05056, over 4827.00 frames. ], tot_loss[loss=0.1655, simple_loss=0.2376, pruned_loss=0.04673, over 954706.31 frames. ], batch size: 51, lr: 3.02e-03, grad_scale: 64.0 2023-04-27 21:46:40,295 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.044e+02 1.540e+02 1.806e+02 2.216e+02 3.604e+02, threshold=3.612e+02, percent-clipped=0.0 2023-04-27 21:47:27,453 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-04-27 21:47:44,176 INFO [finetune.py:976] (1/7) Epoch 25, batch 1350, loss[loss=0.1839, simple_loss=0.2549, pruned_loss=0.05645, over 4866.00 frames. ], tot_loss[loss=0.1645, simple_loss=0.2368, pruned_loss=0.04605, over 956816.95 frames. ], batch size: 34, lr: 3.02e-03, grad_scale: 64.0 2023-04-27 21:48:48,892 INFO [finetune.py:976] (1/7) Epoch 25, batch 1400, loss[loss=0.137, simple_loss=0.2245, pruned_loss=0.02473, over 4783.00 frames. ], tot_loss[loss=0.1656, simple_loss=0.2391, pruned_loss=0.04602, over 956639.89 frames. ], batch size: 29, lr: 3.02e-03, grad_scale: 64.0 2023-04-27 21:48:50,719 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.057e+02 1.518e+02 1.853e+02 2.271e+02 4.217e+02, threshold=3.705e+02, percent-clipped=3.0 2023-04-27 21:48:59,992 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.1060, 0.6599, 0.8809, 0.8316, 1.2602, 0.9922, 0.9157, 0.9382], device='cuda:1'), covar=tensor([0.1623, 0.1478, 0.1961, 0.1619, 0.0961, 0.1399, 0.1616, 0.2204], device='cuda:1'), in_proj_covar=tensor([0.0310, 0.0305, 0.0349, 0.0283, 0.0324, 0.0304, 0.0297, 0.0370], device='cuda:1'), out_proj_covar=tensor([6.3371e-05, 6.2842e-05, 7.3417e-05, 5.6721e-05, 6.6443e-05, 6.3662e-05, 6.1673e-05, 7.8321e-05], device='cuda:1') 2023-04-27 21:49:15,026 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.3826, 1.5679, 1.5200, 1.7376, 1.7062, 1.9231, 1.4610, 3.1172], device='cuda:1'), covar=tensor([0.0592, 0.0718, 0.0687, 0.1106, 0.0560, 0.0589, 0.0706, 0.0174], device='cuda:1'), in_proj_covar=tensor([0.0038, 0.0039, 0.0039, 0.0043, 0.0040, 0.0038, 0.0038, 0.0055], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0013, 0.0013, 0.0015, 0.0014, 0.0013, 0.0013, 0.0016], device='cuda:1') 2023-04-27 21:49:44,982 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=138909.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:49:54,328 INFO [finetune.py:976] (1/7) Epoch 25, batch 1450, loss[loss=0.1786, simple_loss=0.266, pruned_loss=0.04566, over 4885.00 frames. ], tot_loss[loss=0.1669, simple_loss=0.2409, pruned_loss=0.0464, over 954765.23 frames. ], batch size: 35, lr: 3.01e-03, grad_scale: 64.0 2023-04-27 21:50:07,283 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.2967, 2.2621, 1.8381, 2.0392, 2.4558, 1.9521, 2.7898, 1.7011], device='cuda:1'), covar=tensor([0.3625, 0.1876, 0.4782, 0.2713, 0.1536, 0.2395, 0.1543, 0.4160], device='cuda:1'), in_proj_covar=tensor([0.0343, 0.0354, 0.0432, 0.0352, 0.0382, 0.0379, 0.0370, 0.0425], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 21:50:07,874 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=138926.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:50:16,901 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7269, 1.4205, 1.3106, 1.6225, 2.0205, 1.6074, 1.4355, 1.1942], device='cuda:1'), covar=tensor([0.1322, 0.1268, 0.1570, 0.1099, 0.0694, 0.1382, 0.1577, 0.2048], device='cuda:1'), in_proj_covar=tensor([0.0310, 0.0306, 0.0348, 0.0283, 0.0323, 0.0304, 0.0297, 0.0370], device='cuda:1'), out_proj_covar=tensor([6.3313e-05, 6.2892e-05, 7.3372e-05, 5.6763e-05, 6.6383e-05, 6.3650e-05, 6.1704e-05, 7.8397e-05], device='cuda:1') 2023-04-27 21:50:29,973 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5228, 1.6594, 1.3887, 1.1226, 1.1388, 1.1323, 1.3952, 1.1028], device='cuda:1'), covar=tensor([0.1746, 0.1341, 0.1561, 0.1751, 0.2349, 0.2010, 0.1033, 0.2081], device='cuda:1'), in_proj_covar=tensor([0.0197, 0.0209, 0.0168, 0.0204, 0.0199, 0.0186, 0.0156, 0.0185], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-27 21:51:01,100 INFO [finetune.py:976] (1/7) Epoch 25, batch 1500, loss[loss=0.1966, simple_loss=0.2588, pruned_loss=0.06716, over 4844.00 frames. ], tot_loss[loss=0.1683, simple_loss=0.2427, pruned_loss=0.04699, over 955842.12 frames. ], batch size: 47, lr: 3.01e-03, grad_scale: 64.0 2023-04-27 21:51:03,883 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.140e+02 1.504e+02 1.740e+02 2.143e+02 4.195e+02, threshold=3.481e+02, percent-clipped=2.0 2023-04-27 21:51:10,092 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=138970.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 21:51:12,527 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=138974.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:51:19,561 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=138978.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:52:02,852 INFO [finetune.py:976] (1/7) Epoch 25, batch 1550, loss[loss=0.1906, simple_loss=0.2607, pruned_loss=0.06024, over 4823.00 frames. ], tot_loss[loss=0.169, simple_loss=0.2432, pruned_loss=0.04741, over 953363.41 frames. ], batch size: 30, lr: 3.01e-03, grad_scale: 64.0 2023-04-27 21:52:11,153 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=139026.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:52:18,843 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=139037.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:52:22,321 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=139042.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:52:26,101 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.2328, 1.7278, 2.1044, 2.4712, 2.1278, 1.6769, 1.3236, 1.8231], device='cuda:1'), covar=tensor([0.3018, 0.2842, 0.1600, 0.1940, 0.2298, 0.2503, 0.4030, 0.1852], device='cuda:1'), in_proj_covar=tensor([0.0292, 0.0243, 0.0226, 0.0311, 0.0221, 0.0233, 0.0226, 0.0185], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-27 21:52:29,077 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=139053.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:52:41,925 INFO [finetune.py:976] (1/7) Epoch 25, batch 1600, loss[loss=0.1845, simple_loss=0.2573, pruned_loss=0.05586, over 4916.00 frames. ], tot_loss[loss=0.1684, simple_loss=0.2418, pruned_loss=0.04752, over 955232.52 frames. ], batch size: 36, lr: 3.01e-03, grad_scale: 64.0 2023-04-27 21:52:42,645 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([0.8585, 1.0935, 3.2994, 3.0433, 2.9748, 3.1866, 3.2075, 2.8572], device='cuda:1'), covar=tensor([0.7946, 0.6147, 0.1646, 0.2635, 0.1653, 0.2284, 0.1814, 0.2144], device='cuda:1'), in_proj_covar=tensor([0.0315, 0.0310, 0.0411, 0.0414, 0.0352, 0.0416, 0.0322, 0.0372], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-27 21:52:44,232 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.006e+02 1.576e+02 1.832e+02 2.139e+02 4.092e+02, threshold=3.665e+02, percent-clipped=3.0 2023-04-27 21:53:16,627 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7981, 2.1902, 1.7260, 1.5058, 1.3159, 1.3394, 1.8253, 1.2543], device='cuda:1'), covar=tensor([0.1638, 0.1181, 0.1507, 0.1809, 0.2299, 0.1960, 0.0968, 0.2121], device='cuda:1'), in_proj_covar=tensor([0.0197, 0.0209, 0.0168, 0.0203, 0.0199, 0.0185, 0.0155, 0.0185], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-27 21:53:47,319 INFO [finetune.py:976] (1/7) Epoch 25, batch 1650, loss[loss=0.1633, simple_loss=0.2203, pruned_loss=0.05308, over 4702.00 frames. ], tot_loss[loss=0.1666, simple_loss=0.2395, pruned_loss=0.04682, over 953866.96 frames. ], batch size: 23, lr: 3.01e-03, grad_scale: 64.0 2023-04-27 21:54:18,500 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-04-27 21:54:33,489 INFO [finetune.py:976] (1/7) Epoch 25, batch 1700, loss[loss=0.2042, simple_loss=0.2782, pruned_loss=0.06513, over 4185.00 frames. ], tot_loss[loss=0.1665, simple_loss=0.2385, pruned_loss=0.04723, over 953459.12 frames. ], batch size: 65, lr: 3.01e-03, grad_scale: 64.0 2023-04-27 21:54:33,706 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-04-27 21:54:35,338 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.004e+02 1.477e+02 1.739e+02 2.189e+02 4.097e+02, threshold=3.477e+02, percent-clipped=1.0 2023-04-27 21:55:07,051 INFO [finetune.py:976] (1/7) Epoch 25, batch 1750, loss[loss=0.1896, simple_loss=0.268, pruned_loss=0.05563, over 4911.00 frames. ], tot_loss[loss=0.1696, simple_loss=0.2415, pruned_loss=0.04885, over 954443.56 frames. ], batch size: 36, lr: 3.01e-03, grad_scale: 64.0 2023-04-27 21:55:23,349 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=139238.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:55:41,181 INFO [finetune.py:976] (1/7) Epoch 25, batch 1800, loss[loss=0.1832, simple_loss=0.2565, pruned_loss=0.05497, over 4904.00 frames. ], tot_loss[loss=0.171, simple_loss=0.2435, pruned_loss=0.04925, over 953852.77 frames. ], batch size: 37, lr: 3.01e-03, grad_scale: 64.0 2023-04-27 21:55:41,245 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=139265.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 21:55:42,992 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.657e+01 1.551e+02 1.902e+02 2.363e+02 3.513e+02, threshold=3.803e+02, percent-clipped=1.0 2023-04-27 21:55:48,016 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=139276.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:55:52,828 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5755, 0.6990, 1.5170, 1.9661, 1.6344, 1.4807, 1.5224, 1.5158], device='cuda:1'), covar=tensor([0.3980, 0.6145, 0.5251, 0.5131, 0.5048, 0.6473, 0.6503, 0.7634], device='cuda:1'), in_proj_covar=tensor([0.0436, 0.0420, 0.0510, 0.0507, 0.0466, 0.0500, 0.0502, 0.0515], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 21:55:57,460 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=139289.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:56:04,586 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=139299.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:56:14,571 INFO [finetune.py:976] (1/7) Epoch 25, batch 1850, loss[loss=0.1955, simple_loss=0.2666, pruned_loss=0.0622, over 4815.00 frames. ], tot_loss[loss=0.1723, simple_loss=0.2447, pruned_loss=0.04992, over 954781.04 frames. ], batch size: 33, lr: 3.01e-03, grad_scale: 64.0 2023-04-27 21:56:29,232 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=139337.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:56:29,279 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=139337.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 21:56:33,189 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=139342.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:56:33,225 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.2461, 1.7169, 2.1190, 2.5853, 2.1433, 1.6879, 1.4720, 1.9447], device='cuda:1'), covar=tensor([0.2932, 0.3091, 0.1563, 0.2136, 0.2720, 0.2557, 0.4283, 0.1937], device='cuda:1'), in_proj_covar=tensor([0.0292, 0.0244, 0.0226, 0.0311, 0.0220, 0.0232, 0.0226, 0.0185], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-27 21:56:43,882 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=139350.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:56:51,942 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=139353.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:57:05,218 INFO [finetune.py:976] (1/7) Epoch 25, batch 1900, loss[loss=0.1606, simple_loss=0.2437, pruned_loss=0.03873, over 4862.00 frames. ], tot_loss[loss=0.1742, simple_loss=0.2471, pruned_loss=0.05067, over 952468.71 frames. ], batch size: 44, lr: 3.01e-03, grad_scale: 32.0 2023-04-27 21:57:07,684 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.239e+02 1.510e+02 1.790e+02 2.163e+02 4.429e+02, threshold=3.581e+02, percent-clipped=1.0 2023-04-27 21:57:28,904 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=139385.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:57:36,778 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=139390.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:57:49,497 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=139401.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:58:09,750 INFO [finetune.py:976] (1/7) Epoch 25, batch 1950, loss[loss=0.1436, simple_loss=0.2196, pruned_loss=0.03374, over 4813.00 frames. ], tot_loss[loss=0.172, simple_loss=0.245, pruned_loss=0.04945, over 952578.74 frames. ], batch size: 38, lr: 3.01e-03, grad_scale: 32.0 2023-04-27 21:58:22,779 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.14 vs. limit=2.0 2023-04-27 21:59:13,359 INFO [finetune.py:976] (1/7) Epoch 25, batch 2000, loss[loss=0.1717, simple_loss=0.246, pruned_loss=0.04869, over 4739.00 frames. ], tot_loss[loss=0.1712, simple_loss=0.2435, pruned_loss=0.04944, over 954814.85 frames. ], batch size: 54, lr: 3.01e-03, grad_scale: 32.0 2023-04-27 21:59:15,797 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.085e+02 1.572e+02 1.799e+02 2.251e+02 3.942e+02, threshold=3.599e+02, percent-clipped=2.0 2023-04-27 22:00:17,501 INFO [finetune.py:976] (1/7) Epoch 25, batch 2050, loss[loss=0.2064, simple_loss=0.2687, pruned_loss=0.07202, over 4934.00 frames. ], tot_loss[loss=0.168, simple_loss=0.2401, pruned_loss=0.04793, over 955199.36 frames. ], batch size: 38, lr: 3.01e-03, grad_scale: 32.0 2023-04-27 22:01:00,585 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-04-27 22:01:21,673 INFO [finetune.py:976] (1/7) Epoch 25, batch 2100, loss[loss=0.192, simple_loss=0.2437, pruned_loss=0.0701, over 4892.00 frames. ], tot_loss[loss=0.1675, simple_loss=0.2391, pruned_loss=0.04792, over 952229.37 frames. ], batch size: 32, lr: 3.01e-03, grad_scale: 32.0 2023-04-27 22:01:21,776 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=139565.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 22:01:24,120 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.088e+01 1.541e+02 1.795e+02 2.157e+02 3.840e+02, threshold=3.589e+02, percent-clipped=1.0 2023-04-27 22:01:30,514 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=139571.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:01:37,247 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=139582.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:01:44,573 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=139594.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:01:48,307 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-04-27 22:01:51,977 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7938, 2.4380, 1.8725, 1.7664, 1.3137, 1.3497, 1.9027, 1.2616], device='cuda:1'), covar=tensor([0.1837, 0.1409, 0.1348, 0.1760, 0.2400, 0.2097, 0.1009, 0.2195], device='cuda:1'), in_proj_covar=tensor([0.0198, 0.0210, 0.0168, 0.0204, 0.0199, 0.0186, 0.0156, 0.0186], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-27 22:01:57,942 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=139613.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:01:59,118 INFO [finetune.py:976] (1/7) Epoch 25, batch 2150, loss[loss=0.1468, simple_loss=0.2102, pruned_loss=0.04175, over 4763.00 frames. ], tot_loss[loss=0.1685, simple_loss=0.2407, pruned_loss=0.04819, over 953382.68 frames. ], batch size: 27, lr: 3.01e-03, grad_scale: 32.0 2023-04-27 22:02:10,542 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=139632.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 22:02:10,577 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=139632.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:02:17,307 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=139643.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:02:18,458 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=139645.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:02:31,935 INFO [finetune.py:976] (1/7) Epoch 25, batch 2200, loss[loss=0.2005, simple_loss=0.2759, pruned_loss=0.0625, over 4803.00 frames. ], tot_loss[loss=0.1698, simple_loss=0.2423, pruned_loss=0.04862, over 953663.59 frames. ], batch size: 41, lr: 3.01e-03, grad_scale: 32.0 2023-04-27 22:02:34,822 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.151e+02 1.529e+02 1.754e+02 2.258e+02 5.468e+02, threshold=3.509e+02, percent-clipped=4.0 2023-04-27 22:02:54,928 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.1656, 2.4623, 1.1286, 1.4056, 1.9394, 1.2801, 3.0760, 1.7181], device='cuda:1'), covar=tensor([0.0622, 0.0548, 0.0679, 0.1214, 0.0437, 0.0947, 0.0254, 0.0581], device='cuda:1'), in_proj_covar=tensor([0.0050, 0.0064, 0.0047, 0.0046, 0.0049, 0.0051, 0.0072, 0.0050], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:1') 2023-04-27 22:03:04,782 INFO [finetune.py:976] (1/7) Epoch 25, batch 2250, loss[loss=0.1577, simple_loss=0.2435, pruned_loss=0.03595, over 4840.00 frames. ], tot_loss[loss=0.1702, simple_loss=0.2433, pruned_loss=0.04849, over 955444.49 frames. ], batch size: 44, lr: 3.01e-03, grad_scale: 32.0 2023-04-27 22:03:38,678 INFO [finetune.py:976] (1/7) Epoch 25, batch 2300, loss[loss=0.1811, simple_loss=0.2597, pruned_loss=0.0513, over 4810.00 frames. ], tot_loss[loss=0.1708, simple_loss=0.2443, pruned_loss=0.04861, over 952762.66 frames. ], batch size: 41, lr: 3.01e-03, grad_scale: 32.0 2023-04-27 22:03:41,544 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.026e+02 1.526e+02 1.763e+02 2.328e+02 5.368e+02, threshold=3.526e+02, percent-clipped=6.0 2023-04-27 22:03:41,673 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=139769.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:03:53,062 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.16 vs. limit=5.0 2023-04-27 22:03:54,757 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=139789.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:04:03,986 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=139804.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:04:05,253 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=139806.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:04:23,021 INFO [finetune.py:976] (1/7) Epoch 25, batch 2350, loss[loss=0.198, simple_loss=0.2591, pruned_loss=0.06848, over 4823.00 frames. ], tot_loss[loss=0.1685, simple_loss=0.2418, pruned_loss=0.04764, over 954116.63 frames. ], batch size: 30, lr: 3.01e-03, grad_scale: 32.0 2023-04-27 22:04:44,845 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=139830.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:05:06,854 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=139850.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:05:06,863 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=139850.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:05:17,688 INFO [finetune.py:976] (1/7) Epoch 25, batch 2400, loss[loss=0.1574, simple_loss=0.2301, pruned_loss=0.0424, over 4740.00 frames. ], tot_loss[loss=0.1666, simple_loss=0.2391, pruned_loss=0.04709, over 954725.55 frames. ], batch size: 59, lr: 3.01e-03, grad_scale: 32.0 2023-04-27 22:05:17,823 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=139865.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:05:19,518 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=139867.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:05:20,612 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.574e+01 1.558e+02 1.905e+02 2.229e+02 5.557e+02, threshold=3.809e+02, percent-clipped=2.0 2023-04-27 22:05:23,142 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.3246, 1.4856, 1.3741, 1.7912, 1.6126, 1.6859, 1.4275, 2.8243], device='cuda:1'), covar=tensor([0.0563, 0.0715, 0.0707, 0.0989, 0.0535, 0.0462, 0.0627, 0.0201], device='cuda:1'), in_proj_covar=tensor([0.0037, 0.0039, 0.0039, 0.0043, 0.0040, 0.0038, 0.0038, 0.0055], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0013, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:1') 2023-04-27 22:05:37,245 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=139894.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:05:53,372 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=139911.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:05:55,629 INFO [finetune.py:976] (1/7) Epoch 25, batch 2450, loss[loss=0.218, simple_loss=0.2852, pruned_loss=0.07538, over 4804.00 frames. ], tot_loss[loss=0.1648, simple_loss=0.2367, pruned_loss=0.04641, over 955769.95 frames. ], batch size: 40, lr: 3.01e-03, grad_scale: 16.0 2023-04-27 22:06:15,192 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=139927.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:06:18,752 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=139932.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 22:06:27,863 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=139938.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:06:29,764 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8071, 1.3432, 1.4247, 1.5623, 1.9601, 1.6041, 1.3526, 1.3852], device='cuda:1'), covar=tensor([0.1866, 0.1615, 0.1920, 0.1425, 0.0948, 0.1861, 0.2076, 0.2497], device='cuda:1'), in_proj_covar=tensor([0.0312, 0.0308, 0.0351, 0.0286, 0.0328, 0.0306, 0.0298, 0.0373], device='cuda:1'), out_proj_covar=tensor([6.3870e-05, 6.3454e-05, 7.3755e-05, 5.7419e-05, 6.7438e-05, 6.4073e-05, 6.1907e-05, 7.9154e-05], device='cuda:1') 2023-04-27 22:06:35,927 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=139942.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:06:37,758 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=139945.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:06:59,676 INFO [finetune.py:976] (1/7) Epoch 25, batch 2500, loss[loss=0.1755, simple_loss=0.2576, pruned_loss=0.04671, over 4822.00 frames. ], tot_loss[loss=0.1669, simple_loss=0.2385, pruned_loss=0.04761, over 954784.39 frames. ], batch size: 45, lr: 3.01e-03, grad_scale: 16.0 2023-04-27 22:07:04,983 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.044e+02 1.582e+02 1.792e+02 2.204e+02 3.647e+02, threshold=3.585e+02, percent-clipped=0.0 2023-04-27 22:07:11,614 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=139980.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:07:21,148 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=139993.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:07:35,751 INFO [finetune.py:976] (1/7) Epoch 25, batch 2550, loss[loss=0.2549, simple_loss=0.3183, pruned_loss=0.09573, over 4181.00 frames. ], tot_loss[loss=0.1703, simple_loss=0.2426, pruned_loss=0.04901, over 953738.42 frames. ], batch size: 65, lr: 3.01e-03, grad_scale: 16.0 2023-04-27 22:07:37,199 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=3.65 vs. limit=5.0 2023-04-27 22:07:55,542 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.3640, 3.0143, 0.8170, 1.5845, 1.7723, 2.0688, 1.7723, 0.9345], device='cuda:1'), covar=tensor([0.1422, 0.0997, 0.2002, 0.1323, 0.1024, 0.1037, 0.1607, 0.1818], device='cuda:1'), in_proj_covar=tensor([0.0118, 0.0240, 0.0137, 0.0122, 0.0132, 0.0153, 0.0118, 0.0119], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-27 22:08:09,499 INFO [finetune.py:976] (1/7) Epoch 25, batch 2600, loss[loss=0.1528, simple_loss=0.2273, pruned_loss=0.03914, over 4782.00 frames. ], tot_loss[loss=0.1711, simple_loss=0.2441, pruned_loss=0.04906, over 954300.60 frames. ], batch size: 29, lr: 3.01e-03, grad_scale: 16.0 2023-04-27 22:08:12,530 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.025e+02 1.514e+02 1.836e+02 2.215e+02 4.899e+02, threshold=3.672e+02, percent-clipped=3.0 2023-04-27 22:08:35,672 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=140103.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:08:39,911 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=140110.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:08:42,875 INFO [finetune.py:976] (1/7) Epoch 25, batch 2650, loss[loss=0.175, simple_loss=0.2492, pruned_loss=0.0504, over 4787.00 frames. ], tot_loss[loss=0.1716, simple_loss=0.2451, pruned_loss=0.04909, over 954770.26 frames. ], batch size: 29, lr: 3.01e-03, grad_scale: 16.0 2023-04-27 22:08:49,431 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=140125.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:09:04,044 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=140145.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:09:13,148 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=140160.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:09:14,362 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=140162.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:09:15,599 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=140164.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:09:16,092 INFO [finetune.py:976] (1/7) Epoch 25, batch 2700, loss[loss=0.1851, simple_loss=0.26, pruned_loss=0.05515, over 4817.00 frames. ], tot_loss[loss=0.1694, simple_loss=0.2433, pruned_loss=0.04775, over 956488.54 frames. ], batch size: 40, lr: 3.01e-03, grad_scale: 16.0 2023-04-27 22:09:17,428 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.2261, 1.5546, 1.3775, 1.7785, 1.6514, 2.0800, 1.4227, 3.6916], device='cuda:1'), covar=tensor([0.0583, 0.0774, 0.0779, 0.1194, 0.0643, 0.0440, 0.0758, 0.0124], device='cuda:1'), in_proj_covar=tensor([0.0038, 0.0039, 0.0039, 0.0043, 0.0040, 0.0038, 0.0038, 0.0055], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0013, 0.0013, 0.0015, 0.0014, 0.0013, 0.0013, 0.0016], device='cuda:1') 2023-04-27 22:09:19,148 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.968e+01 1.589e+02 1.838e+02 2.247e+02 3.608e+02, threshold=3.675e+02, percent-clipped=0.0 2023-04-27 22:09:19,882 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=140171.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:09:50,088 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=140206.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:10:00,543 INFO [finetune.py:976] (1/7) Epoch 25, batch 2750, loss[loss=0.1482, simple_loss=0.214, pruned_loss=0.04127, over 4783.00 frames. ], tot_loss[loss=0.1672, simple_loss=0.2405, pruned_loss=0.04697, over 955450.15 frames. ], batch size: 25, lr: 3.01e-03, grad_scale: 16.0 2023-04-27 22:10:18,554 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=140227.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:10:32,122 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=140238.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:11:03,254 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=140259.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:11:12,261 INFO [finetune.py:976] (1/7) Epoch 25, batch 2800, loss[loss=0.156, simple_loss=0.2108, pruned_loss=0.05054, over 4273.00 frames. ], tot_loss[loss=0.165, simple_loss=0.2378, pruned_loss=0.04614, over 954669.68 frames. ], batch size: 18, lr: 3.01e-03, grad_scale: 16.0 2023-04-27 22:11:15,320 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.076e+02 1.465e+02 1.747e+02 2.235e+02 4.062e+02, threshold=3.495e+02, percent-clipped=1.0 2023-04-27 22:11:20,368 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=140275.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:11:32,985 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=140286.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:12:14,090 INFO [finetune.py:976] (1/7) Epoch 25, batch 2850, loss[loss=0.1323, simple_loss=0.2094, pruned_loss=0.02759, over 4760.00 frames. ], tot_loss[loss=0.1641, simple_loss=0.2365, pruned_loss=0.04586, over 953743.65 frames. ], batch size: 23, lr: 3.00e-03, grad_scale: 16.0 2023-04-27 22:12:17,252 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=140320.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:12:24,391 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.1730, 1.6044, 1.4497, 1.7210, 1.6372, 1.8625, 1.4028, 3.5008], device='cuda:1'), covar=tensor([0.0614, 0.0744, 0.0782, 0.1161, 0.0608, 0.0513, 0.0730, 0.0148], device='cuda:1'), in_proj_covar=tensor([0.0037, 0.0038, 0.0039, 0.0043, 0.0040, 0.0038, 0.0038, 0.0055], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0013, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:1') 2023-04-27 22:12:32,567 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.2773, 1.1999, 1.5410, 1.5456, 1.2100, 1.1432, 1.2046, 0.7264], device='cuda:1'), covar=tensor([0.0503, 0.0613, 0.0383, 0.0544, 0.0730, 0.1037, 0.0569, 0.0592], device='cuda:1'), in_proj_covar=tensor([0.0070, 0.0067, 0.0066, 0.0068, 0.0074, 0.0095, 0.0073, 0.0063], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-27 22:12:48,658 INFO [finetune.py:976] (1/7) Epoch 25, batch 2900, loss[loss=0.19, simple_loss=0.271, pruned_loss=0.05449, over 4285.00 frames. ], tot_loss[loss=0.1662, simple_loss=0.2389, pruned_loss=0.04673, over 951727.32 frames. ], batch size: 65, lr: 3.00e-03, grad_scale: 16.0 2023-04-27 22:12:51,721 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.049e+02 1.618e+02 1.938e+02 2.286e+02 3.478e+02, threshold=3.877e+02, percent-clipped=0.0 2023-04-27 22:13:02,683 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=140387.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:13:22,623 INFO [finetune.py:976] (1/7) Epoch 25, batch 2950, loss[loss=0.1714, simple_loss=0.2443, pruned_loss=0.04921, over 4790.00 frames. ], tot_loss[loss=0.1692, simple_loss=0.2421, pruned_loss=0.0482, over 951310.93 frames. ], batch size: 51, lr: 3.00e-03, grad_scale: 16.0 2023-04-27 22:13:28,143 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7032, 1.9531, 1.0988, 1.3960, 2.1305, 1.5645, 1.4751, 1.5238], device='cuda:1'), covar=tensor([0.0492, 0.0347, 0.0306, 0.0548, 0.0242, 0.0517, 0.0493, 0.0562], device='cuda:1'), in_proj_covar=tensor([0.0028, 0.0023, 0.0022, 0.0028, 0.0019, 0.0028, 0.0027, 0.0028], device='cuda:1'), out_proj_covar=tensor([0.0051, 0.0046, 0.0038, 0.0052, 0.0039, 0.0050, 0.0050, 0.0052], device='cuda:1') 2023-04-27 22:13:28,749 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=140425.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:13:42,312 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=140445.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:13:44,104 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=140448.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:13:52,225 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=140459.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:13:52,861 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=140460.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:13:54,548 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=140462.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:13:56,276 INFO [finetune.py:976] (1/7) Epoch 25, batch 3000, loss[loss=0.2304, simple_loss=0.2881, pruned_loss=0.08638, over 4334.00 frames. ], tot_loss[loss=0.1708, simple_loss=0.2438, pruned_loss=0.04884, over 950035.47 frames. ], batch size: 19, lr: 3.00e-03, grad_scale: 16.0 2023-04-27 22:13:56,276 INFO [finetune.py:1001] (1/7) Computing validation loss 2023-04-27 22:14:07,208 INFO [finetune.py:1010] (1/7) Epoch 25, validation: loss=0.1531, simple_loss=0.2225, pruned_loss=0.04184, over 2265189.00 frames. 2023-04-27 22:14:07,208 INFO [finetune.py:1011] (1/7) Maximum memory allocated so far is 6391MB 2023-04-27 22:14:07,900 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=140466.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:14:10,185 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.175e+02 1.715e+02 2.080e+02 2.596e+02 3.715e+02, threshold=4.161e+02, percent-clipped=0.0 2023-04-27 22:14:12,095 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=140473.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:14:21,869 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=140489.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:14:24,278 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=140493.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:14:33,147 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=140506.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:14:33,175 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7200, 2.8025, 2.2593, 2.5187, 2.7817, 2.5347, 3.7163, 2.0183], device='cuda:1'), covar=tensor([0.3466, 0.1882, 0.4215, 0.2742, 0.1642, 0.2269, 0.1032, 0.4122], device='cuda:1'), in_proj_covar=tensor([0.0341, 0.0354, 0.0429, 0.0353, 0.0380, 0.0376, 0.0370, 0.0425], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 22:14:34,804 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=140508.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:14:35,463 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=140509.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:14:36,041 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=140510.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:14:36,173 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-04-27 22:14:39,580 INFO [finetune.py:976] (1/7) Epoch 25, batch 3050, loss[loss=0.1468, simple_loss=0.2149, pruned_loss=0.03938, over 4732.00 frames. ], tot_loss[loss=0.1711, simple_loss=0.2444, pruned_loss=0.04887, over 952370.64 frames. ], batch size: 23, lr: 3.00e-03, grad_scale: 16.0 2023-04-27 22:15:02,540 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=140550.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:15:05,406 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=140554.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:15:12,554 INFO [finetune.py:976] (1/7) Epoch 25, batch 3100, loss[loss=0.1298, simple_loss=0.2008, pruned_loss=0.02944, over 4796.00 frames. ], tot_loss[loss=0.1695, simple_loss=0.2422, pruned_loss=0.04834, over 953618.89 frames. ], batch size: 29, lr: 3.00e-03, grad_scale: 16.0 2023-04-27 22:15:12,707 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5831, 0.7083, 1.4632, 1.9992, 1.6661, 1.4792, 1.5257, 1.4911], device='cuda:1'), covar=tensor([0.4022, 0.6206, 0.5487, 0.5411, 0.5018, 0.6311, 0.6535, 0.7766], device='cuda:1'), in_proj_covar=tensor([0.0442, 0.0424, 0.0516, 0.0514, 0.0471, 0.0504, 0.0508, 0.0521], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 22:15:16,050 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.054e+02 1.508e+02 1.726e+02 2.242e+02 3.598e+02, threshold=3.452e+02, percent-clipped=0.0 2023-04-27 22:15:16,198 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=140570.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:15:16,925 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.72 vs. limit=2.0 2023-04-27 22:15:33,874 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.40 vs. limit=5.0 2023-04-27 22:16:07,368 INFO [finetune.py:976] (1/7) Epoch 25, batch 3150, loss[loss=0.1442, simple_loss=0.215, pruned_loss=0.03676, over 4734.00 frames. ], tot_loss[loss=0.1678, simple_loss=0.2401, pruned_loss=0.04772, over 952759.20 frames. ], batch size: 59, lr: 3.00e-03, grad_scale: 16.0 2023-04-27 22:16:07,435 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=140615.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:16:18,500 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.4596, 1.9115, 2.3660, 2.7421, 2.3312, 1.8789, 1.5780, 2.0710], device='cuda:1'), covar=tensor([0.2965, 0.3030, 0.1533, 0.2128, 0.2548, 0.2523, 0.3763, 0.1927], device='cuda:1'), in_proj_covar=tensor([0.0291, 0.0244, 0.0226, 0.0311, 0.0219, 0.0232, 0.0225, 0.0185], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-27 22:17:13,348 INFO [finetune.py:976] (1/7) Epoch 25, batch 3200, loss[loss=0.15, simple_loss=0.227, pruned_loss=0.0365, over 4832.00 frames. ], tot_loss[loss=0.1653, simple_loss=0.2371, pruned_loss=0.04674, over 953611.97 frames. ], batch size: 30, lr: 3.00e-03, grad_scale: 16.0 2023-04-27 22:17:21,870 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.047e+02 1.528e+02 1.799e+02 2.210e+02 5.000e+02, threshold=3.599e+02, percent-clipped=4.0 2023-04-27 22:18:04,974 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-04-27 22:18:19,775 INFO [finetune.py:976] (1/7) Epoch 25, batch 3250, loss[loss=0.1768, simple_loss=0.2636, pruned_loss=0.04501, over 4831.00 frames. ], tot_loss[loss=0.1675, simple_loss=0.2391, pruned_loss=0.04792, over 953290.40 frames. ], batch size: 49, lr: 3.00e-03, grad_scale: 16.0 2023-04-27 22:18:26,960 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=140723.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:18:55,045 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-04-27 22:18:56,001 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=140743.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:19:10,845 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=140759.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:19:20,223 INFO [finetune.py:976] (1/7) Epoch 25, batch 3300, loss[loss=0.1756, simple_loss=0.2553, pruned_loss=0.0479, over 4909.00 frames. ], tot_loss[loss=0.1703, simple_loss=0.2427, pruned_loss=0.04895, over 954397.21 frames. ], batch size: 36, lr: 3.00e-03, grad_scale: 16.0 2023-04-27 22:19:20,938 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=140766.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:19:27,805 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.140e+02 1.701e+02 1.963e+02 2.275e+02 6.006e+02, threshold=3.926e+02, percent-clipped=4.0 2023-04-27 22:19:27,998 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=3.42 vs. limit=5.0 2023-04-27 22:19:48,019 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=140784.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:19:52,006 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.3218, 1.6918, 1.5673, 2.0009, 2.0131, 2.1095, 1.6033, 4.3659], device='cuda:1'), covar=tensor([0.0522, 0.0770, 0.0733, 0.1077, 0.0540, 0.0509, 0.0668, 0.0076], device='cuda:1'), in_proj_covar=tensor([0.0037, 0.0039, 0.0039, 0.0043, 0.0040, 0.0038, 0.0038, 0.0055], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0013, 0.0013, 0.0015, 0.0014, 0.0013, 0.0013, 0.0016], device='cuda:1') 2023-04-27 22:20:13,961 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=140807.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:20:20,737 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=140809.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:20:23,767 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=140814.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:20:24,323 INFO [finetune.py:976] (1/7) Epoch 25, batch 3350, loss[loss=0.2073, simple_loss=0.2769, pruned_loss=0.06883, over 4825.00 frames. ], tot_loss[loss=0.1728, simple_loss=0.2459, pruned_loss=0.04979, over 956715.84 frames. ], batch size: 30, lr: 3.00e-03, grad_scale: 16.0 2023-04-27 22:20:33,712 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([3.9507, 3.8062, 2.8599, 4.5139, 3.8791, 3.8707, 1.5832, 3.9327], device='cuda:1'), covar=tensor([0.1771, 0.1203, 0.3172, 0.1299, 0.2804, 0.1945, 0.5537, 0.2075], device='cuda:1'), in_proj_covar=tensor([0.0247, 0.0220, 0.0254, 0.0309, 0.0300, 0.0250, 0.0275, 0.0275], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 22:20:46,745 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=140845.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:20:51,153 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.3946, 1.6609, 1.8563, 1.9572, 1.7994, 1.8657, 1.8591, 1.9039], device='cuda:1'), covar=tensor([0.4035, 0.5530, 0.4163, 0.4332, 0.5545, 0.6942, 0.5344, 0.4658], device='cuda:1'), in_proj_covar=tensor([0.0342, 0.0376, 0.0328, 0.0340, 0.0350, 0.0392, 0.0359, 0.0333], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 22:20:59,017 INFO [finetune.py:976] (1/7) Epoch 25, batch 3400, loss[loss=0.2018, simple_loss=0.2802, pruned_loss=0.06165, over 4823.00 frames. ], tot_loss[loss=0.1719, simple_loss=0.2454, pruned_loss=0.04919, over 955789.73 frames. ], batch size: 39, lr: 3.00e-03, grad_scale: 16.0 2023-04-27 22:20:59,079 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=140865.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:21:02,010 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.099e+02 1.526e+02 1.810e+02 2.121e+02 3.949e+02, threshold=3.620e+02, percent-clipped=1.0 2023-04-27 22:21:02,135 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=140870.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:21:32,411 INFO [finetune.py:976] (1/7) Epoch 25, batch 3450, loss[loss=0.1283, simple_loss=0.2027, pruned_loss=0.02694, over 4823.00 frames. ], tot_loss[loss=0.1713, simple_loss=0.2448, pruned_loss=0.04894, over 956279.59 frames. ], batch size: 33, lr: 3.00e-03, grad_scale: 16.0 2023-04-27 22:21:32,517 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=140915.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:21:33,740 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=140917.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:21:33,817 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-04-27 22:21:38,675 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.51 vs. limit=2.0 2023-04-27 22:21:50,303 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6526, 1.4547, 0.5418, 1.3415, 1.4664, 1.5142, 1.4459, 1.4052], device='cuda:1'), covar=tensor([0.0482, 0.0405, 0.0383, 0.0556, 0.0277, 0.0518, 0.0517, 0.0565], device='cuda:1'), in_proj_covar=tensor([0.0028, 0.0023, 0.0022, 0.0028, 0.0019, 0.0028, 0.0027, 0.0028], device='cuda:1'), out_proj_covar=tensor([0.0052, 0.0046, 0.0038, 0.0053, 0.0039, 0.0050, 0.0050, 0.0052], device='cuda:1') 2023-04-27 22:21:54,904 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=140947.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:22:05,108 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=140963.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:22:06,248 INFO [finetune.py:976] (1/7) Epoch 25, batch 3500, loss[loss=0.1675, simple_loss=0.2268, pruned_loss=0.05412, over 4927.00 frames. ], tot_loss[loss=0.1692, simple_loss=0.242, pruned_loss=0.04824, over 957399.04 frames. ], batch size: 38, lr: 3.00e-03, grad_scale: 16.0 2023-04-27 22:22:09,321 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.074e+02 1.472e+02 1.838e+02 2.111e+02 1.137e+03, threshold=3.676e+02, percent-clipped=2.0 2023-04-27 22:22:14,760 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=140978.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:22:35,385 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=141008.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:22:39,553 INFO [finetune.py:976] (1/7) Epoch 25, batch 3550, loss[loss=0.1691, simple_loss=0.2404, pruned_loss=0.04895, over 4905.00 frames. ], tot_loss[loss=0.1687, simple_loss=0.2408, pruned_loss=0.04828, over 958582.83 frames. ], batch size: 35, lr: 3.00e-03, grad_scale: 16.0 2023-04-27 22:22:56,968 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4268, 3.3168, 0.8837, 1.7496, 1.8504, 2.4353, 1.8511, 1.0195], device='cuda:1'), covar=tensor([0.1437, 0.0921, 0.2090, 0.1355, 0.1152, 0.0976, 0.1601, 0.1938], device='cuda:1'), in_proj_covar=tensor([0.0118, 0.0240, 0.0136, 0.0122, 0.0132, 0.0153, 0.0118, 0.0119], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-27 22:22:58,188 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=141043.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:23:10,552 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7847, 1.8458, 0.7725, 1.4150, 1.9188, 1.6195, 1.5080, 1.5794], device='cuda:1'), covar=tensor([0.0470, 0.0367, 0.0340, 0.0539, 0.0256, 0.0520, 0.0513, 0.0549], device='cuda:1'), in_proj_covar=tensor([0.0028, 0.0023, 0.0022, 0.0028, 0.0019, 0.0027, 0.0027, 0.0028], device='cuda:1'), out_proj_covar=tensor([0.0051, 0.0046, 0.0038, 0.0053, 0.0039, 0.0050, 0.0050, 0.0052], device='cuda:1') 2023-04-27 22:23:13,442 INFO [finetune.py:976] (1/7) Epoch 25, batch 3600, loss[loss=0.2067, simple_loss=0.2727, pruned_loss=0.07039, over 4831.00 frames. ], tot_loss[loss=0.1671, simple_loss=0.2383, pruned_loss=0.04798, over 956610.53 frames. ], batch size: 33, lr: 3.00e-03, grad_scale: 16.0 2023-04-27 22:23:16,477 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.061e+02 1.519e+02 1.759e+02 2.110e+02 6.340e+02, threshold=3.519e+02, percent-clipped=2.0 2023-04-27 22:23:27,746 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=141079.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:23:46,967 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=141091.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:23:47,643 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.6790, 1.9198, 2.0400, 2.1410, 1.9753, 2.0797, 2.0739, 2.0069], device='cuda:1'), covar=tensor([0.3814, 0.5450, 0.4579, 0.4560, 0.5758, 0.6930, 0.5657, 0.5079], device='cuda:1'), in_proj_covar=tensor([0.0342, 0.0375, 0.0328, 0.0340, 0.0349, 0.0392, 0.0359, 0.0333], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 22:23:55,243 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.2973, 1.4705, 1.3366, 1.7048, 1.6327, 1.9975, 1.3671, 3.3598], device='cuda:1'), covar=tensor([0.0556, 0.0795, 0.0744, 0.1112, 0.0590, 0.0426, 0.0730, 0.0152], device='cuda:1'), in_proj_covar=tensor([0.0037, 0.0039, 0.0039, 0.0042, 0.0040, 0.0038, 0.0038, 0.0055], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0013, 0.0013, 0.0015, 0.0014, 0.0013, 0.0013, 0.0016], device='cuda:1') 2023-04-27 22:23:59,400 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5988, 2.1714, 1.6033, 1.4538, 1.2473, 1.2582, 1.6579, 1.1659], device='cuda:1'), covar=tensor([0.1920, 0.1282, 0.1645, 0.1893, 0.2453, 0.2273, 0.1098, 0.2292], device='cuda:1'), in_proj_covar=tensor([0.0197, 0.0210, 0.0168, 0.0203, 0.0198, 0.0185, 0.0156, 0.0185], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-27 22:24:07,939 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9174, 2.8625, 2.0250, 2.2495, 1.4894, 1.4805, 2.1047, 1.4692], device='cuda:1'), covar=tensor([0.1558, 0.1249, 0.1340, 0.1465, 0.2088, 0.1833, 0.0923, 0.1884], device='cuda:1'), in_proj_covar=tensor([0.0197, 0.0209, 0.0167, 0.0203, 0.0198, 0.0185, 0.0156, 0.0185], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-27 22:24:19,423 INFO [finetune.py:976] (1/7) Epoch 25, batch 3650, loss[loss=0.1505, simple_loss=0.2304, pruned_loss=0.03528, over 4798.00 frames. ], tot_loss[loss=0.1691, simple_loss=0.2406, pruned_loss=0.04882, over 956318.16 frames. ], batch size: 29, lr: 3.00e-03, grad_scale: 16.0 2023-04-27 22:24:30,758 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.3100, 3.2382, 0.8008, 1.6317, 1.5363, 2.2772, 1.7736, 1.0454], device='cuda:1'), covar=tensor([0.1834, 0.1475, 0.2473, 0.1808, 0.1350, 0.1399, 0.1752, 0.2443], device='cuda:1'), in_proj_covar=tensor([0.0118, 0.0241, 0.0137, 0.0122, 0.0132, 0.0154, 0.0118, 0.0120], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-27 22:24:34,388 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4631, 1.2447, 4.1752, 3.9357, 3.6540, 3.9761, 3.7986, 3.6231], device='cuda:1'), covar=tensor([0.7581, 0.6185, 0.1079, 0.1620, 0.1075, 0.1546, 0.2047, 0.1686], device='cuda:1'), in_proj_covar=tensor([0.0312, 0.0309, 0.0409, 0.0411, 0.0350, 0.0414, 0.0319, 0.0370], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-27 22:24:42,746 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=141143.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:24:43,941 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=141145.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:24:53,966 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.3814, 1.7533, 1.8190, 1.9438, 1.8099, 1.8617, 1.9219, 1.9418], device='cuda:1'), covar=tensor([0.3726, 0.4473, 0.4295, 0.3961, 0.5348, 0.6536, 0.4381, 0.4593], device='cuda:1'), in_proj_covar=tensor([0.0342, 0.0376, 0.0328, 0.0341, 0.0350, 0.0393, 0.0359, 0.0333], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 22:24:57,939 INFO [finetune.py:976] (1/7) Epoch 25, batch 3700, loss[loss=0.1772, simple_loss=0.2591, pruned_loss=0.04768, over 4843.00 frames. ], tot_loss[loss=0.1711, simple_loss=0.2434, pruned_loss=0.04946, over 954671.23 frames. ], batch size: 47, lr: 3.00e-03, grad_scale: 16.0 2023-04-27 22:24:58,014 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=141165.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:24:58,031 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=141165.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:24:58,788 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.43 vs. limit=5.0 2023-04-27 22:25:00,977 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.179e+02 1.634e+02 1.921e+02 2.262e+02 4.366e+02, threshold=3.843e+02, percent-clipped=2.0 2023-04-27 22:25:15,804 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=141193.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:25:22,663 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=141204.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:25:30,029 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=141213.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:25:31,150 INFO [finetune.py:976] (1/7) Epoch 25, batch 3750, loss[loss=0.1824, simple_loss=0.2686, pruned_loss=0.0481, over 4820.00 frames. ], tot_loss[loss=0.1715, simple_loss=0.2442, pruned_loss=0.04939, over 955072.54 frames. ], batch size: 33, lr: 3.00e-03, grad_scale: 16.0 2023-04-27 22:26:11,333 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.2375, 1.4447, 1.7691, 1.8768, 1.7583, 1.8709, 1.7712, 1.8026], device='cuda:1'), covar=tensor([0.3872, 0.5110, 0.4058, 0.4141, 0.5340, 0.6695, 0.4742, 0.4276], device='cuda:1'), in_proj_covar=tensor([0.0341, 0.0374, 0.0327, 0.0340, 0.0349, 0.0392, 0.0359, 0.0333], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 22:26:37,434 INFO [finetune.py:976] (1/7) Epoch 25, batch 3800, loss[loss=0.1684, simple_loss=0.2572, pruned_loss=0.03976, over 4817.00 frames. ], tot_loss[loss=0.1727, simple_loss=0.2458, pruned_loss=0.04979, over 956578.01 frames. ], batch size: 40, lr: 3.00e-03, grad_scale: 16.0 2023-04-27 22:26:45,872 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.008e+02 1.561e+02 1.876e+02 2.402e+02 4.277e+02, threshold=3.753e+02, percent-clipped=1.0 2023-04-27 22:26:47,826 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=141273.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:26:48,459 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8136, 1.6171, 1.7424, 2.1695, 2.1401, 1.7949, 1.3879, 1.9016], device='cuda:1'), covar=tensor([0.0771, 0.1115, 0.0836, 0.0468, 0.0613, 0.0746, 0.0737, 0.0520], device='cuda:1'), in_proj_covar=tensor([0.0182, 0.0200, 0.0184, 0.0170, 0.0176, 0.0175, 0.0148, 0.0176], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 22:27:28,693 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=141303.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:27:42,126 INFO [finetune.py:976] (1/7) Epoch 25, batch 3850, loss[loss=0.1865, simple_loss=0.2519, pruned_loss=0.06056, over 4853.00 frames. ], tot_loss[loss=0.1704, simple_loss=0.2436, pruned_loss=0.04864, over 955766.15 frames. ], batch size: 44, lr: 3.00e-03, grad_scale: 16.0 2023-04-27 22:28:02,220 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=141329.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:28:27,561 INFO [finetune.py:976] (1/7) Epoch 25, batch 3900, loss[loss=0.1496, simple_loss=0.2143, pruned_loss=0.04249, over 4293.00 frames. ], tot_loss[loss=0.1682, simple_loss=0.2406, pruned_loss=0.04785, over 955130.85 frames. ], batch size: 19, lr: 3.00e-03, grad_scale: 16.0 2023-04-27 22:28:31,461 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.108e+01 1.538e+02 1.758e+02 2.132e+02 4.697e+02, threshold=3.516e+02, percent-clipped=1.0 2023-04-27 22:28:37,517 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=141379.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:28:44,191 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=141390.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:29:00,435 INFO [finetune.py:976] (1/7) Epoch 25, batch 3950, loss[loss=0.1949, simple_loss=0.2583, pruned_loss=0.06578, over 4822.00 frames. ], tot_loss[loss=0.1655, simple_loss=0.2373, pruned_loss=0.04682, over 953583.47 frames. ], batch size: 30, lr: 3.00e-03, grad_scale: 16.0 2023-04-27 22:29:01,779 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=141417.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 22:29:09,299 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=141427.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:29:10,562 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([0.9595, 1.1251, 0.9714, 1.0811, 0.9624, 0.8971, 0.9580, 0.8005], device='cuda:1'), covar=tensor([0.1016, 0.0885, 0.0643, 0.0763, 0.2182, 0.0757, 0.1121, 0.1525], device='cuda:1'), in_proj_covar=tensor([0.0286, 0.0302, 0.0215, 0.0277, 0.0313, 0.0256, 0.0249, 0.0266], device='cuda:1'), out_proj_covar=tensor([1.1442e-04, 1.1917e-04, 8.4462e-05, 1.0901e-04, 1.2628e-04, 1.0105e-04, 1.0036e-04, 1.0500e-04], device='cuda:1') 2023-04-27 22:29:17,856 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=141441.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:29:29,050 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([0.8936, 1.9161, 1.8354, 1.5330, 2.0875, 1.7356, 2.5674, 1.6173], device='cuda:1'), covar=tensor([0.3728, 0.2136, 0.4345, 0.2982, 0.1542, 0.2405, 0.1308, 0.4283], device='cuda:1'), in_proj_covar=tensor([0.0339, 0.0353, 0.0429, 0.0353, 0.0383, 0.0376, 0.0371, 0.0426], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 22:29:33,818 INFO [finetune.py:976] (1/7) Epoch 25, batch 4000, loss[loss=0.1757, simple_loss=0.2523, pruned_loss=0.04951, over 4780.00 frames. ], tot_loss[loss=0.167, simple_loss=0.2385, pruned_loss=0.04773, over 953316.25 frames. ], batch size: 54, lr: 3.00e-03, grad_scale: 16.0 2023-04-27 22:29:33,897 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=141465.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:29:36,895 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.301e+01 1.493e+02 1.709e+02 2.038e+02 4.561e+02, threshold=3.417e+02, percent-clipped=1.0 2023-04-27 22:29:43,320 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=141478.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 22:29:55,971 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=141499.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:29:57,849 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=141502.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:30:05,459 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=141513.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:30:06,624 INFO [finetune.py:976] (1/7) Epoch 25, batch 4050, loss[loss=0.1384, simple_loss=0.2214, pruned_loss=0.02767, over 4753.00 frames. ], tot_loss[loss=0.1682, simple_loss=0.2403, pruned_loss=0.04803, over 953105.09 frames. ], batch size: 26, lr: 3.00e-03, grad_scale: 16.0 2023-04-27 22:30:06,840 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.09 vs. limit=2.0 2023-04-27 22:30:25,958 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5888, 1.1541, 1.3307, 1.1922, 1.6941, 1.3586, 1.1341, 1.2994], device='cuda:1'), covar=tensor([0.1440, 0.1375, 0.1639, 0.1513, 0.0837, 0.1457, 0.1780, 0.1940], device='cuda:1'), in_proj_covar=tensor([0.0312, 0.0306, 0.0350, 0.0286, 0.0327, 0.0304, 0.0298, 0.0371], device='cuda:1'), out_proj_covar=tensor([6.3764e-05, 6.3031e-05, 7.3507e-05, 5.7443e-05, 6.7176e-05, 6.3599e-05, 6.1759e-05, 7.8660e-05], device='cuda:1') 2023-04-27 22:30:34,953 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7206, 2.3721, 1.8571, 1.6699, 1.3115, 1.3492, 1.8911, 1.2610], device='cuda:1'), covar=tensor([0.1850, 0.1314, 0.1533, 0.1858, 0.2488, 0.2079, 0.1035, 0.2201], device='cuda:1'), in_proj_covar=tensor([0.0197, 0.0209, 0.0167, 0.0203, 0.0198, 0.0185, 0.0156, 0.0187], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-27 22:30:36,167 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2023-04-27 22:30:39,499 INFO [finetune.py:976] (1/7) Epoch 25, batch 4100, loss[loss=0.1877, simple_loss=0.2637, pruned_loss=0.05588, over 4895.00 frames. ], tot_loss[loss=0.1707, simple_loss=0.2434, pruned_loss=0.04902, over 953783.80 frames. ], batch size: 43, lr: 3.00e-03, grad_scale: 16.0 2023-04-27 22:30:40,820 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.2010, 1.6805, 2.1521, 2.6347, 2.1341, 1.6409, 1.4505, 1.9247], device='cuda:1'), covar=tensor([0.3022, 0.2985, 0.1500, 0.1853, 0.2387, 0.2513, 0.3717, 0.1823], device='cuda:1'), in_proj_covar=tensor([0.0290, 0.0243, 0.0226, 0.0311, 0.0219, 0.0232, 0.0225, 0.0184], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-27 22:30:42,488 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.442e+01 1.586e+02 1.949e+02 2.336e+02 4.544e+02, threshold=3.898e+02, percent-clipped=7.0 2023-04-27 22:30:44,329 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=141573.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:31:04,700 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=141603.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:31:17,880 INFO [finetune.py:976] (1/7) Epoch 25, batch 4150, loss[loss=0.1793, simple_loss=0.2533, pruned_loss=0.05261, over 4812.00 frames. ], tot_loss[loss=0.1713, simple_loss=0.2442, pruned_loss=0.04917, over 954625.14 frames. ], batch size: 40, lr: 3.00e-03, grad_scale: 16.0 2023-04-27 22:31:20,600 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 2023-04-27 22:31:21,617 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=141621.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:31:51,175 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.4000, 1.6875, 1.7872, 1.8929, 1.7522, 1.8176, 1.8523, 1.8002], device='cuda:1'), covar=tensor([0.3558, 0.4801, 0.4058, 0.4098, 0.5348, 0.6625, 0.4851, 0.4438], device='cuda:1'), in_proj_covar=tensor([0.0341, 0.0374, 0.0327, 0.0339, 0.0348, 0.0393, 0.0359, 0.0331], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 22:32:02,966 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=141651.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:32:23,211 INFO [finetune.py:976] (1/7) Epoch 25, batch 4200, loss[loss=0.1935, simple_loss=0.2705, pruned_loss=0.05824, over 4749.00 frames. ], tot_loss[loss=0.1719, simple_loss=0.2452, pruned_loss=0.04936, over 953881.78 frames. ], batch size: 54, lr: 3.00e-03, grad_scale: 16.0 2023-04-27 22:32:23,337 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.7113, 2.5314, 2.7708, 3.3996, 3.0477, 2.7529, 2.3971, 3.0279], device='cuda:1'), covar=tensor([0.0892, 0.0974, 0.0631, 0.0500, 0.0668, 0.0855, 0.0700, 0.0493], device='cuda:1'), in_proj_covar=tensor([0.0185, 0.0202, 0.0185, 0.0172, 0.0178, 0.0177, 0.0150, 0.0178], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 22:32:26,264 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.107e+02 1.638e+02 1.894e+02 2.201e+02 5.051e+02, threshold=3.789e+02, percent-clipped=1.0 2023-04-27 22:32:47,255 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=141685.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:33:28,597 INFO [finetune.py:976] (1/7) Epoch 25, batch 4250, loss[loss=0.1906, simple_loss=0.2532, pruned_loss=0.06399, over 4816.00 frames. ], tot_loss[loss=0.1695, simple_loss=0.2421, pruned_loss=0.04841, over 955029.66 frames. ], batch size: 39, lr: 3.00e-03, grad_scale: 16.0 2023-04-27 22:34:03,905 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.8411, 2.1828, 2.1119, 2.2339, 2.0818, 2.1461, 2.1618, 2.1049], device='cuda:1'), covar=tensor([0.4079, 0.6359, 0.4735, 0.4760, 0.5681, 0.6770, 0.6173, 0.5434], device='cuda:1'), in_proj_covar=tensor([0.0343, 0.0375, 0.0328, 0.0341, 0.0349, 0.0394, 0.0360, 0.0333], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 22:34:18,637 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.61 vs. limit=5.0 2023-04-27 22:34:30,331 INFO [finetune.py:976] (1/7) Epoch 25, batch 4300, loss[loss=0.1564, simple_loss=0.2343, pruned_loss=0.03928, over 4765.00 frames. ], tot_loss[loss=0.1676, simple_loss=0.2398, pruned_loss=0.04773, over 956524.97 frames. ], batch size: 27, lr: 2.99e-03, grad_scale: 16.0 2023-04-27 22:34:39,438 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.343e+01 1.507e+02 1.747e+02 2.193e+02 4.425e+02, threshold=3.494e+02, percent-clipped=2.0 2023-04-27 22:34:41,392 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=141773.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 22:35:15,329 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=141797.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:35:16,568 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=141799.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:35:35,615 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([3.3829, 3.3501, 2.5300, 3.9296, 3.4042, 3.3942, 1.3905, 3.3594], device='cuda:1'), covar=tensor([0.2011, 0.1366, 0.2883, 0.2192, 0.3369, 0.2186, 0.6142, 0.2491], device='cuda:1'), in_proj_covar=tensor([0.0246, 0.0219, 0.0253, 0.0306, 0.0300, 0.0250, 0.0275, 0.0275], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 22:35:36,800 INFO [finetune.py:976] (1/7) Epoch 25, batch 4350, loss[loss=0.1358, simple_loss=0.2112, pruned_loss=0.03023, over 4772.00 frames. ], tot_loss[loss=0.1641, simple_loss=0.2361, pruned_loss=0.04611, over 956088.59 frames. ], batch size: 28, lr: 2.99e-03, grad_scale: 16.0 2023-04-27 22:35:46,791 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4951, 1.2845, 4.2252, 3.9130, 3.6613, 4.0655, 3.9470, 3.7146], device='cuda:1'), covar=tensor([0.7622, 0.6157, 0.1224, 0.1921, 0.1355, 0.1737, 0.1758, 0.1570], device='cuda:1'), in_proj_covar=tensor([0.0315, 0.0311, 0.0412, 0.0414, 0.0352, 0.0416, 0.0322, 0.0372], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-27 22:36:20,221 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=141847.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:36:42,215 INFO [finetune.py:976] (1/7) Epoch 25, batch 4400, loss[loss=0.1992, simple_loss=0.2796, pruned_loss=0.05934, over 4804.00 frames. ], tot_loss[loss=0.1663, simple_loss=0.2379, pruned_loss=0.04734, over 955044.35 frames. ], batch size: 45, lr: 2.99e-03, grad_scale: 16.0 2023-04-27 22:36:50,565 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.176e+02 1.541e+02 1.847e+02 2.232e+02 3.824e+02, threshold=3.693e+02, percent-clipped=5.0 2023-04-27 22:37:16,650 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.6232, 2.4295, 2.7614, 3.1868, 3.0595, 2.5765, 2.2727, 2.7925], device='cuda:1'), covar=tensor([0.1009, 0.1033, 0.0603, 0.0621, 0.0613, 0.0913, 0.0767, 0.0595], device='cuda:1'), in_proj_covar=tensor([0.0184, 0.0202, 0.0185, 0.0172, 0.0178, 0.0177, 0.0150, 0.0177], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 22:37:46,060 INFO [finetune.py:976] (1/7) Epoch 25, batch 4450, loss[loss=0.1643, simple_loss=0.237, pruned_loss=0.04582, over 4752.00 frames. ], tot_loss[loss=0.1691, simple_loss=0.2414, pruned_loss=0.04843, over 956653.08 frames. ], batch size: 28, lr: 2.99e-03, grad_scale: 32.0 2023-04-27 22:37:46,786 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=141916.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:38:04,869 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=141928.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:38:51,125 INFO [finetune.py:976] (1/7) Epoch 25, batch 4500, loss[loss=0.1757, simple_loss=0.2475, pruned_loss=0.05194, over 4918.00 frames. ], tot_loss[loss=0.1695, simple_loss=0.242, pruned_loss=0.0485, over 955041.32 frames. ], batch size: 42, lr: 2.99e-03, grad_scale: 32.0 2023-04-27 22:38:59,405 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.003e+02 1.537e+02 1.854e+02 2.228e+02 3.851e+02, threshold=3.709e+02, percent-clipped=1.0 2023-04-27 22:39:09,407 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=141977.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:39:20,816 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([3.9808, 3.8491, 3.0810, 4.5792, 3.7631, 3.9934, 1.8023, 3.9964], device='cuda:1'), covar=tensor([0.1625, 0.1163, 0.3941, 0.1045, 0.3004, 0.1760, 0.5069, 0.1907], device='cuda:1'), in_proj_covar=tensor([0.0247, 0.0220, 0.0253, 0.0307, 0.0301, 0.0251, 0.0276, 0.0275], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 22:39:21,482 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=141985.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:39:23,961 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=141989.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:39:57,194 INFO [finetune.py:976] (1/7) Epoch 25, batch 4550, loss[loss=0.1507, simple_loss=0.2262, pruned_loss=0.03757, over 4742.00 frames. ], tot_loss[loss=0.17, simple_loss=0.2433, pruned_loss=0.04832, over 955652.10 frames. ], batch size: 59, lr: 2.99e-03, grad_scale: 32.0 2023-04-27 22:40:19,391 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=142033.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:41:01,791 INFO [finetune.py:976] (1/7) Epoch 25, batch 4600, loss[loss=0.154, simple_loss=0.2352, pruned_loss=0.03641, over 4864.00 frames. ], tot_loss[loss=0.1688, simple_loss=0.2422, pruned_loss=0.04771, over 955413.34 frames. ], batch size: 34, lr: 2.99e-03, grad_scale: 32.0 2023-04-27 22:41:10,130 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.365e+01 1.636e+02 1.871e+02 2.331e+02 4.472e+02, threshold=3.743e+02, percent-clipped=1.0 2023-04-27 22:41:10,958 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=3.99 vs. limit=5.0 2023-04-27 22:41:12,041 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=142073.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 22:41:43,197 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=142097.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:42:06,628 INFO [finetune.py:976] (1/7) Epoch 25, batch 4650, loss[loss=0.1665, simple_loss=0.2291, pruned_loss=0.052, over 4253.00 frames. ], tot_loss[loss=0.1674, simple_loss=0.2401, pruned_loss=0.0473, over 955261.04 frames. ], batch size: 65, lr: 2.99e-03, grad_scale: 32.0 2023-04-27 22:42:15,791 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=142121.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 22:42:28,643 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5894, 1.2272, 4.0413, 3.7789, 3.4992, 3.7507, 3.6129, 3.5763], device='cuda:1'), covar=tensor([0.7747, 0.6056, 0.1083, 0.1802, 0.1212, 0.1700, 0.3074, 0.1671], device='cuda:1'), in_proj_covar=tensor([0.0313, 0.0308, 0.0408, 0.0411, 0.0350, 0.0414, 0.0319, 0.0368], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-27 22:42:47,435 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=142145.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:43:11,598 INFO [finetune.py:976] (1/7) Epoch 25, batch 4700, loss[loss=0.1501, simple_loss=0.2141, pruned_loss=0.04304, over 4870.00 frames. ], tot_loss[loss=0.1653, simple_loss=0.2375, pruned_loss=0.0466, over 955536.97 frames. ], batch size: 31, lr: 2.99e-03, grad_scale: 32.0 2023-04-27 22:43:19,816 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.009e+02 1.574e+02 1.870e+02 2.251e+02 4.397e+02, threshold=3.741e+02, percent-clipped=1.0 2023-04-27 22:44:17,215 INFO [finetune.py:976] (1/7) Epoch 25, batch 4750, loss[loss=0.1626, simple_loss=0.2322, pruned_loss=0.04646, over 4933.00 frames. ], tot_loss[loss=0.165, simple_loss=0.2363, pruned_loss=0.04682, over 953348.65 frames. ], batch size: 33, lr: 2.99e-03, grad_scale: 32.0 2023-04-27 22:44:37,102 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 2023-04-27 22:44:38,833 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.2981, 1.2872, 1.5733, 1.5505, 1.2258, 1.1855, 1.3023, 0.7195], device='cuda:1'), covar=tensor([0.0518, 0.0604, 0.0366, 0.0498, 0.0701, 0.1038, 0.0470, 0.0585], device='cuda:1'), in_proj_covar=tensor([0.0070, 0.0068, 0.0067, 0.0068, 0.0074, 0.0095, 0.0073, 0.0064], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-27 22:45:18,476 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-04-27 22:45:29,868 INFO [finetune.py:976] (1/7) Epoch 25, batch 4800, loss[loss=0.1684, simple_loss=0.225, pruned_loss=0.05586, over 4466.00 frames. ], tot_loss[loss=0.1673, simple_loss=0.2392, pruned_loss=0.04774, over 954335.25 frames. ], batch size: 19, lr: 2.99e-03, grad_scale: 32.0 2023-04-27 22:45:34,011 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.041e+02 1.626e+02 1.868e+02 2.200e+02 5.189e+02, threshold=3.736e+02, percent-clipped=2.0 2023-04-27 22:45:40,840 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=142272.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:45:54,145 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=142284.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:46:36,610 INFO [finetune.py:976] (1/7) Epoch 25, batch 4850, loss[loss=0.1892, simple_loss=0.2594, pruned_loss=0.05948, over 4866.00 frames. ], tot_loss[loss=0.1702, simple_loss=0.243, pruned_loss=0.04874, over 955255.05 frames. ], batch size: 34, lr: 2.99e-03, grad_scale: 32.0 2023-04-27 22:47:42,702 INFO [finetune.py:976] (1/7) Epoch 25, batch 4900, loss[loss=0.2089, simple_loss=0.2731, pruned_loss=0.07241, over 4823.00 frames. ], tot_loss[loss=0.1709, simple_loss=0.2437, pruned_loss=0.04899, over 954487.47 frames. ], batch size: 49, lr: 2.99e-03, grad_scale: 32.0 2023-04-27 22:47:51,886 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.944e+01 1.647e+02 1.933e+02 2.315e+02 7.407e+02, threshold=3.866e+02, percent-clipped=3.0 2023-04-27 22:48:49,016 INFO [finetune.py:976] (1/7) Epoch 25, batch 4950, loss[loss=0.1738, simple_loss=0.2493, pruned_loss=0.04914, over 4888.00 frames. ], tot_loss[loss=0.1714, simple_loss=0.2448, pruned_loss=0.04905, over 954398.06 frames. ], batch size: 43, lr: 2.99e-03, grad_scale: 32.0 2023-04-27 22:49:57,953 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.1851, 1.6682, 2.1316, 2.7232, 2.0803, 1.6672, 1.5204, 1.9493], device='cuda:1'), covar=tensor([0.3205, 0.3270, 0.1666, 0.2083, 0.2664, 0.2719, 0.4191, 0.1961], device='cuda:1'), in_proj_covar=tensor([0.0291, 0.0245, 0.0227, 0.0313, 0.0221, 0.0234, 0.0226, 0.0186], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-27 22:49:58,431 INFO [finetune.py:976] (1/7) Epoch 25, batch 5000, loss[loss=0.1495, simple_loss=0.2332, pruned_loss=0.03285, over 4850.00 frames. ], tot_loss[loss=0.1696, simple_loss=0.2429, pruned_loss=0.04812, over 956013.21 frames. ], batch size: 44, lr: 2.99e-03, grad_scale: 32.0 2023-04-27 22:50:01,486 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.151e+02 1.520e+02 1.784e+02 2.172e+02 4.625e+02, threshold=3.567e+02, percent-clipped=1.0 2023-04-27 22:50:13,337 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.3035, 2.2071, 2.5225, 2.8636, 2.7989, 2.1333, 1.9911, 2.4652], device='cuda:1'), covar=tensor([0.0898, 0.1041, 0.0656, 0.0593, 0.0654, 0.0890, 0.0774, 0.0635], device='cuda:1'), in_proj_covar=tensor([0.0186, 0.0203, 0.0186, 0.0172, 0.0178, 0.0179, 0.0151, 0.0178], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 22:50:45,481 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.5169, 1.8148, 1.8842, 1.9484, 1.8289, 1.8785, 1.9331, 1.9436], device='cuda:1'), covar=tensor([0.3625, 0.5016, 0.4039, 0.4111, 0.5436, 0.6581, 0.4708, 0.4463], device='cuda:1'), in_proj_covar=tensor([0.0340, 0.0375, 0.0327, 0.0340, 0.0348, 0.0392, 0.0360, 0.0332], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 22:51:03,145 INFO [finetune.py:976] (1/7) Epoch 25, batch 5050, loss[loss=0.1518, simple_loss=0.2264, pruned_loss=0.03855, over 4896.00 frames. ], tot_loss[loss=0.1678, simple_loss=0.2408, pruned_loss=0.04737, over 955967.71 frames. ], batch size: 35, lr: 2.99e-03, grad_scale: 32.0 2023-04-27 22:51:04,456 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.3220, 1.2448, 3.7744, 3.2807, 3.3993, 3.4827, 3.5578, 3.1878], device='cuda:1'), covar=tensor([0.9936, 0.8779, 0.2107, 0.4155, 0.2181, 0.3244, 0.2569, 0.3442], device='cuda:1'), in_proj_covar=tensor([0.0312, 0.0307, 0.0407, 0.0409, 0.0350, 0.0414, 0.0318, 0.0368], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-27 22:51:28,030 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=5.05 vs. limit=5.0 2023-04-27 22:51:48,460 INFO [finetune.py:976] (1/7) Epoch 25, batch 5100, loss[loss=0.153, simple_loss=0.2217, pruned_loss=0.04216, over 4913.00 frames. ], tot_loss[loss=0.1648, simple_loss=0.2372, pruned_loss=0.04621, over 955939.80 frames. ], batch size: 32, lr: 2.99e-03, grad_scale: 32.0 2023-04-27 22:51:51,977 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.162e+02 1.504e+02 1.862e+02 2.404e+02 6.312e+02, threshold=3.723e+02, percent-clipped=2.0 2023-04-27 22:51:53,295 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=142572.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:51:53,944 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=142573.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:52:01,129 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=142584.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:52:21,630 INFO [finetune.py:976] (1/7) Epoch 25, batch 5150, loss[loss=0.1671, simple_loss=0.2373, pruned_loss=0.04843, over 4801.00 frames. ], tot_loss[loss=0.1649, simple_loss=0.2372, pruned_loss=0.04635, over 956626.71 frames. ], batch size: 45, lr: 2.99e-03, grad_scale: 32.0 2023-04-27 22:52:24,036 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 2023-04-27 22:52:26,240 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=142620.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:52:33,586 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=142632.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:52:35,334 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=142634.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:52:43,065 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7860, 1.7662, 2.1868, 2.3571, 1.6083, 1.4757, 1.7704, 0.9291], device='cuda:1'), covar=tensor([0.0571, 0.0609, 0.0415, 0.0665, 0.0799, 0.1080, 0.0701, 0.0724], device='cuda:1'), in_proj_covar=tensor([0.0071, 0.0068, 0.0067, 0.0069, 0.0075, 0.0095, 0.0073, 0.0064], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-27 22:52:56,102 INFO [finetune.py:976] (1/7) Epoch 25, batch 5200, loss[loss=0.1774, simple_loss=0.2465, pruned_loss=0.05419, over 4816.00 frames. ], tot_loss[loss=0.1673, simple_loss=0.2403, pruned_loss=0.04717, over 955658.70 frames. ], batch size: 38, lr: 2.99e-03, grad_scale: 32.0 2023-04-27 22:53:00,175 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.071e+02 1.581e+02 1.903e+02 2.348e+02 3.515e+02, threshold=3.805e+02, percent-clipped=0.0 2023-04-27 22:53:29,657 INFO [finetune.py:976] (1/7) Epoch 25, batch 5250, loss[loss=0.1977, simple_loss=0.2622, pruned_loss=0.06664, over 4185.00 frames. ], tot_loss[loss=0.1691, simple_loss=0.243, pruned_loss=0.0476, over 956176.37 frames. ], batch size: 65, lr: 2.99e-03, grad_scale: 32.0 2023-04-27 22:53:36,262 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=142724.