2023-04-26 10:07:06,675 INFO [finetune.py:1046] (6/7) Training started 2023-04-26 10:07:06,675 INFO [finetune.py:1056] (6/7) Device: cuda:6 2023-04-26 10:07:06,677 INFO [finetune.py:1065] (6/7) {'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 10:07:06,677 INFO [finetune.py:1067] (6/7) About to create model 2023-04-26 10:07:07,059 INFO [zipformer.py:405] (6/7) 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 10:07:07,068 INFO [finetune.py:1071] (6/7) Number of model parameters: 70369391 2023-04-26 10:07:07,068 INFO [finetune.py:626] (6/7) Loading checkpoint from /home/lishaojie/icefall/egs/commonvoice_fr/ASR/pruned_transducer_stateless7_streaming/exp/english_pretrain/pretrained.pt 2023-04-26 10:07:07,258 INFO [finetune.py:647] (6/7) Loading parameters starting with prefix encoder 2023-04-26 10:07:08,630 INFO [finetune.py:1093] (6/7) Using DDP 2023-04-26 10:07:09,244 INFO [commonvoice_fr.py:392] (6/7) About to get train cuts 2023-04-26 10:07:09,245 INFO [commonvoice_fr.py:218] (6/7) Enable MUSAN 2023-04-26 10:07:09,246 INFO [commonvoice_fr.py:219] (6/7) About to get Musan cuts 2023-04-26 10:07:11,046 INFO [commonvoice_fr.py:243] (6/7) Enable SpecAugment 2023-04-26 10:07:11,046 INFO [commonvoice_fr.py:244] (6/7) Time warp factor: 80 2023-04-26 10:07:11,046 INFO [commonvoice_fr.py:254] (6/7) Num frame mask: 10 2023-04-26 10:07:11,046 INFO [commonvoice_fr.py:267] (6/7) About to create train dataset 2023-04-26 10:07:11,046 INFO [commonvoice_fr.py:294] (6/7) Using DynamicBucketingSampler. 2023-04-26 10:07:13,940 INFO [commonvoice_fr.py:309] (6/7) About to create train dataloader 2023-04-26 10:07:13,941 INFO [commonvoice_fr.py:399] (6/7) About to get dev cuts 2023-04-26 10:07:13,942 INFO [commonvoice_fr.py:340] (6/7) About to create dev dataset 2023-04-26 10:07:14,346 INFO [commonvoice_fr.py:357] (6/7) About to create dev dataloader 2023-04-26 10:07:14,346 INFO [finetune.py:1289] (6/7) Sanity check -- see if any of the batches in epoch 1 would cause OOM. 2023-04-26 10:11:06,932 INFO [finetune.py:1317] (6/7) Maximum memory allocated so far is 5162MB 2023-04-26 10:11:07,629 INFO [finetune.py:1317] (6/7) Maximum memory allocated so far is 5675MB 2023-04-26 10:11:08,317 INFO [finetune.py:1317] (6/7) Maximum memory allocated so far is 5675MB 2023-04-26 10:11:08,990 INFO [finetune.py:1317] (6/7) Maximum memory allocated so far is 5675MB 2023-04-26 10:11:09,672 INFO [finetune.py:1317] (6/7) Maximum memory allocated so far is 5675MB 2023-04-26 10:11:10,374 INFO [finetune.py:1317] (6/7) Maximum memory allocated so far is 5675MB 2023-04-26 10:11:19,556 INFO [finetune.py:976] (6/7) Epoch 1, batch 0, loss[loss=7.56, simple_loss=6.86, pruned_loss=6.981, over 4896.00 frames. ], tot_loss[loss=7.56, simple_loss=6.86, pruned_loss=6.981, over 4896.00 frames. ], batch size: 43, lr: 2.00e-03, grad_scale: 2.0 2023-04-26 10:11:19,556 INFO [finetune.py:1001] (6/7) Computing validation loss 2023-04-26 10:11:40,202 INFO [finetune.py:1010] (6/7) 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] (6/7) Maximum memory allocated so far is 5675MB 2023-04-26 10:11:48,692 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=5.0, num_to_drop=2, layers_to_drop={2, 3} 2023-04-26 10:12:11,186 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=23.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 10:12:43,279 INFO [finetune.py:976] (6/7) Epoch 1, batch 50, loss[loss=2.557, simple_loss=2.433, pruned_loss=1.26, over 4818.00 frames. ], tot_loss[loss=4.457, simple_loss=4.035, pruned_loss=4.085, over 215907.26 frames. ], batch size: 40, lr: 2.20e-03, grad_scale: 0.00390625 2023-04-26 10:13:19,830 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=83.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 10:13:41,114 WARNING [finetune.py:966] (6/7) Grad scale is small: 6.103515625e-05 2023-04-26 10:13:41,115 INFO [finetune.py:976] (6/7) Epoch 1, batch 100, loss[loss=2.694, simple_loss=2.572, pruned_loss=1.29, over 4713.00 frames. ], tot_loss[loss=3.518, simple_loss=3.253, pruned_loss=2.589, over 380576.87 frames. ], batch size: 23, lr: 2.40e-03, grad_scale: 0.0001220703125 2023-04-26 10:14:03,495 INFO [optim.py:369] (6/7) 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 10:14:06,182 WARNING [optim.py:389] (6/7) Scaling gradients by 0.014711554162204266, model_norm_threshold=15666.9306640625 2023-04-26 10:14:06,249 INFO [optim.py:451] (6/7) 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 10:14:19,413 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=2.17 vs. limit=2.0 2023-04-26 10:14:23,407 WARNING [optim.py:389] (6/7) Scaling gradients by 0.00018281130178365856, model_norm_threshold=15666.9306640625 2023-04-26 10:14:23,479 INFO [optim.py:451] (6/7) 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 10:14:24,108 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=144.0, num_to_drop=2, layers_to_drop={1, 3} 2023-04-26 10:14:27,642 INFO [finetune.py:976] (6/7) Epoch 1, batch 150, loss[loss=2.278, simple_loss=2.099, pruned_loss=1.541, over 4789.00 frames. ], tot_loss[loss=2.924, simple_loss=2.712, pruned_loss=2.039, over 506739.63 frames. ], batch size: 29, lr: 2.60e-03, grad_scale: 3.0517578125e-05 2023-04-26 10:14:28,149 WARNING [optim.py:389] (6/7) Scaling gradients by 0.00022292081848718226, model_norm_threshold=15666.9306640625 2023-04-26 10:14:28,223 INFO [optim.py:451] (6/7) 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 10:14:34,046 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.3382, 2.8517, 2.3257, 3.1051, 3.3922, 2.7623, 2.9479, 2.1148], device='cuda:6'), covar=tensor([0.0211, 0.0175, 0.0150, 0.0116, 0.0150, 0.0131, 0.0172, 0.0214], device='cuda:6'), in_proj_covar=tensor([0.0318, 0.0337, 0.0244, 0.0308, 0.0324, 0.0283, 0.0287, 0.0299], device='cuda:6'), out_proj_covar=tensor([1.3032e-04, 1.3846e-04, 9.9594e-05, 1.2515e-04, 1.3507e-04, 1.1414e-04, 1.1950e-04, 1.2172e-04], device='cuda:6') 2023-04-26 10:14:40,641 WARNING [optim.py:389] (6/7) Scaling gradients by 0.05655747279524803, model_norm_threshold=15666.9306640625 2023-04-26 10:14:40,713 INFO [optim.py:451] (6/7) 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 10:14:56,526 WARNING [finetune.py:966] (6/7) Grad scale is small: 3.0517578125e-05 2023-04-26 10:14:56,527 INFO [finetune.py:976] (6/7) Epoch 1, batch 200, loss[loss=1.474, simple_loss=1.276, pruned_loss=1.373, over 4912.00 frames. ], tot_loss[loss=2.417, simple_loss=2.219, pruned_loss=1.769, over 608186.56 frames. ], batch size: 37, lr: 2.80e-03, grad_scale: 6.103515625e-05 2023-04-26 10:15:07,788 INFO [optim.py:369] (6/7) 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] (6/7) Scaling gradients by 0.011872046627104282, model_norm_threshold=3679.54541015625 2023-04-26 10:15:13,038 INFO [optim.py:451] (6/7) 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] (6/7) Scaling gradients by 0.08515117317438126, model_norm_threshold=3679.54541015625 2023-04-26 10:15:16,231 INFO [optim.py:451] (6/7) 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 10:15:16,775 WARNING [optim.py:389] (6/7) Scaling gradients by 0.04552413150668144, model_norm_threshold=3679.54541015625 2023-04-26 10:15:16,847 INFO [optim.py:451] (6/7) 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 10:15:25,655 INFO [finetune.py:976] (6/7) Epoch 1, batch 250, loss[loss=1.485, simple_loss=1.26, pruned_loss=1.427, over 4828.00 frames. ], tot_loss[loss=2.093, simple_loss=1.896, pruned_loss=1.624, over 685285.30 frames. ], batch size: 40, lr: 3.00e-03, grad_scale: 6.103515625e-05 2023-04-26 10:15:35,003 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9367, 2.3153, 1.8190, 1.8366, 1.4274, 1.5435, 2.0676, 1.3986], device='cuda:6'), covar=tensor([0.0099, 0.0099, 0.0078, 0.0133, 0.0138, 0.0118, 0.0054, 0.0105], device='cuda:6'), in_proj_covar=tensor([0.0224, 0.0255, 0.0225, 0.0248, 0.0263, 0.0220, 0.0213, 0.0237], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:6') 2023-04-26 10:15:50,805 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=296.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 10:15:52,837 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=300.0, num_to_drop=2, layers_to_drop={0, 1} 2023-04-26 10:15:53,282 WARNING [finetune.py:966] (6/7) Grad scale is small: 6.103515625e-05 2023-04-26 10:15:53,282 INFO [finetune.py:976] (6/7) Epoch 1, batch 300, loss[loss=1.327, simple_loss=1.104, pruned_loss=1.301, over 4817.00 frames. ], tot_loss[loss=1.876, simple_loss=1.676, pruned_loss=1.526, over 744714.84 frames. ], batch size: 30, lr: 3.20e-03, grad_scale: 0.0001220703125 2023-04-26 10:15:58,330 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=2.27 vs. limit=2.0 2023-04-26 10:16:02,527 INFO [optim.py:369] (6/7) 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 10:16:06,403 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=15.14 vs. limit=5.0 2023-04-26 10:16:14,905 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=19.60 vs. limit=5.0 2023-04-26 10:16:21,499 INFO [finetune.py:976] (6/7) Epoch 1, batch 350, loss[loss=1.234, simple_loss=1.015, pruned_loss=1.196, over 4768.00 frames. ], tot_loss[loss=1.725, simple_loss=1.518, pruned_loss=1.455, over 793167.54 frames. ], batch size: 27, lr: 3.40e-03, grad_scale: 0.0001220703125 2023-04-26 10:16:24,660 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=357.0, num_to_drop=2, layers_to_drop={0, 1} 2023-04-26 10:16:26,251 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=17.87 vs. limit=5.0 2023-04-26 10:16:42,457 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=387.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 10:16:56,312 WARNING [finetune.py:966] (6/7) Grad scale is small: 0.0001220703125 2023-04-26 10:16:56,313 INFO [finetune.py:976] (6/7) Epoch 1, batch 400, loss[loss=1.265, simple_loss=1.011, pruned_loss=1.267, over 4888.00 frames. ], tot_loss[loss=1.609, simple_loss=1.394, pruned_loss=1.401, over 829412.97 frames. ], batch size: 35, lr: 3.60e-03, grad_scale: 0.000244140625 2023-04-26 10:17:16,615 INFO [optim.py:369] (6/7) 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 10:17:27,879 WARNING [optim.py:389] (6/7) Scaling gradients by 0.02133115753531456, model_norm_threshold=142.37583923339844 2023-04-26 10:17:27,952 INFO [optim.py:451] (6/7) 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 10:17:41,635 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=439.0, num_to_drop=2, layers_to_drop={0, 3} 2023-04-26 10:17:52,445 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=448.0, num_to_drop=2, layers_to_drop={0, 3} 2023-04-26 10:17:53,954 INFO [finetune.py:976] (6/7) Epoch 1, batch 450, loss[loss=1.153, simple_loss=0.9016, pruned_loss=1.165, over 4821.00 frames. ], tot_loss[loss=1.504, simple_loss=1.281, pruned_loss=1.347, over 856009.54 frames. ], batch size: 39, lr: 3.80e-03, grad_scale: 0.000244140625 2023-04-26 10:18:14,583 WARNING [optim.py:389] (6/7) Scaling gradients by 0.06225070729851723, model_norm_threshold=142.37583923339844 2023-04-26 10:18:14,656 INFO [optim.py:451] (6/7) 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 10:18:19,408 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=31.85 vs. limit=5.0 2023-04-26 10:18:22,846 WARNING [finetune.py:966] (6/7) Grad scale is small: 0.000244140625 2023-04-26 10:18:22,846 INFO [finetune.py:976] (6/7) Epoch 1, batch 500, loss[loss=1.019, simple_loss=0.7775, pruned_loss=1.041, over 4754.00 frames. ], tot_loss[loss=1.409, simple_loss=1.179, pruned_loss=1.295, over 878526.69 frames. ], batch size: 27, lr: 4.00e-03, grad_scale: 0.00048828125 2023-04-26 10:18:31,734 INFO [optim.py:369] (6/7) 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 10:18:47,426 WARNING [optim.py:389] (6/7) Scaling gradients by 0.017591100186109543, model_norm_threshold=62.30100631713867 2023-04-26 10:18:47,499 INFO [optim.py:451] (6/7) 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] (6/7) Scaling gradients by 0.005508477333933115, model_norm_threshold=62.30100631713867 2023-04-26 10:18:56,373 INFO [optim.py:451] (6/7) 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] (6/7) Epoch 1, batch 550, loss[loss=0.8715, simple_loss=0.6472, pruned_loss=0.9007, over 4059.00 frames. ], tot_loss[loss=1.329, simple_loss=1.092, pruned_loss=1.245, over 895480.33 frames. ], batch size: 17, lr: 4.00e-03, grad_scale: 0.00048828125 2023-04-26 10:19:04,286 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=562.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 10:19:13,092 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=568.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 10:19:35,163 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=590.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 10:19:42,771 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=2.08 vs. limit=2.0 2023-04-26 10:19:42,845 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=2.08 vs. limit=2.0 2023-04-26 10:19:45,890 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=600.0, num_to_drop=2, layers_to_drop={1, 2} 2023-04-26 10:19:46,333 WARNING [finetune.py:966] (6/7) Grad scale is small: 0.00048828125 2023-04-26 10:19:46,334 INFO [finetune.py:976] (6/7) Epoch 1, batch 600, loss[loss=1.097, simple_loss=0.8057, pruned_loss=1.113, over 4902.00 frames. ], tot_loss[loss=1.272, simple_loss=1.025, pruned_loss=1.212, over 909524.03 frames. ], batch size: 32, lr: 4.00e-03, grad_scale: 0.0009765625 2023-04-26 10:19:46,614 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=2.26 vs. limit=2.0 2023-04-26 10:20:06,710 INFO [optim.py:369] (6/7) 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 10:20:09,413 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=623.0, num_to_drop=2, layers_to_drop={2, 3} 2023-04-26 10:20:18,144 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=629.0, num_to_drop=2, layers_to_drop={0, 1} 2023-04-26 10:20:26,632 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=8.77 vs. limit=5.0 2023-04-26 10:20:29,920 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 1, batch 650, loss[loss=1.201, simple_loss=0.8835, pruned_loss=1.173, over 4816.00 frames. ], tot_loss[loss=1.239, simple_loss=0.9792, pruned_loss=1.193, over 920504.45 frames. ], batch size: 38, lr: 4.00e-03, grad_scale: 0.0009765625 2023-04-26 10:20:31,536 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=651.0, num_to_drop=2, layers_to_drop={0, 3} 2023-04-26 10:20:32,030 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=652.0, num_to_drop=2, layers_to_drop={0, 1} 2023-04-26 10:20:35,885 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=3.30 vs. limit=2.0 2023-04-26 10:20:41,808 WARNING [optim.py:389] (6/7) Scaling gradients by 0.07653743773698807, model_norm_threshold=52.5806770324707 2023-04-26 10:20:41,878 INFO [optim.py:451] (6/7) 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 10:20:59,953 WARNING [finetune.py:966] (6/7) Grad scale is small: 0.0009765625 2023-04-26 10:20:59,953 INFO [finetune.py:976] (6/7) Epoch 1, batch 700, loss[loss=1.082, simple_loss=0.7624, pruned_loss=1.091, over 4821.00 frames. ], tot_loss[loss=1.209, simple_loss=0.938, pruned_loss=1.171, over 927109.38 frames. ], batch size: 38, lr: 4.00e-03, grad_scale: 0.001953125 2023-04-26 10:21:09,243 INFO [optim.py:369] (6/7) 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 10:21:20,042 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=739.0, num_to_drop=2, layers_to_drop={0, 2} 2023-04-26 10:21:22,083 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 1, batch 750, loss[loss=1.143, simple_loss=0.8242, pruned_loss=1.08, over 4806.00 frames. ], tot_loss[loss=1.179, simple_loss=0.9006, pruned_loss=1.142, over 933844.99 frames. ], batch size: 39, lr: 4.00e-03, grad_scale: 0.001953125 2023-04-26 10:21:48,069 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=787.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 10:21:53,227 WARNING [optim.py:389] (6/7) Scaling gradients by 0.039711207151412964, model_norm_threshold=49.79251480102539 2023-04-26 10:21:53,294 INFO [optim.py:451] (6/7) 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 10:21:53,441 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.3018, 2.2077, 1.2262, 1.5897, 1.2334, 1.2681, 1.4906, 1.2903], device='cuda:6'), covar=tensor([0.0230, 0.0257, 0.0173, 0.0311, 0.0278, 0.0177, 0.0156, 0.0210], device='cuda:6'), in_proj_covar=tensor([0.0224, 0.0255, 0.0225, 0.0248, 0.0263, 0.0220, 0.0213, 0.0237], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:6') 2023-04-26 10:21:55,912 WARNING [finetune.py:966] (6/7) Grad scale is small: 0.001953125 2023-04-26 10:21:55,913 INFO [finetune.py:976] (6/7) Epoch 1, batch 800, loss[loss=1.016, simple_loss=0.719, pruned_loss=0.9587, over 4930.00 frames. ], tot_loss[loss=1.152, simple_loss=0.8679, pruned_loss=1.111, over 938962.89 frames. ], batch size: 33, lr: 4.00e-03, grad_scale: 0.00390625 2023-04-26 10:22:11,118 INFO [optim.py:369] (6/7) 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 10:22:26,849 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=847.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 10:22:28,869 INFO [finetune.py:976] (6/7) Epoch 1, batch 850, loss[loss=0.964, simple_loss=0.6793, pruned_loss=0.8888, over 4827.00 frames. ], tot_loss[loss=1.121, simple_loss=0.8346, pruned_loss=1.07, over 943395.91 frames. ], batch size: 38, lr: 4.00e-03, grad_scale: 0.00390625 2023-04-26 10:23:15,919 WARNING [finetune.py:966] (6/7) Grad scale is small: 0.00390625 2023-04-26 10:23:15,920 INFO [finetune.py:976] (6/7) Epoch 1, batch 900, loss[loss=0.9521, simple_loss=0.6731, pruned_loss=0.852, over 4808.00 frames. ], tot_loss[loss=1.088, simple_loss=0.803, pruned_loss=1.027, over 946144.39 frames. ], batch size: 25, lr: 4.00e-03, grad_scale: 0.0078125 2023-04-26 10:23:20,210 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=908.0, num_to_drop=2, layers_to_drop={1, 3} 2023-04-26 10:23:26,898 INFO [optim.py:369] (6/7) 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 10:23:26,984 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=918.0, num_to_drop=2, layers_to_drop={1, 3} 2023-04-26 10:23:30,105 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=924.0, num_to_drop=2, layers_to_drop={0, 3} 2023-04-26 10:23:32,220 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8432, 1.0515, 1.5601, 0.8142, 1.3508, 1.7530, 1.4546, 1.3960], device='cuda:6'), covar=tensor([0.0158, 0.0248, 0.0161, 0.0196, 0.0195, 0.0161, 0.0151, 0.0165], device='cuda:6'), in_proj_covar=tensor([0.0396, 0.0401, 0.0387, 0.0358, 0.0419, 0.0445, 0.0385, 0.0424], device='cuda:6'), out_proj_covar=tensor([8.7202e-05, 8.6402e-05, 8.3422e-05, 7.4991e-05, 9.0215e-05, 9.7928e-05, 8.4446e-05, 9.1988e-05], device='cuda:6') 2023-04-26 10:23:42,106 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=946.0, num_to_drop=2, layers_to_drop={0, 3} 2023-04-26 10:23:44,696 INFO [finetune.py:976] (6/7) Epoch 1, batch 950, loss[loss=1.084, simple_loss=0.7703, pruned_loss=0.9406, over 4868.00 frames. ], tot_loss[loss=1.066, simple_loss=0.7811, pruned_loss=0.9913, over 949338.58 frames. ], batch size: 34, lr: 4.00e-03, grad_scale: 0.0078125 2023-04-26 10:23:45,289 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=952.0, num_to_drop=2, layers_to_drop={0, 1} 2023-04-26 10:24:41,985 INFO [zipformer.py:1188] (6/7) 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] (6/7) Grad scale is small: 0.0078125 2023-04-26 10:24:42,451 INFO [finetune.py:976] (6/7) Epoch 1, batch 1000, loss[loss=1.063, simple_loss=0.7522, pruned_loss=0.9069, over 4805.00 frames. ], tot_loss[loss=1.064, simple_loss=0.7744, pruned_loss=0.9723, over 952276.63 frames. ], batch size: 41, lr: 4.00e-03, grad_scale: 0.015625 2023-04-26 10:25:00,173 INFO [optim.py:369] (6/7) 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 10:25:02,442 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.50 vs. limit=2.0 2023-04-26 10:25:14,566 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=1043.0, num_to_drop=2, layers_to_drop={0, 1} 2023-04-26 10:25:18,745 INFO [finetune.py:976] (6/7) Epoch 1, batch 1050, loss[loss=1.074, simple_loss=0.7686, pruned_loss=0.8861, over 4863.00 frames. ], tot_loss[loss=1.057, simple_loss=0.7653, pruned_loss=0.9493, over 950750.18 frames. ], batch size: 34, lr: 4.00e-03, grad_scale: 0.015625 2023-04-26 10:25:42,489 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=1091.