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:54:19,567 INFO [finetune.py:976] (1/7) Epoch 25, batch 5300, loss[loss=0.202, simple_loss=0.2723, pruned_loss=0.06589, over 4804.00 frames. ], tot_loss[loss=0.1708, simple_loss=0.2445, pruned_loss=0.04851, over 955779.10 frames. ], batch size: 40, lr: 2.99e-03, grad_scale: 32.0 2023-04-27 22:54:22,608 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.008e+02 1.572e+02 1.803e+02 2.243e+02 5.104e+02, threshold=3.606e+02, percent-clipped=2.0 2023-04-27 22:54:44,816 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=142785.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:55:27,673 INFO [finetune.py:976] (1/7) Epoch 25, batch 5350, loss[loss=0.1208, simple_loss=0.2012, pruned_loss=0.02023, over 4748.00 frames. ], tot_loss[loss=0.1699, simple_loss=0.244, pruned_loss=0.04787, over 955019.27 frames. ], batch size: 54, lr: 2.99e-03, grad_scale: 32.0 2023-04-27 22:55:51,898 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.2121, 2.2188, 1.9233, 1.8439, 2.3968, 2.0137, 2.8888, 1.7402], device='cuda:1'), covar=tensor([0.4267, 0.2023, 0.5366, 0.3840, 0.1847, 0.2571, 0.1467, 0.4984], device='cuda:1'), in_proj_covar=tensor([0.0344, 0.0358, 0.0432, 0.0357, 0.0388, 0.0380, 0.0375, 0.0431], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 22:56:34,014 INFO [finetune.py:976] (1/7) Epoch 25, batch 5400, loss[loss=0.1495, simple_loss=0.225, pruned_loss=0.03703, over 4910.00 frames. ], tot_loss[loss=0.1664, simple_loss=0.2402, pruned_loss=0.04624, over 956445.33 frames. ], batch size: 37, lr: 2.99e-03, grad_scale: 32.0 2023-04-27 22:56:42,414 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.018e+02 1.482e+02 1.784e+02 2.095e+02 4.679e+02, threshold=3.568e+02, percent-clipped=3.0 2023-04-27 22:57:18,643 INFO [finetune.py:976] (1/7) Epoch 25, batch 5450, loss[loss=0.1684, simple_loss=0.2334, pruned_loss=0.05174, over 4911.00 frames. ], tot_loss[loss=0.1649, simple_loss=0.2377, pruned_loss=0.04599, over 954833.99 frames. ], batch size: 32, lr: 2.99e-03, grad_scale: 32.0 2023-04-27 22:57:27,232 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=142929.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:57:33,074 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.0956, 0.7313, 0.9474, 0.7873, 1.1819, 1.0187, 0.8646, 0.9709], device='cuda:1'), covar=tensor([0.1864, 0.1666, 0.2063, 0.1726, 0.1116, 0.1661, 0.1826, 0.2580], device='cuda:1'), in_proj_covar=tensor([0.0312, 0.0306, 0.0350, 0.0284, 0.0326, 0.0304, 0.0297, 0.0372], device='cuda:1'), out_proj_covar=tensor([6.3780e-05, 6.2885e-05, 7.3465e-05, 5.6896e-05, 6.6921e-05, 6.3552e-05, 6.1459e-05, 7.8755e-05], device='cuda:1') 2023-04-27 22:57:36,870 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.72 vs. limit=2.0 2023-04-27 22:57:48,061 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.80 vs. limit=2.0 2023-04-27 22:57:52,213 INFO [finetune.py:976] (1/7) Epoch 25, batch 5500, loss[loss=0.1515, simple_loss=0.215, pruned_loss=0.04394, over 4809.00 frames. ], tot_loss[loss=0.1634, simple_loss=0.2356, pruned_loss=0.04557, over 956804.50 frames. ], batch size: 25, lr: 2.99e-03, grad_scale: 32.0 2023-04-27 22:57:55,647 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.918e+01 1.380e+02 1.724e+02 2.155e+02 4.005e+02, threshold=3.448e+02, percent-clipped=1.0 2023-04-27 22:58:03,813 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2023-04-27 22:58:26,147 INFO [finetune.py:976] (1/7) Epoch 25, batch 5550, loss[loss=0.1599, simple_loss=0.2431, pruned_loss=0.03835, over 4730.00 frames. ], tot_loss[loss=0.1641, simple_loss=0.2365, pruned_loss=0.04585, over 957043.64 frames. ], batch size: 54, lr: 2.99e-03, grad_scale: 32.0 2023-04-27 22:58:37,211 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4554, 1.2125, 4.1571, 3.8870, 3.6574, 3.9844, 3.9029, 3.6443], device='cuda:1'), covar=tensor([0.7212, 0.6072, 0.1093, 0.1756, 0.1145, 0.2124, 0.1782, 0.1709], device='cuda:1'), in_proj_covar=tensor([0.0311, 0.0308, 0.0408, 0.0409, 0.0348, 0.0414, 0.0318, 0.0367], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-27 22:58:37,360 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.10 vs. limit=2.0 2023-04-27 22:58:57,671 INFO [finetune.py:976] (1/7) Epoch 25, batch 5600, loss[loss=0.1662, simple_loss=0.2423, pruned_loss=0.04504, over 4902.00 frames. ], tot_loss[loss=0.1672, simple_loss=0.2404, pruned_loss=0.04702, over 956129.78 frames. ], batch size: 35, lr: 2.99e-03, grad_scale: 32.0 2023-04-27 22:58:57,775 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0942, 2.7470, 2.0293, 2.1050, 1.5745, 1.5262, 2.1603, 1.4372], device='cuda:1'), covar=tensor([0.1568, 0.1311, 0.1274, 0.1547, 0.2126, 0.1710, 0.0932, 0.1927], device='cuda:1'), in_proj_covar=tensor([0.0196, 0.0208, 0.0167, 0.0202, 0.0198, 0.0184, 0.0156, 0.0186], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-27 22:58:58,318 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.3793, 2.9978, 1.0527, 1.8868, 2.3282, 1.5904, 4.2282, 2.2837], device='cuda:1'), covar=tensor([0.0701, 0.0760, 0.0908, 0.1273, 0.0543, 0.1042, 0.0394, 0.0561], device='cuda:1'), in_proj_covar=tensor([0.0051, 0.0064, 0.0047, 0.0046, 0.0049, 0.0051, 0.0073, 0.0051], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:1') 2023-04-27 22:59:00,542 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.011e+02 1.590e+02 1.837e+02 2.169e+02 3.781e+02, threshold=3.675e+02, percent-clipped=1.0 2023-04-27 22:59:06,453 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=143080.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:59:19,185 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.2365, 1.4029, 1.2933, 1.6112, 1.5157, 1.6880, 1.2838, 3.5047], device='cuda:1'), covar=tensor([0.0619, 0.0849, 0.0854, 0.1307, 0.0681, 0.0636, 0.0802, 0.0106], device='cuda:1'), in_proj_covar=tensor([0.0037, 0.0039, 0.0039, 0.0043, 0.0040, 0.0038, 0.0038, 0.0055], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0013, 0.0013, 0.0015, 0.0014, 0.0013, 0.0013, 0.0016], device='cuda:1') 2023-04-27 22:59:26,515 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([5.1167, 5.0536, 3.5969, 5.8219, 5.0906, 5.0300, 2.6105, 5.0915], device='cuda:1'), covar=tensor([0.1461, 0.0907, 0.2559, 0.0770, 0.2726, 0.1709, 0.4783, 0.1938], device='cuda:1'), in_proj_covar=tensor([0.0243, 0.0218, 0.0250, 0.0304, 0.0297, 0.0247, 0.0272, 0.0273], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 22:59:27,655 INFO [finetune.py:976] (1/7) Epoch 25, batch 5650, loss[loss=0.1137, simple_loss=0.1908, pruned_loss=0.01834, over 4519.00 frames. ], tot_loss[loss=0.1694, simple_loss=0.2432, pruned_loss=0.04774, over 955024.32 frames. ], batch size: 20, lr: 2.99e-03, grad_scale: 32.0 2023-04-27 23:00:14,600 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.7048, 2.6515, 2.1344, 3.0567, 2.6930, 2.7027, 1.1762, 2.5497], device='cuda:1'), covar=tensor([0.2122, 0.1649, 0.2797, 0.2492, 0.2806, 0.2084, 0.5514, 0.2800], device='cuda:1'), in_proj_covar=tensor([0.0243, 0.0218, 0.0250, 0.0304, 0.0297, 0.0247, 0.0272, 0.0273], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 23:00:24,805 INFO [finetune.py:976] (1/7) Epoch 25, batch 5700, loss[loss=0.1446, simple_loss=0.2186, pruned_loss=0.03533, over 4377.00 frames. ], tot_loss[loss=0.1672, simple_loss=0.2396, pruned_loss=0.04737, over 936485.49 frames. ], batch size: 19, lr: 2.98e-03, grad_scale: 32.0 2023-04-27 23:00:27,756 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 8.938e+01 1.487e+02 1.759e+02 2.216e+02 4.830e+02, threshold=3.518e+02, percent-clipped=2.0 2023-04-27 23:01:04,602 INFO [finetune.py:976] (1/7) Epoch 26, batch 0, loss[loss=0.1558, simple_loss=0.2342, pruned_loss=0.03868, over 4762.00 frames. ], tot_loss[loss=0.1558, simple_loss=0.2342, pruned_loss=0.03868, over 4762.00 frames. ], batch size: 28, lr: 2.98e-03, grad_scale: 32.0 2023-04-27 23:01:04,602 INFO [finetune.py:1001] (1/7) Computing validation loss 2023-04-27 23:01:10,409 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.3172, 1.6211, 1.4946, 1.8500, 1.7161, 1.7628, 1.4905, 3.1118], device='cuda:1'), covar=tensor([0.0542, 0.0783, 0.0741, 0.1150, 0.0589, 0.0448, 0.0668, 0.0175], device='cuda:1'), in_proj_covar=tensor([0.0037, 0.0038, 0.0039, 0.0042, 0.0040, 0.0037, 0.0038, 0.0055], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0013, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:1') 2023-04-27 23:01:11,505 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7532, 1.6743, 1.8748, 2.1588, 2.1787, 1.7121, 1.3970, 1.9071], device='cuda:1'), covar=tensor([0.0905, 0.1241, 0.0795, 0.0587, 0.0593, 0.0936, 0.0777, 0.0589], device='cuda:1'), in_proj_covar=tensor([0.0183, 0.0201, 0.0184, 0.0170, 0.0175, 0.0176, 0.0149, 0.0176], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 23:01:12,604 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([0.7911, 1.9196, 1.7314, 1.5701, 1.9771, 1.6308, 2.3164, 1.5628], device='cuda:1'), covar=tensor([0.4037, 0.1836, 0.5298, 0.2711, 0.1442, 0.2314, 0.1560, 0.4753], device='cuda:1'), in_proj_covar=tensor([0.0337, 0.0351, 0.0423, 0.0349, 0.0380, 0.0373, 0.0368, 0.0423], device='cuda:1'), out_proj_covar=tensor([9.9582e-05, 1.0454e-04, 1.2819e-04, 1.0477e-04, 1.1284e-04, 1.1102e-04, 1.0775e-04, 1.2729e-04], device='cuda:1') 2023-04-27 23:01:26,498 INFO [finetune.py:1010] (1/7) Epoch 26, validation: loss=0.1543, simple_loss=0.2237, pruned_loss=0.04251, over 2265189.00 frames. 2023-04-27 23:01:26,499 INFO [finetune.py:1011] (1/7) Maximum memory allocated so far is 6391MB 2023-04-27 23:01:41,364 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9067, 2.5238, 1.9159, 1.9828, 1.4309, 1.4132, 2.0203, 1.3486], device='cuda:1'), covar=tensor([0.1535, 0.1248, 0.1241, 0.1528, 0.2106, 0.1865, 0.0898, 0.1978], device='cuda:1'), in_proj_covar=tensor([0.0196, 0.0208, 0.0167, 0.0203, 0.0199, 0.0184, 0.0156, 0.0186], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-27 23:02:03,859 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8848, 1.7549, 2.1197, 2.0601, 1.9081, 1.8090, 1.9649, 1.9729], device='cuda:1'), covar=tensor([0.5938, 0.8742, 0.9194, 0.8283, 0.7865, 1.1613, 1.1041, 1.1897], device='cuda:1'), in_proj_covar=tensor([0.0440, 0.0420, 0.0514, 0.0507, 0.0468, 0.0502, 0.0505, 0.0517], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 23:02:16,887 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=143229.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:02:34,933 INFO [finetune.py:976] (1/7) Epoch 26, batch 50, loss[loss=0.1424, simple_loss=0.2226, pruned_loss=0.0311, over 4805.00 frames. ], tot_loss[loss=0.1653, simple_loss=0.2396, pruned_loss=0.0455, over 215079.24 frames. ], batch size: 33, lr: 2.98e-03, grad_scale: 32.0 2023-04-27 23:03:09,820 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 8.418e+01 1.474e+02 1.785e+02 2.280e+02 3.483e+02, threshold=3.571e+02, percent-clipped=0.0 2023-04-27 23:03:19,705 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=143277.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:03:31,435 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.5277, 2.0385, 2.4532, 3.0270, 2.4511, 1.9624, 1.9261, 2.3467], device='cuda:1'), covar=tensor([0.2783, 0.2897, 0.1420, 0.1963, 0.2547, 0.2396, 0.3523, 0.1900], device='cuda:1'), in_proj_covar=tensor([0.0293, 0.0247, 0.0229, 0.0315, 0.0221, 0.0235, 0.0228, 0.0188], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-27 23:03:40,350 INFO [finetune.py:976] (1/7) Epoch 26, batch 100, loss[loss=0.179, simple_loss=0.2456, pruned_loss=0.05625, over 4770.00 frames. ], tot_loss[loss=0.1618, simple_loss=0.2339, pruned_loss=0.04481, over 379699.17 frames. ], batch size: 27, lr: 2.98e-03, grad_scale: 32.0 2023-04-27 23:04:19,085 INFO [finetune.py:976] (1/7) Epoch 26, batch 150, loss[loss=0.1388, simple_loss=0.2049, pruned_loss=0.03641, over 4832.00 frames. ], tot_loss[loss=0.1586, simple_loss=0.2292, pruned_loss=0.04394, over 508940.61 frames. ], batch size: 47, lr: 2.98e-03, grad_scale: 32.0 2023-04-27 23:04:31,061 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([4.3238, 4.1600, 3.1773, 4.9745, 4.2368, 4.3335, 1.7369, 4.2982], device='cuda:1'), covar=tensor([0.1470, 0.1030, 0.3414, 0.0941, 0.3646, 0.1590, 0.5482, 0.1854], device='cuda:1'), in_proj_covar=tensor([0.0244, 0.0218, 0.0251, 0.0304, 0.0298, 0.0248, 0.0273, 0.0273], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 23:04:37,628 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.031e+02 1.476e+02 1.692e+02 2.051e+02 5.029e+02, threshold=3.384e+02, percent-clipped=1.0 2023-04-27 23:04:43,891 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=143380.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:04:52,922 INFO [finetune.py:976] (1/7) Epoch 26, batch 200, loss[loss=0.1718, simple_loss=0.2344, pruned_loss=0.0546, over 4806.00 frames. ], tot_loss[loss=0.1596, simple_loss=0.2302, pruned_loss=0.04448, over 607642.54 frames. ], batch size: 25, lr: 2.98e-03, grad_scale: 32.0 2023-04-27 23:04:59,067 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=143400.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 23:05:16,578 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=143428.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:05:31,680 INFO [finetune.py:976] (1/7) Epoch 26, batch 250, loss[loss=0.1726, simple_loss=0.2341, pruned_loss=0.05559, over 4703.00 frames. ], tot_loss[loss=0.1622, simple_loss=0.2335, pruned_loss=0.04541, over 685409.51 frames. ], batch size: 23, lr: 2.98e-03, grad_scale: 32.0 2023-04-27 23:05:44,905 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=143461.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 23:05:50,285 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.157e+02 1.556e+02 1.824e+02 2.331e+02 6.380e+02, threshold=3.648e+02, percent-clipped=4.0 2023-04-27 23:05:58,120 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=143473.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 23:06:15,190 INFO [finetune.py:976] (1/7) Epoch 26, batch 300, loss[loss=0.18, simple_loss=0.267, pruned_loss=0.04649, over 4850.00 frames. ], tot_loss[loss=0.165, simple_loss=0.2377, pruned_loss=0.0461, over 743637.13 frames. ], batch size: 44, lr: 2.98e-03, grad_scale: 32.0 2023-04-27 23:06:19,509 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-04-27 23:06:43,060 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=143534.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 23:06:48,238 INFO [finetune.py:976] (1/7) Epoch 26, batch 350, loss[loss=0.1938, simple_loss=0.2767, pruned_loss=0.05543, over 4731.00 frames. ], tot_loss[loss=0.1679, simple_loss=0.2415, pruned_loss=0.04711, over 791165.52 frames. ], batch size: 59, lr: 2.98e-03, grad_scale: 32.0 2023-04-27 23:06:55,955 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([0.6602, 1.6135, 1.5684, 1.2969, 1.8002, 1.4737, 2.2562, 1.4435], device='cuda:1'), covar=tensor([0.3898, 0.2082, 0.5494, 0.2834, 0.1757, 0.2325, 0.1527, 0.4774], device='cuda:1'), in_proj_covar=tensor([0.0340, 0.0354, 0.0427, 0.0351, 0.0383, 0.0377, 0.0370, 0.0425], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 23:07:08,173 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.023e+02 1.499e+02 1.727e+02 2.023e+02 3.986e+02, threshold=3.454e+02, percent-clipped=1.0 2023-04-27 23:07:22,105 INFO [finetune.py:976] (1/7) Epoch 26, batch 400, loss[loss=0.1342, simple_loss=0.2104, pruned_loss=0.02898, over 4747.00 frames. ], tot_loss[loss=0.1689, simple_loss=0.2428, pruned_loss=0.04747, over 826157.63 frames. ], batch size: 27, lr: 2.98e-03, grad_scale: 32.0 2023-04-27 23:07:42,371 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.56 vs. limit=5.0 2023-04-27 23:07:55,519 INFO [finetune.py:976] (1/7) Epoch 26, batch 450, loss[loss=0.1374, simple_loss=0.2181, pruned_loss=0.02834, over 4701.00 frames. ], tot_loss[loss=0.1689, simple_loss=0.2426, pruned_loss=0.04755, over 854679.97 frames. ], batch size: 23, lr: 2.98e-03, grad_scale: 32.0 2023-04-27 23:07:58,552 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8924, 1.6821, 1.8758, 2.2236, 2.2948, 1.8737, 1.6113, 2.0824], device='cuda:1'), covar=tensor([0.0801, 0.1131, 0.0757, 0.0606, 0.0579, 0.0802, 0.0734, 0.0532], device='cuda:1'), in_proj_covar=tensor([0.0184, 0.0203, 0.0185, 0.0171, 0.0177, 0.0177, 0.0150, 0.0177], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 23:08:20,552 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.962e+01 1.532e+02 1.796e+02 2.161e+02 5.029e+02, threshold=3.592e+02, percent-clipped=5.0 2023-04-27 23:08:43,967 INFO [finetune.py:976] (1/7) Epoch 26, batch 500, loss[loss=0.1558, simple_loss=0.2431, pruned_loss=0.03427, over 4893.00 frames. ], tot_loss[loss=0.1675, simple_loss=0.2406, pruned_loss=0.04722, over 876053.07 frames. ], batch size: 35, lr: 2.98e-03, grad_scale: 32.0 2023-04-27 23:09:17,388 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.1193, 2.0513, 2.4680, 2.7927, 1.9555, 1.8228, 2.1455, 1.1288], device='cuda:1'), covar=tensor([0.0573, 0.0636, 0.0338, 0.0530, 0.0626, 0.0965, 0.0585, 0.0631], device='cuda:1'), in_proj_covar=tensor([0.0070, 0.0067, 0.0066, 0.0068, 0.0073, 0.0094, 0.0072, 0.0063], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-27 23:09:27,563 INFO [finetune.py:976] (1/7) Epoch 26, batch 550, loss[loss=0.1841, simple_loss=0.2546, pruned_loss=0.0568, over 4811.00 frames. ], tot_loss[loss=0.1645, simple_loss=0.2368, pruned_loss=0.04609, over 895578.61 frames. ], batch size: 45, lr: 2.98e-03, grad_scale: 32.0 2023-04-27 23:09:31,248 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.3346, 1.2130, 3.8519, 3.6074, 3.3966, 3.7072, 3.7272, 3.3734], device='cuda:1'), covar=tensor([0.7345, 0.5992, 0.1283, 0.2006, 0.1143, 0.1969, 0.1305, 0.1775], device='cuda:1'), in_proj_covar=tensor([0.0312, 0.0308, 0.0408, 0.0411, 0.0349, 0.0415, 0.0318, 0.0366], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-27 23:09:37,699 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=143756.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 23:09:47,112 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 8.995e+01 1.539e+02 1.829e+02 2.212e+02 3.034e+02, threshold=3.659e+02, percent-clipped=1.0 2023-04-27 23:10:00,517 INFO [finetune.py:976] (1/7) Epoch 26, batch 600, loss[loss=0.1835, simple_loss=0.259, pruned_loss=0.05399, over 4817.00 frames. ], tot_loss[loss=0.1668, simple_loss=0.2388, pruned_loss=0.04744, over 908087.63 frames. ], batch size: 47, lr: 2.98e-03, grad_scale: 32.0 2023-04-27 23:10:25,722 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=143829.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 23:10:33,615 INFO [finetune.py:976] (1/7) Epoch 26, batch 650, loss[loss=0.168, simple_loss=0.2482, pruned_loss=0.04387, over 4822.00 frames. ], tot_loss[loss=0.1682, simple_loss=0.2405, pruned_loss=0.04789, over 919169.83 frames. ], batch size: 30, lr: 2.98e-03, grad_scale: 32.0 2023-04-27 23:11:14,500 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.206e+02 1.587e+02 1.916e+02 2.318e+02 7.608e+02, threshold=3.833e+02, percent-clipped=3.0 2023-04-27 23:11:40,022 INFO [finetune.py:976] (1/7) Epoch 26, batch 700, loss[loss=0.1588, simple_loss=0.2261, pruned_loss=0.04572, over 4922.00 frames. ], tot_loss[loss=0.1711, simple_loss=0.2434, pruned_loss=0.0494, over 925705.40 frames. ], batch size: 33, lr: 2.98e-03, grad_scale: 64.0 2023-04-27 23:11:41,498 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-04-27 23:12:45,799 INFO [finetune.py:976] (1/7) Epoch 26, batch 750, loss[loss=0.1968, simple_loss=0.2757, pruned_loss=0.05891, over 4828.00 frames. ], tot_loss[loss=0.1707, simple_loss=0.2435, pruned_loss=0.04898, over 932076.95 frames. ], batch size: 47, lr: 2.98e-03, grad_scale: 64.0 2023-04-27 23:13:05,037 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.2032, 1.5950, 2.0479, 2.3855, 2.0876, 1.6169, 1.3251, 1.8527], device='cuda:1'), covar=tensor([0.3349, 0.3308, 0.1797, 0.2372, 0.2429, 0.2830, 0.4179, 0.2033], device='cuda:1'), in_proj_covar=tensor([0.0291, 0.0245, 0.0227, 0.0312, 0.0219, 0.0234, 0.0226, 0.0185], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-27 23:13:26,583 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.075e+02 1.623e+02 1.843e+02 2.196e+02 3.710e+02, threshold=3.686e+02, percent-clipped=0.0 2023-04-27 23:13:56,910 INFO [finetune.py:976] (1/7) Epoch 26, batch 800, loss[loss=0.2058, simple_loss=0.2733, pruned_loss=0.0692, over 4893.00 frames. ], tot_loss[loss=0.1709, simple_loss=0.244, pruned_loss=0.04884, over 937816.49 frames. ], batch size: 43, lr: 2.98e-03, grad_scale: 64.0 2023-04-27 23:13:57,032 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.2816, 1.9329, 2.5515, 2.8477, 2.0072, 1.8300, 2.1449, 1.1370], device='cuda:1'), covar=tensor([0.0504, 0.0812, 0.0350, 0.0446, 0.0624, 0.1040, 0.0688, 0.0757], device='cuda:1'), in_proj_covar=tensor([0.0070, 0.0067, 0.0066, 0.0069, 0.0074, 0.0094, 0.0073, 0.0063], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-27 23:14:52,467 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9557, 1.3330, 1.5712, 2.3238, 2.3412, 1.8420, 1.5711, 1.9487], device='cuda:1'), covar=tensor([0.0808, 0.1798, 0.1156, 0.0562, 0.0593, 0.0877, 0.0831, 0.0663], device='cuda:1'), in_proj_covar=tensor([0.0184, 0.0202, 0.0185, 0.0171, 0.0177, 0.0177, 0.0150, 0.0178], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 23:15:01,729 INFO [finetune.py:976] (1/7) Epoch 26, batch 850, loss[loss=0.1698, simple_loss=0.245, pruned_loss=0.04733, over 4851.00 frames. ], tot_loss[loss=0.1678, simple_loss=0.2405, pruned_loss=0.04751, over 939254.43 frames. ], batch size: 49, lr: 2.98e-03, grad_scale: 64.0 2023-04-27 23:15:15,467 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=144056.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 23:15:35,341 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.021e+02 1.501e+02 1.661e+02 2.180e+02 3.367e+02, threshold=3.322e+02, percent-clipped=0.0 2023-04-27 23:16:05,579 INFO [finetune.py:976] (1/7) Epoch 26, batch 900, loss[loss=0.1989, simple_loss=0.2655, pruned_loss=0.06618, over 4918.00 frames. ], tot_loss[loss=0.166, simple_loss=0.2386, pruned_loss=0.04672, over 944962.53 frames. ], batch size: 36, lr: 2.97e-03, grad_scale: 32.0 2023-04-27 23:16:12,929 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=144104.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 23:16:17,995 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.48 vs. limit=2.0 2023-04-27 23:16:22,990 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8233, 1.6851, 1.9747, 2.0695, 1.6213, 1.4357, 1.7346, 1.1078], device='cuda:1'), covar=tensor([0.0577, 0.0509, 0.0447, 0.0798, 0.0642, 0.0896, 0.0551, 0.0628], device='cuda:1'), in_proj_covar=tensor([0.0070, 0.0067, 0.0066, 0.0068, 0.0073, 0.0094, 0.0072, 0.0063], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-27 23:16:34,440 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=144129.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 23:16:41,251 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.8072, 3.4859, 3.0507, 3.2244, 2.3932, 2.9964, 3.1052, 2.8036], device='cuda:1'), covar=tensor([0.1723, 0.0949, 0.0635, 0.1066, 0.2875, 0.0844, 0.1525, 0.1914], device='cuda:1'), in_proj_covar=tensor([0.0285, 0.0301, 0.0213, 0.0276, 0.0313, 0.0255, 0.0249, 0.0265], device='cuda:1'), out_proj_covar=tensor([1.1395e-04, 1.1845e-04, 8.3950e-05, 1.0873e-04, 1.2625e-04, 1.0030e-04, 1.0024e-04, 1.0456e-04], device='cuda:1') 2023-04-27 23:16:53,901 INFO [finetune.py:976] (1/7) Epoch 26, batch 950, loss[loss=0.1579, simple_loss=0.2208, pruned_loss=0.04747, over 4745.00 frames. ], tot_loss[loss=0.1643, simple_loss=0.2362, pruned_loss=0.04624, over 948433.20 frames. ], batch size: 23, lr: 2.97e-03, grad_scale: 32.0 2023-04-27 23:17:27,101 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.2663, 1.4843, 1.3292, 1.7803, 1.6521, 1.8813, 1.4468, 3.3443], device='cuda:1'), covar=tensor([0.0610, 0.0820, 0.0840, 0.1194, 0.0618, 0.0474, 0.0719, 0.0136], device='cuda:1'), in_proj_covar=tensor([0.0037, 0.0039, 0.0039, 0.0042, 0.0040, 0.0038, 0.0038, 0.0055], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0013, 0.0013, 0.0015, 0.0014, 0.0013, 0.0013, 0.0016], device='cuda:1') 2023-04-27 23:17:28,112 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.085e+02 1.443e+02 1.800e+02 2.192e+02 5.909e+02, threshold=3.600e+02, percent-clipped=3.0 2023-04-27 23:17:36,053 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6828, 1.4895, 1.3472, 1.5706, 1.9062, 1.5315, 1.4299, 1.2795], device='cuda:1'), covar=tensor([0.1674, 0.1428, 0.1747, 0.1280, 0.0899, 0.1697, 0.1783, 0.2210], device='cuda:1'), in_proj_covar=tensor([0.0314, 0.0308, 0.0350, 0.0286, 0.0329, 0.0303, 0.0297, 0.0373], device='cuda:1'), out_proj_covar=tensor([6.4177e-05, 6.3268e-05, 7.3400e-05, 5.7218e-05, 6.7499e-05, 6.3342e-05, 6.1505e-05, 7.9160e-05], device='cuda:1') 2023-04-27 23:17:37,856 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=144177.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 23:17:59,696 INFO [finetune.py:976] (1/7) Epoch 26, batch 1000, loss[loss=0.1443, simple_loss=0.2165, pruned_loss=0.03601, over 4699.00 frames. ], tot_loss[loss=0.1673, simple_loss=0.2394, pruned_loss=0.04763, over 950313.59 frames. ], batch size: 23, lr: 2.97e-03, grad_scale: 32.0 2023-04-27 23:18:32,531 INFO [finetune.py:976] (1/7) Epoch 26, batch 1050, loss[loss=0.1439, simple_loss=0.2162, pruned_loss=0.03581, over 4775.00 frames. ], tot_loss[loss=0.169, simple_loss=0.2422, pruned_loss=0.04794, over 952348.66 frames. ], batch size: 26, lr: 2.97e-03, grad_scale: 32.0 2023-04-27 23:18:45,251 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9655, 2.4720, 2.2134, 1.9264, 1.4568, 1.5200, 2.3055, 1.4153], device='cuda:1'), covar=tensor([0.1699, 0.1425, 0.1239, 0.1615, 0.2317, 0.1885, 0.0859, 0.2028], device='cuda:1'), in_proj_covar=tensor([0.0197, 0.0210, 0.0168, 0.0204, 0.0200, 0.0185, 0.0156, 0.0186], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-27 23:18:51,248 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.174e+02 1.491e+02 1.893e+02 2.144e+02 7.486e+02, threshold=3.787e+02, percent-clipped=1.0 2023-04-27 23:19:06,557 INFO [finetune.py:976] (1/7) Epoch 26, batch 1100, loss[loss=0.1895, simple_loss=0.2705, pruned_loss=0.05432, over 4795.00 frames. ], tot_loss[loss=0.1712, simple_loss=0.2447, pruned_loss=0.0488, over 954093.40 frames. ], batch size: 51, lr: 2.97e-03, grad_scale: 32.0 2023-04-27 23:19:39,801 INFO [finetune.py:976] (1/7) Epoch 26, batch 1150, loss[loss=0.1761, simple_loss=0.2674, pruned_loss=0.04236, over 4816.00 frames. ], tot_loss[loss=0.1706, simple_loss=0.2445, pruned_loss=0.04834, over 954040.61 frames. ], batch size: 39, lr: 2.97e-03, grad_scale: 32.0 2023-04-27 23:19:46,920 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.1114, 2.6074, 1.1181, 1.5187, 2.2487, 1.2338, 3.5602, 1.8336], device='cuda:1'), covar=tensor([0.0666, 0.0606, 0.0758, 0.1166, 0.0436, 0.1031, 0.0231, 0.0577], device='cuda:1'), in_proj_covar=tensor([0.0050, 0.0064, 0.0047, 0.0046, 0.0049, 0.0051, 0.0073, 0.0050], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:1') 2023-04-27 23:19:58,625 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9691, 4.2019, 0.8701, 2.2666, 2.3839, 2.9665, 2.4242, 1.0870], device='cuda:1'), covar=tensor([0.1273, 0.1043, 0.2151, 0.1164, 0.0958, 0.0923, 0.1413, 0.2091], device='cuda:1'), in_proj_covar=tensor([0.0117, 0.0237, 0.0135, 0.0120, 0.0131, 0.0151, 0.0116, 0.0117], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-27 23:19:58,646 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8611, 2.1206, 1.0471, 1.5237, 2.1768, 1.5654, 1.5800, 1.6603], device='cuda:1'), covar=tensor([0.0466, 0.0341, 0.0296, 0.0537, 0.0234, 0.0506, 0.0478, 0.0543], device='cuda:1'), in_proj_covar=tensor([0.0028, 0.0023, 0.0022, 0.0028, 0.0019, 0.0028, 0.0027, 0.0028], device='cuda:1'), out_proj_covar=tensor([0.0052, 0.0046, 0.0038, 0.0053, 0.0039, 0.0050, 0.0050, 0.0052], device='cuda:1') 2023-04-27 23:19:59,723 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.045e+02 1.546e+02 1.832e+02 2.212e+02 3.348e+02, threshold=3.664e+02, percent-clipped=0.0 2023-04-27 23:20:13,107 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([3.4932, 3.3625, 2.4723, 3.9403, 3.4110, 3.3801, 1.5700, 3.3798], device='cuda:1'), covar=tensor([0.1696, 0.1316, 0.3249, 0.1990, 0.3816, 0.1918, 0.5511, 0.2413], device='cuda:1'), in_proj_covar=tensor([0.0247, 0.0219, 0.0253, 0.0307, 0.0300, 0.0248, 0.0274, 0.0275], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 23:20:14,235 INFO [finetune.py:976] (1/7) Epoch 26, batch 1200, loss[loss=0.1287, simple_loss=0.2068, pruned_loss=0.02533, over 4714.00 frames. ], tot_loss[loss=0.1699, simple_loss=0.2431, pruned_loss=0.04833, over 953439.52 frames. ], batch size: 23, lr: 2.97e-03, grad_scale: 32.0 2023-04-27 23:20:32,361 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=144411.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:21:13,639 INFO [finetune.py:976] (1/7) Epoch 26, batch 1250, loss[loss=0.1591, simple_loss=0.2304, pruned_loss=0.0439, over 4849.00 frames. ], tot_loss[loss=0.1685, simple_loss=0.2408, pruned_loss=0.04809, over 954862.25 frames. ], batch size: 47, lr: 2.97e-03, grad_scale: 32.0 2023-04-27 23:21:45,824 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.274e+01 1.443e+02 1.675e+02 2.129e+02 3.494e+02, threshold=3.349e+02, percent-clipped=0.0 2023-04-27 23:21:51,784 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=144472.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:22:16,357 INFO [finetune.py:976] (1/7) Epoch 26, batch 1300, loss[loss=0.2038, simple_loss=0.2613, pruned_loss=0.07317, over 4877.00 frames. ], tot_loss[loss=0.1665, simple_loss=0.238, pruned_loss=0.04752, over 954167.32 frames. ], batch size: 31, lr: 2.97e-03, grad_scale: 32.0 2023-04-27 23:23:06,076 INFO [finetune.py:976] (1/7) Epoch 26, batch 1350, loss[loss=0.1331, simple_loss=0.2029, pruned_loss=0.03171, over 4770.00 frames. ], tot_loss[loss=0.1659, simple_loss=0.2376, pruned_loss=0.04706, over 954688.19 frames. ], batch size: 28, lr: 2.97e-03, grad_scale: 32.0 2023-04-27 23:23:09,633 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6994, 2.1010, 1.8375, 2.0798, 1.5679, 1.7662, 1.7266, 1.4914], device='cuda:1'), covar=tensor([0.1600, 0.1187, 0.0762, 0.0943, 0.3292, 0.1033, 0.1521, 0.1973], device='cuda:1'), in_proj_covar=tensor([0.0284, 0.0299, 0.0212, 0.0274, 0.0312, 0.0254, 0.0247, 0.0263], device='cuda:1'), out_proj_covar=tensor([1.1339e-04, 1.1781e-04, 8.3549e-05, 1.0802e-04, 1.2573e-04, 9.9741e-05, 9.9468e-05, 1.0403e-04], device='cuda:1') 2023-04-27 23:23:26,215 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.435e+01 1.590e+02 1.903e+02 2.333e+02 4.606e+02, threshold=3.805e+02, percent-clipped=5.0 2023-04-27 23:23:39,995 INFO [finetune.py:976] (1/7) Epoch 26, batch 1400, loss[loss=0.1715, simple_loss=0.2451, pruned_loss=0.04898, over 4849.00 frames. ], tot_loss[loss=0.1674, simple_loss=0.2405, pruned_loss=0.04718, over 956893.83 frames. ], batch size: 47, lr: 2.97e-03, grad_scale: 32.0 2023-04-27 23:23:54,338 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5139, 1.4526, 1.7758, 1.8349, 1.3507, 1.2676, 1.5077, 1.0334], device='cuda:1'), covar=tensor([0.0475, 0.0546, 0.0342, 0.0448, 0.0689, 0.1073, 0.0502, 0.0558], device='cuda:1'), in_proj_covar=tensor([0.0071, 0.0067, 0.0067, 0.0069, 0.0074, 0.0095, 0.0073, 0.0064], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-27 23:23:57,846 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.2718, 1.6833, 2.2052, 2.5497, 2.1522, 1.7074, 1.3631, 1.9981], device='cuda:1'), covar=tensor([0.2887, 0.3117, 0.1615, 0.2209, 0.2507, 0.2507, 0.4053, 0.1895], device='cuda:1'), in_proj_covar=tensor([0.0291, 0.0245, 0.0227, 0.0312, 0.0221, 0.0234, 0.0226, 0.0186], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-27 23:23:59,011 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.2424, 1.4604, 1.3490, 1.7155, 1.6260, 1.9124, 1.4254, 3.3824], device='cuda:1'), covar=tensor([0.0604, 0.0811, 0.0781, 0.1203, 0.0618, 0.0450, 0.0722, 0.0135], device='cuda:1'), in_proj_covar=tensor([0.0037, 0.0039, 0.0039, 0.0042, 0.0040, 0.0037, 0.0038, 0.0055], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0013, 0.0013, 0.0015, 0.0014, 0.0013, 0.0013, 0.0016], device='cuda:1') 2023-04-27 23:24:12,866 INFO [finetune.py:976] (1/7) Epoch 26, batch 1450, loss[loss=0.173, simple_loss=0.2452, pruned_loss=0.05044, over 4761.00 frames. ], tot_loss[loss=0.1688, simple_loss=0.2425, pruned_loss=0.04756, over 956591.14 frames. ], batch size: 54, lr: 2.97e-03, grad_scale: 32.0 2023-04-27 23:24:25,430 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-04-27 23:24:33,310 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.219e+02 1.523e+02 1.914e+02 2.224e+02 4.652e+02, threshold=3.827e+02, percent-clipped=2.0 2023-04-27 23:24:46,127 INFO [finetune.py:976] (1/7) Epoch 26, batch 1500, loss[loss=0.1635, simple_loss=0.2509, pruned_loss=0.03802, over 4806.00 frames. ], tot_loss[loss=0.1682, simple_loss=0.2424, pruned_loss=0.047, over 956937.20 frames. ], batch size: 45, lr: 2.97e-03, grad_scale: 32.0 2023-04-27 23:24:51,034 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.20 vs. limit=5.0 2023-04-27 23:25:19,029 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5726, 1.4800, 1.7811, 1.9140, 1.4520, 1.3028, 1.4999, 1.0428], device='cuda:1'), covar=tensor([0.0515, 0.0593, 0.0343, 0.0477, 0.0675, 0.1083, 0.0522, 0.0486], device='cuda:1'), in_proj_covar=tensor([0.0070, 0.0068, 0.0066, 0.0069, 0.0074, 0.0095, 0.0073, 0.0064], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-27 23:25:20,118 INFO [finetune.py:976] (1/7) Epoch 26, batch 1550, loss[loss=0.1768, simple_loss=0.2538, pruned_loss=0.04986, over 4794.00 frames. ], tot_loss[loss=0.1692, simple_loss=0.2435, pruned_loss=0.04747, over 958406.08 frames. ], batch size: 29, lr: 2.97e-03, grad_scale: 32.0 2023-04-27 23:25:32,515 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.1085, 2.7794, 2.4043, 2.6759, 1.9593, 2.3877, 2.5658, 1.8700], device='cuda:1'), covar=tensor([0.2341, 0.1211, 0.0748, 0.1147, 0.2969, 0.1060, 0.1757, 0.2536], device='cuda:1'), in_proj_covar=tensor([0.0283, 0.0297, 0.0212, 0.0274, 0.0311, 0.0253, 0.0246, 0.0262], device='cuda:1'), out_proj_covar=tensor([1.1279e-04, 1.1698e-04, 8.3330e-05, 1.0783e-04, 1.2533e-04, 9.9375e-05, 9.9081e-05, 1.0329e-04], device='cuda:1') 2023-04-27 23:25:42,641 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=144767.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:25:45,417 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.907e+01 1.462e+02 1.740e+02 2.100e+02 5.227e+02, threshold=3.480e+02, percent-clipped=2.0 2023-04-27 23:26:02,432 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7211, 1.7890, 0.8273, 1.4121, 1.8023, 1.5312, 1.4853, 1.4871], device='cuda:1'), covar=tensor([0.0483, 0.0336, 0.0323, 0.0537, 0.0256, 0.0484, 0.0445, 0.0560], device='cuda:1'), in_proj_covar=tensor([0.0028, 0.0023, 0.0022, 0.0028, 0.0019, 0.0028, 0.0027, 0.0028], device='cuda:1'), out_proj_covar=tensor([0.0051, 0.0046, 0.0038, 0.0053, 0.0039, 0.0050, 0.0050, 0.0052], device='cuda:1') 2023-04-27 23:26:14,643 INFO [finetune.py:976] (1/7) Epoch 26, batch 1600, loss[loss=0.1396, simple_loss=0.209, pruned_loss=0.03505, over 4800.00 frames. ], tot_loss[loss=0.168, simple_loss=0.2417, pruned_loss=0.04719, over 958958.66 frames. ], batch size: 25, lr: 2.97e-03, grad_scale: 32.0 2023-04-27 23:27:00,877 INFO [finetune.py:976] (1/7) Epoch 26, batch 1650, loss[loss=0.1809, simple_loss=0.2598, pruned_loss=0.05099, over 4760.00 frames. ], tot_loss[loss=0.1671, simple_loss=0.2399, pruned_loss=0.04718, over 957177.85 frames. ], batch size: 26, lr: 2.97e-03, grad_scale: 32.0 2023-04-27 23:27:20,930 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.028e+02 1.459e+02 1.734e+02 2.287e+02 5.147e+02, threshold=3.469e+02, percent-clipped=2.0 2023-04-27 23:27:39,743 INFO [finetune.py:976] (1/7) Epoch 26, batch 1700, loss[loss=0.1497, simple_loss=0.2331, pruned_loss=0.03321, over 4790.00 frames. ], tot_loss[loss=0.1642, simple_loss=0.2365, pruned_loss=0.04592, over 957440.12 frames. ], batch size: 29, lr: 2.97e-03, grad_scale: 32.0 2023-04-27 23:28:30,635 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-04-27 23:28:34,823 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6970, 1.4403, 1.3696, 1.4858, 1.8664, 1.5078, 1.3016, 1.3308], device='cuda:1'), covar=tensor([0.1776, 0.1442, 0.1854, 0.1426, 0.0887, 0.1704, 0.1859, 0.2433], device='cuda:1'), in_proj_covar=tensor([0.0314, 0.0306, 0.0349, 0.0285, 0.0328, 0.0304, 0.0297, 0.0374], device='cuda:1'), out_proj_covar=tensor([6.4188e-05, 6.2965e-05, 7.3383e-05, 5.7196e-05, 6.7251e-05, 6.3461e-05, 6.1564e-05, 7.9364e-05], device='cuda:1') 2023-04-27 23:28:44,466 INFO [finetune.py:976] (1/7) Epoch 26, batch 1750, loss[loss=0.1298, simple_loss=0.2066, pruned_loss=0.02651, over 4790.00 frames. ], tot_loss[loss=0.1644, simple_loss=0.2372, pruned_loss=0.0458, over 956244.65 frames. ], batch size: 25, lr: 2.97e-03, grad_scale: 32.0 2023-04-27 23:28:45,290 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-04-27 23:29:07,730 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5077, 2.4019, 2.2338, 2.1133, 2.5899, 2.1839, 3.0863, 1.9306], device='cuda:1'), covar=tensor([0.2770, 0.1386, 0.3100, 0.2113, 0.1175, 0.1975, 0.1301, 0.3436], device='cuda:1'), in_proj_covar=tensor([0.0336, 0.0349, 0.0423, 0.0349, 0.0380, 0.0372, 0.0363, 0.0420], device='cuda:1'), out_proj_covar=tensor([9.9199e-05, 1.0381e-04, 1.2790e-04, 1.0461e-04, 1.1285e-04, 1.1075e-04, 1.0626e-04, 1.2629e-04], device='cuda:1') 2023-04-27 23:29:25,331 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.053e+02 1.536e+02 1.830e+02 2.200e+02 7.306e+02, threshold=3.661e+02, percent-clipped=1.0 2023-04-27 23:29:36,253 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.0920, 0.7468, 0.8426, 0.9361, 1.2408, 1.0067, 0.9588, 0.9194], device='cuda:1'), covar=tensor([0.1843, 0.1527, 0.2174, 0.1664, 0.1242, 0.1484, 0.1963, 0.2612], device='cuda:1'), in_proj_covar=tensor([0.0313, 0.0306, 0.0350, 0.0285, 0.0328, 0.0304, 0.0297, 0.0374], device='cuda:1'), out_proj_covar=tensor([6.4115e-05, 6.2989e-05, 7.3501e-05, 5.7164e-05, 6.7313e-05, 6.3437e-05, 6.1611e-05, 7.9403e-05], device='cuda:1') 2023-04-27 23:29:50,039 INFO [finetune.py:976] (1/7) Epoch 26, batch 1800, loss[loss=0.2169, simple_loss=0.2965, pruned_loss=0.06865, over 4815.00 frames. ], tot_loss[loss=0.1667, simple_loss=0.2405, pruned_loss=0.04641, over 955301.82 frames. ], batch size: 45, lr: 2.97e-03, grad_scale: 32.0 2023-04-27 23:29:53,365 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=144996.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:30:15,061 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=145027.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:30:24,556 INFO [finetune.py:976] (1/7) Epoch 26, batch 1850, loss[loss=0.1906, simple_loss=0.2605, pruned_loss=0.06039, over 4917.00 frames. ], tot_loss[loss=0.1668, simple_loss=0.2413, pruned_loss=0.04621, over 953747.41 frames. ], batch size: 38, lr: 2.97e-03, grad_scale: 32.0 2023-04-27 23:30:34,380 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=145057.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 23:30:41,228 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=145066.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:30:41,819 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=145067.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:30:44,630 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.003e+02 1.513e+02 1.816e+02 2.184e+02 4.128e+02, threshold=3.632e+02, percent-clipped=2.0 2023-04-27 23:30:51,864 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.1704, 1.9448, 2.0552, 2.5685, 2.5026, 2.0909, 1.7087, 2.3162], device='cuda:1'), covar=tensor([0.0820, 0.1133, 0.0784, 0.0512, 0.0595, 0.0810, 0.0751, 0.0517], device='cuda:1'), in_proj_covar=tensor([0.0186, 0.0204, 0.0187, 0.0172, 0.0177, 0.0178, 0.0152, 0.0179], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 23:30:55,917 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=145088.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:30:58,303 INFO [finetune.py:976] (1/7) Epoch 26, batch 1900, loss[loss=0.1333, simple_loss=0.2098, pruned_loss=0.02843, over 4360.00 frames. ], tot_loss[loss=0.1684, simple_loss=0.2427, pruned_loss=0.04701, over 954400.41 frames. ], batch size: 19, lr: 2.97e-03, grad_scale: 32.0 2023-04-27 23:31:13,801 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=145115.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:31:14,447 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.0183, 1.0658, 1.2158, 1.2231, 1.0134, 0.9270, 1.0767, 0.6759], device='cuda:1'), covar=tensor([0.0532, 0.0541, 0.0441, 0.0453, 0.0653, 0.1152, 0.0404, 0.0557], device='cuda:1'), in_proj_covar=tensor([0.0071, 0.0068, 0.0066, 0.0069, 0.0075, 0.0095, 0.0073, 0.0063], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-27 23:31:22,155 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=145127.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:31:32,057 INFO [finetune.py:976] (1/7) Epoch 26, batch 1950, loss[loss=0.1492, simple_loss=0.2187, pruned_loss=0.03986, over 4838.00 frames. ], tot_loss[loss=0.1679, simple_loss=0.2413, pruned_loss=0.0472, over 955757.51 frames. ], batch size: 47, lr: 2.97e-03, grad_scale: 32.0 2023-04-27 23:32:03,280 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.496e+01 1.516e+02 2.001e+02 2.404e+02 4.376e+02, threshold=4.002e+02, percent-clipped=4.0 2023-04-27 23:32:23,670 INFO [finetune.py:976] (1/7) Epoch 26, batch 2000, loss[loss=0.1555, simple_loss=0.2316, pruned_loss=0.03972, over 4758.00 frames. ], tot_loss[loss=0.1663, simple_loss=0.2391, pruned_loss=0.04679, over 956928.01 frames. ], batch size: 26, lr: 2.97e-03, grad_scale: 32.0 2023-04-27 23:33:01,740 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.2710, 1.4729, 1.7734, 1.9043, 1.8043, 1.9056, 1.8077, 1.8067], device='cuda:1'), covar=tensor([0.3789, 0.4601, 0.4025, 0.4154, 0.5130, 0.6315, 0.4430, 0.4114], device='cuda:1'), in_proj_covar=tensor([0.0342, 0.0378, 0.0331, 0.0342, 0.0352, 0.0397, 0.0363, 0.0335], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 23:33:03,263 INFO [finetune.py:976] (1/7) Epoch 26, batch 2050, loss[loss=0.1218, simple_loss=0.199, pruned_loss=0.02224, over 4763.00 frames. ], tot_loss[loss=0.1637, simple_loss=0.2359, pruned_loss=0.04571, over 957012.44 frames. ], batch size: 28, lr: 2.97e-03, grad_scale: 32.0 2023-04-27 23:33:43,224 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.085e+02 1.513e+02 1.827e+02 2.273e+02 3.950e+02, threshold=3.653e+02, percent-clipped=0.0 2023-04-27 23:33:58,463 INFO [finetune.py:976] (1/7) Epoch 26, batch 2100, loss[loss=0.1458, simple_loss=0.222, pruned_loss=0.03477, over 4802.00 frames. ], tot_loss[loss=0.1621, simple_loss=0.2345, pruned_loss=0.04485, over 955680.36 frames. ], batch size: 25, lr: 2.97e-03, grad_scale: 32.0 2023-04-27 23:34:32,515 INFO [finetune.py:976] (1/7) Epoch 26, batch 2150, loss[loss=0.1701, simple_loss=0.2554, pruned_loss=0.04241, over 4733.00 frames. ], tot_loss[loss=0.1659, simple_loss=0.2389, pruned_loss=0.04641, over 955124.53 frames. ], batch size: 54, lr: 2.97e-03, grad_scale: 32.0 2023-04-27 23:34:39,178 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=145352.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 23:34:51,070 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.081e+02 1.597e+02 2.010e+02 2.319e+02 4.178e+02, threshold=4.020e+02, percent-clipped=2.0 2023-04-27 23:34:51,205 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=145371.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:34:59,914 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=145383.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:35:00,498 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.3936, 1.4748, 4.0784, 3.8003, 3.5550, 3.9137, 3.8149, 3.5924], device='cuda:1'), covar=tensor([0.7019, 0.5310, 0.1053, 0.1731, 0.1214, 0.1648, 0.1604, 0.1510], device='cuda:1'), in_proj_covar=tensor([0.0310, 0.0308, 0.0408, 0.0410, 0.0348, 0.0415, 0.0319, 0.0365], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-27 23:35:10,686 INFO [finetune.py:976] (1/7) Epoch 26, batch 2200, loss[loss=0.1788, simple_loss=0.2491, pruned_loss=0.05424, over 4918.00 frames. ], tot_loss[loss=0.1669, simple_loss=0.2407, pruned_loss=0.04655, over 954971.52 frames. ], batch size: 42, lr: 2.97e-03, grad_scale: 32.0 2023-04-27 23:35:30,045 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=145422.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:35:37,101 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=145432.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:35:44,062 INFO [finetune.py:976] (1/7) Epoch 26, batch 2250, loss[loss=0.17, simple_loss=0.2373, pruned_loss=0.05132, over 4799.00 frames. ], tot_loss[loss=0.1683, simple_loss=0.2422, pruned_loss=0.04717, over 956514.17 frames. ], batch size: 51, lr: 2.97e-03, grad_scale: 32.0 2023-04-27 23:36:03,178 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.017e+02 1.640e+02 1.880e+02 2.244e+02 3.153e+02, threshold=3.761e+02, percent-clipped=0.0 2023-04-27 23:36:03,886 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0264, 2.4727, 1.1620, 1.3916, 2.0604, 1.1540, 3.2780, 1.6715], device='cuda:1'), covar=tensor([0.0709, 0.0814, 0.0877, 0.1227, 0.0482, 0.1016, 0.0221, 0.0611], device='cuda:1'), in_proj_covar=tensor([0.0050, 0.0064, 0.0047, 0.0046, 0.0049, 0.0051, 0.0073, 0.0051], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:1') 2023-04-27 23:36:17,798 INFO [finetune.py:976] (1/7) Epoch 26, batch 2300, loss[loss=0.1415, simple_loss=0.2231, pruned_loss=0.02998, over 4792.00 frames. ], tot_loss[loss=0.1679, simple_loss=0.2423, pruned_loss=0.04671, over 956740.28 frames. ], batch size: 26, lr: 2.97e-03, grad_scale: 32.0 2023-04-27 23:36:19,729 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4361, 1.6058, 1.4736, 1.9052, 1.7030, 1.9480, 1.4648, 3.8027], device='cuda:1'), covar=tensor([0.0669, 0.1083, 0.1007, 0.1331, 0.0805, 0.0609, 0.0953, 0.0194], device='cuda:1'), in_proj_covar=tensor([0.0037, 0.0038, 0.0039, 0.0042, 0.0040, 0.0037, 0.0037, 0.0054], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0013, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:1') 2023-04-27 23:36:49,868 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6327, 3.3637, 1.0452, 1.8887, 1.8148, 2.5309, 1.8742, 0.9780], device='cuda:1'), covar=tensor([0.1185, 0.0849, 0.1800, 0.1141, 0.1020, 0.0902, 0.1434, 0.1857], device='cuda:1'), in_proj_covar=tensor([0.0117, 0.0237, 0.0134, 0.0120, 0.0130, 0.0151, 0.0116, 0.0117], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-27 23:36:51,030 INFO [finetune.py:976] (1/7) Epoch 26, batch 2350, loss[loss=0.1169, simple_loss=0.1874, pruned_loss=0.02314, over 4772.00 frames. ], tot_loss[loss=0.1668, simple_loss=0.2406, pruned_loss=0.04647, over 957383.14 frames. ], batch size: 28, lr: 2.96e-03, grad_scale: 32.0 2023-04-27 23:37:10,063 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.053e+02 1.548e+02 1.895e+02 2.111e+02 3.968e+02, threshold=3.791e+02, percent-clipped=2.0 2023-04-27 23:37:40,488 INFO [finetune.py:976] (1/7) Epoch 26, batch 2400, loss[loss=0.1602, simple_loss=0.2309, pruned_loss=0.04472, over 4824.00 frames. ], tot_loss[loss=0.1651, simple_loss=0.2382, pruned_loss=0.04605, over 955676.30 frames. ], batch size: 40, lr: 2.96e-03, grad_scale: 32.0 2023-04-27 23:37:41,819 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=145594.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:38:03,935 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.66 vs. limit=5.0 2023-04-27 23:38:46,679 INFO [finetune.py:976] (1/7) Epoch 26, batch 2450, loss[loss=0.1581, simple_loss=0.2342, pruned_loss=0.04104, over 4926.00 frames. ], tot_loss[loss=0.1636, simple_loss=0.2363, pruned_loss=0.0455, over 957764.01 frames. ], batch size: 33, lr: 2.96e-03, grad_scale: 32.0 2023-04-27 23:38:47,928 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=145643.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:39:00,512 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=145652.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 23:39:07,842 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=145655.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 23:39:30,140 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.068e+02 1.495e+02 1.759e+02 2.001e+02 3.629e+02, threshold=3.517e+02, percent-clipped=0.0 2023-04-27 23:39:42,993 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=145683.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:39:48,404 INFO [finetune.py:976] (1/7) Epoch 26, batch 2500, loss[loss=0.1662, simple_loss=0.2341, pruned_loss=0.04914, over 4851.00 frames. ], tot_loss[loss=0.1638, simple_loss=0.2366, pruned_loss=0.04546, over 958265.16 frames. ], batch size: 44, lr: 2.96e-03, grad_scale: 32.0 2023-04-27 23:39:54,787 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=145700.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:39:57,802 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=145704.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:40:09,661 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=145722.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:40:12,705 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=145727.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:40:15,131 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=145731.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:40:20,101 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=5.14 vs. limit=5.0 2023-04-27 23:40:22,237 INFO [finetune.py:976] (1/7) Epoch 26, batch 2550, loss[loss=0.2141, simple_loss=0.2794, pruned_loss=0.07442, over 4816.00 frames. ], tot_loss[loss=0.1659, simple_loss=0.2398, pruned_loss=0.04601, over 956296.32 frames. ], batch size: 38, lr: 2.96e-03, grad_scale: 32.0 2023-04-27 23:40:41,744 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=145770.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:40:42,276 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.201e+01 1.571e+02 1.844e+02 2.174e+02 6.257e+02, threshold=3.689e+02, percent-clipped=2.0 2023-04-27 23:40:56,106 INFO [finetune.py:976] (1/7) Epoch 26, batch 2600, loss[loss=0.1298, simple_loss=0.2145, pruned_loss=0.02261, over 4795.00 frames. ], tot_loss[loss=0.1672, simple_loss=0.2413, pruned_loss=0.04651, over 954118.83 frames. ], batch size: 29, lr: 2.96e-03, grad_scale: 32.0 2023-04-27 23:41:09,803 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=3.87 vs. limit=5.0 2023-04-27 23:41:14,631 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-04-27 23:41:16,293 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=145821.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 23:41:21,607 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.3704, 1.1054, 3.6841, 3.1876, 3.2705, 3.4458, 3.4762, 3.1073], device='cuda:1'), covar=tensor([0.9902, 0.8471, 0.1905, 0.3354, 0.2462, 0.3361, 0.3398, 0.3358], device='cuda:1'), in_proj_covar=tensor([0.0312, 0.0309, 0.0410, 0.0409, 0.0350, 0.0417, 0.0320, 0.0364], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-27 23:41:29,915 INFO [finetune.py:976] (1/7) Epoch 26, batch 2650, loss[loss=0.1475, simple_loss=0.23, pruned_loss=0.03252, over 4728.00 frames. ], tot_loss[loss=0.1688, simple_loss=0.243, pruned_loss=0.04726, over 955680.05 frames. ], batch size: 23, lr: 2.96e-03, grad_scale: 32.0 2023-04-27 23:41:33,089 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.6687, 2.0546, 2.4187, 3.1595, 2.4086, 1.9028, 2.0340, 2.3418], device='cuda:1'), covar=tensor([0.2905, 0.3196, 0.1548, 0.2088, 0.2600, 0.2565, 0.3430, 0.2111], device='cuda:1'), in_proj_covar=tensor([0.0293, 0.0246, 0.0228, 0.0314, 0.0221, 0.0235, 0.0227, 0.0187], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-27 23:41:49,916 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.145e+02 1.525e+02 1.770e+02 2.195e+02 3.362e+02, threshold=3.540e+02, percent-clipped=0.0 2023-04-27 23:41:57,236 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=145882.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 23:42:03,155 INFO [finetune.py:976] (1/7) Epoch 26, batch 2700, loss[loss=0.161, simple_loss=0.217, pruned_loss=0.05249, over 4595.00 frames. ], tot_loss[loss=0.1684, simple_loss=0.2428, pruned_loss=0.04701, over 955771.65 frames. ], batch size: 19, lr: 2.96e-03, grad_scale: 32.0 2023-04-27 23:42:14,444 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-04-27 23:42:36,791 INFO [finetune.py:976] (1/7) Epoch 26, batch 2750, loss[loss=0.1632, simple_loss=0.2295, pruned_loss=0.0485, over 4821.00 frames. ], tot_loss[loss=0.1674, simple_loss=0.2407, pruned_loss=0.04704, over 957249.62 frames. ], batch size: 33, lr: 2.96e-03, grad_scale: 32.0 2023-04-27 23:42:42,209 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=145950.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 23:42:56,787 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.020e+02 1.527e+02 1.797e+02 2.320e+02 4.826e+02, threshold=3.594e+02, percent-clipped=4.0 2023-04-27 23:43:10,063 INFO [finetune.py:976] (1/7) Epoch 26, batch 2800, loss[loss=0.1708, simple_loss=0.2405, pruned_loss=0.05056, over 4910.00 frames. ], tot_loss[loss=0.1653, simple_loss=0.2381, pruned_loss=0.04623, over 957954.08 frames. ], batch size: 36, lr: 2.96e-03, grad_scale: 32.0 2023-04-27 23:43:14,404 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=145999.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:43:35,186 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=146027.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:43:44,744 INFO [finetune.py:976] (1/7) Epoch 26, batch 2850, loss[loss=0.153, simple_loss=0.2248, pruned_loss=0.04061, over 4764.00 frames. ], tot_loss[loss=0.1648, simple_loss=0.2369, pruned_loss=0.04634, over 957155.54 frames. ], batch size: 59, lr: 2.96e-03, grad_scale: 32.0 2023-04-27 23:44:06,245 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8505, 1.5729, 1.4214, 1.6498, 2.0583, 1.6467, 1.4565, 1.3347], device='cuda:1'), covar=tensor([0.1491, 0.1291, 0.1721, 0.1251, 0.0786, 0.1608, 0.1911, 0.2121], device='cuda:1'), in_proj_covar=tensor([0.0311, 0.0307, 0.0348, 0.0284, 0.0325, 0.0302, 0.0297, 0.0372], device='cuda:1'), out_proj_covar=tensor([6.3470e-05, 6.3253e-05, 7.3149e-05, 5.6773e-05, 6.6632e-05, 6.2975e-05, 6.1431e-05, 7.8762e-05], device='cuda:1') 2023-04-27 23:44:22,215 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.964e+01 1.598e+02 1.831e+02 2.098e+02 3.575e+02, threshold=3.662e+02, percent-clipped=0.0 2023-04-27 23:44:25,237 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=146075.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:44:43,959 INFO [finetune.py:976] (1/7) Epoch 26, batch 2900, loss[loss=0.2084, simple_loss=0.2851, pruned_loss=0.06587, over 4922.00 frames. ], tot_loss[loss=0.1686, simple_loss=0.2413, pruned_loss=0.04791, over 957652.55 frames. ], batch size: 42, lr: 2.96e-03, grad_scale: 64.0 2023-04-27 23:45:49,356 INFO [finetune.py:976] (1/7) Epoch 26, batch 2950, loss[loss=0.1844, simple_loss=0.2517, pruned_loss=0.05855, over 4821.00 frames. ], tot_loss[loss=0.1704, simple_loss=0.244, pruned_loss=0.0484, over 956074.59 frames. ], batch size: 40, lr: 2.96e-03, grad_scale: 64.0 2023-04-27 23:45:59,492 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=146149.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:46:11,510 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6991, 2.0785, 1.6901, 1.9707, 1.4494, 1.6880, 1.6601, 1.3403], device='cuda:1'), covar=tensor([0.1688, 0.1045, 0.0794, 0.0967, 0.3218, 0.0997, 0.1592, 0.2159], device='cuda:1'), in_proj_covar=tensor([0.0285, 0.0299, 0.0215, 0.0276, 0.0314, 0.0254, 0.0248, 0.0263], device='cuda:1'), out_proj_covar=tensor([1.1373e-04, 1.1792e-04, 8.4636e-05, 1.0860e-04, 1.2657e-04, 9.9862e-05, 9.9947e-05, 1.0382e-04], device='cuda:1') 2023-04-27 23:46:13,680 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.156e+02 1.553e+02 1.780e+02 2.193e+02 5.128e+02, threshold=3.559e+02, percent-clipped=3.0 2023-04-27 23:46:17,220 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4752, 1.6761, 1.4765, 1.6479, 1.3502, 1.4393, 1.4706, 1.2081], device='cuda:1'), covar=tensor([0.1471, 0.1242, 0.0828, 0.0961, 0.3372, 0.1035, 0.1501, 0.1974], device='cuda:1'), in_proj_covar=tensor([0.0285, 0.0299, 0.0215, 0.0276, 0.0314, 0.0254, 0.0248, 0.0263], device='cuda:1'), out_proj_covar=tensor([1.1368e-04, 1.1784e-04, 8.4600e-05, 1.0853e-04, 1.2649e-04, 9.9761e-05, 9.9895e-05, 1.0372e-04], device='cuda:1') 2023-04-27 23:46:18,907 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=146177.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 23:46:28,438 INFO [finetune.py:976] (1/7) Epoch 26, batch 3000, loss[loss=0.1262, simple_loss=0.2089, pruned_loss=0.02181, over 4821.00 frames. ], tot_loss[loss=0.1706, simple_loss=0.2443, pruned_loss=0.04842, over 955197.36 frames. ], batch size: 30, lr: 2.96e-03, grad_scale: 64.0 2023-04-27 23:46:28,438 INFO [finetune.py:1001] (1/7) Computing validation loss 2023-04-27 23:46:38,926 INFO [finetune.py:1010] (1/7) Epoch 26, validation: loss=0.1526, simple_loss=0.2216, pruned_loss=0.04183, over 2265189.00 frames. 2023-04-27 23:46:38,927 INFO [finetune.py:1011] (1/7) Maximum memory allocated so far is 6391MB 2023-04-27 23:46:50,702 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=146210.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:46:53,206 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.76 vs. limit=2.0 2023-04-27 23:47:11,567 INFO [finetune.py:976] (1/7) Epoch 26, batch 3050, loss[loss=0.1582, simple_loss=0.2302, pruned_loss=0.04311, over 4797.00 frames. ], tot_loss[loss=0.1704, simple_loss=0.245, pruned_loss=0.04794, over 956462.08 frames. ], batch size: 26, lr: 2.96e-03, grad_scale: 64.0 2023-04-27 23:47:17,396 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=146250.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:47:27,537 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.1387, 2.4512, 2.0398, 2.4159, 1.7511, 2.1106, 2.1694, 1.6428], device='cuda:1'), covar=tensor([0.1846, 0.1480, 0.0839, 0.1136, 0.3176, 0.1130, 0.1991, 0.2537], device='cuda:1'), in_proj_covar=tensor([0.0284, 0.0299, 0.0215, 0.0275, 0.0313, 0.0253, 0.0247, 0.0263], device='cuda:1'), out_proj_covar=tensor([1.1342e-04, 1.1760e-04, 8.4519e-05, 1.0844e-04, 1.2605e-04, 9.9637e-05, 9.9717e-05, 1.0362e-04], device='cuda:1') 2023-04-27 23:47:31,020 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.111e+02 1.541e+02 1.745e+02 2.066e+02 6.833e+02, threshold=3.490e+02, percent-clipped=2.0 2023-04-27 23:47:45,139 INFO [finetune.py:976] (1/7) Epoch 26, batch 3100, loss[loss=0.153, simple_loss=0.2356, pruned_loss=0.03521, over 4716.00 frames. ], tot_loss[loss=0.1678, simple_loss=0.2417, pruned_loss=0.04689, over 956164.40 frames. ], batch size: 23, lr: 2.96e-03, grad_scale: 32.0 2023-04-27 23:47:49,794 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=146298.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:47:50,445 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=146299.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:48:18,966 INFO [finetune.py:976] (1/7) Epoch 26, batch 3150, loss[loss=0.1681, simple_loss=0.2311, pruned_loss=0.05261, over 4213.00 frames. ], tot_loss[loss=0.1663, simple_loss=0.2397, pruned_loss=0.04646, over 956181.73 frames. ], batch size: 65, lr: 2.96e-03, grad_scale: 32.0 2023-04-27 23:48:22,015 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.3479, 1.3329, 1.3888, 1.6456, 1.6283, 1.3642, 0.9560, 1.5412], device='cuda:1'), covar=tensor([0.0858, 0.1318, 0.0975, 0.0626, 0.0751, 0.0841, 0.0843, 0.0609], device='cuda:1'), in_proj_covar=tensor([0.0184, 0.0202, 0.0184, 0.0170, 0.0176, 0.0177, 0.0149, 0.0177], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 23:48:22,566 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=146347.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:48:26,672 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.2616, 2.5727, 0.9710, 1.4431, 2.0453, 1.2059, 3.3846, 1.7377], device='cuda:1'), covar=tensor([0.0640, 0.0760, 0.0833, 0.1267, 0.0487, 0.0980, 0.0261, 0.0614], device='cuda:1'), in_proj_covar=tensor([0.0050, 0.0064, 0.0047, 0.0046, 0.0049, 0.0051, 0.0072, 0.0050], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:1') 2023-04-27 23:48:38,468 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.123e+02 1.566e+02 1.850e+02 2.179e+02 3.570e+02, threshold=3.699e+02, percent-clipped=2.0 2023-04-27 23:48:51,066 INFO [finetune.py:976] (1/7) Epoch 26, batch 3200, loss[loss=0.1295, simple_loss=0.2074, pruned_loss=0.02576, over 4707.00 frames. ], tot_loss[loss=0.1653, simple_loss=0.2379, pruned_loss=0.0463, over 958317.74 frames. ], batch size: 23, lr: 2.96e-03, grad_scale: 32.0 2023-04-27 23:48:51,295 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-04-27 23:48:59,729 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.6824, 2.7185, 2.2558, 3.0780, 2.6676, 2.6906, 1.2956, 2.7036], device='cuda:1'), covar=tensor([0.2596, 0.1832, 0.4243, 0.3594, 0.4348, 0.2549, 0.5033, 0.3192], device='cuda:1'), in_proj_covar=tensor([0.0245, 0.0220, 0.0253, 0.0306, 0.0301, 0.0247, 0.0275, 0.0274], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 23:49:05,527 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8727, 1.4140, 1.9250, 2.3551, 1.9646, 1.8137, 1.9111, 1.8183], device='cuda:1'), covar=tensor([0.4383, 0.6431, 0.5883, 0.5340, 0.5636, 0.7777, 0.7786, 0.8934], device='cuda:1'), in_proj_covar=tensor([0.0441, 0.0422, 0.0515, 0.0507, 0.0470, 0.0506, 0.0506, 0.0518], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-27 23:49:24,697 INFO [finetune.py:976] (1/7) Epoch 26, batch 3250, loss[loss=0.2144, simple_loss=0.2898, pruned_loss=0.06946, over 4743.00 frames. ], tot_loss[loss=0.1671, simple_loss=0.2397, pruned_loss=0.04723, over 956292.29 frames. ], batch size: 59, lr: 2.96e-03, grad_scale: 32.0 2023-04-27 23:49:45,355 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.653e+01 1.477e+02 1.784e+02 2.151e+02 4.577e+02, threshold=3.568e+02, percent-clipped=2.0 2023-04-27 23:49:46,054 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4007, 3.2941, 1.1463, 1.6657, 1.5550, 2.4626, 1.8101, 0.9233], device='cuda:1'), covar=tensor([0.1650, 0.1432, 0.1950, 0.1835, 0.1408, 0.1186, 0.1911, 0.2343], device='cuda:1'), in_proj_covar=tensor([0.0117, 0.0239, 0.0136, 0.0121, 0.0132, 0.0153, 0.0117, 0.0118], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-27 23:49:53,883 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=146477.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 23:49:56,398 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7454, 2.2215, 1.6699, 1.6545, 1.3717, 1.3484, 1.7268, 1.2930], device='cuda:1'), covar=tensor([0.1394, 0.1136, 0.1339, 0.1486, 0.2073, 0.1675, 0.0909, 0.1947], device='cuda:1'), in_proj_covar=tensor([0.0198, 0.0210, 0.0170, 0.0204, 0.0200, 0.0186, 0.0157, 0.0187], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-27 23:49:56,461 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.53 vs. limit=5.0 2023-04-27 23:50:13,002 INFO [finetune.py:976] (1/7) Epoch 26, batch 3300, loss[loss=0.1508, simple_loss=0.2274, pruned_loss=0.03704, over 4809.00 frames. ], tot_loss[loss=0.1685, simple_loss=0.2423, pruned_loss=0.04734, over 955781.40 frames. ], batch size: 25, lr: 2.96e-03, grad_scale: 32.0 2023-04-27 23:50:28,268 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=146505.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:50:57,306 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=146525.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 23:51:19,413 INFO [finetune.py:976] (1/7) Epoch 26, batch 3350, loss[loss=0.1612, simple_loss=0.2282, pruned_loss=0.04708, over 4930.00 frames. ], tot_loss[loss=0.17, simple_loss=0.2441, pruned_loss=0.0479, over 957362.69 frames. ], batch size: 42, lr: 2.96e-03, grad_scale: 32.0 2023-04-27 23:51:36,967 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.2279, 2.2760, 1.8705, 1.9938, 2.4359, 1.8848, 2.9843, 1.7425], device='cuda:1'), covar=tensor([0.4043, 0.2086, 0.4989, 0.3212, 0.1851, 0.2862, 0.1307, 0.4550], device='cuda:1'), in_proj_covar=tensor([0.0334, 0.0348, 0.0418, 0.0345, 0.0377, 0.0370, 0.0362, 0.0418], device='cuda:1'), out_proj_covar=tensor([9.8776e-05, 1.0357e-04, 1.2666e-04, 1.0348e-04, 1.1191e-04, 1.0990e-04, 1.0590e-04, 1.2565e-04], device='cuda:1') 2023-04-27 23:51:45,685 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.686e+01 1.611e+02 1.865e+02 2.186e+02 4.215e+02, threshold=3.729e+02, percent-clipped=3.0 2023-04-27 23:51:57,895 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=146591.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:51:58,398 INFO [finetune.py:976] (1/7) Epoch 26, batch 3400, loss[loss=0.2103, simple_loss=0.2684, pruned_loss=0.0761, over 4877.00 frames. ], tot_loss[loss=0.1708, simple_loss=0.2456, pruned_loss=0.04803, over 959798.21 frames. ], batch size: 32, lr: 2.96e-03, grad_scale: 32.0 2023-04-27 23:52:26,074 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0019, 2.3721, 2.0133, 2.4286, 1.7370, 2.0232, 1.9642, 1.5807], device='cuda:1'), covar=tensor([0.2015, 0.1382, 0.0833, 0.1132, 0.3439, 0.1278, 0.2255, 0.2804], device='cuda:1'), in_proj_covar=tensor([0.0289, 0.0303, 0.0218, 0.0280, 0.0317, 0.0256, 0.0251, 0.0267], device='cuda:1'), out_proj_covar=tensor([1.1530e-04, 1.1926e-04, 8.5730e-05, 1.1026e-04, 1.2767e-04, 1.0095e-04, 1.0138e-04, 1.0529e-04], device='cuda:1') 2023-04-27 23:52:27,922 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=146626.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:52:43,202 INFO [finetune.py:976] (1/7) Epoch 26, batch 3450, loss[loss=0.193, simple_loss=0.2559, pruned_loss=0.06501, over 4839.00 frames. ], tot_loss[loss=0.1693, simple_loss=0.2436, pruned_loss=0.04752, over 954070.81 frames. ], batch size: 44, lr: 2.96e-03, grad_scale: 32.0 2023-04-27 23:52:47,935 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([4.0746, 4.1110, 2.9237, 4.7409, 4.2336, 4.1253, 1.5547, 4.0286], device='cuda:1'), covar=tensor([0.1879, 0.1043, 0.3199, 0.1527, 0.2954, 0.1842, 0.6380, 0.2365], device='cuda:1'), in_proj_covar=tensor([0.0244, 0.0219, 0.0252, 0.0305, 0.0299, 0.0246, 0.0273, 0.0274], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 23:52:49,812 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=146652.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:53:04,800 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.064e+02 1.587e+02 1.818e+02 2.131e+02 4.084e+02, threshold=3.635e+02, percent-clipped=1.0 2023-04-27 23:53:13,321 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([3.3029, 3.3486, 2.5009, 3.8470, 3.3566, 3.2880, 1.4703, 3.2111], device='cuda:1'), covar=tensor([0.2255, 0.1431, 0.3676, 0.2813, 0.3476, 0.2394, 0.5986, 0.2971], device='cuda:1'), in_proj_covar=tensor([0.0246, 0.0220, 0.0254, 0.0306, 0.0300, 0.0247, 0.0274, 0.0275], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-27 23:53:13,978 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=146687.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 23:53:16,862 INFO [finetune.py:976] (1/7) Epoch 26, batch 3500, loss[loss=0.1905, simple_loss=0.2671, pruned_loss=0.05695, over 4911.00 frames. ], tot_loss[loss=0.1679, simple_loss=0.2413, pruned_loss=0.0472, over 955650.02 frames. ], batch size: 37, lr: 2.96e-03, grad_scale: 32.0 2023-04-27 23:53:50,653 INFO [finetune.py:976] (1/7) Epoch 26, batch 3550, loss[loss=0.1436, simple_loss=0.2021, pruned_loss=0.04256, over 4240.00 frames. ], tot_loss[loss=0.164, simple_loss=0.2369, pruned_loss=0.0455, over 955276.42 frames. ], batch size: 65, lr: 2.96e-03, grad_scale: 32.0 2023-04-27 23:54:11,377 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.892e+01 1.463e+02 1.726e+02 2.142e+02 5.887e+02, threshold=3.452e+02, percent-clipped=4.0 2023-04-27 23:54:14,454 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.8584, 2.3021, 2.7295, 3.3407, 2.7134, 2.2028, 2.2887, 2.7896], device='cuda:1'), covar=tensor([0.3407, 0.3475, 0.1803, 0.2492, 0.2741, 0.2816, 0.3827, 0.1822], device='cuda:1'), in_proj_covar=tensor([0.0292, 0.0245, 0.0227, 0.0313, 0.0220, 0.0234, 0.0226, 0.0186], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-27 23:54:24,571 INFO [finetune.py:976] (1/7) Epoch 26, batch 3600, loss[loss=0.1675, simple_loss=0.246, pruned_loss=0.04448, over 4727.00 frames. ], tot_loss[loss=0.1646, simple_loss=0.2364, pruned_loss=0.0464, over 954529.77 frames. ], batch size: 23, lr: 2.96e-03, grad_scale: 32.0 2023-04-27 23:54:32,784 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=146805.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:54:51,005 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-04-27 23:54:59,555 INFO [finetune.py:976] (1/7) Epoch 26, batch 3650, loss[loss=0.1524, simple_loss=0.2336, pruned_loss=0.03557, over 4757.00 frames. ], tot_loss[loss=0.166, simple_loss=0.2383, pruned_loss=0.04688, over 954712.63 frames. ], batch size: 28, lr: 2.96e-03, grad_scale: 32.0 2023-04-27 23:55:00,928 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=146844.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:55:11,622 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=146853.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:55:30,101 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.832e+01 1.619e+02 1.915e+02 2.507e+02 7.747e+02, threshold=3.830e+02, percent-clipped=2.0 2023-04-27 23:55:48,372 INFO [finetune.py:976] (1/7) Epoch 26, batch 3700, loss[loss=0.197, simple_loss=0.2754, pruned_loss=0.0593, over 4808.00 frames. ], tot_loss[loss=0.1677, simple_loss=0.241, pruned_loss=0.04719, over 954754.69 frames. ], batch size: 40, lr: 2.96e-03, grad_scale: 32.0 2023-04-27 23:56:01,858 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=146905.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:56:33,815 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.3683, 1.5462, 1.4192, 1.6978, 1.6532, 1.7419, 1.4171, 2.8313], device='cuda:1'), covar=tensor([0.0553, 0.0696, 0.0695, 0.1073, 0.0545, 0.0592, 0.0697, 0.0227], device='cuda:1'), in_proj_covar=tensor([0.0037, 0.0039, 0.0039, 0.0043, 0.0040, 0.0037, 0.0038, 0.0055], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0013, 0.0013, 0.0015, 0.0014, 0.0013, 0.0013, 0.0016], device='cuda:1') 2023-04-27 23:56:36,153 INFO [finetune.py:976] (1/7) Epoch 26, batch 3750, loss[loss=0.176, simple_loss=0.2294, pruned_loss=0.0613, over 4264.00 frames. ], tot_loss[loss=0.1677, simple_loss=0.2411, pruned_loss=0.04719, over 953510.23 frames. ], batch size: 18, lr: 2.96e-03, grad_scale: 32.0 2023-04-27 23:56:44,461 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=146947.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:57:06,090 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.923e+01 1.552e+02 1.944e+02 2.307e+02 4.306e+02, threshold=3.888e+02, percent-clipped=2.0 2023-04-27 23:57:13,806 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=146982.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 23:57:20,816 INFO [finetune.py:976] (1/7) Epoch 26, batch 3800, loss[loss=0.1697, simple_loss=0.2493, pruned_loss=0.04503, over 4774.00 frames. ], tot_loss[loss=0.169, simple_loss=0.2428, pruned_loss=0.04756, over 955503.86 frames. ], batch size: 29, lr: 2.95e-03, grad_scale: 32.0 2023-04-27 23:57:51,785 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8603, 1.8198, 2.3096, 2.4411, 1.6437, 1.5153, 1.8457, 1.1260], device='cuda:1'), covar=tensor([0.0570, 0.0810, 0.0393, 0.0624, 0.0674, 0.1061, 0.0698, 0.0678], device='cuda:1'), in_proj_covar=tensor([0.0069, 0.0067, 0.0066, 0.0068, 0.0074, 0.0093, 0.0072, 0.0063], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-27 23:58:05,507 INFO [finetune.py:976] (1/7) Epoch 26, batch 3850, loss[loss=0.1807, simple_loss=0.249, pruned_loss=0.05623, over 4827.00 frames. ], tot_loss[loss=0.1681, simple_loss=0.2417, pruned_loss=0.04725, over 958148.52 frames. ], batch size: 38, lr: 2.95e-03, grad_scale: 32.0 2023-04-27 23:58:08,104 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.57 vs. limit=2.0 2023-04-27 23:58:23,781 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.013e+02 1.597e+02 1.804e+02 2.116e+02 3.825e+02, threshold=3.609e+02, percent-clipped=0.0 2023-04-27 23:58:39,180 INFO [finetune.py:976] (1/7) Epoch 26, batch 3900, loss[loss=0.1545, simple_loss=0.2295, pruned_loss=0.03976, over 4393.00 frames. ], tot_loss[loss=0.1657, simple_loss=0.2389, pruned_loss=0.0463, over 959182.69 frames. ], batch size: 65, lr: 2.95e-03, grad_scale: 32.0 2023-04-27 23:59:12,137 INFO [finetune.py:976] (1/7) Epoch 26, batch 3950, loss[loss=0.1972, simple_loss=0.2362, pruned_loss=0.07907, over 4243.00 frames. ], tot_loss[loss=0.163, simple_loss=0.2353, pruned_loss=0.04537, over 954672.10 frames. ], batch size: 18, lr: 2.95e-03, grad_scale: 32.0 2023-04-27 23:59:25,598 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-04-27 23:59:31,439 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.029e+02 1.596e+02 1.912e+02 2.259e+02 3.879e+02, threshold=3.825e+02, percent-clipped=1.0 2023-04-27 23:59:45,578 INFO [finetune.py:976] (1/7) Epoch 26, batch 4000, loss[loss=0.1684, simple_loss=0.248, pruned_loss=0.0444, over 4932.00 frames. ], tot_loss[loss=0.1634, simple_loss=0.2353, pruned_loss=0.04575, over 956872.02 frames. ], batch size: 33, lr: 2.95e-03, grad_scale: 32.0 2023-04-27 23:59:47,501 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=3.36 vs. limit=5.0 2023-04-27 23:59:51,632 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=147200.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:00:18,671 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=147241.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:00:19,197 INFO [finetune.py:976] (1/7) Epoch 26, batch 4050, loss[loss=0.1346, simple_loss=0.206, pruned_loss=0.0316, over 4711.00 frames. ], tot_loss[loss=0.166, simple_loss=0.239, pruned_loss=0.0465, over 956278.33 frames. ], batch size: 23, lr: 2.95e-03, grad_scale: 32.0 2023-04-28 00:00:23,219 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=147247.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:00:41,516 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6270, 1.2246, 4.2905, 4.0178, 3.6846, 4.0702, 3.9728, 3.7734], device='cuda:1'), covar=tensor([0.7351, 0.6367, 0.1066, 0.1818, 0.1327, 0.2081, 0.1809, 0.1634], device='cuda:1'), in_proj_covar=tensor([0.0309, 0.0308, 0.0406, 0.0409, 0.0347, 0.0414, 0.0318, 0.0364], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-28 00:00:44,479 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.189e+02 1.656e+02 1.880e+02 2.300e+02 3.890e+02, threshold=3.760e+02, percent-clipped=1.0 2023-04-28 00:00:55,648 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=147282.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:01:10,422 INFO [finetune.py:976] (1/7) Epoch 26, batch 4100, loss[loss=0.2008, simple_loss=0.2766, pruned_loss=0.06254, over 4826.00 frames. ], tot_loss[loss=0.1677, simple_loss=0.2411, pruned_loss=0.04716, over 955249.92 frames. ], batch size: 51, lr: 2.95e-03, grad_scale: 32.0 2023-04-28 00:01:12,318 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=147295.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:01:23,316 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=147302.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:01:34,373 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=147310.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:01:54,622 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=147330.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:02:13,693 INFO [finetune.py:976] (1/7) Epoch 26, batch 4150, loss[loss=0.1774, simple_loss=0.2611, pruned_loss=0.04682, over 4898.00 frames. ], tot_loss[loss=0.1687, simple_loss=0.2426, pruned_loss=0.04737, over 956788.51 frames. ], batch size: 46, lr: 2.95e-03, grad_scale: 32.0 2023-04-28 00:02:56,063 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=147371.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:02:56,541 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.010e+02 1.642e+02 1.885e+02 2.342e+02 4.093e+02, threshold=3.770e+02, percent-clipped=2.0 2023-04-28 00:03:09,338 INFO [finetune.py:976] (1/7) Epoch 26, batch 4200, loss[loss=0.1496, simple_loss=0.2339, pruned_loss=0.03263, over 4799.00 frames. ], tot_loss[loss=0.1685, simple_loss=0.2428, pruned_loss=0.04708, over 956673.17 frames. ], batch size: 55, lr: 2.95e-03, grad_scale: 32.0 2023-04-28 00:03:21,114 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=147407.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:03:27,742 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-04-28 00:03:37,863 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=147434.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:03:42,685 INFO [finetune.py:976] (1/7) Epoch 26, batch 4250, loss[loss=0.1779, simple_loss=0.2477, pruned_loss=0.05401, over 4936.00 frames. ], tot_loss[loss=0.1678, simple_loss=0.2413, pruned_loss=0.04717, over 958195.27 frames. ], batch size: 33, lr: 2.95e-03, grad_scale: 32.0 2023-04-28 00:03:56,717 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.68 vs. limit=2.0 2023-04-28 00:04:01,789 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=147468.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:04:04,153 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 8.178e+01 1.415e+02 1.781e+02 2.171e+02 8.734e+02, threshold=3.562e+02, percent-clipped=6.0 2023-04-28 00:04:06,101 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.1797, 1.8572, 2.0525, 2.3878, 2.3897, 1.9473, 1.6997, 2.2889], device='cuda:1'), covar=tensor([0.0768, 0.1183, 0.0806, 0.0579, 0.0593, 0.0900, 0.0786, 0.0546], device='cuda:1'), in_proj_covar=tensor([0.0184, 0.0202, 0.0184, 0.0171, 0.0177, 0.0178, 0.0150, 0.0178], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-28 00:04:16,303 INFO [finetune.py:976] (1/7) Epoch 26, batch 4300, loss[loss=0.1421, simple_loss=0.215, pruned_loss=0.03459, over 4777.00 frames. ], tot_loss[loss=0.1654, simple_loss=0.2381, pruned_loss=0.04638, over 957243.34 frames. ], batch size: 26, lr: 2.95e-03, grad_scale: 32.0 2023-04-28 00:04:16,540 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-04-28 00:04:18,256 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=147495.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:04:22,178 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=147500.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:04:49,577 INFO [finetune.py:976] (1/7) Epoch 26, batch 4350, loss[loss=0.1504, simple_loss=0.2155, pruned_loss=0.04263, over 4303.00 frames. ], tot_loss[loss=0.1614, simple_loss=0.2338, pruned_loss=0.0445, over 955585.43 frames. ], batch size: 65, lr: 2.95e-03, grad_scale: 32.0 2023-04-28 00:04:53,339 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=147548.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:05:01,645 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-04-28 00:05:06,781 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([3.0042, 2.4349, 2.1828, 2.3311, 2.1173, 2.2713, 2.3077, 2.2007], device='cuda:1'), covar=tensor([0.3931, 0.5566, 0.5175, 0.4417, 0.5884, 0.6800, 0.6336, 0.5569], device='cuda:1'), in_proj_covar=tensor([0.0341, 0.0376, 0.0329, 0.0341, 0.0351, 0.0396, 0.0361, 0.0332], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 00:05:10,811 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.116e+01 1.521e+02 1.686e+02 2.044e+02 4.229e+02, threshold=3.372e+02, percent-clipped=1.0 2023-04-28 00:05:23,017 INFO [finetune.py:976] (1/7) Epoch 26, batch 4400, loss[loss=0.1683, simple_loss=0.2493, pruned_loss=0.04368, over 4907.00 frames. ], tot_loss[loss=0.1624, simple_loss=0.2353, pruned_loss=0.04477, over 957409.81 frames. ], batch size: 37, lr: 2.95e-03, grad_scale: 32.0 2023-04-28 00:05:26,143 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=147597.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:05:41,270 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-04-28 00:05:56,085 INFO [finetune.py:976] (1/7) Epoch 26, batch 4450, loss[loss=0.2329, simple_loss=0.289, pruned_loss=0.08835, over 4904.00 frames. ], tot_loss[loss=0.1669, simple_loss=0.2401, pruned_loss=0.04683, over 958282.87 frames. ], batch size: 35, lr: 2.95e-03, grad_scale: 32.0 2023-04-28 00:06:17,527 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=147659.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:06:22,292 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=147666.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:06:31,481 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.062e+02 1.516e+02 1.781e+02 2.208e+02 3.417e+02, threshold=3.563e+02, percent-clipped=1.0 2023-04-28 00:06:42,536 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=3.63 vs. limit=5.0 2023-04-28 00:06:44,115 INFO [finetune.py:976] (1/7) Epoch 26, batch 4500, loss[loss=0.17, simple_loss=0.2547, pruned_loss=0.04264, over 4896.00 frames. ], tot_loss[loss=0.1683, simple_loss=0.242, pruned_loss=0.04733, over 959913.91 frames. ], batch size: 35, lr: 2.95e-03, grad_scale: 32.0 2023-04-28 00:07:04,375 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6031, 1.3982, 4.3048, 4.0193, 3.7743, 4.0747, 3.9938, 3.8022], device='cuda:1'), covar=tensor([0.6905, 0.5587, 0.1039, 0.1736, 0.1141, 0.1554, 0.1625, 0.1492], device='cuda:1'), in_proj_covar=tensor([0.0310, 0.0310, 0.0409, 0.0410, 0.0349, 0.0414, 0.0319, 0.0365], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-28 00:07:17,728 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=147720.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:07:37,055 INFO [finetune.py:976] (1/7) Epoch 26, batch 4550, loss[loss=0.1912, simple_loss=0.2639, pruned_loss=0.05925, over 4860.00 frames. ], tot_loss[loss=0.1694, simple_loss=0.2432, pruned_loss=0.04784, over 959575.47 frames. ], batch size: 34, lr: 2.95e-03, grad_scale: 32.0 2023-04-28 00:08:01,006 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=147763.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:08:14,011 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.021e+02 1.503e+02 1.815e+02 2.151e+02 5.912e+02, threshold=3.631e+02, percent-clipped=1.0 2023-04-28 00:08:14,120 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.1119, 2.6958, 1.0843, 1.4567, 2.2107, 1.2507, 3.5325, 1.6725], device='cuda:1'), covar=tensor([0.0680, 0.0677, 0.0857, 0.1220, 0.0497, 0.1024, 0.0224, 0.0653], device='cuda:1'), in_proj_covar=tensor([0.0050, 0.0064, 0.0047, 0.0046, 0.0049, 0.0051, 0.0073, 0.0051], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:1') 2023-04-28 00:08:30,288 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0903, 1.9620, 2.1321, 2.5513, 2.5197, 2.0301, 1.5614, 2.3430], device='cuda:1'), covar=tensor([0.0842, 0.1117, 0.0785, 0.0531, 0.0590, 0.0879, 0.0887, 0.0534], device='cuda:1'), in_proj_covar=tensor([0.0184, 0.0201, 0.0183, 0.0170, 0.0176, 0.0177, 0.0151, 0.0177], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-28 00:08:31,435 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=147790.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:08:32,261 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.53 vs. limit=2.0 2023-04-28 00:08:32,580 INFO [finetune.py:976] (1/7) Epoch 26, batch 4600, loss[loss=0.1437, simple_loss=0.2188, pruned_loss=0.03426, over 4822.00 frames. ], tot_loss[loss=0.1673, simple_loss=0.2413, pruned_loss=0.04663, over 958645.79 frames. ], batch size: 38, lr: 2.95e-03, grad_scale: 32.0 2023-04-28 00:08:47,317 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=147815.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:09:06,109 INFO [finetune.py:976] (1/7) Epoch 26, batch 4650, loss[loss=0.1298, simple_loss=0.2014, pruned_loss=0.02912, over 4825.00 frames. ], tot_loss[loss=0.1655, simple_loss=0.2387, pruned_loss=0.04619, over 958279.13 frames. ], batch size: 33, lr: 2.95e-03, grad_scale: 32.0 2023-04-28 00:09:15,434 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=147857.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:09:20,993 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.2474, 1.4638, 1.7595, 1.8314, 1.7717, 1.8538, 1.8094, 1.7779], device='cuda:1'), covar=tensor([0.3631, 0.4935, 0.3938, 0.3948, 0.4818, 0.6549, 0.4482, 0.4389], device='cuda:1'), in_proj_covar=tensor([0.0340, 0.0374, 0.0328, 0.0339, 0.0350, 0.0395, 0.0360, 0.0332], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 00:09:25,962 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 7.854e+01 1.470e+02 1.652e+02 1.982e+02 3.804e+02, threshold=3.304e+02, percent-clipped=1.0 2023-04-28 00:09:29,058 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=147876.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:09:40,051 INFO [finetune.py:976] (1/7) Epoch 26, batch 4700, loss[loss=0.1427, simple_loss=0.2062, pruned_loss=0.03961, over 4777.00 frames. ], tot_loss[loss=0.1635, simple_loss=0.2359, pruned_loss=0.0455, over 957222.64 frames. ], batch size: 26, lr: 2.95e-03, grad_scale: 32.0 2023-04-28 00:09:43,201 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=147897.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:09:51,618 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=147911.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:09:55,876 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=147918.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:10:07,687 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-04-28 00:10:12,811 INFO [finetune.py:976] (1/7) Epoch 26, batch 4750, loss[loss=0.1645, simple_loss=0.2419, pruned_loss=0.04358, over 4896.00 frames. ], tot_loss[loss=0.1623, simple_loss=0.2344, pruned_loss=0.04504, over 956213.18 frames. ], batch size: 32, lr: 2.95e-03, grad_scale: 32.0 2023-04-28 00:10:15,201 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=147945.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:10:28,083 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=147966.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:10:31,602 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.017e+02 1.547e+02 1.796e+02 2.219e+02 3.890e+02, threshold=3.592e+02, percent-clipped=1.0 2023-04-28 00:10:31,715 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=147972.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:10:46,577 INFO [finetune.py:976] (1/7) Epoch 26, batch 4800, loss[loss=0.1425, simple_loss=0.2372, pruned_loss=0.02392, over 4811.00 frames. ], tot_loss[loss=0.1653, simple_loss=0.238, pruned_loss=0.04631, over 954487.24 frames. ], batch size: 45, lr: 2.95e-03, grad_scale: 32.0 2023-04-28 00:10:50,905 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.4006, 1.9687, 2.3324, 2.6824, 2.3016, 1.9119, 1.7070, 2.0506], device='cuda:1'), covar=tensor([0.3388, 0.3240, 0.1661, 0.2136, 0.2656, 0.2690, 0.3716, 0.1946], device='cuda:1'), in_proj_covar=tensor([0.0295, 0.0247, 0.0229, 0.0315, 0.0222, 0.0236, 0.0229, 0.0188], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-28 00:10:54,035 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4813, 1.6674, 1.4915, 1.9261, 1.7848, 2.0445, 1.5495, 3.7555], device='cuda:1'), covar=tensor([0.0528, 0.0784, 0.0778, 0.1058, 0.0607, 0.0452, 0.0713, 0.0148], device='cuda:1'), in_proj_covar=tensor([0.0037, 0.0039, 0.0039, 0.0042, 0.0040, 0.0037, 0.0038, 0.0054], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0013, 0.0013, 0.0015, 0.0014, 0.0013, 0.0013, 0.0016], device='cuda:1') 2023-04-28 00:11:01,948 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=148014.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:11:02,576 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=148015.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:11:05,221 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-04-28 00:11:21,740 INFO [finetune.py:976] (1/7) Epoch 26, batch 4850, loss[loss=0.1926, simple_loss=0.2823, pruned_loss=0.05145, over 4915.00 frames. ], tot_loss[loss=0.1682, simple_loss=0.2415, pruned_loss=0.0475, over 953687.75 frames. ], batch size: 38, lr: 2.95e-03, grad_scale: 32.0 2023-04-28 00:11:44,635 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=148063.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:11:55,269 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.111e+02 1.622e+02 1.940e+02 2.340e+02 4.108e+02, threshold=3.881e+02, percent-clipped=3.0 2023-04-28 00:12:18,039 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=148090.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:12:24,897 INFO [finetune.py:976] (1/7) Epoch 26, batch 4900, loss[loss=0.2218, simple_loss=0.2843, pruned_loss=0.0797, over 4921.00 frames. ], tot_loss[loss=0.1697, simple_loss=0.2433, pruned_loss=0.04806, over 954272.82 frames. ], batch size: 33, lr: 2.95e-03, grad_scale: 32.0 2023-04-28 00:12:35,327 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.14 vs. limit=2.0 2023-04-28 00:12:47,896 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=148111.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:12:50,542 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.68 vs. limit=2.0 2023-04-28 00:12:51,656 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.3191, 1.5579, 1.4315, 1.8022, 1.6390, 1.7812, 1.4334, 3.1139], device='cuda:1'), covar=tensor([0.0591, 0.0770, 0.0798, 0.1148, 0.0616, 0.0445, 0.0724, 0.0187], device='cuda:1'), in_proj_covar=tensor([0.0037, 0.0039, 0.0039, 0.0043, 0.0040, 0.0037, 0.0038, 0.0055], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0013, 0.0013, 0.0015, 0.0014, 0.0013, 0.0013, 0.0016], device='cuda:1') 2023-04-28 00:13:22,194 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=148138.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:13:30,216 INFO [finetune.py:976] (1/7) Epoch 26, batch 4950, loss[loss=0.1446, simple_loss=0.2079, pruned_loss=0.04062, over 4119.00 frames. ], tot_loss[loss=0.1701, simple_loss=0.244, pruned_loss=0.04807, over 953646.30 frames. ], batch size: 17, lr: 2.95e-03, grad_scale: 32.0 2023-04-28 00:13:52,203 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=148171.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:13:52,733 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.839e+01 1.452e+02 1.801e+02 2.197e+02 4.892e+02, threshold=3.601e+02, percent-clipped=1.0 2023-04-28 00:14:06,564 INFO [finetune.py:976] (1/7) Epoch 26, batch 5000, loss[loss=0.1446, simple_loss=0.2192, pruned_loss=0.03501, over 4892.00 frames. ], tot_loss[loss=0.1686, simple_loss=0.2424, pruned_loss=0.04743, over 950014.96 frames. ], batch size: 43, lr: 2.95e-03, grad_scale: 32.0 2023-04-28 00:14:08,299 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.4920, 2.9898, 0.8801, 1.7266, 2.1751, 1.4844, 4.0742, 2.0315], device='cuda:1'), covar=tensor([0.0626, 0.0775, 0.0948, 0.1233, 0.0559, 0.0978, 0.0200, 0.0583], device='cuda:1'), in_proj_covar=tensor([0.0051, 0.0064, 0.0047, 0.0046, 0.0050, 0.0051, 0.0073, 0.0051], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:1') 2023-04-28 00:14:21,041 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=148213.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:14:39,768 INFO [finetune.py:976] (1/7) Epoch 26, batch 5050, loss[loss=0.2013, simple_loss=0.2647, pruned_loss=0.06891, over 4918.00 frames. ], tot_loss[loss=0.168, simple_loss=0.2407, pruned_loss=0.0477, over 952649.35 frames. ], batch size: 37, lr: 2.95e-03, grad_scale: 32.0 2023-04-28 00:14:56,594 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=148267.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:14:59,522 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.057e+01 1.422e+02 1.750e+02 2.059e+02 5.482e+02, threshold=3.501e+02, percent-clipped=2.0 2023-04-28 00:15:08,791 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=148287.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:15:11,732 INFO [finetune.py:976] (1/7) Epoch 26, batch 5100, loss[loss=0.1666, simple_loss=0.2291, pruned_loss=0.05201, over 3953.00 frames. ], tot_loss[loss=0.1661, simple_loss=0.2384, pruned_loss=0.04688, over 954327.46 frames. ], batch size: 17, lr: 2.95e-03, grad_scale: 64.0 2023-04-28 00:15:28,678 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=148315.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:15:42,747 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.8054, 2.0624, 2.0568, 2.0919, 1.9262, 2.0131, 2.1751, 2.0817], device='cuda:1'), covar=tensor([0.4392, 0.6059, 0.5056, 0.4681, 0.5930, 0.7190, 0.5455, 0.5153], device='cuda:1'), in_proj_covar=tensor([0.0339, 0.0374, 0.0329, 0.0339, 0.0350, 0.0394, 0.0360, 0.0332], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 00:15:45,037 INFO [finetune.py:976] (1/7) Epoch 26, batch 5150, loss[loss=0.1763, simple_loss=0.2538, pruned_loss=0.04945, over 4809.00 frames. ], tot_loss[loss=0.1665, simple_loss=0.2389, pruned_loss=0.04709, over 954078.69 frames. ], batch size: 38, lr: 2.95e-03, grad_scale: 64.0 2023-04-28 00:15:49,310 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=148348.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:15:57,302 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.2502, 1.7238, 2.1502, 2.6787, 2.1460, 1.6752, 1.4535, 1.9718], device='cuda:1'), covar=tensor([0.3011, 0.2954, 0.1617, 0.2024, 0.2409, 0.2614, 0.3776, 0.1896], device='cuda:1'), in_proj_covar=tensor([0.0292, 0.0245, 0.0228, 0.0313, 0.0221, 0.0235, 0.0228, 0.0186], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-28 00:15:58,009 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 2023-04-28 00:16:00,242 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=148363.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:16:06,186 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.267e+01 1.646e+02 1.916e+02 2.227e+02 4.429e+02, threshold=3.832e+02, percent-clipped=2.0 2023-04-28 00:16:18,411 INFO [finetune.py:976] (1/7) Epoch 26, batch 5200, loss[loss=0.1626, simple_loss=0.231, pruned_loss=0.04715, over 4936.00 frames. ], tot_loss[loss=0.1696, simple_loss=0.2422, pruned_loss=0.04851, over 953090.16 frames. ], batch size: 33, lr: 2.95e-03, grad_scale: 64.0 2023-04-28 00:16:22,663 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6628, 1.7946, 0.9286, 1.3464, 1.8181, 1.5171, 1.3966, 1.4704], device='cuda:1'), covar=tensor([0.0523, 0.0378, 0.0362, 0.0608, 0.0293, 0.0551, 0.0589, 0.0596], device='cuda:1'), in_proj_covar=tensor([0.0028, 0.0023, 0.0021, 0.0028, 0.0019, 0.0027, 0.0027, 0.0028], device='cuda:1'), out_proj_covar=tensor([0.0052, 0.0046, 0.0038, 0.0053, 0.0039, 0.0050, 0.0050, 0.0052], device='cuda:1') 2023-04-28 00:16:44,049 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=148429.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:16:51,852 INFO [finetune.py:976] (1/7) Epoch 26, batch 5250, loss[loss=0.1137, simple_loss=0.1876, pruned_loss=0.01984, over 4197.00 frames. ], tot_loss[loss=0.1697, simple_loss=0.243, pruned_loss=0.0482, over 952684.53 frames. ], batch size: 18, lr: 2.94e-03, grad_scale: 64.0 2023-04-28 00:16:54,382 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.1760, 1.5461, 1.4247, 1.7899, 1.6490, 1.8796, 1.4672, 3.4550], device='cuda:1'), covar=tensor([0.0627, 0.0825, 0.0831, 0.1198, 0.0653, 0.0513, 0.0749, 0.0139], device='cuda:1'), in_proj_covar=tensor([0.0037, 0.0039, 0.0039, 0.0043, 0.0040, 0.0038, 0.0038, 0.0055], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0013, 0.0013, 0.0015, 0.0014, 0.0013, 0.0013, 0.0016], device='cuda:1') 2023-04-28 00:17:12,110 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=148471.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:17:12,616 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.069e+02 1.667e+02 1.923e+02 2.369e+02 3.667e+02, threshold=3.847e+02, percent-clipped=0.0 2023-04-28 00:17:16,978 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.1203, 2.4737, 0.7666, 1.4792, 1.5106, 1.8325, 1.5726, 0.8131], device='cuda:1'), covar=tensor([0.1315, 0.0983, 0.1633, 0.1141, 0.0997, 0.0844, 0.1406, 0.1669], device='cuda:1'), in_proj_covar=tensor([0.0116, 0.0235, 0.0134, 0.0120, 0.0130, 0.0150, 0.0116, 0.0117], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-28 00:17:23,732 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=148490.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:17:24,853 INFO [finetune.py:976] (1/7) Epoch 26, batch 5300, loss[loss=0.1674, simple_loss=0.2442, pruned_loss=0.04528, over 4885.00 frames. ], tot_loss[loss=0.1702, simple_loss=0.244, pruned_loss=0.04817, over 955114.15 frames. ], batch size: 43, lr: 2.94e-03, grad_scale: 64.0 2023-04-28 00:17:28,534 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=148498.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:17:31,799 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-04-28 00:17:52,367 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=148513.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:17:53,159 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-04-28 00:17:56,052 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=148519.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:18:24,823 INFO [finetune.py:976] (1/7) Epoch 26, batch 5350, loss[loss=0.1848, simple_loss=0.2477, pruned_loss=0.06098, over 4868.00 frames. ], tot_loss[loss=0.1693, simple_loss=0.2435, pruned_loss=0.04752, over 956616.14 frames. ], batch size: 34, lr: 2.94e-03, grad_scale: 64.0 2023-04-28 00:18:36,144 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-04-28 00:18:46,001 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=148559.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:18:47,175 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=148561.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:18:56,334 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=148567.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:18:59,257 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.079e+02 1.519e+02 1.727e+02 2.044e+02 3.466e+02, threshold=3.453e+02, percent-clipped=0.0 2023-04-28 00:19:22,269 INFO [finetune.py:976] (1/7) Epoch 26, batch 5400, loss[loss=0.1426, simple_loss=0.207, pruned_loss=0.03909, over 4815.00 frames. ], tot_loss[loss=0.1674, simple_loss=0.2407, pruned_loss=0.04705, over 954938.97 frames. ], batch size: 25, lr: 2.94e-03, grad_scale: 64.0 2023-04-28 00:19:33,059 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2023-04-28 00:19:53,633 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=148615.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:20:04,626 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.58 vs. limit=2.0 2023-04-28 00:20:05,821 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=3.43 vs. limit=5.0 2023-04-28 00:20:11,714 INFO [finetune.py:976] (1/7) Epoch 26, batch 5450, loss[loss=0.1401, simple_loss=0.2144, pruned_loss=0.03293, over 4779.00 frames. ], tot_loss[loss=0.1648, simple_loss=0.2372, pruned_loss=0.04619, over 954331.67 frames. ], batch size: 29, lr: 2.94e-03, grad_scale: 64.0 2023-04-28 00:20:12,387 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=148643.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:20:16,107 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.9167, 2.2763, 2.1947, 2.3053, 2.2198, 2.1424, 2.2297, 2.1244], device='cuda:1'), covar=tensor([0.3955, 0.5599, 0.4699, 0.4392, 0.5475, 0.6670, 0.5815, 0.5565], device='cuda:1'), in_proj_covar=tensor([0.0339, 0.0373, 0.0328, 0.0338, 0.0349, 0.0394, 0.0360, 0.0332], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 00:20:17,256 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=148651.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:20:31,290 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.056e+02 1.574e+02 1.926e+02 2.349e+02 4.613e+02, threshold=3.853e+02, percent-clipped=6.0 2023-04-28 00:20:45,353 INFO [finetune.py:976] (1/7) Epoch 26, batch 5500, loss[loss=0.1884, simple_loss=0.262, pruned_loss=0.05743, over 4816.00 frames. ], tot_loss[loss=0.1634, simple_loss=0.2354, pruned_loss=0.04568, over 955362.86 frames. ], batch size: 40, lr: 2.94e-03, grad_scale: 64.0 2023-04-28 00:20:57,696 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=148712.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:21:00,134 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=148716.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:21:18,755 INFO [finetune.py:976] (1/7) Epoch 26, batch 5550, loss[loss=0.1618, simple_loss=0.2516, pruned_loss=0.03599, over 4847.00 frames. ], tot_loss[loss=0.1642, simple_loss=0.2364, pruned_loss=0.04597, over 954190.12 frames. ], batch size: 49, lr: 2.94e-03, grad_scale: 64.0 2023-04-28 00:21:38,040 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.670e+01 1.487e+02 1.767e+02 2.141e+02 3.989e+02, threshold=3.535e+02, percent-clipped=1.0 2023-04-28 00:21:41,045 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=148777.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:21:45,566 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=148785.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:21:49,593 INFO [finetune.py:976] (1/7) Epoch 26, batch 5600, loss[loss=0.1148, simple_loss=0.1872, pruned_loss=0.02118, over 4339.00 frames. ], tot_loss[loss=0.1665, simple_loss=0.2401, pruned_loss=0.04644, over 955538.73 frames. ], batch size: 19, lr: 2.94e-03, grad_scale: 64.0 2023-04-28 00:22:16,097 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8936, 2.3592, 2.3414, 2.7011, 2.5766, 2.4855, 2.3809, 4.9303], device='cuda:1'), covar=tensor([0.0443, 0.0628, 0.0653, 0.0946, 0.0501, 0.0448, 0.0551, 0.0118], device='cuda:1'), in_proj_covar=tensor([0.0037, 0.0039, 0.0039, 0.0043, 0.0040, 0.0038, 0.0038, 0.0055], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0013, 0.0014, 0.0015, 0.0014, 0.0013, 0.0013, 0.0016], device='cuda:1') 2023-04-28 00:22:20,662 INFO [finetune.py:976] (1/7) Epoch 26, batch 5650, loss[loss=0.1712, simple_loss=0.2454, pruned_loss=0.04848, over 4810.00 frames. ], tot_loss[loss=0.1682, simple_loss=0.2428, pruned_loss=0.04678, over 953305.31 frames. ], batch size: 39, lr: 2.94e-03, grad_scale: 32.0 2023-04-28 00:22:23,060 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0180, 2.5888, 1.0853, 1.3550, 2.1879, 1.1286, 3.5342, 1.8813], device='cuda:1'), covar=tensor([0.0751, 0.0618, 0.0788, 0.1335, 0.0493, 0.1105, 0.0203, 0.0618], device='cuda:1'), in_proj_covar=tensor([0.0051, 0.0065, 0.0048, 0.0046, 0.0050, 0.0052, 0.0074, 0.0051], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:1') 2023-04-28 00:22:27,776 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=148854.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:22:38,504 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([4.2341, 4.2044, 3.2079, 4.8264, 4.2995, 4.2269, 2.5032, 4.1473], device='cuda:1'), covar=tensor([0.1735, 0.1210, 0.2570, 0.1379, 0.3048, 0.1916, 0.4765, 0.2252], device='cuda:1'), in_proj_covar=tensor([0.0246, 0.0222, 0.0254, 0.0305, 0.0300, 0.0250, 0.0275, 0.0277], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 00:22:39,029 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.125e+02 1.517e+02 1.815e+02 2.254e+02 4.751e+02, threshold=3.630e+02, percent-clipped=7.0 2023-04-28 00:22:48,560 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([3.2672, 3.2681, 2.5285, 3.8023, 3.3047, 3.3121, 1.5212, 3.1906], device='cuda:1'), covar=tensor([0.2211, 0.1364, 0.3274, 0.2083, 0.3131, 0.2115, 0.6401, 0.2811], device='cuda:1'), in_proj_covar=tensor([0.0246, 0.0221, 0.0253, 0.0305, 0.0300, 0.0249, 0.0275, 0.0276], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 00:22:50,263 INFO [finetune.py:976] (1/7) Epoch 26, batch 5700, loss[loss=0.1749, simple_loss=0.2142, pruned_loss=0.06778, over 4060.00 frames. ], tot_loss[loss=0.1646, simple_loss=0.2374, pruned_loss=0.04587, over 930963.00 frames. ], batch size: 17, lr: 2.94e-03, grad_scale: 32.0 2023-04-28 00:23:20,154 INFO [finetune.py:976] (1/7) Epoch 27, batch 0, loss[loss=0.1632, simple_loss=0.2432, pruned_loss=0.04163, over 4495.00 frames. ], tot_loss[loss=0.1632, simple_loss=0.2432, pruned_loss=0.04163, over 4495.00 frames. ], batch size: 19, lr: 2.94e-03, grad_scale: 32.0 2023-04-28 00:23:20,154 INFO [finetune.py:1001] (1/7) Computing validation loss 2023-04-28 00:23:31,734 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.2528, 2.5052, 1.1179, 1.4944, 2.0002, 1.3467, 3.0216, 1.7414], device='cuda:1'), covar=tensor([0.0566, 0.0769, 0.0636, 0.1058, 0.0352, 0.0768, 0.0285, 0.0497], device='cuda:1'), in_proj_covar=tensor([0.0051, 0.0064, 0.0047, 0.0046, 0.0050, 0.0051, 0.0073, 0.0051], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:1') 2023-04-28 00:23:32,206 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4176, 1.3190, 1.6526, 1.6652, 1.2890, 1.2574, 1.3513, 0.8281], device='cuda:1'), covar=tensor([0.0500, 0.0609, 0.0401, 0.0483, 0.0789, 0.1171, 0.0479, 0.0550], device='cuda:1'), in_proj_covar=tensor([0.0070, 0.0067, 0.0066, 0.0068, 0.0074, 0.0094, 0.0072, 0.0063], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-28 00:23:41,724 INFO [finetune.py:1010] (1/7) Epoch 27, validation: loss=0.1548, simple_loss=0.2237, pruned_loss=0.04298, over 2265189.00 frames. 2023-04-28 00:23:41,724 INFO [finetune.py:1011] (1/7) Maximum memory allocated so far is 6391MB 2023-04-28 00:24:11,140 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=148943.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:24:41,545 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-04-28 00:24:43,146 INFO [finetune.py:976] (1/7) Epoch 27, batch 50, loss[loss=0.1784, simple_loss=0.2522, pruned_loss=0.05234, over 4756.00 frames. ], tot_loss[loss=0.1682, simple_loss=0.2429, pruned_loss=0.04676, over 215740.87 frames. ], batch size: 27, lr: 2.94e-03, grad_scale: 32.0 2023-04-28 00:24:45,491 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.880e+01 1.482e+02 1.782e+02 2.132e+02 4.593e+02, threshold=3.564e+02, percent-clipped=2.0 2023-04-28 00:24:53,728 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6904, 1.3791, 4.5363, 4.2824, 3.9370, 4.3954, 4.1924, 3.9765], device='cuda:1'), covar=tensor([0.7155, 0.5939, 0.1100, 0.1708, 0.1102, 0.1456, 0.1543, 0.1515], device='cuda:1'), in_proj_covar=tensor([0.0312, 0.0310, 0.0410, 0.0411, 0.0349, 0.0415, 0.0320, 0.0367], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-28 00:25:07,485 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=148991.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:25:28,977 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=149007.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:25:37,829 INFO [finetune.py:976] (1/7) Epoch 27, batch 100, loss[loss=0.1475, simple_loss=0.221, pruned_loss=0.03701, over 4755.00 frames. ], tot_loss[loss=0.162, simple_loss=0.2359, pruned_loss=0.044, over 381748.04 frames. ], batch size: 54, lr: 2.94e-03, grad_scale: 32.0 2023-04-28 00:25:48,516 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8651, 1.7031, 1.9944, 2.1967, 2.2793, 1.7590, 1.5918, 2.1070], device='cuda:1'), covar=tensor([0.0774, 0.1077, 0.0657, 0.0529, 0.0556, 0.0881, 0.0708, 0.0523], device='cuda:1'), in_proj_covar=tensor([0.0184, 0.0200, 0.0182, 0.0169, 0.0176, 0.0175, 0.0150, 0.0177], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-28 00:25:58,869 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([4.0766, 4.0084, 2.8771, 4.6445, 4.0956, 3.9619, 1.5050, 3.9400], device='cuda:1'), covar=tensor([0.1602, 0.1263, 0.2998, 0.1470, 0.2655, 0.1787, 0.6113, 0.2167], device='cuda:1'), in_proj_covar=tensor([0.0246, 0.0222, 0.0255, 0.0306, 0.0301, 0.0250, 0.0276, 0.0276], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 00:26:00,972 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.26 vs. limit=5.0 2023-04-28 00:26:11,901 INFO [finetune.py:976] (1/7) Epoch 27, batch 150, loss[loss=0.1356, simple_loss=0.204, pruned_loss=0.03359, over 4821.00 frames. ], tot_loss[loss=0.1588, simple_loss=0.2315, pruned_loss=0.04308, over 510750.14 frames. ], batch size: 40, lr: 2.94e-03, grad_scale: 32.0 2023-04-28 00:26:13,187 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=149072.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:26:13,726 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.057e+02 1.439e+02 1.708e+02 1.990e+02 3.271e+02, threshold=3.417e+02, percent-clipped=0.0 2023-04-28 00:26:18,514 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.31 vs. limit=5.0 2023-04-28 00:26:22,647 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=149085.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:26:45,180 INFO [finetune.py:976] (1/7) Epoch 27, batch 200, loss[loss=0.1174, simple_loss=0.1886, pruned_loss=0.02304, over 4827.00 frames. ], tot_loss[loss=0.1565, simple_loss=0.2286, pruned_loss=0.04218, over 609362.67 frames. ], batch size: 30, lr: 2.94e-03, grad_scale: 32.0 2023-04-28 00:26:55,606 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=149133.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:27:08,547 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=149154.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:27:18,793 INFO [finetune.py:976] (1/7) Epoch 27, batch 250, loss[loss=0.1587, simple_loss=0.2385, pruned_loss=0.0394, over 4748.00 frames. ], tot_loss[loss=0.1609, simple_loss=0.2331, pruned_loss=0.04434, over 685839.33 frames. ], batch size: 27, lr: 2.94e-03, grad_scale: 32.0 2023-04-28 00:27:21,649 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.056e+02 1.484e+02 1.804e+02 2.162e+02 4.404e+02, threshold=3.609e+02, percent-clipped=2.0 2023-04-28 00:27:29,595 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5906, 1.4412, 1.8977, 1.9745, 1.4373, 1.3072, 1.6090, 0.9191], device='cuda:1'), covar=tensor([0.0509, 0.0727, 0.0420, 0.0524, 0.0728, 0.1140, 0.0562, 0.0640], device='cuda:1'), in_proj_covar=tensor([0.0070, 0.0067, 0.0066, 0.0069, 0.0075, 0.0094, 0.0072, 0.0063], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-28 00:27:34,353 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-04-28 00:27:35,643 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.91 vs. limit=5.0 2023-04-28 00:27:41,051 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=149202.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:27:50,687 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-04-28 00:27:52,339 INFO [finetune.py:976] (1/7) Epoch 27, batch 300, loss[loss=0.2117, simple_loss=0.2763, pruned_loss=0.07354, over 4823.00 frames. ], tot_loss[loss=0.167, simple_loss=0.2403, pruned_loss=0.04691, over 746634.57 frames. ], batch size: 40, lr: 2.94e-03, grad_scale: 32.0 2023-04-28 00:27:55,004 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-04-28 00:28:12,510 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.2885, 1.2928, 1.5789, 1.6038, 1.2651, 1.1517, 1.3905, 0.7883], device='cuda:1'), covar=tensor([0.0542, 0.0502, 0.0404, 0.0519, 0.0599, 0.0887, 0.0481, 0.0538], device='cuda:1'), in_proj_covar=tensor([0.0070, 0.0067, 0.0066, 0.0069, 0.0074, 0.0094, 0.0072, 0.0063], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-28 00:28:13,719 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4875, 1.5840, 1.4881, 1.9144, 1.8116, 2.0490, 1.6448, 4.1434], device='cuda:1'), covar=tensor([0.0508, 0.0807, 0.0765, 0.1135, 0.0610, 0.0543, 0.0699, 0.0104], device='cuda:1'), in_proj_covar=tensor([0.0037, 0.0039, 0.0039, 0.0042, 0.0040, 0.0037, 0.0038, 0.0055], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0013, 0.0013, 0.0015, 0.0014, 0.0013, 0.0013, 0.0016], device='cuda:1') 2023-04-28 00:28:22,261 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6699, 1.4468, 4.4304, 4.1554, 3.8315, 4.1867, 4.1001, 3.8847], device='cuda:1'), covar=tensor([0.7204, 0.5570, 0.1006, 0.1606, 0.1071, 0.1540, 0.1383, 0.1426], device='cuda:1'), in_proj_covar=tensor([0.0313, 0.0312, 0.0413, 0.0413, 0.0350, 0.0416, 0.0321, 0.0367], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-28 00:28:25,830 INFO [finetune.py:976] (1/7) Epoch 27, batch 350, loss[loss=0.174, simple_loss=0.2505, pruned_loss=0.04869, over 4825.00 frames. ], tot_loss[loss=0.1694, simple_loss=0.2433, pruned_loss=0.04779, over 794585.94 frames. ], batch size: 30, lr: 2.94e-03, grad_scale: 32.0 2023-04-28 00:28:28,122 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.225e+02 1.612e+02 1.887e+02 2.251e+02 5.949e+02, threshold=3.774e+02, percent-clipped=2.0 2023-04-28 00:28:41,705 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.2936, 1.7583, 2.1957, 2.7513, 2.1693, 1.7429, 1.4710, 2.0160], device='cuda:1'), covar=tensor([0.3248, 0.3179, 0.1725, 0.2036, 0.2725, 0.2702, 0.4072, 0.2027], device='cuda:1'), in_proj_covar=tensor([0.0291, 0.0245, 0.0227, 0.0313, 0.0220, 0.0233, 0.0227, 0.0186], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-28 00:28:51,970 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=149307.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:28:59,851 INFO [finetune.py:976] (1/7) Epoch 27, batch 400, loss[loss=0.1383, simple_loss=0.2251, pruned_loss=0.02569, over 4817.00 frames. ], tot_loss[loss=0.1692, simple_loss=0.2434, pruned_loss=0.04751, over 829217.99 frames. ], batch size: 38, lr: 2.94e-03, grad_scale: 32.0 2023-04-28 00:29:07,755 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4335, 1.1419, 1.2206, 1.1773, 1.5960, 1.2994, 1.1042, 1.1823], device='cuda:1'), covar=tensor([0.1483, 0.1087, 0.1503, 0.1316, 0.0671, 0.1302, 0.1683, 0.1913], device='cuda:1'), in_proj_covar=tensor([0.0315, 0.0308, 0.0351, 0.0287, 0.0326, 0.0306, 0.0299, 0.0374], device='cuda:1'), out_proj_covar=tensor([6.4305e-05, 6.3354e-05, 7.3556e-05, 5.7557e-05, 6.6794e-05, 6.3813e-05, 6.1891e-05, 7.9155e-05], device='cuda:1') 2023-04-28 00:29:08,526 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.94 vs. limit=5.0 2023-04-28 00:29:39,546 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=149355.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:29:39,586 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9214, 2.2731, 0.8892, 1.2606, 1.6507, 1.1800, 2.4954, 1.3983], device='cuda:1'), covar=tensor([0.0673, 0.0500, 0.0643, 0.1208, 0.0458, 0.1019, 0.0307, 0.0660], device='cuda:1'), in_proj_covar=tensor([0.0050, 0.0063, 0.0047, 0.0045, 0.0049, 0.0051, 0.0072, 0.0050], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0007, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:1') 2023-04-28 00:29:48,682 INFO [finetune.py:976] (1/7) Epoch 27, batch 450, loss[loss=0.1565, simple_loss=0.224, pruned_loss=0.04446, over 4699.00 frames. ], tot_loss[loss=0.1683, simple_loss=0.2423, pruned_loss=0.04717, over 856004.02 frames. ], batch size: 23, lr: 2.94e-03, grad_scale: 32.0 2023-04-28 00:29:50,015 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=149372.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:29:50,984 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.462e+01 1.496e+02 1.752e+02 2.050e+02 4.938e+02, threshold=3.505e+02, percent-clipped=1.0 2023-04-28 00:30:41,613 INFO [finetune.py:976] (1/7) Epoch 27, batch 500, loss[loss=0.1399, simple_loss=0.2221, pruned_loss=0.02882, over 4903.00 frames. ], tot_loss[loss=0.1654, simple_loss=0.2388, pruned_loss=0.04606, over 876267.83 frames. ], batch size: 36, lr: 2.93e-03, grad_scale: 32.0 2023-04-28 00:30:41,676 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=149420.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:31:43,468 INFO [finetune.py:976] (1/7) Epoch 27, batch 550, loss[loss=0.1883, simple_loss=0.2626, pruned_loss=0.05697, over 4815.00 frames. ], tot_loss[loss=0.1648, simple_loss=0.2372, pruned_loss=0.04616, over 892308.89 frames. ], batch size: 45, lr: 2.93e-03, grad_scale: 32.0 2023-04-28 00:31:45,296 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.135e+02 1.482e+02 1.685e+02 2.089e+02 3.058e+02, threshold=3.371e+02, percent-clipped=0.0 2023-04-28 00:31:59,891 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.71 vs. limit=2.0 2023-04-28 00:32:13,117 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=149514.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:32:16,577 INFO [finetune.py:976] (1/7) Epoch 27, batch 600, loss[loss=0.1752, simple_loss=0.2492, pruned_loss=0.05063, over 4898.00 frames. ], tot_loss[loss=0.1659, simple_loss=0.238, pruned_loss=0.04688, over 908410.46 frames. ], batch size: 32, lr: 2.93e-03, grad_scale: 32.0 2023-04-28 00:32:24,508 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.2433, 1.7082, 2.0586, 2.2288, 2.0030, 1.7006, 1.1957, 1.7328], device='cuda:1'), covar=tensor([0.3155, 0.2983, 0.1661, 0.2163, 0.2625, 0.2495, 0.4078, 0.1919], device='cuda:1'), in_proj_covar=tensor([0.0292, 0.0246, 0.0228, 0.0314, 0.0221, 0.0234, 0.0228, 0.0187], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-28 00:32:47,530 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=149565.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:32:50,408 INFO [finetune.py:976] (1/7) Epoch 27, batch 650, loss[loss=0.2023, simple_loss=0.2727, pruned_loss=0.06594, over 4848.00 frames. ], tot_loss[loss=0.1682, simple_loss=0.2411, pruned_loss=0.0476, over 918872.38 frames. ], batch size: 44, lr: 2.93e-03, grad_scale: 32.0 2023-04-28 00:32:52,256 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.202e+02 1.546e+02 1.916e+02 2.340e+02 4.465e+02, threshold=3.832e+02, percent-clipped=5.0 2023-04-28 00:32:53,617 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=149575.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:33:23,677 INFO [finetune.py:976] (1/7) Epoch 27, batch 700, loss[loss=0.2228, simple_loss=0.2995, pruned_loss=0.07307, over 4882.00 frames. ], tot_loss[loss=0.1682, simple_loss=0.2417, pruned_loss=0.04741, over 925552.81 frames. ], batch size: 43, lr: 2.93e-03, grad_scale: 32.0 2023-04-28 00:33:27,480 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=149626.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 00:33:39,474 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7924, 3.6238, 0.8868, 1.7115, 1.9911, 2.4132, 1.9584, 0.9817], device='cuda:1'), covar=tensor([0.1340, 0.0852, 0.2090, 0.1420, 0.1058, 0.1185, 0.1551, 0.2060], device='cuda:1'), in_proj_covar=tensor([0.0118, 0.0239, 0.0136, 0.0122, 0.0132, 0.0153, 0.0118, 0.0119], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-28 00:33:56,594 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-04-28 00:33:56,990 INFO [finetune.py:976] (1/7) Epoch 27, batch 750, loss[loss=0.1972, simple_loss=0.2658, pruned_loss=0.06435, over 4848.00 frames. ], tot_loss[loss=0.1691, simple_loss=0.2428, pruned_loss=0.04766, over 930585.20 frames. ], batch size: 44, lr: 2.93e-03, grad_scale: 32.0 2023-04-28 00:33:58,789 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.029e+02 1.603e+02 1.862e+02 2.208e+02 4.099e+02, threshold=3.723e+02, percent-clipped=2.0 2023-04-28 00:34:03,111 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.3318, 1.2473, 3.7411, 3.4954, 3.2954, 3.5585, 3.5110, 3.2643], device='cuda:1'), covar=tensor([0.7187, 0.5491, 0.1126, 0.1835, 0.1193, 0.1600, 0.2610, 0.1527], device='cuda:1'), in_proj_covar=tensor([0.0310, 0.0309, 0.0409, 0.0409, 0.0347, 0.0413, 0.0320, 0.0363], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-28 00:34:15,771 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0428, 2.7253, 1.2165, 1.4649, 2.2061, 1.2061, 3.4223, 1.7365], device='cuda:1'), covar=tensor([0.0683, 0.0657, 0.0680, 0.1102, 0.0423, 0.0956, 0.0201, 0.0580], device='cuda:1'), in_proj_covar=tensor([0.0050, 0.0064, 0.0047, 0.0046, 0.0049, 0.0051, 0.0072, 0.0050], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:1') 2023-04-28 00:34:26,012 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-04-28 00:34:42,992 INFO [finetune.py:976] (1/7) Epoch 27, batch 800, loss[loss=0.1535, simple_loss=0.2223, pruned_loss=0.0424, over 4866.00 frames. ], tot_loss[loss=0.1679, simple_loss=0.2416, pruned_loss=0.04706, over 934146.50 frames. ], batch size: 31, lr: 2.93e-03, grad_scale: 32.0 2023-04-28 00:35:14,876 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=149746.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 00:35:48,631 INFO [finetune.py:976] (1/7) Epoch 27, batch 850, loss[loss=0.1781, simple_loss=0.2421, pruned_loss=0.05703, over 4916.00 frames. ], tot_loss[loss=0.1664, simple_loss=0.2394, pruned_loss=0.04669, over 938134.71 frames. ], batch size: 37, lr: 2.93e-03, grad_scale: 32.0 2023-04-28 00:35:55,545 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.068e+02 1.475e+02 1.749e+02 2.076e+02 4.110e+02, threshold=3.498e+02, percent-clipped=2.0 2023-04-28 00:36:23,790 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([4.5042, 4.3129, 3.1046, 5.1525, 4.4408, 4.4659, 1.6632, 4.4041], device='cuda:1'), covar=tensor([0.1329, 0.1025, 0.3394, 0.0873, 0.2866, 0.1380, 0.6133, 0.1964], device='cuda:1'), in_proj_covar=tensor([0.0248, 0.0222, 0.0255, 0.0306, 0.0302, 0.0250, 0.0278, 0.0277], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 00:36:24,412 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=149807.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 00:36:33,260 INFO [finetune.py:976] (1/7) Epoch 27, batch 900, loss[loss=0.1305, simple_loss=0.2014, pruned_loss=0.02981, over 4732.00 frames. ], tot_loss[loss=0.1647, simple_loss=0.237, pruned_loss=0.04619, over 942724.82 frames. ], batch size: 54, lr: 2.93e-03, grad_scale: 32.0 2023-04-28 00:36:41,326 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.57 vs. limit=2.0 2023-04-28 00:36:42,488 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6932, 2.1101, 1.6824, 2.0463, 1.4519, 1.6870, 1.8044, 1.4277], device='cuda:1'), covar=tensor([0.2201, 0.1748, 0.1210, 0.1417, 0.3615, 0.1523, 0.2014, 0.2555], device='cuda:1'), in_proj_covar=tensor([0.0283, 0.0300, 0.0215, 0.0277, 0.0314, 0.0254, 0.0249, 0.0264], device='cuda:1'), out_proj_covar=tensor([1.1313e-04, 1.1801e-04, 8.4497e-05, 1.0898e-04, 1.2667e-04, 9.9830e-05, 1.0046e-04, 1.0414e-04], device='cuda:1') 2023-04-28 00:37:06,210 INFO [finetune.py:976] (1/7) Epoch 27, batch 950, loss[loss=0.1949, simple_loss=0.2684, pruned_loss=0.06072, over 4812.00 frames. ], tot_loss[loss=0.164, simple_loss=0.2357, pruned_loss=0.04611, over 947599.70 frames. ], batch size: 45, lr: 2.93e-03, grad_scale: 32.0 2023-04-28 00:37:06,274 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=149870.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:37:08,538 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.083e+02 1.514e+02 1.776e+02 2.110e+02 4.018e+02, threshold=3.552e+02, percent-clipped=2.0 2023-04-28 00:37:37,870 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.50 vs. limit=5.0 2023-04-28 00:37:40,040 INFO [finetune.py:976] (1/7) Epoch 27, batch 1000, loss[loss=0.1991, simple_loss=0.2752, pruned_loss=0.0615, over 4898.00 frames. ], tot_loss[loss=0.1668, simple_loss=0.239, pruned_loss=0.04727, over 948144.93 frames. ], batch size: 43, lr: 2.93e-03, grad_scale: 32.0 2023-04-28 00:37:40,688 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=149921.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 00:37:51,047 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-04-28 00:38:13,689 INFO [finetune.py:976] (1/7) Epoch 27, batch 1050, loss[loss=0.256, simple_loss=0.3118, pruned_loss=0.1001, over 4809.00 frames. ], tot_loss[loss=0.1687, simple_loss=0.2421, pruned_loss=0.04764, over 951033.61 frames. ], batch size: 45, lr: 2.93e-03, grad_scale: 32.0 2023-04-28 00:38:15,492 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.212e+02 1.681e+02 2.029e+02 2.486e+02 4.683e+02, threshold=4.057e+02, percent-clipped=3.0 2023-04-28 00:38:26,715 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-04-28 00:38:43,952 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=2.03 vs. limit=2.0 2023-04-28 00:38:48,335 INFO [finetune.py:976] (1/7) Epoch 27, batch 1100, loss[loss=0.1291, simple_loss=0.213, pruned_loss=0.02258, over 4884.00 frames. ], tot_loss[loss=0.1698, simple_loss=0.2438, pruned_loss=0.04788, over 952446.41 frames. ], batch size: 43, lr: 2.93e-03, grad_scale: 32.0 2023-04-28 00:38:59,849 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=3.89 vs. limit=5.0 2023-04-28 00:39:22,074 INFO [finetune.py:976] (1/7) Epoch 27, batch 1150, loss[loss=0.191, simple_loss=0.2726, pruned_loss=0.05475, over 4918.00 frames. ], tot_loss[loss=0.1703, simple_loss=0.2443, pruned_loss=0.04819, over 953311.13 frames. ], batch size: 38, lr: 2.93e-03, grad_scale: 32.0 2023-04-28 00:39:23,883 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.145e+02 1.471e+02 1.817e+02 2.070e+02 3.709e+02, threshold=3.635e+02, percent-clipped=0.0 2023-04-28 00:39:28,116 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=150079.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:39:43,681 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=150102.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 00:39:56,126 INFO [finetune.py:976] (1/7) Epoch 27, batch 1200, loss[loss=0.2057, simple_loss=0.2793, pruned_loss=0.06609, over 4728.00 frames. ], tot_loss[loss=0.169, simple_loss=0.2429, pruned_loss=0.04756, over 955333.48 frames. ], batch size: 54, lr: 2.93e-03, grad_scale: 32.0 2023-04-28 00:40:10,319 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=150140.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:40:45,078 INFO [finetune.py:976] (1/7) Epoch 27, batch 1250, loss[loss=0.1262, simple_loss=0.1926, pruned_loss=0.02988, over 4814.00 frames. ], tot_loss[loss=0.1669, simple_loss=0.24, pruned_loss=0.04694, over 953944.14 frames. ], batch size: 51, lr: 2.93e-03, grad_scale: 32.0 2023-04-28 00:40:45,174 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=150170.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:40:47,938 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 8.413e+01 1.487e+02 1.765e+02 2.046e+02 3.394e+02, threshold=3.529e+02, percent-clipped=0.0 2023-04-28 00:40:50,272 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.45 vs. limit=5.0 2023-04-28 00:41:49,584 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=150218.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:41:50,798 INFO [finetune.py:976] (1/7) Epoch 27, batch 1300, loss[loss=0.1838, simple_loss=0.2417, pruned_loss=0.06297, over 4219.00 frames. ], tot_loss[loss=0.1651, simple_loss=0.238, pruned_loss=0.04615, over 955401.16 frames. ], batch size: 65, lr: 2.93e-03, grad_scale: 32.0 2023-04-28 00:41:51,528 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=150221.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:42:55,588 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7087, 0.7242, 1.5161, 2.0356, 1.7754, 1.5713, 1.5634, 1.5721], device='cuda:1'), covar=tensor([0.4097, 0.6205, 0.6055, 0.5736, 0.5642, 0.6980, 0.7199, 0.8236], device='cuda:1'), in_proj_covar=tensor([0.0441, 0.0422, 0.0516, 0.0505, 0.0469, 0.0506, 0.0507, 0.0521], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-28 00:42:56,733 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=150269.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:42:57,287 INFO [finetune.py:976] (1/7) Epoch 27, batch 1350, loss[loss=0.1808, simple_loss=0.2551, pruned_loss=0.05331, over 4900.00 frames. ], tot_loss[loss=0.1643, simple_loss=0.2369, pruned_loss=0.04581, over 955150.64 frames. ], batch size: 35, lr: 2.93e-03, grad_scale: 32.0 2023-04-28 00:42:59,132 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.101e+02 1.474e+02 1.801e+02 2.070e+02 3.416e+02, threshold=3.602e+02, percent-clipped=0.0 2023-04-28 00:43:05,179 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=150274.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:43:18,563 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5151, 1.5257, 1.8959, 1.9488, 1.3655, 1.2944, 1.5129, 0.9014], device='cuda:1'), covar=tensor([0.0645, 0.0634, 0.0420, 0.0572, 0.0846, 0.1117, 0.0611, 0.0668], device='cuda:1'), in_proj_covar=tensor([0.0070, 0.0067, 0.0066, 0.0069, 0.0074, 0.0094, 0.0072, 0.0062], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-28 00:44:01,099 INFO [finetune.py:976] (1/7) Epoch 27, batch 1400, loss[loss=0.1503, simple_loss=0.2312, pruned_loss=0.0347, over 4902.00 frames. ], tot_loss[loss=0.1674, simple_loss=0.2406, pruned_loss=0.04706, over 957110.24 frames. ], batch size: 35, lr: 2.93e-03, grad_scale: 16.0 2023-04-28 00:44:22,624 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=150335.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:45:00,624 INFO [finetune.py:976] (1/7) Epoch 27, batch 1450, loss[loss=0.2142, simple_loss=0.2859, pruned_loss=0.07125, over 4905.00 frames. ], tot_loss[loss=0.1677, simple_loss=0.2413, pruned_loss=0.04702, over 954144.48 frames. ], batch size: 36, lr: 2.93e-03, grad_scale: 16.0 2023-04-28 00:45:03,057 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.061e+02 1.594e+02 1.851e+02 2.331e+02 4.105e+02, threshold=3.701e+02, percent-clipped=2.0 2023-04-28 00:45:22,902 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=150402.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 00:45:26,593 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7843, 1.4170, 1.9144, 2.2264, 1.8707, 1.7760, 1.8099, 1.7698], device='cuda:1'), covar=tensor([0.4017, 0.6439, 0.5795, 0.5060, 0.5510, 0.7384, 0.7045, 0.7966], device='cuda:1'), in_proj_covar=tensor([0.0441, 0.0421, 0.0515, 0.0505, 0.0469, 0.0506, 0.0507, 0.0520], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-28 00:45:29,026 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.1898, 1.9771, 2.4353, 2.6088, 2.2176, 2.1249, 2.2374, 2.2105], device='cuda:1'), covar=tensor([0.4514, 0.7386, 0.6678, 0.5327, 0.6125, 0.8441, 0.8836, 0.9528], device='cuda:1'), in_proj_covar=tensor([0.0442, 0.0422, 0.0516, 0.0506, 0.0469, 0.0506, 0.0507, 0.0520], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-28 00:45:33,782 INFO [finetune.py:976] (1/7) Epoch 27, batch 1500, loss[loss=0.176, simple_loss=0.2576, pruned_loss=0.04716, over 4834.00 frames. ], tot_loss[loss=0.1694, simple_loss=0.2431, pruned_loss=0.04784, over 952980.60 frames. ], batch size: 47, lr: 2.93e-03, grad_scale: 16.0 2023-04-28 00:45:43,905 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=150435.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:45:54,544 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=150450.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 00:46:06,956 INFO [finetune.py:976] (1/7) Epoch 27, batch 1550, loss[loss=0.173, simple_loss=0.2419, pruned_loss=0.05203, over 4813.00 frames. ], tot_loss[loss=0.1701, simple_loss=0.2436, pruned_loss=0.04829, over 952621.19 frames. ], batch size: 39, lr: 2.93e-03, grad_scale: 16.0 2023-04-28 00:46:09,360 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.042e+02 1.486e+02 1.780e+02 2.149e+02 3.630e+02, threshold=3.561e+02, percent-clipped=0.0 2023-04-28 00:46:19,674 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-04-28 00:46:22,364 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.1582, 2.6389, 1.1278, 1.4851, 2.1151, 1.1161, 3.5139, 1.7559], device='cuda:1'), covar=tensor([0.0665, 0.0555, 0.0732, 0.1221, 0.0502, 0.1075, 0.0244, 0.0649], device='cuda:1'), in_proj_covar=tensor([0.0050, 0.0063, 0.0046, 0.0045, 0.0049, 0.0051, 0.0072, 0.0050], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0007, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:1') 2023-04-28 00:46:40,768 INFO [finetune.py:976] (1/7) Epoch 27, batch 1600, loss[loss=0.1514, simple_loss=0.2199, pruned_loss=0.04146, over 4874.00 frames. ], tot_loss[loss=0.1676, simple_loss=0.2413, pruned_loss=0.04693, over 955919.24 frames. ], batch size: 34, lr: 2.93e-03, grad_scale: 16.0 2023-04-28 00:46:59,333 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.5917, 1.9363, 1.6844, 2.3664, 2.4684, 2.0063, 2.0674, 1.8624], device='cuda:1'), covar=tensor([0.1511, 0.1523, 0.2061, 0.1553, 0.1108, 0.1952, 0.1915, 0.2374], device='cuda:1'), in_proj_covar=tensor([0.0312, 0.0306, 0.0348, 0.0285, 0.0325, 0.0305, 0.0297, 0.0372], device='cuda:1'), out_proj_covar=tensor([6.3604e-05, 6.2921e-05, 7.3116e-05, 5.6946e-05, 6.6419e-05, 6.3567e-05, 6.1442e-05, 7.8719e-05], device='cuda:1') 2023-04-28 00:47:01,156 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6336, 1.5219, 0.6715, 1.3091, 1.6477, 1.4981, 1.4187, 1.4476], device='cuda:1'), covar=tensor([0.0494, 0.0383, 0.0361, 0.0557, 0.0290, 0.0533, 0.0502, 0.0574], device='cuda:1'), in_proj_covar=tensor([0.0028, 0.0023, 0.0022, 0.0028, 0.0019, 0.0027, 0.0027, 0.0028], device='cuda:1'), out_proj_covar=tensor([0.0052, 0.0046, 0.0039, 0.0053, 0.0039, 0.0051, 0.0051, 0.0053], device='cuda:1') 2023-04-28 00:47:08,134 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=150537.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:47:41,251 INFO [finetune.py:976] (1/7) Epoch 27, batch 1650, loss[loss=0.1423, simple_loss=0.2192, pruned_loss=0.03267, over 4809.00 frames. ], tot_loss[loss=0.1661, simple_loss=0.2393, pruned_loss=0.04643, over 955125.88 frames. ], batch size: 25, lr: 2.93e-03, grad_scale: 16.0 2023-04-28 00:47:41,377 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8540, 2.4327, 2.0568, 2.2715, 1.8203, 1.9377, 1.9773, 1.6368], device='cuda:1'), covar=tensor([0.1830, 0.0980, 0.0820, 0.1007, 0.3020, 0.1171, 0.1858, 0.2465], device='cuda:1'), in_proj_covar=tensor([0.0286, 0.0302, 0.0217, 0.0279, 0.0316, 0.0255, 0.0251, 0.0266], device='cuda:1'), out_proj_covar=tensor([1.1432e-04, 1.1903e-04, 8.5539e-05, 1.0992e-04, 1.2739e-04, 1.0027e-04, 1.0115e-04, 1.0506e-04], device='cuda:1') 2023-04-28 00:47:43,699 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.015e+02 1.594e+02 1.886e+02 2.369e+02 4.479e+02, threshold=3.773e+02, percent-clipped=3.0 2023-04-28 00:48:00,504 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=150598.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:48:14,791 INFO [finetune.py:976] (1/7) Epoch 27, batch 1700, loss[loss=0.18, simple_loss=0.2452, pruned_loss=0.05737, over 4848.00 frames. ], tot_loss[loss=0.1633, simple_loss=0.2361, pruned_loss=0.04525, over 956422.11 frames. ], batch size: 49, lr: 2.93e-03, grad_scale: 16.0 2023-04-28 00:48:20,943 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=150630.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:48:48,160 INFO [finetune.py:976] (1/7) Epoch 27, batch 1750, loss[loss=0.1662, simple_loss=0.2431, pruned_loss=0.04464, over 4898.00 frames. ], tot_loss[loss=0.1663, simple_loss=0.2391, pruned_loss=0.04674, over 957948.18 frames. ], batch size: 35, lr: 2.93e-03, grad_scale: 16.0 2023-04-28 00:48:50,601 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.161e+02 1.571e+02 1.900e+02 2.319e+02 4.181e+02, threshold=3.799e+02, percent-clipped=4.0 2023-04-28 00:49:21,910 INFO [finetune.py:976] (1/7) Epoch 27, batch 1800, loss[loss=0.1877, simple_loss=0.2611, pruned_loss=0.05719, over 4813.00 frames. ], tot_loss[loss=0.1688, simple_loss=0.2421, pruned_loss=0.04771, over 958140.38 frames. ], batch size: 40, lr: 2.93e-03, grad_scale: 16.0 2023-04-28 00:49:22,678 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.5596, 2.2110, 2.4948, 2.9652, 2.4782, 1.9623, 1.9511, 2.4852], device='cuda:1'), covar=tensor([0.3265, 0.3007, 0.1515, 0.2426, 0.2606, 0.2580, 0.3728, 0.1929], device='cuda:1'), in_proj_covar=tensor([0.0294, 0.0247, 0.0229, 0.0315, 0.0221, 0.0235, 0.0228, 0.0187], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-28 00:49:29,197 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=150732.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:49:31,014 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=150735.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:49:43,300 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9102, 2.4127, 1.8518, 1.8385, 1.3209, 1.3595, 1.9719, 1.2543], device='cuda:1'), covar=tensor([0.1605, 0.1323, 0.1294, 0.1537, 0.2190, 0.1889, 0.0873, 0.1961], device='cuda:1'), in_proj_covar=tensor([0.0197, 0.0209, 0.0169, 0.0203, 0.0200, 0.0186, 0.0156, 0.0187], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-28 00:49:54,545 INFO [finetune.py:976] (1/7) Epoch 27, batch 1850, loss[loss=0.1755, simple_loss=0.2525, pruned_loss=0.04921, over 4922.00 frames. ], tot_loss[loss=0.1708, simple_loss=0.244, pruned_loss=0.04882, over 957966.02 frames. ], batch size: 38, lr: 2.93e-03, grad_scale: 16.0 2023-04-28 00:50:02,285 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.087e+02 1.541e+02 1.934e+02 2.290e+02 4.009e+02, threshold=3.867e+02, percent-clipped=2.0 2023-04-28 00:50:13,360 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=150783.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:50:25,736 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=150793.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 00:50:56,516 INFO [finetune.py:976] (1/7) Epoch 27, batch 1900, loss[loss=0.1686, simple_loss=0.2468, pruned_loss=0.04518, over 4776.00 frames. ], tot_loss[loss=0.1702, simple_loss=0.2437, pruned_loss=0.04841, over 954092.64 frames. ], batch size: 51, lr: 2.93e-03, grad_scale: 16.0 2023-04-28 00:51:17,561 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.2143, 1.7190, 2.0617, 2.1858, 2.0855, 1.6881, 1.2608, 1.7881], device='cuda:1'), covar=tensor([0.2926, 0.2938, 0.1593, 0.2037, 0.2348, 0.2401, 0.3747, 0.1865], device='cuda:1'), in_proj_covar=tensor([0.0295, 0.0248, 0.0230, 0.0317, 0.0222, 0.0237, 0.0230, 0.0188], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-28 00:51:18,798 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-04-28 00:51:29,796 INFO [finetune.py:976] (1/7) Epoch 27, batch 1950, loss[loss=0.1407, simple_loss=0.2184, pruned_loss=0.03149, over 4750.00 frames. ], tot_loss[loss=0.168, simple_loss=0.2416, pruned_loss=0.04722, over 952994.50 frames. ], batch size: 27, lr: 2.92e-03, grad_scale: 16.0 2023-04-28 00:51:32,193 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.093e+02 1.535e+02 1.747e+02 2.093e+02 5.445e+02, threshold=3.494e+02, percent-clipped=3.0 2023-04-28 00:51:44,420 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=150893.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:51:45,720 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.2945, 1.5106, 1.8198, 1.9510, 1.8458, 1.9255, 1.8463, 1.8234], device='cuda:1'), covar=tensor([0.3472, 0.4680, 0.3898, 0.4007, 0.4891, 0.6424, 0.4788, 0.4514], device='cuda:1'), in_proj_covar=tensor([0.0340, 0.0375, 0.0329, 0.0340, 0.0350, 0.0394, 0.0359, 0.0332], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 00:51:46,272 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.2244, 4.3795, 0.8621, 2.3676, 2.6389, 2.9578, 2.5718, 1.1102], device='cuda:1'), covar=tensor([0.1259, 0.0980, 0.2180, 0.1184, 0.0975, 0.1020, 0.1487, 0.2082], device='cuda:1'), in_proj_covar=tensor([0.0117, 0.0236, 0.0135, 0.0120, 0.0131, 0.0151, 0.0117, 0.0117], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-28 00:51:49,215 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=150900.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:51:51,074 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=3.28 vs. limit=5.0 2023-04-28 00:52:03,059 INFO [finetune.py:976] (1/7) Epoch 27, batch 2000, loss[loss=0.171, simple_loss=0.2452, pruned_loss=0.04843, over 4855.00 frames. ], tot_loss[loss=0.167, simple_loss=0.24, pruned_loss=0.04697, over 954292.59 frames. ], batch size: 49, lr: 2.92e-03, grad_scale: 16.0 2023-04-28 00:52:09,181 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=150930.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:52:44,345 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6862, 2.1051, 1.8822, 2.0267, 1.5593, 1.7375, 1.7233, 1.4244], device='cuda:1'), covar=tensor([0.1827, 0.1143, 0.0821, 0.0993, 0.3407, 0.1048, 0.1835, 0.2384], device='cuda:1'), in_proj_covar=tensor([0.0285, 0.0303, 0.0217, 0.0278, 0.0317, 0.0254, 0.0249, 0.0266], device='cuda:1'), out_proj_covar=tensor([1.1376e-04, 1.1918e-04, 8.5237e-05, 1.0960e-04, 1.2781e-04, 9.9984e-05, 1.0043e-04, 1.0485e-04], device='cuda:1') 2023-04-28 00:52:51,903 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=150961.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:52:55,532 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0364, 2.3441, 0.9799, 1.3411, 1.7562, 1.1998, 3.0060, 1.5885], device='cuda:1'), covar=tensor([0.0701, 0.0572, 0.0775, 0.1252, 0.0532, 0.1025, 0.0245, 0.0662], device='cuda:1'), in_proj_covar=tensor([0.0050, 0.0063, 0.0046, 0.0046, 0.0049, 0.0051, 0.0072, 0.0050], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0007, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:1') 2023-04-28 00:53:02,955 INFO [finetune.py:976] (1/7) Epoch 27, batch 2050, loss[loss=0.1369, simple_loss=0.2023, pruned_loss=0.03574, over 4759.00 frames. ], tot_loss[loss=0.165, simple_loss=0.2371, pruned_loss=0.04643, over 954544.89 frames. ], batch size: 28, lr: 2.92e-03, grad_scale: 16.0 2023-04-28 00:53:05,048 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=3.84 vs. limit=5.0 2023-04-28 00:53:05,397 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.068e+02 1.536e+02 1.853e+02 2.161e+02 4.480e+02, threshold=3.705e+02, percent-clipped=3.0 2023-04-28 00:53:12,609 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=150978.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:54:07,735 INFO [finetune.py:976] (1/7) Epoch 27, batch 2100, loss[loss=0.1653, simple_loss=0.2426, pruned_loss=0.04405, over 4933.00 frames. ], tot_loss[loss=0.1653, simple_loss=0.2374, pruned_loss=0.04659, over 954855.44 frames. ], batch size: 38, lr: 2.92e-03, grad_scale: 16.0 2023-04-28 00:54:10,325 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-04-28 00:55:11,731 INFO [finetune.py:976] (1/7) Epoch 27, batch 2150, loss[loss=0.1806, simple_loss=0.2667, pruned_loss=0.04723, over 4905.00 frames. ], tot_loss[loss=0.1677, simple_loss=0.2405, pruned_loss=0.04745, over 955660.72 frames. ], batch size: 36, lr: 2.92e-03, grad_scale: 16.0 2023-04-28 00:55:19,340 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.036e+02 1.480e+02 1.779e+02 2.139e+02 3.586e+02, threshold=3.558e+02, percent-clipped=0.0 2023-04-28 00:55:33,247 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=151088.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 00:55:40,987 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.14 vs. limit=2.0 2023-04-28 00:56:16,162 INFO [finetune.py:976] (1/7) Epoch 27, batch 2200, loss[loss=0.2011, simple_loss=0.2664, pruned_loss=0.06787, over 4828.00 frames. ], tot_loss[loss=0.1688, simple_loss=0.2414, pruned_loss=0.04804, over 953369.09 frames. ], batch size: 30, lr: 2.92e-03, grad_scale: 16.0 2023-04-28 00:57:19,653 INFO [finetune.py:976] (1/7) Epoch 27, batch 2250, loss[loss=0.1587, simple_loss=0.2348, pruned_loss=0.04129, over 4827.00 frames. ], tot_loss[loss=0.169, simple_loss=0.2421, pruned_loss=0.04791, over 951916.85 frames. ], batch size: 47, lr: 2.92e-03, grad_scale: 16.0 2023-04-28 00:57:27,225 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.948e+01 1.668e+02 1.911e+02 2.463e+02 4.700e+02, threshold=3.822e+02, percent-clipped=3.0 2023-04-28 00:57:27,951 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.2630, 1.4976, 1.4262, 1.7107, 1.7081, 2.0384, 1.4441, 3.5314], device='cuda:1'), covar=tensor([0.0603, 0.0785, 0.0765, 0.1180, 0.0618, 0.0473, 0.0733, 0.0133], device='cuda:1'), in_proj_covar=tensor([0.0037, 0.0039, 0.0039, 0.0042, 0.0040, 0.0037, 0.0038, 0.0054], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0013, 0.0013, 0.0015, 0.0014, 0.0013, 0.0013, 0.0016], device='cuda:1') 2023-04-28 00:57:37,934 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([3.9734, 3.8470, 2.8824, 4.5377, 3.9072, 3.8944, 1.6723, 3.8638], device='cuda:1'), covar=tensor([0.1648, 0.1069, 0.3181, 0.1387, 0.2510, 0.1663, 0.5775, 0.2428], device='cuda:1'), in_proj_covar=tensor([0.0243, 0.0217, 0.0250, 0.0300, 0.0296, 0.0247, 0.0273, 0.0272], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 00:57:38,568 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=151184.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:57:49,170 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=151193.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:58:22,083 INFO [finetune.py:976] (1/7) Epoch 27, batch 2300, loss[loss=0.1961, simple_loss=0.2632, pruned_loss=0.06448, over 4912.00 frames. ], tot_loss[loss=0.1687, simple_loss=0.242, pruned_loss=0.04771, over 951826.69 frames. ], batch size: 38, lr: 2.92e-03, grad_scale: 16.0 2023-04-28 00:58:28,303 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.0468, 2.0020, 1.7987, 1.6612, 2.1386, 1.7888, 2.6653, 1.6709], device='cuda:1'), covar=tensor([0.3612, 0.1858, 0.4696, 0.3079, 0.1536, 0.2315, 0.1231, 0.4171], device='cuda:1'), in_proj_covar=tensor([0.0338, 0.0353, 0.0424, 0.0350, 0.0380, 0.0374, 0.0366, 0.0419], device='cuda:1'), out_proj_covar=tensor([9.9833e-05, 1.0526e-04, 1.2824e-04, 1.0473e-04, 1.1249e-04, 1.1124e-04, 1.0690e-04, 1.2602e-04], device='cuda:1') 2023-04-28 00:58:42,717 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=151241.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:58:42,779 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6768, 2.0586, 1.8065, 1.9568, 1.5723, 1.6415, 1.6409, 1.3912], device='cuda:1'), covar=tensor([0.1658, 0.1190, 0.0759, 0.1072, 0.3379, 0.1155, 0.1611, 0.2112], device='cuda:1'), in_proj_covar=tensor([0.0285, 0.0303, 0.0217, 0.0279, 0.0318, 0.0255, 0.0250, 0.0266], device='cuda:1'), out_proj_covar=tensor([1.1395e-04, 1.1937e-04, 8.5377e-05, 1.1000e-04, 1.2822e-04, 1.0037e-04, 1.0080e-04, 1.0495e-04], device='cuda:1') 2023-04-28 00:58:45,173 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=151245.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:58:47,040 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7906, 1.5663, 1.8967, 2.1501, 1.6245, 1.5083, 1.6821, 1.0863], device='cuda:1'), covar=tensor([0.0410, 0.0685, 0.0414, 0.0421, 0.0577, 0.0903, 0.0565, 0.0537], device='cuda:1'), in_proj_covar=tensor([0.0070, 0.0067, 0.0066, 0.0069, 0.0074, 0.0094, 0.0072, 0.0063], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-28 00:58:52,340 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=151256.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:59:01,205 INFO [finetune.py:976] (1/7) Epoch 27, batch 2350, loss[loss=0.1768, simple_loss=0.2512, pruned_loss=0.05119, over 4756.00 frames. ], tot_loss[loss=0.1676, simple_loss=0.2405, pruned_loss=0.04733, over 953473.47 frames. ], batch size: 54, lr: 2.92e-03, grad_scale: 16.0 2023-04-28 00:59:04,086 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.014e+02 1.580e+02 1.872e+02 2.240e+02 4.120e+02, threshold=3.744e+02, percent-clipped=1.0 2023-04-28 00:59:04,795 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([3.9847, 3.9070, 2.8036, 4.6094, 4.0154, 3.9881, 1.8803, 3.8841], device='cuda:1'), covar=tensor([0.1700, 0.1214, 0.3090, 0.1555, 0.2925, 0.1948, 0.5536, 0.2331], device='cuda:1'), in_proj_covar=tensor([0.0243, 0.0217, 0.0250, 0.0300, 0.0296, 0.0247, 0.0273, 0.0272], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 00:59:10,150 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=151282.