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 10:25:46,285 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=2.69 vs. limit=2.0 2023-04-26 10:25:48,290 INFO [finetune.py:976] (6/7) Epoch 1, batch 1100, loss[loss=1.106, simple_loss=0.7809, pruned_loss=0.9065, over 4928.00 frames. ], tot_loss[loss=1.055, simple_loss=0.7598, pruned_loss=0.93, over 953531.62 frames. ], batch size: 33, lr: 4.00e-03, grad_scale: 0.03125 2023-04-26 10:25:57,274 INFO [optim.py:369] (6/7) 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 10:26:17,721 INFO [finetune.py:976] (6/7) Epoch 1, batch 1150, loss[loss=1.034, simple_loss=0.73, pruned_loss=0.8316, over 4854.00 frames. ], tot_loss[loss=1.052, simple_loss=0.7542, pruned_loss=0.9114, over 952413.72 frames. ], batch size: 31, lr: 4.00e-03, grad_scale: 0.03125 2023-04-26 10:26:18,896 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=1153.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 10:26:46,977 INFO [finetune.py:976] (6/7) Epoch 1, batch 1200, loss[loss=1.021, simple_loss=0.7209, pruned_loss=0.8069, over 4923.00 frames. ], tot_loss[loss=1.04, simple_loss=0.7422, pruned_loss=0.8866, over 953459.49 frames. ], batch size: 46, lr: 4.00e-03, grad_scale: 0.0625 2023-04-26 10:26:48,592 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=1203.0, num_to_drop=2, layers_to_drop={0, 1} 2023-04-26 10:26:54,329 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=1214.0, num_to_drop=2, layers_to_drop={1, 2} 2023-04-26 10:26:56,332 INFO [optim.py:369] (6/7) 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 10:26:56,428 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=1218.0, num_to_drop=2, layers_to_drop={0, 1} 2023-04-26 10:26:59,563 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=1224.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 10:27:19,085 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=1246.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 10:27:21,699 INFO [finetune.py:976] (6/7) Epoch 1, batch 1250, loss[loss=1, simple_loss=0.6934, pruned_loss=0.7899, over 4907.00 frames. ], tot_loss[loss=1.023, simple_loss=0.7273, pruned_loss=0.8594, over 953326.40 frames. ], batch size: 32, lr: 4.00e-03, grad_scale: 0.0625 2023-04-26 10:27:40,181 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=1266.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 10:27:49,528 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=1272.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 10:28:15,545 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=1294.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 10:28:24,920 INFO [finetune.py:976] (6/7) Epoch 1, batch 1300, loss[loss=1.02, simple_loss=0.708, pruned_loss=0.7909, over 4825.00 frames. ], tot_loss[loss=1.008, simple_loss=0.713, pruned_loss=0.8341, over 953764.39 frames. ], batch size: 39, lr: 4.00e-03, grad_scale: 0.125 2023-04-26 10:28:40,244 INFO [optim.py:369] (6/7) 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] (6/7) Epoch 1, batch 1350, loss[loss=1.09, simple_loss=0.7517, pruned_loss=0.8358, over 4868.00 frames. ], tot_loss[loss=1.007, simple_loss=0.7085, pruned_loss=0.8206, over 954903.18 frames. ], batch size: 34, lr: 4.00e-03, grad_scale: 0.125 2023-04-26 10:29:50,224 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.1181, 2.4259, 1.8071, 1.7132, 1.4715, 1.5635, 1.5655, 1.2970], device='cuda:6'), covar=tensor([0.0834, 0.1099, 0.0760, 0.1591, 0.1442, 0.1291, 0.0646, 0.1006], device='cuda:6'), in_proj_covar=tensor([0.0224, 0.0256, 0.0225, 0.0248, 0.0264, 0.0220, 0.0214, 0.0237], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:6') 2023-04-26 10:29:52,993 INFO [finetune.py:976] (6/7) Epoch 1, batch 1400, loss[loss=1.1, simple_loss=0.7586, pruned_loss=0.8313, over 4811.00 frames. ], tot_loss[loss=1.02, simple_loss=0.714, pruned_loss=0.8192, over 955520.71 frames. ], batch size: 45, lr: 4.00e-03, grad_scale: 0.25 2023-04-26 10:30:11,384 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-04-26 10:30:14,311 INFO [optim.py:369] (6/7) 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 10:30:22,777 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.5780, 1.8215, 1.8280, 1.9610, 2.1005, 1.8154, 1.6805, 1.9058], device='cuda:6'), covar=tensor([0.5392, 1.5898, 1.1379, 1.3344, 0.8334, 0.9872, 1.2364, 1.0114], device='cuda:6'), in_proj_covar=tensor([0.0335, 0.0489, 0.0394, 0.0383, 0.0426, 0.0431, 0.0478, 0.0416], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-26 10:30:34,173 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 1, batch 1450, loss[loss=1.048, simple_loss=0.7197, pruned_loss=0.7823, over 4919.00 frames. ], tot_loss[loss=1.024, simple_loss=0.7133, pruned_loss=0.8104, over 955517.49 frames. ], batch size: 38, lr: 4.00e-03, grad_scale: 0.25 2023-04-26 10:30:59,468 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=2.22 vs. limit=2.0 2023-04-26 10:31:42,610 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=1495.0, num_to_drop=2, layers_to_drop={0, 2} 2023-04-26 10:31:53,098 INFO [finetune.py:976] (6/7) Epoch 1, batch 1500, loss[loss=1.022, simple_loss=0.7026, pruned_loss=0.7512, over 4926.00 frames. ], tot_loss[loss=1.024, simple_loss=0.7109, pruned_loss=0.7988, over 956323.16 frames. ], batch size: 42, lr: 4.00e-03, grad_scale: 0.5 2023-04-26 10:31:59,751 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=1503.0, num_to_drop=1, layers_to_drop={2} 2023-04-26 10:32:03,107 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=1509.0, num_to_drop=2, layers_to_drop={1, 3} 2023-04-26 10:32:11,041 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=14.17 vs. limit=5.0 2023-04-26 10:32:13,078 INFO [optim.py:369] (6/7) 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 10:32:23,904 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.4634, 1.6762, 1.6499, 1.6765, 1.7445, 1.8144, 1.8296, 1.7702], device='cuda:6'), covar=tensor([3.7171, 6.5144, 8.0049, 6.3208, 3.6940, 4.6897, 6.3998, 5.4513], device='cuda:6'), in_proj_covar=tensor([0.0344, 0.0501, 0.0406, 0.0394, 0.0437, 0.0443, 0.0491, 0.0429], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-26 10:32:36,550 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.09 vs. limit=5.0 2023-04-26 10:32:48,627 INFO [finetune.py:976] (6/7) Epoch 1, batch 1550, loss[loss=0.9449, simple_loss=0.6559, pruned_loss=0.6808, over 4922.00 frames. ], tot_loss[loss=1.015, simple_loss=0.7046, pruned_loss=0.7798, over 956911.87 frames. ], batch size: 38, lr: 4.00e-03, grad_scale: 0.5 2023-04-26 10:32:53,427 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=1551.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 10:33:08,381 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=1566.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 10:33:48,995 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-04-26 10:33:50,405 INFO [finetune.py:976] (6/7) Epoch 1, batch 1600, loss[loss=0.8047, simple_loss=0.5734, pruned_loss=0.5617, over 4819.00 frames. ], tot_loss[loss=0.9892, simple_loss=0.6891, pruned_loss=0.7481, over 956160.03 frames. ], batch size: 25, lr: 4.00e-03, grad_scale: 1.0 2023-04-26 10:34:03,940 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=6.83 vs. limit=5.0 2023-04-26 10:34:04,229 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=1611.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 10:34:13,525 INFO [optim.py:369] (6/7) 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 10:34:24,076 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=1627.0, num_to_drop=2, layers_to_drop={1, 3} 2023-04-26 10:34:47,807 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=5.57 vs. limit=5.0 2023-04-26 10:34:48,570 INFO [finetune.py:976] (6/7) Epoch 1, batch 1650, loss[loss=0.8045, simple_loss=0.5864, pruned_loss=0.5458, over 4827.00 frames. ], tot_loss[loss=0.9592, simple_loss=0.6722, pruned_loss=0.7129, over 956248.91 frames. ], batch size: 25, lr: 4.00e-03, grad_scale: 1.0 2023-04-26 10:35:08,291 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=1672.0, num_to_drop=2, layers_to_drop={1, 2} 2023-04-26 10:35:36,516 INFO [finetune.py:976] (6/7) Epoch 1, batch 1700, loss[loss=0.8668, simple_loss=0.6394, pruned_loss=0.5769, over 4938.00 frames. ], tot_loss[loss=0.9254, simple_loss=0.6544, pruned_loss=0.675, over 956610.27 frames. ], batch size: 33, lr: 4.00e-03, grad_scale: 1.0 2023-04-26 10:35:59,306 INFO [optim.py:369] (6/7) 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 10:36:24,060 INFO [finetune.py:976] (6/7) Epoch 1, batch 1750, loss[loss=0.8927, simple_loss=0.6849, pruned_loss=0.5717, over 4851.00 frames. ], tot_loss[loss=0.8979, simple_loss=0.6433, pruned_loss=0.6411, over 958047.12 frames. ], batch size: 44, lr: 4.00e-03, grad_scale: 1.0 2023-04-26 10:36:38,470 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=27.49 vs. limit=5.0 2023-04-26 10:36:58,581 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 1, batch 1800, loss[loss=0.7601, simple_loss=0.5928, pruned_loss=0.4769, over 4891.00 frames. ], tot_loss[loss=0.8721, simple_loss=0.6343, pruned_loss=0.6089, over 957117.21 frames. ], batch size: 32, lr: 4.00e-03, grad_scale: 1.0 2023-04-26 10:37:21,910 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=1809.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 10:37:28,277 INFO [optim.py:369] (6/7) 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 10:37:40,494 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=1838.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 10:37:47,697 INFO [finetune.py:976] (6/7) Epoch 1, batch 1850, loss[loss=0.7465, simple_loss=0.6064, pruned_loss=0.4509, over 4918.00 frames. ], tot_loss[loss=0.8398, simple_loss=0.6214, pruned_loss=0.5734, over 956960.27 frames. ], batch size: 38, lr: 4.00e-03, grad_scale: 1.0 2023-04-26 10:37:51,629 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=1857.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 10:37:51,656 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=1857.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 10:38:00,468 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=1870.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 10:38:16,955 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=5.06 vs. limit=5.0 2023-04-26 10:38:17,754 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=1899.0, num_to_drop=2, layers_to_drop={0, 1} 2023-04-26 10:38:18,740 INFO [finetune.py:976] (6/7) Epoch 1, batch 1900, loss[loss=0.6444, simple_loss=0.537, pruned_loss=0.3795, over 4812.00 frames. ], tot_loss[loss=0.8073, simple_loss=0.6077, pruned_loss=0.5393, over 955815.26 frames. ], batch size: 40, lr: 4.00e-03, grad_scale: 1.0 2023-04-26 10:38:21,826 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.79 vs. limit=2.0 2023-04-26 10:38:28,757 INFO [optim.py:369] (6/7) 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 10:38:28,884 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=1918.0, num_to_drop=2, layers_to_drop={1, 3} 2023-04-26 10:38:32,525 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=1922.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 10:38:38,999 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=1931.0, num_to_drop=2, layers_to_drop={1, 2} 2023-04-26 10:38:50,104 INFO [finetune.py:976] (6/7) Epoch 1, batch 1950, loss[loss=0.6819, simple_loss=0.5467, pruned_loss=0.4109, over 4868.00 frames. ], tot_loss[loss=0.7686, simple_loss=0.5887, pruned_loss=0.5028, over 956141.43 frames. ], batch size: 34, lr: 4.00e-03, grad_scale: 1.0 2023-04-26 10:38:59,747 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=1967.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 10:39:22,531 INFO [finetune.py:976] (6/7) Epoch 1, batch 2000, loss[loss=0.5401, simple_loss=0.4644, pruned_loss=0.308, over 4721.00 frames. ], tot_loss[loss=0.7308, simple_loss=0.5688, pruned_loss=0.4687, over 957491.41 frames. ], batch size: 23, lr: 4.00e-03, grad_scale: 2.0 2023-04-26 10:39:39,762 INFO [optim.py:369] (6/7) 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 10:40:01,843 INFO [finetune.py:976] (6/7) Epoch 1, batch 2050, loss[loss=0.4812, simple_loss=0.4292, pruned_loss=0.2666, over 4891.00 frames. ], tot_loss[loss=0.6892, simple_loss=0.5463, pruned_loss=0.4334, over 958360.19 frames. ], batch size: 32, lr: 4.00e-03, grad_scale: 2.0 2023-04-26 10:40:11,913 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.0672, 1.1902, 1.2033, 1.2791, 1.6442, 1.0110, 0.6479, 1.2719], device='cuda:6'), covar=tensor([0.1033, 0.1388, 0.1115, 0.0839, 0.0546, 0.1011, 0.1185, 0.0781], device='cuda:6'), in_proj_covar=tensor([0.0201, 0.0211, 0.0190, 0.0169, 0.0169, 0.0186, 0.0164, 0.0181], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-26 10:40:38,144 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=2090.0, num_to_drop=2, layers_to_drop={1, 2} 2023-04-26 10:40:50,243 INFO [finetune.py:976] (6/7) Epoch 1, batch 2100, loss[loss=0.4718, simple_loss=0.4282, pruned_loss=0.2577, over 4752.00 frames. ], tot_loss[loss=0.6567, simple_loss=0.5303, pruned_loss=0.4051, over 956570.33 frames. ], batch size: 27, lr: 4.00e-03, grad_scale: 2.0 2023-04-26 10:41:00,192 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([0.8782, 0.5864, 0.6829, 0.7210, 0.5303, 0.5744, 0.4049, 0.4802], device='cuda:6'), covar=tensor([ 9.3179, 13.6415, 7.1070, 13.1379, 19.9752, 11.1989, 13.8917, 14.7242], device='cuda:6'), in_proj_covar=tensor([0.0280, 0.0295, 0.0236, 0.0373, 0.0257, 0.0245, 0.0292, 0.0241], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-26 10:41:11,499 INFO [optim.py:369] (6/7) 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 10:41:32,454 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=2138.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 10:41:39,118 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.79 vs. limit=2.0 2023-04-26 10:41:41,040 INFO [finetune.py:976] (6/7) Epoch 1, batch 2150, loss[loss=0.6239, simple_loss=0.5348, pruned_loss=0.3565, over 4826.00 frames. ], tot_loss[loss=0.6343, simple_loss=0.5217, pruned_loss=0.384, over 955275.92 frames. ], batch size: 33, lr: 4.00e-03, grad_scale: 4.0 2023-04-26 10:41:46,960 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([0.7509, 0.6487, 0.8662, 1.1752, 0.9704, 0.6966, 0.7000, 0.7123], device='cuda:6'), covar=tensor([36.3763, 56.8431, 54.2668, 34.4833, 46.7527, 79.6268, 96.0066, 52.6474], device='cuda:6'), in_proj_covar=tensor([0.0473, 0.0534, 0.0601, 0.0578, 0.0513, 0.0577, 0.0596, 0.0593], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-26 10:42:25,867 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=2194.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 10:42:30,518 INFO [finetune.py:976] (6/7) Epoch 1, batch 2200, loss[loss=0.4441, simple_loss=0.4205, pruned_loss=0.2339, over 4751.00 frames. ], tot_loss[loss=0.6116, simple_loss=0.5125, pruned_loss=0.3636, over 955150.45 frames. ], batch size: 27, lr: 4.00e-03, grad_scale: 4.0 2023-04-26 10:42:37,619 INFO [zipformer.py:1188] (6/7) 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] (6/7) 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 10:42:42,827 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=2222.0, num_to_drop=2, layers_to_drop={0, 1} 2023-04-26 10:42:45,140 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=2226.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 10:43:02,985 INFO [finetune.py:976] (6/7) Epoch 1, batch 2250, loss[loss=0.5153, simple_loss=0.4798, pruned_loss=0.2754, over 4918.00 frames. ], tot_loss[loss=0.5889, simple_loss=0.5019, pruned_loss=0.3444, over 953586.44 frames. ], batch size: 33, lr: 4.00e-03, grad_scale: 4.0 2023-04-26 10:43:12,390 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=2267.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 10:43:14,138 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=2270.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 10:43:22,334 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=5.09 vs. limit=5.0 2023-04-26 10:43:57,381 INFO [finetune.py:976] (6/7) Epoch 1, batch 2300, loss[loss=0.5092, simple_loss=0.4696, pruned_loss=0.2745, over 4917.00 frames. ], tot_loss[loss=0.5699, simple_loss=0.4931, pruned_loss=0.3283, over 953363.54 frames. ], batch size: 42, lr: 4.00e-03, grad_scale: 4.0 2023-04-26 10:44:17,315 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=2315.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 10:44:19,006 INFO [optim.py:369] (6/7) 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 10:44:43,396 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=2340.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 10:44:51,817 INFO [finetune.py:976] (6/7) Epoch 1, batch 2350, loss[loss=0.4023, simple_loss=0.3862, pruned_loss=0.2092, over 4856.00 frames. ], tot_loss[loss=0.5446, simple_loss=0.4784, pruned_loss=0.3093, over 955119.23 frames. ], batch size: 49, lr: 4.00e-03, grad_scale: 4.0 2023-04-26 10:45:40,788 INFO [finetune.py:976] (6/7) Epoch 1, batch 2400, loss[loss=0.4732, simple_loss=0.4308, pruned_loss=0.2578, over 4832.00 frames. ], tot_loss[loss=0.5242, simple_loss=0.4664, pruned_loss=0.2939, over 957084.90 frames. ], batch size: 33, lr: 4.00e-03, grad_scale: 8.0 2023-04-26 10:45:40,918 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=2401.0, num_to_drop=2, layers_to_drop={0, 2} 2023-04-26 10:45:51,693 INFO [optim.py:369] (6/7) 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 10:46:11,805 INFO [finetune.py:976] (6/7) Epoch 1, batch 2450, loss[loss=0.4648, simple_loss=0.4364, pruned_loss=0.2466, over 4826.00 frames. ], tot_loss[loss=0.5047, simple_loss=0.4546, pruned_loss=0.2797, over 956086.78 frames. ], batch size: 30, lr: 4.00e-03, grad_scale: 8.0 2023-04-26 10:46:26,837 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.3750, 3.3595, 0.8296, 1.9141, 1.8217, 2.3533, 2.1199, 1.0368], device='cuda:6'), covar=tensor([0.1188, 0.0666, 0.2017, 0.1100, 0.0969, 0.0952, 0.1219, 0.1944], device='cuda:6'), in_proj_covar=tensor([0.0121, 0.0264, 0.0146, 0.0130, 0.0140, 0.0161, 0.0128, 0.0130], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-26 10:46:39,130 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=2494.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 10:46:44,083 INFO [finetune.py:976] (6/7) Epoch 1, batch 2500, loss[loss=0.4594, simple_loss=0.4155, pruned_loss=0.2516, over 4215.00 frames. ], tot_loss[loss=0.4931, simple_loss=0.449, pruned_loss=0.2705, over 954514.80 frames. ], batch size: 18, lr: 4.00e-03, grad_scale: 8.0 2023-04-26 10:46:47,724 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.0247, 1.4860, 1.4314, 1.6123, 1.5289, 1.4126, 1.4442, 2.5396], device='cuda:6'), covar=tensor([0.0792, 0.0844, 0.0885, 0.1568, 0.0824, 0.0587, 0.0773, 0.0280], device='cuda:6'), in_proj_covar=tensor([0.0038, 0.0041, 0.0041, 0.0047, 0.0042, 0.0040, 0.0041, 0.0065], device='cuda:6'), out_proj_covar=tensor([0.0011, 0.0012, 0.0013, 0.0015, 0.0013, 0.0012, 0.0013, 0.0017], device='cuda:6') 2023-04-26 10:46:52,665 INFO [zipformer.py:1188] (6/7) 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] (6/7) 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 10:47:00,963 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=2526.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 10:47:01,539 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6822, 3.9661, 1.0153, 2.2269, 2.2562, 2.5534, 2.6325, 0.9507], device='cuda:6'), covar=tensor([0.1221, 0.0685, 0.1909, 0.1230, 0.0909, 0.1052, 0.1188, 0.2086], device='cuda:6'), in_proj_covar=tensor([0.0121, 0.0264, 0.0146, 0.0130, 0.0140, 0.0161, 0.0128, 0.0130], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-26 10:47:10,443 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=2542.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 10:47:12,201 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([0.9798, 1.3033, 3.3292, 3.1189, 3.0145, 3.2062, 3.2626, 2.9510], device='cuda:6'), covar=tensor([0.6273, 0.4992, 0.1283, 0.1706, 0.1187, 0.1587, 0.1285, 0.1538], device='cuda:6'), in_proj_covar=tensor([0.0345, 0.0318, 0.0456, 0.0468, 0.0382, 0.0435, 0.0350, 0.0404], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0002], device='cuda:6') 2023-04-26 10:47:16,162 INFO [finetune.py:976] (6/7) Epoch 1, batch 2550, loss[loss=0.4439, simple_loss=0.4387, pruned_loss=0.2245, over 4898.00 frames. ], tot_loss[loss=0.4866, simple_loss=0.4482, pruned_loss=0.2639, over 953794.48 frames. ], batch size: 37, lr: 4.00e-03, grad_scale: 8.0 2023-04-26 10:47:35,364 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=2561.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 10:47:50,050 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=2574.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 10:48:00,206 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=6.94 vs. limit=5.0 2023-04-26 10:48:03,126 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-04-26 10:48:06,449 INFO [finetune.py:976] (6/7) Epoch 1, batch 2600, loss[loss=0.4878, simple_loss=0.4646, pruned_loss=0.2554, over 4754.00 frames. ], tot_loss[loss=0.4788, simple_loss=0.4452, pruned_loss=0.2573, over 951277.00 frames. ], batch size: 27, lr: 4.00e-03, grad_scale: 8.0 2023-04-26 10:48:12,988 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.1904, 2.6109, 1.0261, 1.6230, 1.9274, 1.3661, 3.6464, 1.8537], device='cuda:6'), covar=tensor([0.0791, 0.0749, 0.0978, 0.1147, 0.0599, 0.1048, 0.0157, 0.0621], device='cuda:6'), in_proj_covar=tensor([0.0057, 0.0072, 0.0052, 0.0049, 0.0054, 0.0055, 0.0088, 0.0053], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0011, 0.0007], device='cuda:6') 2023-04-26 10:48:18,463 INFO [optim.py:369] (6/7) 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 10:48:26,699 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7092, 1.0416, 1.0365, 1.3852, 1.5998, 1.5778, 1.3589, 0.9855], device='cuda:6'), covar=tensor([0.1307, 0.3055, 0.3686, 0.1520, 0.1478, 0.1705, 0.2332, 0.3165], device='cuda:6'), in_proj_covar=tensor([0.0350, 0.0366, 0.0361, 0.0322, 0.0378, 0.0402, 0.0347, 0.0388], device='cuda:6'), out_proj_covar=tensor([7.6812e-05, 7.8918e-05, 7.7753e-05, 6.7247e-05, 8.0666e-05, 8.7996e-05, 7.5699e-05, 8.3949e-05], device='cuda:6') 2023-04-26 10:48:38,255 INFO [finetune.py:976] (6/7) Epoch 1, batch 2650, loss[loss=0.4512, simple_loss=0.4562, pruned_loss=0.2231, over 4905.00 frames. ], tot_loss[loss=0.4691, simple_loss=0.441, pruned_loss=0.2495, over 951934.15 frames. ], batch size: 46, lr: 4.00e-03, grad_scale: 8.0 2023-04-26 10:48:40,562 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.4368, 1.2308, 0.4943, 1.0645, 1.4414, 1.2875, 1.2481, 1.2217], device='cuda:6'), covar=tensor([0.0710, 0.0633, 0.0626, 0.0801, 0.0437, 0.0725, 0.0707, 0.1011], device='cuda:6'), in_proj_covar=tensor([0.0032, 0.0026, 0.0024, 0.0031, 0.0021, 0.0031, 0.0030, 0.0033], device='cuda:6'), out_proj_covar=tensor([0.0048, 0.0042, 0.0037, 0.0049, 0.0036, 0.0048, 0.0047, 0.0051], device='cuda:6') 2023-04-26 10:49:01,823 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=2685.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 10:49:14,016 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=2696.0, num_to_drop=1, layers_to_drop={2} 2023-04-26 10:49:16,397 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.1627, 0.7892, 0.8933, 1.1124, 0.7798, 0.8122, 0.6397, 0.6928], device='cuda:6'), covar=tensor([ 5.3720, 7.7858, 3.7849, 9.5256, 10.3348, 6.0912, 9.5362, 10.1018], device='cuda:6'), in_proj_covar=tensor([0.0269, 0.0287, 0.0228, 0.0358, 0.0249, 0.0238, 0.0284, 0.0236], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-26 10:49:16,821 INFO [finetune.py:976] (6/7) Epoch 1, batch 2700, loss[loss=0.427, simple_loss=0.4166, pruned_loss=0.2187, over 4820.00 frames. ], tot_loss[loss=0.4589, simple_loss=0.4353, pruned_loss=0.2419, over 952428.80 frames. ], batch size: 33, lr: 4.00e-03, grad_scale: 8.0 2023-04-26 10:49:39,497 INFO [optim.py:369] (6/7) 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 10:49:55,004 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2023-04-26 10:50:04,143 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=2746.0, num_to_drop=2, layers_to_drop={2, 3} 2023-04-26 10:50:13,026 INFO [finetune.py:976] (6/7) Epoch 1, batch 2750, loss[loss=0.423, simple_loss=0.4181, pruned_loss=0.214, over 4819.00 frames. ], tot_loss[loss=0.446, simple_loss=0.4268, pruned_loss=0.2331, over 953421.15 frames. ], batch size: 40, lr: 4.00e-03, grad_scale: 8.0 2023-04-26 10:50:14,324 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.4266, 1.3732, 1.5937, 1.7138, 1.7621, 1.3308, 0.8675, 1.5272], device='cuda:6'), covar=tensor([0.1201, 0.1371, 0.0904, 0.0908, 0.0712, 0.1114, 0.1446, 0.0898], device='cuda:6'), in_proj_covar=tensor([0.0204, 0.0211, 0.0191, 0.0174, 0.0173, 0.0189, 0.0168, 0.0185], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-26 10:51:08,916 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=5.87 vs. limit=5.0 2023-04-26 10:51:16,384 INFO [finetune.py:976] (6/7) Epoch 1, batch 2800, loss[loss=0.3139, simple_loss=0.321, pruned_loss=0.1534, over 4156.00 frames. ], tot_loss[loss=0.4331, simple_loss=0.4175, pruned_loss=0.2248, over 953820.66 frames. ], batch size: 18, lr: 4.00e-03, grad_scale: 8.0 2023-04-26 10:51:31,234 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.3698, 1.2030, 1.8542, 1.8831, 1.1874, 1.0344, 1.2864, 0.8808], device='cuda:6'), covar=tensor([0.1956, 0.1213, 0.0576, 0.0642, 0.1687, 0.2324, 0.1388, 0.1751], device='cuda:6'), in_proj_covar=tensor([0.0079, 0.0087, 0.0080, 0.0084, 0.0101, 0.0104, 0.0101, 0.0087], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:6') 2023-04-26 10:51:38,598 INFO [optim.py:369] (6/7) 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 10:51:49,423 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=2824.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 10:52:01,759 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7524, 2.6987, 1.3762, 1.5907, 1.2169, 1.2004, 1.3625, 1.0593], device='cuda:6'), covar=tensor([0.3865, 0.4036, 0.6080, 0.6883, 0.6238, 0.5362, 0.4636, 0.5687], device='cuda:6'), in_proj_covar=tensor([0.0206, 0.0230, 0.0210, 0.0226, 0.0245, 0.0205, 0.0201, 0.0218], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-26 10:52:24,239 INFO [finetune.py:976] (6/7) Epoch 1, batch 2850, loss[loss=0.396, simple_loss=0.3653, pruned_loss=0.2133, over 3975.00 frames. ], tot_loss[loss=0.4259, simple_loss=0.4128, pruned_loss=0.2198, over 952920.26 frames. ], batch size: 17, lr: 4.00e-03, grad_scale: 8.0 2023-04-26 10:53:07,947 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=2885.0, num_to_drop=2, layers_to_drop={1, 2} 2023-04-26 10:53:29,244 INFO [finetune.py:976] (6/7) Epoch 1, batch 2900, loss[loss=0.4126, simple_loss=0.4118, pruned_loss=0.2067, over 4122.00 frames. ], tot_loss[loss=0.4277, simple_loss=0.4165, pruned_loss=0.2197, over 954205.16 frames. ], batch size: 65, lr: 4.00e-03, grad_scale: 8.0 2023-04-26 10:53:42,437 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.4384, 2.8566, 1.1257, 1.8196, 2.1458, 1.6469, 3.8935, 2.0935], device='cuda:6'), covar=tensor([0.0614, 0.0611, 0.0924, 0.1112, 0.0577, 0.0929, 0.0155, 0.0600], device='cuda:6'), in_proj_covar=tensor([0.0057, 0.0072, 0.0052, 0.0049, 0.0054, 0.0055, 0.0088, 0.0053], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0011, 0.0007], device='cuda:6') 2023-04-26 10:53:46,014 INFO [optim.py:369] (6/7) 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 10:53:58,688 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=8.28 vs. limit=5.0 2023-04-26 10:54:00,313 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=2937.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 10:54:08,441 INFO [finetune.py:976] (6/7) Epoch 1, batch 2950, loss[loss=0.4261, simple_loss=0.4259, pruned_loss=0.2131, over 4849.00 frames. ], tot_loss[loss=0.4266, simple_loss=0.4183, pruned_loss=0.2176, over 955976.99 frames. ], batch size: 47, lr: 4.00e-03, grad_scale: 8.0 2023-04-26 10:54:20,424 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8295, 2.4507, 1.6498, 2.2447, 1.7470, 1.7842, 2.3519, 1.5326], device='cuda:6'), covar=tensor([0.2322, 0.1442, 0.1860, 0.1667, 0.3114, 0.1611, 0.1958, 0.3437], device='cuda:6'), in_proj_covar=tensor([0.0291, 0.0305, 0.0225, 0.0285, 0.0298, 0.0260, 0.0261, 0.0275], device='cuda:6'), out_proj_covar=tensor([1.1909e-04, 1.2504e-04, 9.2402e-05, 1.1528e-04, 1.2390e-04, 1.0542e-04, 1.0803e-04, 1.1199e-04], device='cuda:6') 2023-04-26 10:54:29,442 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.1811, 1.5479, 1.3788, 1.6015, 1.4641, 1.7766, 1.4953, 3.1775], device='cuda:6'), covar=tensor([0.0860, 0.0829, 0.0917, 0.1533, 0.0807, 0.0753, 0.0861, 0.0193], device='cuda:6'), in_proj_covar=tensor([0.0038, 0.0040, 0.0041, 0.0046, 0.0041, 0.0040, 0.0040, 0.0064], device='cuda:6'), out_proj_covar=tensor([0.0011, 0.0012, 0.0013, 0.0015, 0.0013, 0.0012, 0.0013, 0.0017], device='cuda:6') 2023-04-26 10:54:37,935 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=2996.0, num_to_drop=1, layers_to_drop={2} 2023-04-26 10:54:39,149 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=2998.0, num_to_drop=2, layers_to_drop={2, 3} 2023-04-26 10:54:40,895 INFO [finetune.py:976] (6/7) Epoch 1, batch 3000, loss[loss=0.4204, simple_loss=0.4253, pruned_loss=0.2078, over 4922.00 frames. ], tot_loss[loss=0.4256, simple_loss=0.4185, pruned_loss=0.2165, over 953728.48 frames. ], batch size: 33, lr: 4.00e-03, grad_scale: 8.0 2023-04-26 10:54:40,895 INFO [finetune.py:1001] (6/7) Computing validation loss 2023-04-26 10:54:51,383 INFO [finetune.py:1010] (6/7) Epoch 1, validation: loss=0.4217, simple_loss=0.4614, pruned_loss=0.191, over 2265189.00 frames. 2023-04-26 10:54:51,383 INFO [finetune.py:1011] (6/7) Maximum memory allocated so far is 5675MB 2023-04-26 10:54:53,294 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([3.7462, 3.6596, 2.7947, 4.2333, 3.7220, 3.6843, 1.5746, 3.6590], device='cuda:6'), covar=tensor([0.1487, 0.1041, 0.2789, 0.1528, 0.3247, 0.1677, 0.5068, 0.2007], device='cuda:6'), in_proj_covar=tensor([0.0261, 0.0232, 0.0283, 0.0326, 0.0320, 0.0271, 0.0283, 0.0284], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-26 10:54:53,908 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([4.8821, 4.8890, 3.3244, 5.5323, 4.9483, 4.8101, 2.5455, 4.7641], device='cuda:6'), covar=tensor([0.1202, 0.0682, 0.2521, 0.0876, 0.3163, 0.1478, 0.4456, 0.1614], device='cuda:6'), in_proj_covar=tensor([0.0261, 0.0232, 0.0283, 0.0326, 0.0320, 0.0271, 0.0283, 0.0284], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-26 10:54:58,727 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5788, 1.9558, 1.4326, 1.9749, 1.5606, 1.3723, 1.8913, 1.2156], device='cuda:6'), covar=tensor([0.2087, 0.1537, 0.1607, 0.1494, 0.2607, 0.1651, 0.1690, 0.2944], device='cuda:6'), in_proj_covar=tensor([0.0292, 0.0307, 0.0227, 0.0286, 0.0300, 0.0262, 0.0262, 0.0276], device='cuda:6'), out_proj_covar=tensor([1.1967e-04, 1.2588e-04, 9.2987e-05, 1.1589e-04, 1.2463e-04, 1.0599e-04, 1.0868e-04, 1.1263e-04], device='cuda:6') 2023-04-26 10:55:01,571 INFO [optim.py:369] (6/7) 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 10:55:02,286 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=3019.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 10:55:16,257 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=3041.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 10:55:18,055 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=3044.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 10:55:19,960 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=5.18 vs. limit=5.0 2023-04-26 10:55:23,631 INFO [finetune.py:976] (6/7) Epoch 1, batch 3050, loss[loss=0.4196, simple_loss=0.4201, pruned_loss=0.2095, over 4889.00 frames. ], tot_loss[loss=0.4196, simple_loss=0.4155, pruned_loss=0.212, over 955010.27 frames. ], batch size: 32, lr: 4.00e-03, grad_scale: 8.0 2023-04-26 10:55:42,255 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 1, batch 3100, loss[loss=0.3672, simple_loss=0.3753, pruned_loss=0.1796, over 4815.00 frames. ], tot_loss[loss=0.41, simple_loss=0.4085, pruned_loss=0.2058, over 956291.08 frames. ], batch size: 39, lr: 4.00e-03, grad_scale: 8.0 2023-04-26 10:56:12,962 INFO [optim.py:369] (6/7) 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 10:56:55,096 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([0.9195, 1.8446, 1.6512, 1.5795, 1.9813, 1.6323, 2.4325, 1.4528], device='cuda:6'), covar=tensor([0.3337, 0.1121, 0.3158, 0.2030, 0.1315, 0.1792, 0.0736, 0.2917], device='cuda:6'), in_proj_covar=tensor([0.0306, 0.0316, 0.0386, 0.0328, 0.0363, 0.0339, 0.0354, 0.0367], device='cuda:6'), out_proj_covar=tensor([9.4607e-05, 9.7524e-05, 1.1947e-04, 1.0252e-04, 1.1148e-04, 1.0349e-04, 1.0686e-04, 1.1379e-04], device='cuda:6') 2023-04-26 10:56:55,554 INFO [finetune.py:976] (6/7) Epoch 1, batch 3150, loss[loss=0.4038, simple_loss=0.403, pruned_loss=0.2023, over 4860.00 frames. ], tot_loss[loss=0.403, simple_loss=0.403, pruned_loss=0.2015, over 954651.69 frames. ], batch size: 49, lr: 4.00e-03, grad_scale: 8.0 2023-04-26 10:57:32,973 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=3180.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 10:57:46,512 INFO [finetune.py:976] (6/7) Epoch 1, batch 3200, loss[loss=0.3193, simple_loss=0.3454, pruned_loss=0.1467, over 4805.00 frames. ], tot_loss[loss=0.3932, simple_loss=0.3954, pruned_loss=0.1956, over 955352.50 frames. ], batch size: 29, lr: 4.00e-03, grad_scale: 8.0 2023-04-26 10:58:16,455 INFO [optim.py:369] (6/7) 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 10:58:52,917 INFO [finetune.py:976] (6/7) Epoch 1, batch 3250, loss[loss=0.3723, simple_loss=0.3811, pruned_loss=0.1818, over 4762.00 frames. ], tot_loss[loss=0.3897, simple_loss=0.3931, pruned_loss=0.1932, over 952023.03 frames. ], batch size: 28, lr: 4.00e-03, grad_scale: 8.0 2023-04-26 10:59:37,247 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([0.8714, 1.8495, 1.5308, 1.5538, 1.9868, 1.5644, 2.4050, 1.3867], device='cuda:6'), covar=tensor([0.3747, 0.1191, 0.3637, 0.2456, 0.1579, 0.2092, 0.0887, 0.3336], device='cuda:6'), in_proj_covar=tensor([0.0308, 0.0319, 0.0390, 0.0332, 0.0365, 0.0342, 0.0358, 0.0370], device='cuda:6'), out_proj_covar=tensor([9.5281e-05, 9.8495e-05, 1.2076e-04, 1.0381e-04, 1.1193e-04, 1.0441e-04, 1.0821e-04, 1.1491e-04], device='cuda:6') 2023-04-26 10:59:47,327 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=3288.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 10:59:50,311 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=3293.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:00:00,689 INFO [finetune.py:976] (6/7) Epoch 1, batch 3300, loss[loss=0.4033, simple_loss=0.4204, pruned_loss=0.1931, over 4922.00 frames. ], tot_loss[loss=0.3911, simple_loss=0.3967, pruned_loss=0.1928, over 953608.06 frames. ], batch size: 38, lr: 4.00e-03, grad_scale: 8.0 2023-04-26 11:00:17,468 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.3746, 2.0722, 1.2403, 1.2129, 1.0482, 1.0977, 1.1639, 0.9530], device='cuda:6'), covar=tensor([0.3051, 0.2847, 0.4251, 0.4696, 0.4718, 0.3577, 0.3139, 0.4360], device='cuda:6'), in_proj_covar=tensor([0.0203, 0.0225, 0.0207, 0.0222, 0.0241, 0.0202, 0.0198, 0.0216], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-26 11:00:19,090 INFO [optim.py:369] (6/7) 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 11:00:33,903 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=3341.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:00:38,725 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=3349.0, num_to_drop=2, layers_to_drop={0, 2} 2023-04-26 11:00:39,794 INFO [finetune.py:976] (6/7) Epoch 1, batch 3350, loss[loss=0.443, simple_loss=0.4366, pruned_loss=0.2247, over 4816.00 frames. ], tot_loss[loss=0.3906, simple_loss=0.3975, pruned_loss=0.1918, over 952136.02 frames. ], batch size: 38, lr: 4.00e-03, grad_scale: 8.0 2023-04-26 11:00:57,409 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=3375.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:01:02,851 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([0.9114, 0.9006, 1.0570, 1.4057, 1.1304, 1.0421, 1.1027, 1.0655], device='cuda:6'), covar=tensor([24.4837, 33.5085, 33.4207, 23.4942, 37.5030, 38.5220, 39.0263, 21.8948], device='cuda:6'), in_proj_covar=tensor([0.0472, 0.0540, 0.0616, 0.0586, 0.0511, 0.0578, 0.0588, 0.0589], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-26 11:01:05,773 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=3389.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:01:09,961 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.2048, 1.2677, 5.1760, 4.8479, 4.5856, 4.8864, 4.7032, 4.5682], device='cuda:6'), covar=tensor([0.5929, 0.6092, 0.0897, 0.1592, 0.0932, 0.0793, 0.1028, 0.1285], device='cuda:6'), in_proj_covar=tensor([0.0344, 0.0320, 0.0458, 0.0467, 0.0386, 0.0439, 0.0350, 0.0409], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0002], device='cuda:6') 2023-04-26 11:01:13,016 INFO [finetune.py:976] (6/7) Epoch 1, batch 3400, loss[loss=0.4229, simple_loss=0.4311, pruned_loss=0.2074, over 4894.00 frames. ], tot_loss[loss=0.3887, simple_loss=0.3975, pruned_loss=0.1899, over 953478.97 frames. ], batch size: 35, lr: 4.00e-03, grad_scale: 8.0 2023-04-26 11:01:16,125 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7157, 1.1878, 1.2301, 1.1692, 1.8241, 1.7308, 1.2318, 1.2454], device='cuda:6'), covar=tensor([0.0755, 0.1030, 0.1471, 0.1081, 0.0427, 0.0675, 0.1168, 0.1266], device='cuda:6'), in_proj_covar=tensor([0.0335, 0.0346, 0.0347, 0.0308, 0.0358, 0.0380, 0.0330, 0.0366], device='cuda:6'), out_proj_covar=tensor([7.3311e-05, 7.4487e-05, 7.4806e-05, 6.4502e-05, 7.6383e-05, 8.3162e-05, 7.1909e-05, 7.9209e-05], device='cuda:6') 2023-04-26 11:01:23,590 INFO [optim.py:369] (6/7) 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 11:01:29,298 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=5.16 vs. limit=5.0 2023-04-26 11:01:57,865 INFO [finetune.py:976] (6/7) Epoch 1, batch 3450, loss[loss=0.3829, simple_loss=0.3952, pruned_loss=0.1853, over 4815.00 frames. ], tot_loss[loss=0.3851, simple_loss=0.3952, pruned_loss=0.1875, over 953856.73 frames. ], batch size: 25, lr: 4.00e-03, grad_scale: 8.0 2023-04-26 11:02:05,185 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6327, 1.0403, 1.3211, 1.3982, 1.2605, 1.1028, 0.6165, 1.0367], device='cuda:6'), covar=tensor([0.6901, 0.8909, 0.3985, 0.9301, 0.7981, 0.6167, 1.0683, 0.7233], device='cuda:6'), in_proj_covar=tensor([0.0264, 0.0285, 0.0225, 0.0350, 0.0245, 0.0236, 0.0280, 0.0233], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-26 11:02:36,479 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=3480.