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:59:20,463 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=151298.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:59:34,699 INFO [finetune.py:976] (1/7) Epoch 27, batch 2400, loss[loss=0.127, simple_loss=0.1906, pruned_loss=0.03172, over 4488.00 frames. ], tot_loss[loss=0.1658, simple_loss=0.2383, pruned_loss=0.0467, over 954729.78 frames. ], batch size: 19, lr: 2.92e-03, grad_scale: 16.0 2023-04-28 00:59:38,937 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.2063, 2.7387, 1.2140, 1.4504, 2.1955, 1.2260, 3.7118, 1.9248], device='cuda:1'), covar=tensor([0.0638, 0.0552, 0.0719, 0.1191, 0.0471, 0.0996, 0.0351, 0.0576], device='cuda:1'), in_proj_covar=tensor([0.0050, 0.0064, 0.0047, 0.0046, 0.0049, 0.0051, 0.0072, 0.0051], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:1') 2023-04-28 00:59:50,888 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=151343.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:59:51,625 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-04-28 01:00:01,135 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=151359.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:00:07,713 INFO [finetune.py:976] (1/7) Epoch 27, batch 2450, loss[loss=0.1586, simple_loss=0.2275, pruned_loss=0.04483, over 4816.00 frames. ], tot_loss[loss=0.1637, simple_loss=0.2361, pruned_loss=0.04567, over 955631.84 frames. ], batch size: 51, lr: 2.92e-03, grad_scale: 16.0 2023-04-28 01:00:10,590 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.893e+01 1.571e+02 1.897e+02 2.275e+02 8.002e+02, threshold=3.795e+02, percent-clipped=4.0 2023-04-28 01:00:21,203 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=151388.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:00:34,102 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9100, 2.5570, 1.8947, 1.9589, 1.4527, 1.4656, 1.9617, 1.3415], device='cuda:1'), covar=tensor([0.1580, 0.1377, 0.1350, 0.1581, 0.2130, 0.1864, 0.0931, 0.1925], device='cuda:1'), in_proj_covar=tensor([0.0198, 0.0209, 0.0170, 0.0204, 0.0200, 0.0187, 0.0156, 0.0187], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-28 01:00:41,810 INFO [finetune.py:976] (1/7) Epoch 27, batch 2500, loss[loss=0.1968, simple_loss=0.2644, pruned_loss=0.06461, over 4904.00 frames. ], tot_loss[loss=0.1662, simple_loss=0.2384, pruned_loss=0.04704, over 953137.38 frames. ], batch size: 32, lr: 2.92e-03, grad_scale: 16.0 2023-04-28 01:00:53,953 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=151436.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:01:15,577 INFO [finetune.py:976] (1/7) Epoch 27, batch 2550, loss[loss=0.1331, simple_loss=0.2048, pruned_loss=0.03074, over 4770.00 frames. ], tot_loss[loss=0.1675, simple_loss=0.2404, pruned_loss=0.04729, over 953176.08 frames. ], batch size: 26, lr: 2.92e-03, grad_scale: 16.0 2023-04-28 01:01:17,946 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.064e+02 1.630e+02 1.877e+02 2.163e+02 4.851e+02, threshold=3.753e+02, percent-clipped=1.0 2023-04-28 01:01:30,171 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.71 vs. limit=2.0 2023-04-28 01:01:33,140 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.69 vs. limit=2.0 2023-04-28 01:01:47,852 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8534, 2.3860, 1.8035, 1.9122, 1.3811, 1.3879, 1.9115, 1.3111], device='cuda:1'), covar=tensor([0.1603, 0.1408, 0.1422, 0.1652, 0.2266, 0.1890, 0.0975, 0.2069], device='cuda:1'), in_proj_covar=tensor([0.0198, 0.0209, 0.0170, 0.0204, 0.0200, 0.0187, 0.0156, 0.0187], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-28 01:01:48,925 INFO [finetune.py:976] (1/7) Epoch 27, batch 2600, loss[loss=0.212, simple_loss=0.2788, pruned_loss=0.07257, over 4824.00 frames. ], tot_loss[loss=0.1694, simple_loss=0.2423, pruned_loss=0.04826, over 953734.61 frames. ], batch size: 39, lr: 2.92e-03, grad_scale: 16.0 2023-04-28 01:02:02,083 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=151540.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:02:19,375 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=151556.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:02:33,622 INFO [finetune.py:976] (1/7) Epoch 27, batch 2650, loss[loss=0.1679, simple_loss=0.2443, pruned_loss=0.04572, over 4889.00 frames. ], tot_loss[loss=0.1708, simple_loss=0.2441, pruned_loss=0.04874, over 953816.01 frames. ], batch size: 43, lr: 2.92e-03, grad_scale: 16.0 2023-04-28 01:02:41,497 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.080e+02 1.560e+02 1.768e+02 2.112e+02 4.460e+02, threshold=3.536e+02, percent-clipped=1.0 2023-04-28 01:03:23,442 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=151604.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:03:39,056 INFO [finetune.py:976] (1/7) Epoch 27, batch 2700, loss[loss=0.1481, simple_loss=0.215, pruned_loss=0.04055, over 4232.00 frames. ], tot_loss[loss=0.1707, simple_loss=0.2443, pruned_loss=0.04853, over 954424.89 frames. ], batch size: 18, lr: 2.92e-03, grad_scale: 16.0 2023-04-28 01:03:46,138 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.4572, 2.9187, 1.0382, 1.6758, 2.1300, 1.2647, 3.8363, 1.8815], device='cuda:1'), covar=tensor([0.0600, 0.0768, 0.0843, 0.1129, 0.0467, 0.0930, 0.0175, 0.0566], device='cuda:1'), in_proj_covar=tensor([0.0050, 0.0064, 0.0047, 0.0046, 0.0049, 0.0051, 0.0072, 0.0051], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:1') 2023-04-28 01:03:57,102 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.1689, 2.6926, 2.4279, 2.5625, 2.0373, 2.4430, 2.2225, 1.8214], device='cuda:1'), covar=tensor([0.1779, 0.1009, 0.0652, 0.1056, 0.2905, 0.0920, 0.1804, 0.2169], device='cuda:1'), in_proj_covar=tensor([0.0287, 0.0303, 0.0217, 0.0280, 0.0318, 0.0255, 0.0250, 0.0267], device='cuda:1'), out_proj_covar=tensor([1.1460e-04, 1.1917e-04, 8.5467e-05, 1.1017e-04, 1.2810e-04, 1.0039e-04, 1.0097e-04, 1.0528e-04], device='cuda:1') 2023-04-28 01:04:05,034 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=151638.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:04:21,249 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=151654.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:04:28,346 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-04-28 01:04:40,568 INFO [finetune.py:976] (1/7) Epoch 27, batch 2750, loss[loss=0.1486, simple_loss=0.2203, pruned_loss=0.03845, over 4860.00 frames. ], tot_loss[loss=0.1686, simple_loss=0.2419, pruned_loss=0.0476, over 956467.31 frames. ], batch size: 49, lr: 2.92e-03, grad_scale: 16.0 2023-04-28 01:04:48,522 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.458e+01 1.381e+02 1.737e+02 2.155e+02 3.699e+02, threshold=3.473e+02, percent-clipped=1.0 2023-04-28 01:05:02,019 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=151688.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:05:09,114 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-04-28 01:05:44,740 INFO [finetune.py:976] (1/7) Epoch 27, batch 2800, loss[loss=0.2109, simple_loss=0.2767, pruned_loss=0.07252, over 4312.00 frames. ], tot_loss[loss=0.1667, simple_loss=0.2391, pruned_loss=0.0471, over 955828.01 frames. ], batch size: 65, lr: 2.92e-03, grad_scale: 16.0 2023-04-28 01:06:24,879 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=151749.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:06:27,827 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.1493, 2.7896, 2.4490, 2.6900, 1.9945, 2.5089, 2.5746, 1.8970], device='cuda:1'), covar=tensor([0.2243, 0.1057, 0.0783, 0.1209, 0.3237, 0.1079, 0.1950, 0.2735], device='cuda:1'), in_proj_covar=tensor([0.0287, 0.0304, 0.0217, 0.0280, 0.0318, 0.0256, 0.0250, 0.0267], device='cuda:1'), out_proj_covar=tensor([1.1467e-04, 1.1944e-04, 8.5382e-05, 1.1032e-04, 1.2824e-04, 1.0064e-04, 1.0097e-04, 1.0534e-04], device='cuda:1') 2023-04-28 01:06:53,981 INFO [finetune.py:976] (1/7) Epoch 27, batch 2850, loss[loss=0.1511, simple_loss=0.2262, pruned_loss=0.03798, over 4893.00 frames. ], tot_loss[loss=0.1655, simple_loss=0.2377, pruned_loss=0.0466, over 952102.53 frames. ], batch size: 35, lr: 2.92e-03, grad_scale: 16.0 2023-04-28 01:06:56,484 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.135e+02 1.506e+02 1.770e+02 2.075e+02 3.794e+02, threshold=3.540e+02, percent-clipped=1.0 2023-04-28 01:07:25,201 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.79 vs. limit=2.0 2023-04-28 01:07:29,470 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=151801.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:07:58,281 INFO [finetune.py:976] (1/7) Epoch 27, batch 2900, loss[loss=0.1768, simple_loss=0.2531, pruned_loss=0.05027, over 4813.00 frames. ], tot_loss[loss=0.1687, simple_loss=0.2413, pruned_loss=0.04807, over 954020.03 frames. ], batch size: 33, lr: 2.92e-03, grad_scale: 16.0 2023-04-28 01:08:11,763 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=151840.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:08:15,303 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6127, 1.8410, 0.8771, 1.3245, 2.1802, 1.4882, 1.4127, 1.5216], device='cuda:1'), covar=tensor([0.0511, 0.0349, 0.0291, 0.0520, 0.0237, 0.0481, 0.0469, 0.0569], device='cuda:1'), in_proj_covar=tensor([0.0027, 0.0023, 0.0021, 0.0028, 0.0019, 0.0027, 0.0027, 0.0028], device='cuda:1'), out_proj_covar=tensor([0.0052, 0.0046, 0.0038, 0.0053, 0.0039, 0.0050, 0.0050, 0.0052], device='cuda:1') 2023-04-28 01:08:32,189 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=151862.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:08:42,905 INFO [finetune.py:976] (1/7) Epoch 27, batch 2950, loss[loss=0.1465, simple_loss=0.2321, pruned_loss=0.03043, over 4806.00 frames. ], tot_loss[loss=0.1705, simple_loss=0.2437, pruned_loss=0.0487, over 954158.42 frames. ], batch size: 45, lr: 2.92e-03, grad_scale: 16.0 2023-04-28 01:08:50,555 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.751e+01 1.623e+02 1.878e+02 2.443e+02 4.733e+02, threshold=3.756e+02, percent-clipped=2.0 2023-04-28 01:09:04,984 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=151888.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:09:12,291 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-04-28 01:09:25,949 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8640, 2.3032, 1.9204, 1.7692, 1.3800, 1.3738, 1.9924, 1.3321], device='cuda:1'), covar=tensor([0.1671, 0.1253, 0.1334, 0.1592, 0.2250, 0.1960, 0.0961, 0.1997], device='cuda:1'), in_proj_covar=tensor([0.0198, 0.0208, 0.0170, 0.0204, 0.0200, 0.0186, 0.0156, 0.0187], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-28 01:09:47,998 INFO [finetune.py:976] (1/7) Epoch 27, batch 3000, loss[loss=0.1348, simple_loss=0.2063, pruned_loss=0.0317, over 4746.00 frames. ], tot_loss[loss=0.1696, simple_loss=0.2432, pruned_loss=0.04797, over 954091.60 frames. ], batch size: 26, lr: 2.92e-03, grad_scale: 16.0 2023-04-28 01:09:47,998 INFO [finetune.py:1001] (1/7) Computing validation loss 2023-04-28 01:09:54,794 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9611, 2.3482, 1.8772, 1.6731, 1.4986, 1.4769, 1.8793, 1.4511], device='cuda:1'), covar=tensor([0.1570, 0.1188, 0.1430, 0.1568, 0.2204, 0.1869, 0.1005, 0.1942], device='cuda:1'), in_proj_covar=tensor([0.0198, 0.0208, 0.0170, 0.0204, 0.0200, 0.0186, 0.0156, 0.0187], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-28 01:09:55,293 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5384, 3.0127, 0.9435, 1.8448, 1.8918, 2.3088, 1.9304, 1.0502], device='cuda:1'), covar=tensor([0.1213, 0.0858, 0.1741, 0.1144, 0.0998, 0.0877, 0.1407, 0.1702], device='cuda:1'), in_proj_covar=tensor([0.0118, 0.0240, 0.0136, 0.0122, 0.0132, 0.0153, 0.0118, 0.0120], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-28 01:10:08,676 INFO [finetune.py:1010] (1/7) Epoch 27, validation: loss=0.1539, simple_loss=0.2224, pruned_loss=0.04268, over 2265189.00 frames. 2023-04-28 01:10:08,676 INFO [finetune.py:1011] (1/7) Maximum memory allocated so far is 6391MB 2023-04-28 01:10:23,031 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=3.61 vs. limit=5.0 2023-04-28 01:10:24,778 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=151938.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:10:32,800 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-04-28 01:10:35,093 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=151954.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:10:45,721 INFO [finetune.py:976] (1/7) Epoch 27, batch 3050, loss[loss=0.2066, simple_loss=0.2721, pruned_loss=0.07056, over 4922.00 frames. ], tot_loss[loss=0.1694, simple_loss=0.2434, pruned_loss=0.04773, over 951927.52 frames. ], batch size: 33, lr: 2.92e-03, grad_scale: 16.0 2023-04-28 01:10:48,112 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.167e+02 1.615e+02 1.912e+02 2.194e+02 5.604e+02, threshold=3.825e+02, percent-clipped=1.0 2023-04-28 01:10:57,067 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=151986.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:11:00,773 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7157, 1.3403, 1.3450, 1.5156, 1.8786, 1.5145, 1.3295, 1.3045], device='cuda:1'), covar=tensor([0.1744, 0.1565, 0.1971, 0.1311, 0.0911, 0.1770, 0.2023, 0.2520], device='cuda:1'), in_proj_covar=tensor([0.0318, 0.0311, 0.0355, 0.0290, 0.0331, 0.0309, 0.0303, 0.0379], device='cuda:1'), out_proj_covar=tensor([6.4963e-05, 6.3877e-05, 7.4671e-05, 5.8039e-05, 6.7573e-05, 6.4570e-05, 6.2674e-05, 8.0278e-05], device='cuda:1') 2023-04-28 01:11:08,925 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=152002.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:11:10,364 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2023-04-28 01:11:12,109 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.2559, 1.6799, 2.1376, 2.4073, 2.0861, 1.6533, 1.3401, 1.8427], device='cuda:1'), covar=tensor([0.2837, 0.3065, 0.1566, 0.2091, 0.2461, 0.2640, 0.4189, 0.2100], device='cuda:1'), in_proj_covar=tensor([0.0291, 0.0245, 0.0227, 0.0313, 0.0220, 0.0233, 0.0226, 0.0185], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-28 01:11:20,459 INFO [finetune.py:976] (1/7) Epoch 27, batch 3100, loss[loss=0.1279, simple_loss=0.1998, pruned_loss=0.02795, over 4724.00 frames. ], tot_loss[loss=0.1682, simple_loss=0.242, pruned_loss=0.04726, over 951625.69 frames. ], batch size: 23, lr: 2.92e-03, grad_scale: 16.0 2023-04-28 01:11:37,595 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=152044.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:11:54,337 INFO [finetune.py:976] (1/7) Epoch 27, batch 3150, loss[loss=0.1879, simple_loss=0.2585, pruned_loss=0.05864, over 4918.00 frames. ], tot_loss[loss=0.1661, simple_loss=0.2394, pruned_loss=0.0464, over 953302.11 frames. ], batch size: 43, lr: 2.92e-03, grad_scale: 16.0 2023-04-28 01:11:56,743 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.740e+01 1.539e+02 1.884e+02 2.286e+02 5.253e+02, threshold=3.767e+02, percent-clipped=2.0 2023-04-28 01:12:27,068 INFO [finetune.py:976] (1/7) Epoch 27, batch 3200, loss[loss=0.1615, simple_loss=0.2363, pruned_loss=0.04337, over 4889.00 frames. ], tot_loss[loss=0.1635, simple_loss=0.2364, pruned_loss=0.04528, over 953126.50 frames. ], batch size: 32, lr: 2.92e-03, grad_scale: 16.0 2023-04-28 01:12:28,998 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([3.8236, 3.7719, 2.6654, 4.4037, 3.8456, 3.8097, 1.7152, 3.7994], device='cuda:1'), covar=tensor([0.1489, 0.1148, 0.3568, 0.1676, 0.3619, 0.1849, 0.5256, 0.2262], device='cuda:1'), in_proj_covar=tensor([0.0244, 0.0217, 0.0250, 0.0300, 0.0297, 0.0246, 0.0272, 0.0272], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 01:12:46,658 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=3.14 vs. limit=5.0 2023-04-28 01:12:47,305 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.56 vs. limit=2.0 2023-04-28 01:12:52,036 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=152157.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:13:00,433 INFO [finetune.py:976] (1/7) Epoch 27, batch 3250, loss[loss=0.1457, simple_loss=0.2166, pruned_loss=0.03738, over 4769.00 frames. ], tot_loss[loss=0.1648, simple_loss=0.2373, pruned_loss=0.04621, over 953486.84 frames. ], batch size: 28, lr: 2.92e-03, grad_scale: 16.0 2023-04-28 01:13:02,806 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.216e+01 1.479e+02 1.800e+02 2.164e+02 4.753e+02, threshold=3.600e+02, percent-clipped=3.0 2023-04-28 01:13:14,509 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([3.3831, 3.3684, 2.5267, 3.8633, 3.4102, 3.3126, 1.5207, 3.2972], device='cuda:1'), covar=tensor([0.1719, 0.1310, 0.3248, 0.2197, 0.4353, 0.1895, 0.5823, 0.2650], device='cuda:1'), in_proj_covar=tensor([0.0244, 0.0217, 0.0249, 0.0300, 0.0297, 0.0246, 0.0272, 0.0272], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 01:13:33,550 INFO [finetune.py:976] (1/7) Epoch 27, batch 3300, loss[loss=0.1762, simple_loss=0.2602, pruned_loss=0.04611, over 4821.00 frames. ], tot_loss[loss=0.168, simple_loss=0.2412, pruned_loss=0.04742, over 954910.90 frames. ], batch size: 40, lr: 2.92e-03, grad_scale: 16.0 2023-04-28 01:13:33,700 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.5071, 2.0083, 2.3693, 2.9081, 2.3724, 1.8789, 1.8972, 2.2120], device='cuda:1'), covar=tensor([0.3060, 0.3133, 0.1564, 0.2437, 0.2726, 0.2614, 0.3545, 0.2126], device='cuda:1'), in_proj_covar=tensor([0.0291, 0.0246, 0.0227, 0.0313, 0.0220, 0.0233, 0.0227, 0.0185], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-28 01:13:36,738 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=152225.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:13:46,837 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.17 vs. limit=5.0 2023-04-28 01:14:13,061 INFO [finetune.py:976] (1/7) Epoch 27, batch 3350, loss[loss=0.1671, simple_loss=0.2433, pruned_loss=0.04542, over 4766.00 frames. ], tot_loss[loss=0.1676, simple_loss=0.2412, pruned_loss=0.04694, over 953464.13 frames. ], batch size: 28, lr: 2.92e-03, grad_scale: 16.0 2023-04-28 01:14:14,903 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=152273.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:14:15,409 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.182e+02 1.492e+02 1.749e+02 2.149e+02 5.486e+02, threshold=3.498e+02, percent-clipped=3.0 2023-04-28 01:14:34,636 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=152286.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:14:46,566 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5185, 1.3612, 1.7546, 1.8070, 1.3393, 1.2957, 1.4933, 1.0118], device='cuda:1'), covar=tensor([0.0511, 0.0670, 0.0357, 0.0573, 0.0779, 0.1084, 0.0510, 0.0567], device='cuda:1'), in_proj_covar=tensor([0.0070, 0.0068, 0.0066, 0.0069, 0.0075, 0.0095, 0.0073, 0.0063], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-28 01:15:17,722 INFO [finetune.py:976] (1/7) Epoch 27, batch 3400, loss[loss=0.167, simple_loss=0.2453, pruned_loss=0.04432, over 4889.00 frames. ], tot_loss[loss=0.1673, simple_loss=0.241, pruned_loss=0.04684, over 950956.98 frames. ], batch size: 32, lr: 2.92e-03, grad_scale: 32.0 2023-04-28 01:15:36,890 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=152334.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:15:49,715 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=152344.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:16:01,848 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.1001, 1.9059, 2.5051, 2.6342, 1.7856, 1.6350, 2.0116, 1.2047], device='cuda:1'), covar=tensor([0.0516, 0.0708, 0.0331, 0.0623, 0.0771, 0.1044, 0.0668, 0.0677], device='cuda:1'), in_proj_covar=tensor([0.0070, 0.0068, 0.0066, 0.0069, 0.0075, 0.0095, 0.0073, 0.0063], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-28 01:16:22,888 INFO [finetune.py:976] (1/7) Epoch 27, batch 3450, loss[loss=0.1479, simple_loss=0.2257, pruned_loss=0.03501, over 4819.00 frames. ], tot_loss[loss=0.1656, simple_loss=0.2396, pruned_loss=0.0458, over 948805.24 frames. ], batch size: 39, lr: 2.91e-03, grad_scale: 32.0 2023-04-28 01:16:31,452 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.088e+02 1.589e+02 1.871e+02 2.255e+02 4.038e+02, threshold=3.742e+02, percent-clipped=2.0 2023-04-28 01:16:53,820 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=152392.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:17:28,481 INFO [finetune.py:976] (1/7) Epoch 27, batch 3500, loss[loss=0.2082, simple_loss=0.2716, pruned_loss=0.0724, over 4892.00 frames. ], tot_loss[loss=0.1644, simple_loss=0.238, pruned_loss=0.04544, over 951434.75 frames. ], batch size: 32, lr: 2.91e-03, grad_scale: 32.0 2023-04-28 01:17:37,762 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([0.7859, 1.6970, 1.7151, 1.3727, 1.7949, 1.5042, 2.3194, 1.5446], device='cuda:1'), covar=tensor([0.3739, 0.1966, 0.4896, 0.2856, 0.1561, 0.2406, 0.1598, 0.4411], device='cuda:1'), in_proj_covar=tensor([0.0336, 0.0352, 0.0420, 0.0349, 0.0378, 0.0373, 0.0365, 0.0419], device='cuda:1'), out_proj_covar=tensor([9.9175e-05, 1.0492e-04, 1.2712e-04, 1.0440e-04, 1.1199e-04, 1.1081e-04, 1.0669e-04, 1.2595e-04], device='cuda:1') 2023-04-28 01:18:00,017 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=152445.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:18:18,983 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=152457.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:18:32,262 INFO [finetune.py:976] (1/7) Epoch 27, batch 3550, loss[loss=0.1322, simple_loss=0.209, pruned_loss=0.02772, over 4773.00 frames. ], tot_loss[loss=0.1615, simple_loss=0.2346, pruned_loss=0.04418, over 950896.95 frames. ], batch size: 27, lr: 2.91e-03, grad_scale: 32.0 2023-04-28 01:18:34,714 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.620e+01 1.515e+02 1.760e+02 2.193e+02 3.921e+02, threshold=3.521e+02, percent-clipped=1.0 2023-04-28 01:19:25,032 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=152505.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:19:25,759 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=152506.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:19:29,430 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5379, 1.4162, 1.7877, 1.8418, 1.3583, 1.3175, 1.4957, 0.8993], device='cuda:1'), covar=tensor([0.0510, 0.0521, 0.0381, 0.0405, 0.0779, 0.1091, 0.0480, 0.0555], device='cuda:1'), in_proj_covar=tensor([0.0070, 0.0068, 0.0066, 0.0069, 0.0075, 0.0095, 0.0073, 0.0063], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-28 01:19:39,436 INFO [finetune.py:976] (1/7) Epoch 27, batch 3600, loss[loss=0.1667, simple_loss=0.2549, pruned_loss=0.03927, over 4915.00 frames. ], tot_loss[loss=0.1586, simple_loss=0.2313, pruned_loss=0.04294, over 952425.04 frames. ], batch size: 37, lr: 2.91e-03, grad_scale: 32.0 2023-04-28 01:20:33,895 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.3771, 1.7201, 2.2753, 2.6222, 2.2852, 1.7788, 1.6238, 2.0524], device='cuda:1'), covar=tensor([0.3071, 0.3009, 0.1584, 0.2266, 0.2430, 0.2526, 0.3874, 0.1898], device='cuda:1'), in_proj_covar=tensor([0.0297, 0.0249, 0.0230, 0.0318, 0.0224, 0.0236, 0.0230, 0.0188], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-28 01:20:44,730 INFO [finetune.py:976] (1/7) Epoch 27, batch 3650, loss[loss=0.1576, simple_loss=0.2355, pruned_loss=0.0398, over 4741.00 frames. ], tot_loss[loss=0.1604, simple_loss=0.2332, pruned_loss=0.04385, over 952930.29 frames. ], batch size: 27, lr: 2.91e-03, grad_scale: 32.0 2023-04-28 01:20:46,021 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([0.9084, 1.2328, 3.3120, 3.0910, 2.9785, 3.2616, 3.2586, 2.9386], device='cuda:1'), covar=tensor([0.7006, 0.5236, 0.1390, 0.1974, 0.1353, 0.1982, 0.1386, 0.1621], device='cuda:1'), in_proj_covar=tensor([0.0314, 0.0309, 0.0411, 0.0409, 0.0350, 0.0418, 0.0321, 0.0367], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-28 01:20:51,634 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.035e+02 1.532e+02 1.875e+02 2.206e+02 4.612e+02, threshold=3.749e+02, percent-clipped=4.0 2023-04-28 01:20:56,029 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=152581.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:21:02,384 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=152582.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:21:07,227 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=152589.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:21:49,538 INFO [finetune.py:976] (1/7) Epoch 27, batch 3700, loss[loss=0.1282, simple_loss=0.2093, pruned_loss=0.02357, over 4782.00 frames. ], tot_loss[loss=0.1624, simple_loss=0.2361, pruned_loss=0.04437, over 954452.44 frames. ], batch size: 26, lr: 2.91e-03, grad_scale: 32.0 2023-04-28 01:22:00,777 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=152629.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:22:21,227 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=152643.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:22:31,082 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=152650.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:22:56,764 INFO [finetune.py:976] (1/7) Epoch 27, batch 3750, loss[loss=0.1909, simple_loss=0.243, pruned_loss=0.0694, over 4722.00 frames. ], tot_loss[loss=0.1643, simple_loss=0.2386, pruned_loss=0.04499, over 955501.50 frames. ], batch size: 23, lr: 2.91e-03, grad_scale: 32.0 2023-04-28 01:23:04,991 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.032e+02 1.529e+02 1.747e+02 2.097e+02 4.287e+02, threshold=3.495e+02, percent-clipped=2.0 2023-04-28 01:23:37,927 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0697, 1.8156, 2.1751, 2.4998, 2.5296, 2.0568, 1.7690, 2.2532], device='cuda:1'), covar=tensor([0.0946, 0.1178, 0.0740, 0.0649, 0.0648, 0.0869, 0.0769, 0.0609], device='cuda:1'), in_proj_covar=tensor([0.0182, 0.0199, 0.0182, 0.0168, 0.0176, 0.0176, 0.0149, 0.0176], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-28 01:24:08,007 INFO [finetune.py:976] (1/7) Epoch 27, batch 3800, loss[loss=0.2068, simple_loss=0.2664, pruned_loss=0.07363, over 4242.00 frames. ], tot_loss[loss=0.165, simple_loss=0.2394, pruned_loss=0.04531, over 955251.81 frames. ], batch size: 66, lr: 2.91e-03, grad_scale: 32.0 2023-04-28 01:24:11,818 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=152726.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:24:29,224 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.2251, 1.1589, 3.8434, 3.6006, 3.3719, 3.7153, 3.6891, 3.4003], device='cuda:1'), covar=tensor([0.7670, 0.6137, 0.1212, 0.1766, 0.1272, 0.1776, 0.1537, 0.1647], device='cuda:1'), in_proj_covar=tensor([0.0314, 0.0310, 0.0411, 0.0409, 0.0350, 0.0418, 0.0321, 0.0366], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-28 01:24:31,130 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8453, 1.5092, 1.9730, 2.3035, 1.9496, 1.8462, 1.9294, 1.8415], device='cuda:1'), covar=tensor([0.4815, 0.7402, 0.6789, 0.5906, 0.6372, 0.8758, 0.8679, 1.0140], device='cuda:1'), in_proj_covar=tensor([0.0442, 0.0421, 0.0517, 0.0507, 0.0469, 0.0506, 0.0505, 0.0522], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-28 01:25:13,179 INFO [finetune.py:976] (1/7) Epoch 27, batch 3850, loss[loss=0.1572, simple_loss=0.2311, pruned_loss=0.04165, over 4816.00 frames. ], tot_loss[loss=0.1644, simple_loss=0.2388, pruned_loss=0.04501, over 957318.42 frames. ], batch size: 41, lr: 2.91e-03, grad_scale: 32.0 2023-04-28 01:25:15,589 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.111e+02 1.420e+02 1.784e+02 2.045e+02 3.500e+02, threshold=3.567e+02, percent-clipped=1.0 2023-04-28 01:25:24,506 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.3666, 1.7260, 1.8609, 1.9693, 1.8540, 1.8825, 1.8950, 1.8871], device='cuda:1'), covar=tensor([0.3931, 0.4982, 0.4098, 0.4262, 0.5182, 0.6645, 0.4648, 0.4675], device='cuda:1'), in_proj_covar=tensor([0.0343, 0.0375, 0.0330, 0.0341, 0.0350, 0.0394, 0.0361, 0.0335], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 01:25:33,565 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=152787.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:25:47,900 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=152801.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:26:06,637 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.7726, 3.3778, 3.0963, 3.1630, 2.3541, 2.9390, 3.0774, 2.2797], device='cuda:1'), covar=tensor([0.1611, 0.0995, 0.0557, 0.1016, 0.2590, 0.0961, 0.1417, 0.2326], device='cuda:1'), in_proj_covar=tensor([0.0282, 0.0300, 0.0215, 0.0275, 0.0314, 0.0253, 0.0246, 0.0263], device='cuda:1'), out_proj_covar=tensor([1.1261e-04, 1.1805e-04, 8.4708e-05, 1.0846e-04, 1.2637e-04, 9.9421e-05, 9.9246e-05, 1.0392e-04], device='cuda:1') 2023-04-28 01:26:08,467 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=5.06 vs. limit=5.0 2023-04-28 01:26:17,504 INFO [finetune.py:976] (1/7) Epoch 27, batch 3900, loss[loss=0.1273, simple_loss=0.2033, pruned_loss=0.02563, over 4899.00 frames. ], tot_loss[loss=0.164, simple_loss=0.2374, pruned_loss=0.04535, over 955406.66 frames. ], batch size: 35, lr: 2.91e-03, grad_scale: 32.0 2023-04-28 01:26:53,006 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.54 vs. limit=2.0 2023-04-28 01:27:01,617 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-04-28 01:27:12,211 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0028, 2.6360, 1.0752, 1.3784, 1.9913, 1.1324, 3.3441, 1.6676], device='cuda:1'), covar=tensor([0.0863, 0.0779, 0.0890, 0.1472, 0.0632, 0.1240, 0.0291, 0.0845], device='cuda:1'), in_proj_covar=tensor([0.0050, 0.0064, 0.0047, 0.0046, 0.0049, 0.0051, 0.0072, 0.0050], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:1') 2023-04-28 01:27:22,910 INFO [finetune.py:976] (1/7) Epoch 27, batch 3950, loss[loss=0.1448, simple_loss=0.2104, pruned_loss=0.03959, over 4837.00 frames. ], tot_loss[loss=0.1618, simple_loss=0.2344, pruned_loss=0.0446, over 953419.33 frames. ], batch size: 25, lr: 2.91e-03, grad_scale: 32.0 2023-04-28 01:27:26,329 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.030e+02 1.447e+02 1.748e+02 2.157e+02 5.230e+02, threshold=3.496e+02, percent-clipped=2.0 2023-04-28 01:27:36,700 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=152881.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:28:27,783 INFO [finetune.py:976] (1/7) Epoch 27, batch 4000, loss[loss=0.1446, simple_loss=0.2279, pruned_loss=0.03059, over 4755.00 frames. ], tot_loss[loss=0.1615, simple_loss=0.2338, pruned_loss=0.04456, over 952845.92 frames. ], batch size: 54, lr: 2.91e-03, grad_scale: 32.0 2023-04-28 01:28:39,697 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=152929.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:28:39,741 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=152929.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:28:50,581 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=152938.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:29:00,125 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=152945.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:29:32,517 INFO [finetune.py:976] (1/7) Epoch 27, batch 4050, loss[loss=0.1545, simple_loss=0.2283, pruned_loss=0.04034, over 4865.00 frames. ], tot_loss[loss=0.1658, simple_loss=0.2382, pruned_loss=0.04673, over 951720.15 frames. ], batch size: 31, lr: 2.91e-03, grad_scale: 32.0 2023-04-28 01:29:35,454 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.138e+02 1.640e+02 1.918e+02 2.261e+02 4.754e+02, threshold=3.835e+02, percent-clipped=2.0 2023-04-28 01:29:43,788 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=152977.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:30:36,411 INFO [finetune.py:976] (1/7) Epoch 27, batch 4100, loss[loss=0.1551, simple_loss=0.2346, pruned_loss=0.0378, over 4915.00 frames. ], tot_loss[loss=0.1678, simple_loss=0.2409, pruned_loss=0.04734, over 952168.50 frames. ], batch size: 36, lr: 2.91e-03, grad_scale: 32.0 2023-04-28 01:30:58,686 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.1724, 1.7477, 2.0817, 2.4887, 2.4413, 2.0987, 1.8361, 2.2438], device='cuda:1'), covar=tensor([0.0830, 0.1124, 0.0670, 0.0557, 0.0588, 0.0846, 0.0671, 0.0542], device='cuda:1'), in_proj_covar=tensor([0.0183, 0.0199, 0.0182, 0.0169, 0.0176, 0.0176, 0.0149, 0.0176], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-28 01:31:40,934 INFO [finetune.py:976] (1/7) Epoch 27, batch 4150, loss[loss=0.2035, simple_loss=0.2696, pruned_loss=0.06868, over 4841.00 frames. ], tot_loss[loss=0.1687, simple_loss=0.2422, pruned_loss=0.04757, over 952290.15 frames. ], batch size: 49, lr: 2.91e-03, grad_scale: 32.0 2023-04-28 01:31:43,380 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.012e+02 1.659e+02 1.933e+02 2.318e+02 5.641e+02, threshold=3.866e+02, percent-clipped=2.0 2023-04-28 01:31:54,614 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=153082.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:32:22,859 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=153101.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:32:43,321 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=153117.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:32:45,074 INFO [finetune.py:976] (1/7) Epoch 27, batch 4200, loss[loss=0.1617, simple_loss=0.2359, pruned_loss=0.04373, over 4840.00 frames. ], tot_loss[loss=0.1685, simple_loss=0.2426, pruned_loss=0.04718, over 952997.65 frames. ], batch size: 49, lr: 2.91e-03, grad_scale: 32.0 2023-04-28 01:33:26,251 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=153149.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:33:49,691 INFO [finetune.py:976] (1/7) Epoch 27, batch 4250, loss[loss=0.1679, simple_loss=0.2485, pruned_loss=0.04363, over 4904.00 frames. ], tot_loss[loss=0.1661, simple_loss=0.24, pruned_loss=0.0461, over 954271.69 frames. ], batch size: 36, lr: 2.91e-03, grad_scale: 32.0 2023-04-28 01:33:57,257 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.090e+02 1.449e+02 1.716e+02 2.053e+02 3.736e+02, threshold=3.432e+02, percent-clipped=0.0 2023-04-28 01:33:59,832 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=153178.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:34:54,745 INFO [finetune.py:976] (1/7) Epoch 27, batch 4300, loss[loss=0.1433, simple_loss=0.2184, pruned_loss=0.03408, over 4895.00 frames. ], tot_loss[loss=0.1637, simple_loss=0.2369, pruned_loss=0.04525, over 954908.40 frames. ], batch size: 32, lr: 2.91e-03, grad_scale: 32.0 2023-04-28 01:35:03,530 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7681, 1.0192, 1.7672, 2.2110, 1.8390, 1.7025, 1.7391, 1.6819], device='cuda:1'), covar=tensor([0.4159, 0.6766, 0.5627, 0.4955, 0.5409, 0.6895, 0.7272, 0.8884], device='cuda:1'), in_proj_covar=tensor([0.0442, 0.0422, 0.0517, 0.0507, 0.0469, 0.0506, 0.0507, 0.0523], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-28 01:35:17,363 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=153238.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:35:23,024 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.4729, 1.6513, 1.8562, 1.9861, 1.8037, 1.9534, 1.8774, 1.9522], device='cuda:1'), covar=tensor([0.3686, 0.5310, 0.4351, 0.4169, 0.5417, 0.6523, 0.5088, 0.4780], device='cuda:1'), in_proj_covar=tensor([0.0341, 0.0373, 0.0329, 0.0340, 0.0349, 0.0393, 0.0360, 0.0333], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 01:35:23,569 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=153245.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:35:38,705 INFO [finetune.py:976] (1/7) Epoch 27, batch 4350, loss[loss=0.1475, simple_loss=0.2277, pruned_loss=0.0337, over 4828.00 frames. ], tot_loss[loss=0.1622, simple_loss=0.2351, pruned_loss=0.04472, over 955351.52 frames. ], batch size: 33, lr: 2.91e-03, grad_scale: 32.0 2023-04-28 01:35:38,838 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5163, 1.4364, 1.6799, 1.8046, 1.3544, 1.2672, 1.4163, 0.9066], device='cuda:1'), covar=tensor([0.0484, 0.0551, 0.0379, 0.0575, 0.0764, 0.1012, 0.0538, 0.0551], device='cuda:1'), in_proj_covar=tensor([0.0070, 0.0067, 0.0065, 0.0069, 0.0074, 0.0094, 0.0072, 0.0063], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-28 01:35:41,097 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.135e+02 1.585e+02 1.848e+02 2.377e+02 5.336e+02, threshold=3.696e+02, percent-clipped=3.0 2023-04-28 01:35:44,150 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=153278.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:35:48,966 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=153286.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:35:53,607 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([4.4650, 4.3914, 3.0927, 5.1765, 4.5347, 4.4238, 1.9234, 4.4228], device='cuda:1'), covar=tensor([0.1669, 0.1061, 0.3311, 0.1024, 0.4073, 0.1730, 0.6072, 0.2260], device='cuda:1'), in_proj_covar=tensor([0.0244, 0.0218, 0.0251, 0.0303, 0.0299, 0.0247, 0.0274, 0.0273], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 01:35:54,194 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=153293.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:36:12,553 INFO [finetune.py:976] (1/7) Epoch 27, batch 4400, loss[loss=0.1736, simple_loss=0.2495, pruned_loss=0.04883, over 4913.00 frames. ], tot_loss[loss=0.1644, simple_loss=0.2371, pruned_loss=0.04588, over 955671.01 frames. ], batch size: 37, lr: 2.91e-03, grad_scale: 32.0 2023-04-28 01:36:15,823 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-04-28 01:36:33,333 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=153339.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:37:16,504 INFO [finetune.py:976] (1/7) Epoch 27, batch 4450, loss[loss=0.148, simple_loss=0.2335, pruned_loss=0.03123, over 4905.00 frames. ], tot_loss[loss=0.1656, simple_loss=0.2396, pruned_loss=0.04583, over 956113.12 frames. ], batch size: 37, lr: 2.91e-03, grad_scale: 32.0 2023-04-28 01:37:18,370 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5468, 3.3720, 1.0324, 1.9780, 1.8135, 2.3519, 2.0240, 1.1517], device='cuda:1'), covar=tensor([0.1415, 0.0964, 0.1880, 0.1180, 0.1108, 0.1095, 0.1463, 0.2027], device='cuda:1'), in_proj_covar=tensor([0.0118, 0.0240, 0.0136, 0.0121, 0.0132, 0.0153, 0.0118, 0.0119], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-28 01:37:18,874 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.171e+02 1.598e+02 1.820e+02 2.307e+02 3.312e+02, threshold=3.640e+02, percent-clipped=0.0 2023-04-28 01:37:30,039 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=153382.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:38:10,740 INFO [finetune.py:976] (1/7) Epoch 27, batch 4500, loss[loss=0.1976, simple_loss=0.2756, pruned_loss=0.05974, over 4795.00 frames. ], tot_loss[loss=0.1662, simple_loss=0.2403, pruned_loss=0.04608, over 955966.30 frames. ], batch size: 45, lr: 2.91e-03, grad_scale: 32.0 2023-04-28 01:38:11,452 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=153421.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:38:16,906 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=153430.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:38:43,736 INFO [finetune.py:976] (1/7) Epoch 27, batch 4550, loss[loss=0.1803, simple_loss=0.266, pruned_loss=0.0473, over 4756.00 frames. ], tot_loss[loss=0.1665, simple_loss=0.241, pruned_loss=0.04595, over 953548.41 frames. ], batch size: 54, lr: 2.91e-03, grad_scale: 32.0 2023-04-28 01:38:45,569 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=153473.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:38:46,109 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.179e+01 1.577e+02 1.859e+02 2.196e+02 3.775e+02, threshold=3.717e+02, percent-clipped=3.0 2023-04-28 01:38:51,074 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=153482.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:39:15,744 INFO [finetune.py:976] (1/7) Epoch 27, batch 4600, loss[loss=0.1207, simple_loss=0.1867, pruned_loss=0.02732, over 4338.00 frames. ], tot_loss[loss=0.1671, simple_loss=0.2415, pruned_loss=0.04636, over 953962.24 frames. ], batch size: 18, lr: 2.91e-03, grad_scale: 32.0 2023-04-28 01:39:17,144 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6370, 0.6885, 1.5411, 1.9730, 1.7366, 1.5358, 1.5581, 1.5759], device='cuda:1'), covar=tensor([0.4334, 0.6591, 0.6332, 0.5730, 0.5744, 0.7298, 0.7852, 0.8339], device='cuda:1'), in_proj_covar=tensor([0.0444, 0.0424, 0.0520, 0.0509, 0.0471, 0.0509, 0.0510, 0.0525], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-28 01:39:48,983 INFO [finetune.py:976] (1/7) Epoch 27, batch 4650, loss[loss=0.1384, simple_loss=0.211, pruned_loss=0.0329, over 4779.00 frames. ], tot_loss[loss=0.1664, simple_loss=0.24, pruned_loss=0.04644, over 956163.15 frames. ], batch size: 29, lr: 2.91e-03, grad_scale: 32.0 2023-04-28 01:39:51,379 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.905e+01 1.489e+02 1.829e+02 2.275e+02 4.563e+02, threshold=3.657e+02, percent-clipped=2.0 2023-04-28 01:39:54,513 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7285, 2.1367, 1.9217, 2.1001, 1.5629, 1.8686, 1.7787, 1.4786], device='cuda:1'), covar=tensor([0.1807, 0.1270, 0.0768, 0.0981, 0.3351, 0.1103, 0.1807, 0.2297], device='cuda:1'), in_proj_covar=tensor([0.0284, 0.0300, 0.0216, 0.0276, 0.0315, 0.0254, 0.0248, 0.0264], device='cuda:1'), out_proj_covar=tensor([1.1315e-04, 1.1792e-04, 8.4781e-05, 1.0867e-04, 1.2705e-04, 9.9619e-05, 9.9874e-05, 1.0408e-04], device='cuda:1') 2023-04-28 01:39:59,494 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.64 vs. limit=5.0 2023-04-28 01:40:09,739 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-04-28 01:40:27,427 INFO [finetune.py:976] (1/7) Epoch 27, batch 4700, loss[loss=0.1478, simple_loss=0.2184, pruned_loss=0.03864, over 4778.00 frames. ], tot_loss[loss=0.1642, simple_loss=0.2372, pruned_loss=0.04565, over 958073.66 frames. ], batch size: 26, lr: 2.91e-03, grad_scale: 32.0 2023-04-28 01:40:28,909 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-04-28 01:40:41,146 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=153634.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:41:06,299 INFO [finetune.py:976] (1/7) Epoch 27, batch 4750, loss[loss=0.2006, simple_loss=0.2623, pruned_loss=0.06943, over 4825.00 frames. ], tot_loss[loss=0.163, simple_loss=0.2353, pruned_loss=0.0453, over 958703.19 frames. ], batch size: 39, lr: 2.91e-03, grad_scale: 32.0 2023-04-28 01:41:08,706 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.046e+02 1.498e+02 1.801e+02 2.144e+02 5.776e+02, threshold=3.603e+02, percent-clipped=1.0 2023-04-28 01:41:39,703 INFO [finetune.py:976] (1/7) Epoch 27, batch 4800, loss[loss=0.1959, simple_loss=0.2907, pruned_loss=0.0506, over 4800.00 frames. ], tot_loss[loss=0.1649, simple_loss=0.2377, pruned_loss=0.04605, over 958720.87 frames. ], batch size: 45, lr: 2.91e-03, grad_scale: 32.0 2023-04-28 01:41:47,007 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.1459, 2.6565, 2.1513, 2.0587, 1.6124, 1.6177, 2.2274, 1.5507], device='cuda:1'), covar=tensor([0.1556, 0.1256, 0.1355, 0.1500, 0.2178, 0.1748, 0.0887, 0.1873], device='cuda:1'), in_proj_covar=tensor([0.0197, 0.0209, 0.0169, 0.0204, 0.0200, 0.0186, 0.0155, 0.0187], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-28 01:42:13,213 INFO [finetune.py:976] (1/7) Epoch 27, batch 4850, loss[loss=0.2023, simple_loss=0.2754, pruned_loss=0.06459, over 4739.00 frames. ], tot_loss[loss=0.1683, simple_loss=0.2419, pruned_loss=0.04733, over 958630.53 frames. ], batch size: 59, lr: 2.91e-03, grad_scale: 32.0 2023-04-28 01:42:15,131 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=153773.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:42:16,120 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.154e+02 1.558e+02 1.785e+02 2.146e+02 3.572e+02, threshold=3.570e+02, percent-clipped=0.0 2023-04-28 01:42:17,991 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=153777.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:42:44,720 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([0.9145, 1.8610, 1.7268, 1.4481, 1.9825, 1.6296, 2.5364, 1.5400], device='cuda:1'), covar=tensor([0.3848, 0.2169, 0.5012, 0.3247, 0.1778, 0.2419, 0.1302, 0.4725], device='cuda:1'), in_proj_covar=tensor([0.0338, 0.0355, 0.0424, 0.0351, 0.0380, 0.0374, 0.0369, 0.0422], device='cuda:1'), out_proj_covar=tensor([9.9878e-05, 1.0563e-04, 1.2844e-04, 1.0511e-04, 1.1236e-04, 1.1121e-04, 1.0798e-04, 1.2682e-04], device='cuda:1') 2023-04-28 01:43:05,705 INFO [finetune.py:976] (1/7) Epoch 27, batch 4900, loss[loss=0.1708, simple_loss=0.2313, pruned_loss=0.05511, over 4673.00 frames. ], tot_loss[loss=0.1688, simple_loss=0.2431, pruned_loss=0.04727, over 958001.16 frames. ], batch size: 23, lr: 2.91e-03, grad_scale: 32.0 2023-04-28 01:43:06,910 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=153821.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:43:36,565 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([4.6964, 4.5905, 3.0879, 5.3353, 4.7206, 4.6679, 1.9306, 4.5630], device='cuda:1'), covar=tensor([0.1664, 0.1007, 0.3357, 0.0847, 0.2780, 0.1459, 0.5785, 0.2351], device='cuda:1'), in_proj_covar=tensor([0.0244, 0.0218, 0.0250, 0.0302, 0.0297, 0.0246, 0.0273, 0.0273], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 01:44:09,302 INFO [finetune.py:976] (1/7) Epoch 27, batch 4950, loss[loss=0.1517, simple_loss=0.2338, pruned_loss=0.03476, over 4850.00 frames. ], tot_loss[loss=0.1698, simple_loss=0.2443, pruned_loss=0.04763, over 955998.25 frames. ], batch size: 31, lr: 2.90e-03, grad_scale: 32.0 2023-04-28 01:44:18,025 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.075e+02 1.601e+02 1.869e+02 2.254e+02 5.628e+02, threshold=3.738e+02, percent-clipped=5.0 2023-04-28 01:44:37,774 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.30 vs. limit=5.0 2023-04-28 01:44:51,815 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.1989, 2.5430, 1.1058, 1.3741, 1.9571, 1.2777, 3.0871, 1.7261], device='cuda:1'), covar=tensor([0.0662, 0.0591, 0.0745, 0.1222, 0.0486, 0.1007, 0.0327, 0.0628], device='cuda:1'), in_proj_covar=tensor([0.0051, 0.0064, 0.0047, 0.0046, 0.0050, 0.0051, 0.0073, 0.0051], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:1') 2023-04-28 01:45:13,176 INFO [finetune.py:976] (1/7) Epoch 27, batch 5000, loss[loss=0.1343, simple_loss=0.205, pruned_loss=0.03179, over 4782.00 frames. ], tot_loss[loss=0.1694, simple_loss=0.2433, pruned_loss=0.04778, over 955522.73 frames. ], batch size: 26, lr: 2.90e-03, grad_scale: 32.0 2023-04-28 01:45:13,353 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=3.70 vs. limit=5.0 2023-04-28 01:45:15,636 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=153923.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:45:34,545 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=153934.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:45:55,356 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.6858, 2.1560, 2.5983, 3.1016, 2.5062, 2.0246, 2.0152, 2.4343], device='cuda:1'), covar=tensor([0.3014, 0.2799, 0.1457, 0.2244, 0.2563, 0.2423, 0.3518, 0.1864], device='cuda:1'), in_proj_covar=tensor([0.0292, 0.0246, 0.0228, 0.0314, 0.0222, 0.0234, 0.0228, 0.0186], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-28 01:46:17,424 INFO [finetune.py:976] (1/7) Epoch 27, batch 5050, loss[loss=0.1102, simple_loss=0.1711, pruned_loss=0.02464, over 4080.00 frames. ], tot_loss[loss=0.1671, simple_loss=0.2399, pruned_loss=0.04714, over 953872.36 frames. ], batch size: 17, lr: 2.90e-03, grad_scale: 32.0 2023-04-28 01:46:25,376 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.912e+01 1.615e+02 1.861e+02 2.291e+02 3.934e+02, threshold=3.722e+02, percent-clipped=1.0 2023-04-28 01:46:32,334 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=153982.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:46:33,619 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=153984.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:46:40,251 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7342, 1.6202, 1.8495, 2.0739, 2.1405, 1.6056, 1.3767, 1.8655], device='cuda:1'), covar=tensor([0.0859, 0.1222, 0.0775, 0.0666, 0.0641, 0.0846, 0.0758, 0.0603], device='cuda:1'), in_proj_covar=tensor([0.0187, 0.0204, 0.0186, 0.0173, 0.0180, 0.0181, 0.0153, 0.0180], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-28 01:46:45,489 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=3.80 vs. limit=5.0 2023-04-28 01:46:57,619 INFO [finetune.py:976] (1/7) Epoch 27, batch 5100, loss[loss=0.1668, simple_loss=0.2337, pruned_loss=0.04991, over 4828.00 frames. ], tot_loss[loss=0.1651, simple_loss=0.2373, pruned_loss=0.04642, over 955311.83 frames. ], batch size: 30, lr: 2.90e-03, grad_scale: 32.0 2023-04-28 01:47:18,522 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=154049.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 01:47:26,084 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.52 vs. limit=5.0 2023-04-28 01:47:27,702 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5178, 1.3118, 3.9187, 3.6471, 3.3895, 3.6492, 3.6308, 3.4437], device='cuda:1'), covar=tensor([0.7344, 0.5897, 0.1027, 0.1698, 0.1158, 0.1846, 0.2265, 0.1628], device='cuda:1'), in_proj_covar=tensor([0.0312, 0.0309, 0.0408, 0.0408, 0.0348, 0.0417, 0.0318, 0.0364], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-28 01:47:31,249 INFO [finetune.py:976] (1/7) Epoch 27, batch 5150, loss[loss=0.09852, simple_loss=0.1682, pruned_loss=0.0144, over 4693.00 frames. ], tot_loss[loss=0.1638, simple_loss=0.2362, pruned_loss=0.04571, over 954225.49 frames. ], batch size: 23, lr: 2.90e-03, grad_scale: 32.0 2023-04-28 01:47:33,690 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.527e+01 1.424e+02 1.690e+02 2.200e+02 4.419e+02, threshold=3.381e+02, percent-clipped=1.0 2023-04-28 01:47:35,035 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=154076.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:47:35,618 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=154077.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:47:40,849 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=154084.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:47:53,828 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.5770, 2.9920, 1.9217, 2.2821, 3.0734, 2.4275, 2.3221, 2.5076], device='cuda:1'), covar=tensor([0.0373, 0.0257, 0.0229, 0.0445, 0.0186, 0.0440, 0.0385, 0.0457], device='cuda:1'), in_proj_covar=tensor([0.0027, 0.0023, 0.0022, 0.0028, 0.0019, 0.0027, 0.0027, 0.0028], device='cuda:1'), out_proj_covar=tensor([0.0052, 0.0046, 0.0038, 0.0053, 0.0039, 0.0051, 0.0050, 0.0052], device='cuda:1') 2023-04-28 01:48:05,734 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.58 vs. limit=2.0 2023-04-28 01:48:15,442 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=154110.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 01:48:26,894 INFO [finetune.py:976] (1/7) Epoch 27, batch 5200, loss[loss=0.1156, simple_loss=0.1983, pruned_loss=0.01645, over 4775.00 frames. ], tot_loss[loss=0.1661, simple_loss=0.2397, pruned_loss=0.04628, over 953850.53 frames. ], batch size: 26, lr: 2.90e-03, grad_scale: 32.0 2023-04-28 01:48:35,422 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=154125.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:48:49,923 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=154137.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 01:49:00,230 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=154145.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:49:20,755 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.8942, 2.9295, 2.1436, 3.3128, 2.9367, 2.9249, 1.1713, 2.7805], device='cuda:1'), covar=tensor([0.2065, 0.1473, 0.3324, 0.2734, 0.3272, 0.2060, 0.5614, 0.2995], device='cuda:1'), in_proj_covar=tensor([0.0244, 0.0218, 0.0250, 0.0303, 0.0298, 0.0247, 0.0272, 0.0272], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 01:49:21,558 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.87 vs. limit=5.0 2023-04-28 01:49:31,678 INFO [finetune.py:976] (1/7) Epoch 27, batch 5250, loss[loss=0.182, simple_loss=0.2588, pruned_loss=0.05258, over 4819.00 frames. ], tot_loss[loss=0.1681, simple_loss=0.2421, pruned_loss=0.04704, over 952438.57 frames. ], batch size: 38, lr: 2.90e-03, grad_scale: 32.0 2023-04-28 01:49:34,127 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.088e+02 1.655e+02 1.963e+02 2.223e+02 5.480e+02, threshold=3.927e+02, percent-clipped=5.0 2023-04-28 01:50:28,037 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([0.7165, 1.6661, 1.7204, 1.3368, 1.7419, 1.5083, 2.2592, 1.4913], device='cuda:1'), covar=tensor([0.3376, 0.1907, 0.4176, 0.2728, 0.1515, 0.2161, 0.1370, 0.4388], device='cuda:1'), in_proj_covar=tensor([0.0337, 0.0353, 0.0423, 0.0349, 0.0378, 0.0374, 0.0368, 0.0421], device='cuda:1'), out_proj_covar=tensor([9.9585e-05, 1.0508e-04, 1.2804e-04, 1.0459e-04, 1.1200e-04, 1.1120e-04, 1.0758e-04, 1.2663e-04], device='cuda:1') 2023-04-28 01:50:35,789 INFO [finetune.py:976] (1/7) Epoch 27, batch 5300, loss[loss=0.1478, simple_loss=0.2332, pruned_loss=0.03125, over 4910.00 frames. ], tot_loss[loss=0.1683, simple_loss=0.2428, pruned_loss=0.04693, over 952983.50 frames. ], batch size: 42, lr: 2.90e-03, grad_scale: 32.0 2023-04-28 01:50:36,525 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=154221.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:51:15,621 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-04-28 01:51:16,103 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.1382, 1.9290, 1.5932, 1.5922, 1.8091, 1.6717, 2.0366, 1.3695], device='cuda:1'), covar=tensor([0.3131, 0.1241, 0.3997, 0.2101, 0.1435, 0.1664, 0.1755, 0.4459], device='cuda:1'), in_proj_covar=tensor([0.0339, 0.0354, 0.0426, 0.0351, 0.0380, 0.0376, 0.0369, 0.0423], device='cuda:1'), out_proj_covar=tensor([9.9951e-05, 1.0546e-04, 1.2872e-04, 1.0497e-04, 1.1249e-04, 1.1173e-04, 1.0801e-04, 1.2722e-04], device='cuda:1') 2023-04-28 01:51:41,568 INFO [finetune.py:976] (1/7) Epoch 27, batch 5350, loss[loss=0.2058, simple_loss=0.2738, pruned_loss=0.06885, over 4908.00 frames. ], tot_loss[loss=0.1692, simple_loss=0.2437, pruned_loss=0.04737, over 952530.28 frames. ], batch size: 38, lr: 2.90e-03, grad_scale: 32.0 2023-04-28 01:51:49,112 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.757e+01 1.510e+02 1.847e+02 2.221e+02 4.452e+02, threshold=3.694e+02, percent-clipped=2.0 2023-04-28 01:51:52,261 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=154279.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:51:54,097 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=154282.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 01:51:56,603 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=154286.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:52:00,107 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=154291.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:52:19,624 INFO [finetune.py:976] (1/7) Epoch 27, batch 5400, loss[loss=0.1323, simple_loss=0.2138, pruned_loss=0.02536, over 4826.00 frames. ], tot_loss[loss=0.1686, simple_loss=0.2421, pruned_loss=0.04758, over 955005.54 frames. ], batch size: 30, lr: 2.90e-03, grad_scale: 64.0 2023-04-28 01:52:30,057 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=154337.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 01:52:37,691 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=154347.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:52:41,208 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=154352.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:52:44,937 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8611, 1.0712, 1.8129, 2.2596, 1.9277, 1.8233, 1.8455, 1.8188], device='cuda:1'), covar=tensor([0.4365, 0.6327, 0.5640, 0.5630, 0.5311, 0.7377, 0.7101, 0.7470], device='cuda:1'), in_proj_covar=tensor([0.0441, 0.0423, 0.0517, 0.0505, 0.0471, 0.0506, 0.0507, 0.0523], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-28 01:52:52,200 INFO [finetune.py:976] (1/7) Epoch 27, batch 5450, loss[loss=0.1782, simple_loss=0.2461, pruned_loss=0.05519, over 4756.00 frames. ], tot_loss[loss=0.166, simple_loss=0.239, pruned_loss=0.04653, over 955627.95 frames. ], batch size: 54, lr: 2.90e-03, grad_scale: 64.0 2023-04-28 01:52:54,629 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.131e+02 1.557e+02 1.812e+02 2.063e+02 3.462e+02, threshold=3.624e+02, percent-clipped=0.0 2023-04-28 01:53:09,446 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=154398.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 01:53:16,764 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=154405.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 01:53:20,460 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.2403, 1.5699, 1.4034, 1.5563, 1.3346, 1.3291, 1.3727, 1.1180], device='cuda:1'), covar=tensor([0.1712, 0.1287, 0.0897, 0.1163, 0.3535, 0.1232, 0.1679, 0.1981], device='cuda:1'), in_proj_covar=tensor([0.0283, 0.0302, 0.0216, 0.0277, 0.0314, 0.0254, 0.0248, 0.0262], device='cuda:1'), out_proj_covar=tensor([1.1299e-04, 1.1871e-04, 8.4936e-05, 1.0889e-04, 1.2676e-04, 1.0001e-04, 9.9976e-05, 1.0340e-04], device='cuda:1') 2023-04-28 01:53:31,462 INFO [finetune.py:976] (1/7) Epoch 27, batch 5500, loss[loss=0.1473, simple_loss=0.207, pruned_loss=0.04375, over 4892.00 frames. ], tot_loss[loss=0.1633, simple_loss=0.2355, pruned_loss=0.04557, over 955667.17 frames. ], batch size: 32, lr: 2.90e-03, grad_scale: 64.0 2023-04-28 01:53:44,351 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=154432.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 01:53:52,963 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.11 vs. limit=2.0 2023-04-28 01:53:54,046 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=154440.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:54:35,187 INFO [finetune.py:976] (1/7) Epoch 27, batch 5550, loss[loss=0.1472, simple_loss=0.2219, pruned_loss=0.03628, over 4794.00 frames. ], tot_loss[loss=0.1651, simple_loss=0.2374, pruned_loss=0.04644, over 953662.45 frames. ], batch size: 29, lr: 2.90e-03, grad_scale: 64.0 2023-04-28 01:54:37,624 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.014e+02 1.508e+02 1.736e+02 2.106e+02 4.174e+02, threshold=3.472e+02, percent-clipped=1.0 2023-04-28 01:54:37,735 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=154474.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:55:07,207 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.0140, 1.0072, 1.1986, 1.1789, 1.0347, 0.8766, 0.9603, 0.3572], device='cuda:1'), covar=tensor([0.0526, 0.0445, 0.0439, 0.0443, 0.0641, 0.1402, 0.0414, 0.0745], device='cuda:1'), in_proj_covar=tensor([0.0070, 0.0067, 0.0065, 0.0069, 0.0075, 0.0094, 0.0072, 0.0062], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-28 01:55:31,962 INFO [finetune.py:976] (1/7) Epoch 27, batch 5600, loss[loss=0.1897, simple_loss=0.2674, pruned_loss=0.05598, over 4897.00 frames. ], tot_loss[loss=0.1675, simple_loss=0.2411, pruned_loss=0.04693, over 955198.79 frames. ], batch size: 43, lr: 2.90e-03, grad_scale: 64.0 2023-04-28 01:55:50,540 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=154535.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:56:00,968 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 2023-04-28 01:56:32,734 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5642, 1.6908, 0.8712, 1.2955, 1.5968, 1.4442, 1.3275, 1.4393], device='cuda:1'), covar=tensor([0.0465, 0.0328, 0.0330, 0.0520, 0.0279, 0.0461, 0.0458, 0.0522], device='cuda:1'), in_proj_covar=tensor([0.0028, 0.0023, 0.0022, 0.0028, 0.0019, 0.0028, 0.0027, 0.0028], device='cuda:1'), out_proj_covar=tensor([0.0052, 0.0046, 0.0039, 0.0053, 0.0039, 0.0051, 0.0051, 0.0053], device='cuda:1') 2023-04-28 01:56:35,594 INFO [finetune.py:976] (1/7) Epoch 27, batch 5650, loss[loss=0.1908, simple_loss=0.2584, pruned_loss=0.06159, over 4870.00 frames. ], tot_loss[loss=0.1686, simple_loss=0.2432, pruned_loss=0.04702, over 952721.67 frames. ], batch size: 34, lr: 2.90e-03, grad_scale: 64.0 2023-04-28 01:56:42,716 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.767e+01 1.509e+02 1.920e+02 2.295e+02 5.018e+02, threshold=3.840e+02, percent-clipped=4.0 2023-04-28 01:56:44,486 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=154577.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 01:56:45,660 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=154579.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:57:17,094 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.1361, 2.6022, 2.3029, 2.5602, 1.9891, 2.2882, 2.1964, 1.8284], device='cuda:1'), covar=tensor([0.1784, 0.1203, 0.0719, 0.1117, 0.3015, 0.1083, 0.1740, 0.2285], device='cuda:1'), in_proj_covar=tensor([0.0284, 0.0304, 0.0217, 0.0277, 0.0315, 0.0256, 0.0248, 0.0263], device='cuda:1'), out_proj_covar=tensor([1.1351e-04, 1.1955e-04, 8.5337e-05, 1.0896e-04, 1.2716e-04, 1.0041e-04, 1.0022e-04, 1.0381e-04], device='cuda:1') 2023-04-28 01:57:36,099 INFO [finetune.py:976] (1/7) Epoch 27, batch 5700, loss[loss=0.1554, simple_loss=0.2168, pruned_loss=0.04701, over 3963.00 frames. ], tot_loss[loss=0.1665, simple_loss=0.2395, pruned_loss=0.04672, over 932041.05 frames. ], batch size: 17, lr: 2.90e-03, grad_scale: 32.0 2023-04-28 01:57:36,179 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0695, 3.0969, 2.6538, 2.8978, 3.1962, 2.6640, 4.0237, 2.5249], device='cuda:1'), covar=tensor([0.3001, 0.1682, 0.3151, 0.2327, 0.1341, 0.2439, 0.1112, 0.3107], device='cuda:1'), in_proj_covar=tensor([0.0338, 0.0354, 0.0425, 0.0351, 0.0380, 0.0377, 0.0370, 0.0423], device='cuda:1'), out_proj_covar=tensor([9.9833e-05, 1.0529e-04, 1.2857e-04, 1.0501e-04, 1.1237e-04, 1.1208e-04, 1.0813e-04, 1.2723e-04], device='cuda:1') 2023-04-28 01:57:45,551 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=154627.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:57:46,207 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=154628.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:57:59,108 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=154642.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:58:12,012 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=154647.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:58:12,561 INFO [finetune.py:976] (1/7) Epoch 28, batch 0, loss[loss=0.1667, simple_loss=0.2339, pruned_loss=0.04979, over 4906.00 frames. ], tot_loss[loss=0.1667, simple_loss=0.2339, pruned_loss=0.04979, over 4906.00 frames. ], batch size: 36, lr: 2.90e-03, grad_scale: 32.0 2023-04-28 01:58:12,561 INFO [finetune.py:1001] (1/7) Computing validation loss 2023-04-28 01:58:23,873 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5021, 1.2348, 1.2579, 1.2347, 1.6006, 1.3913, 1.1879, 1.2721], device='cuda:1'), covar=tensor([0.1564, 0.1177, 0.1582, 0.1371, 0.0716, 0.1118, 0.1572, 0.1902], device='cuda:1'), in_proj_covar=tensor([0.0313, 0.0306, 0.0350, 0.0286, 0.0325, 0.0304, 0.0300, 0.0375], device='cuda:1'), out_proj_covar=tensor([6.3781e-05, 6.2941e-05, 7.3412e-05, 5.7246e-05, 6.6436e-05, 6.3401e-05, 6.2023e-05, 7.9490e-05], device='cuda:1') 2023-04-28 01:58:29,428 INFO [finetune.py:1010] (1/7) Epoch 28, validation: loss=0.1549, simple_loss=0.224, pruned_loss=0.04297, over 2265189.00 frames. 2023-04-28 01:58:29,429 INFO [finetune.py:1011] (1/7) Maximum memory allocated so far is 6391MB 2023-04-28 01:58:33,238 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.2210, 2.7034, 1.1986, 1.4208, 2.0249, 1.2306, 3.4728, 1.9548], device='cuda:1'), covar=tensor([0.0661, 0.0641, 0.0789, 0.1297, 0.0546, 0.1070, 0.0435, 0.0609], device='cuda:1'), in_proj_covar=tensor([0.0051, 0.0064, 0.0047, 0.0046, 0.0050, 0.0051, 0.0073, 0.0051], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:1') 2023-04-28 01:58:55,769 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.710e+01 1.461e+02 1.801e+02 2.323e+02 7.600e+02, threshold=3.601e+02, percent-clipped=3.0 2023-04-28 01:59:04,489 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=154689.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:59:06,854 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=154693.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 01:59:09,839 INFO [finetune.py:976] (1/7) Epoch 28, batch 50, loss[loss=0.1212, simple_loss=0.1929, pruned_loss=0.02474, over 4783.00 frames. ], tot_loss[loss=0.1694, simple_loss=0.2438, pruned_loss=0.04754, over 216869.09 frames. ], batch size: 26, lr: 2.90e-03, grad_scale: 32.0 2023-04-28 01:59:17,130 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=154705.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 01:59:20,228 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.8702, 2.2534, 2.1221, 2.2592, 2.0669, 2.1277, 2.1475, 2.1496], device='cuda:1'), covar=tensor([0.4037, 0.5716, 0.4615, 0.4194, 0.5586, 0.6424, 0.6230, 0.5823], device='cuda:1'), in_proj_covar=tensor([0.0341, 0.0374, 0.0329, 0.0341, 0.0350, 0.0394, 0.0360, 0.0333], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 01:59:26,295 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=154720.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:59:33,636 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=154732.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:59:38,567 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=154740.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:59:43,315 INFO [finetune.py:976] (1/7) Epoch 28, batch 100, loss[loss=0.1777, simple_loss=0.2479, pruned_loss=0.05371, over 4939.00 frames. ], tot_loss[loss=0.163, simple_loss=0.2361, pruned_loss=0.04495, over 381612.19 frames. ], batch size: 33, lr: 2.90e-03, grad_scale: 32.0 2023-04-28 01:59:48,357 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=154753.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 02:00:02,205 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.686e+01 1.538e+02 1.845e+02 2.159e+02 3.671e+02, threshold=3.690e+02, percent-clipped=1.0 2023-04-28 02:00:05,345 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=154780.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:00:06,011 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=154781.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:00:09,738 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-04-28 02:00:10,223 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=154788.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:00:16,273 INFO [finetune.py:976] (1/7) Epoch 28, batch 150, loss[loss=0.154, simple_loss=0.2289, pruned_loss=0.03959, over 4859.00 frames. ], tot_loss[loss=0.1606, simple_loss=0.2329, pruned_loss=0.04421, over 508547.80 frames. ], batch size: 31, lr: 2.90e-03, grad_scale: 32.0 2023-04-28 02:00:38,339 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=154830.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:00:49,253 INFO [finetune.py:976] (1/7) Epoch 28, batch 200, loss[loss=0.1414, simple_loss=0.2098, pruned_loss=0.03653, over 4821.00 frames. ], tot_loss[loss=0.1579, simple_loss=0.2297, pruned_loss=0.04303, over 608524.93 frames. ], batch size: 30, lr: 2.89e-03, grad_scale: 32.0 2023-04-28 02:01:08,233 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.084e+02 1.505e+02 1.793e+02 2.296e+02 3.844e+02, threshold=3.586e+02, percent-clipped=2.0 2023-04-28 02:01:09,538 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=154877.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 02:01:22,150 INFO [finetune.py:976] (1/7) Epoch 28, batch 250, loss[loss=0.1779, simple_loss=0.2592, pruned_loss=0.04828, over 4819.00 frames. ], tot_loss[loss=0.1602, simple_loss=0.2328, pruned_loss=0.04381, over 685890.93 frames. ], batch size: 33, lr: 2.89e-03, grad_scale: 32.0 2023-04-28 02:01:27,018 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.85 vs. limit=5.0 2023-04-28 02:01:28,636 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6272, 1.7381, 0.7456, 1.2593, 1.8344, 1.4665, 1.3424, 1.4177], device='cuda:1'), covar=tensor([0.0500, 0.0369, 0.0321, 0.0547, 0.0255, 0.0515, 0.0474, 0.0613], device='cuda:1'), in_proj_covar=tensor([0.0027, 0.0023, 0.0021, 0.0028, 0.0019, 0.0027, 0.0027, 0.0028], device='cuda:1'), out_proj_covar=tensor([0.0052, 0.0046, 0.0038, 0.0053, 0.0039, 0.0051, 0.0050, 0.0052], device='cuda:1') 2023-04-28 02:01:41,469 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=154925.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:01:50,100 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7327, 2.4347, 1.9619, 2.2332, 2.4225, 2.0074, 2.9450, 1.8476], device='cuda:1'), covar=tensor([0.3238, 0.1938, 0.3837, 0.3048, 0.1741, 0.2532, 0.1936, 0.3996], device='cuda:1'), in_proj_covar=tensor([0.0340, 0.0356, 0.0424, 0.0352, 0.0381, 0.0377, 0.0370, 0.0424], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-28 02:01:51,915 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=154942.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:01:54,976 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=154947.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:02:00,647 INFO [finetune.py:976] (1/7) Epoch 28, batch 300, loss[loss=0.2011, simple_loss=0.2804, pruned_loss=0.06084, over 4265.00 frames. ], tot_loss[loss=0.1651, simple_loss=0.2391, pruned_loss=0.04557, over 745556.90 frames. ], batch size: 65, lr: 2.89e-03, grad_scale: 32.0 2023-04-28 02:02:11,953 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.11 vs. limit=2.0 2023-04-28 02:02:36,609 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.011e+02 1.547e+02 1.784e+02 2.160e+02 3.859e+02, threshold=3.568e+02, percent-clipped=1.0 2023-04-28 02:02:37,344 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=154976.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:02:43,879 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.2962, 2.1077, 1.7619, 1.8732, 2.1991, 1.7199, 2.6091, 1.5480], device='cuda:1'), covar=tensor([0.3667, 0.1889, 0.4685, 0.2831, 0.1748, 0.2460, 0.1957, 0.4617], device='cuda:1'), in_proj_covar=tensor([0.0340, 0.0356, 0.0425, 0.0351, 0.0381, 0.0377, 0.0371, 0.0424], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-28 02:02:47,486 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=154984.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:02:56,111 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=154990.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:02:57,990 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=154993.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 02:02:59,162 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=154995.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:03:06,551 INFO [finetune.py:976] (1/7) Epoch 28, batch 350, loss[loss=0.1608, simple_loss=0.2359, pruned_loss=0.04284, over 4846.00 frames. ], tot_loss[loss=0.1658, simple_loss=0.2406, pruned_loss=0.04547, over 793983.86 frames. ], batch size: 44, lr: 2.89e-03, grad_scale: 32.0 2023-04-28 02:03:45,297 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=155037.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:03:47,614 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=155041.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 02:03:51,698 INFO [finetune.py:976] (1/7) Epoch 28, batch 400, loss[loss=0.2327, simple_loss=0.3006, pruned_loss=0.08245, over 4831.00 frames. ], tot_loss[loss=0.1668, simple_loss=0.2418, pruned_loss=0.04588, over 830059.05 frames. ], batch size: 47, lr: 2.89e-03, grad_scale: 32.0 2023-04-28 02:04:03,384 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.62 vs. limit=2.0 2023-04-28 02:04:06,194 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=155069.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:04:11,280 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.089e+02 1.524e+02 1.961e+02 2.438e+02 3.962e+02, threshold=3.922e+02, percent-clipped=3.0 2023-04-28 02:04:11,990 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=155076.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:04:25,381 INFO [finetune.py:976] (1/7) Epoch 28, batch 450, loss[loss=0.1728, simple_loss=0.24, pruned_loss=0.05279, over 4785.00 frames. ], tot_loss[loss=0.1653, simple_loss=0.2401, pruned_loss=0.04531, over 858103.71 frames. ], batch size: 25, lr: 2.89e-03, grad_scale: 32.0 2023-04-28 02:04:48,042 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=155130.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:04:48,069 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=155130.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:04:58,929 INFO [finetune.py:976] (1/7) Epoch 28, batch 500, loss[loss=0.164, simple_loss=0.2357, pruned_loss=0.04617, over 4800.00 frames. ], tot_loss[loss=0.1626, simple_loss=0.2369, pruned_loss=0.04418, over 879216.54 frames. ], batch size: 51, lr: 2.89e-03, grad_scale: 32.0 2023-04-28 02:05:17,835 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.043e+02 1.583e+02 1.761e+02 2.203e+02 6.801e+02, threshold=3.523e+02, percent-clipped=1.0 2023-04-28 02:05:20,240 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=155178.