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:02:38,924 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.4040, 1.4956, 1.7307, 1.6664, 1.9051, 1.3270, 0.9066, 1.5501], device='cuda:6'), covar=tensor([0.1202, 0.1187, 0.0780, 0.0985, 0.0679, 0.1218, 0.1459, 0.0870], device='cuda:6'), in_proj_covar=tensor([0.0204, 0.0208, 0.0188, 0.0175, 0.0173, 0.0190, 0.0169, 0.0185], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-26 11:02:46,126 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5031, 1.5498, 0.7506, 1.1832, 1.8398, 1.3785, 1.3337, 1.3783], device='cuda:6'), covar=tensor([0.0655, 0.0530, 0.0574, 0.0705, 0.0374, 0.0678, 0.0603, 0.0855], device='cuda:6'), in_proj_covar=tensor([0.0032, 0.0027, 0.0025, 0.0031, 0.0021, 0.0031, 0.0030, 0.0033], device='cuda:6'), out_proj_covar=tensor([0.0048, 0.0043, 0.0038, 0.0049, 0.0037, 0.0048, 0.0047, 0.0051], device='cuda:6') 2023-04-26 11:02:49,060 INFO [finetune.py:976] (6/7) Epoch 1, batch 3500, loss[loss=0.3653, simple_loss=0.3724, pruned_loss=0.1791, over 4827.00 frames. ], tot_loss[loss=0.383, simple_loss=0.3925, pruned_loss=0.1867, over 953643.25 frames. ], batch size: 33, lr: 4.00e-03, grad_scale: 8.0 2023-04-26 11:03:00,464 INFO [optim.py:369] (6/7) 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 11:03:07,524 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=3528.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:03:24,413 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=3546.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 11:03:27,267 INFO [finetune.py:976] (6/7) Epoch 1, batch 3550, loss[loss=0.3604, simple_loss=0.3696, pruned_loss=0.1756, over 4756.00 frames. ], tot_loss[loss=0.3789, simple_loss=0.3883, pruned_loss=0.1847, over 952050.91 frames. ], batch size: 54, lr: 4.00e-03, grad_scale: 8.0 2023-04-26 11:03:34,928 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7012, 1.6274, 1.9075, 1.8908, 2.1046, 1.5427, 1.1667, 1.7648], device='cuda:6'), covar=tensor([0.1120, 0.1181, 0.0702, 0.0906, 0.0719, 0.1211, 0.1379, 0.0796], device='cuda:6'), in_proj_covar=tensor([0.0205, 0.0208, 0.0189, 0.0176, 0.0174, 0.0191, 0.0170, 0.0186], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-26 11:04:02,487 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=5.43 vs. limit=5.0 2023-04-26 11:04:11,237 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=3593.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:04:16,043 INFO [finetune.py:976] (6/7) Epoch 1, batch 3600, loss[loss=0.4069, simple_loss=0.4196, pruned_loss=0.1971, over 4916.00 frames. ], tot_loss[loss=0.3735, simple_loss=0.384, pruned_loss=0.1816, over 955016.59 frames. ], batch size: 37, lr: 4.00e-03, grad_scale: 8.0 2023-04-26 11:04:19,846 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=3607.0, num_to_drop=2, layers_to_drop={2, 3} 2023-04-26 11:04:22,858 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([0.9268, 1.3910, 1.2322, 1.6357, 1.5073, 1.4349, 1.3908, 2.5579], device='cuda:6'), covar=tensor([0.0758, 0.0835, 0.0861, 0.1376, 0.0743, 0.0572, 0.0759, 0.0227], device='cuda:6'), in_proj_covar=tensor([0.0039, 0.0040, 0.0041, 0.0046, 0.0041, 0.0040, 0.0041, 0.0064], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0017], device='cuda:6') 2023-04-26 11:04:24,669 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.0654, 1.9017, 1.5165, 1.8141, 2.0651, 1.6476, 2.3703, 1.3337], device='cuda:6'), covar=tensor([0.3315, 0.1266, 0.3630, 0.2111, 0.1381, 0.1962, 0.0935, 0.3329], device='cuda:6'), in_proj_covar=tensor([0.0313, 0.0324, 0.0395, 0.0336, 0.0370, 0.0346, 0.0364, 0.0377], device='cuda:6'), out_proj_covar=tensor([9.6684e-05, 9.9992e-05, 1.2250e-04, 1.0491e-04, 1.1361e-04, 1.0587e-04, 1.0990e-04, 1.1695e-04], device='cuda:6') 2023-04-26 11:04:26,349 INFO [optim.py:369] (6/7) 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 11:04:42,924 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=3641.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:04:42,962 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0116, 2.3565, 1.0447, 1.3385, 1.6505, 1.3321, 3.0405, 1.6774], device='cuda:6'), covar=tensor([0.0743, 0.0579, 0.0817, 0.1164, 0.0576, 0.0984, 0.0186, 0.0635], device='cuda:6'), in_proj_covar=tensor([0.0057, 0.0073, 0.0053, 0.0050, 0.0055, 0.0055, 0.0088, 0.0053], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0012, 0.0008], device='cuda:6') 2023-04-26 11:04:44,784 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=3644.0, num_to_drop=1, layers_to_drop={2} 2023-04-26 11:04:48,957 INFO [finetune.py:976] (6/7) Epoch 1, batch 3650, loss[loss=0.2872, simple_loss=0.3199, pruned_loss=0.1272, over 4758.00 frames. ], tot_loss[loss=0.3745, simple_loss=0.3859, pruned_loss=0.1816, over 954685.39 frames. ], batch size: 28, lr: 4.00e-03, grad_scale: 8.0 2023-04-26 11:04:59,850 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-04-26 11:05:04,835 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=3675.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:05:05,412 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=3676.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 11:05:23,023 INFO [finetune.py:976] (6/7) Epoch 1, batch 3700, loss[loss=0.3321, simple_loss=0.3529, pruned_loss=0.1556, over 4803.00 frames. ], tot_loss[loss=0.3758, simple_loss=0.389, pruned_loss=0.1814, over 956535.17 frames. ], batch size: 25, lr: 4.00e-03, grad_scale: 8.0 2023-04-26 11:05:29,743 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([0.7892, 0.5146, 0.5303, 0.4707, 0.5512, 0.5950, 0.4850, 0.5410], device='cuda:6'), covar=tensor([ 8.3525, 25.5332, 16.3460, 15.4783, 13.6485, 17.4936, 25.0590, 18.8414], device='cuda:6'), in_proj_covar=tensor([0.0271, 0.0387, 0.0315, 0.0312, 0.0342, 0.0351, 0.0379, 0.0344], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-26 11:05:44,465 INFO [optim.py:369] (6/7) 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 11:05:53,052 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=3723.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:06:12,559 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=3737.0, num_to_drop=2, layers_to_drop={0, 2} 2023-04-26 11:06:14,860 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([3.3429, 3.3025, 2.6686, 3.7806, 3.2825, 3.2639, 1.5042, 3.1724], device='cuda:6'), covar=tensor([0.1732, 0.1237, 0.3111, 0.2221, 0.2196, 0.1969, 0.4902, 0.2352], device='cuda:6'), in_proj_covar=tensor([0.0264, 0.0233, 0.0285, 0.0329, 0.0323, 0.0273, 0.0287, 0.0287], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-26 11:06:21,955 INFO [finetune.py:976] (6/7) Epoch 1, batch 3750, loss[loss=0.4209, simple_loss=0.4297, pruned_loss=0.2061, over 4807.00 frames. ], tot_loss[loss=0.3748, simple_loss=0.3892, pruned_loss=0.1802, over 957260.93 frames. ], batch size: 41, lr: 4.00e-03, grad_scale: 8.0 2023-04-26 11:06:58,913 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-04-26 11:07:07,655 INFO [finetune.py:976] (6/7) Epoch 1, batch 3800, loss[loss=0.3624, simple_loss=0.3733, pruned_loss=0.1757, over 4724.00 frames. ], tot_loss[loss=0.3712, simple_loss=0.3875, pruned_loss=0.1774, over 956154.30 frames. ], batch size: 54, lr: 4.00e-03, grad_scale: 8.0 2023-04-26 11:07:29,603 INFO [optim.py:369] (6/7) 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] (6/7) Epoch 1, batch 3850, loss[loss=0.3235, simple_loss=0.3567, pruned_loss=0.1451, over 4914.00 frames. ], tot_loss[loss=0.3651, simple_loss=0.3831, pruned_loss=0.1736, over 955308.89 frames. ], batch size: 36, lr: 4.00e-03, grad_scale: 8.0 2023-04-26 11:08:47,568 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.3987, 0.7129, 0.7929, 0.7210, 0.8005, 0.9421, 0.7262, 0.8630], device='cuda:6'), covar=tensor([ 6.1940, 18.7719, 12.4444, 11.0556, 11.1569, 14.6474, 18.3567, 13.9065], device='cuda:6'), in_proj_covar=tensor([0.0258, 0.0369, 0.0301, 0.0298, 0.0327, 0.0338, 0.0360, 0.0328], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-26 11:09:22,800 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-04-26 11:09:22,875 INFO [finetune.py:976] (6/7) Epoch 1, batch 3900, loss[loss=0.3163, simple_loss=0.3503, pruned_loss=0.1411, over 4822.00 frames. ], tot_loss[loss=0.3599, simple_loss=0.3783, pruned_loss=0.1708, over 956051.03 frames. ], batch size: 40, lr: 4.00e-03, grad_scale: 8.0 2023-04-26 11:09:29,502 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=3902.0, num_to_drop=1, layers_to_drop={3} 2023-04-26 11:09:32,083 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-04-26 11:09:44,795 INFO [optim.py:369] (6/7) 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 11:09:49,800 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=3926.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:10:02,217 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=3944.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 11:10:07,403 INFO [finetune.py:976] (6/7) Epoch 1, batch 3950, loss[loss=0.2896, simple_loss=0.3264, pruned_loss=0.1263, over 4748.00 frames. ], tot_loss[loss=0.3529, simple_loss=0.3722, pruned_loss=0.1668, over 954366.64 frames. ], batch size: 27, lr: 4.00e-03, grad_scale: 8.0 2023-04-26 11:10:30,558 INFO [zipformer.py:1188] (6/7) 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] (6/7) 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] (6/7) Epoch 1, batch 4000, loss[loss=0.272, simple_loss=0.3156, pruned_loss=0.1142, over 4768.00 frames. ], tot_loss[loss=0.3459, simple_loss=0.3667, pruned_loss=0.1625, over 953316.81 frames. ], batch size: 28, lr: 4.00e-03, grad_scale: 8.0 2023-04-26 11:10:54,138 INFO [optim.py:369] (6/7) 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 11:11:02,817 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=4032.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 11:11:15,866 INFO [finetune.py:976] (6/7) Epoch 1, batch 4050, loss[loss=0.3372, simple_loss=0.3782, pruned_loss=0.1482, over 4828.00 frames. ], tot_loss[loss=0.3485, simple_loss=0.3699, pruned_loss=0.1636, over 953710.70 frames. ], batch size: 33, lr: 4.00e-03, grad_scale: 8.0 2023-04-26 11:11:41,986 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.97 vs. limit=2.0 2023-04-26 11:11:49,877 INFO [finetune.py:976] (6/7) Epoch 1, batch 4100, loss[loss=0.3603, simple_loss=0.3923, pruned_loss=0.1642, over 4928.00 frames. ], tot_loss[loss=0.3505, simple_loss=0.3733, pruned_loss=0.1638, over 955599.69 frames. ], batch size: 38, lr: 4.00e-03, grad_scale: 8.0 2023-04-26 11:12:08,259 INFO [optim.py:369] (6/7) 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 11:12:19,226 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.60 vs. limit=2.0 2023-04-26 11:12:34,766 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=4141.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:12:34,790 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8875, 1.0425, 1.3350, 2.1010, 1.5417, 1.1407, 0.9478, 1.5490], device='cuda:6'), covar=tensor([0.6670, 1.0112, 0.4765, 1.4848, 1.0259, 0.7740, 1.9828, 0.9956], device='cuda:6'), in_proj_covar=tensor([0.0252, 0.0273, 0.0213, 0.0330, 0.0232, 0.0225, 0.0267, 0.0220], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-26 11:12:41,353 INFO [finetune.py:976] (6/7) Epoch 1, batch 4150, loss[loss=0.3076, simple_loss=0.3564, pruned_loss=0.1294, over 4812.00 frames. ], tot_loss[loss=0.3493, simple_loss=0.3732, pruned_loss=0.1627, over 956887.92 frames. ], batch size: 40, lr: 4.00e-03, grad_scale: 16.0 2023-04-26 11:12:56,139 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-04-26 11:13:01,577 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.1395, 2.5643, 1.0878, 1.3145, 1.7861, 1.2284, 3.4919, 1.6304], device='cuda:6'), covar=tensor([0.0700, 0.0692, 0.0913, 0.1211, 0.0592, 0.1033, 0.0159, 0.0662], device='cuda:6'), in_proj_covar=tensor([0.0057, 0.0073, 0.0054, 0.0050, 0.0055, 0.0056, 0.0088, 0.0054], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0008, 0.0008, 0.0008, 0.0008, 0.0012, 0.0008], device='cuda:6') 2023-04-26 11:13:15,086 INFO [finetune.py:976] (6/7) Epoch 1, batch 4200, loss[loss=0.2997, simple_loss=0.3397, pruned_loss=0.1299, over 4891.00 frames. ], tot_loss[loss=0.346, simple_loss=0.3712, pruned_loss=0.1604, over 956804.78 frames. ], batch size: 32, lr: 4.00e-03, grad_scale: 16.0 2023-04-26 11:13:16,198 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=4202.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 11:13:16,222 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=4202.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:13:27,101 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([0.9761, 1.5104, 1.3668, 1.7364, 1.5537, 1.8145, 1.5031, 3.1234], device='cuda:6'), covar=tensor([0.0785, 0.0745, 0.0773, 0.1277, 0.0671, 0.0500, 0.0711, 0.0192], device='cuda:6'), in_proj_covar=tensor([0.0039, 0.0040, 0.0041, 0.0046, 0.0041, 0.0040, 0.0041, 0.0064], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0017], device='cuda:6') 2023-04-26 11:13:28,643 INFO [optim.py:369] (6/7) 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 11:13:31,371 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=5.17 vs. limit=5.0 2023-04-26 11:13:33,115 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8362, 2.1441, 1.6061, 2.1051, 1.6682, 1.6373, 1.9669, 1.3811], device='cuda:6'), covar=tensor([0.2234, 0.2044, 0.1867, 0.1793, 0.3753, 0.2565, 0.2151, 0.4048], device='cuda:6'), in_proj_covar=tensor([0.0303, 0.0320, 0.0237, 0.0300, 0.0306, 0.0274, 0.0272, 0.0292], device='cuda:6'), out_proj_covar=tensor([1.2424e-04, 1.3092e-04, 9.7104e-05, 1.2125e-04, 1.2681e-04, 1.1090e-04, 1.1307e-04, 1.1900e-04], device='cuda:6') 2023-04-26 11:13:41,444 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([0.5430, 0.4823, 0.6031, 0.8684, 0.7385, 0.6093, 0.6123, 0.6511], device='cuda:6'), covar=tensor([16.1631, 23.8815, 20.6219, 16.9760, 19.7830, 23.6374, 24.3707, 14.9488], device='cuda:6'), in_proj_covar=tensor([0.0418, 0.0477, 0.0545, 0.0519, 0.0445, 0.0503, 0.0510, 0.0518], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-26 11:14:05,470 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=4250.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 11:14:05,987 INFO [finetune.py:976] (6/7) Epoch 1, batch 4250, loss[loss=0.3912, simple_loss=0.3892, pruned_loss=0.1966, over 4880.00 frames. ], tot_loss[loss=0.3404, simple_loss=0.3664, pruned_loss=0.1572, over 957878.13 frames. ], batch size: 32, lr: 4.00e-03, grad_scale: 16.0 2023-04-26 11:14:27,363 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=5.72 vs. limit=5.0 2023-04-26 11:14:51,195 INFO [zipformer.py:1188] (6/7) 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] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=4288.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:15:07,277 INFO [finetune.py:976] (6/7) Epoch 1, batch 4300, loss[loss=0.2915, simple_loss=0.3219, pruned_loss=0.1306, over 4851.00 frames. ], tot_loss[loss=0.3343, simple_loss=0.3608, pruned_loss=0.1539, over 958345.85 frames. ], batch size: 44, lr: 4.00e-03, grad_scale: 16.0 2023-04-26 11:15:19,787 INFO [optim.py:369] (6/7) 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 11:15:23,487 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([0.6304, 1.6980, 1.7224, 1.4087, 1.8644, 1.4307, 2.3317, 1.5654], device='cuda:6'), covar=tensor([0.4123, 0.1404, 0.3884, 0.2306, 0.1505, 0.2151, 0.0947, 0.3751], device='cuda:6'), in_proj_covar=tensor([0.0320, 0.0332, 0.0406, 0.0344, 0.0378, 0.0352, 0.0373, 0.0384], device='cuda:6'), out_proj_covar=tensor([9.8857e-05, 1.0266e-04, 1.2573e-04, 1.0745e-04, 1.1598e-04, 1.0745e-04, 1.1281e-04, 1.1885e-04], device='cuda:6') 2023-04-26 11:15:29,473 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=4332.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 11:15:31,985 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5925, 1.5335, 0.6090, 1.2675, 1.8411, 1.4873, 1.3811, 1.4284], device='cuda:6'), covar=tensor([0.0652, 0.0511, 0.0581, 0.0680, 0.0368, 0.0655, 0.0617, 0.0803], device='cuda:6'), in_proj_covar=tensor([0.0032, 0.0027, 0.0025, 0.0031, 0.0022, 0.0031, 0.0030, 0.0033], device='cuda:6'), out_proj_covar=tensor([0.0048, 0.0043, 0.0038, 0.0049, 0.0037, 0.0048, 0.0047, 0.0051], device='cuda:6') 2023-04-26 11:15:39,948 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=4349.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:15:41,062 INFO [finetune.py:976] (6/7) Epoch 1, batch 4350, loss[loss=0.339, simple_loss=0.3517, pruned_loss=0.1631, over 4718.00 frames. ], tot_loss[loss=0.3281, simple_loss=0.3552, pruned_loss=0.1505, over 958942.32 frames. ], batch size: 59, lr: 4.00e-03, grad_scale: 16.0 2023-04-26 11:16:01,300 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=5.12 vs. limit=5.0 2023-04-26 11:16:02,413 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=4380.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 11:16:15,322 INFO [finetune.py:976] (6/7) Epoch 1, batch 4400, loss[loss=0.2685, simple_loss=0.3072, pruned_loss=0.1149, over 4807.00 frames. ], tot_loss[loss=0.33, simple_loss=0.3567, pruned_loss=0.1517, over 958464.30 frames. ], batch size: 25, lr: 4.00e-03, grad_scale: 16.0 2023-04-26 11:16:26,705 INFO [optim.py:369] (6/7) 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 11:16:48,905 INFO [finetune.py:976] (6/7) Epoch 1, batch 4450, loss[loss=0.302, simple_loss=0.3519, pruned_loss=0.1261, over 4795.00 frames. ], tot_loss[loss=0.3332, simple_loss=0.3606, pruned_loss=0.1529, over 958834.06 frames. ], batch size: 29, lr: 4.00e-03, grad_scale: 16.0 2023-04-26 11:16:55,145 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.4429, 1.2293, 1.8780, 1.6465, 1.3435, 1.1146, 1.3894, 0.9219], device='cuda:6'), covar=tensor([0.1047, 0.1217, 0.0558, 0.0843, 0.1253, 0.1637, 0.0956, 0.1471], device='cuda:6'), in_proj_covar=tensor([0.0074, 0.0080, 0.0076, 0.0078, 0.0093, 0.0097, 0.0094, 0.0082], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:6') 2023-04-26 11:17:19,620 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=4497.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:17:22,029 INFO [finetune.py:976] (6/7) Epoch 1, batch 4500, loss[loss=0.3807, simple_loss=0.4056, pruned_loss=0.1779, over 4796.00 frames. ], tot_loss[loss=0.3347, simple_loss=0.3629, pruned_loss=0.1532, over 957890.28 frames. ], batch size: 51, lr: 4.00e-03, grad_scale: 16.0 2023-04-26 11:17:32,938 INFO [optim.py:369] (6/7) 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 11:18:14,109 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.63 vs. limit=2.0 2023-04-26 11:18:15,788 INFO [finetune.py:976] (6/7) Epoch 1, batch 4550, loss[loss=0.3356, simple_loss=0.3677, pruned_loss=0.1517, over 4719.00 frames. ], tot_loss[loss=0.3353, simple_loss=0.3634, pruned_loss=0.1536, over 955479.59 frames. ], batch size: 59, lr: 4.00e-03, grad_scale: 16.0 2023-04-26 11:18:35,400 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.84 vs. limit=5.0 2023-04-26 11:19:01,275 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=4582.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:19:13,033 INFO [finetune.py:976] (6/7) Epoch 1, batch 4600, loss[loss=0.3084, simple_loss=0.3419, pruned_loss=0.1375, over 4814.00 frames. ], tot_loss[loss=0.3309, simple_loss=0.3601, pruned_loss=0.1508, over 955715.63 frames. ], batch size: 40, lr: 4.00e-03, grad_scale: 16.0 2023-04-26 11:19:23,471 INFO [optim.py:369] (6/7) 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 11:19:33,262 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=4630.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:19:36,727 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([0.9774, 1.9786, 1.7656, 1.6570, 2.1700, 1.6470, 2.5837, 1.5778], device='cuda:6'), covar=tensor([0.4138, 0.1403, 0.4612, 0.2668, 0.1536, 0.2419, 0.1097, 0.3628], device='cuda:6'), in_proj_covar=tensor([0.0324, 0.0336, 0.0411, 0.0348, 0.0383, 0.0356, 0.0378, 0.0389], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-26 11:19:37,963 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5369, 1.2398, 0.7151, 1.1850, 1.4011, 1.4445, 1.2932, 1.3505], device='cuda:6'), covar=tensor([0.0640, 0.0532, 0.0548, 0.0691, 0.0391, 0.0651, 0.0607, 0.0787], device='cuda:6'), in_proj_covar=tensor([0.0032, 0.0027, 0.0025, 0.0031, 0.0022, 0.0031, 0.0030, 0.0033], device='cuda:6'), out_proj_covar=tensor([0.0048, 0.0043, 0.0038, 0.0049, 0.0037, 0.0048, 0.0047, 0.0051], device='cuda:6') 2023-04-26 11:19:42,843 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 1, batch 4650, loss[loss=0.3197, simple_loss=0.3428, pruned_loss=0.1483, over 4684.00 frames. ], tot_loss[loss=0.324, simple_loss=0.3537, pruned_loss=0.1472, over 955029.22 frames. ], batch size: 23, lr: 4.00e-03, grad_scale: 16.0 2023-04-26 11:19:54,077 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9070, 1.7017, 2.4210, 2.6579, 1.7652, 1.2714, 1.9060, 1.1144], device='cuda:6'), covar=tensor([0.1102, 0.1189, 0.0614, 0.0592, 0.1367, 0.1688, 0.1123, 0.1850], device='cuda:6'), in_proj_covar=tensor([0.0074, 0.0081, 0.0077, 0.0078, 0.0093, 0.0097, 0.0094, 0.0082], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:6') 2023-04-26 11:19:54,650 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5357, 1.2817, 4.1318, 3.7886, 3.6725, 3.8642, 3.9128, 3.6415], device='cuda:6'), covar=tensor([0.6857, 0.6096, 0.1133, 0.1927, 0.1087, 0.1440, 0.1536, 0.1656], device='cuda:6'), in_proj_covar=tensor([0.0344, 0.0320, 0.0456, 0.0465, 0.0384, 0.0440, 0.0347, 0.0411], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0002], device='cuda:6') 2023-04-26 11:20:19,007 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8307, 1.0521, 1.3667, 1.8467, 1.3333, 1.0784, 0.8109, 1.2565], device='cuda:6'), covar=tensor([0.9364, 1.0234, 0.5045, 1.4624, 1.0995, 0.8366, 1.7597, 1.0770], device='cuda:6'), in_proj_covar=tensor([0.0251, 0.0271, 0.0212, 0.0329, 0.0230, 0.0224, 0.0266, 0.0218], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-26 11:20:22,578 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.85 vs. limit=5.0 2023-04-26 11:20:42,429 INFO [finetune.py:976] (6/7) Epoch 1, batch 4700, loss[loss=0.2982, simple_loss=0.3272, pruned_loss=0.1345, over 4928.00 frames. ], tot_loss[loss=0.3172, simple_loss=0.3479, pruned_loss=0.1433, over 957253.49 frames. ], batch size: 33, lr: 4.00e-03, grad_scale: 16.0 2023-04-26 11:21:03,816 INFO [optim.py:369] (6/7) 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 11:21:39,265 INFO [finetune.py:976] (6/7) Epoch 1, batch 4750, loss[loss=0.2741, simple_loss=0.305, pruned_loss=0.1216, over 4725.00 frames. ], tot_loss[loss=0.314, simple_loss=0.3451, pruned_loss=0.1415, over 953929.45 frames. ], batch size: 23, lr: 4.00e-03, grad_scale: 16.0 2023-04-26 11:22:07,688 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.95 vs. limit=5.0 2023-04-26 11:22:11,038 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 1, batch 4800, loss[loss=0.3153, simple_loss=0.3464, pruned_loss=0.1421, over 4871.00 frames. ], tot_loss[loss=0.3178, simple_loss=0.3487, pruned_loss=0.1434, over 954668.18 frames. ], batch size: 31, lr: 4.00e-03, grad_scale: 16.0 2023-04-26 11:22:24,009 INFO [optim.py:369] (6/7) 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 11:22:43,636 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=4845.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:22:47,737 INFO [finetune.py:976] (6/7) Epoch 1, batch 4850, loss[loss=0.3294, simple_loss=0.3561, pruned_loss=0.1514, over 4821.00 frames. ], tot_loss[loss=0.3194, simple_loss=0.3512, pruned_loss=0.1438, over 953600.71 frames. ], batch size: 45, lr: 4.00e-03, grad_scale: 16.0 2023-04-26 11:23:08,810 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.98 vs. limit=2.0 2023-04-26 11:23:25,284 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-04-26 11:23:33,605 INFO [finetune.py:976] (6/7) Epoch 1, batch 4900, loss[loss=0.34, simple_loss=0.3801, pruned_loss=0.15, over 4803.00 frames. ], tot_loss[loss=0.3207, simple_loss=0.3528, pruned_loss=0.1443, over 953762.99 frames. ], batch size: 41, lr: 4.00e-03, grad_scale: 16.0 2023-04-26 11:23:49,559 INFO [optim.py:369] (6/7) 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 11:24:17,630 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([3.5571, 3.5606, 2.8512, 4.1029, 3.4380, 3.5385, 1.7795, 3.4494], device='cuda:6'), covar=tensor([0.1611, 0.1035, 0.3364, 0.1468, 0.2389, 0.1684, 0.4933, 0.2265], device='cuda:6'), in_proj_covar=tensor([0.0263, 0.0233, 0.0284, 0.0328, 0.0323, 0.0272, 0.0287, 0.0287], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-26 11:24:29,545 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=4944.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:24:41,251 INFO [finetune.py:976] (6/7) Epoch 1, batch 4950, loss[loss=0.3006, simple_loss=0.3387, pruned_loss=0.1312, over 4895.00 frames. ], tot_loss[loss=0.3192, simple_loss=0.3523, pruned_loss=0.1431, over 953728.08 frames. ], batch size: 35, lr: 4.00e-03, grad_scale: 16.0 2023-04-26 11:25:10,630 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.3520, 1.6014, 1.1091, 1.0731, 1.0667, 1.0576, 1.1266, 0.9793], device='cuda:6'), covar=tensor([0.2094, 0.2093, 0.2924, 0.3099, 0.3472, 0.2426, 0.1927, 0.2957], device='cuda:6'), in_proj_covar=tensor([0.0197, 0.0221, 0.0200, 0.0216, 0.0236, 0.0197, 0.0193, 0.0209], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-26 11:25:19,473 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=4992.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:25:23,047 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([0.2901, 1.3582, 1.4121, 1.1593, 1.4561, 1.0257, 1.8497, 1.3588], device='cuda:6'), covar=tensor([0.3962, 0.1656, 0.4632, 0.2274, 0.1623, 0.2216, 0.1398, 0.4025], device='cuda:6'), in_proj_covar=tensor([0.0328, 0.0339, 0.0417, 0.0351, 0.0387, 0.0360, 0.0384, 0.0393], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-26 11:25:26,586 INFO [finetune.py:976] (6/7) Epoch 1, batch 5000, loss[loss=0.3017, simple_loss=0.3361, pruned_loss=0.1337, over 4774.00 frames. ], tot_loss[loss=0.3165, simple_loss=0.3501, pruned_loss=0.1414, over 954010.76 frames. ], batch size: 28, lr: 4.00e-03, grad_scale: 16.0 2023-04-26 11:25:37,463 INFO [optim.py:369] (6/7) 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] (6/7) Epoch 1, batch 5050, loss[loss=0.2459, simple_loss=0.294, pruned_loss=0.09892, over 4757.00 frames. ], tot_loss[loss=0.3109, simple_loss=0.3451, pruned_loss=0.1384, over 955786.22 frames. ], batch size: 28, lr: 4.00e-03, grad_scale: 16.0 2023-04-26 11:26:23,274 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=5.01 vs. limit=5.0 2023-04-26 11:26:24,299 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.90 vs. limit=5.0 2023-04-26 11:26:58,680 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([3.3776, 3.3551, 2.6502, 3.8689, 3.3708, 3.3602, 1.5276, 3.1504], device='cuda:6'), covar=tensor([0.1687, 0.1182, 0.2753, 0.1998, 0.2440, 0.1927, 0.6079, 0.2454], device='cuda:6'), in_proj_covar=tensor([0.0262, 0.0232, 0.0281, 0.0326, 0.0322, 0.0271, 0.0287, 0.0287], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-26 11:27:11,310 INFO [finetune.py:976] (6/7) Epoch 1, batch 5100, loss[loss=0.2751, simple_loss=0.3007, pruned_loss=0.1247, over 4725.00 frames. ], tot_loss[loss=0.3045, simple_loss=0.3391, pruned_loss=0.135, over 952681.22 frames. ], batch size: 23, lr: 4.00e-03, grad_scale: 16.0 2023-04-26 11:27:20,368 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6062, 1.7797, 1.4518, 1.8055, 1.4973, 1.3696, 1.5179, 1.2057], device='cuda:6'), covar=tensor([0.1853, 0.1627, 0.1510, 0.1581, 0.3386, 0.1696, 0.1971, 0.2981], device='cuda:6'), in_proj_covar=tensor([0.0307, 0.0326, 0.0241, 0.0304, 0.0309, 0.0279, 0.0275, 0.0295], device='cuda:6'), out_proj_covar=tensor([1.2588e-04, 1.3398e-04, 9.8737e-05, 1.2342e-04, 1.2816e-04, 1.1272e-04, 1.1419e-04, 1.2040e-04], device='cuda:6') 2023-04-26 11:27:40,806 INFO [optim.py:369] (6/7) 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 11:28:13,747 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.1406, 1.6626, 1.5755, 1.8448, 1.6987, 1.9258, 1.6146, 3.6485], device='cuda:6'), covar=tensor([0.0711, 0.0731, 0.0749, 0.1270, 0.0668, 0.0549, 0.0718, 0.0155], device='cuda:6'), in_proj_covar=tensor([0.0039, 0.0040, 0.0041, 0.0046, 0.0041, 0.0040, 0.0041, 0.0064], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0017], device='cuda:6') 2023-04-26 11:28:17,912 INFO [finetune.py:976] (6/7) Epoch 1, batch 5150, loss[loss=0.2762, simple_loss=0.3108, pruned_loss=0.1208, over 4687.00 frames. ], tot_loss[loss=0.3035, simple_loss=0.3381, pruned_loss=0.1345, over 952277.51 frames. ], batch size: 23, lr: 4.00e-03, grad_scale: 16.0 2023-04-26 11:28:49,486 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.3724, 1.0581, 0.9638, 1.1441, 1.6358, 1.3135, 1.1683, 0.9061], device='cuda:6'), covar=tensor([0.1593, 0.1891, 0.2653, 0.1953, 0.0808, 0.1704, 0.1744, 0.2403], device='cuda:6'), in_proj_covar=tensor([0.0333, 0.0346, 0.0349, 0.0315, 0.0357, 0.0381, 0.0329, 0.0364], device='cuda:6'), out_proj_covar=tensor([7.3040e-05, 7.4620e-05, 7.5497e-05, 6.6048e-05, 7.6260e-05, 8.3492e-05, 7.1830e-05, 7.8787e-05], device='cuda:6') 2023-04-26 11:29:07,360 INFO [finetune.py:976] (6/7) Epoch 1, batch 5200, loss[loss=0.3958, simple_loss=0.4184, pruned_loss=0.1866, over 4902.00 frames. ], tot_loss[loss=0.3078, simple_loss=0.3428, pruned_loss=0.1364, over 953383.28 frames. ], batch size: 36, lr: 4.00e-03, grad_scale: 16.0 2023-04-26 11:29:07,686 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.80 vs. limit=5.0 2023-04-26 11:29:21,207 INFO [optim.py:369] (6/7) 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 11:29:41,593 INFO [finetune.py:976] (6/7) Epoch 1, batch 5250, loss[loss=0.2911, simple_loss=0.3387, pruned_loss=0.1217, over 4924.00 frames. ], tot_loss[loss=0.3079, simple_loss=0.3439, pruned_loss=0.1359, over 952991.47 frames. ], batch size: 38, lr: 4.00e-03, grad_scale: 16.0 2023-04-26 11:29:57,006 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.47 vs. limit=2.0 2023-04-26 11:30:03,047 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-04-26 11:30:14,473 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([4.5138, 4.4582, 3.3001, 5.0667, 4.5134, 4.4203, 1.9658, 4.3281], device='cuda:6'), covar=tensor([0.1355, 0.0900, 0.2800, 0.0926, 0.2531, 0.1515, 0.5737, 0.1846], device='cuda:6'), in_proj_covar=tensor([0.0267, 0.0237, 0.0287, 0.0333, 0.0329, 0.0277, 0.0293, 0.0292], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-26 11:30:15,018 INFO [finetune.py:976] (6/7) Epoch 1, batch 5300, loss[loss=0.3022, simple_loss=0.3638, pruned_loss=0.1203, over 4924.00 frames. ], tot_loss[loss=0.311, simple_loss=0.3472, pruned_loss=0.1374, over 954086.87 frames. ], batch size: 42, lr: 4.00e-03, grad_scale: 16.0 2023-04-26 11:30:20,668 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-04-26 11:30:27,273 INFO [optim.py:369] (6/7) 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 11:30:46,756 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=6.48 vs. limit=5.0 2023-04-26 11:30:49,037 INFO [finetune.py:976] (6/7) Epoch 1, batch 5350, loss[loss=0.3381, simple_loss=0.3718, pruned_loss=0.1522, over 4800.00 frames. ], tot_loss[loss=0.3073, simple_loss=0.345, pruned_loss=0.1347, over 952975.33 frames. ], batch size: 41, lr: 4.00e-03, grad_scale: 16.0 2023-04-26 11:31:35,568 INFO [finetune.py:976] (6/7) Epoch 1, batch 5400, loss[loss=0.2546, simple_loss=0.2908, pruned_loss=0.1092, over 4769.00 frames. ], tot_loss[loss=0.304, simple_loss=0.3412, pruned_loss=0.1334, over 954112.69 frames. ], batch size: 28, lr: 4.00e-03, grad_scale: 16.0 2023-04-26 11:31:52,841 INFO [optim.py:369] (6/7) 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 11:32:15,538 INFO [finetune.py:976] (6/7) Epoch 1, batch 5450, loss[loss=0.2695, simple_loss=0.3127, pruned_loss=0.1132, over 4815.00 frames. ], tot_loss[loss=0.2995, simple_loss=0.3367, pruned_loss=0.1311, over 955281.61 frames. ], batch size: 39, lr: 4.00e-03, grad_scale: 16.0 2023-04-26 11:32:22,985 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.4340, 0.9984, 1.1147, 0.8734, 1.6271, 1.3651, 1.0778, 1.1528], device='cuda:6'), covar=tensor([0.1741, 0.1845, 0.2824, 0.2447, 0.0957, 0.1812, 0.2065, 0.2245], device='cuda:6'), in_proj_covar=tensor([0.0333, 0.0348, 0.0350, 0.0317, 0.0359, 0.0382, 0.0330, 0.0365], device='cuda:6'), out_proj_covar=tensor([7.3000e-05, 7.4922e-05, 7.5535e-05, 6.6518e-05, 7.6606e-05, 8.3652e-05, 7.1897e-05, 7.8957e-05], device='cuda:6') 2023-04-26 11:32:49,240 INFO [finetune.py:976] (6/7) Epoch 1, batch 5500, loss[loss=0.3301, simple_loss=0.3674, pruned_loss=0.1464, over 4826.00 frames. ], tot_loss[loss=0.2962, simple_loss=0.3334, pruned_loss=0.1295, over 956064.16 frames. ], batch size: 33, lr: 4.00e-03, grad_scale: 16.0 2023-04-26 11:32:59,730 INFO [optim.py:369] (6/7) 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] (6/7) Epoch 1, batch 5550, loss[loss=0.3067, simple_loss=0.3523, pruned_loss=0.1306, over 4874.00 frames. ], tot_loss[loss=0.297, simple_loss=0.3336, pruned_loss=0.1302, over 952925.44 frames. ], batch size: 34, lr: 4.00e-03, grad_scale: 16.0 2023-04-26 11:33:57,227 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.77 vs. limit=5.0 2023-04-26 11:34:30,897 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.2018, 1.7477, 1.5089, 1.8820, 1.8856, 2.0619, 1.5351, 4.2897], device='cuda:6'), covar=tensor([0.0754, 0.0758, 0.0860, 0.1328, 0.0666, 0.0634, 0.0801, 0.0128], device='cuda:6'), in_proj_covar=tensor([0.0040, 0.0040, 0.0041, 0.0046, 0.0042, 0.0041, 0.0041, 0.0065], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0017], device='cuda:6') 2023-04-26 11:34:53,185 INFO [finetune.py:976] (6/7) Epoch 1, batch 5600, loss[loss=0.3296, simple_loss=0.3626, pruned_loss=0.1483, over 4907.00 frames. ], tot_loss[loss=0.2992, simple_loss=0.3372, pruned_loss=0.1306, over 953286.86 frames. ], batch size: 36, lr: 4.00e-03, grad_scale: 16.0 2023-04-26 11:35:13,913 INFO [optim.py:369] (6/7) 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 11:35:44,907 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=5650.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 11:35:45,418 INFO [finetune.py:976] (6/7) Epoch 1, batch 5650, loss[loss=0.3322, simple_loss=0.3706, pruned_loss=0.1469, over 4799.00 frames. ], tot_loss[loss=0.3021, simple_loss=0.3408, pruned_loss=0.1317, over 952472.68 frames. ], batch size: 41, lr: 4.00e-03, grad_scale: 16.0 2023-04-26 11:36:15,235 INFO [finetune.py:976] (6/7) Epoch 1, batch 5700, loss[loss=0.2458, simple_loss=0.2796, pruned_loss=0.106, over 4369.00 frames. ], tot_loss[loss=0.2991, simple_loss=0.3361, pruned_loss=0.131, over 933120.24 frames. ], batch size: 18, lr: 4.00e-03, grad_scale: 16.0 2023-04-26 11:36:21,269 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=5711.0, num_to_drop=1, layers_to_drop={2} 2023-04-26 11:36:31,941 INFO [optim.py:369] (6/7) 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] (6/7) Epoch 2, batch 0, loss[loss=0.3179, simple_loss=0.3545, pruned_loss=0.1407, over 4895.00 frames. ], tot_loss[loss=0.3179, simple_loss=0.3545, pruned_loss=0.1407, over 4895.00 frames. ], batch size: 43, lr: 4.00e-03, grad_scale: 16.0 2023-04-26 11:37:01,400 INFO [finetune.py:1001] (6/7) Computing validation loss 2023-04-26 11:37:23,938 INFO [finetune.py:1010] (6/7) 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] (6/7) Maximum memory allocated so far is 6326MB 2023-04-26 11:37:46,593 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 2, batch 50, loss[loss=0.2937, simple_loss=0.3297, pruned_loss=0.1289, over 4853.00 frames. ], tot_loss[loss=0.3031, simple_loss=0.3399, pruned_loss=0.1331, over 217202.42 frames. ], batch size: 31, lr: 4.00e-03, grad_scale: 16.0 2023-04-26 11:38:14,013 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0919, 2.4950, 1.0811, 1.4005, 1.8208, 1.2311, 2.9697, 1.4552], device='cuda:6'), covar=tensor([0.0648, 0.0558, 0.0841, 0.1117, 0.0490, 0.0944, 0.0205, 0.0656], device='cuda:6'), in_proj_covar=tensor([0.0058, 0.0074, 0.0054, 0.0051, 0.0056, 0.0056, 0.0088, 0.0054], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0008, 0.0008, 0.0008, 0.0008, 0.0012, 0.0008], device='cuda:6') 2023-04-26 11:38:22,408 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.79 vs. limit=2.0 2023-04-26 11:38:27,692 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=5816.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:38:29,285 INFO [optim.py:369] (6/7) 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] (6/7) Epoch 2, batch 100, loss[loss=0.2804, simple_loss=0.3261, pruned_loss=0.1173, over 4814.00 frames. ], tot_loss[loss=0.2929, simple_loss=0.3305, pruned_loss=0.1276, over 381441.01 frames. ], batch size: 38, lr: 4.00e-03, grad_scale: 16.0 2023-04-26 11:38:55,149 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=5858.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:38:56,702 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-04-26 11:39:09,187 INFO [finetune.py:976] (6/7) Epoch 2, batch 150, loss[loss=0.2463, simple_loss=0.2899, pruned_loss=0.1014, over 4910.00 frames. ], tot_loss[loss=0.2859, simple_loss=0.3247, pruned_loss=0.1235, over 510680.78 frames. ], batch size: 43, lr: 4.00e-03, grad_scale: 16.0 2023-04-26 11:39:34,932 INFO [optim.py:369] (6/7) 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 11:39:35,693 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=5919.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:39:48,163 INFO [finetune.py:976] (6/7) Epoch 2, batch 200, loss[loss=0.2761, simple_loss=0.3236, pruned_loss=0.1143, over 4928.00 frames. ], tot_loss[loss=0.288, simple_loss=0.3257, pruned_loss=0.1252, over 611011.71 frames. ], batch size: 33, lr: 4.00e-03, grad_scale: 16.0 2023-04-26 11:39:57,018 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.67 vs. limit=2.0 2023-04-26 11:40:52,012 INFO [finetune.py:976] (6/7) Epoch 2, batch 250, loss[loss=0.262, simple_loss=0.3053, pruned_loss=0.1094, over 4913.00 frames. ], tot_loss[loss=0.293, simple_loss=0.3307, pruned_loss=0.1276, over 688074.46 frames. ], batch size: 32, lr: 4.00e-03, grad_scale: 16.0 2023-04-26 11:41:06,872 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=5990.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:41:18,235 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6006.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 11:41:25,491 INFO [optim.py:369] (6/7) 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 11:41:33,102 INFO [finetune.py:976] (6/7) Epoch 2, batch 300, loss[loss=0.2865, simple_loss=0.3306, pruned_loss=0.1212, over 4868.00 frames. ], tot_loss[loss=0.2962, simple_loss=0.3349, pruned_loss=0.1287, over 748398.14 frames. ], batch size: 31, lr: 4.00e-03, grad_scale: 16.0 2023-04-26 11:41:49,301 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=6043.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:41:54,630 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 2, batch 350, loss[loss=0.3039, simple_loss=0.3492, pruned_loss=0.1294, over 4865.00 frames. ], tot_loss[loss=0.2959, simple_loss=0.3355, pruned_loss=0.1282, over 795005.51 frames. ], batch size: 34, lr: 4.00e-03, grad_scale: 16.0 2023-04-26 11:42:47,469 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=6104.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:42:57,934 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6111.