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:05:32,275 INFO [finetune.py:976] (1/7) Epoch 28, batch 550, loss[loss=0.1318, simple_loss=0.2044, pruned_loss=0.02963, over 4908.00 frames. ], tot_loss[loss=0.1602, simple_loss=0.2337, pruned_loss=0.04332, over 896487.82 frames. ], batch size: 36, lr: 2.89e-03, grad_scale: 32.0 2023-04-28 02:06:05,934 INFO [finetune.py:976] (1/7) Epoch 28, batch 600, loss[loss=0.1981, simple_loss=0.2698, pruned_loss=0.06324, over 4912.00 frames. ], tot_loss[loss=0.1626, simple_loss=0.2359, pruned_loss=0.04463, over 910522.11 frames. ], batch size: 37, lr: 2.89e-03, grad_scale: 32.0 2023-04-28 02:06:23,749 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.531e+01 1.578e+02 2.066e+02 2.357e+02 4.358e+02, threshold=4.132e+02, percent-clipped=3.0 2023-04-28 02:06:28,484 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4396, 2.4475, 2.0990, 2.1616, 2.3918, 1.7904, 3.0040, 1.6977], device='cuda:1'), covar=tensor([0.3735, 0.1517, 0.3608, 0.2830, 0.1866, 0.3159, 0.1212, 0.4623], device='cuda:1'), in_proj_covar=tensor([0.0340, 0.0357, 0.0424, 0.0351, 0.0381, 0.0377, 0.0372, 0.0422], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-28 02:06:30,745 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=155284.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:06:31,373 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9226, 2.5602, 1.9281, 1.8077, 1.4096, 1.4080, 1.9077, 1.3311], device='cuda:1'), covar=tensor([0.1725, 0.1321, 0.1418, 0.1754, 0.2361, 0.1933, 0.1052, 0.2094], device='cuda:1'), in_proj_covar=tensor([0.0199, 0.0211, 0.0171, 0.0206, 0.0202, 0.0187, 0.0157, 0.0189], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-28 02:06:39,097 INFO [finetune.py:976] (1/7) Epoch 28, batch 650, loss[loss=0.2126, simple_loss=0.2848, pruned_loss=0.07015, over 4906.00 frames. ], tot_loss[loss=0.1647, simple_loss=0.2384, pruned_loss=0.04545, over 920603.78 frames. ], batch size: 35, lr: 2.89e-03, grad_scale: 32.0 2023-04-28 02:06:48,957 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6968, 2.3955, 1.6068, 1.6936, 1.3027, 1.2727, 1.6450, 1.1719], device='cuda:1'), covar=tensor([0.1629, 0.1097, 0.1456, 0.1610, 0.2270, 0.1977, 0.1011, 0.2063], device='cuda:1'), in_proj_covar=tensor([0.0200, 0.0211, 0.0172, 0.0207, 0.0202, 0.0188, 0.0157, 0.0190], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-28 02:07:02,743 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=155332.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:07:02,752 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=155332.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:07:12,883 INFO [finetune.py:976] (1/7) Epoch 28, batch 700, loss[loss=0.1991, simple_loss=0.2647, pruned_loss=0.06677, over 4832.00 frames. ], tot_loss[loss=0.1671, simple_loss=0.2411, pruned_loss=0.04651, over 926606.43 frames. ], batch size: 47, lr: 2.89e-03, grad_scale: 32.0 2023-04-28 02:07:34,633 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.108e+02 1.574e+02 1.852e+02 2.166e+02 4.284e+02, threshold=3.703e+02, percent-clipped=1.0 2023-04-28 02:07:40,175 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=155376.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:08:09,445 INFO [finetune.py:976] (1/7) Epoch 28, batch 750, loss[loss=0.1944, simple_loss=0.2864, pruned_loss=0.0512, over 4926.00 frames. ], tot_loss[loss=0.1687, simple_loss=0.2428, pruned_loss=0.04733, over 933894.26 frames. ], batch size: 41, lr: 2.89e-03, grad_scale: 32.0 2023-04-28 02:08:33,695 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5244, 3.0791, 0.8520, 1.5943, 1.7779, 2.2139, 1.7319, 1.0057], device='cuda:1'), covar=tensor([0.1387, 0.1068, 0.1964, 0.1408, 0.1122, 0.1027, 0.1643, 0.1935], device='cuda:1'), in_proj_covar=tensor([0.0118, 0.0240, 0.0135, 0.0121, 0.0132, 0.0153, 0.0117, 0.0118], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-28 02:08:41,162 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=155424.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:08:41,830 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=155425.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:09:12,224 INFO [finetune.py:976] (1/7) Epoch 28, batch 800, loss[loss=0.1399, simple_loss=0.2123, pruned_loss=0.03374, over 4730.00 frames. ], tot_loss[loss=0.1667, simple_loss=0.2414, pruned_loss=0.04597, over 940452.95 frames. ], batch size: 54, lr: 2.89e-03, grad_scale: 32.0 2023-04-28 02:09:22,983 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=155457.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:09:42,704 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 8.932e+01 1.490e+02 1.749e+02 2.216e+02 3.602e+02, threshold=3.499e+02, percent-clipped=0.0 2023-04-28 02:09:59,775 INFO [finetune.py:976] (1/7) Epoch 28, batch 850, loss[loss=0.1711, simple_loss=0.2328, pruned_loss=0.05472, over 4825.00 frames. ], tot_loss[loss=0.1652, simple_loss=0.2397, pruned_loss=0.04535, over 943547.23 frames. ], batch size: 40, lr: 2.89e-03, grad_scale: 32.0 2023-04-28 02:10:11,981 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=155518.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:10:26,708 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.51 vs. limit=5.0 2023-04-28 02:10:32,818 INFO [finetune.py:976] (1/7) Epoch 28, batch 900, loss[loss=0.1391, simple_loss=0.206, pruned_loss=0.03613, over 4837.00 frames. ], tot_loss[loss=0.163, simple_loss=0.2367, pruned_loss=0.04462, over 947759.74 frames. ], batch size: 30, lr: 2.89e-03, grad_scale: 32.0 2023-04-28 02:10:49,542 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 8.500e+01 1.455e+02 1.735e+02 2.204e+02 5.182e+02, threshold=3.469e+02, percent-clipped=3.0 2023-04-28 02:10:50,286 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.3050, 1.6457, 1.5029, 1.9418, 1.8834, 2.0382, 1.6154, 4.1937], device='cuda:1'), covar=tensor([0.0569, 0.0804, 0.0780, 0.1201, 0.0614, 0.0586, 0.0727, 0.0114], device='cuda:1'), in_proj_covar=tensor([0.0037, 0.0038, 0.0039, 0.0042, 0.0040, 0.0037, 0.0038, 0.0054], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0013, 0.0013, 0.0015, 0.0014, 0.0013, 0.0013, 0.0016], device='cuda:1') 2023-04-28 02:10:52,717 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4128, 1.3817, 1.7030, 1.6745, 1.2679, 1.2116, 1.4465, 1.0021], device='cuda:1'), covar=tensor([0.0526, 0.0567, 0.0391, 0.0618, 0.0751, 0.0927, 0.0570, 0.0524], device='cuda:1'), in_proj_covar=tensor([0.0071, 0.0068, 0.0066, 0.0069, 0.0075, 0.0095, 0.0073, 0.0062], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-28 02:11:03,902 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-04-28 02:11:05,996 INFO [finetune.py:976] (1/7) Epoch 28, batch 950, loss[loss=0.164, simple_loss=0.2274, pruned_loss=0.05029, over 4923.00 frames. ], tot_loss[loss=0.1629, simple_loss=0.2357, pruned_loss=0.04505, over 949987.65 frames. ], batch size: 43, lr: 2.89e-03, grad_scale: 32.0 2023-04-28 02:11:10,936 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=155605.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:11:23,107 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5233, 1.5201, 1.8418, 1.8982, 1.3644, 1.2902, 1.5358, 0.9444], device='cuda:1'), covar=tensor([0.0547, 0.0706, 0.0403, 0.0658, 0.0724, 0.1086, 0.0638, 0.0637], device='cuda:1'), in_proj_covar=tensor([0.0070, 0.0068, 0.0066, 0.0069, 0.0075, 0.0095, 0.0073, 0.0062], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-28 02:11:27,349 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=155632.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:11:39,449 INFO [finetune.py:976] (1/7) Epoch 28, batch 1000, loss[loss=0.2805, simple_loss=0.33, pruned_loss=0.1155, over 4222.00 frames. ], tot_loss[loss=0.1652, simple_loss=0.2383, pruned_loss=0.04603, over 951740.66 frames. ], batch size: 65, lr: 2.89e-03, grad_scale: 32.0 2023-04-28 02:11:45,540 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=155657.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:11:51,051 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=155666.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:11:56,469 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 8.942e+01 1.547e+02 1.789e+02 2.038e+02 3.472e+02, threshold=3.578e+02, percent-clipped=1.0 2023-04-28 02:11:59,549 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=155680.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:12:12,444 INFO [finetune.py:976] (1/7) Epoch 28, batch 1050, loss[loss=0.1685, simple_loss=0.2323, pruned_loss=0.05231, over 4719.00 frames. ], tot_loss[loss=0.167, simple_loss=0.2409, pruned_loss=0.04655, over 952021.75 frames. ], batch size: 23, lr: 2.89e-03, grad_scale: 32.0 2023-04-28 02:12:25,711 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=155718.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:12:29,953 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=155725.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:12:45,200 INFO [finetune.py:976] (1/7) Epoch 28, batch 1100, loss[loss=0.1601, simple_loss=0.2402, pruned_loss=0.03999, over 4865.00 frames. ], tot_loss[loss=0.1683, simple_loss=0.2422, pruned_loss=0.04719, over 952466.71 frames. ], batch size: 35, lr: 2.89e-03, grad_scale: 32.0 2023-04-28 02:13:13,273 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=155773.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:13:19,855 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.115e+02 1.609e+02 1.883e+02 2.305e+02 3.828e+02, threshold=3.767e+02, percent-clipped=2.0 2023-04-28 02:13:30,960 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=155784.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:13:33,480 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-04-28 02:13:42,052 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.1468, 1.3296, 5.5627, 5.2353, 4.8824, 5.3748, 4.8269, 5.0401], device='cuda:1'), covar=tensor([0.6138, 0.6078, 0.0789, 0.1372, 0.0889, 0.1414, 0.1198, 0.1059], device='cuda:1'), in_proj_covar=tensor([0.0314, 0.0312, 0.0411, 0.0412, 0.0352, 0.0421, 0.0321, 0.0368], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-28 02:13:46,059 INFO [finetune.py:976] (1/7) Epoch 28, batch 1150, loss[loss=0.1424, simple_loss=0.2269, pruned_loss=0.0289, over 4838.00 frames. ], tot_loss[loss=0.1694, simple_loss=0.2437, pruned_loss=0.04756, over 955288.32 frames. ], batch size: 30, lr: 2.89e-03, grad_scale: 32.0 2023-04-28 02:13:58,496 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=155813.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:14:18,104 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=155845.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 02:14:19,787 INFO [finetune.py:976] (1/7) Epoch 28, batch 1200, loss[loss=0.1475, simple_loss=0.2317, pruned_loss=0.0316, over 4747.00 frames. ], tot_loss[loss=0.168, simple_loss=0.2418, pruned_loss=0.04708, over 955901.95 frames. ], batch size: 27, lr: 2.89e-03, grad_scale: 32.0 2023-04-28 02:14:22,269 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-04-28 02:14:54,753 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.082e+02 1.563e+02 1.837e+02 2.158e+02 4.161e+02, threshold=3.673e+02, percent-clipped=1.0 2023-04-28 02:15:25,826 INFO [finetune.py:976] (1/7) Epoch 28, batch 1250, loss[loss=0.1383, simple_loss=0.2286, pruned_loss=0.02401, over 4826.00 frames. ], tot_loss[loss=0.1663, simple_loss=0.2396, pruned_loss=0.04648, over 956880.57 frames. ], batch size: 33, lr: 2.89e-03, grad_scale: 32.0 2023-04-28 02:15:50,205 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-04-28 02:16:33,390 INFO [finetune.py:976] (1/7) Epoch 28, batch 1300, loss[loss=0.1451, simple_loss=0.2097, pruned_loss=0.04021, over 4903.00 frames. ], tot_loss[loss=0.1637, simple_loss=0.2366, pruned_loss=0.04536, over 958091.33 frames. ], batch size: 43, lr: 2.89e-03, grad_scale: 32.0 2023-04-28 02:16:53,061 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=155961.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:16:54,395 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-04-28 02:17:07,930 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.397e+01 1.571e+02 1.820e+02 2.327e+02 4.182e+02, threshold=3.640e+02, percent-clipped=2.0 2023-04-28 02:17:37,319 INFO [finetune.py:976] (1/7) Epoch 28, batch 1350, loss[loss=0.2004, simple_loss=0.2652, pruned_loss=0.06785, over 4825.00 frames. ], tot_loss[loss=0.1648, simple_loss=0.2375, pruned_loss=0.04601, over 957384.81 frames. ], batch size: 33, lr: 2.89e-03, grad_scale: 32.0 2023-04-28 02:17:56,768 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=156013.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:18:40,867 INFO [finetune.py:976] (1/7) Epoch 28, batch 1400, loss[loss=0.1751, simple_loss=0.2511, pruned_loss=0.04952, over 4828.00 frames. ], tot_loss[loss=0.1677, simple_loss=0.241, pruned_loss=0.04724, over 956224.62 frames. ], batch size: 38, lr: 2.89e-03, grad_scale: 32.0 2023-04-28 02:18:59,795 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9642, 2.4894, 2.1711, 2.4144, 1.6723, 2.1639, 2.1001, 1.6203], device='cuda:1'), covar=tensor([0.1805, 0.0894, 0.0704, 0.0968, 0.3061, 0.0882, 0.1654, 0.2266], device='cuda:1'), in_proj_covar=tensor([0.0283, 0.0302, 0.0217, 0.0276, 0.0315, 0.0254, 0.0248, 0.0263], device='cuda:1'), out_proj_covar=tensor([1.1273e-04, 1.1878e-04, 8.5281e-05, 1.0862e-04, 1.2699e-04, 9.9920e-05, 1.0001e-04, 1.0377e-04], device='cuda:1') 2023-04-28 02:19:12,179 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=156070.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:19:20,863 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.099e+02 1.674e+02 1.966e+02 2.354e+02 4.293e+02, threshold=3.933e+02, percent-clipped=2.0 2023-04-28 02:19:22,885 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=3.85 vs. limit=5.0 2023-04-28 02:19:36,469 INFO [finetune.py:976] (1/7) Epoch 28, batch 1450, loss[loss=0.1529, simple_loss=0.2332, pruned_loss=0.0363, over 4106.00 frames. ], tot_loss[loss=0.1692, simple_loss=0.2432, pruned_loss=0.04759, over 955698.06 frames. ], batch size: 65, lr: 2.89e-03, grad_scale: 32.0 2023-04-28 02:19:46,155 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=156113.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:19:52,496 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.0763, 1.3833, 1.2307, 1.6109, 1.5283, 1.4722, 1.3224, 2.4484], device='cuda:1'), covar=tensor([0.0652, 0.0844, 0.0850, 0.1252, 0.0648, 0.0496, 0.0756, 0.0215], device='cuda:1'), in_proj_covar=tensor([0.0037, 0.0039, 0.0039, 0.0042, 0.0040, 0.0037, 0.0038, 0.0054], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0013, 0.0013, 0.0015, 0.0014, 0.0013, 0.0013, 0.0016], device='cuda:1') 2023-04-28 02:19:59,759 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=156131.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 02:20:05,152 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=156140.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 02:20:07,690 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-04-28 02:20:09,933 INFO [finetune.py:976] (1/7) Epoch 28, batch 1500, loss[loss=0.1657, simple_loss=0.2464, pruned_loss=0.04251, over 4871.00 frames. ], tot_loss[loss=0.17, simple_loss=0.2444, pruned_loss=0.04784, over 956535.81 frames. ], batch size: 35, lr: 2.89e-03, grad_scale: 32.0 2023-04-28 02:20:11,249 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.2357, 1.5162, 1.8202, 1.9076, 1.8479, 1.9601, 1.8568, 1.8778], device='cuda:1'), covar=tensor([0.3384, 0.4392, 0.4010, 0.3994, 0.4492, 0.5721, 0.4073, 0.3870], device='cuda:1'), in_proj_covar=tensor([0.0343, 0.0377, 0.0332, 0.0343, 0.0351, 0.0394, 0.0362, 0.0336], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 02:20:23,242 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=156161.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:20:43,694 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.067e+02 1.591e+02 1.859e+02 2.303e+02 4.450e+02, threshold=3.717e+02, percent-clipped=1.0 2023-04-28 02:21:08,784 INFO [finetune.py:976] (1/7) Epoch 28, batch 1550, loss[loss=0.1498, simple_loss=0.2178, pruned_loss=0.04091, over 4755.00 frames. ], tot_loss[loss=0.1685, simple_loss=0.2426, pruned_loss=0.04721, over 954339.55 frames. ], batch size: 23, lr: 2.89e-03, grad_scale: 32.0 2023-04-28 02:21:31,123 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=156216.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:22:04,805 INFO [finetune.py:976] (1/7) Epoch 28, batch 1600, loss[loss=0.1585, simple_loss=0.2206, pruned_loss=0.04816, over 4786.00 frames. ], tot_loss[loss=0.1659, simple_loss=0.2393, pruned_loss=0.04628, over 954746.27 frames. ], batch size: 26, lr: 2.89e-03, grad_scale: 32.0 2023-04-28 02:22:12,232 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6362, 1.4648, 1.6480, 1.9677, 2.0634, 1.5902, 1.3389, 1.7896], device='cuda:1'), covar=tensor([0.0787, 0.1201, 0.0776, 0.0553, 0.0544, 0.0771, 0.0727, 0.0555], device='cuda:1'), in_proj_covar=tensor([0.0185, 0.0202, 0.0184, 0.0170, 0.0178, 0.0178, 0.0151, 0.0178], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-28 02:22:12,840 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=156261.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:22:24,231 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.128e+02 1.624e+02 1.857e+02 2.215e+02 3.679e+02, threshold=3.713e+02, percent-clipped=0.0 2023-04-28 02:22:25,629 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=156277.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:22:31,113 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4217, 1.3152, 1.6527, 1.6452, 1.3136, 1.2634, 1.3705, 0.8574], device='cuda:1'), covar=tensor([0.0543, 0.0676, 0.0400, 0.0586, 0.0807, 0.1126, 0.0500, 0.0545], device='cuda:1'), in_proj_covar=tensor([0.0071, 0.0068, 0.0066, 0.0070, 0.0076, 0.0095, 0.0073, 0.0063], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-28 02:22:34,018 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=156290.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:22:38,824 INFO [finetune.py:976] (1/7) Epoch 28, batch 1650, loss[loss=0.1511, simple_loss=0.227, pruned_loss=0.03758, over 4828.00 frames. ], tot_loss[loss=0.1629, simple_loss=0.236, pruned_loss=0.04496, over 955178.64 frames. ], batch size: 39, lr: 2.89e-03, grad_scale: 32.0 2023-04-28 02:22:45,574 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=156309.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:22:48,032 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=156313.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:23:12,276 INFO [finetune.py:976] (1/7) Epoch 28, batch 1700, loss[loss=0.1741, simple_loss=0.2486, pruned_loss=0.04981, over 4859.00 frames. ], tot_loss[loss=0.1621, simple_loss=0.2345, pruned_loss=0.04479, over 956086.74 frames. ], batch size: 44, lr: 2.88e-03, grad_scale: 32.0 2023-04-28 02:23:14,227 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.2736, 1.5277, 1.4057, 1.5472, 1.3277, 1.3168, 1.4071, 1.0995], device='cuda:1'), covar=tensor([0.1777, 0.1274, 0.0836, 0.1173, 0.3679, 0.1200, 0.1789, 0.2207], device='cuda:1'), in_proj_covar=tensor([0.0282, 0.0303, 0.0217, 0.0276, 0.0314, 0.0254, 0.0247, 0.0263], device='cuda:1'), out_proj_covar=tensor([1.1232e-04, 1.1902e-04, 8.5175e-05, 1.0845e-04, 1.2667e-04, 9.9818e-05, 9.9613e-05, 1.0359e-04], device='cuda:1') 2023-04-28 02:23:14,243 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=156351.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:23:16,704 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.4625, 1.6745, 1.9808, 2.0744, 1.9457, 1.9851, 2.0248, 2.0187], device='cuda:1'), covar=tensor([0.3612, 0.5383, 0.4171, 0.4448, 0.5171, 0.6788, 0.4905, 0.4757], device='cuda:1'), in_proj_covar=tensor([0.0341, 0.0376, 0.0331, 0.0341, 0.0350, 0.0392, 0.0361, 0.0335], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 02:23:19,066 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4830, 1.4048, 4.2054, 3.9982, 3.6988, 3.9989, 3.9725, 3.7258], device='cuda:1'), covar=tensor([0.7294, 0.5541, 0.1107, 0.1650, 0.1132, 0.1774, 0.1216, 0.1648], device='cuda:1'), in_proj_covar=tensor([0.0309, 0.0307, 0.0406, 0.0406, 0.0347, 0.0415, 0.0317, 0.0363], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-28 02:23:19,114 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6527, 1.3828, 1.8220, 1.8991, 1.4987, 1.3839, 1.5085, 0.9212], device='cuda:1'), covar=tensor([0.0436, 0.0777, 0.0365, 0.0518, 0.0621, 0.1053, 0.0493, 0.0537], device='cuda:1'), in_proj_covar=tensor([0.0070, 0.0068, 0.0066, 0.0069, 0.0075, 0.0095, 0.0073, 0.0062], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-28 02:23:20,259 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=156361.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:23:28,664 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.471e+01 1.503e+02 1.857e+02 2.212e+02 3.931e+02, threshold=3.714e+02, percent-clipped=1.0 2023-04-28 02:23:33,408 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-04-28 02:23:45,145 INFO [finetune.py:976] (1/7) Epoch 28, batch 1750, loss[loss=0.1765, simple_loss=0.2526, pruned_loss=0.05022, over 4888.00 frames. ], tot_loss[loss=0.165, simple_loss=0.2379, pruned_loss=0.04601, over 955938.58 frames. ], batch size: 35, lr: 2.88e-03, grad_scale: 32.0 2023-04-28 02:23:52,713 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7188, 2.1153, 1.6527, 1.5367, 1.2637, 1.2829, 1.6906, 1.2125], device='cuda:1'), covar=tensor([0.1483, 0.1254, 0.1387, 0.1554, 0.2223, 0.1813, 0.0997, 0.2008], device='cuda:1'), in_proj_covar=tensor([0.0198, 0.0209, 0.0170, 0.0204, 0.0200, 0.0186, 0.0156, 0.0188], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-28 02:24:08,543 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=156426.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 02:24:11,476 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7064, 2.1596, 1.6456, 1.5546, 1.2279, 1.2783, 1.7247, 1.2355], device='cuda:1'), covar=tensor([0.1621, 0.1217, 0.1432, 0.1607, 0.2230, 0.1879, 0.0956, 0.2058], device='cuda:1'), in_proj_covar=tensor([0.0199, 0.0209, 0.0170, 0.0205, 0.0201, 0.0186, 0.0156, 0.0188], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-28 02:24:29,079 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=156440.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 02:24:39,114 INFO [finetune.py:976] (1/7) Epoch 28, batch 1800, loss[loss=0.2221, simple_loss=0.2835, pruned_loss=0.08037, over 4799.00 frames. ], tot_loss[loss=0.1666, simple_loss=0.2402, pruned_loss=0.04646, over 955335.46 frames. ], batch size: 51, lr: 2.88e-03, grad_scale: 32.0 2023-04-28 02:24:54,212 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.3241, 1.5386, 1.4551, 1.6820, 1.5937, 1.9452, 1.4294, 3.6185], device='cuda:1'), covar=tensor([0.0631, 0.0871, 0.0835, 0.1288, 0.0689, 0.0502, 0.0802, 0.0115], device='cuda:1'), in_proj_covar=tensor([0.0038, 0.0039, 0.0039, 0.0043, 0.0040, 0.0038, 0.0038, 0.0054], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0013, 0.0013, 0.0015, 0.0014, 0.0013, 0.0013, 0.0016], device='cuda:1') 2023-04-28 02:25:11,114 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.353e+01 1.610e+02 1.909e+02 2.260e+02 4.171e+02, threshold=3.817e+02, percent-clipped=3.0 2023-04-28 02:25:31,033 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=156488.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:25:42,451 INFO [finetune.py:976] (1/7) Epoch 28, batch 1850, loss[loss=0.1439, simple_loss=0.2357, pruned_loss=0.02607, over 4811.00 frames. ], tot_loss[loss=0.1672, simple_loss=0.2411, pruned_loss=0.04665, over 954497.85 frames. ], batch size: 40, lr: 2.88e-03, grad_scale: 32.0 2023-04-28 02:26:48,594 INFO [finetune.py:976] (1/7) Epoch 28, batch 1900, loss[loss=0.1759, simple_loss=0.247, pruned_loss=0.05238, over 4893.00 frames. ], tot_loss[loss=0.1688, simple_loss=0.2428, pruned_loss=0.0474, over 954801.89 frames. ], batch size: 36, lr: 2.88e-03, grad_scale: 32.0 2023-04-28 02:27:20,586 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=156572.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:27:22,347 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.009e+02 1.575e+02 1.976e+02 2.372e+02 4.648e+02, threshold=3.953e+02, percent-clipped=2.0 2023-04-28 02:27:45,061 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.16 vs. limit=5.0 2023-04-28 02:27:45,603 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7676, 2.1279, 1.2299, 1.4749, 2.3329, 1.6630, 1.5632, 1.6360], device='cuda:1'), covar=tensor([0.0469, 0.0322, 0.0259, 0.0527, 0.0223, 0.0462, 0.0441, 0.0527], device='cuda:1'), in_proj_covar=tensor([0.0028, 0.0023, 0.0022, 0.0028, 0.0019, 0.0027, 0.0027, 0.0028], device='cuda:1'), out_proj_covar=tensor([0.0052, 0.0046, 0.0039, 0.0053, 0.0039, 0.0051, 0.0051, 0.0053], device='cuda:1') 2023-04-28 02:27:53,133 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.4523, 1.7846, 1.8990, 1.9921, 1.8798, 1.9279, 1.9408, 1.9644], device='cuda:1'), covar=tensor([0.3663, 0.4886, 0.4241, 0.4143, 0.5144, 0.6737, 0.4870, 0.4488], device='cuda:1'), in_proj_covar=tensor([0.0341, 0.0374, 0.0330, 0.0340, 0.0350, 0.0392, 0.0360, 0.0334], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 02:27:54,790 INFO [finetune.py:976] (1/7) Epoch 28, batch 1950, loss[loss=0.1548, simple_loss=0.2217, pruned_loss=0.04393, over 4754.00 frames. ], tot_loss[loss=0.1677, simple_loss=0.2413, pruned_loss=0.04703, over 953357.01 frames. ], batch size: 27, lr: 2.88e-03, grad_scale: 32.0 2023-04-28 02:28:06,413 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.2819, 1.2752, 1.3711, 1.5571, 1.5364, 1.2693, 0.9887, 1.4901], device='cuda:1'), covar=tensor([0.0875, 0.1373, 0.0976, 0.0607, 0.0743, 0.0889, 0.0827, 0.0590], device='cuda:1'), in_proj_covar=tensor([0.0186, 0.0204, 0.0186, 0.0171, 0.0179, 0.0178, 0.0152, 0.0179], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-28 02:28:47,851 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6531, 1.7198, 1.5216, 1.1638, 1.2324, 1.2469, 1.5019, 1.1792], device='cuda:1'), covar=tensor([0.1538, 0.1216, 0.1403, 0.1589, 0.2152, 0.1854, 0.0932, 0.1950], device='cuda:1'), in_proj_covar=tensor([0.0198, 0.0208, 0.0169, 0.0204, 0.0200, 0.0186, 0.0155, 0.0187], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-28 02:29:00,505 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=156646.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:29:01,645 INFO [finetune.py:976] (1/7) Epoch 28, batch 2000, loss[loss=0.1647, simple_loss=0.23, pruned_loss=0.04972, over 4902.00 frames. ], tot_loss[loss=0.1668, simple_loss=0.2397, pruned_loss=0.04696, over 955257.47 frames. ], batch size: 36, lr: 2.88e-03, grad_scale: 64.0 2023-04-28 02:29:34,745 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.047e+02 1.509e+02 1.833e+02 2.149e+02 4.980e+02, threshold=3.665e+02, percent-clipped=1.0 2023-04-28 02:29:41,953 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5713, 1.3884, 4.3056, 4.0725, 3.8062, 4.2005, 4.1044, 3.8408], device='cuda:1'), covar=tensor([0.7144, 0.6012, 0.1108, 0.1702, 0.1227, 0.1696, 0.1345, 0.1478], device='cuda:1'), in_proj_covar=tensor([0.0309, 0.0307, 0.0405, 0.0405, 0.0346, 0.0414, 0.0317, 0.0363], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-28 02:30:06,050 INFO [finetune.py:976] (1/7) Epoch 28, batch 2050, loss[loss=0.1223, simple_loss=0.1954, pruned_loss=0.02462, over 4781.00 frames. ], tot_loss[loss=0.1624, simple_loss=0.235, pruned_loss=0.04492, over 956126.71 frames. ], batch size: 28, lr: 2.88e-03, grad_scale: 64.0 2023-04-28 02:30:35,517 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5821, 0.7624, 1.5472, 1.9687, 1.6788, 1.5294, 1.5388, 1.5479], device='cuda:1'), covar=tensor([0.3832, 0.6009, 0.5012, 0.4922, 0.5115, 0.6360, 0.6070, 0.6754], device='cuda:1'), in_proj_covar=tensor([0.0445, 0.0424, 0.0520, 0.0508, 0.0473, 0.0510, 0.0512, 0.0527], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-28 02:30:43,823 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=156726.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 02:31:05,850 INFO [finetune.py:976] (1/7) Epoch 28, batch 2100, loss[loss=0.1243, simple_loss=0.194, pruned_loss=0.02726, over 4761.00 frames. ], tot_loss[loss=0.1629, simple_loss=0.2354, pruned_loss=0.04519, over 957255.79 frames. ], batch size: 27, lr: 2.88e-03, grad_scale: 64.0 2023-04-28 02:31:07,079 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=156749.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:31:18,120 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.66 vs. limit=5.0 2023-04-28 02:31:39,407 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=156774.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:31:39,961 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.904e+01 1.566e+02 1.862e+02 2.081e+02 4.387e+02, threshold=3.725e+02, percent-clipped=1.0 2023-04-28 02:32:10,049 INFO [finetune.py:976] (1/7) Epoch 28, batch 2150, loss[loss=0.1734, simple_loss=0.2435, pruned_loss=0.05168, over 4854.00 frames. ], tot_loss[loss=0.1656, simple_loss=0.239, pruned_loss=0.04612, over 956512.77 frames. ], batch size: 31, lr: 2.88e-03, grad_scale: 32.0 2023-04-28 02:32:22,733 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=156810.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:32:46,415 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-04-28 02:32:48,109 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7834, 2.1001, 1.0697, 1.4801, 2.2470, 1.6109, 1.4930, 1.6227], device='cuda:1'), covar=tensor([0.0443, 0.0335, 0.0286, 0.0517, 0.0246, 0.0488, 0.0474, 0.0526], device='cuda:1'), in_proj_covar=tensor([0.0028, 0.0023, 0.0022, 0.0028, 0.0019, 0.0028, 0.0027, 0.0028], device='cuda:1'), out_proj_covar=tensor([0.0052, 0.0047, 0.0039, 0.0054, 0.0039, 0.0051, 0.0051, 0.0053], device='cuda:1') 2023-04-28 02:32:57,253 INFO [finetune.py:976] (1/7) Epoch 28, batch 2200, loss[loss=0.1681, simple_loss=0.2495, pruned_loss=0.04332, over 4753.00 frames. ], tot_loss[loss=0.1661, simple_loss=0.2401, pruned_loss=0.0461, over 955416.93 frames. ], batch size: 28, lr: 2.88e-03, grad_scale: 32.0 2023-04-28 02:33:14,527 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.8660, 2.5719, 2.9101, 3.1682, 2.8242, 2.6719, 2.8171, 2.6812], device='cuda:1'), covar=tensor([0.3653, 0.6060, 0.5665, 0.4518, 0.5177, 0.6935, 0.7243, 0.7230], device='cuda:1'), in_proj_covar=tensor([0.0446, 0.0425, 0.0522, 0.0509, 0.0475, 0.0512, 0.0512, 0.0529], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-28 02:33:15,052 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=156872.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:33:17,875 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.069e+02 1.564e+02 1.773e+02 2.100e+02 3.301e+02, threshold=3.547e+02, percent-clipped=0.0 2023-04-28 02:33:23,530 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=156885.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:33:29,721 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.4086, 1.8826, 2.2451, 2.9708, 2.3740, 1.8006, 1.9528, 2.2891], device='cuda:1'), covar=tensor([0.2661, 0.2834, 0.1454, 0.1936, 0.2545, 0.2370, 0.3444, 0.1829], device='cuda:1'), in_proj_covar=tensor([0.0292, 0.0246, 0.0229, 0.0314, 0.0222, 0.0234, 0.0228, 0.0186], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-28 02:33:31,016 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.64 vs. limit=2.0 2023-04-28 02:33:31,444 INFO [finetune.py:976] (1/7) Epoch 28, batch 2250, loss[loss=0.2031, simple_loss=0.2846, pruned_loss=0.06082, over 4842.00 frames. ], tot_loss[loss=0.1668, simple_loss=0.2409, pruned_loss=0.04638, over 955116.78 frames. ], batch size: 49, lr: 2.88e-03, grad_scale: 32.0 2023-04-28 02:33:47,971 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=156920.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:34:04,149 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=156946.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:34:04,194 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=156946.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:34:05,294 INFO [finetune.py:976] (1/7) Epoch 28, batch 2300, loss[loss=0.1496, simple_loss=0.2262, pruned_loss=0.03651, over 4758.00 frames. ], tot_loss[loss=0.1661, simple_loss=0.2404, pruned_loss=0.0459, over 952995.52 frames. ], batch size: 28, lr: 2.88e-03, grad_scale: 32.0 2023-04-28 02:34:10,643 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6358, 2.5028, 2.0135, 2.2389, 2.4917, 2.1544, 3.2619, 1.9627], device='cuda:1'), covar=tensor([0.3558, 0.2226, 0.4665, 0.3074, 0.1769, 0.2642, 0.1714, 0.4288], device='cuda:1'), in_proj_covar=tensor([0.0340, 0.0356, 0.0426, 0.0352, 0.0383, 0.0377, 0.0371, 0.0424], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-28 02:34:25,202 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.091e+02 1.469e+02 1.728e+02 2.151e+02 3.571e+02, threshold=3.456e+02, percent-clipped=1.0 2023-04-28 02:34:36,690 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=156994.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:34:39,083 INFO [finetune.py:976] (1/7) Epoch 28, batch 2350, loss[loss=0.1518, simple_loss=0.2212, pruned_loss=0.04119, over 4034.00 frames. ], tot_loss[loss=0.1644, simple_loss=0.2385, pruned_loss=0.04516, over 954376.13 frames. ], batch size: 17, lr: 2.88e-03, grad_scale: 32.0 2023-04-28 02:34:46,315 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0751, 3.1031, 2.6874, 2.9112, 3.2033, 2.8382, 4.1045, 2.6875], device='cuda:1'), covar=tensor([0.3300, 0.1742, 0.3371, 0.2742, 0.1531, 0.2268, 0.1017, 0.3073], device='cuda:1'), in_proj_covar=tensor([0.0340, 0.0355, 0.0425, 0.0351, 0.0382, 0.0377, 0.0371, 0.0423], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-28 02:34:47,984 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=157010.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:35:23,596 INFO [finetune.py:976] (1/7) Epoch 28, batch 2400, loss[loss=0.1938, simple_loss=0.2714, pruned_loss=0.05804, over 4758.00 frames. ], tot_loss[loss=0.1635, simple_loss=0.2368, pruned_loss=0.0451, over 953038.07 frames. ], batch size: 26, lr: 2.88e-03, grad_scale: 32.0 2023-04-28 02:35:48,424 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-04-28 02:35:50,208 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=157071.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:35:53,120 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.080e+02 1.553e+02 1.893e+02 2.191e+02 4.408e+02, threshold=3.786e+02, percent-clipped=4.0 2023-04-28 02:36:07,059 INFO [finetune.py:976] (1/7) Epoch 28, batch 2450, loss[loss=0.1426, simple_loss=0.2189, pruned_loss=0.03311, over 4786.00 frames. ], tot_loss[loss=0.162, simple_loss=0.2345, pruned_loss=0.04478, over 951792.96 frames. ], batch size: 29, lr: 2.88e-03, grad_scale: 32.0 2023-04-28 02:36:11,402 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=157105.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:36:13,891 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.3995, 1.3408, 1.7618, 1.6889, 1.3147, 1.1715, 1.3605, 0.8934], device='cuda:1'), covar=tensor([0.0628, 0.0650, 0.0379, 0.0739, 0.0861, 0.1157, 0.0688, 0.0688], device='cuda:1'), in_proj_covar=tensor([0.0070, 0.0068, 0.0066, 0.0069, 0.0076, 0.0095, 0.0073, 0.0063], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-28 02:36:26,628 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([3.8593, 3.8006, 2.7980, 4.5090, 3.8441, 3.9308, 1.7493, 3.8744], device='cuda:1'), covar=tensor([0.1799, 0.1206, 0.3287, 0.1337, 0.4420, 0.1817, 0.5612, 0.2189], device='cuda:1'), in_proj_covar=tensor([0.0245, 0.0220, 0.0252, 0.0303, 0.0299, 0.0248, 0.0274, 0.0272], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 02:36:58,723 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=157142.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:37:07,638 INFO [finetune.py:976] (1/7) Epoch 28, batch 2500, loss[loss=0.1408, simple_loss=0.2144, pruned_loss=0.0336, over 4829.00 frames. ], tot_loss[loss=0.1645, simple_loss=0.237, pruned_loss=0.04601, over 954525.40 frames. ], batch size: 25, lr: 2.88e-03, grad_scale: 32.0 2023-04-28 02:37:39,252 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 2023-04-28 02:37:44,333 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.070e+02 1.608e+02 1.980e+02 2.407e+02 6.066e+02, threshold=3.960e+02, percent-clipped=5.0 2023-04-28 02:38:14,042 INFO [finetune.py:976] (1/7) Epoch 28, batch 2550, loss[loss=0.188, simple_loss=0.2592, pruned_loss=0.0584, over 4929.00 frames. ], tot_loss[loss=0.1664, simple_loss=0.2396, pruned_loss=0.04656, over 953788.71 frames. ], batch size: 33, lr: 2.88e-03, grad_scale: 32.0 2023-04-28 02:38:22,915 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=157203.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:39:11,685 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=157241.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:39:16,444 INFO [finetune.py:976] (1/7) Epoch 28, batch 2600, loss[loss=0.1807, simple_loss=0.2612, pruned_loss=0.05013, over 4808.00 frames. ], tot_loss[loss=0.1667, simple_loss=0.2405, pruned_loss=0.04646, over 952786.15 frames. ], batch size: 51, lr: 2.88e-03, grad_scale: 32.0 2023-04-28 02:39:33,956 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-04-28 02:39:50,661 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.059e+02 1.630e+02 1.851e+02 2.303e+02 4.332e+02, threshold=3.701e+02, percent-clipped=1.0 2023-04-28 02:40:05,059 INFO [finetune.py:976] (1/7) Epoch 28, batch 2650, loss[loss=0.1687, simple_loss=0.2368, pruned_loss=0.05026, over 4865.00 frames. ], tot_loss[loss=0.1674, simple_loss=0.2415, pruned_loss=0.04664, over 951914.96 frames. ], batch size: 34, lr: 2.88e-03, grad_scale: 32.0 2023-04-28 02:40:11,732 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9061, 1.7516, 1.8716, 2.2629, 2.2700, 1.6955, 1.5544, 1.9311], device='cuda:1'), covar=tensor([0.0751, 0.1037, 0.0736, 0.0453, 0.0552, 0.0821, 0.0715, 0.0546], device='cuda:1'), in_proj_covar=tensor([0.0184, 0.0201, 0.0184, 0.0169, 0.0178, 0.0177, 0.0151, 0.0177], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-28 02:40:53,194 INFO [finetune.py:976] (1/7) Epoch 28, batch 2700, loss[loss=0.1336, simple_loss=0.2066, pruned_loss=0.03027, over 4924.00 frames. ], tot_loss[loss=0.1657, simple_loss=0.2399, pruned_loss=0.0457, over 951832.99 frames. ], batch size: 33, lr: 2.88e-03, grad_scale: 32.0 2023-04-28 02:40:58,710 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=157356.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:41:04,709 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=157366.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:41:10,643 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.027e+02 1.425e+02 1.670e+02 2.111e+02 3.242e+02, threshold=3.340e+02, percent-clipped=0.0 2023-04-28 02:41:15,932 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=157382.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:41:26,099 INFO [finetune.py:976] (1/7) Epoch 28, batch 2750, loss[loss=0.1477, simple_loss=0.2278, pruned_loss=0.03381, over 4863.00 frames. ], tot_loss[loss=0.1652, simple_loss=0.2386, pruned_loss=0.04592, over 952905.41 frames. ], batch size: 44, lr: 2.88e-03, grad_scale: 32.0 2023-04-28 02:41:30,006 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=157404.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 02:41:31,100 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=157405.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:41:38,348 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=157417.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:42:02,271 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=157443.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:42:05,179 INFO [finetune.py:976] (1/7) Epoch 28, batch 2800, loss[loss=0.1581, simple_loss=0.2333, pruned_loss=0.04152, over 4908.00 frames. ], tot_loss[loss=0.1635, simple_loss=0.2365, pruned_loss=0.04523, over 954123.20 frames. ], batch size: 36, lr: 2.88e-03, grad_scale: 32.0 2023-04-28 02:42:13,773 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=157453.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:42:18,596 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.3751, 1.8105, 1.6514, 2.1396, 1.8784, 2.0031, 1.6230, 4.4263], device='cuda:1'), covar=tensor([0.0540, 0.0764, 0.0777, 0.1139, 0.0617, 0.0545, 0.0701, 0.0109], device='cuda:1'), in_proj_covar=tensor([0.0037, 0.0038, 0.0039, 0.0042, 0.0040, 0.0037, 0.0038, 0.0054], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0013, 0.0013, 0.0015, 0.0014, 0.0013, 0.0013, 0.0016], device='cuda:1') 2023-04-28 02:42:26,573 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=157465.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 02:42:38,114 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.163e+02 1.553e+02 1.775e+02 2.169e+02 3.521e+02, threshold=3.549e+02, percent-clipped=1.0 2023-04-28 02:43:07,286 INFO [finetune.py:976] (1/7) Epoch 28, batch 2850, loss[loss=0.196, simple_loss=0.2592, pruned_loss=0.0664, over 4913.00 frames. ], tot_loss[loss=0.1628, simple_loss=0.2354, pruned_loss=0.04507, over 955188.00 frames. ], batch size: 36, lr: 2.88e-03, grad_scale: 32.0 2023-04-28 02:43:07,356 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=157498.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:43:07,476 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.94 vs. limit=5.0 2023-04-28 02:43:17,201 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=157506.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 02:44:00,011 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=157541.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:44:09,515 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.75 vs. limit=2.0 2023-04-28 02:44:09,939 INFO [finetune.py:976] (1/7) Epoch 28, batch 2900, loss[loss=0.169, simple_loss=0.2486, pruned_loss=0.04468, over 4219.00 frames. ], tot_loss[loss=0.1643, simple_loss=0.2376, pruned_loss=0.04548, over 955150.92 frames. ], batch size: 65, lr: 2.88e-03, grad_scale: 32.0 2023-04-28 02:44:10,746 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.49 vs. limit=2.0 2023-04-28 02:44:33,684 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=157567.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 02:44:33,707 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.1277, 1.4001, 1.9104, 2.3966, 1.9962, 1.5424, 1.3441, 1.6845], device='cuda:1'), covar=tensor([0.3395, 0.3870, 0.1987, 0.2693, 0.2880, 0.2831, 0.4256, 0.2213], device='cuda:1'), in_proj_covar=tensor([0.0291, 0.0245, 0.0229, 0.0315, 0.0222, 0.0235, 0.0228, 0.0186], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-28 02:44:34,287 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=157568.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:44:44,347 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.234e+02 1.564e+02 1.889e+02 2.208e+02 3.535e+02, threshold=3.777e+02, percent-clipped=0.0 2023-04-28 02:45:03,636 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=157589.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:45:14,533 INFO [finetune.py:976] (1/7) Epoch 28, batch 2950, loss[loss=0.1561, simple_loss=0.2074, pruned_loss=0.0524, over 4022.00 frames. ], tot_loss[loss=0.1668, simple_loss=0.241, pruned_loss=0.04635, over 955610.20 frames. ], batch size: 17, lr: 2.88e-03, grad_scale: 32.0 2023-04-28 02:45:50,100 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=157629.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:46:09,061 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6512, 1.4635, 0.6297, 1.4030, 1.5089, 1.4935, 1.3925, 1.4256], device='cuda:1'), covar=tensor([0.0484, 0.0383, 0.0372, 0.0528, 0.0281, 0.0500, 0.0494, 0.0554], device='cuda:1'), in_proj_covar=tensor([0.0028, 0.0023, 0.0022, 0.0028, 0.0019, 0.0027, 0.0027, 0.0028], device='cuda:1'), out_proj_covar=tensor([0.0052, 0.0046, 0.0039, 0.0053, 0.0039, 0.0051, 0.0051, 0.0053], device='cuda:1') 2023-04-28 02:46:18,331 INFO [finetune.py:976] (1/7) Epoch 28, batch 3000, loss[loss=0.1445, simple_loss=0.2249, pruned_loss=0.03203, over 4825.00 frames. ], tot_loss[loss=0.1688, simple_loss=0.243, pruned_loss=0.04731, over 956625.13 frames. ], batch size: 30, lr: 2.88e-03, grad_scale: 32.0 2023-04-28 02:46:18,331 INFO [finetune.py:1001] (1/7) Computing validation loss 2023-04-28 02:46:26,678 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7755, 1.1279, 1.7531, 2.2396, 1.8684, 1.7224, 1.7200, 1.6914], device='cuda:1'), covar=tensor([0.4502, 0.7128, 0.6375, 0.5785, 0.6272, 0.8032, 0.8594, 0.9259], device='cuda:1'), in_proj_covar=tensor([0.0447, 0.0425, 0.0522, 0.0509, 0.0475, 0.0512, 0.0513, 0.0527], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-28 02:46:34,885 INFO [finetune.py:1010] (1/7) Epoch 28, validation: loss=0.153, simple_loss=0.2217, pruned_loss=0.04213, over 2265189.00 frames. 2023-04-28 02:46:34,885 INFO [finetune.py:1011] (1/7) Maximum memory allocated so far is 6391MB 2023-04-28 02:46:49,123 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=157666.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:46:55,059 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.120e+02 1.557e+02 1.829e+02 2.170e+02 4.324e+02, threshold=3.658e+02, percent-clipped=1.0 2023-04-28 02:47:03,621 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9218, 1.5212, 1.9774, 2.2271, 1.9379, 1.8518, 1.9039, 1.9231], device='cuda:1'), covar=tensor([0.5688, 0.8060, 0.8559, 0.8465, 0.7358, 1.1499, 1.0332, 1.1367], device='cuda:1'), in_proj_covar=tensor([0.0447, 0.0425, 0.0522, 0.0509, 0.0475, 0.0512, 0.0512, 0.0527], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-28 02:47:09,008 INFO [finetune.py:976] (1/7) Epoch 28, batch 3050, loss[loss=0.1622, simple_loss=0.2304, pruned_loss=0.04695, over 4723.00 frames. ], tot_loss[loss=0.1687, simple_loss=0.243, pruned_loss=0.04715, over 956400.23 frames. ], batch size: 59, lr: 2.88e-03, grad_scale: 32.0 2023-04-28 02:47:27,243 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=157712.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:47:28,442 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=157714.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:47:28,567 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-04-28 02:47:58,444 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=157738.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:48:09,325 INFO [finetune.py:976] (1/7) Epoch 28, batch 3100, loss[loss=0.1528, simple_loss=0.217, pruned_loss=0.0443, over 4846.00 frames. ], tot_loss[loss=0.1671, simple_loss=0.2413, pruned_loss=0.04644, over 955799.21 frames. ], batch size: 49, lr: 2.88e-03, grad_scale: 32.0 2023-04-28 02:48:21,067 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-04-28 02:48:22,596 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=157760.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 02:48:32,272 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.835e+01 1.560e+02 1.859e+02 2.154e+02 3.848e+02, threshold=3.719e+02, percent-clipped=2.0 2023-04-28 02:48:38,672 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.35 vs. limit=5.0 2023-04-28 02:48:46,291 INFO [finetune.py:976] (1/7) Epoch 28, batch 3150, loss[loss=0.187, simple_loss=0.2538, pruned_loss=0.06009, over 4816.00 frames. ], tot_loss[loss=0.1653, simple_loss=0.2386, pruned_loss=0.046, over 955115.29 frames. ], batch size: 45, lr: 2.88e-03, grad_scale: 32.0 2023-04-28 02:48:46,404 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=157798.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:48:49,610 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8851, 2.3634, 2.0667, 2.2702, 1.5418, 2.0726, 1.9725, 1.5855], device='cuda:1'), covar=tensor([0.1871, 0.1087, 0.0727, 0.1141, 0.3613, 0.1102, 0.1867, 0.2415], device='cuda:1'), in_proj_covar=tensor([0.0284, 0.0303, 0.0218, 0.0278, 0.0316, 0.0255, 0.0249, 0.0264], device='cuda:1'), out_proj_covar=tensor([1.1307e-04, 1.1928e-04, 8.5686e-05, 1.0953e-04, 1.2741e-04, 1.0003e-04, 1.0024e-04, 1.0387e-04], device='cuda:1') 2023-04-28 02:49:18,160 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=157846.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:49:19,357 INFO [finetune.py:976] (1/7) Epoch 28, batch 3200, loss[loss=0.1608, simple_loss=0.2139, pruned_loss=0.05385, over 4878.00 frames. ], tot_loss[loss=0.1633, simple_loss=0.2362, pruned_loss=0.04525, over 956535.12 frames. ], batch size: 31, lr: 2.88e-03, grad_scale: 32.0 2023-04-28 02:49:22,555 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=157853.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:49:27,981 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=157862.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 02:49:38,913 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.062e+02 1.560e+02 1.811e+02 2.192e+02 5.644e+02, threshold=3.622e+02, percent-clipped=2.0 2023-04-28 02:50:07,745 INFO [finetune.py:976] (1/7) Epoch 28, batch 3250, loss[loss=0.1616, simple_loss=0.2359, pruned_loss=0.04368, over 4803.00 frames. ], tot_loss[loss=0.1635, simple_loss=0.2366, pruned_loss=0.04524, over 953900.54 frames. ], batch size: 45, lr: 2.87e-03, grad_scale: 32.0 2023-04-28 02:50:29,691 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=157914.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:50:32,045 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6442, 1.2737, 1.2980, 1.4414, 1.8109, 1.4823, 1.1962, 1.1967], device='cuda:1'), covar=tensor([0.1504, 0.1394, 0.1625, 0.1189, 0.0774, 0.1507, 0.2364, 0.2309], device='cuda:1'), in_proj_covar=tensor([0.0314, 0.0305, 0.0349, 0.0284, 0.0323, 0.0304, 0.0298, 0.0373], device='cuda:1'), out_proj_covar=tensor([6.3960e-05, 6.2464e-05, 7.3068e-05, 5.6941e-05, 6.5807e-05, 6.3343e-05, 6.1581e-05, 7.8882e-05], device='cuda:1') 2023-04-28 02:50:43,189 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=157924.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:50:50,337 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 2023-04-28 02:51:14,742 INFO [finetune.py:976] (1/7) Epoch 28, batch 3300, loss[loss=0.1819, simple_loss=0.2696, pruned_loss=0.04708, over 4853.00 frames. ], tot_loss[loss=0.1674, simple_loss=0.2411, pruned_loss=0.04681, over 953978.30 frames. ], batch size: 44, lr: 2.87e-03, grad_scale: 32.0 2023-04-28 02:51:49,059 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-04-28 02:51:51,135 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.051e+02 1.721e+02 1.967e+02 2.277e+02 5.584e+02, threshold=3.933e+02, percent-clipped=3.0 2023-04-28 02:51:57,978 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.3779, 3.0182, 2.5173, 2.9121, 2.0806, 2.6755, 2.6662, 1.9145], device='cuda:1'), covar=tensor([0.1945, 0.1367, 0.0769, 0.1037, 0.3106, 0.1139, 0.1774, 0.2688], device='cuda:1'), in_proj_covar=tensor([0.0284, 0.0304, 0.0219, 0.0278, 0.0317, 0.0255, 0.0250, 0.0265], device='cuda:1'), out_proj_covar=tensor([1.1336e-04, 1.1949e-04, 8.6007e-05, 1.0937e-04, 1.2765e-04, 1.0035e-04, 1.0045e-04, 1.0423e-04], device='cuda:1') 2023-04-28 02:52:20,397 INFO [finetune.py:976] (1/7) Epoch 28, batch 3350, loss[loss=0.1394, simple_loss=0.2135, pruned_loss=0.03265, over 4825.00 frames. ], tot_loss[loss=0.1687, simple_loss=0.2428, pruned_loss=0.04729, over 955912.73 frames. ], batch size: 33, lr: 2.87e-03, grad_scale: 32.0 2023-04-28 02:52:32,100 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-04-28 02:52:41,476 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=158012.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:53:12,149 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=158038.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:53:12,768 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.1717, 2.6959, 2.3232, 2.6750, 2.0015, 2.4062, 2.4206, 1.8033], device='cuda:1'), covar=tensor([0.1912, 0.1214, 0.0733, 0.1071, 0.2749, 0.0994, 0.1891, 0.2569], device='cuda:1'), in_proj_covar=tensor([0.0282, 0.0302, 0.0218, 0.0276, 0.0315, 0.0254, 0.0248, 0.0263], device='cuda:1'), out_proj_covar=tensor([1.1263e-04, 1.1893e-04, 8.5540e-05, 1.0875e-04, 1.2687e-04, 9.9698e-05, 9.9951e-05, 1.0370e-04], device='cuda:1') 2023-04-28 02:53:23,378 INFO [finetune.py:976] (1/7) Epoch 28, batch 3400, loss[loss=0.1354, simple_loss=0.2142, pruned_loss=0.02825, over 4757.00 frames. ], tot_loss[loss=0.1685, simple_loss=0.2426, pruned_loss=0.04717, over 956316.58 frames. ], batch size: 27, lr: 2.87e-03, grad_scale: 32.0 2023-04-28 02:53:41,206 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=158060.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:53:41,232 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=158060.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 02:53:44,447 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-04-28 02:53:55,703 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4730, 3.1128, 0.9367, 1.7503, 1.8412, 2.2334, 1.8395, 1.0829], device='cuda:1'), covar=tensor([0.1465, 0.0957, 0.1923, 0.1217, 0.1065, 0.0984, 0.1555, 0.1696], device='cuda:1'), in_proj_covar=tensor([0.0118, 0.0239, 0.0135, 0.0121, 0.0132, 0.0153, 0.0117, 0.0119], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-28 02:53:56,199 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.114e+02 1.582e+02 1.900e+02 2.246e+02 3.787e+02, threshold=3.800e+02, percent-clipped=0.0 2023-04-28 02:54:10,626 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=158086.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:54:23,032 INFO [finetune.py:976] (1/7) Epoch 28, batch 3450, loss[loss=0.1352, simple_loss=0.2062, pruned_loss=0.03217, over 4710.00 frames. ], tot_loss[loss=0.168, simple_loss=0.2425, pruned_loss=0.04672, over 957066.61 frames. ], batch size: 23, lr: 2.87e-03, grad_scale: 32.0 2023-04-28 02:54:34,533 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=158108.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 02:54:42,053 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=3.99 vs. limit=5.0 2023-04-28 02:54:54,869 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=158122.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:55:28,047 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0056, 2.4669, 1.1350, 1.3384, 1.9805, 1.1526, 3.0749, 1.7050], device='cuda:1'), covar=tensor([0.0702, 0.0555, 0.0723, 0.1303, 0.0463, 0.1082, 0.0265, 0.0595], device='cuda:1'), in_proj_covar=tensor([0.0050, 0.0064, 0.0047, 0.0046, 0.0049, 0.0051, 0.0072, 0.0050], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:1') 2023-04-28 02:55:29,823 INFO [finetune.py:976] (1/7) Epoch 28, batch 3500, loss[loss=0.1427, simple_loss=0.2105, pruned_loss=0.03747, over 4784.00 frames. ], tot_loss[loss=0.1669, simple_loss=0.2407, pruned_loss=0.0466, over 955408.24 frames. ], batch size: 23, lr: 2.87e-03, grad_scale: 32.0 2023-04-28 02:55:48,557 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=158162.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 02:56:07,658 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.013e+02 1.474e+02 1.724e+02 2.129e+02 5.829e+02, threshold=3.447e+02, percent-clipped=1.0 2023-04-28 02:56:12,613 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=158183.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:56:23,134 INFO [finetune.py:976] (1/7) Epoch 28, batch 3550, loss[loss=0.151, simple_loss=0.2253, pruned_loss=0.03834, over 4912.00 frames. ], tot_loss[loss=0.166, simple_loss=0.239, pruned_loss=0.04657, over 955759.99 frames. ], batch size: 43, lr: 2.87e-03, grad_scale: 32.0 2023-04-28 02:56:30,177 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=158209.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:56:30,777 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=158210.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 02:56:30,829 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.2645, 1.8589, 1.5465, 2.1547, 2.3478, 1.9143, 1.8591, 1.6261], device='cuda:1'), covar=tensor([0.1955, 0.1574, 0.1938, 0.1589, 0.1095, 0.1851, 0.2171, 0.2387], device='cuda:1'), in_proj_covar=tensor([0.0315, 0.0307, 0.0350, 0.0285, 0.0324, 0.0306, 0.0300, 0.0375], device='cuda:1'), out_proj_covar=tensor([6.4162e-05, 6.2949e-05, 7.3297e-05, 5.7094e-05, 6.6100e-05, 6.3581e-05, 6.1949e-05, 7.9205e-05], device='cuda:1') 2023-04-28 02:56:39,839 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=158224.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:56:56,634 INFO [finetune.py:976] (1/7) Epoch 28, batch 3600, loss[loss=0.1895, simple_loss=0.2562, pruned_loss=0.06139, over 4848.00 frames. ], tot_loss[loss=0.1655, simple_loss=0.2375, pruned_loss=0.04672, over 954193.06 frames. ], batch size: 49, lr: 2.87e-03, grad_scale: 32.0 2023-04-28 02:57:11,756 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=158272.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:57:14,138 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.946e+01 1.583e+02 1.935e+02 2.237e+02 3.659e+02, threshold=3.870e+02, percent-clipped=1.0 2023-04-28 02:57:24,874 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([0.8943, 0.9042, 1.0385, 1.0030, 0.9127, 0.7725, 0.8918, 0.5841], device='cuda:1'), covar=tensor([0.0458, 0.0385, 0.0475, 0.0480, 0.0565, 0.0814, 0.0383, 0.0525], device='cuda:1'), in_proj_covar=tensor([0.0070, 0.0067, 0.0065, 0.0069, 0.0075, 0.0094, 0.0072, 0.0062], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-28 02:57:29,418 INFO [finetune.py:976] (1/7) Epoch 28, batch 3650, loss[loss=0.1324, simple_loss=0.2049, pruned_loss=0.02992, over 4762.00 frames. ], tot_loss[loss=0.1667, simple_loss=0.2389, pruned_loss=0.04721, over 951114.58 frames. ], batch size: 26, lr: 2.87e-03, grad_scale: 32.0 2023-04-28 02:57:34,399 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0311, 1.6314, 4.0505, 3.8085, 3.5611, 3.8407, 3.6935, 3.5705], device='cuda:1'), covar=tensor([0.6567, 0.5214, 0.1202, 0.1753, 0.1231, 0.1582, 0.3060, 0.1665], device='cuda:1'), in_proj_covar=tensor([0.0311, 0.0311, 0.0408, 0.0409, 0.0352, 0.0418, 0.0319, 0.0366], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-28 02:57:41,233 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.1889, 1.5944, 1.6461, 1.7400, 1.6454, 1.7796, 1.7502, 1.6390], device='cuda:1'), covar=tensor([0.3721, 0.4374, 0.3914, 0.3838, 0.5430, 0.6903, 0.4572, 0.4495], device='cuda:1'), in_proj_covar=tensor([0.0344, 0.0377, 0.0331, 0.0343, 0.0352, 0.0393, 0.0362, 0.0335], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 02:57:44,852 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-04-28 02:57:47,776 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.1961, 1.5021, 1.3239, 1.6405, 1.5739, 1.9024, 1.3899, 3.3614], device='cuda:1'), covar=tensor([0.0597, 0.0785, 0.0796, 0.1205, 0.0616, 0.0498, 0.0732, 0.0138], device='cuda:1'), in_proj_covar=tensor([0.0037, 0.0038, 0.0039, 0.0042, 0.0040, 0.0037, 0.0038, 0.0054], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0013, 0.0013, 0.0015, 0.0014, 0.0013, 0.0013, 0.0016], device='cuda:1') 2023-04-28 02:58:02,904 INFO [finetune.py:976] (1/7) Epoch 28, batch 3700, loss[loss=0.1775, simple_loss=0.2656, pruned_loss=0.0447, over 4764.00 frames. ], tot_loss[loss=0.1685, simple_loss=0.2415, pruned_loss=0.04772, over 950899.77 frames. ], batch size: 28, lr: 2.87e-03, grad_scale: 32.0 2023-04-28 02:58:06,623 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.3831, 1.7212, 1.6185, 1.8443, 1.8094, 2.1424, 1.4957, 3.7990], device='cuda:1'), covar=tensor([0.0566, 0.0806, 0.0777, 0.1228, 0.0640, 0.0448, 0.0751, 0.0118], device='cuda:1'), in_proj_covar=tensor([0.0037, 0.0039, 0.0039, 0.0043, 0.0040, 0.0038, 0.0038, 0.0054], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0013, 0.0013, 0.0015, 0.0014, 0.0013, 0.0013, 0.0016], device='cuda:1') 2023-04-28 02:58:09,674 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5354, 1.3372, 0.6623, 1.2556, 1.3997, 1.4129, 1.2982, 1.3370], device='cuda:1'), covar=tensor([0.0492, 0.0380, 0.0362, 0.0540, 0.0307, 0.0474, 0.0491, 0.0568], device='cuda:1'), in_proj_covar=tensor([0.0028, 0.0023, 0.0022, 0.0028, 0.0019, 0.0027, 0.0027, 0.0028], device='cuda:1'), out_proj_covar=tensor([0.0052, 0.0046, 0.0039, 0.0053, 0.0039, 0.0051, 0.0051, 0.0053], device='cuda:1') 2023-04-28 02:58:19,866 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.152e+02 1.537e+02 1.915e+02 2.211e+02 4.327e+02, threshold=3.830e+02, percent-clipped=1.0 2023-04-28 02:58:25,399 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=158384.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:58:35,209 INFO [finetune.py:976] (1/7) Epoch 28, batch 3750, loss[loss=0.1808, simple_loss=0.2646, pruned_loss=0.04848, over 4898.00 frames. ], tot_loss[loss=0.1686, simple_loss=0.2423, pruned_loss=0.04752, over 953815.62 frames. ], batch size: 36, lr: 2.87e-03, grad_scale: 32.0 2023-04-28 02:58:36,583 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 2023-04-28 02:58:40,027 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7821, 1.7426, 2.0641, 2.2083, 1.6399, 1.5258, 1.7152, 1.0471], device='cuda:1'), covar=tensor([0.0627, 0.0579, 0.0441, 0.0855, 0.0762, 0.0992, 0.0710, 0.0646], device='cuda:1'), in_proj_covar=tensor([0.0070, 0.0067, 0.0065, 0.0069, 0.0074, 0.0094, 0.0072, 0.0061], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-28 02:59:02,016 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.86 vs. limit=5.0 2023-04-28 02:59:05,533 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=158445.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:59:07,664 INFO [finetune.py:976] (1/7) Epoch 28, batch 3800, loss[loss=0.149, simple_loss=0.2381, pruned_loss=0.02989, over 4693.00 frames. ], tot_loss[loss=0.1687, simple_loss=0.2432, pruned_loss=0.04713, over 954808.22 frames. ], batch size: 59, lr: 2.87e-03, grad_scale: 32.0 2023-04-28 02:59:41,648 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.2792, 1.5546, 1.8123, 1.8759, 1.8235, 1.8618, 1.8601, 1.8063], device='cuda:1'), covar=tensor([0.3507, 0.4927, 0.4038, 0.4038, 0.5172, 0.6527, 0.4636, 0.4315], device='cuda:1'), in_proj_covar=tensor([0.0343, 0.0377, 0.0330, 0.0342, 0.0351, 0.0393, 0.0361, 0.0334], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 02:59:42,093 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 8.322e+01 1.536e+02 1.803e+02 2.140e+02 3.917e+02, threshold=3.606e+02, percent-clipped=1.0 2023-04-28 02:59:43,423 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=158478.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:00:05,721 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0989, 2.6478, 1.2155, 1.4442, 2.1301, 1.2866, 3.5293, 1.9459], device='cuda:1'), covar=tensor([0.0673, 0.0555, 0.0691, 0.1180, 0.0459, 0.0940, 0.0234, 0.0560], device='cuda:1'), in_proj_covar=tensor([0.0050, 0.0064, 0.0047, 0.0046, 0.0049, 0.0051, 0.0073, 0.0050], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:1') 2023-04-28 03:00:07,899 INFO [finetune.py:976] (1/7) Epoch 28, batch 3850, loss[loss=0.1488, simple_loss=0.2343, pruned_loss=0.0317, over 4755.00 frames. ], tot_loss[loss=0.1672, simple_loss=0.2415, pruned_loss=0.0465, over 952472.83 frames. ], batch size: 27, lr: 2.87e-03, grad_scale: 32.0 2023-04-28 03:00:26,826 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=158509.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:01:06,732 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.2306, 2.7338, 2.2340, 2.2163, 1.5644, 1.6101, 2.3975, 1.5593], device='cuda:1'), covar=tensor([0.1653, 0.1415, 0.1369, 0.1767, 0.2407, 0.2112, 0.0948, 0.2133], device='cuda:1'), in_proj_covar=tensor([0.0200, 0.0211, 0.0171, 0.0206, 0.0202, 0.0188, 0.0158, 0.0189], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-28 03:01:10,982 INFO [finetune.py:976] (1/7) Epoch 28, batch 3900, loss[loss=0.1493, simple_loss=0.2155, pruned_loss=0.04155, over 4784.00 frames. ], tot_loss[loss=0.1643, simple_loss=0.2381, pruned_loss=0.04529, over 954457.43 frames. ], batch size: 29, lr: 2.87e-03, grad_scale: 32.0 2023-04-28 03:01:28,061 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=158557.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:01:51,820 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.216e+01 1.537e+02 1.820e+02 2.187e+02 3.711e+02, threshold=3.639e+02, percent-clipped=1.0 2023-04-28 03:02:23,135 INFO [finetune.py:976] (1/7) Epoch 28, batch 3950, loss[loss=0.152, simple_loss=0.2229, pruned_loss=0.04059, over 4797.00 frames. ], tot_loss[loss=0.1619, simple_loss=0.235, pruned_loss=0.04437, over 954878.44 frames. ], batch size: 29, lr: 2.87e-03, grad_scale: 32.0 2023-04-28 03:03:31,309 INFO [finetune.py:976] (1/7) Epoch 28, batch 4000, loss[loss=0.155, simple_loss=0.2375, pruned_loss=0.03624, over 4754.00 frames. ], tot_loss[loss=0.161, simple_loss=0.2337, pruned_loss=0.04419, over 951884.47 frames. ], batch size: 23, lr: 2.87e-03, grad_scale: 32.0 2023-04-28 03:04:13,586 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.149e+01 1.505e+02 1.815e+02 2.168e+02 4.885e+02, threshold=3.630e+02, percent-clipped=2.0 2023-04-28 03:04:37,072 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5046, 0.7935, 1.5132, 1.8623, 1.6114, 1.4213, 1.4819, 1.4419], device='cuda:1'), covar=tensor([0.3184, 0.4868, 0.3910, 0.4104, 0.4152, 0.5341, 0.5202, 0.6031], device='cuda:1'), in_proj_covar=tensor([0.0444, 0.0424, 0.0520, 0.0509, 0.0473, 0.0510, 0.0511, 0.0526], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-28 03:04:38,159 INFO [finetune.py:976] (1/7) Epoch 28, batch 4050, loss[loss=0.2223, simple_loss=0.2948, pruned_loss=0.07484, over 4935.00 frames. ], tot_loss[loss=0.1637, simple_loss=0.2373, pruned_loss=0.04499, over 953257.17 frames. ], batch size: 33, lr: 2.87e-03, grad_scale: 32.0 2023-04-28 03:05:22,481 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7321, 2.9636, 2.4399, 2.5720, 2.9979, 2.4980, 3.9158, 2.2934], device='cuda:1'), covar=tensor([0.4089, 0.2171, 0.4135, 0.3485, 0.1829, 0.2737, 0.1407, 0.3784], device='cuda:1'), in_proj_covar=tensor([0.0340, 0.0356, 0.0425, 0.0353, 0.0384, 0.0379, 0.0372, 0.0425], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-28 03:05:33,917 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=158740.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:05:43,860 INFO [finetune.py:976] (1/7) Epoch 28, batch 4100, loss[loss=0.1648, simple_loss=0.2404, pruned_loss=0.04456, over 4050.00 frames. ], tot_loss[loss=0.1644, simple_loss=0.2389, pruned_loss=0.04495, over 952517.07 frames. ], batch size: 65, lr: 2.87e-03, grad_scale: 32.0 2023-04-28 03:06:25,666 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.124e+02 1.629e+02 1.870e+02 2.384e+02 6.257e+02, threshold=3.741e+02, percent-clipped=1.0 2023-04-28 03:06:26,417 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=158777.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:06:26,986 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=158778.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:06:35,227 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=158783.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:06:49,753 INFO [finetune.py:976] (1/7) Epoch 28, batch 4150, loss[loss=0.2044, simple_loss=0.2738, pruned_loss=0.06753, over 4717.00 frames. ], tot_loss[loss=0.1661, simple_loss=0.2405, pruned_loss=0.0459, over 950073.48 frames. ], batch size: 59, lr: 2.87e-03, grad_scale: 64.0 2023-04-28 03:07:30,858 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=158826.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:07:41,990 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=158835.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 03:07:43,879 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=158838.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:07:52,643 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=158844.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:07:54,973 INFO [finetune.py:976] (1/7) Epoch 28, batch 4200, loss[loss=0.1563, simple_loss=0.2352, pruned_loss=0.03868, over 4809.00 frames. ], tot_loss[loss=0.1647, simple_loss=0.2396, pruned_loss=0.0449, over 951725.41 frames. ], batch size: 40, lr: 2.87e-03, grad_scale: 32.0 2023-04-28 03:08:37,326 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.042e+02 1.525e+02 1.688e+02 1.998e+02 3.338e+02, threshold=3.377e+02, percent-clipped=0.0 2023-04-28 03:09:04,962 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=158896.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 03:09:06,081 INFO [finetune.py:976] (1/7) Epoch 28, batch 4250, loss[loss=0.2049, simple_loss=0.2655, pruned_loss=0.0721, over 4230.00 frames. ], tot_loss[loss=0.1633, simple_loss=0.2375, pruned_loss=0.04459, over 952167.81 frames. ], batch size: 65, lr: 2.87e-03, grad_scale: 32.0 2023-04-28 03:10:02,523 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8429, 2.0451, 1.2050, 1.5533, 2.2500, 1.7090, 1.5663, 1.6913], device='cuda:1'), covar=tensor([0.0458, 0.0345, 0.0260, 0.0543, 0.0224, 0.0476, 0.0453, 0.0535], device='cuda:1'), in_proj_covar=tensor([0.0027, 0.0023, 0.0022, 0.0028, 0.0019, 0.0027, 0.0027, 0.0028], device='cuda:1'), out_proj_covar=tensor([0.0052, 0.0046, 0.0039, 0.0053, 0.0039, 0.0051, 0.0051, 0.0053], device='cuda:1') 2023-04-28 03:10:12,866 INFO [finetune.py:976] (1/7) Epoch 28, batch 4300, loss[loss=0.1334, simple_loss=0.1989, pruned_loss=0.034, over 4857.00 frames. ], tot_loss[loss=0.1619, simple_loss=0.2353, pruned_loss=0.04423, over 953819.34 frames. ], batch size: 47, lr: 2.87e-03, grad_scale: 32.0 2023-04-28 03:10:48,763 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 8.467e+01 1.425e+02 1.689e+02 1.969e+02 4.397e+02, threshold=3.377e+02, percent-clipped=1.0 2023-04-28 03:11:18,966 INFO [finetune.py:976] (1/7) Epoch 28, batch 4350, loss[loss=0.1535, simple_loss=0.2363, pruned_loss=0.03531, over 4897.00 frames. ], tot_loss[loss=0.1606, simple_loss=0.2335, pruned_loss=0.04391, over 953012.58 frames. ], batch size: 43, lr: 2.87e-03, grad_scale: 32.0 2023-04-28 03:11:51,780 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5131, 2.4832, 1.9306, 2.0798, 2.4968, 2.0951, 3.2807, 1.8245], device='cuda:1'), covar=tensor([0.3910, 0.2637, 0.4878, 0.3832, 0.2048, 0.3025, 0.1713, 0.4532], device='cuda:1'), in_proj_covar=tensor([0.0339, 0.0356, 0.0423, 0.0353, 0.0384, 0.0377, 0.0373, 0.0424], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-28 03:12:15,696 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=159040.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:12:26,355 INFO [finetune.py:976] (1/7) Epoch 28, batch 4400, loss[loss=0.2021, simple_loss=0.291, pruned_loss=0.05665, over 4807.00 frames. ], tot_loss[loss=0.1604, simple_loss=0.2336, pruned_loss=0.04364, over 952577.62 frames. ], batch size: 41, lr: 2.87e-03, grad_scale: 32.0 2023-04-28 03:12:54,445 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=159069.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:13:05,084 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.784e+01 1.599e+02 1.852e+02 2.200e+02 4.625e+02, threshold=3.703e+02, percent-clipped=1.0 2023-04-28 03:13:05,815 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9357, 1.4571, 4.0442, 3.8132, 3.5646, 3.7879, 3.6790, 3.5988], device='cuda:1'), covar=tensor([0.6738, 0.5502, 0.0997, 0.1450, 0.1070, 0.1799, 0.3838, 0.1412], device='cuda:1'), in_proj_covar=tensor([0.0310, 0.0310, 0.0404, 0.0408, 0.0350, 0.0415, 0.0317, 0.0363], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-28 03:13:19,036 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=159088.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:13:29,730 INFO [finetune.py:976] (1/7) Epoch 28, batch 4450, loss[loss=0.2035, simple_loss=0.2696, pruned_loss=0.0687, over 4859.00 frames. ], tot_loss[loss=0.1631, simple_loss=0.237, pruned_loss=0.04457, over 952838.26 frames. ], batch size: 31, lr: 2.87e-03, grad_scale: 32.0 2023-04-28 03:13:38,907 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-04-28 03:14:09,972 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=159130.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:14:12,213 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=159133.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:14:21,844 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=159139.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:14:27,807 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.4396, 3.0145, 2.2762, 2.4042, 1.7043, 1.7075, 2.3762, 1.8335], device='cuda:1'), covar=tensor([0.1538, 0.1197, 0.1343, 0.1572, 0.2138, 0.1779, 0.1004, 0.1830], device='cuda:1'), in_proj_covar=tensor([0.0198, 0.0210, 0.0170, 0.0205, 0.0201, 0.0186, 0.0156, 0.0188], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-28 03:14:38,434 INFO [finetune.py:976] (1/7) Epoch 28, batch 4500, loss[loss=0.1501, simple_loss=0.2315, pruned_loss=0.03436, over 4923.00 frames. ], tot_loss[loss=0.1636, simple_loss=0.2376, pruned_loss=0.04483, over 950717.77 frames. ], batch size: 33, lr: 2.87e-03, grad_scale: 32.0 2023-04-28 03:14:49,790 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=159159.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:15:11,107 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.154e+02 1.553e+02 1.824e+02 2.169e+02 4.146e+02, threshold=3.648e+02, percent-clipped=1.0 2023-04-28 03:15:31,025 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=159191.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 03:15:41,723 INFO [finetune.py:976] (1/7) Epoch 28, batch 4550, loss[loss=0.1833, simple_loss=0.2471, pruned_loss=0.05971, over 4814.00 frames. ], tot_loss[loss=0.1654, simple_loss=0.2398, pruned_loss=0.04551, over 953173.24 frames. ], batch size: 38, lr: 2.87e-03, grad_scale: 32.0 2023-04-28 03:16:05,616 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=159220.