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:43:02,077 INFO [optim.py:369] (6/7) 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 11:43:14,254 INFO [finetune.py:976] (6/7) Epoch 2, batch 400, loss[loss=0.3016, simple_loss=0.3349, pruned_loss=0.1341, over 4865.00 frames. ], tot_loss[loss=0.2955, simple_loss=0.3353, pruned_loss=0.1279, over 830877.53 frames. ], batch size: 34, lr: 4.00e-03, grad_scale: 16.0 2023-04-26 11:43:20,180 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=6135.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:43:29,736 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.0467, 0.9760, 1.2198, 1.1559, 1.0299, 0.8397, 0.9250, 0.5043], device='cuda:6'), covar=tensor([0.0835, 0.0857, 0.0641, 0.0820, 0.1104, 0.1602, 0.0739, 0.1415], device='cuda:6'), in_proj_covar=tensor([0.0072, 0.0079, 0.0075, 0.0075, 0.0089, 0.0095, 0.0090, 0.0081], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003], device='cuda:6') 2023-04-26 11:43:43,641 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 2, batch 450, loss[loss=0.2381, simple_loss=0.2857, pruned_loss=0.09527, over 4745.00 frames. ], tot_loss[loss=0.2931, simple_loss=0.3334, pruned_loss=0.1264, over 857753.78 frames. ], batch size: 54, lr: 4.00e-03, grad_scale: 32.0 2023-04-26 11:44:00,963 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=6196.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:44:06,755 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=6203.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:44:11,666 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=6211.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:44:13,452 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([3.9200, 3.8553, 2.9436, 4.5032, 3.9266, 3.9217, 1.8932, 3.8119], device='cuda:6'), covar=tensor([0.1475, 0.1024, 0.2903, 0.1380, 0.3104, 0.1719, 0.5143, 0.2195], device='cuda:6'), in_proj_covar=tensor([0.0264, 0.0233, 0.0284, 0.0331, 0.0326, 0.0273, 0.0289, 0.0289], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-26 11:44:13,458 INFO [zipformer.py:1188] (6/7) 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] (6/7) 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 11:44:21,877 INFO [finetune.py:976] (6/7) Epoch 2, batch 500, loss[loss=0.2437, simple_loss=0.2881, pruned_loss=0.09971, over 4798.00 frames. ], tot_loss[loss=0.2891, simple_loss=0.3292, pruned_loss=0.1245, over 879994.18 frames. ], batch size: 51, lr: 4.00e-03, grad_scale: 32.0 2023-04-26 11:44:23,810 INFO [zipformer.py:1188] (6/7) 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] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=6264.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:44:48,025 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=6266.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:44:51,726 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=6272.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:44:55,242 INFO [finetune.py:976] (6/7) Epoch 2, batch 550, loss[loss=0.3125, simple_loss=0.3346, pruned_loss=0.1452, over 4782.00 frames. ], tot_loss[loss=0.2853, simple_loss=0.3251, pruned_loss=0.1228, over 897359.93 frames. ], batch size: 26, lr: 4.00e-03, grad_scale: 32.0 2023-04-26 11:45:14,999 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=6306.0, num_to_drop=1, layers_to_drop={2} 2023-04-26 11:45:28,233 INFO [optim.py:369] (6/7) 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 11:45:37,845 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 2, batch 600, loss[loss=0.3125, simple_loss=0.3477, pruned_loss=0.1387, over 4813.00 frames. ], tot_loss[loss=0.2847, simple_loss=0.3242, pruned_loss=0.1226, over 910174.13 frames. ], batch size: 51, lr: 4.00e-03, grad_scale: 32.0 2023-04-26 11:46:00,722 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6346.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:46:13,051 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=6354.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 11:46:39,284 INFO [finetune.py:976] (6/7) Epoch 2, batch 650, loss[loss=0.322, simple_loss=0.342, pruned_loss=0.151, over 4789.00 frames. ], tot_loss[loss=0.2911, simple_loss=0.3305, pruned_loss=0.1258, over 920515.31 frames. ], batch size: 26, lr: 4.00e-03, grad_scale: 32.0 2023-04-26 11:46:53,629 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6399.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:47:02,605 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=6411.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:47:07,254 INFO [optim.py:369] (6/7) 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] (6/7) Epoch 2, batch 700, loss[loss=0.3072, simple_loss=0.3347, pruned_loss=0.1399, over 4863.00 frames. ], tot_loss[loss=0.2918, simple_loss=0.3313, pruned_loss=0.1261, over 927107.47 frames. ], batch size: 31, lr: 4.00e-03, grad_scale: 32.0 2023-04-26 11:47:15,077 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-04-26 11:47:46,401 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=6459.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:47:48,748 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=6462.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:47:51,721 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=6467.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:47:58,778 INFO [finetune.py:976] (6/7) Epoch 2, batch 750, loss[loss=0.2481, simple_loss=0.2936, pruned_loss=0.1013, over 4783.00 frames. ], tot_loss[loss=0.2935, simple_loss=0.3333, pruned_loss=0.1268, over 934630.08 frames. ], batch size: 25, lr: 4.00e-03, grad_scale: 32.0 2023-04-26 11:48:18,140 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6491.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:48:30,371 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.72 vs. limit=2.0 2023-04-26 11:48:51,401 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=6514.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:48:53,699 INFO [optim.py:369] (6/7) 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 11:49:03,669 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=6523.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 11:49:05,428 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 2, batch 800, loss[loss=0.2566, simple_loss=0.2945, pruned_loss=0.1094, over 4745.00 frames. ], tot_loss[loss=0.2919, simple_loss=0.3323, pruned_loss=0.1257, over 939604.25 frames. ], batch size: 59, lr: 4.00e-03, grad_scale: 32.0 2023-04-26 11:49:06,723 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=6528.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:49:27,195 INFO [zipformer.py:1188] (6/7) 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] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=6562.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:49:33,152 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6567.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:49:40,231 INFO [finetune.py:976] (6/7) Epoch 2, batch 850, loss[loss=0.2297, simple_loss=0.2855, pruned_loss=0.08695, over 4912.00 frames. ], tot_loss[loss=0.2899, simple_loss=0.3301, pruned_loss=0.1249, over 941494.56 frames. ], batch size: 46, lr: 4.00e-03, grad_scale: 32.0 2023-04-26 11:50:18,616 INFO [optim.py:369] (6/7) 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 11:50:21,173 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6622.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:50:25,265 INFO [finetune.py:976] (6/7) Epoch 2, batch 900, loss[loss=0.2616, simple_loss=0.3094, pruned_loss=0.1069, over 4768.00 frames. ], tot_loss[loss=0.2846, simple_loss=0.3255, pruned_loss=0.1218, over 942234.21 frames. ], batch size: 28, lr: 4.00e-03, grad_scale: 32.0 2023-04-26 11:50:25,977 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=6629.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:50:36,411 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=6646.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:50:42,808 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-04-26 11:50:58,409 INFO [finetune.py:976] (6/7) Epoch 2, batch 950, loss[loss=0.3682, simple_loss=0.3933, pruned_loss=0.1716, over 4864.00 frames. ], tot_loss[loss=0.2835, simple_loss=0.3243, pruned_loss=0.1214, over 945468.76 frames. ], batch size: 44, lr: 4.00e-03, grad_scale: 32.0 2023-04-26 11:51:06,463 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=6690.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:51:08,887 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=6694.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:51:17,798 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=6699.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:51:41,858 INFO [optim.py:369] (6/7) 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 11:51:48,518 INFO [finetune.py:976] (6/7) Epoch 2, batch 1000, loss[loss=0.314, simple_loss=0.3737, pruned_loss=0.1271, over 4857.00 frames. ], tot_loss[loss=0.2852, simple_loss=0.3262, pruned_loss=0.1221, over 946142.39 frames. ], batch size: 44, lr: 4.00e-03, grad_scale: 32.0 2023-04-26 11:52:11,819 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=6747.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:52:55,094 INFO [finetune.py:976] (6/7) Epoch 2, batch 1050, loss[loss=0.3075, simple_loss=0.3484, pruned_loss=0.1333, over 4331.00 frames. ], tot_loss[loss=0.2872, simple_loss=0.3292, pruned_loss=0.1226, over 949634.56 frames. ], batch size: 65, lr: 4.00e-03, grad_scale: 32.0 2023-04-26 11:53:03,724 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=6791.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:53:04,542 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 2023-04-26 11:53:32,766 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9078, 2.2653, 1.8482, 2.2198, 1.6238, 1.8277, 2.0851, 1.4264], device='cuda:6'), covar=tensor([0.2038, 0.1480, 0.1343, 0.1533, 0.3300, 0.1564, 0.1972, 0.2781], device='cuda:6'), in_proj_covar=tensor([0.0313, 0.0333, 0.0245, 0.0310, 0.0312, 0.0283, 0.0279, 0.0300], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-26 11:53:33,226 INFO [optim.py:369] (6/7) 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 11:53:33,314 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6818.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 11:53:42,754 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6823.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:53:44,648 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=6826.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:53:51,476 INFO [finetune.py:976] (6/7) Epoch 2, batch 1100, loss[loss=0.2894, simple_loss=0.3439, pruned_loss=0.1174, over 4924.00 frames. ], tot_loss[loss=0.2873, simple_loss=0.3298, pruned_loss=0.1224, over 951592.76 frames. ], batch size: 33, lr: 4.00e-03, grad_scale: 32.0 2023-04-26 11:54:04,441 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=6839.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:54:16,595 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=6859.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:54:22,454 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=6867.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:54:28,122 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=6874.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:54:30,497 INFO [finetune.py:976] (6/7) Epoch 2, batch 1150, loss[loss=0.2634, simple_loss=0.3278, pruned_loss=0.0995, over 4863.00 frames. ], tot_loss[loss=0.2873, simple_loss=0.3305, pruned_loss=0.122, over 950778.71 frames. ], batch size: 34, lr: 4.00e-03, grad_scale: 32.0 2023-04-26 11:54:30,631 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=6878.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 11:54:45,852 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.95 vs. limit=5.0 2023-04-26 11:54:49,253 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=6907.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:54:49,927 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8479, 2.2424, 1.7497, 2.1475, 1.7252, 1.7890, 1.9410, 1.4321], device='cuda:6'), covar=tensor([0.2401, 0.1592, 0.1518, 0.1690, 0.3375, 0.1804, 0.2015, 0.3544], device='cuda:6'), in_proj_covar=tensor([0.0312, 0.0333, 0.0244, 0.0309, 0.0312, 0.0282, 0.0279, 0.0299], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-26 11:54:50,563 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=6909.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:54:54,161 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=6915.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:54:56,915 INFO [optim.py:369] (6/7) 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 11:54:57,603 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([3.3450, 3.2768, 2.5747, 3.7738, 3.2471, 3.3031, 1.4941, 3.1789], device='cuda:6'), covar=tensor([0.1680, 0.1329, 0.3511, 0.2431, 0.2674, 0.2013, 0.5301, 0.2507], device='cuda:6'), in_proj_covar=tensor([0.0260, 0.0231, 0.0280, 0.0329, 0.0323, 0.0272, 0.0286, 0.0287], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-26 11:54:59,937 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=6922.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:55:01,006 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([3.5026, 3.4473, 2.7491, 3.9335, 3.4622, 3.4706, 1.5429, 3.3519], device='cuda:6'), covar=tensor([0.1976, 0.1318, 0.3019, 0.2276, 0.2910, 0.2057, 0.5679, 0.2534], device='cuda:6'), in_proj_covar=tensor([0.0260, 0.0230, 0.0280, 0.0329, 0.0323, 0.0272, 0.0286, 0.0287], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-26 11:55:02,778 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6176, 1.6116, 0.8962, 1.2422, 1.7119, 1.5637, 1.4018, 1.4565], device='cuda:6'), covar=tensor([0.0620, 0.0501, 0.0510, 0.0694, 0.0351, 0.0613, 0.0621, 0.0785], device='cuda:6'), in_proj_covar=tensor([0.0032, 0.0027, 0.0025, 0.0031, 0.0022, 0.0031, 0.0030, 0.0033], device='cuda:6'), out_proj_covar=tensor([0.0049, 0.0044, 0.0039, 0.0049, 0.0037, 0.0048, 0.0047, 0.0051], device='cuda:6') 2023-04-26 11:55:04,519 INFO [finetune.py:976] (6/7) Epoch 2, batch 1200, loss[loss=0.3021, simple_loss=0.3446, pruned_loss=0.1298, over 4850.00 frames. ], tot_loss[loss=0.2845, simple_loss=0.328, pruned_loss=0.1205, over 950176.88 frames. ], batch size: 44, lr: 4.00e-03, grad_scale: 32.0 2023-04-26 11:55:08,776 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8955, 2.2567, 1.7838, 2.1971, 1.7049, 1.8202, 2.0280, 1.4963], device='cuda:6'), covar=tensor([0.2221, 0.1615, 0.1351, 0.1710, 0.2940, 0.1601, 0.2116, 0.3040], device='cuda:6'), in_proj_covar=tensor([0.0313, 0.0333, 0.0244, 0.0309, 0.0312, 0.0283, 0.0279, 0.0300], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-26 11:55:08,811 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.0629, 0.9700, 1.1214, 1.4289, 1.3201, 1.0163, 1.0523, 1.0136], device='cuda:6'), covar=tensor([ 6.7695, 10.3465, 11.5484, 10.6892, 6.8478, 11.8516, 12.5105, 7.8272], device='cuda:6'), in_proj_covar=tensor([0.0442, 0.0507, 0.0589, 0.0573, 0.0472, 0.0529, 0.0535, 0.0544], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-26 11:55:11,755 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=6939.0, num_to_drop=1, layers_to_drop={2} 2023-04-26 11:55:32,156 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=6970.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:55:32,207 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 2, batch 1250, loss[loss=0.2015, simple_loss=0.2605, pruned_loss=0.07128, over 4826.00 frames. ], tot_loss[loss=0.2813, simple_loss=0.3244, pruned_loss=0.1191, over 950369.26 frames. ], batch size: 30, lr: 4.00e-03, grad_scale: 32.0 2023-04-26 11:55:43,401 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6985.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:56:04,201 INFO [optim.py:369] (6/7) 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 11:56:11,777 INFO [finetune.py:976] (6/7) Epoch 2, batch 1300, loss[loss=0.2448, simple_loss=0.2741, pruned_loss=0.1077, over 3997.00 frames. ], tot_loss[loss=0.2773, simple_loss=0.3201, pruned_loss=0.1173, over 950902.13 frames. ], batch size: 17, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 11:56:21,351 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.62 vs. limit=5.0 2023-04-26 11:56:25,534 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=7048.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 11:56:55,868 INFO [finetune.py:976] (6/7) Epoch 2, batch 1350, loss[loss=0.2822, simple_loss=0.344, pruned_loss=0.1102, over 4909.00 frames. ], tot_loss[loss=0.2768, simple_loss=0.3194, pruned_loss=0.1171, over 952412.90 frames. ], batch size: 43, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 11:57:34,482 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=7109.0, num_to_drop=1, layers_to_drop={3} 2023-04-26 11:57:39,839 INFO [optim.py:369] (6/7) 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 11:57:45,821 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=7118.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 11:57:48,841 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=7123.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:57:56,137 INFO [finetune.py:976] (6/7) Epoch 2, batch 1400, loss[loss=0.3859, simple_loss=0.4077, pruned_loss=0.182, over 4758.00 frames. ], tot_loss[loss=0.2826, simple_loss=0.325, pruned_loss=0.1201, over 951892.40 frames. ], batch size: 28, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 11:58:36,328 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=7163.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:58:38,104 INFO [zipformer.py:1188] (6/7) 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] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=7171.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:58:49,113 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7014, 1.3232, 4.5956, 4.2599, 4.0216, 4.2788, 4.2477, 4.0229], device='cuda:6'), covar=tensor([0.6659, 0.6253, 0.0936, 0.1632, 0.1059, 0.1494, 0.1288, 0.1436], device='cuda:6'), in_proj_covar=tensor([0.0341, 0.0319, 0.0458, 0.0462, 0.0383, 0.0441, 0.0349, 0.0406], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0002], device='cuda:6') 2023-04-26 11:58:50,023 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.99 vs. limit=5.0 2023-04-26 11:58:51,413 INFO [finetune.py:976] (6/7) Epoch 2, batch 1450, loss[loss=0.2468, simple_loss=0.3102, pruned_loss=0.09173, over 4764.00 frames. ], tot_loss[loss=0.2834, simple_loss=0.326, pruned_loss=0.1204, over 950289.74 frames. ], batch size: 54, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 11:59:19,363 INFO [optim.py:369] (6/7) 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 11:59:23,174 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=7224.0, num_to_drop=1, layers_to_drop={2} 2023-04-26 11:59:23,796 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.0948, 0.7776, 0.8977, 0.7046, 1.2386, 0.9972, 0.8582, 1.0057], device='cuda:6'), covar=tensor([0.1559, 0.1782, 0.2244, 0.2115, 0.1239, 0.1477, 0.1942, 0.2458], device='cuda:6'), in_proj_covar=tensor([0.0329, 0.0343, 0.0347, 0.0316, 0.0355, 0.0377, 0.0328, 0.0360], device='cuda:6'), out_proj_covar=tensor([7.1993e-05, 7.3725e-05, 7.4895e-05, 6.6197e-05, 7.5562e-05, 8.2583e-05, 7.1418e-05, 7.7717e-05], device='cuda:6') 2023-04-26 11:59:25,498 INFO [finetune.py:976] (6/7) Epoch 2, batch 1500, loss[loss=0.2708, simple_loss=0.3293, pruned_loss=0.1061, over 4890.00 frames. ], tot_loss[loss=0.2831, simple_loss=0.3262, pruned_loss=0.1199, over 949230.10 frames. ], batch size: 43, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 11:59:30,167 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=7234.0, num_to_drop=1, layers_to_drop={2} 2023-04-26 11:59:39,477 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([3.7773, 3.6444, 2.8493, 4.3024, 3.7615, 3.7585, 1.7233, 3.5603], device='cuda:6'), covar=tensor([0.1570, 0.1077, 0.2988, 0.1633, 0.3017, 0.1881, 0.5586, 0.2301], device='cuda:6'), in_proj_covar=tensor([0.0259, 0.0231, 0.0278, 0.0328, 0.0322, 0.0272, 0.0285, 0.0287], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-26 11:59:51,525 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=7265.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 11:59:57,726 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 2, batch 1550, loss[loss=0.2273, simple_loss=0.2861, pruned_loss=0.08431, over 4824.00 frames. ], tot_loss[loss=0.2833, simple_loss=0.3265, pruned_loss=0.12, over 949731.42 frames. ], batch size: 30, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 12:00:04,483 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=7285.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:00:27,651 INFO [optim.py:369] (6/7) 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 12:00:33,049 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.77 vs. limit=5.0 2023-04-26 12:00:33,829 INFO [finetune.py:976] (6/7) Epoch 2, batch 1600, loss[loss=0.2351, simple_loss=0.2838, pruned_loss=0.09318, over 4823.00 frames. ], tot_loss[loss=0.2796, simple_loss=0.3232, pruned_loss=0.118, over 949817.13 frames. ], batch size: 38, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 12:00:36,971 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=7333.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:00:38,912 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=7336.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:01:00,096 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6506, 1.1648, 1.3433, 1.3876, 1.2228, 1.0858, 0.5131, 1.0536], device='cuda:6'), covar=tensor([0.6431, 0.7567, 0.3538, 0.7247, 0.7113, 0.5835, 1.1040, 0.7006], device='cuda:6'), in_proj_covar=tensor([0.0266, 0.0281, 0.0223, 0.0348, 0.0238, 0.0236, 0.0275, 0.0223], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-26 12:01:07,695 INFO [finetune.py:976] (6/7) Epoch 2, batch 1650, loss[loss=0.2887, simple_loss=0.3262, pruned_loss=0.1256, over 4905.00 frames. ], tot_loss[loss=0.2736, simple_loss=0.3174, pruned_loss=0.1148, over 949720.82 frames. ], batch size: 36, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 12:01:26,435 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=7404.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 12:01:28,246 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=7407.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:01:31,930 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.1164, 2.0719, 2.2884, 2.4446, 2.4134, 1.8779, 1.5259, 2.1273], device='cuda:6'), covar=tensor([0.1182, 0.0981, 0.0551, 0.0693, 0.0708, 0.1152, 0.1252, 0.0703], device='cuda:6'), in_proj_covar=tensor([0.0208, 0.0207, 0.0188, 0.0181, 0.0178, 0.0196, 0.0173, 0.0192], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-26 12:01:34,835 INFO [optim.py:369] (6/7) 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 12:01:41,462 INFO [finetune.py:976] (6/7) Epoch 2, batch 1700, loss[loss=0.2317, simple_loss=0.2802, pruned_loss=0.09159, over 4756.00 frames. ], tot_loss[loss=0.2709, simple_loss=0.3149, pruned_loss=0.1135, over 951767.60 frames. ], batch size: 26, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 12:01:56,547 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.1602, 2.8296, 1.0382, 1.3118, 1.9792, 1.2230, 3.6215, 1.8323], device='cuda:6'), covar=tensor([0.0711, 0.0664, 0.0919, 0.1255, 0.0589, 0.1044, 0.0216, 0.0654], device='cuda:6'), in_proj_covar=tensor([0.0057, 0.0073, 0.0054, 0.0051, 0.0056, 0.0056, 0.0087, 0.0054], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0008, 0.0008, 0.0008, 0.0008, 0.0011, 0.0008], device='cuda:6') 2023-04-26 12:02:09,806 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 2, batch 1750, loss[loss=0.253, simple_loss=0.3159, pruned_loss=0.09504, over 4811.00 frames. ], tot_loss[loss=0.273, simple_loss=0.3169, pruned_loss=0.1145, over 951865.17 frames. ], batch size: 51, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 12:02:32,384 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.3889, 0.9943, 1.1238, 0.9512, 1.5454, 1.2713, 0.9922, 1.1258], device='cuda:6'), covar=tensor([0.2060, 0.2130, 0.2546, 0.2383, 0.1061, 0.2110, 0.2526, 0.2338], device='cuda:6'), in_proj_covar=tensor([0.0327, 0.0340, 0.0345, 0.0316, 0.0352, 0.0374, 0.0326, 0.0357], device='cuda:6'), out_proj_covar=tensor([7.1638e-05, 7.3253e-05, 7.4551e-05, 6.6342e-05, 7.4918e-05, 8.1901e-05, 7.0910e-05, 7.6998e-05], device='cuda:6') 2023-04-26 12:03:11,381 INFO [optim.py:369] (6/7) 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 12:03:12,102 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=7519.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 12:03:23,647 INFO [finetune.py:976] (6/7) Epoch 2, batch 1800, loss[loss=0.2582, simple_loss=0.3089, pruned_loss=0.1038, over 4895.00 frames. ], tot_loss[loss=0.2764, simple_loss=0.3215, pruned_loss=0.1157, over 954517.79 frames. ], batch size: 43, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 12:03:33,047 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=7534.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 12:04:17,657 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=7565.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:04:31,082 INFO [finetune.py:976] (6/7) Epoch 2, batch 1850, loss[loss=0.2729, simple_loss=0.3173, pruned_loss=0.1143, over 4763.00 frames. ], tot_loss[loss=0.2791, simple_loss=0.3237, pruned_loss=0.1173, over 954824.44 frames. ], batch size: 28, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 12:04:40,137 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=7582.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 12:05:07,102 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=7613.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:05:10,538 INFO [optim.py:369] (6/7) 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 12:05:14,373 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=7624.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:05:16,721 INFO [finetune.py:976] (6/7) Epoch 2, batch 1900, loss[loss=0.2312, simple_loss=0.2782, pruned_loss=0.09205, over 4720.00 frames. ], tot_loss[loss=0.2788, simple_loss=0.3237, pruned_loss=0.1169, over 955410.08 frames. ], batch size: 23, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 12:05:18,632 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=7631.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:05:48,978 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5756, 1.3242, 1.2562, 1.2312, 1.8457, 1.4994, 1.1244, 1.2655], device='cuda:6'), covar=tensor([0.1546, 0.1603, 0.2192, 0.1814, 0.0801, 0.1655, 0.2176, 0.1961], device='cuda:6'), in_proj_covar=tensor([0.0326, 0.0340, 0.0345, 0.0316, 0.0352, 0.0372, 0.0324, 0.0355], device='cuda:6'), out_proj_covar=tensor([7.1275e-05, 7.3118e-05, 7.4615e-05, 6.6168e-05, 7.4927e-05, 8.1611e-05, 7.0530e-05, 7.6690e-05], device='cuda:6') 2023-04-26 12:05:50,046 INFO [finetune.py:976] (6/7) Epoch 2, batch 1950, loss[loss=0.264, simple_loss=0.3048, pruned_loss=0.1116, over 4783.00 frames. ], tot_loss[loss=0.2765, simple_loss=0.322, pruned_loss=0.1156, over 956904.52 frames. ], batch size: 51, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 12:05:54,928 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=7685.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:06:06,996 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=7704.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 12:06:16,404 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.3347, 1.3188, 3.8311, 3.5426, 3.4449, 3.6346, 3.6326, 3.4145], device='cuda:6'), covar=tensor([0.6898, 0.5612, 0.1134, 0.1860, 0.1159, 0.1680, 0.1946, 0.1565], device='cuda:6'), in_proj_covar=tensor([0.0335, 0.0314, 0.0449, 0.0458, 0.0378, 0.0433, 0.0344, 0.0399], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0002], device='cuda:6') 2023-04-26 12:06:17,408 INFO [optim.py:369] (6/7) 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 12:06:23,980 INFO [finetune.py:976] (6/7) Epoch 2, batch 2000, loss[loss=0.2838, simple_loss=0.3301, pruned_loss=0.1187, over 4822.00 frames. ], tot_loss[loss=0.272, simple_loss=0.3173, pruned_loss=0.1133, over 957052.77 frames. ], batch size: 41, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 12:06:26,544 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7912, 1.1842, 1.6216, 2.0382, 1.4390, 1.2175, 0.8277, 1.3916], device='cuda:6'), covar=tensor([0.6076, 0.7468, 0.3328, 0.6299, 0.7352, 0.5565, 1.1175, 0.6950], device='cuda:6'), in_proj_covar=tensor([0.0267, 0.0281, 0.0224, 0.0348, 0.0239, 0.0236, 0.0275, 0.0223], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-26 12:06:33,709 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=7743.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:06:39,111 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=7752.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 12:06:42,700 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8151, 1.2071, 1.5600, 1.9089, 1.4496, 1.1573, 0.8248, 1.3204], device='cuda:6'), covar=tensor([0.6646, 0.8138, 0.3691, 0.7960, 0.8857, 0.6097, 1.2346, 0.8391], device='cuda:6'), in_proj_covar=tensor([0.0268, 0.0282, 0.0224, 0.0349, 0.0240, 0.0237, 0.0275, 0.0224], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-26 12:06:47,289 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 2, batch 2050, loss[loss=0.2342, simple_loss=0.2783, pruned_loss=0.09507, over 4702.00 frames. ], tot_loss[loss=0.2689, simple_loss=0.3138, pruned_loss=0.112, over 956980.39 frames. ], batch size: 23, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 12:07:13,098 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=7802.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:07:14,355 INFO [zipformer.py:1188] (6/7) 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] (6/7) 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 12:07:25,338 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 2, batch 2100, loss[loss=0.3464, simple_loss=0.3729, pruned_loss=0.1599, over 4829.00 frames. ], tot_loss[loss=0.2679, simple_loss=0.3131, pruned_loss=0.1113, over 958445.57 frames. ], batch size: 39, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 12:07:46,726 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.72 vs. limit=2.0 2023-04-26 12:07:54,057 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=7863.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:07:55,224 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.1186, 1.4000, 5.4370, 5.0840, 4.7571, 5.0897, 4.7661, 4.7720], device='cuda:6'), covar=tensor([0.6502, 0.6270, 0.0936, 0.1813, 0.0995, 0.1939, 0.1021, 0.1626], device='cuda:6'), in_proj_covar=tensor([0.0338, 0.0318, 0.0453, 0.0460, 0.0380, 0.0437, 0.0345, 0.0403], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0002], device='cuda:6') 2023-04-26 12:07:56,931 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 2, batch 2150, loss[loss=0.2758, simple_loss=0.3241, pruned_loss=0.1137, over 4822.00 frames. ], tot_loss[loss=0.2715, simple_loss=0.3173, pruned_loss=0.1129, over 959938.40 frames. ], batch size: 30, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 12:09:02,677 INFO [optim.py:369] (6/7) 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 12:09:20,979 INFO [finetune.py:976] (6/7) Epoch 2, batch 2200, loss[loss=0.31, simple_loss=0.329, pruned_loss=0.1456, over 4146.00 frames. ], tot_loss[loss=0.2738, simple_loss=0.32, pruned_loss=0.1138, over 956938.21 frames. ], batch size: 65, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 12:09:22,964 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=7931.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:10:21,351 INFO [finetune.py:976] (6/7) Epoch 2, batch 2250, loss[loss=0.2436, simple_loss=0.3009, pruned_loss=0.09311, over 4765.00 frames. ], tot_loss[loss=0.2743, simple_loss=0.3212, pruned_loss=0.1137, over 955782.72 frames. ], batch size: 26, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 12:10:22,509 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=7979.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:10:23,128 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=7980.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:10:47,703 INFO [optim.py:369] (6/7) 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 12:10:55,298 INFO [finetune.py:976] (6/7) Epoch 2, batch 2300, loss[loss=0.2592, simple_loss=0.285, pruned_loss=0.1167, over 4084.00 frames. ], tot_loss[loss=0.2738, simple_loss=0.3209, pruned_loss=0.1134, over 955568.17 frames. ], batch size: 17, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 12:11:01,709 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([0.7945, 1.7820, 1.7992, 1.4947, 2.0305, 1.5122, 2.5810, 1.4917], device='cuda:6'), covar=tensor([0.4540, 0.1829, 0.5159, 0.3283, 0.1975, 0.2664, 0.1351, 0.4548], device='cuda:6'), in_proj_covar=tensor([0.0344, 0.0352, 0.0436, 0.0369, 0.0404, 0.0375, 0.0400, 0.0410], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-26 12:11:18,731 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=8063.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:11:28,369 INFO [finetune.py:976] (6/7) Epoch 2, batch 2350, loss[loss=0.2543, simple_loss=0.304, pruned_loss=0.1023, over 4889.00 frames. ], tot_loss[loss=0.271, simple_loss=0.3179, pruned_loss=0.1121, over 955092.43 frames. ], batch size: 32, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 12:11:40,367 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.87 vs. limit=5.0 2023-04-26 12:11:42,662 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=8099.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:11:49,273 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8349, 1.7296, 1.9000, 2.1230, 2.1761, 1.6841, 1.3276, 1.8555], device='cuda:6'), covar=tensor([0.1213, 0.1269, 0.0815, 0.0841, 0.0672, 0.1324, 0.1338, 0.0859], device='cuda:6'), in_proj_covar=tensor([0.0213, 0.0211, 0.0191, 0.0184, 0.0181, 0.0200, 0.0178, 0.0194], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-26 12:11:50,391 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=8111.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:11:54,657 INFO [optim.py:369] (6/7) 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 12:12:01,656 INFO [finetune.py:976] (6/7) Epoch 2, batch 2400, loss[loss=0.2431, simple_loss=0.3027, pruned_loss=0.09172, over 4935.00 frames. ], tot_loss[loss=0.2665, simple_loss=0.3138, pruned_loss=0.1096, over 957059.64 frames. ], batch size: 33, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 12:12:17,601 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0767, 2.4751, 1.0703, 1.3023, 1.8932, 1.2008, 2.9592, 1.5327], device='cuda:6'), covar=tensor([0.0717, 0.0663, 0.0783, 0.1230, 0.0506, 0.1006, 0.0247, 0.0690], device='cuda:6'), in_proj_covar=tensor([0.0058, 0.0074, 0.0055, 0.0051, 0.0056, 0.0057, 0.0088, 0.0055], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0008, 0.0008, 0.0008, 0.0008, 0.0011, 0.0008], device='cuda:6') 2023-04-26 12:12:21,842 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=8158.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:12:34,355 INFO [finetune.py:976] (6/7) Epoch 2, batch 2450, loss[loss=0.2527, simple_loss=0.3096, pruned_loss=0.09791, over 4852.00 frames. ], tot_loss[loss=0.2641, simple_loss=0.3106, pruned_loss=0.1088, over 956708.59 frames. ], batch size: 49, lr: 3.99e-03, grad_scale: 64.0 2023-04-26 12:13:01,849 INFO [optim.py:369] (6/7) 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] (6/7) Epoch 2, batch 2500, loss[loss=0.272, simple_loss=0.3285, pruned_loss=0.1078, over 4791.00 frames. ], tot_loss[loss=0.2671, simple_loss=0.313, pruned_loss=0.1106, over 955509.77 frames. ], batch size: 51, lr: 3.99e-03, grad_scale: 64.0 2023-04-26 12:13:49,878 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=8265.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:14:03,523 INFO [finetune.py:976] (6/7) Epoch 2, batch 2550, loss[loss=0.2362, simple_loss=0.2924, pruned_loss=0.08995, over 4796.00 frames. ], tot_loss[loss=0.2682, simple_loss=0.3147, pruned_loss=0.1109, over 955867.66 frames. ], batch size: 25, lr: 3.99e-03, grad_scale: 64.0 2023-04-26 12:14:10,627 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=8280.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:14:38,557 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([3.4147, 3.3417, 2.5728, 3.8586, 3.3562, 3.3906, 1.3504, 3.3012], device='cuda:6'), covar=tensor([0.1682, 0.1114, 0.2947, 0.2067, 0.2454, 0.1943, 0.5674, 0.2253], device='cuda:6'), in_proj_covar=tensor([0.0259, 0.0233, 0.0277, 0.0327, 0.0322, 0.0272, 0.0286, 0.0286], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-26 12:15:02,139 INFO [optim.py:369] (6/7) 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 12:15:13,568 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=8326.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:15:20,188 INFO [finetune.py:976] (6/7) Epoch 2, batch 2600, loss[loss=0.2381, simple_loss=0.2902, pruned_loss=0.09297, over 4744.00 frames. ], tot_loss[loss=0.2685, simple_loss=0.3159, pruned_loss=0.1106, over 957749.50 frames. ], batch size: 26, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 12:15:20,254 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=8328.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:16:05,459 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.71 vs. limit=2.0 2023-04-26 12:16:23,379 INFO [finetune.py:976] (6/7) Epoch 2, batch 2650, loss[loss=0.288, simple_loss=0.3334, pruned_loss=0.1214, over 4782.00 frames. ], tot_loss[loss=0.2705, simple_loss=0.3182, pruned_loss=0.1114, over 956905.38 frames. ], batch size: 51, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 12:16:49,626 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=8399.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:17:09,391 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.1989, 0.8561, 1.0466, 1.4728, 1.4481, 1.0980, 1.1169, 1.0841], device='cuda:6'), covar=tensor([4.0676, 5.9078, 6.6581, 6.7510, 4.1156, 7.2658, 6.9501, 5.1289], device='cuda:6'), in_proj_covar=tensor([0.0445, 0.0509, 0.0594, 0.0585, 0.0477, 0.0527, 0.0535, 0.0547], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-26 12:17:13,453 INFO [optim.py:369] (6/7) 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] (6/7) Epoch 2, batch 2700, loss[loss=0.2905, simple_loss=0.3329, pruned_loss=0.1241, over 4816.00 frames. ], tot_loss[loss=0.2707, simple_loss=0.3181, pruned_loss=0.1116, over 955803.61 frames. ], batch size: 40, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 12:17:33,005 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=8447.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:17:37,994 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.77 vs. limit=2.0 2023-04-26 12:17:40,750 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=8458.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:17:53,353 INFO [finetune.py:976] (6/7) Epoch 2, batch 2750, loss[loss=0.2402, simple_loss=0.2822, pruned_loss=0.09906, over 4767.00 frames. ], tot_loss[loss=0.2661, simple_loss=0.3134, pruned_loss=0.1093, over 955769.18 frames. ], batch size: 23, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 12:18:13,287 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=8506.