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:16:27,835 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.0915, 1.3654, 1.2169, 1.5666, 1.4938, 1.5322, 1.3158, 2.4022], device='cuda:1'), covar=tensor([0.0606, 0.0794, 0.0802, 0.1196, 0.0625, 0.0502, 0.0750, 0.0204], device='cuda:1'), in_proj_covar=tensor([0.0038, 0.0039, 0.0039, 0.0043, 0.0040, 0.0038, 0.0038, 0.0055], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0013, 0.0014, 0.0015, 0.0014, 0.0013, 0.0013, 0.0016], device='cuda:1') 2023-04-28 03:16:45,688 INFO [finetune.py:976] (1/7) Epoch 28, batch 4600, loss[loss=0.2036, simple_loss=0.2796, pruned_loss=0.06378, over 4839.00 frames. ], tot_loss[loss=0.1653, simple_loss=0.2396, pruned_loss=0.04548, over 953072.35 frames. ], batch size: 49, lr: 2.87e-03, grad_scale: 32.0 2023-04-28 03:17:06,274 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8799, 2.0601, 1.8441, 1.6069, 1.3895, 1.3906, 1.8801, 1.3702], device='cuda:1'), covar=tensor([0.1604, 0.1313, 0.1309, 0.1619, 0.2342, 0.1923, 0.0943, 0.2027], device='cuda:1'), in_proj_covar=tensor([0.0199, 0.0210, 0.0171, 0.0206, 0.0202, 0.0187, 0.0157, 0.0189], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-28 03:17:18,839 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.827e+01 1.517e+02 1.769e+02 2.322e+02 3.935e+02, threshold=3.539e+02, percent-clipped=1.0 2023-04-28 03:17:29,018 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([3.9247, 3.8624, 2.7697, 4.4995, 4.0060, 3.9395, 1.4537, 3.9020], device='cuda:1'), covar=tensor([0.1525, 0.1225, 0.3009, 0.1561, 0.3586, 0.1681, 0.6357, 0.2216], device='cuda:1'), in_proj_covar=tensor([0.0246, 0.0221, 0.0252, 0.0305, 0.0301, 0.0251, 0.0277, 0.0275], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 03:17:38,393 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=159290.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:17:49,023 INFO [finetune.py:976] (1/7) Epoch 28, batch 4650, loss[loss=0.1733, simple_loss=0.2531, pruned_loss=0.0467, over 4823.00 frames. ], tot_loss[loss=0.164, simple_loss=0.2378, pruned_loss=0.04511, over 953806.56 frames. ], batch size: 40, lr: 2.87e-03, grad_scale: 32.0 2023-04-28 03:18:18,590 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.04 vs. limit=5.0 2023-04-28 03:18:19,759 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.2622, 2.7687, 2.1681, 2.2793, 1.5240, 1.5790, 2.3067, 1.6065], device='cuda:1'), covar=tensor([0.1594, 0.1423, 0.1398, 0.1692, 0.2218, 0.1901, 0.0972, 0.2071], device='cuda:1'), in_proj_covar=tensor([0.0198, 0.0210, 0.0171, 0.0206, 0.0201, 0.0187, 0.0157, 0.0189], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-28 03:18:21,170 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-04-28 03:18:54,507 INFO [finetune.py:976] (1/7) Epoch 28, batch 4700, loss[loss=0.1477, simple_loss=0.225, pruned_loss=0.03526, over 4775.00 frames. ], tot_loss[loss=0.1625, simple_loss=0.2361, pruned_loss=0.04452, over 956261.45 frames. ], batch size: 28, lr: 2.87e-03, grad_scale: 32.0 2023-04-28 03:19:03,244 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=159351.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:19:36,242 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.039e+02 1.471e+02 1.707e+02 1.992e+02 4.348e+02, threshold=3.414e+02, percent-clipped=1.0 2023-04-28 03:20:00,616 INFO [finetune.py:976] (1/7) Epoch 28, batch 4750, loss[loss=0.2003, simple_loss=0.2737, pruned_loss=0.06347, over 4919.00 frames. ], tot_loss[loss=0.163, simple_loss=0.2358, pruned_loss=0.04512, over 955964.43 frames. ], batch size: 38, lr: 2.87e-03, grad_scale: 32.0 2023-04-28 03:20:11,263 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4263, 1.5608, 1.4480, 1.7430, 1.7628, 2.0073, 1.5388, 3.3970], device='cuda:1'), covar=tensor([0.0587, 0.0802, 0.0831, 0.1248, 0.0624, 0.0436, 0.0714, 0.0159], device='cuda:1'), in_proj_covar=tensor([0.0038, 0.0039, 0.0039, 0.0043, 0.0040, 0.0038, 0.0038, 0.0055], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0013, 0.0014, 0.0015, 0.0014, 0.0013, 0.0013, 0.0016], device='cuda:1') 2023-04-28 03:20:41,287 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=159425.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:20:46,225 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=159433.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:20:55,228 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=159439.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:21:06,689 INFO [finetune.py:976] (1/7) Epoch 28, batch 4800, loss[loss=0.1498, simple_loss=0.2307, pruned_loss=0.03449, over 4814.00 frames. ], tot_loss[loss=0.1643, simple_loss=0.2376, pruned_loss=0.04549, over 954159.73 frames. ], batch size: 25, lr: 2.86e-03, grad_scale: 32.0 2023-04-28 03:21:16,467 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.2760, 1.3249, 3.8820, 3.6139, 3.3752, 3.7463, 3.7520, 3.4425], device='cuda:1'), covar=tensor([0.7214, 0.5355, 0.1168, 0.1844, 0.1270, 0.1809, 0.1264, 0.1714], device='cuda:1'), in_proj_covar=tensor([0.0309, 0.0309, 0.0403, 0.0406, 0.0349, 0.0414, 0.0316, 0.0364], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-28 03:21:48,622 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.050e+02 1.679e+02 2.036e+02 2.333e+02 4.105e+02, threshold=4.072e+02, percent-clipped=2.0 2023-04-28 03:21:51,195 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=159481.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:21:59,476 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=159487.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:22:01,995 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=159491.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 03:22:08,783 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7384, 2.4999, 1.8868, 2.1907, 2.4618, 2.1318, 3.0826, 1.8207], device='cuda:1'), covar=tensor([0.3286, 0.2042, 0.4469, 0.2764, 0.1678, 0.2260, 0.1529, 0.4191], device='cuda:1'), in_proj_covar=tensor([0.0338, 0.0355, 0.0422, 0.0352, 0.0382, 0.0376, 0.0372, 0.0423], device='cuda:1'), out_proj_covar=tensor([9.9757e-05, 1.0569e-04, 1.2770e-04, 1.0524e-04, 1.1310e-04, 1.1142e-04, 1.0874e-04, 1.2717e-04], device='cuda:1') 2023-04-28 03:22:12,296 INFO [finetune.py:976] (1/7) Epoch 28, batch 4850, loss[loss=0.1827, simple_loss=0.2613, pruned_loss=0.05201, over 4804.00 frames. ], tot_loss[loss=0.165, simple_loss=0.239, pruned_loss=0.04553, over 954295.94 frames. ], batch size: 41, lr: 2.86e-03, grad_scale: 32.0 2023-04-28 03:22:40,591 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=159515.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:22:56,336 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 2023-04-28 03:23:05,841 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=159539.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 03:23:16,522 INFO [finetune.py:976] (1/7) Epoch 28, batch 4900, loss[loss=0.1436, simple_loss=0.2258, pruned_loss=0.03066, over 4929.00 frames. ], tot_loss[loss=0.1663, simple_loss=0.2404, pruned_loss=0.04609, over 955109.65 frames. ], batch size: 38, lr: 2.86e-03, grad_scale: 32.0 2023-04-28 03:23:56,948 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.158e+02 1.586e+02 1.930e+02 2.270e+02 5.838e+02, threshold=3.860e+02, percent-clipped=1.0 2023-04-28 03:24:05,186 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7427, 2.2735, 1.8012, 1.6700, 1.3041, 1.3278, 1.7915, 1.2472], device='cuda:1'), covar=tensor([0.1567, 0.1193, 0.1256, 0.1610, 0.2176, 0.1829, 0.0907, 0.2043], device='cuda:1'), in_proj_covar=tensor([0.0198, 0.0210, 0.0170, 0.0205, 0.0201, 0.0186, 0.0157, 0.0188], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-28 03:24:20,243 INFO [finetune.py:976] (1/7) Epoch 28, batch 4950, loss[loss=0.1721, simple_loss=0.2452, pruned_loss=0.0495, over 4723.00 frames. ], tot_loss[loss=0.1683, simple_loss=0.2426, pruned_loss=0.04703, over 954476.89 frames. ], batch size: 54, lr: 2.86e-03, grad_scale: 32.0 2023-04-28 03:25:08,571 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=159631.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:25:23,009 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=159646.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:25:24,156 INFO [finetune.py:976] (1/7) Epoch 28, batch 5000, loss[loss=0.1675, simple_loss=0.2395, pruned_loss=0.04778, over 4801.00 frames. ], tot_loss[loss=0.1672, simple_loss=0.2413, pruned_loss=0.0465, over 953941.90 frames. ], batch size: 25, lr: 2.86e-03, grad_scale: 32.0 2023-04-28 03:26:04,851 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-04-28 03:26:05,240 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.105e+02 1.569e+02 1.814e+02 2.278e+02 4.708e+02, threshold=3.628e+02, percent-clipped=1.0 2023-04-28 03:26:24,443 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=159692.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:26:32,516 INFO [finetune.py:976] (1/7) Epoch 28, batch 5050, loss[loss=0.1344, simple_loss=0.2034, pruned_loss=0.03269, over 4708.00 frames. ], tot_loss[loss=0.1645, simple_loss=0.2377, pruned_loss=0.04569, over 953620.87 frames. ], batch size: 23, lr: 2.86e-03, grad_scale: 32.0 2023-04-28 03:26:34,618 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 2023-04-28 03:26:58,588 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.2040, 2.6340, 1.2224, 1.4038, 2.3870, 1.2995, 3.7531, 1.8344], device='cuda:1'), covar=tensor([0.0651, 0.0696, 0.0781, 0.1239, 0.0443, 0.0987, 0.0196, 0.0596], device='cuda:1'), in_proj_covar=tensor([0.0050, 0.0064, 0.0047, 0.0046, 0.0049, 0.0051, 0.0073, 0.0050], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:1') 2023-04-28 03:27:06,784 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=159725.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:27:07,448 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7925, 2.3957, 1.7078, 1.7086, 1.3277, 1.3746, 1.7416, 1.2823], device='cuda:1'), covar=tensor([0.1923, 0.1359, 0.1647, 0.1816, 0.2467, 0.2253, 0.1068, 0.2249], device='cuda:1'), in_proj_covar=tensor([0.0198, 0.0210, 0.0170, 0.0205, 0.0201, 0.0187, 0.0157, 0.0189], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-28 03:27:35,606 INFO [finetune.py:976] (1/7) Epoch 28, batch 5100, loss[loss=0.1865, simple_loss=0.2515, pruned_loss=0.06079, over 4927.00 frames. ], tot_loss[loss=0.1615, simple_loss=0.2341, pruned_loss=0.04442, over 956103.97 frames. ], batch size: 33, lr: 2.86e-03, grad_scale: 32.0 2023-04-28 03:28:09,664 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=159773.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:28:13,002 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.052e+01 1.474e+02 1.729e+02 2.168e+02 4.974e+02, threshold=3.458e+02, percent-clipped=1.0 2023-04-28 03:28:29,606 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 2023-04-28 03:28:41,781 INFO [finetune.py:976] (1/7) Epoch 28, batch 5150, loss[loss=0.1566, simple_loss=0.2354, pruned_loss=0.03892, over 4766.00 frames. ], tot_loss[loss=0.1618, simple_loss=0.2344, pruned_loss=0.04458, over 956603.51 frames. ], batch size: 26, lr: 2.86e-03, grad_scale: 32.0 2023-04-28 03:29:03,550 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=159815.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:29:13,561 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-04-28 03:29:45,762 INFO [finetune.py:976] (1/7) Epoch 28, batch 5200, loss[loss=0.1858, simple_loss=0.2534, pruned_loss=0.05906, over 4864.00 frames. ], tot_loss[loss=0.1643, simple_loss=0.2377, pruned_loss=0.0455, over 957598.60 frames. ], batch size: 31, lr: 2.86e-03, grad_scale: 32.0 2023-04-28 03:29:55,482 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=159855.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:30:05,368 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=159863.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:30:26,725 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.215e+02 1.682e+02 1.955e+02 2.433e+02 4.346e+02, threshold=3.909e+02, percent-clipped=3.0 2023-04-28 03:30:51,955 INFO [finetune.py:976] (1/7) Epoch 28, batch 5250, loss[loss=0.1604, simple_loss=0.2396, pruned_loss=0.04065, over 4816.00 frames. ], tot_loss[loss=0.166, simple_loss=0.2398, pruned_loss=0.04608, over 953445.42 frames. ], batch size: 39, lr: 2.86e-03, grad_scale: 32.0 2023-04-28 03:31:08,801 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-04-28 03:31:09,953 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=159910.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:31:13,575 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7957, 1.3089, 4.1635, 3.9061, 3.6138, 3.8559, 3.8231, 3.6878], device='cuda:1'), covar=tensor([0.6694, 0.5469, 0.0953, 0.1531, 0.1144, 0.1499, 0.2665, 0.1459], device='cuda:1'), in_proj_covar=tensor([0.0310, 0.0310, 0.0406, 0.0409, 0.0351, 0.0416, 0.0319, 0.0365], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-28 03:31:13,622 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=159916.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:31:30,221 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([0.9642, 0.9826, 1.2160, 1.1436, 0.9788, 0.9515, 0.9190, 0.4890], device='cuda:1'), covar=tensor([0.0579, 0.0490, 0.0394, 0.0529, 0.0647, 0.0956, 0.0397, 0.0578], device='cuda:1'), in_proj_covar=tensor([0.0070, 0.0067, 0.0065, 0.0069, 0.0075, 0.0094, 0.0072, 0.0062], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-28 03:31:31,435 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4054, 1.1579, 3.6901, 3.1920, 3.2856, 3.4193, 3.4257, 3.0873], device='cuda:1'), covar=tensor([0.8777, 0.7870, 0.1822, 0.3170, 0.2247, 0.3218, 0.4071, 0.3204], device='cuda:1'), in_proj_covar=tensor([0.0310, 0.0310, 0.0406, 0.0409, 0.0351, 0.0416, 0.0319, 0.0365], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-28 03:31:55,411 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=159946.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:32:02,069 INFO [finetune.py:976] (1/7) Epoch 28, batch 5300, loss[loss=0.1998, simple_loss=0.264, pruned_loss=0.06778, over 4888.00 frames. ], tot_loss[loss=0.1684, simple_loss=0.2422, pruned_loss=0.04727, over 952547.31 frames. ], batch size: 32, lr: 2.86e-03, grad_scale: 32.0 2023-04-28 03:32:26,782 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=159971.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:32:35,450 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.404e+01 1.544e+02 1.829e+02 2.287e+02 5.054e+02, threshold=3.659e+02, percent-clipped=1.0 2023-04-28 03:32:48,723 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=159987.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:32:58,789 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=159994.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:33:06,690 INFO [finetune.py:976] (1/7) Epoch 28, batch 5350, loss[loss=0.1559, simple_loss=0.2286, pruned_loss=0.04167, over 4815.00 frames. ], tot_loss[loss=0.1685, simple_loss=0.2429, pruned_loss=0.04704, over 955048.12 frames. ], batch size: 33, lr: 2.86e-03, grad_scale: 32.0 2023-04-28 03:33:11,342 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.5516, 1.8104, 1.9171, 2.0107, 1.8532, 1.8535, 2.0177, 1.9278], device='cuda:1'), covar=tensor([0.3600, 0.5280, 0.4117, 0.4111, 0.5397, 0.7571, 0.4612, 0.4366], device='cuda:1'), in_proj_covar=tensor([0.0343, 0.0377, 0.0331, 0.0344, 0.0353, 0.0394, 0.0364, 0.0335], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 03:34:13,401 INFO [finetune.py:976] (1/7) Epoch 28, batch 5400, loss[loss=0.1229, simple_loss=0.1964, pruned_loss=0.02464, over 4803.00 frames. ], tot_loss[loss=0.1673, simple_loss=0.2408, pruned_loss=0.04689, over 956386.14 frames. ], batch size: 29, lr: 2.86e-03, grad_scale: 32.0 2023-04-28 03:34:24,960 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.1424, 1.9551, 2.3652, 2.5923, 2.2278, 2.0715, 2.2571, 2.2035], device='cuda:1'), covar=tensor([0.4520, 0.7126, 0.6688, 0.5172, 0.5845, 0.8637, 0.8478, 0.9746], device='cuda:1'), in_proj_covar=tensor([0.0445, 0.0425, 0.0520, 0.0509, 0.0474, 0.0510, 0.0512, 0.0527], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-28 03:34:46,559 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.246e+02 1.579e+02 1.912e+02 2.220e+02 4.270e+02, threshold=3.825e+02, percent-clipped=1.0 2023-04-28 03:35:18,039 INFO [finetune.py:976] (1/7) Epoch 28, batch 5450, loss[loss=0.172, simple_loss=0.2348, pruned_loss=0.05458, over 4770.00 frames. ], tot_loss[loss=0.1649, simple_loss=0.2378, pruned_loss=0.04602, over 954941.53 frames. ], batch size: 27, lr: 2.86e-03, grad_scale: 32.0 2023-04-28 03:36:22,774 INFO [finetune.py:976] (1/7) Epoch 28, batch 5500, loss[loss=0.141, simple_loss=0.2111, pruned_loss=0.03544, over 4830.00 frames. ], tot_loss[loss=0.1637, simple_loss=0.2358, pruned_loss=0.04578, over 954746.65 frames. ], batch size: 39, lr: 2.86e-03, grad_scale: 32.0 2023-04-28 03:37:00,971 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 7.993e+01 1.507e+02 1.787e+02 2.170e+02 4.329e+02, threshold=3.573e+02, percent-clipped=2.0 2023-04-28 03:37:01,111 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.1981, 2.7558, 2.1618, 2.2859, 1.6001, 1.6178, 2.2228, 1.5811], device='cuda:1'), covar=tensor([0.1478, 0.1329, 0.1335, 0.1449, 0.2166, 0.1784, 0.0936, 0.1892], device='cuda:1'), in_proj_covar=tensor([0.0197, 0.0208, 0.0169, 0.0204, 0.0200, 0.0185, 0.0156, 0.0187], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-28 03:37:04,207 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.4794, 2.0309, 2.3581, 2.7621, 2.3921, 1.8666, 1.5751, 2.2004], device='cuda:1'), covar=tensor([0.2780, 0.2814, 0.1429, 0.1948, 0.2376, 0.2422, 0.3806, 0.1766], device='cuda:1'), in_proj_covar=tensor([0.0290, 0.0246, 0.0227, 0.0313, 0.0222, 0.0234, 0.0228, 0.0186], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-28 03:37:27,649 INFO [finetune.py:976] (1/7) Epoch 28, batch 5550, loss[loss=0.1777, simple_loss=0.2571, pruned_loss=0.04909, over 4811.00 frames. ], tot_loss[loss=0.163, simple_loss=0.2355, pruned_loss=0.04523, over 953120.10 frames. ], batch size: 39, lr: 2.86e-03, grad_scale: 32.0 2023-04-28 03:37:41,099 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=160211.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:38:03,242 INFO [finetune.py:976] (1/7) Epoch 28, batch 5600, loss[loss=0.1929, simple_loss=0.2652, pruned_loss=0.06029, over 4833.00 frames. ], tot_loss[loss=0.166, simple_loss=0.239, pruned_loss=0.04652, over 952064.29 frames. ], batch size: 33, lr: 2.86e-03, grad_scale: 32.0 2023-04-28 03:38:13,712 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=160266.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:38:17,925 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9089, 1.4054, 1.7427, 1.6949, 1.7232, 1.3738, 0.8370, 1.3985], device='cuda:1'), covar=tensor([0.3062, 0.3104, 0.1664, 0.2011, 0.2360, 0.2611, 0.3786, 0.1892], device='cuda:1'), in_proj_covar=tensor([0.0290, 0.0245, 0.0226, 0.0312, 0.0222, 0.0234, 0.0227, 0.0185], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-28 03:38:20,147 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.023e+02 1.566e+02 2.005e+02 2.303e+02 3.818e+02, threshold=4.010e+02, percent-clipped=2.0 2023-04-28 03:38:24,379 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.67 vs. limit=5.0 2023-04-28 03:38:26,527 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=160287.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:38:32,860 INFO [finetune.py:976] (1/7) Epoch 28, batch 5650, loss[loss=0.2119, simple_loss=0.2738, pruned_loss=0.07497, over 4908.00 frames. ], tot_loss[loss=0.1675, simple_loss=0.2414, pruned_loss=0.04678, over 952851.23 frames. ], batch size: 37, lr: 2.86e-03, grad_scale: 32.0 2023-04-28 03:38:38,224 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=160307.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 03:38:54,999 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=160335.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:39:02,716 INFO [finetune.py:976] (1/7) Epoch 28, batch 5700, loss[loss=0.1343, simple_loss=0.1985, pruned_loss=0.03504, over 4177.00 frames. ], tot_loss[loss=0.1654, simple_loss=0.2381, pruned_loss=0.04635, over 934273.22 frames. ], batch size: 17, lr: 2.86e-03, grad_scale: 32.0 2023-04-28 03:39:10,416 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.3021, 2.7779, 1.7153, 2.1041, 2.7194, 2.2053, 2.0810, 2.2824], device='cuda:1'), covar=tensor([0.0420, 0.0271, 0.0227, 0.0445, 0.0205, 0.0411, 0.0389, 0.0469], device='cuda:1'), in_proj_covar=tensor([0.0028, 0.0023, 0.0022, 0.0028, 0.0019, 0.0028, 0.0027, 0.0028], device='cuda:1'), out_proj_covar=tensor([0.0052, 0.0046, 0.0039, 0.0053, 0.0039, 0.0051, 0.0051, 0.0053], device='cuda:1') 2023-04-28 03:39:14,623 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=160368.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 03:39:31,532 INFO [finetune.py:976] (1/7) Epoch 29, batch 0, loss[loss=0.1552, simple_loss=0.2384, pruned_loss=0.03598, over 4783.00 frames. ], tot_loss[loss=0.1552, simple_loss=0.2384, pruned_loss=0.03598, over 4783.00 frames. ], batch size: 29, lr: 2.86e-03, grad_scale: 32.0 2023-04-28 03:39:31,532 INFO [finetune.py:1001] (1/7) Computing validation loss 2023-04-28 03:39:42,779 INFO [finetune.py:1010] (1/7) Epoch 29, validation: loss=0.1546, simple_loss=0.2236, pruned_loss=0.04278, over 2265189.00 frames. 2023-04-28 03:39:42,779 INFO [finetune.py:1011] (1/7) Maximum memory allocated so far is 6391MB 2023-04-28 03:39:44,399 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.451e+01 1.469e+02 1.726e+02 2.023e+02 3.272e+02, threshold=3.452e+02, percent-clipped=0.0 2023-04-28 03:39:45,694 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.3431, 4.6069, 0.9849, 2.6542, 3.0062, 3.2173, 3.0016, 1.2560], device='cuda:1'), covar=tensor([0.1193, 0.0878, 0.2031, 0.1043, 0.0789, 0.0965, 0.1166, 0.1988], device='cuda:1'), in_proj_covar=tensor([0.0117, 0.0238, 0.0136, 0.0120, 0.0131, 0.0152, 0.0117, 0.0119], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-28 03:40:05,896 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.2981, 1.2301, 1.3343, 1.5513, 1.5565, 1.2565, 1.0078, 1.3890], device='cuda:1'), covar=tensor([0.0907, 0.1497, 0.0992, 0.0687, 0.0761, 0.0909, 0.0867, 0.0683], device='cuda:1'), in_proj_covar=tensor([0.0186, 0.0206, 0.0185, 0.0172, 0.0180, 0.0179, 0.0152, 0.0178], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-28 03:40:16,113 INFO [finetune.py:976] (1/7) Epoch 29, batch 50, loss[loss=0.1753, simple_loss=0.2531, pruned_loss=0.0488, over 4817.00 frames. ], tot_loss[loss=0.1715, simple_loss=0.247, pruned_loss=0.048, over 218756.77 frames. ], batch size: 38, lr: 2.85e-03, grad_scale: 32.0 2023-04-28 03:41:16,375 INFO [finetune.py:976] (1/7) Epoch 29, batch 100, loss[loss=0.1435, simple_loss=0.2251, pruned_loss=0.03098, over 4769.00 frames. ], tot_loss[loss=0.1627, simple_loss=0.2363, pruned_loss=0.04455, over 383185.07 frames. ], batch size: 26, lr: 2.85e-03, grad_scale: 16.0 2023-04-28 03:41:24,289 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.006e+02 1.575e+02 1.942e+02 2.371e+02 3.324e+02, threshold=3.883e+02, percent-clipped=0.0 2023-04-28 03:41:24,505 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.58 vs. limit=2.0 2023-04-28 03:41:47,181 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.3588, 2.0137, 2.3092, 2.7450, 2.6830, 2.2108, 2.0833, 2.5051], device='cuda:1'), covar=tensor([0.0695, 0.1094, 0.0597, 0.0481, 0.0598, 0.0696, 0.0587, 0.0438], device='cuda:1'), in_proj_covar=tensor([0.0184, 0.0204, 0.0184, 0.0171, 0.0178, 0.0178, 0.0151, 0.0176], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-28 03:42:06,919 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-04-28 03:42:06,950 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-04-28 03:42:07,341 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=160511.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:42:22,986 INFO [finetune.py:976] (1/7) Epoch 29, batch 150, loss[loss=0.166, simple_loss=0.2353, pruned_loss=0.04831, over 4936.00 frames. ], tot_loss[loss=0.1593, simple_loss=0.2316, pruned_loss=0.04347, over 511603.85 frames. ], batch size: 33, lr: 2.85e-03, grad_scale: 16.0 2023-04-28 03:43:05,313 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=160559.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:43:12,966 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([3.7431, 3.5988, 2.7911, 4.3097, 3.6785, 3.7285, 1.7245, 3.7135], device='cuda:1'), covar=tensor([0.1643, 0.1339, 0.3729, 0.1552, 0.3471, 0.1741, 0.5403, 0.2197], device='cuda:1'), in_proj_covar=tensor([0.0247, 0.0222, 0.0252, 0.0307, 0.0301, 0.0252, 0.0278, 0.0276], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 03:43:14,854 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=160566.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:43:27,102 INFO [finetune.py:976] (1/7) Epoch 29, batch 200, loss[loss=0.1752, simple_loss=0.2501, pruned_loss=0.05022, over 4811.00 frames. ], tot_loss[loss=0.1598, simple_loss=0.231, pruned_loss=0.04425, over 608620.21 frames. ], batch size: 41, lr: 2.85e-03, grad_scale: 16.0 2023-04-28 03:43:34,236 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.002e+02 1.562e+02 1.851e+02 2.230e+02 3.985e+02, threshold=3.702e+02, percent-clipped=1.0 2023-04-28 03:44:17,862 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=160614.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:44:30,031 INFO [finetune.py:976] (1/7) Epoch 29, batch 250, loss[loss=0.2339, simple_loss=0.2954, pruned_loss=0.08619, over 4893.00 frames. ], tot_loss[loss=0.1625, simple_loss=0.2346, pruned_loss=0.04526, over 687445.01 frames. ], batch size: 35, lr: 2.85e-03, grad_scale: 16.0 2023-04-28 03:45:20,095 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=160663.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 03:45:32,745 INFO [finetune.py:976] (1/7) Epoch 29, batch 300, loss[loss=0.1584, simple_loss=0.2259, pruned_loss=0.0454, over 4931.00 frames. ], tot_loss[loss=0.1659, simple_loss=0.2395, pruned_loss=0.04609, over 749352.66 frames. ], batch size: 38, lr: 2.85e-03, grad_scale: 16.0 2023-04-28 03:45:38,924 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.030e+02 1.518e+02 1.888e+02 2.303e+02 4.692e+02, threshold=3.776e+02, percent-clipped=1.0 2023-04-28 03:46:37,389 INFO [finetune.py:976] (1/7) Epoch 29, batch 350, loss[loss=0.1705, simple_loss=0.2436, pruned_loss=0.04873, over 4933.00 frames. ], tot_loss[loss=0.1673, simple_loss=0.2413, pruned_loss=0.04667, over 794724.31 frames. ], batch size: 29, lr: 2.85e-03, grad_scale: 16.0 2023-04-28 03:46:43,916 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 2023-04-28 03:47:07,146 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=3.31 vs. limit=5.0 2023-04-28 03:47:28,987 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.64 vs. limit=2.0 2023-04-28 03:47:45,907 INFO [finetune.py:976] (1/7) Epoch 29, batch 400, loss[loss=0.1852, simple_loss=0.2716, pruned_loss=0.04938, over 4814.00 frames. ], tot_loss[loss=0.1674, simple_loss=0.2415, pruned_loss=0.04658, over 827317.23 frames. ], batch size: 33, lr: 2.85e-03, grad_scale: 16.0 2023-04-28 03:47:47,780 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.194e+02 1.618e+02 1.920e+02 2.400e+02 5.071e+02, threshold=3.839e+02, percent-clipped=2.0 2023-04-28 03:48:31,273 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.70 vs. limit=2.0 2023-04-28 03:48:48,104 INFO [finetune.py:976] (1/7) Epoch 29, batch 450, loss[loss=0.138, simple_loss=0.2116, pruned_loss=0.03222, over 4815.00 frames. ], tot_loss[loss=0.1667, simple_loss=0.2409, pruned_loss=0.04628, over 857254.02 frames. ], batch size: 39, lr: 2.85e-03, grad_scale: 16.0 2023-04-28 03:48:50,684 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9371, 2.4186, 2.0128, 1.8512, 1.4081, 1.4670, 2.0195, 1.4167], device='cuda:1'), covar=tensor([0.1659, 0.1343, 0.1266, 0.1602, 0.2332, 0.1956, 0.0954, 0.2036], device='cuda:1'), in_proj_covar=tensor([0.0199, 0.0211, 0.0170, 0.0205, 0.0202, 0.0187, 0.0157, 0.0188], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-28 03:49:01,428 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7061, 1.7407, 0.8773, 1.4214, 1.9146, 1.5956, 1.4729, 1.5027], device='cuda:1'), covar=tensor([0.0453, 0.0335, 0.0315, 0.0506, 0.0257, 0.0445, 0.0434, 0.0497], device='cuda:1'), in_proj_covar=tensor([0.0028, 0.0023, 0.0022, 0.0028, 0.0019, 0.0028, 0.0027, 0.0028], device='cuda:1'), out_proj_covar=tensor([0.0052, 0.0046, 0.0039, 0.0053, 0.0039, 0.0051, 0.0051, 0.0053], device='cuda:1') 2023-04-28 03:49:19,309 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8497, 2.1731, 1.8377, 1.5859, 1.4133, 1.4256, 1.8906, 1.3548], device='cuda:1'), covar=tensor([0.1693, 0.1215, 0.1337, 0.1697, 0.2382, 0.1954, 0.0997, 0.1999], device='cuda:1'), in_proj_covar=tensor([0.0200, 0.0211, 0.0170, 0.0206, 0.0202, 0.0187, 0.0157, 0.0188], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-28 03:49:51,531 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-04-28 03:49:51,959 INFO [finetune.py:976] (1/7) Epoch 29, batch 500, loss[loss=0.1841, simple_loss=0.2572, pruned_loss=0.05548, over 4818.00 frames. ], tot_loss[loss=0.165, simple_loss=0.2386, pruned_loss=0.04574, over 880446.55 frames. ], batch size: 39, lr: 2.85e-03, grad_scale: 16.0 2023-04-28 03:49:53,853 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.077e+02 1.560e+02 1.871e+02 2.250e+02 5.218e+02, threshold=3.742e+02, percent-clipped=1.0 2023-04-28 03:50:23,517 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([3.8798, 3.8904, 2.8701, 4.5107, 3.9949, 3.8882, 1.6992, 3.7864], device='cuda:1'), covar=tensor([0.1721, 0.1348, 0.3049, 0.1482, 0.3042, 0.1614, 0.6218, 0.2525], device='cuda:1'), in_proj_covar=tensor([0.0249, 0.0223, 0.0254, 0.0307, 0.0303, 0.0254, 0.0280, 0.0278], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 03:50:25,293 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=160901.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:50:56,030 INFO [finetune.py:976] (1/7) Epoch 29, batch 550, loss[loss=0.1351, simple_loss=0.2126, pruned_loss=0.02882, over 4901.00 frames. ], tot_loss[loss=0.1619, simple_loss=0.235, pruned_loss=0.04442, over 897227.54 frames. ], batch size: 35, lr: 2.85e-03, grad_scale: 16.0 2023-04-28 03:51:43,473 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=160962.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:51:49,194 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=160963.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 03:51:50,417 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=160965.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:51:52,374 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.52 vs. limit=5.0 2023-04-28 03:52:01,370 INFO [finetune.py:976] (1/7) Epoch 29, batch 600, loss[loss=0.1627, simple_loss=0.2353, pruned_loss=0.04506, over 4934.00 frames. ], tot_loss[loss=0.1616, simple_loss=0.2345, pruned_loss=0.0444, over 910885.30 frames. ], batch size: 38, lr: 2.85e-03, grad_scale: 16.0 2023-04-28 03:52:03,148 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.040e+02 1.581e+02 1.979e+02 2.385e+02 4.353e+02, threshold=3.958e+02, percent-clipped=2.0 2023-04-28 03:52:35,724 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=161011.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 03:52:44,112 INFO [finetune.py:976] (1/7) Epoch 29, batch 650, loss[loss=0.1788, simple_loss=0.2498, pruned_loss=0.05393, over 4752.00 frames. ], tot_loss[loss=0.1629, simple_loss=0.2361, pruned_loss=0.04485, over 922067.31 frames. ], batch size: 27, lr: 2.85e-03, grad_scale: 16.0 2023-04-28 03:52:44,863 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=161026.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:53:17,108 INFO [finetune.py:976] (1/7) Epoch 29, batch 700, loss[loss=0.2045, simple_loss=0.2768, pruned_loss=0.06611, over 4811.00 frames. ], tot_loss[loss=0.1647, simple_loss=0.2382, pruned_loss=0.04562, over 928164.50 frames. ], batch size: 33, lr: 2.85e-03, grad_scale: 16.0 2023-04-28 03:53:18,919 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.470e+01 1.599e+02 1.884e+02 2.263e+02 4.345e+02, threshold=3.768e+02, percent-clipped=2.0 2023-04-28 03:53:50,553 INFO [finetune.py:976] (1/7) Epoch 29, batch 750, loss[loss=0.1803, simple_loss=0.2566, pruned_loss=0.05198, over 4856.00 frames. ], tot_loss[loss=0.1663, simple_loss=0.2402, pruned_loss=0.0462, over 934178.05 frames. ], batch size: 31, lr: 2.85e-03, grad_scale: 16.0 2023-04-28 03:54:00,519 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.3353, 1.3009, 1.3769, 1.6098, 1.6338, 1.2993, 1.0244, 1.5137], device='cuda:1'), covar=tensor([0.0760, 0.1357, 0.0872, 0.0537, 0.0598, 0.0739, 0.0759, 0.0542], device='cuda:1'), in_proj_covar=tensor([0.0184, 0.0203, 0.0183, 0.0170, 0.0177, 0.0177, 0.0150, 0.0176], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-28 03:54:12,589 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-04-28 03:54:24,403 INFO [finetune.py:976] (1/7) Epoch 29, batch 800, loss[loss=0.1498, simple_loss=0.2213, pruned_loss=0.03914, over 4762.00 frames. ], tot_loss[loss=0.1657, simple_loss=0.2399, pruned_loss=0.04569, over 940477.12 frames. ], batch size: 26, lr: 2.85e-03, grad_scale: 16.0 2023-04-28 03:54:26,204 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.992e+01 1.674e+02 1.972e+02 2.388e+02 4.465e+02, threshold=3.944e+02, percent-clipped=4.0 2023-04-28 03:54:52,566 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=161216.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:54:55,743 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=5.51 vs. limit=5.0 2023-04-28 03:54:57,898 INFO [finetune.py:976] (1/7) Epoch 29, batch 850, loss[loss=0.1599, simple_loss=0.2246, pruned_loss=0.04761, over 4820.00 frames. ], tot_loss[loss=0.1641, simple_loss=0.2381, pruned_loss=0.04511, over 943659.14 frames. ], batch size: 39, lr: 2.85e-03, grad_scale: 16.0 2023-04-28 03:55:01,668 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=161231.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:55:29,543 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=161257.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:55:38,786 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0310, 2.4523, 0.9981, 1.3059, 1.7767, 1.1577, 3.0331, 1.6160], device='cuda:1'), covar=tensor([0.0671, 0.0543, 0.0757, 0.1220, 0.0477, 0.1013, 0.0254, 0.0612], device='cuda:1'), in_proj_covar=tensor([0.0051, 0.0064, 0.0047, 0.0046, 0.0049, 0.0051, 0.0073, 0.0051], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:1') 2023-04-28 03:55:49,585 INFO [finetune.py:976] (1/7) Epoch 29, batch 900, loss[loss=0.1529, simple_loss=0.217, pruned_loss=0.04439, over 4897.00 frames. ], tot_loss[loss=0.1617, simple_loss=0.2351, pruned_loss=0.04417, over 946031.53 frames. ], batch size: 35, lr: 2.85e-03, grad_scale: 16.0 2023-04-28 03:55:50,945 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=161277.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:55:51,420 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.567e+01 1.487e+02 1.850e+02 2.194e+02 4.508e+02, threshold=3.700e+02, percent-clipped=1.0 2023-04-28 03:56:00,185 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=161292.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:56:03,248 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.3843, 1.7922, 1.7462, 1.8849, 1.7601, 1.8785, 1.8759, 1.7905], device='cuda:1'), covar=tensor([0.4107, 0.4711, 0.4248, 0.3980, 0.5289, 0.6776, 0.4689, 0.4525], device='cuda:1'), in_proj_covar=tensor([0.0343, 0.0375, 0.0332, 0.0343, 0.0351, 0.0394, 0.0363, 0.0335], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 03:56:03,834 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9270, 2.6283, 1.8734, 1.9658, 1.4574, 1.4772, 1.8720, 1.4380], device='cuda:1'), covar=tensor([0.1868, 0.1276, 0.1557, 0.1568, 0.2426, 0.2243, 0.1029, 0.2119], device='cuda:1'), in_proj_covar=tensor([0.0199, 0.0211, 0.0171, 0.0205, 0.0202, 0.0187, 0.0157, 0.0188], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-28 03:56:21,348 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=161321.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:56:23,750 INFO [finetune.py:976] (1/7) Epoch 29, batch 950, loss[loss=0.197, simple_loss=0.2678, pruned_loss=0.06304, over 4811.00 frames. ], tot_loss[loss=0.1613, simple_loss=0.2341, pruned_loss=0.04431, over 948840.19 frames. ], batch size: 45, lr: 2.85e-03, grad_scale: 16.0 2023-04-28 03:56:40,930 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7421, 1.3035, 1.4647, 1.4771, 1.8777, 1.5171, 1.3028, 1.4108], device='cuda:1'), covar=tensor([0.1889, 0.1648, 0.1866, 0.1402, 0.1090, 0.1695, 0.1797, 0.2585], device='cuda:1'), in_proj_covar=tensor([0.0316, 0.0309, 0.0351, 0.0288, 0.0328, 0.0307, 0.0301, 0.0377], device='cuda:1'), out_proj_covar=tensor([6.4413e-05, 6.3350e-05, 7.3468e-05, 5.7738e-05, 6.6839e-05, 6.3762e-05, 6.2195e-05, 7.9684e-05], device='cuda:1') 2023-04-28 03:57:07,673 INFO [finetune.py:976] (1/7) Epoch 29, batch 1000, loss[loss=0.1449, simple_loss=0.2365, pruned_loss=0.02662, over 4942.00 frames. ], tot_loss[loss=0.1634, simple_loss=0.2364, pruned_loss=0.04522, over 948966.78 frames. ], batch size: 33, lr: 2.85e-03, grad_scale: 16.0 2023-04-28 03:57:09,493 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.091e+02 1.605e+02 1.983e+02 2.337e+02 3.564e+02, threshold=3.965e+02, percent-clipped=0.0 2023-04-28 03:57:09,639 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.2800, 1.6977, 1.5276, 2.0871, 2.2646, 1.8937, 1.7933, 1.5798], device='cuda:1'), covar=tensor([0.1814, 0.1714, 0.2059, 0.1557, 0.1334, 0.1811, 0.2107, 0.2374], device='cuda:1'), in_proj_covar=tensor([0.0316, 0.0310, 0.0351, 0.0288, 0.0328, 0.0307, 0.0302, 0.0377], device='cuda:1'), out_proj_covar=tensor([6.4511e-05, 6.3445e-05, 7.3536e-05, 5.7750e-05, 6.6902e-05, 6.3818e-05, 6.2252e-05, 7.9723e-05], device='cuda:1') 2023-04-28 03:57:39,559 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=161401.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:57:42,135 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=3.92 vs. limit=5.0 2023-04-28 03:58:12,309 INFO [finetune.py:976] (1/7) Epoch 29, batch 1050, loss[loss=0.133, simple_loss=0.2208, pruned_loss=0.02262, over 4928.00 frames. ], tot_loss[loss=0.1646, simple_loss=0.2384, pruned_loss=0.04537, over 950602.00 frames. ], batch size: 42, lr: 2.85e-03, grad_scale: 16.0 2023-04-28 03:58:55,482 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=161460.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:58:56,732 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=161462.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:59:17,654 INFO [finetune.py:976] (1/7) Epoch 29, batch 1100, loss[loss=0.1813, simple_loss=0.2333, pruned_loss=0.06459, over 4426.00 frames. ], tot_loss[loss=0.1666, simple_loss=0.2402, pruned_loss=0.04648, over 951477.43 frames. ], batch size: 19, lr: 2.85e-03, grad_scale: 16.0 2023-04-28 03:59:20,410 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.123e+02 1.507e+02 1.843e+02 2.398e+02 4.910e+02, threshold=3.687e+02, percent-clipped=4.0 2023-04-28 03:59:28,765 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-04-28 04:00:20,264 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=161521.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:00:23,056 INFO [finetune.py:976] (1/7) Epoch 29, batch 1150, loss[loss=0.1855, simple_loss=0.2582, pruned_loss=0.05645, over 4778.00 frames. ], tot_loss[loss=0.1678, simple_loss=0.2415, pruned_loss=0.04703, over 953716.59 frames. ], batch size: 51, lr: 2.85e-03, grad_scale: 16.0 2023-04-28 04:01:04,601 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=161557.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:01:25,598 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=161572.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:01:27,867 INFO [finetune.py:976] (1/7) Epoch 29, batch 1200, loss[loss=0.1409, simple_loss=0.2073, pruned_loss=0.03727, over 4917.00 frames. ], tot_loss[loss=0.1674, simple_loss=0.2409, pruned_loss=0.04692, over 954815.27 frames. ], batch size: 46, lr: 2.85e-03, grad_scale: 16.0 2023-04-28 04:01:29,684 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.012e+02 1.552e+02 1.827e+02 2.258e+02 5.032e+02, threshold=3.654e+02, percent-clipped=3.0 2023-04-28 04:01:47,191 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=161587.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:02:09,297 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=161605.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:02:23,980 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=161621.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:02:33,463 INFO [finetune.py:976] (1/7) Epoch 29, batch 1250, loss[loss=0.1326, simple_loss=0.207, pruned_loss=0.0291, over 4890.00 frames. ], tot_loss[loss=0.1648, simple_loss=0.2378, pruned_loss=0.04588, over 953299.33 frames. ], batch size: 32, lr: 2.85e-03, grad_scale: 16.0 2023-04-28 04:02:40,893 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5317, 1.4700, 1.7663, 1.7985, 1.3536, 1.2817, 1.4901, 0.9021], device='cuda:1'), covar=tensor([0.0573, 0.0537, 0.0401, 0.0614, 0.0787, 0.0986, 0.0598, 0.0619], device='cuda:1'), in_proj_covar=tensor([0.0071, 0.0067, 0.0065, 0.0069, 0.0075, 0.0095, 0.0073, 0.0062], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-28 04:03:25,560 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.09 vs. limit=5.0 2023-04-28 04:03:28,496 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=161669.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:03:38,164 INFO [finetune.py:976] (1/7) Epoch 29, batch 1300, loss[loss=0.1991, simple_loss=0.2662, pruned_loss=0.06601, over 4931.00 frames. ], tot_loss[loss=0.1622, simple_loss=0.2348, pruned_loss=0.04485, over 953652.37 frames. ], batch size: 38, lr: 2.85e-03, grad_scale: 16.0 2023-04-28 04:03:46,263 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.807e+01 1.476e+02 1.722e+02 2.175e+02 4.011e+02, threshold=3.444e+02, percent-clipped=1.0 2023-04-28 04:03:49,909 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.3834, 1.7115, 1.6055, 1.8539, 1.8151, 1.9043, 1.6274, 3.0559], device='cuda:1'), covar=tensor([0.0560, 0.0653, 0.0663, 0.1016, 0.0513, 0.0600, 0.0637, 0.0198], device='cuda:1'), in_proj_covar=tensor([0.0037, 0.0038, 0.0039, 0.0043, 0.0040, 0.0037, 0.0038, 0.0054], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0013, 0.0013, 0.0015, 0.0014, 0.0013, 0.0013, 0.0016], device='cuda:1') 2023-04-28 04:04:42,648 INFO [finetune.py:976] (1/7) Epoch 29, batch 1350, loss[loss=0.1721, simple_loss=0.2559, pruned_loss=0.04409, over 4860.00 frames. ], tot_loss[loss=0.1631, simple_loss=0.2355, pruned_loss=0.04541, over 952272.61 frames. ], batch size: 44, lr: 2.85e-03, grad_scale: 16.0 2023-04-28 04:04:50,075 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5821, 3.6949, 1.0312, 2.1069, 2.0581, 2.6146, 2.2985, 1.1187], device='cuda:1'), covar=tensor([0.1331, 0.0946, 0.1777, 0.1070, 0.1021, 0.0962, 0.1333, 0.2105], device='cuda:1'), in_proj_covar=tensor([0.0117, 0.0236, 0.0135, 0.0119, 0.0131, 0.0152, 0.0117, 0.0118], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-28 04:05:27,901 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=161757.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:05:34,393 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.83 vs. limit=5.0 2023-04-28 04:05:50,110 INFO [finetune.py:976] (1/7) Epoch 29, batch 1400, loss[loss=0.18, simple_loss=0.2433, pruned_loss=0.05835, over 4766.00 frames. ], tot_loss[loss=0.1651, simple_loss=0.2382, pruned_loss=0.04606, over 952792.41 frames. ], batch size: 27, lr: 2.85e-03, grad_scale: 16.0 2023-04-28 04:05:58,008 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.100e+02 1.598e+02 1.895e+02 2.318e+02 6.343e+02, threshold=3.789e+02, percent-clipped=7.0 2023-04-28 04:06:33,095 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-04-28 04:06:51,783 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=161815.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:06:52,362 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=161816.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:07:03,300 INFO [finetune.py:976] (1/7) Epoch 29, batch 1450, loss[loss=0.1936, simple_loss=0.2709, pruned_loss=0.05813, over 4830.00 frames. ], tot_loss[loss=0.1651, simple_loss=0.2386, pruned_loss=0.04575, over 951564.93 frames. ], batch size: 49, lr: 2.85e-03, grad_scale: 16.0 2023-04-28 04:07:13,970 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-04-28 04:07:51,597 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6439, 1.7926, 0.9143, 1.3605, 2.0067, 1.5083, 1.3938, 1.4962], device='cuda:1'), covar=tensor([0.0496, 0.0366, 0.0323, 0.0569, 0.0249, 0.0517, 0.0493, 0.0576], device='cuda:1'), in_proj_covar=tensor([0.0028, 0.0023, 0.0021, 0.0028, 0.0019, 0.0027, 0.0027, 0.0028], device='cuda:1'), out_proj_covar=tensor([0.0052, 0.0046, 0.0039, 0.0053, 0.0039, 0.0051, 0.0051, 0.0053], device='cuda:1') 2023-04-28 04:08:00,384 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=161872.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:08:02,915 INFO [finetune.py:976] (1/7) Epoch 29, batch 1500, loss[loss=0.1879, simple_loss=0.2628, pruned_loss=0.05652, over 4801.00 frames. ], tot_loss[loss=0.1657, simple_loss=0.2395, pruned_loss=0.04598, over 953223.73 frames. ], batch size: 25, lr: 2.85e-03, grad_scale: 16.0 2023-04-28 04:08:03,665 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=161876.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:08:04,722 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.141e+02 1.641e+02 1.910e+02 2.350e+02 4.691e+02, threshold=3.820e+02, percent-clipped=1.0 2023-04-28 04:08:16,016 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=161887.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:09:01,030 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=161920.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:09:08,062 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.28 vs. limit=5.0 2023-04-28 04:09:09,620 INFO [finetune.py:976] (1/7) Epoch 29, batch 1550, loss[loss=0.1791, simple_loss=0.2532, pruned_loss=0.05243, over 4890.00 frames. ], tot_loss[loss=0.1649, simple_loss=0.2394, pruned_loss=0.04521, over 953795.16 frames. ], batch size: 35, lr: 2.85e-03, grad_scale: 16.0 2023-04-28 04:09:21,319 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=161935.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:09:30,605 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5753, 1.8493, 1.7616, 2.1499, 2.0863, 2.0191, 1.9068, 4.6113], device='cuda:1'), covar=tensor([0.0533, 0.0788, 0.0750, 0.1170, 0.0579, 0.0555, 0.0676, 0.0096], device='cuda:1'), in_proj_covar=tensor([0.0037, 0.0038, 0.0039, 0.0043, 0.0040, 0.0037, 0.0038, 0.0054], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0013, 0.0013, 0.0015, 0.0014, 0.0013, 0.0013, 0.0016], device='cuda:1') 2023-04-28 04:10:14,712 INFO [finetune.py:976] (1/7) Epoch 29, batch 1600, loss[loss=0.1874, simple_loss=0.247, pruned_loss=0.06393, over 4868.00 frames. ], tot_loss[loss=0.1643, simple_loss=0.2379, pruned_loss=0.0454, over 953762.35 frames. ], batch size: 34, lr: 2.84e-03, grad_scale: 16.0 2023-04-28 04:10:16,469 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 7.098e+01 1.531e+02 1.823e+02 2.197e+02 4.043e+02, threshold=3.646e+02, percent-clipped=1.0 2023-04-28 04:11:20,386 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.2738, 1.5510, 1.4819, 1.7744, 1.7392, 1.9525, 1.4863, 3.4228], device='cuda:1'), covar=tensor([0.0636, 0.0787, 0.0749, 0.1139, 0.0572, 0.0614, 0.0713, 0.0137], device='cuda:1'), in_proj_covar=tensor([0.0037, 0.0038, 0.0039, 0.0042, 0.0040, 0.0037, 0.0038, 0.0054], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0013, 0.0013, 0.0015, 0.0014, 0.0013, 0.0013, 0.0016], device='cuda:1') 2023-04-28 04:11:22,117 INFO [finetune.py:976] (1/7) Epoch 29, batch 1650, loss[loss=0.157, simple_loss=0.2234, pruned_loss=0.04528, over 4930.00 frames. ], tot_loss[loss=0.1628, simple_loss=0.2357, pruned_loss=0.04495, over 953876.51 frames. ], batch size: 33, lr: 2.84e-03, grad_scale: 16.0 2023-04-28 04:12:07,055 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=162057.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:12:29,334 INFO [finetune.py:976] (1/7) Epoch 29, batch 1700, loss[loss=0.1749, simple_loss=0.2437, pruned_loss=0.05306, over 4921.00 frames. ], tot_loss[loss=0.1596, simple_loss=0.2325, pruned_loss=0.04339, over 956330.53 frames. ], batch size: 37, lr: 2.84e-03, grad_scale: 16.0 2023-04-28 04:12:36,115 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=162077.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:12:36,610 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.915e+01 1.582e+02 1.975e+02 2.280e+02 6.731e+02, threshold=3.951e+02, percent-clipped=4.0 2023-04-28 04:12:36,738 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8650, 2.2782, 1.9376, 2.1832, 1.5692, 1.8444, 1.9073, 1.5217], device='cuda:1'), covar=tensor([0.1859, 0.1172, 0.0892, 0.1134, 0.3827, 0.1168, 0.1874, 0.2440], device='cuda:1'), in_proj_covar=tensor([0.0281, 0.0299, 0.0215, 0.0275, 0.0313, 0.0252, 0.0247, 0.0262], device='cuda:1'), out_proj_covar=tensor([1.1186e-04, 1.1746e-04, 8.4312e-05, 1.0800e-04, 1.2596e-04, 9.9013e-05, 9.9356e-05, 1.0326e-04], device='cuda:1') 2023-04-28 04:13:11,489 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=162105.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:13:24,511 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=162116.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:13:29,854 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([4.8102, 4.4985, 3.1336, 5.4809, 4.6371, 4.7812, 1.9716, 4.7307], device='cuda:1'), covar=tensor([0.1589, 0.0904, 0.3256, 0.0910, 0.3724, 0.1486, 0.5828, 0.2068], device='cuda:1'), in_proj_covar=tensor([0.0249, 0.0222, 0.0253, 0.0307, 0.0303, 0.0253, 0.0278, 0.0277], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 04:13:34,110 INFO [finetune.py:976] (1/7) Epoch 29, batch 1750, loss[loss=0.1621, simple_loss=0.2303, pruned_loss=0.04697, over 4829.00 frames. ], tot_loss[loss=0.1629, simple_loss=0.236, pruned_loss=0.04485, over 956938.64 frames. ], batch size: 25, lr: 2.84e-03, grad_scale: 16.0 2023-04-28 04:13:51,927 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=162138.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:13:53,132 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0303, 1.7161, 1.9153, 2.3778, 2.3679, 1.8287, 1.8528, 2.0226], device='cuda:1'), covar=tensor([0.0805, 0.1189, 0.0795, 0.0548, 0.0600, 0.0890, 0.0682, 0.0589], device='cuda:1'), in_proj_covar=tensor([0.0185, 0.0204, 0.0186, 0.0171, 0.0178, 0.0178, 0.0152, 0.0178], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-28 04:14:27,315 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=162164.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:14:37,342 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=162171.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:14:45,214 INFO [finetune.py:976] (1/7) Epoch 29, batch 1800, loss[loss=0.1802, simple_loss=0.2561, pruned_loss=0.05213, over 4935.00 frames. ], tot_loss[loss=0.1644, simple_loss=0.2381, pruned_loss=0.04536, over 957720.77 frames. ], batch size: 33, lr: 2.84e-03, grad_scale: 16.0 2023-04-28 04:14:47,027 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.098e+02 1.596e+02 1.871e+02 2.266e+02 6.327e+02, threshold=3.743e+02, percent-clipped=2.0 2023-04-28 04:15:08,194 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.1934, 1.4985, 1.3557, 1.5167, 1.2452, 1.3047, 1.2455, 1.0580], device='cuda:1'), covar=tensor([0.1736, 0.1284, 0.0936, 0.1176, 0.3779, 0.1151, 0.1951, 0.2197], device='cuda:1'), in_proj_covar=tensor([0.0283, 0.0301, 0.0217, 0.0277, 0.0314, 0.0253, 0.0249, 0.0264], device='cuda:1'), out_proj_covar=tensor([1.1280e-04, 1.1839e-04, 8.4972e-05, 1.0875e-04, 1.2665e-04, 9.9436e-05, 1.0002e-04, 1.0400e-04], device='cuda:1') 2023-04-28 04:15:49,792 INFO [finetune.py:976] (1/7) Epoch 29, batch 1850, loss[loss=0.1346, simple_loss=0.2019, pruned_loss=0.03364, over 4725.00 frames. ], tot_loss[loss=0.1661, simple_loss=0.2396, pruned_loss=0.04627, over 954159.70 frames. ], batch size: 23, lr: 2.84e-03, grad_scale: 16.0 2023-04-28 04:15:53,560 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=162231.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:16:53,967 INFO [finetune.py:976] (1/7) Epoch 29, batch 1900, loss[loss=0.1811, simple_loss=0.2499, pruned_loss=0.0562, over 4866.00 frames. ], tot_loss[loss=0.1668, simple_loss=0.2405, pruned_loss=0.04658, over 953322.97 frames. ], batch size: 34, lr: 2.84e-03, grad_scale: 16.0 2023-04-28 04:16:55,788 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.192e+01 1.518e+02 1.776e+02 2.128e+02 3.542e+02, threshold=3.553e+02, percent-clipped=0.0 2023-04-28 04:17:02,688 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6311, 1.2935, 4.6040, 4.3469, 4.0139, 4.4386, 4.3094, 4.0522], device='cuda:1'), covar=tensor([0.7510, 0.6138, 0.0892, 0.1434, 0.1138, 0.1778, 0.1326, 0.1427], device='cuda:1'), in_proj_covar=tensor([0.0309, 0.0308, 0.0405, 0.0408, 0.0349, 0.0413, 0.0316, 0.0362], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-28 04:17:14,632 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=162292.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:17:48,798 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7523, 2.0493, 1.7353, 1.3624, 1.3110, 1.3096, 1.7153, 1.2914], device='cuda:1'), covar=tensor([0.1535, 0.1287, 0.1410, 0.1660, 0.2247, 0.1902, 0.0929, 0.2047], device='cuda:1'), in_proj_covar=tensor([0.0197, 0.0208, 0.0170, 0.0203, 0.0200, 0.0186, 0.0156, 0.0186], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-28 04:17:58,314 INFO [finetune.py:976] (1/7) Epoch 29, batch 1950, loss[loss=0.1912, simple_loss=0.262, pruned_loss=0.06024, over 4804.00 frames. ], tot_loss[loss=0.1658, simple_loss=0.2397, pruned_loss=0.04597, over 955219.78 frames. ], batch size: 45, lr: 2.84e-03, grad_scale: 16.0 2023-04-28 04:17:58,386 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([3.6584, 3.5413, 2.6754, 4.2096, 3.6927, 3.6132, 1.5755, 3.5779], device='cuda:1'), covar=tensor([0.1804, 0.1281, 0.3060, 0.1783, 0.2844, 0.1866, 0.5963, 0.2561], device='cuda:1'), in_proj_covar=tensor([0.0249, 0.0222, 0.0253, 0.0307, 0.0303, 0.0253, 0.0277, 0.0277], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 04:19:03,738 INFO [finetune.py:976] (1/7) Epoch 29, batch 2000, loss[loss=0.1334, simple_loss=0.2094, pruned_loss=0.02875, over 4789.00 frames. ], tot_loss[loss=0.1636, simple_loss=0.237, pruned_loss=0.04516, over 953675.63 frames. ], batch size: 25, lr: 2.84e-03, grad_scale: 16.0 2023-04-28 04:19:05,551 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.225e+01 1.505e+02 1.783e+02 2.109e+02 5.594e+02, threshold=3.566e+02, percent-clipped=2.0 2023-04-28 04:19:12,320 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.3513, 3.0482, 0.8646, 1.8738, 1.8225, 2.1595, 1.8642, 0.8959], device='cuda:1'), covar=tensor([0.1468, 0.0879, 0.1921, 0.1120, 0.1005, 0.1030, 0.1426, 0.1888], device='cuda:1'), in_proj_covar=tensor([0.0117, 0.0236, 0.0135, 0.0119, 0.0131, 0.0152, 0.0116, 0.0117], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-28 04:19:19,812 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-04-28 04:19:43,204 INFO [finetune.py:976] (1/7) Epoch 29, batch 2050, loss[loss=0.1204, simple_loss=0.1976, pruned_loss=0.0216, over 4713.00 frames. ], tot_loss[loss=0.1615, simple_loss=0.2343, pruned_loss=0.04434, over 954112.41 frames. ], batch size: 23, lr: 2.84e-03, grad_scale: 16.0 2023-04-28 04:19:48,142 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=162433.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:19:52,020 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.29 vs. limit=5.0 2023-04-28 04:19:53,124 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.3991, 1.9312, 2.3434, 2.7412, 2.2688, 1.7876, 1.5123, 2.1078], device='cuda:1'), covar=tensor([0.3116, 0.3054, 0.1538, 0.1951, 0.2599, 0.2535, 0.4022, 0.1810], device='cuda:1'), in_proj_covar=tensor([0.0290, 0.0246, 0.0227, 0.0312, 0.0222, 0.0235, 0.0227, 0.0185], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-28 04:20:13,319 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=162471.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:20:16,687 INFO [finetune.py:976] (1/7) Epoch 29, batch 2100, loss[loss=0.1454, simple_loss=0.218, pruned_loss=0.03638, over 4902.00 frames. ], tot_loss[loss=0.1625, simple_loss=0.235, pruned_loss=0.04499, over 953615.12 frames. ], batch size: 32, lr: 2.84e-03, grad_scale: 32.0 2023-04-28 04:20:19,020 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 8.161e+01 1.530e+02 1.785e+02 2.249e+02 3.474e+02, threshold=3.570e+02, percent-clipped=1.0 2023-04-28 04:20:39,363 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=162511.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:20:45,709 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=162519.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:20:50,374 INFO [finetune.py:976] (1/7) Epoch 29, batch 2150, loss[loss=0.1397, simple_loss=0.2108, pruned_loss=0.03431, over 4821.00 frames. ], tot_loss[loss=0.1646, simple_loss=0.2377, pruned_loss=0.04573, over 954383.83 frames. ], batch size: 25, lr: 2.84e-03, grad_scale: 32.0 2023-04-28 04:21:00,776 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.97 vs. limit=2.0 2023-04-28 04:21:21,583 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=162572.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 04:21:23,763 INFO [finetune.py:976] (1/7) Epoch 29, batch 2200, loss[loss=0.1882, simple_loss=0.2646, pruned_loss=0.0559, over 4812.00 frames. ], tot_loss[loss=0.165, simple_loss=0.2391, pruned_loss=0.04545, over 956482.12 frames. ], batch size: 51, lr: 2.84e-03, grad_scale: 32.0 2023-04-28 04:21:26,031 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.398e+01 1.632e+02 1.891e+02 2.223e+02 3.490e+02, threshold=3.782e+02, percent-clipped=0.0 2023-04-28 04:21:29,673 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([0.9718, 1.0353, 1.1909, 1.1563, 0.9709, 0.9135, 1.0058, 0.5921], device='cuda:1'), covar=tensor([0.0525, 0.0489, 0.0409, 0.0504, 0.0688, 0.1014, 0.0428, 0.0564], device='cuda:1'), in_proj_covar=tensor([0.0071, 0.0067, 0.0065, 0.0069, 0.0075, 0.0094, 0.0072, 0.0062], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-28 04:21:37,939 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=162587.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:22:03,314 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([4.3797, 4.2285, 3.0188, 5.0279, 4.3923, 4.2891, 1.6309, 4.3184], device='cuda:1'), covar=tensor([0.1600, 0.1215, 0.3687, 0.0984, 0.2959, 0.1558, 0.6225, 0.2334], device='cuda:1'), in_proj_covar=tensor([0.0248, 0.0221, 0.0254, 0.0307, 0.0303, 0.0253, 0.0278, 0.0277], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 04:22:23,020 INFO [finetune.py:976] (1/7) Epoch 29, batch 2250, loss[loss=0.1454, simple_loss=0.2424, pruned_loss=0.02423, over 4753.00 frames. ], tot_loss[loss=0.1666, simple_loss=0.2413, pruned_loss=0.04595, over 957098.45 frames. ], batch size: 28, lr: 2.84e-03, grad_scale: 32.0 2023-04-28 04:22:23,134 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6751, 1.6882, 0.6640, 1.3750, 1.8469, 1.5375, 1.4524, 1.5365], device='cuda:1'), covar=tensor([0.0495, 0.0358, 0.0350, 0.0531, 0.0280, 0.0492, 0.0469, 0.0561], device='cuda:1'), in_proj_covar=tensor([0.0028, 0.0023, 0.0022, 0.0028, 0.0019, 0.0028, 0.0027, 0.0028], device='cuda:1'), out_proj_covar=tensor([0.0052, 0.0047, 0.0039, 0.0054, 0.0039, 0.0051, 0.0051, 0.0053], device='cuda:1') 2023-04-28 04:23:28,881 INFO [finetune.py:976] (1/7) Epoch 29, batch 2300, loss[loss=0.1765, simple_loss=0.2514, pruned_loss=0.05076, over 4817.00 frames. ], tot_loss[loss=0.1679, simple_loss=0.243, pruned_loss=0.0464, over 957126.05 frames. ], batch size: 38, lr: 2.84e-03, grad_scale: 32.0 2023-04-28 04:23:36,409 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 8.802e+01 1.609e+02 1.813e+02 2.047e+02 3.617e+02, threshold=3.626e+02, percent-clipped=0.0 2023-04-28 04:24:17,788 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([0.8656, 1.1179, 3.2768, 3.0359, 2.9518, 3.1930, 3.1818, 2.8737], device='cuda:1'), covar=tensor([0.7505, 0.5560, 0.1490, 0.2243, 0.1439, 0.1818, 0.1856, 0.1747], device='cuda:1'), in_proj_covar=tensor([0.0307, 0.0307, 0.0404, 0.0406, 0.0348, 0.0413, 0.0316, 0.0361], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-28 04:24:31,808 INFO [finetune.py:976] (1/7) Epoch 29, batch 2350, loss[loss=0.1717, simple_loss=0.2411, pruned_loss=0.05111, over 4926.00 frames. ], tot_loss[loss=0.1666, simple_loss=0.2407, pruned_loss=0.04622, over 958156.12 frames. ], batch size: 33, lr: 2.84e-03, grad_scale: 32.0 2023-04-28 04:24:41,990 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=162733.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:25:01,097 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-04-28 04:25:30,907 INFO [finetune.py:976] (1/7) Epoch 29, batch 2400, loss[loss=0.1411, simple_loss=0.2063, pruned_loss=0.03798, over 4761.00 frames. ], tot_loss[loss=0.1641, simple_loss=0.2376, pruned_loss=0.04527, over 958568.08 frames. ], batch size: 28, lr: 2.84e-03, grad_scale: 32.0 2023-04-28 04:25:32,703 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5978, 1.7331, 1.4818, 1.1935, 1.2197, 1.1965, 1.4435, 1.1540], device='cuda:1'), covar=tensor([0.1559, 0.1188, 0.1394, 0.1540, 0.2052, 0.1843, 0.0916, 0.1876], device='cuda:1'), in_proj_covar=tensor([0.0198, 0.0209, 0.0170, 0.0203, 0.0200, 0.0186, 0.0156, 0.0186], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-28 04:25:33,146 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.191e+01 1.527e+02 1.810e+02 2.223e+02 4.938e+02, threshold=3.619e+02, percent-clipped=3.0 2023-04-28 04:25:36,108 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=162781.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:25:51,304 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4140, 1.3259, 1.5990, 1.6429, 1.2640, 1.2082, 1.3378, 0.8122], device='cuda:1'), covar=tensor([0.0517, 0.0544, 0.0375, 0.0475, 0.0633, 0.1001, 0.0461, 0.0504], device='cuda:1'), in_proj_covar=tensor([0.0071, 0.0067, 0.0065, 0.0069, 0.0075, 0.0094, 0.0072, 0.0062], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-28 04:26:04,085 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8550, 1.4722, 1.4770, 1.7107, 2.1081, 1.7116, 1.4685, 1.3658], device='cuda:1'), covar=tensor([0.1771, 0.1542, 0.2038, 0.1271, 0.0863, 0.1700, 0.2053, 0.2538], device='cuda:1'), in_proj_covar=tensor([0.0315, 0.0309, 0.0350, 0.0286, 0.0326, 0.0305, 0.0301, 0.0377], device='cuda:1'), out_proj_covar=tensor([6.4365e-05, 6.3260e-05, 7.3154e-05, 5.7269e-05, 6.6318e-05, 6.3381e-05, 6.2159e-05, 7.9589e-05], device='cuda:1') 2023-04-28 04:26:04,585 INFO [finetune.py:976] (1/7) Epoch 29, batch 2450, loss[loss=0.161, simple_loss=0.232, pruned_loss=0.04496, over 4837.00 frames. ], tot_loss[loss=0.1624, simple_loss=0.2351, pruned_loss=0.04487, over 955542.32 frames. ], batch size: 30, lr: 2.84e-03, grad_scale: 32.0 2023-04-28 04:26:05,928 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=162827.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:26:33,771 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=162867.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 04:26:38,578 INFO [finetune.py:976] (1/7) Epoch 29, batch 2500, loss[loss=0.1855, simple_loss=0.2522, pruned_loss=0.05937, over 4858.00 frames. ], tot_loss[loss=0.1627, simple_loss=0.2357, pruned_loss=0.04489, over 954807.32 frames. ], batch size: 31, lr: 2.84e-03, grad_scale: 32.0 2023-04-28 04:26:40,363 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.423e+01 1.461e+02 1.777e+02 2.062e+02 3.489e+02, threshold=3.555e+02, percent-clipped=0.0 2023-04-28 04:26:47,950 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=162887.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:26:48,602 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=162888.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:26:49,171 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8047, 4.0417, 0.8982, 2.1683, 2.2932, 2.6172, 2.4640, 0.9947], device='cuda:1'), covar=tensor([0.1401, 0.0869, 0.1993, 0.1205, 0.0945, 0.1114, 0.1413, 0.2174], device='cuda:1'), in_proj_covar=tensor([0.0118, 0.0238, 0.0136, 0.0120, 0.0132, 0.0153, 0.0117, 0.0119], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-28 04:27:12,365 INFO [finetune.py:976] (1/7) Epoch 29, batch 2550, loss[loss=0.1998, simple_loss=0.2717, pruned_loss=0.06394, over 4897.00 frames. ], tot_loss[loss=0.1653, simple_loss=0.2392, pruned_loss=0.0457, over 955134.78 frames. ], batch size: 35, lr: 2.84e-03, grad_scale: 32.0 2023-04-28 04:27:19,033 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=162935.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:27:24,172 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6072, 1.5562, 0.9222, 1.3683, 1.6232, 1.5030, 1.4077, 1.4763], device='cuda:1'), covar=tensor([0.0423, 0.0327, 0.0378, 0.0476, 0.0318, 0.0424, 0.0405, 0.0479], device='cuda:1'), in_proj_covar=tensor([0.0028, 0.0023, 0.0022, 0.0028, 0.0019, 0.0028, 0.0027, 0.0028], device='cuda:1'), out_proj_covar=tensor([0.0052, 0.0047, 0.0039, 0.0054, 0.0039, 0.0051, 0.0051, 0.0053], device='cuda:1') 2023-04-28 04:27:46,026 INFO [finetune.py:976] (1/7) Epoch 29, batch 2600, loss[loss=0.1847, simple_loss=0.2572, pruned_loss=0.05613, over 4808.00 frames. ], tot_loss[loss=0.1674, simple_loss=0.2417, pruned_loss=0.04659, over 954620.66 frames. ], batch size: 40, lr: 2.84e-03, grad_scale: 32.0 2023-04-28 04:27:47,806 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.782e+01 1.602e+02 1.933e+02 2.328e+02 3.675e+02, threshold=3.867e+02, percent-clipped=1.0 2023-04-28 04:28:13,566 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.0993, 0.7317, 0.9672, 0.7501, 1.1781, 0.9730, 0.8306, 1.0032], device='cuda:1'), covar=tensor([0.1606, 0.1490, 0.1793, 0.1376, 0.0972, 0.1361, 0.1704, 0.2244], device='cuda:1'), in_proj_covar=tensor([0.0315, 0.0308, 0.0349, 0.0285, 0.0324, 0.0305, 0.0300, 0.0376], device='cuda:1'), out_proj_covar=tensor([6.4319e-05, 6.3077e-05, 7.3058e-05, 5.7056e-05, 6.6005e-05, 6.3243e-05, 6.1978e-05, 7.9430e-05], device='cuda:1') 2023-04-28 04:28:19,510 INFO [finetune.py:976] (1/7) Epoch 29, batch 2650, loss[loss=0.1637, simple_loss=0.2475, pruned_loss=0.03991, over 4907.00 frames. ], tot_loss[loss=0.1683, simple_loss=0.2433, pruned_loss=0.04662, over 956393.01 frames. ], batch size: 37, lr: 2.84e-03, grad_scale: 32.0 2023-04-28 04:28:22,061 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9592, 1.7972, 1.9548, 2.2265, 2.3477, 1.9077, 1.7754, 2.0573], device='cuda:1'), covar=tensor([0.0715, 0.0981, 0.0603, 0.0479, 0.0506, 0.0761, 0.0615, 0.0465], device='cuda:1'), in_proj_covar=tensor([0.0184, 0.0202, 0.0183, 0.0170, 0.0176, 0.0177, 0.0150, 0.0176], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-28 04:28:36,809 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6883, 1.4172, 4.4094, 4.1447, 3.8371, 4.2172, 4.1429, 3.8654], device='cuda:1'), covar=tensor([0.7462, 0.5721, 0.1150, 0.1707, 0.1175, 0.1501, 0.1423, 0.1512], device='cuda:1'), in_proj_covar=tensor([0.0307, 0.0306, 0.0403, 0.0406, 0.0347, 0.0412, 0.0315, 0.0360], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-28 04:28:57,567 INFO [finetune.py:976] (1/7) Epoch 29, batch 2700, loss[loss=0.1778, simple_loss=0.2392, pruned_loss=0.05823, over 4816.00 frames. ], tot_loss[loss=0.167, simple_loss=0.2418, pruned_loss=0.04609, over 952652.81 frames. ], batch size: 33, lr: 2.84e-03, grad_scale: 32.0 2023-04-28 04:29:04,329 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.042e+02 1.516e+02 1.799e+02 2.193e+02 4.304e+02, threshold=3.599e+02, percent-clipped=1.0 2023-04-28 04:29:17,295 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=163090.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:29:51,557 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.2876, 1.6559, 1.5773, 1.8420, 1.7871, 2.0099, 1.5263, 3.7744], device='cuda:1'), covar=tensor([0.0595, 0.0755, 0.0735, 0.1141, 0.0598, 0.0469, 0.0674, 0.0125], device='cuda:1'), in_proj_covar=tensor([0.0037, 0.0038, 0.0039, 0.0042, 0.0040, 0.0037, 0.0037, 0.0054], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0013, 0.0013, 0.0015, 0.0014, 0.0013, 0.0013, 0.0016], device='cuda:1') 2023-04-28 04:30:02,215 INFO [finetune.py:976] (1/7) Epoch 29, batch 2750, loss[loss=0.164, simple_loss=0.2313, pruned_loss=0.04832, over 4813.00 frames. ], tot_loss[loss=0.1647, simple_loss=0.2385, pruned_loss=0.04548, over 950992.03 frames. ], batch size: 41, lr: 2.84e-03, grad_scale: 32.0 2023-04-28 04:30:08,864 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.3824, 1.9044, 2.3105, 2.6883, 2.3030, 1.8344, 1.4770, 2.1504], device='cuda:1'), covar=tensor([0.3161, 0.2992, 0.1685, 0.2307, 0.2321, 0.2638, 0.3940, 0.1803], device='cuda:1'), in_proj_covar=tensor([0.0292, 0.0246, 0.0228, 0.0313, 0.0222, 0.0235, 0.0227, 0.0185], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-28 04:30:38,977 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=163151.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:30:39,571 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9242, 2.2938, 0.9558, 1.2688, 1.6530, 1.1229, 2.5375, 1.4445], device='cuda:1'), covar=tensor([0.0616, 0.0462, 0.0578, 0.1173, 0.0387, 0.0950, 0.0261, 0.0617], device='cuda:1'), in_proj_covar=tensor([0.0050, 0.0064, 0.0047, 0.0046, 0.0049, 0.0051, 0.0072, 0.0050], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:1') 2023-04-28 04:30:39,611 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.1102, 0.8030, 0.9263, 0.8695, 1.2376, 0.9933, 0.9312, 0.9734], device='cuda:1'), covar=tensor([0.1890, 0.1606, 0.2209, 0.1679, 0.1176, 0.1611, 0.1671, 0.2445], device='cuda:1'), in_proj_covar=tensor([0.0314, 0.0307, 0.0348, 0.0284, 0.0323, 0.0304, 0.0299, 0.0374], device='cuda:1'), out_proj_covar=tensor([6.4102e-05, 6.2806e-05, 7.2775e-05, 5.6794e-05, 6.5805e-05, 6.3117e-05, 6.1720e-05, 7.9068e-05], device='cuda:1') 2023-04-28 04:30:50,677 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=163167.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 04:31:01,088 INFO [finetune.py:976] (1/7) Epoch 29, batch 2800, loss[loss=0.1309, simple_loss=0.2167, pruned_loss=0.02255, over 4934.00 frames. ], tot_loss[loss=0.1624, simple_loss=0.2356, pruned_loss=0.04458, over 952976.02 frames. ], batch size: 38, lr: 2.84e-03, grad_scale: 32.0 2023-04-28 04:31:08,097 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.000e+02 1.442e+02 1.723e+02 2.053e+02 5.173e+02, threshold=3.446e+02, percent-clipped=1.0 2023-04-28 04:31:11,255 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=163183.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:31:52,459 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=163215.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:31:59,015 INFO [finetune.py:976] (1/7) Epoch 29, batch 2850, loss[loss=0.1387, simple_loss=0.213, pruned_loss=0.03221, over 4766.00 frames. ], tot_loss[loss=0.1603, simple_loss=0.2333, pruned_loss=0.04367, over 953127.32 frames. ], batch size: 27, lr: 2.84e-03, grad_scale: 32.0 2023-04-28 04:32:08,288 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=163240.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:32:31,404 INFO [finetune.py:976] (1/7) Epoch 29, batch 2900, loss[loss=0.2085, simple_loss=0.2955, pruned_loss=0.06071, over 4905.00 frames. ], tot_loss[loss=0.1652, simple_loss=0.2385, pruned_loss=0.04592, over 952617.94 frames. ], batch size: 43, lr: 2.84e-03, grad_scale: 32.0 2023-04-28 04:32:33,726 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.059e+02 1.435e+02 1.823e+02 2.184e+02 4.986e+02, threshold=3.647e+02, percent-clipped=1.0 2023-04-28 04:32:35,719 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9517, 1.4105, 1.7274, 1.7158, 1.7484, 1.3936, 0.8254, 1.3962], device='cuda:1'), covar=tensor([0.2819, 0.3056, 0.1645, 0.1928, 0.2147, 0.2551, 0.4021, 0.1892], device='cuda:1'), in_proj_covar=tensor([0.0293, 0.0246, 0.0228, 0.0313, 0.0222, 0.0235, 0.0228, 0.0186], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-28 04:32:47,881 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=163301.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:33:04,178 INFO [finetune.py:976] (1/7) Epoch 29, batch 2950, loss[loss=0.1573, simple_loss=0.2348, pruned_loss=0.03985, over 4840.00 frames. ], tot_loss[loss=0.1673, simple_loss=0.2412, pruned_loss=0.04664, over 954255.48 frames. ], batch size: 47, lr: 2.84e-03, grad_scale: 32.0 2023-04-28 04:33:09,605 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5497, 1.6427, 0.7566, 1.2525, 1.7861, 1.4222, 1.3196, 1.4126], device='cuda:1'), covar=tensor([0.0478, 0.0352, 0.0348, 0.0541, 0.0260, 0.0499, 0.0477, 0.0534], device='cuda:1'), in_proj_covar=tensor([0.0028, 0.0023, 0.0022, 0.0028, 0.0019, 0.0027, 0.0027, 0.0028], device='cuda:1'), out_proj_covar=tensor([0.0052, 0.0046, 0.0039, 0.0053, 0.0039, 0.0051, 0.0051, 0.0053], device='cuda:1') 2023-04-28 04:33:37,002 INFO [finetune.py:976] (1/7) Epoch 29, batch 3000, loss[loss=0.2053, simple_loss=0.2777, pruned_loss=0.06647, over 4840.00 frames. ], tot_loss[loss=0.1674, simple_loss=0.2416, pruned_loss=0.04667, over 953428.40 frames. ], batch size: 47, lr: 2.84e-03, grad_scale: 32.0 2023-04-28 04:33:37,002 INFO [finetune.py:1001] (1/7) Computing validation loss 2023-04-28 04:33:39,441 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([0.9152, 1.9011, 1.7720, 1.5260, 1.9753, 1.6559, 2.2918, 1.5748], device='cuda:1'), covar=tensor([0.3346, 0.1829, 0.5122, 0.2828, 0.1397, 0.2166, 0.1575, 0.4691], device='cuda:1'), in_proj_covar=tensor([0.0340, 0.0354, 0.0422, 0.0355, 0.0385, 0.0377, 0.0372, 0.0425], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-28 04:33:40,610 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.2785, 1.6413, 1.4813, 1.8014, 1.7148, 1.7474, 1.4715, 3.0984], device='cuda:1'), covar=tensor([0.0607, 0.0738, 0.0700, 0.1179, 0.0577, 0.0422, 0.0688, 0.0169], device='cuda:1'), in_proj_covar=tensor([0.0037, 0.0038, 0.0039, 0.0042, 0.0040, 0.0037, 0.0037, 0.0054], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0013, 0.0013, 0.0015, 0.0014, 0.0013, 0.0013, 0.0016], device='cuda:1') 2023-04-28 04:33:43,891 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7465, 1.0902, 1.6540, 2.2538, 1.8417, 1.6579, 1.6522, 1.6799], device='cuda:1'), covar=tensor([0.4454, 0.6821, 0.6465, 0.5352, 0.5753, 0.8164, 0.7876, 0.8882], device='cuda:1'), in_proj_covar=tensor([0.0448, 0.0424, 0.0520, 0.0508, 0.0475, 0.0510, 0.0514, 0.0530], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-28 04:33:47,841 INFO [finetune.py:1010] (1/7) Epoch 29, validation: loss=0.1535, simple_loss=0.222, pruned_loss=0.04252, over 2265189.00 frames. 2023-04-28 04:33:47,842 INFO [finetune.py:1011] (1/7) Maximum memory allocated so far is 6391MB 2023-04-28 04:33:49,648 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.160e+02 1.593e+02 1.913e+02 2.268e+02 4.606e+02, threshold=3.825e+02, percent-clipped=1.0 2023-04-28 04:34:11,697 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.4364, 2.0265, 2.3709, 2.8397, 2.3318, 1.9040, 1.6758, 2.1693], device='cuda:1'), covar=tensor([0.3281, 0.2761, 0.1602, 0.2137, 0.2439, 0.2603, 0.3659, 0.1743], device='cuda:1'), in_proj_covar=tensor([0.0291, 0.0244, 0.0227, 0.0311, 0.0221, 0.0234, 0.0227, 0.0185], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-28 04:34:14,181 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-04-28 04:34:19,348 INFO [finetune.py:976] (1/7) Epoch 29, batch 3050, loss[loss=0.1816, simple_loss=0.2492, pruned_loss=0.05706, over 4868.00 frames. ], tot_loss[loss=0.1661, simple_loss=0.2408, pruned_loss=0.04566, over 954004.79 frames. ], batch size: 34, lr: 2.84e-03, grad_scale: 32.0 2023-04-28 04:34:33,250 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.81 vs. limit=5.0 2023-04-28 04:34:34,883 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=163446.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:34:53,038 INFO [finetune.py:976] (1/7) Epoch 29, batch 3100, loss[loss=0.1815, simple_loss=0.2371, pruned_loss=0.06299, over 4892.00 frames. ], tot_loss[loss=0.1652, simple_loss=0.2395, pruned_loss=0.04545, over 954023.02 frames. ], batch size: 32, lr: 2.84e-03, grad_scale: 32.0 2023-04-28 04:34:55,820 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.898e+01 1.464e+02 1.707e+02 2.164e+02 5.622e+02, threshold=3.413e+02, percent-clipped=1.0 2023-04-28 04:34:59,458 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=163483.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:35:27,054 INFO [finetune.py:976] (1/7) Epoch 29, batch 3150, loss[loss=0.1641, simple_loss=0.2268, pruned_loss=0.05066, over 4814.00 frames. ], tot_loss[loss=0.1641, simple_loss=0.2377, pruned_loss=0.04528, over 956977.40 frames. ], batch size: 38, lr: 2.84e-03, grad_scale: 32.0 2023-04-28 04:35:31,735 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=163531.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:36:10,391 INFO [finetune.py:976] (1/7) Epoch 29, batch 3200, loss[loss=0.1958, simple_loss=0.2641, pruned_loss=0.06372, over 4244.00 frames. ], tot_loss[loss=0.1625, simple_loss=0.2351, pruned_loss=0.04496, over 956214.02 frames. ], batch size: 65, lr: 2.83e-03, grad_scale: 32.0 2023-04-28 04:36:12,739 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.620e+01 1.506e+02 1.753e+02 2.068e+02 8.624e+02, threshold=3.506e+02, percent-clipped=5.0 2023-04-28 04:36:42,502 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=163596.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:36:43,787 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([3.1885, 3.7158, 3.2029, 3.5254, 2.7988, 3.2346, 3.3600, 2.7291], device='cuda:1'), covar=tensor([0.1318, 0.0930, 0.0613, 0.0781, 0.2369, 0.0798, 0.1227, 0.1904], device='cuda:1'), in_proj_covar=tensor([0.0281, 0.0297, 0.0215, 0.0274, 0.0311, 0.0251, 0.0245, 0.0262], device='cuda:1'), out_proj_covar=tensor([1.1199e-04, 1.1675e-04, 8.4473e-05, 1.0755e-04, 1.2529e-04, 9.8608e-05, 9.8767e-05, 1.0313e-04], device='cuda:1') 2023-04-28 04:37:17,128 INFO [finetune.py:976] (1/7) Epoch 29, batch 3250, loss[loss=0.1735, simple_loss=0.2587, pruned_loss=0.04413, over 4843.00 frames. ], tot_loss[loss=0.163, simple_loss=0.236, pruned_loss=0.04498, over 955557.08 frames. ], batch size: 49, lr: 2.83e-03, grad_scale: 32.0 2023-04-28 04:37:51,856 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.2205, 1.3807, 1.1831, 1.9221, 1.5352, 1.8837, 1.3138, 4.0936], device='cuda:1'), covar=tensor([0.0751, 0.1109, 0.1155, 0.1395, 0.0864, 0.0692, 0.1090, 0.0258], device='cuda:1'), in_proj_covar=tensor([0.0037, 0.0038, 0.0039, 0.0042, 0.0040, 0.0037, 0.0038, 0.0054], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0013, 0.0014, 0.0015, 0.0014, 0.0013, 0.0013, 0.0016], device='cuda:1') 2023-04-28 04:38:20,406 INFO [finetune.py:976] (1/7) Epoch 29, batch 3300, loss[loss=0.1903, simple_loss=0.266, pruned_loss=0.05731, over 4936.00 frames. ], tot_loss[loss=0.1671, simple_loss=0.2408, pruned_loss=0.04671, over 955341.24 frames. ], batch size: 38, lr: 2.83e-03, grad_scale: 32.0 2023-04-28 04:38:22,207 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.182e+02 1.605e+02 1.841e+02 2.298e+02 3.971e+02, threshold=3.681e+02, percent-clipped=3.0 2023-04-28 04:38:22,934 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4014, 3.3060, 1.0141, 1.6486, 1.9813, 2.3657, 1.8731, 1.0662], device='cuda:1'), covar=tensor([0.1591, 0.0996, 0.2055, 0.1488, 0.1061, 0.1119, 0.1611, 0.1925], device='cuda:1'), in_proj_covar=tensor([0.0119, 0.0240, 0.0137, 0.0121, 0.0133, 0.0154, 0.0118, 0.0120], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-28 04:39:22,574 INFO [finetune.py:976] (1/7) Epoch 29, batch 3350, loss[loss=0.1397, simple_loss=0.2177, pruned_loss=0.03082, over 4221.00 frames. ], tot_loss[loss=0.1679, simple_loss=0.2418, pruned_loss=0.04702, over 954102.14 frames. ], batch size: 18, lr: 2.83e-03, grad_scale: 32.0 2023-04-28 04:39:47,013 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=163746.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:40:21,578 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([4.9992, 4.8464, 3.1372, 5.5953, 4.8599, 4.8581, 2.2932, 4.8169], device='cuda:1'), covar=tensor([0.1392, 0.0839, 0.3150, 0.0773, 0.3085, 0.1478, 0.5237, 0.2103], device='cuda:1'), in_proj_covar=tensor([0.0248, 0.0221, 0.0253, 0.0305, 0.0302, 0.0252, 0.0277, 0.0276], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 04:40:27,596 INFO [finetune.py:976] (1/7) Epoch 29, batch 3400, loss[loss=0.1465, simple_loss=0.2241, pruned_loss=0.03449, over 4821.00 frames. ], tot_loss[loss=0.1675, simple_loss=0.2416, pruned_loss=0.04668, over 952647.06 frames. ], batch size: 25, lr: 2.83e-03, grad_scale: 32.0 2023-04-28 04:40:29,472 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.152e+02 1.589e+02 1.829e+02 2.322e+02 5.159e+02, threshold=3.657e+02, percent-clipped=5.0 2023-04-28 04:40:50,558 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=163794.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:41:25,120 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-04-28 04:41:32,567 INFO [finetune.py:976] (1/7) Epoch 29, batch 3450, loss[loss=0.1274, simple_loss=0.212, pruned_loss=0.02146, over 4760.00 frames. ], tot_loss[loss=0.1667, simple_loss=0.2412, pruned_loss=0.0461, over 954202.85 frames. ], batch size: 26, lr: 2.83e-03, grad_scale: 32.0 2023-04-28 04:42:36,322 INFO [finetune.py:976] (1/7) Epoch 29, batch 3500, loss[loss=0.1626, simple_loss=0.2436, pruned_loss=0.04078, over 4920.00 frames. ], tot_loss[loss=0.1651, simple_loss=0.2389, pruned_loss=0.04562, over 954191.44 frames. ], batch size: 38, lr: 2.83e-03, grad_scale: 32.0 2023-04-28 04:42:38,148 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.042e+02 1.447e+02 1.753e+02 2.110e+02 3.235e+02, threshold=3.507e+02, percent-clipped=0.0 2023-04-28 04:42:59,210 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=163896.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:43:09,357 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.2743, 1.7362, 1.6996, 2.2109, 2.4088, 1.9686, 1.9555, 1.6512], device='cuda:1'), covar=tensor([0.2182, 0.1890, 0.1627, 0.1582, 0.1072, 0.2221, 0.2195, 0.2321], device='cuda:1'), in_proj_covar=tensor([0.0313, 0.0307, 0.0347, 0.0285, 0.0324, 0.0304, 0.0298, 0.0373], device='cuda:1'), out_proj_covar=tensor([6.3757e-05, 6.2870e-05, 7.2612e-05, 5.7016e-05, 6.6126e-05, 6.3051e-05, 6.1542e-05, 7.8746e-05], device='cuda:1') 2023-04-28 04:43:09,962 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4340, 1.5991, 1.5126, 1.8471, 1.6537, 2.0770, 1.4820, 3.5730], device='cuda:1'), covar=tensor([0.0556, 0.0741, 0.0748, 0.1146, 0.0620, 0.0466, 0.0699, 0.0124], device='cuda:1'), in_proj_covar=tensor([0.0037, 0.0038, 0.0039, 0.0042, 0.0040, 0.0037, 0.0038, 0.0054], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0013, 0.0013, 0.0015, 0.0014, 0.0013, 0.0013, 0.0016], device='cuda:1') 2023-04-28 04:43:23,816 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.50 vs. limit=5.0 2023-04-28 04:43:28,395 INFO [finetune.py:976] (1/7) Epoch 29, batch 3550, loss[loss=0.1297, simple_loss=0.2013, pruned_loss=0.029, over 4758.00 frames. ], tot_loss[loss=0.1632, simple_loss=0.2368, pruned_loss=0.04485, over 954853.58 frames. ], batch size: 27, lr: 2.83e-03, grad_scale: 32.0 2023-04-28 04:43:30,369 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.4174, 3.1594, 2.3553, 2.7770, 1.8421, 1.8238, 2.5027, 1.8017], device='cuda:1'), covar=tensor([0.1674, 0.1405, 0.1369, 0.1450, 0.2296, 0.1882, 0.0955, 0.2008], device='cuda:1'), in_proj_covar=tensor([0.0199, 0.0210, 0.0171, 0.0204, 0.0201, 0.0187, 0.0157, 0.0188], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-28 04:43:39,968 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=163944.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:44:01,287 INFO [finetune.py:976] (1/7) Epoch 29, batch 3600, loss[loss=0.1505, simple_loss=0.2238, pruned_loss=0.03861, over 4887.00 frames. ], tot_loss[loss=0.1615, simple_loss=0.2345, pruned_loss=0.04424, over 953529.67 frames. ], batch size: 32, lr: 2.83e-03, grad_scale: 32.0 2023-04-28 04:44:01,967 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.80 vs. limit=2.0 2023-04-28 04:44:03,550 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 8.301e+01 1.508e+02 1.790e+02 2.030e+02 3.807e+02, threshold=3.580e+02, percent-clipped=2.0 2023-04-28 04:44:19,188 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6639, 1.3445, 4.7776, 4.5366, 4.1792, 4.5378, 4.3114, 4.2683], device='cuda:1'), covar=tensor([0.7977, 0.6010, 0.1161, 0.1831, 0.1161, 0.1325, 0.1685, 0.1606], device='cuda:1'), in_proj_covar=tensor([0.0307, 0.0307, 0.0405, 0.0407, 0.0348, 0.0412, 0.0317, 0.0360], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-28 04:44:36,729 INFO [finetune.py:976] (1/7) Epoch 29, batch 3650, loss[loss=0.1782, simple_loss=0.2573, pruned_loss=0.04954, over 4830.00 frames. ], tot_loss[loss=0.1646, simple_loss=0.238, pruned_loss=0.04566, over 953939.20 frames. ], batch size: 47, lr: 2.83e-03, grad_scale: 32.0 2023-04-28 04:45:08,911 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0093, 2.5290, 1.0138, 1.3941, 1.9083, 1.3020, 3.0848, 1.7789], device='cuda:1'), covar=tensor([0.0705, 0.0497, 0.0718, 0.1291, 0.0481, 0.0959, 0.0362, 0.0571], device='cuda:1'), in_proj_covar=tensor([0.0050, 0.0063, 0.0047, 0.0046, 0.0049, 0.0051, 0.0072, 0.0050], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:1') 2023-04-28 04:45:09,521 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.3180, 2.9974, 0.8658, 1.7733, 1.7556, 2.1354, 1.7470, 0.9433], device='cuda:1'), covar=tensor([0.1372, 0.0884, 0.1877, 0.1152, 0.1059, 0.0989, 0.1475, 0.1891], device='cuda:1'), in_proj_covar=tensor([0.0117, 0.0238, 0.0135, 0.0120, 0.0132, 0.0153, 0.0117, 0.0119], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-28 04:45:10,039 INFO [finetune.py:976] (1/7) Epoch 29, batch 3700, loss[loss=0.1844, simple_loss=0.2715, pruned_loss=0.04869, over 4821.00 frames. ], tot_loss[loss=0.1663, simple_loss=0.2404, pruned_loss=0.04611, over 956132.31 frames. ], batch size: 39, lr: 2.83e-03, grad_scale: 32.0 2023-04-28 04:45:11,858 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.127e+02 1.572e+02 1.919e+02 2.476e+02 4.831e+02, threshold=3.838e+02, percent-clipped=5.0 2023-04-28 04:45:19,704 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=164090.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:45:43,377 INFO [finetune.py:976] (1/7) Epoch 29, batch 3750, loss[loss=0.1815, simple_loss=0.2621, pruned_loss=0.05041, over 4887.00 frames. ], tot_loss[loss=0.1665, simple_loss=0.2408, pruned_loss=0.04603, over 957179.83 frames. ], batch size: 43, lr: 2.83e-03, grad_scale: 32.0 2023-04-28 04:45:58,186 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([0.7745, 1.7432, 1.7425, 1.3892, 1.8883, 1.5952, 2.3537, 1.5814], device='cuda:1'), covar=tensor([0.3556, 0.1809, 0.4611, 0.2658, 0.1500, 0.2101, 0.1478, 0.4373], device='cuda:1'), in_proj_covar=tensor([0.0340, 0.0354, 0.0422, 0.0353, 0.0385, 0.0375, 0.0372, 0.0425], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-28 04:46:11,047 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=164151.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:46:13,496 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=164155.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:46:19,408 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=3.99 vs. limit=5.0 2023-04-28 04:46:43,097 INFO [finetune.py:976] (1/7) Epoch 29, batch 3800, loss[loss=0.1099, simple_loss=0.1853, pruned_loss=0.01721, over 4725.00 frames. ], tot_loss[loss=0.1665, simple_loss=0.2413, pruned_loss=0.0459, over 957188.59 frames. ], batch size: 23, lr: 2.83e-03, grad_scale: 32.0 2023-04-28 04:46:47,404 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.309e+01 1.497e+02 1.761e+02 2.099e+02 4.096e+02, threshold=3.523e+02, percent-clipped=1.0 2023-04-28 04:46:59,688 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7628, 1.7794, 2.2815, 2.3056, 1.5667, 1.4854, 1.7711, 1.0042], device='cuda:1'), covar=tensor([0.0651, 0.0641, 0.0383, 0.0731, 0.0741, 0.1086, 0.0701, 0.0675], device='cuda:1'), in_proj_covar=tensor([0.0071, 0.0067, 0.0065, 0.0069, 0.0076, 0.0095, 0.0073, 0.0062], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-28 04:47:17,880 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=164216.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:47:24,758 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-04-28 04:47:34,750 INFO [finetune.py:976] (1/7) Epoch 29, batch 3850, loss[loss=0.1654, simple_loss=0.2255, pruned_loss=0.05263, over 4685.00 frames. ], tot_loss[loss=0.165, simple_loss=0.2395, pruned_loss=0.04526, over 954917.89 frames. ], batch size: 23, lr: 2.83e-03, grad_scale: 32.0 2023-04-28 04:48:38,394 INFO [finetune.py:976] (1/7) Epoch 29, batch 3900, loss[loss=0.1226, simple_loss=0.1989, pruned_loss=0.02316, over 4759.00 frames. ], tot_loss[loss=0.1623, simple_loss=0.2365, pruned_loss=0.04406, over 955431.08 frames. ], batch size: 28, lr: 2.83e-03, grad_scale: 32.0 2023-04-28 04:48:40,203 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.003e+02 1.498e+02 1.753e+02 2.124e+02 6.066e+02, threshold=3.506e+02, percent-clipped=2.0 2023-04-28 04:49:00,460 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6546, 1.3239, 4.3568, 4.1163, 3.7736, 4.1297, 4.0196, 3.8709], device='cuda:1'), covar=tensor([0.7112, 0.5590, 0.1002, 0.1458, 0.1102, 0.1767, 0.2160, 0.1372], device='cuda:1'), in_proj_covar=tensor([0.0306, 0.0306, 0.0403, 0.0405, 0.0347, 0.0409, 0.0315, 0.0358], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-28 04:49:39,858 INFO [finetune.py:976] (1/7) Epoch 29, batch 3950, loss[loss=0.1392, simple_loss=0.2184, pruned_loss=0.03001, over 4773.00 frames. ], tot_loss[loss=0.1602, simple_loss=0.2333, pruned_loss=0.04355, over 954163.83 frames. ], batch size: 26, lr: 2.83e-03, grad_scale: 32.0 2023-04-28 04:50:24,376 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=164360.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:50:45,027 INFO [finetune.py:976] (1/7) Epoch 29, batch 4000, loss[loss=0.1326, simple_loss=0.2184, pruned_loss=0.0234, over 4198.00 frames. ], tot_loss[loss=0.1592, simple_loss=0.2322, pruned_loss=0.04307, over 953643.52 frames. ], batch size: 65, lr: 2.83e-03, grad_scale: 32.0 2023-04-28 04:50:47,323 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.986e+01 1.466e+02 1.722e+02 2.043e+02 3.565e+02, threshold=3.444e+02, percent-clipped=1.0 2023-04-28 04:51:39,480 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-04-28 04:51:41,750 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.3020, 1.2981, 3.8501, 3.6220, 3.3894, 3.7089, 3.7465, 3.4332], device='cuda:1'), covar=tensor([0.7706, 0.5692, 0.1447, 0.2192, 0.1422, 0.2184, 0.1447, 0.1868], device='cuda:1'), in_proj_covar=tensor([0.0311, 0.0310, 0.0408, 0.0410, 0.0351, 0.0414, 0.0319, 0.0364], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-28 04:51:41,803 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=164421.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:51:48,792 INFO [finetune.py:976] (1/7) Epoch 29, batch 4050, loss[loss=0.1443, simple_loss=0.2269, pruned_loss=0.0308, over 4791.00 frames. ], tot_loss[loss=0.1629, simple_loss=0.2365, pruned_loss=0.04461, over 952246.18 frames. ], batch size: 29, lr: 2.83e-03, grad_scale: 32.0 2023-04-28 04:52:09,570 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-04-28 04:52:20,687 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=164446.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:52:55,386 INFO [finetune.py:976] (1/7) Epoch 29, batch 4100, loss[loss=0.1754, simple_loss=0.2621, pruned_loss=0.04429, over 4820.00 frames. ], tot_loss[loss=0.165, simple_loss=0.2393, pruned_loss=0.0454, over 952294.55 frames. ], batch size: 40, lr: 2.83e-03, grad_scale: 32.0 2023-04-28 04:53:02,652 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.013e+02 1.616e+02 1.841e+02 2.144e+02 4.180e+02, threshold=3.683e+02, percent-clipped=3.0 2023-04-28 04:53:41,177 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=164511.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:53:51,243 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4299, 1.3361, 1.7017, 1.6323, 1.2856, 1.2913, 1.3703, 0.8984], device='cuda:1'), covar=tensor([0.0517, 0.0559, 0.0359, 0.0531, 0.0759, 0.1027, 0.0453, 0.0535], device='cuda:1'), in_proj_covar=tensor([0.0071, 0.0067, 0.0065, 0.0069, 0.0076, 0.0095, 0.0072, 0.0062], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-28 04:53:54,828 INFO [finetune.py:976] (1/7) Epoch 29, batch 4150, loss[loss=0.1598, simple_loss=0.2372, pruned_loss=0.04122, over 4909.00 frames. ], tot_loss[loss=0.1662, simple_loss=0.2408, pruned_loss=0.04575, over 953478.16 frames. ], batch size: 36, lr: 2.83e-03, grad_scale: 32.0 2023-04-28 04:54:35,885 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=164568.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:54:40,037 INFO [finetune.py:976] (1/7) Epoch 29, batch 4200, loss[loss=0.166, simple_loss=0.2363, pruned_loss=0.04787, over 4743.00 frames. ], tot_loss[loss=0.1657, simple_loss=0.2407, pruned_loss=0.0454, over 955909.90 frames. ], batch size: 54, lr: 2.83e-03, grad_scale: 32.0 2023-04-28 04:54:41,966 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=164578.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:54:42,442 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.830e+01 1.587e+02 1.797e+02 2.330e+02 9.173e+02, threshold=3.593e+02, percent-clipped=2.0 2023-04-28 04:54:59,999 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=3.75 vs. limit=5.0 2023-04-28 04:55:08,269 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([3.5207, 3.4290, 2.7583, 4.0944, 3.4908, 3.5039, 1.6011, 3.4668], device='cuda:1'), covar=tensor([0.1748, 0.1289, 0.3511, 0.1870, 0.2892, 0.1826, 0.5553, 0.2527], device='cuda:1'), in_proj_covar=tensor([0.0246, 0.0221, 0.0253, 0.0305, 0.0301, 0.0252, 0.0276, 0.0276], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 04:55:14,299 INFO [finetune.py:976] (1/7) Epoch 29, batch 4250, loss[loss=0.1831, simple_loss=0.2502, pruned_loss=0.05802, over 4910.00 frames. ], tot_loss[loss=0.1645, simple_loss=0.2387, pruned_loss=0.04509, over 955753.98 frames. ], batch size: 37, lr: 2.83e-03, grad_scale: 32.0 2023-04-28 04:55:16,888 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=164629.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:55:22,823 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.1600, 1.4539, 1.2977, 1.7327, 1.5754, 1.6469, 1.3742, 3.0082], device='cuda:1'), covar=tensor([0.0613, 0.0746, 0.0750, 0.1124, 0.0600, 0.0504, 0.0700, 0.0165], device='cuda:1'), in_proj_covar=tensor([0.0037, 0.0038, 0.0039, 0.0042, 0.0040, 0.0037, 0.0038, 0.0054], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0013, 0.0013, 0.0015, 0.0014, 0.0013, 0.0013, 0.0016], device='cuda:1') 2023-04-28 04:55:24,091 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=164639.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 04:55:37,641 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=164658.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:55:48,429 INFO [finetune.py:976] (1/7) Epoch 29, batch 4300, loss[loss=0.1433, simple_loss=0.2207, pruned_loss=0.03292, over 4906.00 frames. ], tot_loss[loss=0.1632, simple_loss=0.2366, pruned_loss=0.04494, over 956184.23 frames. ], batch size: 36, lr: 2.83e-03, grad_scale: 32.0 2023-04-28 04:55:50,845 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.176e+01 1.533e+02 1.704e+02 2.198e+02 4.636e+02, threshold=3.409e+02, percent-clipped=3.0 2023-04-28 04:55:52,909 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.77 vs. limit=2.0 2023-04-28 04:55:56,464 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.79 vs. limit=2.0 2023-04-28 04:56:08,723 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.67 vs. limit=2.0 2023-04-28 04:56:16,856 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=164716.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:56:18,747 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=164719.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:56:21,264 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-04-28 04:56:22,290 INFO [finetune.py:976] (1/7) Epoch 29, batch 4350, loss[loss=0.1624, simple_loss=0.2336, pruned_loss=0.04556, over 4916.00 frames. ], tot_loss[loss=0.1609, simple_loss=0.2337, pruned_loss=0.04402, over 954946.20 frames. ], batch size: 37, lr: 2.83e-03, grad_scale: 32.0 2023-04-28 04:56:34,728 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.52 vs. limit=2.0 2023-04-28 04:56:36,272 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=164746.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:56:36,312 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=164746.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:57:17,722 INFO [finetune.py:976] (1/7) Epoch 29, batch 4400, loss[loss=0.1221, simple_loss=0.1957, pruned_loss=0.02427, over 4774.00 frames. ], tot_loss[loss=0.1612, simple_loss=0.2339, pruned_loss=0.04419, over 955642.59 frames. ], batch size: 26, lr: 2.83e-03, grad_scale: 32.0 2023-04-28 04:57:20,181 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.037e+02 1.542e+02 1.919e+02 2.288e+02 3.219e+02, threshold=3.837e+02, percent-clipped=0.0 2023-04-28 04:57:41,272 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=164794.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:58:01,553 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=164807.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:58:04,446 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=164811.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:58:24,162 INFO [finetune.py:976] (1/7) Epoch 29, batch 4450, loss[loss=0.2091, simple_loss=0.2826, pruned_loss=0.06787, over 4904.00 frames. ], tot_loss[loss=0.1642, simple_loss=0.2384, pruned_loss=0.04499, over 957147.60 frames. ], batch size: 36, lr: 2.83e-03, grad_scale: 32.0 2023-04-28 04:58:36,791 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.0555, 2.5497, 0.8552, 1.4114, 1.4320, 1.9314, 1.5323, 0.8455], device='cuda:1'), covar=tensor([0.1448, 0.1012, 0.1688, 0.1319, 0.1147, 0.0912, 0.1530, 0.1701], device='cuda:1'), in_proj_covar=tensor([0.0117, 0.0237, 0.0135, 0.0120, 0.0131, 0.0152, 0.0117, 0.0118], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-28 04:58:59,954 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=164859.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:59:10,509 INFO [finetune.py:976] (1/7) Epoch 29, batch 4500, loss[loss=0.195, simple_loss=0.258, pruned_loss=0.06597, over 4902.00 frames. ], tot_loss[loss=0.1659, simple_loss=0.2405, pruned_loss=0.04568, over 958259.08 frames. ], batch size: 46, lr: 2.83e-03, grad_scale: 32.0 2023-04-28 04:59:18,304 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.040e+02 1.480e+02 1.804e+02 2.196e+02 5.471e+02, threshold=3.609e+02, percent-clipped=1.0 2023-04-28 05:00:15,087 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=164924.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:00:20,955 INFO [finetune.py:976] (1/7) Epoch 29, batch 4550, loss[loss=0.161, simple_loss=0.2281, pruned_loss=0.04691, over 4930.00 frames. ], tot_loss[loss=0.165, simple_loss=0.2399, pruned_loss=0.04503, over 957206.99 frames. ], batch size: 33, lr: 2.83e-03, grad_scale: 32.0 2023-04-28 05:00:32,596 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=164934.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 05:01:06,393 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6819, 1.5488, 1.6746, 1.9696, 2.0471, 1.6217, 1.3099, 1.7892], device='cuda:1'), covar=tensor([0.0721, 0.1060, 0.0739, 0.0508, 0.0488, 0.0746, 0.0703, 0.0518], device='cuda:1'), in_proj_covar=tensor([0.0182, 0.0200, 0.0182, 0.0170, 0.0176, 0.0175, 0.0150, 0.0176], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-28 05:01:26,332 INFO [finetune.py:976] (1/7) Epoch 29, batch 4600, loss[loss=0.1726, simple_loss=0.2346, pruned_loss=0.05524, over 4822.00 frames. ], tot_loss[loss=0.165, simple_loss=0.2399, pruned_loss=0.04503, over 957185.13 frames. ], batch size: 40, lr: 2.83e-03, grad_scale: 32.0 2023-04-28 05:01:29,216 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.013e+02 1.461e+02 1.731e+02 1.968e+02 2.942e+02, threshold=3.463e+02, percent-clipped=1.0 2023-04-28 05:02:13,343 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.1504, 2.6988, 2.3918, 2.5843, 2.0408, 2.4292, 2.4707, 1.9535], device='cuda:1'), covar=tensor([0.2073, 0.1138, 0.0804, 0.1228, 0.2909, 0.1068, 0.2043, 0.2527], device='cuda:1'), in_proj_covar=tensor([0.0284, 0.0303, 0.0218, 0.0277, 0.0315, 0.0255, 0.0249, 0.0264], device='cuda:1'), out_proj_covar=tensor([1.1303e-04, 1.1890e-04, 8.5375e-05, 1.0884e-04, 1.2679e-04, 1.0011e-04, 1.0005e-04, 1.0402e-04], device='cuda:1') 2023-04-28 05:02:14,410 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=165014.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:02:20,895 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=165016.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:02:32,274 INFO [finetune.py:976] (1/7) Epoch 29, batch 4650, loss[loss=0.1868, simple_loss=0.249, pruned_loss=0.06231, over 4893.00 frames. ], tot_loss[loss=0.1648, simple_loss=0.2385, pruned_loss=0.0455, over 956828.94 frames. ], batch size: 32, lr: 2.83e-03, grad_scale: 32.0 2023-04-28 05:02:34,339 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-04-28 05:02:55,402 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.09 vs. limit=5.0 2023-04-28 05:03:08,792 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.2819, 1.8111, 2.1012, 2.4480, 2.1017, 1.7104, 1.4677, 1.9537], device='cuda:1'), covar=tensor([0.2479, 0.2521, 0.1411, 0.1641, 0.2155, 0.2298, 0.3903, 0.1659], device='cuda:1'), in_proj_covar=tensor([0.0294, 0.0248, 0.0230, 0.0317, 0.0224, 0.0237, 0.0231, 0.0187], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-28 05:03:18,781 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=165064.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:03:37,007 INFO [finetune.py:976] (1/7) Epoch 29, batch 4700, loss[loss=0.1326, simple_loss=0.2138, pruned_loss=0.02567, over 4896.00 frames. ], tot_loss[loss=0.1626, simple_loss=0.2358, pruned_loss=0.04472, over 957002.60 frames. ], batch size: 35, lr: 2.83e-03, grad_scale: 16.0 2023-04-28 05:03:39,644 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.64 vs. limit=2.0 2023-04-28 05:03:40,047 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.099e+02 1.538e+02 1.838e+02 2.257e+02 4.770e+02, threshold=3.676e+02, percent-clipped=2.0 2023-04-28 05:03:57,510 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.5769, 2.4405, 2.5110, 2.9528, 2.9364, 2.3233, 2.1334, 2.6338], device='cuda:1'), covar=tensor([0.0721, 0.0916, 0.0663, 0.0489, 0.0516, 0.0788, 0.0649, 0.0501], device='cuda:1'), in_proj_covar=tensor([0.0183, 0.0200, 0.0183, 0.0170, 0.0176, 0.0175, 0.0149, 0.0176], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-28 05:04:09,169 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=165102.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:04:40,828 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=165123.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:04:42,451 INFO [finetune.py:976] (1/7) Epoch 29, batch 4750, loss[loss=0.1527, simple_loss=0.2206, pruned_loss=0.04242, over 4928.00 frames. ], tot_loss[loss=0.1614, simple_loss=0.234, pruned_loss=0.04437, over 957935.11 frames. ], batch size: 33, lr: 2.83e-03, grad_scale: 16.0 2023-04-28 05:05:11,236 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8257, 2.4415, 1.9602, 1.9332, 1.3198, 1.3460, 2.0346, 1.3010], device='cuda:1'), covar=tensor([0.1571, 0.1258, 0.1244, 0.1514, 0.2225, 0.1854, 0.0880, 0.2012], device='cuda:1'), in_proj_covar=tensor([0.0198, 0.0209, 0.0171, 0.0204, 0.0201, 0.0187, 0.0157, 0.0188], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-28 05:05:22,434 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.1964, 1.4565, 1.4447, 2.0728, 2.2338, 1.7832, 1.8352, 1.5035], device='cuda:1'), covar=tensor([0.1951, 0.2232, 0.2097, 0.1685, 0.1338, 0.2218, 0.2494, 0.2511], device='cuda:1'), in_proj_covar=tensor([0.0313, 0.0308, 0.0348, 0.0287, 0.0325, 0.0304, 0.0300, 0.0376], device='cuda:1'), out_proj_covar=tensor([6.3704e-05, 6.3127e-05, 7.2827e-05, 5.7344e-05, 6.6151e-05, 6.3042e-05, 6.1806e-05, 7.9423e-05], device='cuda:1') 2023-04-28 05:05:24,372 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-04-28 05:05:25,308 INFO [finetune.py:976] (1/7) Epoch 29, batch 4800, loss[loss=0.1559, simple_loss=0.2235, pruned_loss=0.04413, over 4756.00 frames. ], tot_loss[loss=0.1623, simple_loss=0.2353, pruned_loss=0.04463, over 956399.28 frames. ], batch size: 27, lr: 2.82e-03, grad_scale: 16.0 2023-04-28 05:05:28,758 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.123e+02 1.552e+02 1.836e+02 2.145e+02 4.672e+02, threshold=3.672e+02, percent-clipped=1.0 2023-04-28 05:05:31,331 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=165184.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:05:48,476 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-04-28 05:05:57,703 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=165224.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:05:58,220 INFO [finetune.py:976] (1/7) Epoch 29, batch 4850, loss[loss=0.2256, simple_loss=0.2962, pruned_loss=0.07755, over 4815.00 frames. ], tot_loss[loss=0.1649, simple_loss=0.2387, pruned_loss=0.04558, over 955671.82 frames. ], batch size: 38, lr: 2.82e-03, grad_scale: 16.0 2023-04-28 05:06:05,323 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=165234.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 05:06:29,519 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=165272.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:06:31,766 INFO [finetune.py:976] (1/7) Epoch 29, batch 4900, loss[loss=0.1912, simple_loss=0.283, pruned_loss=0.04971, over 4901.00 frames. ], tot_loss[loss=0.1649, simple_loss=0.239, pruned_loss=0.04546, over 954839.21 frames. ], batch size: 43, lr: 2.82e-03, grad_scale: 16.0 2023-04-28 05:06:35,317 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.134e+02 1.561e+02 1.845e+02 2.104e+02 4.657e+02, threshold=3.691e+02, percent-clipped=1.0 2023-04-28 05:06:37,063 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=165282.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:06:57,526 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=165314.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:07:04,967 INFO [finetune.py:976] (1/7) Epoch 29, batch 4950, loss[loss=0.1368, simple_loss=0.2204, pruned_loss=0.02662, over 4773.00 frames. ], tot_loss[loss=0.1658, simple_loss=0.2406, pruned_loss=0.04551, over 953580.10 frames. ], batch size: 29, lr: 2.82e-03, grad_scale: 16.0 2023-04-28 05:07:15,576 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.51 vs. limit=2.0 2023-04-28 05:07:29,754 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=165362.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:07:32,163 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.8391, 2.7717, 2.1416, 3.2710, 2.8178, 2.7945, 1.2095, 2.7986], device='cuda:1'), covar=tensor([0.2103, 0.1834, 0.3649, 0.3118, 0.3547, 0.2261, 0.5601, 0.2803], device='cuda:1'), in_proj_covar=tensor([0.0245, 0.0218, 0.0251, 0.0303, 0.0300, 0.0249, 0.0273, 0.0274], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 05:07:38,075 INFO [finetune.py:976] (1/7) Epoch 29, batch 5000, loss[loss=0.1562, simple_loss=0.2439, pruned_loss=0.03423, over 4913.00 frames. ], tot_loss[loss=0.1639, simple_loss=0.2382, pruned_loss=0.04481, over 952503.54 frames. ], batch size: 37, lr: 2.82e-03, grad_scale: 16.0 2023-04-28 05:07:41,610 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.839e+01 1.556e+02 1.839e+02 2.298e+02 4.035e+02, threshold=3.677e+02, percent-clipped=3.0 2023-04-28 05:07:57,188 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=165402.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:08:12,132 INFO [finetune.py:976] (1/7) Epoch 29, batch 5050, loss[loss=0.1866, simple_loss=0.2505, pruned_loss=0.06137, over 4733.00 frames. ], tot_loss[loss=0.1621, simple_loss=0.2358, pruned_loss=0.0442, over 953657.99 frames. ], batch size: 23, lr: 2.82e-03, grad_scale: 16.0 2023-04-28 05:08:22,717 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([4.8600, 4.7600, 3.2209, 5.5828, 4.8927, 4.7621, 2.3944, 4.7282], device='cuda:1'), covar=tensor([0.1418, 0.0840, 0.2855, 0.0808, 0.2877, 0.1656, 0.5224, 0.2119], device='cuda:1'), in_proj_covar=tensor([0.0245, 0.0218, 0.0251, 0.0303, 0.0300, 0.0249, 0.0273, 0.0274], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 05:08:29,765 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=165450.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:08:45,559 INFO [finetune.py:976] (1/7) Epoch 29, batch 5100, loss[loss=0.1554, simple_loss=0.2291, pruned_loss=0.04084, over 4917.00 frames. ], tot_loss[loss=0.1607, simple_loss=0.2335, pruned_loss=0.04395, over 954966.66 frames. ], batch size: 43, lr: 2.82e-03, grad_scale: 16.0 2023-04-28 05:08:46,906 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.1801, 2.6245, 2.2252, 2.5569, 1.8446, 2.2967, 2.1740, 1.7012], device='cuda:1'), covar=tensor([0.1593, 0.1119, 0.0752, 0.0944, 0.3031, 0.1135, 0.1762, 0.2603], device='cuda:1'), in_proj_covar=tensor([0.0280, 0.0299, 0.0215, 0.0274, 0.0311, 0.0252, 0.0245, 0.0261], device='cuda:1'), out_proj_covar=tensor([1.1153e-04, 1.1749e-04, 8.4243e-05, 1.0738e-04, 1.2503e-04, 9.8940e-05, 9.8694e-05, 1.0267e-04], device='cuda:1') 2023-04-28 05:08:48,020 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4912, 1.4994, 1.8523, 1.8316, 1.3638, 1.2464, 1.4818, 0.8317], device='cuda:1'), covar=tensor([0.0488, 0.0541, 0.0353, 0.0554, 0.0657, 0.0931, 0.0499, 0.0624], device='cuda:1'), in_proj_covar=tensor([0.0070, 0.0067, 0.0065, 0.0069, 0.0076, 0.0095, 0.0072, 0.0062], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-28 05:08:48,558 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=165479.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:08:49,075 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 8.177e+01 1.550e+02 1.878e+02 2.337e+02 3.681e+02, threshold=3.756e+02, percent-clipped=1.0 2023-04-28 05:09:48,070 INFO [finetune.py:976] (1/7) Epoch 29, batch 5150, loss[loss=0.1539, simple_loss=0.2296, pruned_loss=0.03907, over 4816.00 frames. ], tot_loss[loss=0.1619, simple_loss=0.2346, pruned_loss=0.04461, over 955445.08 frames. ], batch size: 39, lr: 2.82e-03, grad_scale: 16.0 2023-04-28 05:10:21,699 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=165551.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:10:22,352 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0818, 1.6745, 1.4702, 2.0052, 2.1550, 1.8185, 1.8919, 1.5428], device='cuda:1'), covar=tensor([0.2038, 0.1697, 0.1866, 0.1405, 0.1228, 0.1901, 0.1821, 0.2412], device='cuda:1'), in_proj_covar=tensor([0.0315, 0.0310, 0.0351, 0.0288, 0.0328, 0.0306, 0.0302, 0.0378], device='cuda:1'), out_proj_covar=tensor([6.4215e-05, 6.3398e-05, 7.3342e-05, 5.7570e-05, 6.6764e-05, 6.3473e-05, 6.2260e-05, 7.9964e-05], device='cuda:1') 2023-04-28 05:10:52,971 INFO [finetune.py:976] (1/7) Epoch 29, batch 5200, loss[loss=0.1863, simple_loss=0.2609, pruned_loss=0.05588, over 4828.00 frames. ], tot_loss[loss=0.1649, simple_loss=0.2385, pruned_loss=0.04568, over 952917.85 frames. ], batch size: 30, lr: 2.82e-03, grad_scale: 16.0 2023-04-28 05:10:55,998 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.013e+02 1.620e+02 1.941e+02 2.269e+02 3.767e+02, threshold=3.883e+02, percent-clipped=2.0 2023-04-28 05:11:45,805 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=165612.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:11:59,710 INFO [finetune.py:976] (1/7) Epoch 29, batch 5250, loss[loss=0.1924, simple_loss=0.2635, pruned_loss=0.06065, over 4902.00 frames. ], tot_loss[loss=0.1658, simple_loss=0.2403, pruned_loss=0.04565, over 955156.26 frames. ], batch size: 43, lr: 2.82e-03, grad_scale: 16.0 2023-04-28 05:12:02,283 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9286, 1.7178, 2.0914, 2.3325, 2.0715, 1.9300, 2.0000, 1.9874], device='cuda:1'), covar=tensor([0.3495, 0.5360, 0.4875, 0.4335, 0.4519, 0.6134, 0.5899, 0.6940], device='cuda:1'), in_proj_covar=tensor([0.0447, 0.0426, 0.0521, 0.0509, 0.0475, 0.0515, 0.0516, 0.0531], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-28 05:12:23,587 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=3.79 vs. limit=5.0 2023-04-28 05:13:07,356 INFO [finetune.py:976] (1/7) Epoch 29, batch 5300, loss[loss=0.2013, simple_loss=0.272, pruned_loss=0.06527, over 4892.00 frames. ], tot_loss[loss=0.1676, simple_loss=0.242, pruned_loss=0.04663, over 955206.95 frames. ], batch size: 35, lr: 2.82e-03, grad_scale: 16.0 2023-04-28 05:13:16,015 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.838e+01 1.557e+02 1.826e+02 2.262e+02 6.421e+02, threshold=3.651e+02, percent-clipped=2.0 2023-04-28 05:13:27,876 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.2189, 1.3884, 1.7232, 1.8640, 1.7516, 1.7948, 1.7296, 1.7804], device='cuda:1'), covar=tensor([0.3433, 0.4114, 0.3363, 0.3425, 0.4514, 0.5885, 0.3989, 0.3489], device='cuda:1'), in_proj_covar=tensor([0.0343, 0.0374, 0.0331, 0.0343, 0.0351, 0.0393, 0.0362, 0.0333], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 05:14:12,929 INFO [finetune.py:976] (1/7) Epoch 29, batch 5350, loss[loss=0.1586, simple_loss=0.2366, pruned_loss=0.04035, over 4832.00 frames. ], tot_loss[loss=0.1681, simple_loss=0.2425, pruned_loss=0.04683, over 954818.78 frames. ], batch size: 30, lr: 2.82e-03, grad_scale: 16.0 2023-04-28 05:14:19,344 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0628, 1.8678, 2.0283, 2.3479, 2.4636, 1.9501, 1.6602, 2.1077], device='cuda:1'), covar=tensor([0.0872, 0.1164, 0.0805, 0.0561, 0.0526, 0.0954, 0.0764, 0.0599], device='cuda:1'), in_proj_covar=tensor([0.0183, 0.0201, 0.0184, 0.0170, 0.0177, 0.0177, 0.0149, 0.0176], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-28 05:14:43,454 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=165750.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:15:21,307 INFO [finetune.py:976] (1/7) Epoch 29, batch 5400, loss[loss=0.2077, simple_loss=0.2719, pruned_loss=0.07179, over 4868.00 frames. ], tot_loss[loss=0.1666, simple_loss=0.2403, pruned_loss=0.04648, over 954471.87 frames. ], batch size: 34, lr: 2.82e-03, grad_scale: 16.0 2023-04-28 05:15:24,331 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=165779.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:15:24,845 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.026e+02 1.500e+02 1.780e+02 2.167e+02 4.402e+02, threshold=3.560e+02, percent-clipped=1.0 2023-04-28 05:15:55,077 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=165803.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:16:07,208 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=165811.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:16:20,314 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([0.4256, 1.2515, 1.4906, 1.0935, 1.3858, 1.2300, 1.7554, 1.4617], device='cuda:1'), covar=tensor([0.3688, 0.2162, 0.4879, 0.2585, 0.1618, 0.2148, 0.1659, 0.4750], device='cuda:1'), in_proj_covar=tensor([0.0343, 0.0359, 0.0429, 0.0355, 0.0389, 0.0379, 0.0374, 0.0429], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-28 05:16:21,414 INFO [finetune.py:976] (1/7) Epoch 29, batch 5450, loss[loss=0.1474, simple_loss=0.2228, pruned_loss=0.03599, over 4829.00 frames. ], tot_loss[loss=0.1642, simple_loss=0.2374, pruned_loss=0.04549, over 954416.88 frames. ], batch size: 33, lr: 2.82e-03, grad_scale: 16.0 2023-04-28 05:16:21,566 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9410, 1.1489, 1.6234, 1.7324, 1.7047, 1.7422, 1.6456, 1.6146], device='cuda:1'), covar=tensor([0.3600, 0.4577, 0.3996, 0.4026, 0.4807, 0.6636, 0.4312, 0.4183], device='cuda:1'), in_proj_covar=tensor([0.0345, 0.0376, 0.0332, 0.0344, 0.0352, 0.0395, 0.0364, 0.0335], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 05:16:22,707 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=165827.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:16:30,715 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2023-04-28 05:16:48,219 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=165864.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:16:50,047 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5451, 1.1771, 1.2859, 1.2301, 1.6100, 1.3321, 1.1302, 1.2764], device='cuda:1'), covar=tensor([0.1763, 0.1441, 0.1990, 0.1521, 0.0927, 0.1716, 0.1854, 0.2491], device='cuda:1'), in_proj_covar=tensor([0.0317, 0.0312, 0.0352, 0.0289, 0.0329, 0.0308, 0.0303, 0.0380], device='cuda:1'), out_proj_covar=tensor([6.4652e-05, 6.3801e-05, 7.3600e-05, 5.7644e-05, 6.6993e-05, 6.3833e-05, 6.2566e-05, 8.0412e-05], device='cuda:1') 2023-04-28 05:16:55,273 INFO [finetune.py:976] (1/7) Epoch 29, batch 5500, loss[loss=0.1693, simple_loss=0.2337, pruned_loss=0.05249, over 4148.00 frames. ], tot_loss[loss=0.1619, simple_loss=0.2348, pruned_loss=0.04457, over 954792.28 frames. ], batch size: 65, lr: 2.82e-03, grad_scale: 16.0 2023-04-28 05:16:58,257 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.287e+01 1.513e+02 1.807e+02 2.218e+02 5.669e+02, threshold=3.614e+02, percent-clipped=2.0 2023-04-28 05:17:16,260 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=165907.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:17:29,074 INFO [finetune.py:976] (1/7) Epoch 29, batch 5550, loss[loss=0.1482, simple_loss=0.2292, pruned_loss=0.03358, over 4768.00 frames. ], tot_loss[loss=0.1623, simple_loss=0.2354, pruned_loss=0.04456, over 954158.15 frames. ], batch size: 28, lr: 2.82e-03, grad_scale: 16.0 2023-04-28 05:18:28,708 INFO [finetune.py:976] (1/7) Epoch 29, batch 5600, loss[loss=0.1685, simple_loss=0.2467, pruned_loss=0.04516, over 4904.00 frames. ], tot_loss[loss=0.1635, simple_loss=0.2375, pruned_loss=0.04475, over 952294.02 frames. ], batch size: 37, lr: 2.82e-03, grad_scale: 16.0 2023-04-28 05:18:37,366 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.927e+01 1.529e+02 1.829e+02 2.121e+02 4.998e+02, threshold=3.659e+02, percent-clipped=4.0 2023-04-28 05:19:09,921 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.1765, 1.6446, 2.0157, 2.2176, 1.9239, 1.6399, 1.1106, 1.7317], device='cuda:1'), covar=tensor([0.3392, 0.3332, 0.1846, 0.2136, 0.2671, 0.2776, 0.4391, 0.2021], device='cuda:1'), in_proj_covar=tensor([0.0291, 0.0245, 0.0228, 0.0314, 0.0222, 0.0235, 0.0229, 0.0184], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-28 05:19:32,708 INFO [finetune.py:976] (1/7) Epoch 29, batch 5650, loss[loss=0.1458, simple_loss=0.224, pruned_loss=0.03379, over 4871.00 frames. ], tot_loss[loss=0.1657, simple_loss=0.2401, pruned_loss=0.04564, over 951275.64 frames. ], batch size: 34, lr: 2.82e-03, grad_scale: 16.0 2023-04-28 05:19:43,882 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.1380, 1.2864, 3.8129, 3.5963, 3.3964, 3.6804, 3.7092, 3.3623], device='cuda:1'), covar=tensor([0.7666, 0.5752, 0.1254, 0.1845, 0.1266, 0.2032, 0.1651, 0.1642], device='cuda:1'), in_proj_covar=tensor([0.0309, 0.0309, 0.0406, 0.0409, 0.0350, 0.0414, 0.0316, 0.0363], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-28 05:20:11,617 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=166054.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 05:20:31,781 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 2023-04-28 05:20:33,889 INFO [finetune.py:976] (1/7) Epoch 29, batch 5700, loss[loss=0.1072, simple_loss=0.1714, pruned_loss=0.02149, over 4429.00 frames. ], tot_loss[loss=0.1637, simple_loss=0.237, pruned_loss=0.04515, over 935217.38 frames. ], batch size: 19, lr: 2.82e-03, grad_scale: 16.0 2023-04-28 05:20:42,336 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 6.981e+01 1.433e+02 1.713e+02 2.148e+02 4.581e+02, threshold=3.425e+02, percent-clipped=1.0 2023-04-28 05:21:19,175 INFO [finetune.py:976] (1/7) Epoch 30, batch 0, loss[loss=0.2119, simple_loss=0.2714, pruned_loss=0.07623, over 4752.00 frames. ], tot_loss[loss=0.2119, simple_loss=0.2714, pruned_loss=0.07623, over 4752.00 frames. ], batch size: 54, lr: 2.82e-03, grad_scale: 16.0 2023-04-28 05:21:19,176 INFO [finetune.py:1001] (1/7) Computing validation loss 2023-04-28 05:21:20,963 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8703, 1.1049, 1.7991, 2.3665, 1.9525, 1.7676, 1.7709, 1.7718], device='cuda:1'), covar=tensor([0.4574, 0.7336, 0.6143, 0.5744, 0.6365, 0.8223, 0.8397, 0.8897], device='cuda:1'), in_proj_covar=tensor([0.0448, 0.0427, 0.0523, 0.0511, 0.0477, 0.0517, 0.0519, 0.0533], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-28 05:21:26,945 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.2579, 1.5490, 1.8136, 1.9550, 1.8834, 1.8894, 1.7806, 1.8212], device='cuda:1'), covar=tensor([0.3982, 0.5449, 0.4427, 0.4218, 0.5595, 0.7205, 0.5573, 0.4921], device='cuda:1'), in_proj_covar=tensor([0.0347, 0.0377, 0.0333, 0.0346, 0.0354, 0.0396, 0.0365, 0.0336], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 05:21:34,681 INFO [finetune.py:1010] (1/7) Epoch 30, validation: loss=0.1551, simple_loss=0.2236, pruned_loss=0.04334, over 2265189.00 frames. 2023-04-28 05:21:34,682 INFO [finetune.py:1011] (1/7) Maximum memory allocated so far is 6391MB 2023-04-28 05:21:37,612 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=166106.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:21:48,590 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=166115.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 05:22:31,225 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=166147.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:22:35,224 INFO [finetune.py:976] (1/7) Epoch 30, batch 50, loss[loss=0.1801, simple_loss=0.2514, pruned_loss=0.0544, over 4852.00 frames. ], tot_loss[loss=0.1652, simple_loss=0.2397, pruned_loss=0.04532, over 215061.64 frames. ], batch size: 49, lr: 2.81e-03, grad_scale: 16.0 2023-04-28 05:22:46,129 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=166159.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:23:06,032 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7898, 1.4963, 1.4426, 1.6041, 1.9868, 1.6028, 1.4134, 1.3480], device='cuda:1'), covar=tensor([0.1534, 0.1524, 0.1793, 0.1303, 0.0781, 0.1624, 0.1963, 0.2366], device='cuda:1'), in_proj_covar=tensor([0.0316, 0.0310, 0.0351, 0.0287, 0.0327, 0.0306, 0.0302, 0.0379], device='cuda:1'), out_proj_covar=tensor([6.4382e-05, 6.3487e-05, 7.3320e-05, 5.7392e-05, 6.6687e-05, 6.3455e-05, 6.2253e-05, 8.0039e-05], device='cuda:1') 2023-04-28 05:23:14,383 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 7.993e+01 1.354e+02 1.707e+02 2.159e+02 3.038e+02, threshold=3.414e+02, percent-clipped=0.0 2023-04-28 05:23:47,365 INFO [finetune.py:976] (1/7) Epoch 30, batch 100, loss[loss=0.2185, simple_loss=0.2764, pruned_loss=0.08026, over 4872.00 frames. ], tot_loss[loss=0.1606, simple_loss=0.2344, pruned_loss=0.04337, over 380526.62 frames. ], batch size: 34, lr: 2.81e-03, grad_scale: 16.0 2023-04-28 05:23:50,898 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=166207.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:23:56,828 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=166208.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:24:43,852 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([3.7034, 3.5781, 2.6643, 4.3561, 3.7581, 3.6725, 1.9062, 3.6883], device='cuda:1'), covar=tensor([0.1923, 0.1284, 0.3884, 0.1617, 0.3024, 0.1926, 0.5494, 0.2422], device='cuda:1'), in_proj_covar=tensor([0.0244, 0.0217, 0.0251, 0.0302, 0.0297, 0.0248, 0.0273, 0.0274], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 05:24:53,826 INFO [finetune.py:976] (1/7) Epoch 30, batch 150, loss[loss=0.1328, simple_loss=0.2075, pruned_loss=0.02901, over 4768.00 frames. ], tot_loss[loss=0.1597, simple_loss=0.2317, pruned_loss=0.04381, over 507341.48 frames. ], batch size: 28, lr: 2.81e-03, grad_scale: 16.0 2023-04-28 05:24:55,116 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=166255.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:25:28,828 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 8.799e+01 1.515e+02 1.784e+02 2.130e+02 3.221e+02, threshold=3.568e+02, percent-clipped=0.0 2023-04-28 05:25:59,857 INFO [finetune.py:976] (1/7) Epoch 30, batch 200, loss[loss=0.1375, simple_loss=0.2162, pruned_loss=0.02938, over 4791.00 frames. ], tot_loss[loss=0.1608, simple_loss=0.2323, pruned_loss=0.04465, over 605556.50 frames. ], batch size: 51, lr: 2.81e-03, grad_scale: 16.0 2023-04-28 05:26:02,462 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=166307.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:27:02,967 INFO [finetune.py:976] (1/7) Epoch 30, batch 250, loss[loss=0.1577, simple_loss=0.2264, pruned_loss=0.04445, over 4773.00 frames. ], tot_loss[loss=0.1631, simple_loss=0.2354, pruned_loss=0.04544, over 682906.74 frames. ], batch size: 27, lr: 2.81e-03, grad_scale: 16.0 2023-04-28 05:27:19,832 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=166368.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:27:38,573 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.540e+01 1.580e+02 1.903e+02 2.256e+02 3.495e+02, threshold=3.807e+02, percent-clipped=0.0 2023-04-28 05:28:02,277 INFO [finetune.py:976] (1/7) Epoch 30, batch 300, loss[loss=0.1407, simple_loss=0.2143, pruned_loss=0.0336, over 4914.00 frames. ], tot_loss[loss=0.1666, simple_loss=0.2404, pruned_loss=0.04633, over 745300.87 frames. ], batch size: 36, lr: 2.81e-03, grad_scale: 16.0 2023-04-28 05:28:07,536 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=166406.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:28:08,962 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.49 vs. limit=2.0 2023-04-28 05:28:09,962 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=166410.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 05:29:01,973 INFO [finetune.py:976] (1/7) Epoch 30, batch 350, loss[loss=0.2039, simple_loss=0.2779, pruned_loss=0.06497, over 4920.00 frames. ], tot_loss[loss=0.1695, simple_loss=0.2436, pruned_loss=0.04765, over 793092.26 frames. ], batch size: 42, lr: 2.81e-03, grad_scale: 16.0 2023-04-28 05:29:08,237 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=166454.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:29:11,886 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=166459.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:29:31,640 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.632e+01 1.563e+02 1.883e+02 2.303e+02 4.800e+02, threshold=3.766e+02, percent-clipped=2.0 2023-04-28 05:29:32,389 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=166481.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:29:46,150 INFO [finetune.py:976] (1/7) Epoch 30, batch 400, loss[loss=0.1841, simple_loss=0.2539, pruned_loss=0.05716, over 4862.00 frames. ], tot_loss[loss=0.1691, simple_loss=0.2439, pruned_loss=0.0472, over 827938.57 frames. ], batch size: 34, lr: 2.81e-03, grad_scale: 16.0 2023-04-28 05:29:46,223 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=166503.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:29:49,095 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=166507.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:29:50,826 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=166509.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:29:50,903 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-04-28 05:30:06,151 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=166531.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:30:13,247 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=166542.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:30:19,818 INFO [finetune.py:976] (1/7) Epoch 30, batch 450, loss[loss=0.1696, simple_loss=0.2609, pruned_loss=0.03921, over 4887.00 frames. ], tot_loss[loss=0.1674, simple_loss=0.2419, pruned_loss=0.04647, over 855634.02 frames. ], batch size: 35, lr: 2.81e-03, grad_scale: 16.0 2023-04-28 05:30:31,307 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=166570.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:30:37,715 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.2775, 1.7142, 2.1439, 2.2558, 2.0279, 1.7104, 1.1604, 1.8181], device='cuda:1'), covar=tensor([0.2950, 0.3088, 0.1557, 0.1955, 0.2419, 0.2665, 0.3959, 0.1825], device='cuda:1'), in_proj_covar=tensor([0.0293, 0.0248, 0.0229, 0.0316, 0.0223, 0.0238, 0.0230, 0.0186], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-28 05:30:39,258 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.011e+02 1.427e+02 1.810e+02 2.217e+02 8.020e+02, threshold=3.620e+02, percent-clipped=2.0 2023-04-28 05:30:47,079 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=166592.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:30:53,784 INFO [finetune.py:976] (1/7) Epoch 30, batch 500, loss[loss=0.1604, simple_loss=0.2347, pruned_loss=0.04308, over 4873.00 frames. ], tot_loss[loss=0.1645, simple_loss=0.2383, pruned_loss=0.04537, over 876884.78 frames. ], batch size: 34, lr: 2.81e-03, grad_scale: 16.0 2023-04-28 05:31:07,086 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4395, 1.2018, 4.1623, 3.9436, 3.6412, 3.9859, 3.9464, 3.7262], device='cuda:1'), covar=tensor([0.7486, 0.6115, 0.1125, 0.1693, 0.1266, 0.1953, 0.1450, 0.1577], device='cuda:1'), in_proj_covar=tensor([0.0312, 0.0311, 0.0409, 0.0412, 0.0353, 0.0416, 0.0319, 0.0365], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-28 05:31:27,596 INFO [finetune.py:976] (1/7) Epoch 30, batch 550, loss[loss=0.154, simple_loss=0.2301, pruned_loss=0.03894, over 4730.00 frames. ], tot_loss[loss=0.1636, simple_loss=0.2363, pruned_loss=0.04545, over 893967.94 frames. ], batch size: 23, lr: 2.81e-03, grad_scale: 16.0 2023-04-28 05:31:34,201 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=166663.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:31:45,971 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.022e+02 1.489e+02 1.745e+02 2.160e+02 4.947e+02, threshold=3.489e+02, percent-clipped=2.0 2023-04-28 05:32:01,358 INFO [finetune.py:976] (1/7) Epoch 30, batch 600, loss[loss=0.1349, simple_loss=0.2085, pruned_loss=0.03064, over 4797.00 frames. ], tot_loss[loss=0.1629, simple_loss=0.2357, pruned_loss=0.04506, over 908817.32 frames. ], batch size: 26, lr: 2.81e-03, grad_scale: 16.0 2023-04-28 05:32:05,746 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=166710.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 05:32:07,097 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-04-28 05:32:23,056 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=166735.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:32:34,881 INFO [finetune.py:976] (1/7) Epoch 30, batch 650, loss[loss=0.1597, simple_loss=0.2378, pruned_loss=0.04078, over 4874.00 frames. ], tot_loss[loss=0.1642, simple_loss=0.2379, pruned_loss=0.04528, over 918747.52 frames. ], batch size: 34, lr: 2.81e-03, grad_scale: 16.0 2023-04-28 05:32:37,945 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=166758.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 05:32:47,099 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=166764.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:32:55,012 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5837, 0.6830, 1.5338, 1.9367, 1.6459, 1.4796, 1.5103, 1.5103], device='cuda:1'), covar=tensor([0.4198, 0.6307, 0.5836, 0.5396, 0.5354, 0.6812, 0.7057, 0.8192], device='cuda:1'), in_proj_covar=tensor([0.0449, 0.0426, 0.0523, 0.0510, 0.0476, 0.0516, 0.0519, 0.0532], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-28 05:33:07,160 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.129e+02 1.523e+02 1.858e+02 2.285e+02 4.759e+02, threshold=3.715e+02, percent-clipped=1.0 2023-04-28 05:33:29,721 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=166796.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:33:39,477 INFO [finetune.py:976] (1/7) Epoch 30, batch 700, loss[loss=0.1706, simple_loss=0.2431, pruned_loss=0.04903, over 4931.00 frames. ], tot_loss[loss=0.1665, simple_loss=0.2407, pruned_loss=0.04617, over 927105.57 frames. ], batch size: 41, lr: 2.81e-03, grad_scale: 16.0 2023-04-28 05:33:39,581 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=166803.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:34:03,937 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=166825.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:34:22,682 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=166837.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:34:37,356 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=166851.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:34:43,121 INFO [finetune.py:976] (1/7) Epoch 30, batch 750, loss[loss=0.1645, simple_loss=0.2478, pruned_loss=0.04058, over 4844.00 frames. ], tot_loss[loss=0.1663, simple_loss=0.2406, pruned_loss=0.04597, over 932683.37 frames. ], batch size: 47, lr: 2.81e-03, grad_scale: 16.0 2023-04-28 05:34:51,920 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=166865.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:35:11,847 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.075e+02 1.441e+02 1.685e+02 2.007e+02 5.927e+02, threshold=3.370e+02, percent-clipped=1.0 2023-04-28 05:35:21,895 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=166887.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:35:44,639 INFO [finetune.py:976] (1/7) Epoch 30, batch 800, loss[loss=0.1836, simple_loss=0.2508, pruned_loss=0.05825, over 4820.00 frames. ], tot_loss[loss=0.1653, simple_loss=0.2396, pruned_loss=0.04543, over 938156.03 frames. ], batch size: 33, lr: 2.81e-03, grad_scale: 16.0 2023-04-28 05:36:17,511 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=166929.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:36:38,369 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([0.7265, 1.6565, 1.7857, 1.4170, 1.8514, 1.5338, 2.3198, 1.6407], device='cuda:1'), covar=tensor([0.3649, 0.1965, 0.4653, 0.2679, 0.1603, 0.2097, 0.1335, 0.4163], device='cuda:1'), in_proj_covar=tensor([0.0342, 0.0356, 0.0427, 0.0353, 0.0388, 0.0377, 0.0373, 0.0427], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-28 05:36:49,757 INFO [finetune.py:976] (1/7) Epoch 30, batch 850, loss[loss=0.1466, simple_loss=0.2201, pruned_loss=0.03653, over 4928.00 frames. ], tot_loss[loss=0.1647, simple_loss=0.2386, pruned_loss=0.04542, over 942751.69 frames. ], batch size: 33, lr: 2.81e-03, grad_scale: 16.0 2023-04-28 05:36:56,498 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=3.85 vs. limit=5.0 2023-04-28 05:36:59,947 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=166963.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:37:11,778 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0253, 2.4609, 1.0743, 1.2494, 1.9201, 1.1645, 3.2177, 1.4757], device='cuda:1'), covar=tensor([0.0730, 0.0620, 0.0809, 0.1328, 0.0514, 0.1085, 0.0250, 0.0724], device='cuda:1'), in_proj_covar=tensor([0.0051, 0.0064, 0.0047, 0.0046, 0.0049, 0.0051, 0.0073, 0.0051], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:1') 2023-04-28 05:37:20,968 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.032e+02 1.548e+02 1.839e+02 2.172e+02 6.436e+02, threshold=3.678e+02, percent-clipped=1.0 2023-04-28 05:37:33,076 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=166990.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:37:42,686 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.2732, 1.7323, 1.6400, 2.1651, 2.3340, 1.9881, 1.9540, 1.6973], device='cuda:1'), covar=tensor([0.1792, 0.1788, 0.1492, 0.1731, 0.1041, 0.1899, 0.1983, 0.2267], device='cuda:1'), in_proj_covar=tensor([0.0316, 0.0309, 0.0350, 0.0287, 0.0325, 0.0305, 0.0301, 0.0378], device='cuda:1'), out_proj_covar=tensor([6.4217e-05, 6.3167e-05, 7.3301e-05, 5.7265e-05, 6.6175e-05, 6.3244e-05, 6.1999e-05, 7.9834e-05], device='cuda:1') 2023-04-28 05:37:47,445 INFO [finetune.py:976] (1/7) Epoch 30, batch 900, loss[loss=0.1677, simple_loss=0.2494, pruned_loss=0.04294, over 4914.00 frames. ], tot_loss[loss=0.163, simple_loss=0.2365, pruned_loss=0.04481, over 946625.58 frames. ], batch size: 43, lr: 2.81e-03, grad_scale: 16.0 2023-04-28 05:37:52,360 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=167011.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:38:21,329 INFO [finetune.py:976] (1/7) Epoch 30, batch 950, loss[loss=0.1328, simple_loss=0.2108, pruned_loss=0.02744, over 4784.00 frames. ], tot_loss[loss=0.1626, simple_loss=0.2358, pruned_loss=0.0447, over 948436.60 frames. ], batch size: 28, lr: 2.81e-03, grad_scale: 32.0 2023-04-28 05:38:38,159 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.089e+02 1.627e+02 1.847e+02 2.138e+02 3.460e+02, threshold=3.694e+02, percent-clipped=0.0 2023-04-28 05:38:45,981 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=167091.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:38:52,937 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=167100.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:38:55,144 INFO [finetune.py:976] (1/7) Epoch 30, batch 1000, loss[loss=0.2035, simple_loss=0.2794, pruned_loss=0.06375, over 4750.00 frames. ], tot_loss[loss=0.1637, simple_loss=0.2373, pruned_loss=0.04503, over 948675.45 frames. ], batch size: 54, lr: 2.81e-03, grad_scale: 32.0 2023-04-28 05:38:55,864 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=167104.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:39:06,114 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=167120.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:39:17,576 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=167137.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:39:28,662 INFO [finetune.py:976] (1/7) Epoch 30, batch 1050, loss[loss=0.1919, simple_loss=0.2593, pruned_loss=0.06223, over 4818.00 frames. ], tot_loss[loss=0.165, simple_loss=0.2394, pruned_loss=0.04532, over 949305.89 frames. ], batch size: 30, lr: 2.81e-03, grad_scale: 32.0 2023-04-28 05:39:29,876 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.2071, 2.0111, 1.7511, 1.6948, 2.0950, 1.7757, 2.4840, 1.5807], device='cuda:1'), covar=tensor([0.3665, 0.2022, 0.4021, 0.3199, 0.1692, 0.2320, 0.1517, 0.4309], device='cuda:1'), in_proj_covar=tensor([0.0340, 0.0355, 0.0425, 0.0353, 0.0386, 0.0375, 0.0371, 0.0425], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-28 05:39:34,163 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=167161.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:39:36,562 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=167165.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:39:36,606 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=167165.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:39:46,131 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.104e+02 1.567e+02 1.873e+02 2.278e+02 3.738e+02, threshold=3.746e+02, percent-clipped=2.0 2023-04-28 05:39:49,156 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=167185.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:39:50,862 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=167187.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:40:01,670 INFO [finetune.py:976] (1/7) Epoch 30, batch 1100, loss[loss=0.1828, simple_loss=0.2541, pruned_loss=0.05579, over 4911.00 frames. ], tot_loss[loss=0.1662, simple_loss=0.2408, pruned_loss=0.04574, over 950894.36 frames. ], batch size: 36, lr: 2.81e-03, grad_scale: 32.0 2023-04-28 05:40:08,806 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=167213.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:40:08,873 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4767, 1.4610, 1.7690, 1.7538, 1.2674, 1.1756, 1.5115, 0.9328], device='cuda:1'), covar=tensor([0.0525, 0.0562, 0.0395, 0.0529, 0.0699, 0.1059, 0.0490, 0.0589], device='cuda:1'), in_proj_covar=tensor([0.0071, 0.0067, 0.0065, 0.0069, 0.0076, 0.0095, 0.0072, 0.0062], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-28 05:40:15,590 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9544, 1.0815, 1.6040, 1.6910, 1.6329, 1.7402, 1.6029, 1.5745], device='cuda:1'), covar=tensor([0.3732, 0.5007, 0.4196, 0.3851, 0.5001, 0.6460, 0.4421, 0.4379], device='cuda:1'), in_proj_covar=tensor([0.0344, 0.0374, 0.0330, 0.0342, 0.0351, 0.0393, 0.0361, 0.0334], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 05:40:20,278 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7618, 2.3060, 0.8055, 1.0735, 1.4164, 1.1285, 2.4666, 1.2001], device='cuda:1'), covar=tensor([0.0899, 0.0654, 0.0842, 0.1614, 0.0627, 0.1322, 0.0367, 0.0992], device='cuda:1'), in_proj_covar=tensor([0.0051, 0.0064, 0.0047, 0.0046, 0.0049, 0.0051, 0.0073, 0.0050], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:1') 2023-04-28 05:40:22,661 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=167235.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:40:34,490 INFO [finetune.py:976] (1/7) Epoch 30, batch 1150, loss[loss=0.1653, simple_loss=0.2451, pruned_loss=0.04273, over 4830.00 frames. ], tot_loss[loss=0.1662, simple_loss=0.2413, pruned_loss=0.04552, over 952439.29 frames. ], batch size: 47, lr: 2.81e-03, grad_scale: 32.0 2023-04-28 05:41:03,287 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.007e+02 1.590e+02 1.832e+02 2.168e+02 3.581e+02, threshold=3.663e+02, percent-clipped=0.0 2023-04-28 05:41:11,483 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=167285.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:41:33,517 INFO [finetune.py:976] (1/7) Epoch 30, batch 1200, loss[loss=0.1361, simple_loss=0.2267, pruned_loss=0.02274, over 4792.00 frames. ], tot_loss[loss=0.166, simple_loss=0.2404, pruned_loss=0.04586, over 951882.36 frames. ], batch size: 29, lr: 2.81e-03, grad_scale: 32.0 2023-04-28 05:41:43,257 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 2023-04-28 05:42:37,504 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-04-28 05:42:37,935 INFO [finetune.py:976] (1/7) Epoch 30, batch 1250, loss[loss=0.1733, simple_loss=0.2357, pruned_loss=0.05545, over 4932.00 frames. ], tot_loss[loss=0.1641, simple_loss=0.2375, pruned_loss=0.04535, over 953023.49 frames. ], batch size: 33, lr: 2.81e-03, grad_scale: 32.0 2023-04-28 05:42:38,154 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-04-28 05:42:39,888 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7524, 1.6294, 4.5267, 4.3121, 3.8871, 4.2606, 4.2006, 3.9707], device='cuda:1'), covar=tensor([0.6592, 0.5156, 0.1081, 0.1596, 0.1118, 0.1465, 0.1094, 0.1563], device='cuda:1'), in_proj_covar=tensor([0.0309, 0.0310, 0.0407, 0.0409, 0.0352, 0.0414, 0.0316, 0.0364], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-28 05:43:19,232 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 7.674e+01 1.528e+02 1.825e+02 2.243e+02 4.898e+02, threshold=3.650e+02, percent-clipped=3.0 2023-04-28 05:43:32,143 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=167391.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:43:42,854 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8955, 2.2418, 0.9213, 1.2557, 1.5292, 1.2258, 2.4671, 1.4124], device='cuda:1'), covar=tensor([0.0719, 0.0534, 0.0628, 0.1268, 0.0487, 0.0986, 0.0300, 0.0689], device='cuda:1'), in_proj_covar=tensor([0.0051, 0.0064, 0.0047, 0.0046, 0.0049, 0.0051, 0.0073, 0.0050], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:1') 2023-04-28 05:43:45,346 INFO [finetune.py:976] (1/7) Epoch 30, batch 1300, loss[loss=0.1906, simple_loss=0.2512, pruned_loss=0.06494, over 4796.00 frames. ], tot_loss[loss=0.1628, simple_loss=0.2357, pruned_loss=0.04497, over 954733.31 frames. ], batch size: 45, lr: 2.81e-03, grad_scale: 32.0 2023-04-28 05:44:12,871 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=167420.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:44:25,754 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=167432.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:44:35,809 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=167439.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:44:45,739 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([0.9962, 0.9916, 1.1595, 1.1397, 0.9800, 0.8821, 0.9566, 0.5009], device='cuda:1'), covar=tensor([0.0537, 0.0512, 0.0430, 0.0511, 0.0663, 0.1053, 0.0463, 0.0640], device='cuda:1'), in_proj_covar=tensor([0.0071, 0.0067, 0.0065, 0.0068, 0.0075, 0.0094, 0.0072, 0.0062], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-28 05:44:56,435 INFO [finetune.py:976] (1/7) Epoch 30, batch 1350, loss[loss=0.2291, simple_loss=0.291, pruned_loss=0.08361, over 4832.00 frames. ], tot_loss[loss=0.1615, simple_loss=0.2346, pruned_loss=0.04422, over 954624.50 frames. ], batch size: 47, lr: 2.81e-03, grad_scale: 32.0 2023-04-28 05:44:58,335 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=167456.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:45:00,774 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=167460.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:45:12,121 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=167468.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:45:25,586 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 8.814e+01 1.503e+02 1.795e+02 2.104e+02 5.763e+02, threshold=3.590e+02, percent-clipped=1.0 2023-04-28 05:45:34,218 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=167493.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:45:41,215 INFO [finetune.py:976] (1/7) Epoch 30, batch 1400, loss[loss=0.1872, simple_loss=0.2625, pruned_loss=0.056, over 4807.00 frames. ], tot_loss[loss=0.1636, simple_loss=0.2375, pruned_loss=0.04488, over 954480.20 frames. ], batch size: 45, lr: 2.81e-03, grad_scale: 32.0 2023-04-28 05:46:14,262 INFO [finetune.py:976] (1/7) Epoch 30, batch 1450, loss[loss=0.1528, simple_loss=0.2257, pruned_loss=0.03998, over 4783.00 frames. ], tot_loss[loss=0.1635, simple_loss=0.238, pruned_loss=0.04447, over 956595.86 frames. ], batch size: 51, lr: 2.81e-03, grad_scale: 32.0 2023-04-28 05:46:34,035 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5164, 2.5985, 2.0906, 2.2684, 2.5440, 2.1838, 3.4000, 1.9389], device='cuda:1'), covar=tensor([0.3811, 0.2489, 0.4340, 0.3281, 0.1997, 0.2626, 0.1649, 0.4494], device='cuda:1'), in_proj_covar=tensor([0.0337, 0.0352, 0.0420, 0.0349, 0.0382, 0.0371, 0.0369, 0.0421], device='cuda:1'), out_proj_covar=tensor([9.9381e-05, 1.0460e-04, 1.2700e-04, 1.0392e-04, 1.1299e-04, 1.1024e-04, 1.0730e-04, 1.2642e-04], device='cuda:1') 2023-04-28 05:46:52,053 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.057e+02 1.458e+02 1.734e+02 2.083e+02 3.176e+02, threshold=3.469e+02, percent-clipped=0.0 2023-04-28 05:47:00,606 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=167585.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:47:21,141 INFO [finetune.py:976] (1/7) Epoch 30, batch 1500, loss[loss=0.1265, simple_loss=0.196, pruned_loss=0.02846, over 4783.00 frames. ], tot_loss[loss=0.1651, simple_loss=0.2399, pruned_loss=0.04512, over 957165.74 frames. ], batch size: 25, lr: 2.81e-03, grad_scale: 32.0 2023-04-28 05:47:52,029 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-04-28 05:47:56,721 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=167633.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:48:20,795 INFO [finetune.py:976] (1/7) Epoch 30, batch 1550, loss[loss=0.155, simple_loss=0.2263, pruned_loss=0.04189, over 4816.00 frames. ], tot_loss[loss=0.1642, simple_loss=0.2392, pruned_loss=0.04461, over 957666.36 frames. ], batch size: 40, lr: 2.81e-03, grad_scale: 32.0 2023-04-28 05:48:40,247 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.357e+01 1.558e+02 1.802e+02 2.131e+02 3.986e+02, threshold=3.604e+02, percent-clipped=1.0 2023-04-28 05:48:54,269 INFO [finetune.py:976] (1/7) Epoch 30, batch 1600, loss[loss=0.1417, simple_loss=0.2086, pruned_loss=0.03738, over 4770.00 frames. ], tot_loss[loss=0.1626, simple_loss=0.2368, pruned_loss=0.0442, over 956689.76 frames. ], batch size: 28, lr: 2.81e-03, grad_scale: 32.0 2023-04-28 05:49:05,585 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=167719.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:49:28,176 INFO [finetune.py:976] (1/7) Epoch 30, batch 1650, loss[loss=0.1444, simple_loss=0.205, pruned_loss=0.04193, over 4740.00 frames. ], tot_loss[loss=0.1607, simple_loss=0.2346, pruned_loss=0.04343, over 959022.61 frames. ], batch size: 54, lr: 2.81e-03, grad_scale: 32.0 2023-04-28 05:49:30,080 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=167756.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:49:32,985 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=167760.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:49:37,863 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=167767.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:49:47,071 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.070e+02 1.494e+02 1.761e+02 2.054e+02 5.504e+02, threshold=3.522e+02, percent-clipped=6.0 2023-04-28 05:49:47,205 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=167780.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:49:51,118 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8921, 1.6325, 1.8068, 2.2460, 2.1981, 1.7504, 1.5592, 2.0126], device='cuda:1'), covar=tensor([0.0715, 0.1140, 0.0733, 0.0494, 0.0567, 0.0761, 0.0730, 0.0493], device='cuda:1'), in_proj_covar=tensor([0.0182, 0.0201, 0.0182, 0.0170, 0.0177, 0.0177, 0.0148, 0.0175], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-28 05:49:52,949 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=167788.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:50:02,176 INFO [finetune.py:976] (1/7) Epoch 30, batch 1700, loss[loss=0.1385, simple_loss=0.1983, pruned_loss=0.03937, over 4090.00 frames. ], tot_loss[loss=0.1588, simple_loss=0.232, pruned_loss=0.04275, over 958387.71 frames. ], batch size: 17, lr: 2.80e-03, grad_scale: 32.0 2023-04-28 05:50:02,850 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=167804.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:50:04,120 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.1842, 2.1070, 2.0104, 1.9378, 2.3471, 1.8541, 2.8735, 1.7401], device='cuda:1'), covar=tensor([0.3501, 0.1876, 0.4188, 0.2461, 0.1454, 0.2435, 0.1344, 0.4417], device='cuda:1'), in_proj_covar=tensor([0.0337, 0.0351, 0.0421, 0.0349, 0.0382, 0.0372, 0.0368, 0.0421], device='cuda:1'), out_proj_covar=tensor([9.9512e-05, 1.0453e-04, 1.2712e-04, 1.0390e-04, 1.1282e-04, 1.1035e-04, 1.0698e-04, 1.2623e-04], device='cuda:1') 2023-04-28 05:50:05,282 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=167808.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:50:18,853 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=167828.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:50:35,260 INFO [finetune.py:976] (1/7) Epoch 30, batch 1750, loss[loss=0.1874, simple_loss=0.2615, pruned_loss=0.05669, over 4892.00 frames. ], tot_loss[loss=0.1604, simple_loss=0.2343, pruned_loss=0.04331, over 957528.72 frames. ], batch size: 35, lr: 2.80e-03, grad_scale: 32.0 2023-04-28 05:50:40,295 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7478, 1.5121, 1.7043, 2.0210, 2.0944, 1.5904, 1.3066, 1.8379], device='cuda:1'), covar=tensor([0.0709, 0.1127, 0.0685, 0.0494, 0.0530, 0.0827, 0.0737, 0.0495], device='cuda:1'), in_proj_covar=tensor([0.0182, 0.0202, 0.0182, 0.0171, 0.0178, 0.0177, 0.0149, 0.0176], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-28 05:50:48,444 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6847, 3.7900, 0.9853, 1.8000, 2.0913, 2.7114, 2.0182, 1.1279], device='cuda:1'), covar=tensor([0.1398, 0.0885, 0.1979, 0.1309, 0.1177, 0.0994, 0.1457, 0.1941], device='cuda:1'), in_proj_covar=tensor([0.0119, 0.0241, 0.0138, 0.0122, 0.0134, 0.0155, 0.0120, 0.0119], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-28 05:50:50,300 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4941, 1.4233, 1.8743, 1.8443, 1.3411, 1.2176, 1.4554, 0.9635], device='cuda:1'), covar=tensor([0.0616, 0.0592, 0.0339, 0.0520, 0.0734, 0.1157, 0.0607, 0.0582], device='cuda:1'), in_proj_covar=tensor([0.0071, 0.0067, 0.0065, 0.0069, 0.0075, 0.0094, 0.0072, 0.0062], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-28 05:50:53,215 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.596e+01 1.560e+02 1.890e+02 2.217e+02 7.846e+02, threshold=3.780e+02, percent-clipped=0.0 2023-04-28 05:51:03,408 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.3752, 2.8200, 1.0669, 1.4948, 2.1244, 1.3672, 3.9416, 1.8378], device='cuda:1'), covar=tensor([0.0706, 0.0880, 0.0929, 0.1289, 0.0543, 0.1043, 0.0194, 0.0627], device='cuda:1'), in_proj_covar=tensor([0.0051, 0.0064, 0.0047, 0.0046, 0.0049, 0.0052, 0.0073, 0.0051], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:1') 2023-04-28 05:51:08,163 INFO [finetune.py:976] (1/7) Epoch 30, batch 1800, loss[loss=0.1342, simple_loss=0.215, pruned_loss=0.02674, over 4780.00 frames. ], tot_loss[loss=0.1634, simple_loss=0.2382, pruned_loss=0.04429, over 958123.47 frames. ], batch size: 29, lr: 2.80e-03, grad_scale: 32.0 2023-04-28 05:51:41,397 INFO [finetune.py:976] (1/7) Epoch 30, batch 1850, loss[loss=0.1752, simple_loss=0.2582, pruned_loss=0.04615, over 4819.00 frames. ], tot_loss[loss=0.1652, simple_loss=0.2402, pruned_loss=0.0451, over 957772.62 frames. ], batch size: 39, lr: 2.80e-03, grad_scale: 32.0 2023-04-28 05:52:04,613 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.089e+02 1.506e+02 1.821e+02 2.216e+02 4.142e+02, threshold=3.642e+02, percent-clipped=2.0 2023-04-28 05:52:26,788 INFO [finetune.py:976] (1/7) Epoch 30, batch 1900, loss[loss=0.1701, simple_loss=0.2579, pruned_loss=0.0412, over 4820.00 frames. ], tot_loss[loss=0.1653, simple_loss=0.2409, pruned_loss=0.0448, over 957182.35 frames. ], batch size: 33, lr: 2.80e-03, grad_scale: 32.0 2023-04-28 05:52:48,423 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([4.6175, 4.5482, 3.0343, 5.2556, 4.6143, 4.5468, 1.7835, 4.5686], device='cuda:1'), covar=tensor([0.1437, 0.0852, 0.3284, 0.0975, 0.2570, 0.1532, 0.5759, 0.1971], device='cuda:1'), in_proj_covar=tensor([0.0248, 0.0218, 0.0253, 0.0303, 0.0300, 0.0249, 0.0274, 0.0274], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 05:53:09,295 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=168035.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:53:09,896 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8228, 2.1826, 0.8749, 1.2024, 1.4741, 1.1253, 2.4935, 1.3629], device='cuda:1'), covar=tensor([0.0752, 0.0582, 0.0656, 0.1227, 0.0475, 0.1097, 0.0291, 0.0695], device='cuda:1'), in_proj_covar=tensor([0.0051, 0.0064, 0.0047, 0.0046, 0.0050, 0.0052, 0.0073, 0.0051], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:1') 2023-04-28 05:53:11,143 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0776, 2.3898, 1.0575, 1.3270, 1.8181, 1.2712, 3.0992, 1.5796], device='cuda:1'), covar=tensor([0.0727, 0.0602, 0.0751, 0.1377, 0.0490, 0.1069, 0.0254, 0.0649], device='cuda:1'), in_proj_covar=tensor([0.0051, 0.0064, 0.0047, 0.0046, 0.0050, 0.0052, 0.0073, 0.0051], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:1') 2023-04-28 05:53:31,849 INFO [finetune.py:976] (1/7) Epoch 30, batch 1950, loss[loss=0.1492, simple_loss=0.2217, pruned_loss=0.03829, over 4889.00 frames. ], tot_loss[loss=0.1646, simple_loss=0.2399, pruned_loss=0.04464, over 955561.39 frames. ], batch size: 43, lr: 2.80e-03, grad_scale: 32.0 2023-04-28 05:53:56,251 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=168075.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:54:04,670 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.011e+02 1.564e+02 1.911e+02 2.576e+02 9.918e+02, threshold=3.821e+02, percent-clipped=1.0 2023-04-28 05:54:13,523 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=168088.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:54:24,477 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=168096.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 05:54:32,404 INFO [finetune.py:976] (1/7) Epoch 30, batch 2000, loss[loss=0.1449, simple_loss=0.214, pruned_loss=0.03795, over 4392.00 frames. ], tot_loss[loss=0.1628, simple_loss=0.2372, pruned_loss=0.04421, over 956128.49 frames. ], batch size: 19, lr: 2.80e-03, grad_scale: 32.0 2023-04-28 05:54:55,109 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=168123.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:55:13,714 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=168136.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:55:30,830 INFO [finetune.py:976] (1/7) Epoch 30, batch 2050, loss[loss=0.1459, simple_loss=0.2167, pruned_loss=0.03757, over 4324.00 frames. ], tot_loss[loss=0.1604, simple_loss=0.234, pruned_loss=0.04336, over 954487.78 frames. ], batch size: 65, lr: 2.80e-03, grad_scale: 32.0 2023-04-28 05:55:40,693 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.1122, 2.6705, 2.1103, 2.5433, 1.8486, 2.3610, 2.2126, 1.6173], device='cuda:1'), covar=tensor([0.1835, 0.1149, 0.0843, 0.1080, 0.3283, 0.1047, 0.1915, 0.2758], device='cuda:1'), in_proj_covar=tensor([0.0279, 0.0298, 0.0216, 0.0271, 0.0306, 0.0250, 0.0245, 0.0258], device='cuda:1'), out_proj_covar=tensor([1.1114e-04, 1.1711e-04, 8.4729e-05, 1.0651e-04, 1.2303e-04, 9.8215e-05, 9.8635e-05, 1.0136e-04], device='cuda:1') 2023-04-28 05:55:43,765 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9084, 2.4966, 2.0102, 2.0209, 1.4015, 1.4576, 2.0293, 1.3496], device='cuda:1'), covar=tensor([0.1529, 0.1164, 0.1280, 0.1468, 0.2106, 0.1855, 0.0880, 0.1945], device='cuda:1'), in_proj_covar=tensor([0.0198, 0.0210, 0.0170, 0.0205, 0.0201, 0.0188, 0.0157, 0.0188], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-28 05:55:44,981 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8107, 1.6165, 1.7821, 2.1532, 2.2037, 1.7240, 1.4437, 1.9633], device='cuda:1'), covar=tensor([0.0793, 0.1266, 0.0778, 0.0586, 0.0588, 0.0850, 0.0702, 0.0539], device='cuda:1'), in_proj_covar=tensor([0.0181, 0.0201, 0.0181, 0.0169, 0.0176, 0.0176, 0.0148, 0.0174], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-28 05:55:47,312 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.348e+01 1.497e+02 1.684e+02 2.075e+02 3.378e+02, threshold=3.368e+02, percent-clipped=0.0 2023-04-28 05:56:04,271 INFO [finetune.py:976] (1/7) Epoch 30, batch 2100, loss[loss=0.1285, simple_loss=0.2042, pruned_loss=0.02635, over 4762.00 frames. ], tot_loss[loss=0.1602, simple_loss=0.2338, pruned_loss=0.04334, over 954365.61 frames. ], batch size: 26, lr: 2.80e-03, grad_scale: 32.0 2023-04-28 05:56:23,097 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=168233.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:56:37,578 INFO [finetune.py:976] (1/7) Epoch 30, batch 2150, loss[loss=0.1746, simple_loss=0.234, pruned_loss=0.0576, over 4839.00 frames. ], tot_loss[loss=0.1626, simple_loss=0.2371, pruned_loss=0.04404, over 955286.37 frames. ], batch size: 30, lr: 2.80e-03, grad_scale: 32.0 2023-04-28 05:56:54,477 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.048e+02 1.582e+02 1.868e+02 2.335e+02 3.831e+02, threshold=3.737e+02, percent-clipped=1.0 2023-04-28 05:57:04,547 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=168294.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:57:09,879 INFO [finetune.py:976] (1/7) Epoch 30, batch 2200, loss[loss=0.1907, simple_loss=0.2576, pruned_loss=0.06188, over 4814.00 frames. ], tot_loss[loss=0.165, simple_loss=0.2398, pruned_loss=0.04508, over 953017.61 frames. ], batch size: 38, lr: 2.80e-03, grad_scale: 32.0 2023-04-28 05:57:18,061 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([0.9974, 1.0211, 1.1685, 1.1674, 0.9810, 0.8616, 0.9301, 0.5455], device='cuda:1'), covar=tensor([0.0498, 0.0554, 0.0415, 0.0486, 0.0679, 0.1191, 0.0458, 0.0590], device='cuda:1'), in_proj_covar=tensor([0.0071, 0.0067, 0.0065, 0.0069, 0.0076, 0.0095, 0.0072, 0.0062], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-28 05:58:09,636 INFO [finetune.py:976] (1/7) Epoch 30, batch 2250, loss[loss=0.1573, simple_loss=0.2413, pruned_loss=0.03668, over 4876.00 frames. ], tot_loss[loss=0.1662, simple_loss=0.241, pruned_loss=0.04568, over 953295.40 frames. ], batch size: 35, lr: 2.80e-03, grad_scale: 32.0 2023-04-28 05:58:40,749 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=168375.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:58:43,714 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.119e+02 1.481e+02 1.883e+02 2.203e+02 3.675e+02, threshold=3.765e+02, percent-clipped=0.0 2023-04-28 05:58:55,049 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=168391.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 05:59:13,986 INFO [finetune.py:976] (1/7) Epoch 30, batch 2300, loss[loss=0.1543, simple_loss=0.2283, pruned_loss=0.04014, over 4838.00 frames. ], tot_loss[loss=0.1661, simple_loss=0.2414, pruned_loss=0.04543, over 954247.51 frames. ], batch size: 47, lr: 2.80e-03, grad_scale: 32.0 2023-04-28 05:59:23,242 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.4436, 2.1706, 2.3420, 2.7584, 2.7169, 2.3851, 2.0549, 2.5203], device='cuda:1'), covar=tensor([0.0637, 0.1013, 0.0655, 0.0443, 0.0520, 0.0710, 0.0595, 0.0469], device='cuda:1'), in_proj_covar=tensor([0.0181, 0.0201, 0.0182, 0.0169, 0.0176, 0.0176, 0.0148, 0.0175], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-28 05:59:38,050 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=168423.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:59:38,072 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=168423.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:00:16,725 INFO [finetune.py:976] (1/7) Epoch 30, batch 2350, loss[loss=0.1231, simple_loss=0.1998, pruned_loss=0.02319, over 4831.00 frames. ], tot_loss[loss=0.1636, simple_loss=0.238, pruned_loss=0.04461, over 952182.59 frames. ], batch size: 30, lr: 2.80e-03, grad_scale: 32.0 2023-04-28 06:00:37,861 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=168471.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:00:48,854 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 8.790e+01 1.485e+02 1.765e+02 2.221e+02 4.572e+02, threshold=3.529e+02, percent-clipped=1.0 2023-04-28 06:01:21,065 INFO [finetune.py:976] (1/7) Epoch 30, batch 2400, loss[loss=0.1272, simple_loss=0.198, pruned_loss=0.02821, over 4730.00 frames. ], tot_loss[loss=0.1629, simple_loss=0.2367, pruned_loss=0.04448, over 953433.56 frames. ], batch size: 23, lr: 2.80e-03, grad_scale: 32.0 2023-04-28 06:02:28,239 INFO [finetune.py:976] (1/7) Epoch 30, batch 2450, loss[loss=0.2028, simple_loss=0.2708, pruned_loss=0.06737, over 4910.00 frames. ], tot_loss[loss=0.1612, simple_loss=0.2342, pruned_loss=0.04411, over 953854.65 frames. ], batch size: 37, lr: 2.80e-03, grad_scale: 32.0 2023-04-28 06:02:40,503 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.3069, 1.6287, 1.4740, 1.8082, 1.7371, 1.8933, 1.4991, 3.5984], device='cuda:1'), covar=tensor([0.0636, 0.0785, 0.0766, 0.1221, 0.0618, 0.0518, 0.0723, 0.0137], device='cuda:1'), in_proj_covar=tensor([0.0037, 0.0038, 0.0039, 0.0043, 0.0040, 0.0037, 0.0038, 0.0054], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0013, 0.0013, 0.0015, 0.0014, 0.0013, 0.0013, 0.0016], device='cuda:1') 2023-04-28 06:03:10,380 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.038e+02 1.524e+02 1.793e+02 2.156e+02 5.076e+02, threshold=3.587e+02, percent-clipped=1.0 2023-04-28 06:03:21,581 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=168589.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:03:36,319 INFO [finetune.py:976] (1/7) Epoch 30, batch 2500, loss[loss=0.1685, simple_loss=0.2368, pruned_loss=0.0501, over 4936.00 frames. ], tot_loss[loss=0.1636, simple_loss=0.2367, pruned_loss=0.04526, over 953220.79 frames. ], batch size: 33, lr: 2.80e-03, grad_scale: 32.0 2023-04-28 06:04:06,210 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4452, 1.2989, 1.7191, 1.6976, 1.2702, 1.1478, 1.4076, 0.9087], device='cuda:1'), covar=tensor([0.0461, 0.0582, 0.0340, 0.0508, 0.0635, 0.1136, 0.0501, 0.0544], device='cuda:1'), in_proj_covar=tensor([0.0070, 0.0067, 0.0065, 0.0068, 0.0075, 0.0094, 0.0072, 0.0062], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-28 06:04:31,451 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.68 vs. limit=5.0 2023-04-28 06:04:49,840 INFO [finetune.py:976] (1/7) Epoch 30, batch 2550, loss[loss=0.1871, simple_loss=0.2627, pruned_loss=0.05577, over 4885.00 frames. ], tot_loss[loss=0.1673, simple_loss=0.2411, pruned_loss=0.04678, over 954194.39 frames. ], batch size: 32, lr: 2.80e-03, grad_scale: 32.0 2023-04-28 06:05:22,982 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5022, 1.2861, 0.5131, 1.2331, 1.3310, 1.3720, 1.3089, 1.2980], device='cuda:1'), covar=tensor([0.0569, 0.0351, 0.0376, 0.0596, 0.0278, 0.0601, 0.0597, 0.0633], device='cuda:1'), in_proj_covar=tensor([0.0027, 0.0023, 0.0021, 0.0028, 0.0019, 0.0027, 0.0027, 0.0028], device='cuda:1'), out_proj_covar=tensor([0.0052, 0.0046, 0.0039, 0.0053, 0.0039, 0.0051, 0.0051, 0.0053], device='cuda:1') 2023-04-28 06:05:23,461 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.119e+02 1.608e+02 1.886e+02 2.283e+02 4.830e+02, threshold=3.772e+02, percent-clipped=5.0 2023-04-28 06:05:31,405 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=168691.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:05:38,672 INFO [finetune.py:976] (1/7) Epoch 30, batch 2600, loss[loss=0.1709, simple_loss=0.2575, pruned_loss=0.0422, over 4791.00 frames. ], tot_loss[loss=0.168, simple_loss=0.2419, pruned_loss=0.04701, over 953857.15 frames. ], batch size: 45, lr: 2.80e-03, grad_scale: 32.0 2023-04-28 06:05:38,793 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9281, 2.4007, 2.0056, 2.3581, 1.7173, 2.0794, 1.9809, 1.4747], device='cuda:1'), covar=tensor([0.1812, 0.1099, 0.0820, 0.0979, 0.3341, 0.1068, 0.1774, 0.2521], device='cuda:1'), in_proj_covar=tensor([0.0279, 0.0296, 0.0215, 0.0269, 0.0305, 0.0250, 0.0244, 0.0256], device='cuda:1'), out_proj_covar=tensor([1.1094e-04, 1.1637e-04, 8.4232e-05, 1.0573e-04, 1.2264e-04, 9.8118e-05, 9.8084e-05, 1.0077e-04], device='cuda:1') 2023-04-28 06:06:13,777 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9159, 1.0483, 1.5979, 1.6896, 1.6626, 1.6763, 1.5824, 1.5595], device='cuda:1'), covar=tensor([0.3865, 0.5159, 0.3883, 0.4379, 0.5038, 0.6908, 0.4420, 0.4255], device='cuda:1'), in_proj_covar=tensor([0.0346, 0.0376, 0.0332, 0.0344, 0.0353, 0.0396, 0.0363, 0.0336], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 06:06:26,239 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=168739.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:06:40,211 INFO [finetune.py:976] (1/7) Epoch 30, batch 2650, loss[loss=0.1754, simple_loss=0.2259, pruned_loss=0.06248, over 4118.00 frames. ], tot_loss[loss=0.1681, simple_loss=0.2422, pruned_loss=0.04697, over 952288.73 frames. ], batch size: 17, lr: 2.80e-03, grad_scale: 32.0 2023-04-28 06:07:17,598 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.117e+02 1.537e+02 1.777e+02 2.075e+02 3.537e+02, threshold=3.553e+02, percent-clipped=0.0 2023-04-28 06:07:17,689 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.6194, 2.5619, 2.1622, 3.0255, 2.6176, 2.7034, 1.1465, 2.5999], device='cuda:1'), covar=tensor([0.2075, 0.1675, 0.3164, 0.2452, 0.3885, 0.2103, 0.5511, 0.2749], device='cuda:1'), in_proj_covar=tensor([0.0249, 0.0219, 0.0255, 0.0304, 0.0301, 0.0251, 0.0277, 0.0275], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 06:07:44,072 INFO [finetune.py:976] (1/7) Epoch 30, batch 2700, loss[loss=0.1226, simple_loss=0.1963, pruned_loss=0.02445, over 4817.00 frames. ], tot_loss[loss=0.1663, simple_loss=0.2405, pruned_loss=0.04606, over 952852.91 frames. ], batch size: 39, lr: 2.80e-03, grad_scale: 32.0 2023-04-28 06:07:54,410 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4309, 1.1930, 0.4172, 1.2015, 1.1787, 1.2981, 1.2458, 1.2161], device='cuda:1'), covar=tensor([0.0617, 0.0380, 0.0430, 0.0624, 0.0337, 0.0642, 0.0644, 0.0701], device='cuda:1'), in_proj_covar=tensor([0.0028, 0.0023, 0.0022, 0.0028, 0.0019, 0.0027, 0.0027, 0.0028], device='cuda:1'), out_proj_covar=tensor([0.0052, 0.0046, 0.0039, 0.0053, 0.0039, 0.0051, 0.0051, 0.0053], device='cuda:1') 2023-04-28 06:08:10,423 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8842, 1.3412, 4.8186, 4.5343, 4.1742, 4.5591, 4.2944, 4.2705], device='cuda:1'), covar=tensor([0.7229, 0.6417, 0.1276, 0.2265, 0.1178, 0.1961, 0.1735, 0.1842], device='cuda:1'), in_proj_covar=tensor([0.0312, 0.0311, 0.0408, 0.0412, 0.0351, 0.0417, 0.0316, 0.0365], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-28 06:08:17,674 INFO [finetune.py:976] (1/7) Epoch 30, batch 2750, loss[loss=0.1353, simple_loss=0.2164, pruned_loss=0.02712, over 4757.00 frames. ], tot_loss[loss=0.1642, simple_loss=0.2377, pruned_loss=0.04536, over 953925.64 frames. ], batch size: 27, lr: 2.80e-03, grad_scale: 32.0 2023-04-28 06:08:32,433 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9029, 1.4049, 1.5488, 1.5999, 2.0195, 1.7091, 1.4031, 1.4975], device='cuda:1'), covar=tensor([0.1740, 0.1625, 0.1680, 0.1326, 0.0823, 0.1648, 0.2327, 0.2254], device='cuda:1'), in_proj_covar=tensor([0.0313, 0.0307, 0.0349, 0.0285, 0.0324, 0.0303, 0.0299, 0.0375], device='cuda:1'), out_proj_covar=tensor([6.3641e-05, 6.2566e-05, 7.2860e-05, 5.6774e-05, 6.5926e-05, 6.2776e-05, 6.1608e-05, 7.9165e-05], device='cuda:1') 2023-04-28 06:08:35,309 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.781e+01 1.528e+02 1.886e+02 2.380e+02 6.138e+02, threshold=3.773e+02, percent-clipped=2.0 2023-04-28 06:08:41,775 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=168889.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:08:44,204 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.67 vs. limit=5.0 2023-04-28 06:08:50,765 INFO [finetune.py:976] (1/7) Epoch 30, batch 2800, loss[loss=0.1105, simple_loss=0.1737, pruned_loss=0.0236, over 3880.00 frames. ], tot_loss[loss=0.1612, simple_loss=0.2339, pruned_loss=0.04423, over 951924.56 frames. ], batch size: 16, lr: 2.80e-03, grad_scale: 32.0 2023-04-28 06:09:12,961 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=168937.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:09:24,482 INFO [finetune.py:976] (1/7) Epoch 30, batch 2850, loss[loss=0.1786, simple_loss=0.2477, pruned_loss=0.05475, over 4821.00 frames. ], tot_loss[loss=0.1607, simple_loss=0.2327, pruned_loss=0.04435, over 953572.92 frames. ], batch size: 30, lr: 2.80e-03, grad_scale: 32.0 2023-04-28 06:09:33,141 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=168967.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:09:34,350 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=168969.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 06:09:41,981 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.589e+01 1.497e+02 1.758e+02 2.128e+02 6.037e+02, threshold=3.517e+02, percent-clipped=2.0 2023-04-28 06:09:58,542 INFO [finetune.py:976] (1/7) Epoch 30, batch 2900, loss[loss=0.1544, simple_loss=0.2398, pruned_loss=0.03446, over 4820.00 frames. ], tot_loss[loss=0.1644, simple_loss=0.2367, pruned_loss=0.04603, over 951648.20 frames. ], batch size: 39, lr: 2.80e-03, grad_scale: 32.0 2023-04-28 06:10:02,970 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5852, 1.6974, 0.7063, 1.2579, 1.7502, 1.4429, 1.3310, 1.4425], device='cuda:1'), covar=tensor([0.0485, 0.0361, 0.0342, 0.0557, 0.0260, 0.0505, 0.0495, 0.0556], device='cuda:1'), in_proj_covar=tensor([0.0027, 0.0023, 0.0021, 0.0028, 0.0019, 0.0027, 0.0027, 0.0028], device='cuda:1'), out_proj_covar=tensor([0.0052, 0.0046, 0.0039, 0.0053, 0.0039, 0.0051, 0.0051, 0.0053], device='cuda:1') 2023-04-28 06:10:08,486 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7775, 1.3155, 1.8534, 2.2415, 1.8163, 1.7026, 1.7793, 1.7566], device='cuda:1'), covar=tensor([0.4507, 0.6939, 0.6429, 0.5499, 0.5923, 0.7746, 0.8106, 0.9279], device='cuda:1'), in_proj_covar=tensor([0.0448, 0.0426, 0.0519, 0.0507, 0.0474, 0.0514, 0.0514, 0.0529], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-28 06:10:10,247 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.45 vs. limit=2.0 2023-04-28 06:10:14,815 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=169028.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:10:16,014 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=169030.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 06:10:31,852 INFO [finetune.py:976] (1/7) Epoch 30, batch 2950, loss[loss=0.1702, simple_loss=0.2481, pruned_loss=0.04616, over 4750.00 frames. ], tot_loss[loss=0.1673, simple_loss=0.2403, pruned_loss=0.0471, over 951802.00 frames. ], batch size: 54, lr: 2.80e-03, grad_scale: 32.0 2023-04-28 06:10:43,664 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.70 vs. limit=2.0 2023-04-28 06:10:49,916 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.313e+01 1.646e+02 1.860e+02 2.214e+02 4.123e+02, threshold=3.720e+02, percent-clipped=2.0 2023-04-28 06:10:50,646 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=169082.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:11:05,768 INFO [finetune.py:976] (1/7) Epoch 30, batch 3000, loss[loss=0.2713, simple_loss=0.3254, pruned_loss=0.1086, over 4112.00 frames. ], tot_loss[loss=0.1684, simple_loss=0.242, pruned_loss=0.04743, over 951200.15 frames. ], batch size: 65, lr: 2.80e-03, grad_scale: 32.0 2023-04-28 06:11:05,768 INFO [finetune.py:1001] (1/7) Computing validation loss 2023-04-28 06:11:10,978 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8469, 2.1777, 1.8262, 1.5681, 1.3902, 1.4349, 1.8124, 1.3798], device='cuda:1'), covar=tensor([0.1787, 0.1260, 0.1480, 0.1684, 0.2392, 0.1996, 0.1052, 0.2043], device='cuda:1'), in_proj_covar=tensor([0.0200, 0.0210, 0.0171, 0.0205, 0.0202, 0.0189, 0.0158, 0.0188], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-28 06:11:16,542 INFO [finetune.py:1010] (1/7) Epoch 30, validation: loss=0.1534, simple_loss=0.2215, pruned_loss=0.04259, over 2265189.00 frames. 2023-04-28 06:11:16,543 INFO [finetune.py:1011] (1/7) Maximum memory allocated so far is 6391MB 2023-04-28 06:11:24,701 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=169116.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:11:52,086 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=169143.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:12:04,394 INFO [finetune.py:976] (1/7) Epoch 30, batch 3050, loss[loss=0.149, simple_loss=0.2213, pruned_loss=0.0383, over 4806.00 frames. ], tot_loss[loss=0.1669, simple_loss=0.2416, pruned_loss=0.04611, over 952648.41 frames. ], batch size: 39, lr: 2.80e-03, grad_scale: 32.0 2023-04-28 06:12:13,893 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=169159.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:12:36,608 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=169177.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:12:42,082 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.026e+02 1.505e+02 1.782e+02 2.187e+02 4.062e+02, threshold=3.563e+02, percent-clipped=1.0 2023-04-28 06:13:06,207 INFO [finetune.py:976] (1/7) Epoch 30, batch 3100, loss[loss=0.1287, simple_loss=0.2035, pruned_loss=0.02692, over 4813.00 frames. ], tot_loss[loss=0.1645, simple_loss=0.2385, pruned_loss=0.04524, over 953344.14 frames. ], batch size: 41, lr: 2.80e-03, grad_scale: 32.0 2023-04-28 06:13:29,659 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=169220.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:13:38,180 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.2243, 1.3911, 1.7359, 1.8443, 1.6970, 1.8066, 1.7185, 1.7292], device='cuda:1'), covar=tensor([0.3499, 0.4840, 0.3790, 0.3864, 0.5078, 0.6575, 0.4357, 0.4518], device='cuda:1'), in_proj_covar=tensor([0.0342, 0.0372, 0.0329, 0.0341, 0.0349, 0.0392, 0.0360, 0.0334], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 06:13:50,608 INFO [finetune.py:976] (1/7) Epoch 30, batch 3150, loss[loss=0.1877, simple_loss=0.2652, pruned_loss=0.05509, over 4810.00 frames. ], tot_loss[loss=0.164, simple_loss=0.2369, pruned_loss=0.04551, over 953893.94 frames. ], batch size: 45, lr: 2.80e-03, grad_scale: 32.0 2023-04-28 06:14:10,082 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.051e+01 1.353e+02 1.739e+02 2.048e+02 6.582e+02, threshold=3.478e+02, percent-clipped=1.0 2023-04-28 06:14:23,996 INFO [finetune.py:976] (1/7) Epoch 30, batch 3200, loss[loss=0.1244, simple_loss=0.212, pruned_loss=0.01842, over 4910.00 frames. ], tot_loss[loss=0.1602, simple_loss=0.2331, pruned_loss=0.04367, over 955862.84 frames. ], batch size: 35, lr: 2.80e-03, grad_scale: 32.0 2023-04-28 06:14:26,113 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-04-28 06:14:32,083 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-04-28 06:14:38,827 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=169323.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:14:40,569 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=169325.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 06:14:43,043 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0327, 1.6305, 1.9085, 2.3334, 2.4907, 1.8345, 1.7103, 2.1582], device='cuda:1'), covar=tensor([0.0754, 0.1434, 0.0900, 0.0578, 0.0570, 0.0949, 0.0750, 0.0582], device='cuda:1'), in_proj_covar=tensor([0.0185, 0.0205, 0.0185, 0.0172, 0.0179, 0.0180, 0.0151, 0.0178], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-28 06:14:48,506 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6527, 3.6372, 0.9043, 2.0097, 1.9479, 2.5158, 1.9104, 1.0208], device='cuda:1'), covar=tensor([0.1358, 0.0828, 0.2044, 0.1194, 0.1077, 0.1024, 0.1606, 0.1934], device='cuda:1'), in_proj_covar=tensor([0.0117, 0.0237, 0.0135, 0.0120, 0.0131, 0.0153, 0.0117, 0.0117], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-28 06:14:57,967 INFO [finetune.py:976] (1/7) Epoch 30, batch 3250, loss[loss=0.1493, simple_loss=0.228, pruned_loss=0.03534, over 4771.00 frames. ], tot_loss[loss=0.1606, simple_loss=0.2336, pruned_loss=0.04378, over 956103.43 frames. ], batch size: 26, lr: 2.80e-03, grad_scale: 32.0 2023-04-28 06:15:18,031 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.661e+01 1.547e+02 1.939e+02 2.313e+02 4.246e+02, threshold=3.878e+02, percent-clipped=2.0 2023-04-28 06:15:32,127 INFO [finetune.py:976] (1/7) Epoch 30, batch 3300, loss[loss=0.2035, simple_loss=0.272, pruned_loss=0.06752, over 4852.00 frames. ], tot_loss[loss=0.1647, simple_loss=0.2385, pruned_loss=0.04547, over 953722.45 frames. ], batch size: 44, lr: 2.79e-03, grad_scale: 32.0 2023-04-28 06:15:56,489 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=169438.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:16:05,662 INFO [finetune.py:976] (1/7) Epoch 30, batch 3350, loss[loss=0.1441, simple_loss=0.2251, pruned_loss=0.03153, over 4850.00 frames. ], tot_loss[loss=0.166, simple_loss=0.2405, pruned_loss=0.04571, over 953373.92 frames. ], batch size: 49, lr: 2.79e-03, grad_scale: 32.0 2023-04-28 06:16:05,881 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-04-28 06:16:11,670 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([0.7845, 1.7116, 1.6872, 1.4172, 1.8011, 1.5863, 2.2768, 1.4786], device='cuda:1'), covar=tensor([0.3316, 0.1877, 0.4427, 0.2600, 0.1456, 0.2131, 0.1362, 0.4468], device='cuda:1'), in_proj_covar=tensor([0.0340, 0.0357, 0.0427, 0.0353, 0.0386, 0.0376, 0.0372, 0.0427], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-28 06:16:15,934 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=6.23 vs. limit=5.0 2023-04-28 06:16:18,729 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=169472.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:16:26,057 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.059e+02 1.565e+02 1.911e+02 2.262e+02 9.005e+02, threshold=3.822e+02, percent-clipped=1.0 2023-04-28 06:16:30,458 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7528, 1.2294, 1.8454, 2.1626, 1.8115, 1.7224, 1.7795, 1.7919], device='cuda:1'), covar=tensor([0.4728, 0.7145, 0.6353, 0.6146, 0.6239, 0.8273, 0.8939, 0.9715], device='cuda:1'), in_proj_covar=tensor([0.0447, 0.0426, 0.0520, 0.0507, 0.0473, 0.0514, 0.0514, 0.0530], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-28 06:16:39,411 INFO [finetune.py:976] (1/7) Epoch 30, batch 3400, loss[loss=0.2172, simple_loss=0.2945, pruned_loss=0.06996, over 4890.00 frames. ], tot_loss[loss=0.1672, simple_loss=0.2418, pruned_loss=0.04625, over 954031.37 frames. ], batch size: 43, lr: 2.79e-03, grad_scale: 32.0 2023-04-28 06:16:47,268 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=169515.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:17:12,721 INFO [finetune.py:976] (1/7) Epoch 30, batch 3450, loss[loss=0.1306, simple_loss=0.2154, pruned_loss=0.02286, over 4751.00 frames. ], tot_loss[loss=0.1655, simple_loss=0.2405, pruned_loss=0.04525, over 952810.47 frames. ], batch size: 27, lr: 2.79e-03, grad_scale: 32.0 2023-04-28 06:17:15,233 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8525, 1.7839, 2.2976, 2.6274, 1.6797, 1.4188, 1.6409, 0.8450], device='cuda:1'), covar=tensor([0.0589, 0.0579, 0.0386, 0.0521, 0.0641, 0.1438, 0.0746, 0.0872], device='cuda:1'), in_proj_covar=tensor([0.0071, 0.0067, 0.0065, 0.0068, 0.0075, 0.0094, 0.0072, 0.0062], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-28 06:17:21,147 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6258, 1.4670, 1.8667, 1.8763, 1.4730, 1.3628, 1.4846, 0.9466], device='cuda:1'), covar=tensor([0.0502, 0.0616, 0.0352, 0.0495, 0.0703, 0.1021, 0.0507, 0.0508], device='cuda:1'), in_proj_covar=tensor([0.0071, 0.0067, 0.0065, 0.0068, 0.0075, 0.0094, 0.0072, 0.0062], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-04-28 06:17:36,961 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.050e+02 1.504e+02 1.780e+02 2.155e+02 3.271e+02, threshold=3.560e+02, percent-clipped=0.0 2023-04-28 06:18:07,736 INFO [finetune.py:976] (1/7) Epoch 30, batch 3500, loss[loss=0.1989, simple_loss=0.2582, pruned_loss=0.06978, over 4822.00 frames. ], tot_loss[loss=0.1638, simple_loss=0.2384, pruned_loss=0.04457, over 953595.14 frames. ], batch size: 30, lr: 2.79e-03, grad_scale: 32.0 2023-04-28 06:18:29,317 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=169621.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:18:31,005 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=169623.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:18:32,263 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=169625.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 06:19:13,502 INFO [finetune.py:976] (1/7) Epoch 30, batch 3550, loss[loss=0.1859, simple_loss=0.2484, pruned_loss=0.0617, over 4775.00 frames. ], tot_loss[loss=0.1611, simple_loss=0.2351, pruned_loss=0.04358, over 954013.00 frames. ], batch size: 59, lr: 2.79e-03, grad_scale: 32.0 2023-04-28 06:19:25,259 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.2265, 4.3135, 0.8439, 2.3248, 2.6414, 2.7758, 2.4604, 1.1226], device='cuda:1'), covar=tensor([0.1248, 0.1105, 0.2187, 0.1251, 0.0967, 0.1239, 0.1499, 0.2136], device='cuda:1'), in_proj_covar=tensor([0.0118, 0.0238, 0.0136, 0.0121, 0.0131, 0.0154, 0.0118, 0.0118], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-28 06:19:30,804 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=169671.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:19:32,048 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=169673.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 06:19:42,308 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.126e+01 1.430e+02 1.728e+02 2.038e+02 3.209e+02, threshold=3.455e+02, percent-clipped=0.0 2023-04-28 06:19:43,558 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=169682.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:20:12,670 INFO [finetune.py:976] (1/7) Epoch 30, batch 3600, loss[loss=0.1589, simple_loss=0.2286, pruned_loss=0.04458, over 4923.00 frames. ], tot_loss[loss=0.16, simple_loss=0.2329, pruned_loss=0.04353, over 954835.98 frames. ], batch size: 36, lr: 2.79e-03, grad_scale: 32.0 2023-04-28 06:20:34,875 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8260, 1.4102, 1.9533, 2.3591, 1.9312, 1.8174, 1.9174, 1.8358], device='cuda:1'), covar=tensor([0.4605, 0.7184, 0.6684, 0.5575, 0.6193, 0.7904, 0.8193, 0.9166], device='cuda:1'), in_proj_covar=tensor([0.0447, 0.0428, 0.0523, 0.0508, 0.0476, 0.0516, 0.0516, 0.0532], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-28 06:20:45,855 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=169729.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:20:57,516 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=169738.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:21:10,882 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=169747.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:21:20,087 INFO [finetune.py:976] (1/7) Epoch 30, batch 3650, loss[loss=0.1885, simple_loss=0.2726, pruned_loss=0.05221, over 4874.00 frames. ], tot_loss[loss=0.1607, simple_loss=0.2339, pruned_loss=0.04374, over 954080.87 frames. ], batch size: 34, lr: 2.79e-03, grad_scale: 32.0 2023-04-28 06:21:42,870 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.52 vs. limit=2.0 2023-04-28 06:21:43,273 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=169772.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:21:53,879 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.801e+01 1.645e+02 1.930e+02 2.224e+02 4.571e+02, threshold=3.861e+02, percent-clipped=2.0 2023-04-28 06:22:03,120 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=169786.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:22:03,192 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=169786.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:22:11,307 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=169790.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:22:25,954 INFO [finetune.py:976] (1/7) Epoch 30, batch 3700, loss[loss=0.195, simple_loss=0.2714, pruned_loss=0.05934, over 4897.00 frames. ], tot_loss[loss=0.1642, simple_loss=0.2379, pruned_loss=0.04521, over 953440.82 frames. ], batch size: 43, lr: 2.79e-03, grad_scale: 32.0 2023-04-28 06:22:35,254 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=169808.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:22:44,936 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=169815.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:22:47,945 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=169820.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:23:22,314 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=169847.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:23:31,472 INFO [finetune.py:976] (1/7) Epoch 30, batch 3750, loss[loss=0.2053, simple_loss=0.2823, pruned_loss=0.06418, over 4861.00 frames. ], tot_loss[loss=0.1644, simple_loss=0.2383, pruned_loss=0.04522, over 954974.34 frames. ], batch size: 34, lr: 2.79e-03, grad_scale: 32.0 2023-04-28 06:23:42,221 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=169863.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:24:04,805 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.130e+02 1.513e+02 1.804e+02 2.185e+02 4.010e+02, threshold=3.607e+02, percent-clipped=1.0 2023-04-28 06:24:23,986 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5800, 1.6325, 0.7770, 1.2925, 1.8557, 1.4378, 1.3194, 1.4438], device='cuda:1'), covar=tensor([0.0502, 0.0363, 0.0331, 0.0546, 0.0249, 0.0488, 0.0494, 0.0571], device='cuda:1'), in_proj_covar=tensor([0.0027, 0.0023, 0.0021, 0.0028, 0.0019, 0.0027, 0.0027, 0.0028], device='cuda:1'), out_proj_covar=tensor([0.0052, 0.0046, 0.0039, 0.0053, 0.0039, 0.0051, 0.0051, 0.0053], device='cuda:1') 2023-04-28 06:24:36,631 INFO [finetune.py:976] (1/7) Epoch 30, batch 3800, loss[loss=0.201, simple_loss=0.274, pruned_loss=0.064, over 4823.00 frames. ], tot_loss[loss=0.1652, simple_loss=0.2397, pruned_loss=0.04539, over 955121.06 frames. ], batch size: 47, lr: 2.79e-03, grad_scale: 32.0 2023-04-28 06:25:40,863 INFO [finetune.py:976] (1/7) Epoch 30, batch 3850, loss[loss=0.1827, simple_loss=0.2434, pruned_loss=0.061, over 4909.00 frames. ], tot_loss[loss=0.1645, simple_loss=0.2391, pruned_loss=0.04488, over 954010.75 frames. ], batch size: 37, lr: 2.79e-03, grad_scale: 32.0 2023-04-28 06:25:50,580 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-04-28 06:26:12,315 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=169977.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:26:14,681 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.720e+01 1.456e+02 1.655e+02 1.971e+02 4.065e+02, threshold=3.309e+02, percent-clipped=1.0 2023-04-28 06:26:43,558 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9396, 2.5183, 1.9761, 1.9779, 1.4192, 1.4925, 2.0900, 1.4329], device='cuda:1'), covar=tensor([0.1652, 0.1305, 0.1355, 0.1589, 0.2165, 0.1842, 0.0943, 0.1949], device='cuda:1'), in_proj_covar=tensor([0.0198, 0.0208, 0.0169, 0.0202, 0.0199, 0.0186, 0.0156, 0.0186], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-28 06:26:45,748 INFO [finetune.py:976] (1/7) Epoch 30, batch 3900, loss[loss=0.2159, simple_loss=0.2692, pruned_loss=0.08135, over 4933.00 frames. ], tot_loss[loss=0.163, simple_loss=0.2366, pruned_loss=0.04469, over 955890.12 frames. ], batch size: 38, lr: 2.79e-03, grad_scale: 32.0 2023-04-28 06:27:02,647 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.1343, 1.3724, 1.2078, 1.6194, 1.4898, 1.5194, 1.3217, 2.4192], device='cuda:1'), covar=tensor([0.0561, 0.0785, 0.0777, 0.1159, 0.0615, 0.0481, 0.0699, 0.0214], device='cuda:1'), in_proj_covar=tensor([0.0037, 0.0039, 0.0039, 0.0043, 0.0040, 0.0037, 0.0038, 0.0055], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0013, 0.0014, 0.0015, 0.0014, 0.0013, 0.0013, 0.0016], device='cuda:1') 2023-04-28 06:27:25,755 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([0.5052, 1.3438, 1.3635, 1.0217, 1.3617, 1.1974, 1.6172, 1.3827], device='cuda:1'), covar=tensor([0.2889, 0.1684, 0.3958, 0.2179, 0.1297, 0.2064, 0.1450, 0.3710], device='cuda:1'), in_proj_covar=tensor([0.0339, 0.0356, 0.0426, 0.0352, 0.0384, 0.0375, 0.0370, 0.0424], device='cuda:1'), out_proj_covar=tensor([9.9879e-05, 1.0584e-04, 1.2881e-04, 1.0475e-04, 1.1338e-04, 1.1123e-04, 1.0746e-04, 1.2714e-04], device='cuda:1') 2023-04-28 06:27:36,571 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-04-28 06:27:40,100 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=170045.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:27:48,653 INFO [finetune.py:976] (1/7) Epoch 30, batch 3950, loss[loss=0.1539, simple_loss=0.2218, pruned_loss=0.04306, over 4832.00 frames. ], tot_loss[loss=0.1611, simple_loss=0.2343, pruned_loss=0.04395, over 956850.67 frames. ], batch size: 33, lr: 2.79e-03, grad_scale: 32.0 2023-04-28 06:27:53,449 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5647, 2.5896, 2.0820, 2.2587, 2.4791, 2.2238, 3.2633, 1.8903], device='cuda:1'), covar=tensor([0.3697, 0.2237, 0.4065, 0.3414, 0.1998, 0.2582, 0.1921, 0.4725], device='cuda:1'), in_proj_covar=tensor([0.0339, 0.0356, 0.0426, 0.0352, 0.0384, 0.0375, 0.0370, 0.0424], device='cuda:1'), out_proj_covar=tensor([9.9975e-05, 1.0580e-04, 1.2888e-04, 1.0488e-04, 1.1349e-04, 1.1125e-04, 1.0749e-04, 1.2720e-04], device='cuda:1') 2023-04-28 06:28:24,768 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.030e+02 1.459e+02 1.752e+02 2.144e+02 4.109e+02, threshold=3.505e+02, percent-clipped=2.0 2023-04-28 06:28:27,239 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=170085.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:28:46,917 INFO [finetune.py:976] (1/7) Epoch 30, batch 4000, loss[loss=0.1541, simple_loss=0.239, pruned_loss=0.03461, over 4821.00 frames. ], tot_loss[loss=0.1622, simple_loss=0.2352, pruned_loss=0.04459, over 956210.11 frames. ], batch size: 45, lr: 2.79e-03, grad_scale: 32.0 2023-04-28 06:28:46,981 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=170103.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:28:53,802 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=170106.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:29:26,533 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8831, 4.0463, 0.7459, 2.2637, 2.3593, 2.6942, 2.4367, 0.8695], device='cuda:1'), covar=tensor([0.1200, 0.0750, 0.2035, 0.1146, 0.0948, 0.0996, 0.1295, 0.2136], device='cuda:1'), in_proj_covar=tensor([0.0116, 0.0236, 0.0135, 0.0119, 0.0130, 0.0152, 0.0117, 0.0117], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-28 06:29:39,696 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=170142.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:29:50,077 INFO [finetune.py:976] (1/7) Epoch 30, batch 4050, loss[loss=0.1709, simple_loss=0.2723, pruned_loss=0.03469, over 4824.00 frames. ], tot_loss[loss=0.1651, simple_loss=0.2389, pruned_loss=0.04567, over 954955.09 frames. ], batch size: 49, lr: 2.79e-03, grad_scale: 32.0 2023-04-28 06:30:17,025 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 2023-04-28 06:30:22,175 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.182e+02 1.558e+02 1.902e+02 2.221e+02 4.560e+02, threshold=3.805e+02, percent-clipped=3.0 2023-04-28 06:30:36,425 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.63 vs. limit=5.0 2023-04-28 06:30:36,789 INFO [finetune.py:976] (1/7) Epoch 30, batch 4100, loss[loss=0.1742, simple_loss=0.2546, pruned_loss=0.04689, over 4822.00 frames. ], tot_loss[loss=0.1668, simple_loss=0.2415, pruned_loss=0.04609, over 954932.65 frames. ], batch size: 39, lr: 2.79e-03, grad_scale: 32.0 2023-04-28 06:30:58,756 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.2255, 1.7272, 1.5175, 2.0744, 2.2720, 1.9050, 1.8676, 1.5243], device='cuda:1'), covar=tensor([0.2047, 0.1899, 0.2017, 0.2039, 0.1138, 0.2161, 0.2321, 0.2375], device='cuda:1'), in_proj_covar=tensor([0.0315, 0.0309, 0.0352, 0.0288, 0.0325, 0.0308, 0.0301, 0.0377], device='cuda:1'), out_proj_covar=tensor([6.3984e-05, 6.2943e-05, 7.3543e-05, 5.7450e-05, 6.6065e-05, 6.3751e-05, 6.2092e-05, 7.9485e-05], device='cuda:1') 2023-04-28 06:31:10,085 INFO [finetune.py:976] (1/7) Epoch 30, batch 4150, loss[loss=0.1655, simple_loss=0.2222, pruned_loss=0.05438, over 4165.00 frames. ], tot_loss[loss=0.1669, simple_loss=0.2416, pruned_loss=0.04608, over 951884.76 frames. ], batch size: 18, lr: 2.79e-03, grad_scale: 32.0 2023-04-28 06:31:26,665 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=170277.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:31:29,027 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.851e+01 1.521e+02 1.781e+02 2.076e+02 3.722e+02, threshold=3.562e+02, percent-clipped=0.0 2023-04-28 06:31:42,965 INFO [finetune.py:976] (1/7) Epoch 30, batch 4200, loss[loss=0.1409, simple_loss=0.2171, pruned_loss=0.0324, over 4824.00 frames. ], tot_loss[loss=0.1662, simple_loss=0.2412, pruned_loss=0.04563, over 953212.58 frames. ], batch size: 30, lr: 2.79e-03, grad_scale: 32.0 2023-04-28 06:31:58,820 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=170325.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:32:15,646 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0941, 2.3734, 1.2673, 1.5748, 1.9211, 1.3787, 2.8073, 1.7747], device='cuda:1'), covar=tensor([0.0587, 0.0773, 0.0633, 0.0874, 0.0353, 0.0789, 0.0222, 0.0480], device='cuda:1'), in_proj_covar=tensor([0.0050, 0.0064, 0.0046, 0.0046, 0.0049, 0.0051, 0.0072, 0.0050], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:1') 2023-04-28 06:32:16,768 INFO [finetune.py:976] (1/7) Epoch 30, batch 4250, loss[loss=0.2142, simple_loss=0.2708, pruned_loss=0.0788, over 4701.00 frames. ], tot_loss[loss=0.165, simple_loss=0.239, pruned_loss=0.04555, over 952340.70 frames. ], batch size: 23, lr: 2.79e-03, grad_scale: 32.0 2023-04-28 06:32:18,772 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-04-28 06:32:36,243 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.843e+01 1.455e+02 1.729e+02 2.061e+02 4.081e+02, threshold=3.459e+02, percent-clipped=1.0 2023-04-28 06:32:38,796 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=170385.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:32:48,582 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=170401.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:32:50,164 INFO [finetune.py:976] (1/7) Epoch 30, batch 4300, loss[loss=0.1435, simple_loss=0.219, pruned_loss=0.03395, over 4788.00 frames. ], tot_loss[loss=0.1624, simple_loss=0.236, pruned_loss=0.04437, over 953272.44 frames. ], batch size: 29, lr: 2.79e-03, grad_scale: 32.0 2023-04-28 06:32:50,278 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=170403.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:33:10,859 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=170433.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:33:16,815 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=170442.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:33:22,270 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=170451.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:33:23,431 INFO [finetune.py:976] (1/7) Epoch 30, batch 4350, loss[loss=0.167, simple_loss=0.2284, pruned_loss=0.05279, over 4919.00 frames. ], tot_loss[loss=0.1603, simple_loss=0.2335, pruned_loss=0.04357, over 954350.97 frames. ], batch size: 36, lr: 2.79e-03, grad_scale: 32.0 2023-04-28 06:33:28,871 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=170461.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:33:52,313 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.003e+02 1.526e+02 1.784e+02 2.388e+02 5.022e+02, threshold=3.568e+02, percent-clipped=3.0 2023-04-28 06:34:04,605 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=170490.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:34:23,440 INFO [finetune.py:976] (1/7) Epoch 30, batch 4400, loss[loss=0.162, simple_loss=0.2359, pruned_loss=0.04404, over 4766.00 frames. ], tot_loss[loss=0.1609, simple_loss=0.234, pruned_loss=0.04389, over 955778.92 frames. ], batch size: 26, lr: 2.79e-03, grad_scale: 32.0 2023-04-28 06:34:47,080 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=170522.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 06:35:05,428 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6586, 2.1096, 1.6532, 1.5921, 1.2617, 1.3298, 1.7306, 1.1939], device='cuda:1'), covar=tensor([0.1765, 0.1243, 0.1591, 0.1691, 0.2416, 0.2060, 0.1085, 0.2173], device='cuda:1'), in_proj_covar=tensor([0.0199, 0.0209, 0.0170, 0.0204, 0.0201, 0.0187, 0.0157, 0.0187], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-28 06:35:08,312 INFO [finetune.py:976] (1/7) Epoch 30, batch 4450, loss[loss=0.1764, simple_loss=0.2464, pruned_loss=0.05318, over 4810.00 frames. ], tot_loss[loss=0.1633, simple_loss=0.2372, pruned_loss=0.04464, over 955902.12 frames. ], batch size: 45, lr: 2.79e-03, grad_scale: 32.0 2023-04-28 06:35:18,638 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=170569.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:35:26,201 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.085e+02 1.572e+02 1.800e+02 2.072e+02 3.635e+02, threshold=3.600e+02, percent-clipped=1.0 2023-04-28 06:35:37,926 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.69 vs. limit=2.0 2023-04-28 06:35:41,957 INFO [finetune.py:976] (1/7) Epoch 30, batch 4500, loss[loss=0.1558, simple_loss=0.2315, pruned_loss=0.04005, over 4761.00 frames. ], tot_loss[loss=0.1648, simple_loss=0.2392, pruned_loss=0.04519, over 956336.88 frames. ], batch size: 27, lr: 2.79e-03, grad_scale: 32.0 2023-04-28 06:35:45,759 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0893, 1.5524, 1.9380, 2.3806, 1.9128, 1.5567, 1.2355, 1.7642], device='cuda:1'), covar=tensor([0.3255, 0.3290, 0.1762, 0.2229, 0.2669, 0.2674, 0.4290, 0.1963], device='cuda:1'), in_proj_covar=tensor([0.0296, 0.0248, 0.0232, 0.0318, 0.0226, 0.0238, 0.0232, 0.0188], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-28 06:35:55,338 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.11 vs. limit=2.0 2023-04-28 06:35:59,494 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=170630.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:36:15,352 INFO [finetune.py:976] (1/7) Epoch 30, batch 4550, loss[loss=0.1232, simple_loss=0.2085, pruned_loss=0.01891, over 4815.00 frames. ], tot_loss[loss=0.1664, simple_loss=0.241, pruned_loss=0.04591, over 955673.92 frames. ], batch size: 38, lr: 2.79e-03, grad_scale: 32.0 2023-04-28 06:36:24,378 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0408, 1.3071, 5.1657, 4.8748, 4.4690, 5.0509, 4.5289, 4.5474], device='cuda:1'), covar=tensor([0.6807, 0.5933, 0.0996, 0.1825, 0.1112, 0.1602, 0.1374, 0.1578], device='cuda:1'), in_proj_covar=tensor([0.0308, 0.0306, 0.0402, 0.0407, 0.0346, 0.0412, 0.0315, 0.0361], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-28 06:36:25,042 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.6639, 2.5396, 1.9947, 2.2399, 2.5389, 2.0616, 3.2252, 1.9162], device='cuda:1'), covar=tensor([0.3520, 0.2177, 0.4303, 0.3348, 0.1974, 0.2893, 0.1870, 0.4468], device='cuda:1'), in_proj_covar=tensor([0.0341, 0.0358, 0.0429, 0.0354, 0.0388, 0.0377, 0.0372, 0.0425], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-28 06:36:33,303 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.995e+01 1.537e+02 1.792e+02 2.273e+02 3.619e+02, threshold=3.584e+02, percent-clipped=1.0 2023-04-28 06:36:47,830 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=170701.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:36:48,972 INFO [finetune.py:976] (1/7) Epoch 30, batch 4600, loss[loss=0.1559, simple_loss=0.2316, pruned_loss=0.04009, over 4907.00 frames. ], tot_loss[loss=0.1657, simple_loss=0.2403, pruned_loss=0.04552, over 956458.60 frames. ], batch size: 37, lr: 2.79e-03, grad_scale: 32.0 2023-04-28 06:37:10,104 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0385, 1.5799, 1.5969, 1.7957, 2.1791, 1.7993, 1.5287, 1.5501], device='cuda:1'), covar=tensor([0.1529, 0.1404, 0.1963, 0.1157, 0.0785, 0.1464, 0.2043, 0.2148], device='cuda:1'), in_proj_covar=tensor([0.0315, 0.0307, 0.0350, 0.0286, 0.0323, 0.0305, 0.0300, 0.0377], device='cuda:1'), out_proj_covar=tensor([6.3933e-05, 6.2649e-05, 7.3093e-05, 5.7044e-05, 6.5776e-05, 6.3235e-05, 6.1846e-05, 7.9499e-05], device='cuda:1') 2023-04-28 06:37:19,720 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=170749.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:37:22,624 INFO [finetune.py:976] (1/7) Epoch 30, batch 4650, loss[loss=0.1412, simple_loss=0.2182, pruned_loss=0.03208, over 4818.00 frames. ], tot_loss[loss=0.1633, simple_loss=0.2373, pruned_loss=0.04467, over 954840.73 frames. ], batch size: 30, lr: 2.79e-03, grad_scale: 32.0 2023-04-28 06:37:23,292 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8133, 2.1223, 1.8227, 2.1595, 1.5142, 1.8815, 1.9752, 1.4619], device='cuda:1'), covar=tensor([0.1688, 0.1186, 0.0884, 0.1054, 0.3482, 0.1068, 0.1626, 0.2153], device='cuda:1'), in_proj_covar=tensor([0.0282, 0.0300, 0.0217, 0.0273, 0.0308, 0.0255, 0.0248, 0.0261], device='cuda:1'), out_proj_covar=tensor([1.1211e-04, 1.1761e-04, 8.4831e-05, 1.0689e-04, 1.2387e-04, 1.0023e-04, 9.9482e-05, 1.0251e-04], device='cuda:1') 2023-04-28 06:37:34,235 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9097, 2.4186, 1.9038, 1.8581, 1.4133, 1.4811, 1.9645, 1.3636], device='cuda:1'), covar=tensor([0.1666, 0.1214, 0.1477, 0.1681, 0.2323, 0.1944, 0.0991, 0.2133], device='cuda:1'), in_proj_covar=tensor([0.0198, 0.0207, 0.0169, 0.0203, 0.0199, 0.0186, 0.0156, 0.0187], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-28 06:37:39,986 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.849e+01 1.520e+02 1.798e+02 2.217e+02 4.405e+02, threshold=3.597e+02, percent-clipped=2.0 2023-04-28 06:37:55,525 INFO [finetune.py:976] (1/7) Epoch 30, batch 4700, loss[loss=0.1501, simple_loss=0.212, pruned_loss=0.04412, over 4790.00 frames. ], tot_loss[loss=0.161, simple_loss=0.2344, pruned_loss=0.04382, over 956186.15 frames. ], batch size: 25, lr: 2.79e-03, grad_scale: 32.0 2023-04-28 06:38:05,047 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=170817.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 06:38:29,291 INFO [finetune.py:976] (1/7) Epoch 30, batch 4750, loss[loss=0.1461, simple_loss=0.2166, pruned_loss=0.03779, over 4804.00 frames. ], tot_loss[loss=0.1594, simple_loss=0.2323, pruned_loss=0.04323, over 956632.50 frames. ], batch size: 45, lr: 2.79e-03, grad_scale: 32.0 2023-04-28 06:38:47,640 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.061e+02 1.520e+02 1.829e+02 2.099e+02 3.984e+02, threshold=3.658e+02, percent-clipped=1.0 2023-04-28 06:38:59,650 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0792, 2.3693, 1.1000, 1.3801, 1.8310, 1.2723, 3.0577, 1.7279], device='cuda:1'), covar=tensor([0.0685, 0.0564, 0.0669, 0.1261, 0.0454, 0.0997, 0.0373, 0.0609], device='cuda:1'), in_proj_covar=tensor([0.0051, 0.0064, 0.0046, 0.0046, 0.0049, 0.0051, 0.0072, 0.0050], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:1') 2023-04-28 06:39:16,798 INFO [finetune.py:976] (1/7) Epoch 30, batch 4800, loss[loss=0.1575, simple_loss=0.2299, pruned_loss=0.04258, over 4799.00 frames. ], tot_loss[loss=0.1614, simple_loss=0.2344, pruned_loss=0.04419, over 954643.85 frames. ], batch size: 25, lr: 2.79e-03, grad_scale: 32.0 2023-04-28 06:39:45,897 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=170925.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:39:47,788 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=170928.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:39:50,694 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.3065, 1.5801, 1.4347, 1.7475, 1.6189, 1.8717, 1.4426, 3.5030], device='cuda:1'), covar=tensor([0.0592, 0.0832, 0.0802, 0.1208, 0.0669, 0.0500, 0.0742, 0.0134], device='cuda:1'), in_proj_covar=tensor([0.0037, 0.0038, 0.0039, 0.0042, 0.0040, 0.0037, 0.0038, 0.0054], device='cuda:1'), out_proj_covar=tensor([0.0012, 0.0013, 0.0014, 0.0015, 0.0014, 0.0013, 0.0013, 0.0016], device='cuda:1') 2023-04-28 06:40:18,951 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.5134, 1.2768, 4.2920, 4.0449, 3.7921, 4.1535, 4.0072, 3.7801], device='cuda:1'), covar=tensor([0.7522, 0.6340, 0.1167, 0.1852, 0.1163, 0.2001, 0.1671, 0.1570], device='cuda:1'), in_proj_covar=tensor([0.0313, 0.0310, 0.0408, 0.0412, 0.0349, 0.0418, 0.0318, 0.0365], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-28 06:40:21,162 INFO [finetune.py:976] (1/7) Epoch 30, batch 4850, loss[loss=0.1626, simple_loss=0.2425, pruned_loss=0.04136, over 4896.00 frames. ], tot_loss[loss=0.1626, simple_loss=0.2369, pruned_loss=0.04417, over 955160.93 frames. ], batch size: 46, lr: 2.79e-03, grad_scale: 32.0 2023-04-28 06:40:52,178 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 8.978e+01 1.534e+02 1.869e+02 2.233e+02 4.849e+02, threshold=3.739e+02, percent-clipped=1.0 2023-04-28 06:40:59,515 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=170989.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:41:19,848 INFO [finetune.py:976] (1/7) Epoch 30, batch 4900, loss[loss=0.1395, simple_loss=0.2285, pruned_loss=0.0252, over 4744.00 frames. ], tot_loss[loss=0.1633, simple_loss=0.238, pruned_loss=0.04431, over 957080.06 frames. ], batch size: 28, lr: 2.79e-03, grad_scale: 32.0 2023-04-28 06:41:42,000 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2023-04-28 06:42:25,542 INFO [finetune.py:976] (1/7) Epoch 30, batch 4950, loss[loss=0.1553, simple_loss=0.2237, pruned_loss=0.04346, over 4753.00 frames. ], tot_loss[loss=0.1628, simple_loss=0.2378, pruned_loss=0.04393, over 954985.14 frames. ], batch size: 23, lr: 2.78e-03, grad_scale: 64.0 2023-04-28 06:42:32,642 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.0079, 2.3225, 1.0429, 1.3460, 1.7659, 1.1824, 3.0294, 1.6240], device='cuda:1'), covar=tensor([0.0660, 0.0530, 0.0663, 0.1245, 0.0466, 0.1034, 0.0205, 0.0586], device='cuda:1'), in_proj_covar=tensor([0.0050, 0.0064, 0.0046, 0.0046, 0.0049, 0.0051, 0.0072, 0.0050], device='cuda:1'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:1') 2023-04-28 06:42:45,070 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.189e+02 1.630e+02 1.843e+02 2.166e+02 3.655e+02, threshold=3.685e+02, percent-clipped=0.0 2023-04-28 06:42:47,181 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.55 vs. limit=5.0 2023-04-28 06:42:52,360 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=171092.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:42:59,424 INFO [finetune.py:976] (1/7) Epoch 30, batch 5000, loss[loss=0.1218, simple_loss=0.2081, pruned_loss=0.01773, over 4755.00 frames. ], tot_loss[loss=0.1615, simple_loss=0.2358, pruned_loss=0.04354, over 954495.85 frames. ], batch size: 28, lr: 2.78e-03, grad_scale: 64.0 2023-04-28 06:43:09,464 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=171117.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 06:43:33,016 INFO [finetune.py:976] (1/7) Epoch 30, batch 5050, loss[loss=0.1759, simple_loss=0.2373, pruned_loss=0.05728, over 4831.00 frames. ], tot_loss[loss=0.1616, simple_loss=0.2349, pruned_loss=0.04413, over 957265.88 frames. ], batch size: 41, lr: 2.78e-03, grad_scale: 64.0 2023-04-28 06:43:33,148 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=171153.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:43:37,406 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.7216, 2.1221, 2.2140, 2.3156, 2.1400, 2.1798, 2.2649, 2.1998], device='cuda:1'), covar=tensor([0.3610, 0.4981, 0.4360, 0.4054, 0.5084, 0.6218, 0.4881, 0.4634], device='cuda:1'), in_proj_covar=tensor([0.0340, 0.0373, 0.0330, 0.0342, 0.0350, 0.0393, 0.0361, 0.0334], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-04-28 06:43:40,806 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=171165.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:43:52,462 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.935e+01 1.498e+02 1.792e+02 2.110e+02 3.798e+02, threshold=3.585e+02, percent-clipped=1.0 2023-04-28 06:43:56,850 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.8748, 2.2012, 1.8152, 1.6262, 1.3887, 1.4136, 1.7951, 1.3387], device='cuda:1'), covar=tensor([0.1691, 0.1253, 0.1375, 0.1659, 0.2287, 0.1951, 0.0981, 0.2070], device='cuda:1'), in_proj_covar=tensor([0.0198, 0.0209, 0.0170, 0.0204, 0.0200, 0.0187, 0.0156, 0.0188], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-28 06:44:02,204 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=171196.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:44:06,508 INFO [finetune.py:976] (1/7) Epoch 30, batch 5100, loss[loss=0.2081, simple_loss=0.2662, pruned_loss=0.07502, over 4826.00 frames. ], tot_loss[loss=0.1587, simple_loss=0.2315, pruned_loss=0.04299, over 955761.34 frames. ], batch size: 40, lr: 2.78e-03, grad_scale: 64.0 2023-04-28 06:44:21,434 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=171225.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:44:40,211 INFO [finetune.py:976] (1/7) Epoch 30, batch 5150, loss[loss=0.17, simple_loss=0.2372, pruned_loss=0.05138, over 4863.00 frames. ], tot_loss[loss=0.1614, simple_loss=0.2339, pruned_loss=0.04442, over 955286.09 frames. ], batch size: 34, lr: 2.78e-03, grad_scale: 64.0 2023-04-28 06:44:42,805 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=171257.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:44:43,984 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.09 vs. limit=2.0 2023-04-28 06:44:51,132 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.7210, 1.3981, 1.3592, 1.4969, 1.8497, 1.5269, 1.2847, 1.3292], device='cuda:1'), covar=tensor([0.1507, 0.1178, 0.1762, 0.1196, 0.0819, 0.1382, 0.1715, 0.1996], device='cuda:1'), in_proj_covar=tensor([0.0313, 0.0307, 0.0348, 0.0285, 0.0323, 0.0306, 0.0299, 0.0374], device='cuda:1'), out_proj_covar=tensor([6.3489e-05, 6.2644e-05, 7.2677e-05, 5.6749e-05, 6.5735e-05, 6.3311e-05, 6.1616e-05, 7.8902e-05], device='cuda:1') 2023-04-28 06:44:53,819 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=171273.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:44:57,394 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9498, 1.3786, 4.5949, 4.3465, 3.9651, 4.4128, 4.1456, 4.0280], device='cuda:1'), covar=tensor([0.7188, 0.5861, 0.1372, 0.2089, 0.1254, 0.1401, 0.2207, 0.1617], device='cuda:1'), in_proj_covar=tensor([0.0313, 0.0309, 0.0408, 0.0412, 0.0350, 0.0418, 0.0318, 0.0366], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-04-28 06:45:05,608 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.060e+02 1.555e+02 1.961e+02 2.367e+02 3.631e+02, threshold=3.922e+02, percent-clipped=1.0 2023-04-28 06:45:07,342 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=171284.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:45:35,544 INFO [finetune.py:976] (1/7) Epoch 30, batch 5200, loss[loss=0.2148, simple_loss=0.2926, pruned_loss=0.06851, over 4751.00 frames. ], tot_loss[loss=0.1642, simple_loss=0.238, pruned_loss=0.04522, over 955505.04 frames. ], batch size: 54, lr: 2.78e-03, grad_scale: 32.0 2023-04-28 06:45:48,143 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 2023-04-28 06:45:58,277 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.73 vs. limit=5.0 2023-04-28 06:46:31,215 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=171345.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:46:36,594 INFO [finetune.py:976] (1/7) Epoch 30, batch 5250, loss[loss=0.1534, simple_loss=0.2285, pruned_loss=0.03916, over 4759.00 frames. ], tot_loss[loss=0.1666, simple_loss=0.2411, pruned_loss=0.04603, over 957839.63 frames. ], batch size: 27, lr: 2.78e-03, grad_scale: 32.0 2023-04-28 06:46:49,527 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([2.2893, 1.7438, 2.2035, 2.6103, 2.2180, 1.6603, 1.4735, 2.0530], device='cuda:1'), covar=tensor([0.2783, 0.2905, 0.1386, 0.2027, 0.2256, 0.2393, 0.3877, 0.1709], device='cuda:1'), in_proj_covar=tensor([0.0293, 0.0247, 0.0229, 0.0315, 0.0225, 0.0236, 0.0230, 0.0186], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-04-28 06:46:59,195 INFO [scaling.py:679] (1/7) Whitening: num_groups=1, num_channels=384, metric=4.27 vs. limit=5.0 2023-04-28 06:47:18,738 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.056e+02 1.554e+02 1.778e+02 2.227e+02 5.005e+02, threshold=3.556e+02, percent-clipped=1.0 2023-04-28 06:47:43,263 INFO [finetune.py:976] (1/7) Epoch 30, batch 5300, loss[loss=0.1833, simple_loss=0.2652, pruned_loss=0.05075, over 4812.00 frames. ], tot_loss[loss=0.1668, simple_loss=0.2415, pruned_loss=0.04607, over 956122.69 frames. ], batch size: 39, lr: 2.78e-03, grad_scale: 32.0 2023-04-28 06:47:50,518 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=171406.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:48:41,622 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=171448.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:48:44,664 INFO [finetune.py:976] (1/7) Epoch 30, batch 5350, loss[loss=0.2016, simple_loss=0.2727, pruned_loss=0.06528, over 4815.00 frames. ], tot_loss[loss=0.1668, simple_loss=0.2417, pruned_loss=0.04591, over 956926.61 frames. ], batch size: 33, lr: 2.78e-03, grad_scale: 32.0 2023-04-28 06:48:54,916 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9904, 1.7967, 1.9705, 2.3821, 2.5108, 1.9132, 1.7113, 2.0579], device='cuda:1'), covar=tensor([0.0862, 0.1084, 0.0812, 0.0582, 0.0533, 0.0936, 0.0747, 0.0585], device='cuda:1'), in_proj_covar=tensor([0.0182, 0.0202, 0.0183, 0.0171, 0.0177, 0.0178, 0.0149, 0.0176], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-04-28 06:49:03,281 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.027e+02 1.512e+02 1.839e+02 2.201e+02 3.854e+02, threshold=3.679e+02, percent-clipped=1.0 2023-04-28 06:49:15,016 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9373, 2.1242, 1.1027, 1.6201, 2.3549, 1.7441, 1.6230, 1.7371], device='cuda:1'), covar=tensor([0.0445, 0.0333, 0.0277, 0.0508, 0.0218, 0.0488, 0.0471, 0.0542], device='cuda:1'), in_proj_covar=tensor([0.0027, 0.0023, 0.0021, 0.0028, 0.0019, 0.0027, 0.0027, 0.0028], device='cuda:1'), out_proj_covar=tensor([0.0052, 0.0047, 0.0039, 0.0054, 0.0039, 0.0051, 0.0051, 0.0053], device='cuda:1') 2023-04-28 06:49:15,618 INFO [zipformer.py:1188] (1/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=171498.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:49:18,573 INFO [finetune.py:976] (1/7) Epoch 30, batch 5400, loss[loss=0.1686, simple_loss=0.2349, pruned_loss=0.05117, over 4735.00 frames. ], tot_loss[loss=0.1649, simple_loss=0.2392, pruned_loss=0.04529, over 956548.29 frames. ], batch size: 59, lr: 2.78e-03, grad_scale: 32.0 2023-04-28 06:49:51,781 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=171552.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:49:52,349 INFO [finetune.py:976] (1/7) Epoch 30, batch 5450, loss[loss=0.1738, simple_loss=0.233, pruned_loss=0.0573, over 4909.00 frames. ], tot_loss[loss=0.1633, simple_loss=0.2365, pruned_loss=0.04502, over 956413.26 frames. ], batch size: 32, lr: 2.78e-03, grad_scale: 32.0 2023-04-28 06:49:56,126 INFO [zipformer.py:1188] (1/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=171559.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:50:10,993 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 9.901e+01 1.444e+02 1.753e+02 2.019e+02 4.846e+02, threshold=3.505e+02, percent-clipped=2.0 2023-04-28 06:50:11,287 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 2023-04-28 06:50:12,307 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=171584.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:50:14,278 INFO [scaling.py:679] (1/7) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 2023-04-28 06:50:25,849 INFO [finetune.py:976] (1/7) Epoch 30, batch 5500, loss[loss=0.1379, simple_loss=0.2138, pruned_loss=0.031, over 4775.00 frames. ], tot_loss[loss=0.161, simple_loss=0.2337, pruned_loss=0.04419, over 955270.93 frames. ], batch size: 26, lr: 2.78e-03, grad_scale: 32.0 2023-04-28 06:50:31,400 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.9684, 2.5321, 1.9979, 1.9825, 1.4366, 1.4508, 2.0652, 1.4066], device='cuda:1'), covar=tensor([0.1741, 0.1365, 0.1425, 0.1747, 0.2373, 0.1942, 0.1019, 0.2183], device='cuda:1'), in_proj_covar=tensor([0.0200, 0.0210, 0.0171, 0.0205, 0.0202, 0.0188, 0.0157, 0.0189], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-04-28 06:50:33,193 INFO [zipformer.py:2441] (1/7) attn_weights_entropy = tensor([1.4648, 3.3382, 1.0024, 1.9703, 1.8794, 2.4386, 1.9322, 1.0899], device='cuda:1'), covar=tensor([0.1246, 0.0805, 0.1638, 0.1086, 0.0999, 0.0870, 0.1305, 0.1791], device='cuda:1'), in_proj_covar=tensor([0.0116, 0.0235, 0.0135, 0.0120, 0.0130, 0.0151, 0.0116, 0.0116], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-04-28 06:50:45,014 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=171632.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:50:59,647 INFO [finetune.py:976] (1/7) Epoch 30, batch 5550, loss[loss=0.195, simple_loss=0.2679, pruned_loss=0.0611, over 4825.00 frames. ], tot_loss[loss=0.1611, simple_loss=0.2344, pruned_loss=0.04387, over 956772.64 frames. ], batch size: 51, lr: 2.78e-03, grad_scale: 32.0 2023-04-28 06:51:38,774 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.059e+02 1.632e+02 1.883e+02 2.258e+02 4.653e+02, threshold=3.766e+02, percent-clipped=3.0 2023-04-28 06:52:00,712 INFO [zipformer.py:1188] (1/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=171701.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:52:01,828 INFO [finetune.py:976] (1/7) Epoch 30, batch 5600, loss[loss=0.1828, simple_loss=0.2554, pruned_loss=0.0551, over 4833.00 frames. ], tot_loss[loss=0.1626, simple_loss=0.2371, pruned_loss=0.04401, over 957284.54 frames. ], batch size: 33, lr: 2.78e-03, grad_scale: 32.0 2023-04-28 06:52:55,572 INFO [zipformer.py:1188] (1/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=171748.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:52:58,358 INFO [finetune.py:976] (1/7) Epoch 30, batch 5650, loss[loss=0.1552, simple_loss=0.2145, pruned_loss=0.04797, over 4082.00 frames. ], tot_loss[loss=0.1644, simple_loss=0.2397, pruned_loss=0.04451, over 956657.69 frames. ], batch size: 17, lr: 2.78e-03, grad_scale: 32.0 2023-04-28 06:53:37,492 INFO [optim.py:369] (1/7) Clipping_scale=2.0, grad-norm quartiles 1.071e+02 1.430e+02 1.735e+02 1.948e+02 3.086e+02, threshold=3.470e+02, percent-clipped=0.0 2023-04-28 06:53:50,968 INFO [zipformer.py:1188] (1/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=171796.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:54:00,717 INFO [finetune.py:976] (1/7) Epoch 30, batch 5700, loss[loss=0.1506, simple_loss=0.2112, pruned_loss=0.04501, over 4202.00 frames. ], tot_loss[loss=0.162, simple_loss=0.2364, pruned_loss=0.04386, over 936117.16 frames. ], batch size: 18, lr: 2.78e-03, grad_scale: 32.0 2023-04-28 06:54:33,224 INFO [finetune.py:1241] (1/7) Done!