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:18:20,780 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=5.21 vs. limit=5.0 2023-04-26 12:18:21,016 INFO [optim.py:369] (6/7) 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] (6/7) Epoch 2, batch 2800, loss[loss=0.2339, simple_loss=0.2915, pruned_loss=0.08816, over 4894.00 frames. ], tot_loss[loss=0.2611, simple_loss=0.3088, pruned_loss=0.1067, over 958161.60 frames. ], batch size: 35, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 12:18:33,550 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.85 vs. limit=5.0 2023-04-26 12:19:00,283 INFO [finetune.py:976] (6/7) Epoch 2, batch 2850, loss[loss=0.265, simple_loss=0.3207, pruned_loss=0.1047, over 4822.00 frames. ], tot_loss[loss=0.2579, simple_loss=0.3059, pruned_loss=0.105, over 957565.63 frames. ], batch size: 40, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 12:19:04,445 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6087, 3.9132, 0.7859, 1.9554, 2.0949, 2.4790, 2.4109, 0.8497], device='cuda:6'), covar=tensor([0.1385, 0.0898, 0.2350, 0.1451, 0.1096, 0.1275, 0.1438, 0.2431], device='cuda:6'), in_proj_covar=tensor([0.0125, 0.0274, 0.0153, 0.0134, 0.0145, 0.0168, 0.0130, 0.0135], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-26 12:19:45,302 INFO [optim.py:369] (6/7) 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 12:19:52,096 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=8621.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:19:56,367 INFO [finetune.py:976] (6/7) Epoch 2, batch 2900, loss[loss=0.3234, simple_loss=0.358, pruned_loss=0.1443, over 4837.00 frames. ], tot_loss[loss=0.2631, simple_loss=0.3109, pruned_loss=0.1077, over 958043.24 frames. ], batch size: 49, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 12:20:15,799 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.4343, 0.9092, 1.1058, 0.9745, 1.5644, 1.2876, 0.9600, 1.0762], device='cuda:6'), covar=tensor([0.1742, 0.1816, 0.2430, 0.1976, 0.1014, 0.1673, 0.2209, 0.2385], device='cuda:6'), in_proj_covar=tensor([0.0326, 0.0340, 0.0348, 0.0316, 0.0353, 0.0373, 0.0324, 0.0357], device='cuda:6'), out_proj_covar=tensor([7.1159e-05, 7.3159e-05, 7.5200e-05, 6.6333e-05, 7.5238e-05, 8.1612e-05, 7.0628e-05, 7.7099e-05], device='cuda:6') 2023-04-26 12:20:28,610 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.0095, 1.3377, 1.1951, 1.6307, 1.4518, 1.6158, 1.3528, 2.4840], device='cuda:6'), covar=tensor([0.0694, 0.0795, 0.0871, 0.1298, 0.0699, 0.0495, 0.0777, 0.0272], device='cuda:6'), in_proj_covar=tensor([0.0041, 0.0041, 0.0042, 0.0047, 0.0042, 0.0041, 0.0041, 0.0066], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0013, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0017], device='cuda:6') 2023-04-26 12:20:56,457 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9143, 2.6745, 2.0032, 2.4675, 1.9763, 2.1892, 2.4277, 1.8986], device='cuda:6'), covar=tensor([0.2516, 0.1576, 0.1291, 0.1481, 0.2906, 0.1346, 0.1981, 0.2632], device='cuda:6'), in_proj_covar=tensor([0.0317, 0.0336, 0.0247, 0.0313, 0.0317, 0.0287, 0.0281, 0.0304], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-26 12:21:07,305 INFO [finetune.py:976] (6/7) Epoch 2, batch 2950, loss[loss=0.2419, simple_loss=0.2856, pruned_loss=0.09911, over 4726.00 frames. ], tot_loss[loss=0.2668, simple_loss=0.3148, pruned_loss=0.1095, over 958333.60 frames. ], batch size: 23, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 12:22:03,511 INFO [optim.py:369] (6/7) 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 12:22:15,447 INFO [finetune.py:976] (6/7) Epoch 2, batch 3000, loss[loss=0.2868, simple_loss=0.337, pruned_loss=0.1183, over 4815.00 frames. ], tot_loss[loss=0.269, simple_loss=0.3167, pruned_loss=0.1107, over 958448.47 frames. ], batch size: 45, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 12:22:15,447 INFO [finetune.py:1001] (6/7) Computing validation loss 2023-04-26 12:22:27,475 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6139, 1.7145, 1.8414, 1.9993, 1.9377, 1.5361, 1.2143, 1.7949], device='cuda:6'), covar=tensor([0.1091, 0.1100, 0.0678, 0.0717, 0.0683, 0.1021, 0.1165, 0.0666], device='cuda:6'), in_proj_covar=tensor([0.0210, 0.0209, 0.0189, 0.0182, 0.0180, 0.0198, 0.0176, 0.0193], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-26 12:22:32,548 INFO [finetune.py:1010] (6/7) Epoch 2, validation: loss=0.1863, simple_loss=0.2571, pruned_loss=0.0578, over 2265189.00 frames. 2023-04-26 12:22:32,548 INFO [finetune.py:1011] (6/7) Maximum memory allocated so far is 6326MB 2023-04-26 12:23:09,092 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=8760.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:23:21,915 INFO [finetune.py:976] (6/7) Epoch 2, batch 3050, loss[loss=0.2376, simple_loss=0.3046, pruned_loss=0.08529, over 4824.00 frames. ], tot_loss[loss=0.2689, simple_loss=0.3172, pruned_loss=0.1103, over 957851.12 frames. ], batch size: 38, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 12:23:49,151 INFO [optim.py:369] (6/7) 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 12:23:51,014 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 2, batch 3100, loss[loss=0.2497, simple_loss=0.2911, pruned_loss=0.1041, over 4734.00 frames. ], tot_loss[loss=0.2655, simple_loss=0.3142, pruned_loss=0.1084, over 959005.77 frames. ], batch size: 54, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 12:24:04,013 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.2717, 1.4245, 1.3455, 1.3976, 1.4567, 1.6021, 1.4421, 1.4564], device='cuda:6'), covar=tensor([2.2087, 4.1208, 3.1667, 2.6656, 2.8199, 4.7120, 4.1693, 3.5037], device='cuda:6'), in_proj_covar=tensor([0.0278, 0.0378, 0.0303, 0.0305, 0.0333, 0.0368, 0.0365, 0.0331], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-26 12:24:29,317 INFO [finetune.py:976] (6/7) Epoch 2, batch 3150, loss[loss=0.2137, simple_loss=0.2683, pruned_loss=0.07951, over 4874.00 frames. ], tot_loss[loss=0.2637, simple_loss=0.3114, pruned_loss=0.108, over 958451.69 frames. ], batch size: 31, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 12:24:56,258 INFO [optim.py:369] (6/7) 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 12:24:57,623 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=8921.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:25:01,743 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 2, batch 3200, loss[loss=0.3274, simple_loss=0.3497, pruned_loss=0.1526, over 4853.00 frames. ], tot_loss[loss=0.2599, simple_loss=0.3075, pruned_loss=0.1062, over 958324.69 frames. ], batch size: 47, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 12:25:18,839 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0920, 2.6299, 2.2016, 2.5583, 2.0656, 2.1576, 2.2511, 1.9182], device='cuda:6'), covar=tensor([0.2092, 0.1461, 0.1132, 0.1341, 0.3243, 0.1670, 0.1811, 0.2740], device='cuda:6'), in_proj_covar=tensor([0.0318, 0.0338, 0.0248, 0.0313, 0.0318, 0.0290, 0.0281, 0.0305], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-26 12:25:30,262 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=8969.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:25:41,808 INFO [finetune.py:976] (6/7) Epoch 2, batch 3250, loss[loss=0.3082, simple_loss=0.3547, pruned_loss=0.1309, over 4899.00 frames. ], tot_loss[loss=0.2612, simple_loss=0.3083, pruned_loss=0.107, over 955009.37 frames. ], batch size: 37, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 12:25:53,853 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=8986.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:25:55,587 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=8988.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:26:04,941 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.4148, 1.2874, 4.1485, 3.8742, 3.7121, 3.9119, 3.9089, 3.6547], device='cuda:6'), covar=tensor([0.6942, 0.5691, 0.0963, 0.1705, 0.1014, 0.2050, 0.1472, 0.1583], device='cuda:6'), in_proj_covar=tensor([0.0335, 0.0315, 0.0449, 0.0457, 0.0380, 0.0434, 0.0343, 0.0401], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0002], device='cuda:6') 2023-04-26 12:26:27,430 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=9013.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:26:37,528 INFO [optim.py:369] (6/7) 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 12:26:39,505 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5288, 1.1784, 1.0407, 1.2528, 1.6814, 1.4362, 1.2267, 0.9716], device='cuda:6'), covar=tensor([0.1816, 0.1775, 0.2514, 0.1671, 0.1062, 0.1580, 0.2243, 0.2196], device='cuda:6'), in_proj_covar=tensor([0.0327, 0.0341, 0.0349, 0.0318, 0.0354, 0.0374, 0.0327, 0.0358], device='cuda:6'), out_proj_covar=tensor([7.1345e-05, 7.3262e-05, 7.5440e-05, 6.6719e-05, 7.5444e-05, 8.1868e-05, 7.1207e-05, 7.7205e-05], device='cuda:6') 2023-04-26 12:26:48,552 INFO [finetune.py:976] (6/7) Epoch 2, batch 3300, loss[loss=0.2941, simple_loss=0.3482, pruned_loss=0.12, over 4914.00 frames. ], tot_loss[loss=0.2655, simple_loss=0.3128, pruned_loss=0.1091, over 955040.99 frames. ], batch size: 36, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 12:26:51,737 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.75 vs. limit=2.0 2023-04-26 12:27:08,775 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=9047.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:27:26,233 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 2, batch 3350, loss[loss=0.2514, simple_loss=0.3035, pruned_loss=0.0996, over 4792.00 frames. ], tot_loss[loss=0.2677, simple_loss=0.3152, pruned_loss=0.1101, over 952687.17 frames. ], batch size: 51, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 12:27:58,642 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-04-26 12:28:05,403 INFO [zipformer.py:1188] (6/7) 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] (6/7) 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 12:28:17,534 INFO [finetune.py:976] (6/7) Epoch 2, batch 3400, loss[loss=0.253, simple_loss=0.2952, pruned_loss=0.1054, over 4857.00 frames. ], tot_loss[loss=0.2683, simple_loss=0.3162, pruned_loss=0.1102, over 954167.62 frames. ], batch size: 31, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 12:28:53,596 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.2862, 2.6410, 0.9854, 1.3604, 1.8453, 1.3669, 3.6180, 1.8047], device='cuda:6'), covar=tensor([0.0680, 0.0713, 0.0891, 0.1341, 0.0600, 0.0999, 0.0281, 0.0650], device='cuda:6'), in_proj_covar=tensor([0.0057, 0.0073, 0.0054, 0.0051, 0.0056, 0.0056, 0.0087, 0.0055], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0008, 0.0008, 0.0008, 0.0008, 0.0011, 0.0008], device='cuda:6') 2023-04-26 12:29:01,800 INFO [finetune.py:976] (6/7) Epoch 2, batch 3450, loss[loss=0.278, simple_loss=0.3102, pruned_loss=0.1229, over 4926.00 frames. ], tot_loss[loss=0.2657, simple_loss=0.3137, pruned_loss=0.1088, over 954781.17 frames. ], batch size: 28, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 12:29:16,109 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.3049, 2.6796, 1.1404, 1.3930, 1.9794, 1.3331, 3.5061, 1.7086], device='cuda:6'), covar=tensor([0.0673, 0.0795, 0.0926, 0.1215, 0.0563, 0.0994, 0.0237, 0.0664], device='cuda:6'), in_proj_covar=tensor([0.0057, 0.0073, 0.0054, 0.0051, 0.0056, 0.0056, 0.0087, 0.0054], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0008, 0.0008, 0.0008, 0.0008, 0.0011, 0.0008], device='cuda:6') 2023-04-26 12:29:29,512 INFO [optim.py:369] (6/7) 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 12:29:30,595 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-04-26 12:29:35,077 INFO [finetune.py:976] (6/7) Epoch 2, batch 3500, loss[loss=0.3219, simple_loss=0.3465, pruned_loss=0.1486, over 4915.00 frames. ], tot_loss[loss=0.2625, simple_loss=0.3102, pruned_loss=0.1074, over 956235.23 frames. ], batch size: 37, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 12:29:39,616 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.64 vs. limit=5.0 2023-04-26 12:29:43,441 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([4.6318, 4.5935, 3.3516, 5.2931, 4.6252, 4.5996, 2.4765, 4.5151], device='cuda:6'), covar=tensor([0.1405, 0.0783, 0.2725, 0.0846, 0.3213, 0.1427, 0.4962, 0.1795], device='cuda:6'), in_proj_covar=tensor([0.0256, 0.0231, 0.0275, 0.0325, 0.0320, 0.0268, 0.0284, 0.0284], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-26 12:29:49,576 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.77 vs. limit=2.0 2023-04-26 12:29:53,753 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 2023-04-26 12:30:08,949 INFO [finetune.py:976] (6/7) Epoch 2, batch 3550, loss[loss=0.2466, simple_loss=0.2968, pruned_loss=0.09821, over 4706.00 frames. ], tot_loss[loss=0.2581, simple_loss=0.3058, pruned_loss=0.1052, over 955964.59 frames. ], batch size: 59, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 12:30:12,106 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=9283.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:30:37,458 INFO [optim.py:369] (6/7) 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 12:30:48,977 INFO [finetune.py:976] (6/7) Epoch 2, batch 3600, loss[loss=0.3194, simple_loss=0.3506, pruned_loss=0.1441, over 4839.00 frames. ], tot_loss[loss=0.2551, simple_loss=0.3029, pruned_loss=0.1036, over 957478.92 frames. ], batch size: 49, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 12:31:03,422 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=9342.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:31:44,878 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=9369.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:31:55,589 INFO [finetune.py:976] (6/7) Epoch 2, batch 3650, loss[loss=0.2666, simple_loss=0.3021, pruned_loss=0.1155, over 4721.00 frames. ], tot_loss[loss=0.2576, simple_loss=0.3057, pruned_loss=0.1048, over 955901.29 frames. ], batch size: 23, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 12:32:25,875 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=9414.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:32:27,071 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=9416.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:32:28,290 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([3.8029, 3.6605, 2.7757, 4.3136, 3.7227, 3.7675, 1.7396, 3.6499], device='cuda:6'), covar=tensor([0.1716, 0.1161, 0.3553, 0.1804, 0.2842, 0.2032, 0.5754, 0.2533], device='cuda:6'), in_proj_covar=tensor([0.0257, 0.0232, 0.0274, 0.0325, 0.0320, 0.0269, 0.0283, 0.0284], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-26 12:32:28,805 INFO [optim.py:369] (6/7) 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 12:32:34,830 INFO [finetune.py:976] (6/7) Epoch 2, batch 3700, loss[loss=0.2454, simple_loss=0.2762, pruned_loss=0.1073, over 3908.00 frames. ], tot_loss[loss=0.2612, simple_loss=0.3105, pruned_loss=0.106, over 955799.98 frames. ], batch size: 17, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 12:32:51,703 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6858, 1.2429, 1.1261, 1.1897, 1.7947, 1.4851, 1.2090, 1.0594], device='cuda:6'), covar=tensor([0.2276, 0.2925, 0.3802, 0.3303, 0.1558, 0.2818, 0.3372, 0.3414], device='cuda:6'), in_proj_covar=tensor([0.0328, 0.0342, 0.0350, 0.0320, 0.0357, 0.0375, 0.0326, 0.0359], device='cuda:6'), out_proj_covar=tensor([7.1676e-05, 7.3537e-05, 7.5710e-05, 6.7221e-05, 7.6063e-05, 8.2019e-05, 7.0970e-05, 7.7547e-05], device='cuda:6') 2023-04-26 12:32:58,828 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=9464.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:33:07,020 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=9475.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:33:08,732 INFO [finetune.py:976] (6/7) Epoch 2, batch 3750, loss[loss=0.2435, simple_loss=0.2712, pruned_loss=0.1079, over 4014.00 frames. ], tot_loss[loss=0.2616, simple_loss=0.3115, pruned_loss=0.1058, over 956535.19 frames. ], batch size: 17, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 12:33:47,551 INFO [optim.py:369] (6/7) 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 12:33:59,190 INFO [finetune.py:976] (6/7) Epoch 2, batch 3800, loss[loss=0.275, simple_loss=0.2996, pruned_loss=0.1252, over 4030.00 frames. ], tot_loss[loss=0.2656, simple_loss=0.3151, pruned_loss=0.1081, over 954029.73 frames. ], batch size: 17, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 12:34:13,264 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.3951, 1.4685, 1.4614, 2.0836, 2.3662, 1.9573, 1.8023, 1.6268], device='cuda:6'), covar=tensor([0.2238, 0.2791, 0.2645, 0.2653, 0.1731, 0.3030, 0.3249, 0.2394], device='cuda:6'), in_proj_covar=tensor([0.0326, 0.0341, 0.0349, 0.0319, 0.0354, 0.0372, 0.0325, 0.0357], device='cuda:6'), out_proj_covar=tensor([7.1165e-05, 7.3311e-05, 7.5429e-05, 6.6906e-05, 7.5542e-05, 8.1471e-05, 7.0733e-05, 7.7118e-05], device='cuda:6') 2023-04-26 12:34:42,757 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.3747, 1.4263, 3.8959, 3.6341, 3.4911, 3.7588, 3.8017, 3.4430], device='cuda:6'), covar=tensor([0.6413, 0.5174, 0.1046, 0.1659, 0.1024, 0.1516, 0.1227, 0.1417], device='cuda:6'), in_proj_covar=tensor([0.0333, 0.0312, 0.0446, 0.0450, 0.0376, 0.0428, 0.0340, 0.0396], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0002], device='cuda:6') 2023-04-26 12:34:55,619 INFO [finetune.py:976] (6/7) Epoch 2, batch 3850, loss[loss=0.2447, simple_loss=0.2996, pruned_loss=0.0949, over 4722.00 frames. ], tot_loss[loss=0.263, simple_loss=0.3126, pruned_loss=0.1067, over 954446.48 frames. ], batch size: 54, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 12:34:58,842 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=9583.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:34:59,461 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=9584.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:35:00,309 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.94 vs. limit=5.0 2023-04-26 12:35:03,800 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7109, 1.6410, 1.8174, 2.1153, 2.1229, 1.6676, 1.2308, 1.8371], device='cuda:6'), covar=tensor([0.1108, 0.1262, 0.0805, 0.0765, 0.0675, 0.1098, 0.1302, 0.0736], device='cuda:6'), in_proj_covar=tensor([0.0211, 0.0210, 0.0189, 0.0184, 0.0180, 0.0199, 0.0177, 0.0194], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-26 12:35:22,631 INFO [optim.py:369] (6/7) 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 12:35:25,978 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2023-04-26 12:35:29,161 INFO [finetune.py:976] (6/7) Epoch 2, batch 3900, loss[loss=0.2397, simple_loss=0.2925, pruned_loss=0.09351, over 4790.00 frames. ], tot_loss[loss=0.2581, simple_loss=0.3077, pruned_loss=0.1042, over 956493.25 frames. ], batch size: 29, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 12:35:38,177 INFO [zipformer.py:1188] (6/7) 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] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=9642.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:35:52,653 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=9645.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:36:02,310 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.0300, 1.9486, 1.7714, 1.6481, 2.1336, 1.6629, 2.7848, 1.6106], device='cuda:6'), covar=tensor([0.5199, 0.2311, 0.5341, 0.4439, 0.2129, 0.3189, 0.1457, 0.4840], device='cuda:6'), in_proj_covar=tensor([0.0345, 0.0348, 0.0435, 0.0368, 0.0402, 0.0371, 0.0397, 0.0411], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-26 12:36:12,977 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=9658.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:36:14,186 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.1834, 1.5444, 1.3482, 1.7925, 1.6352, 1.9081, 1.4189, 3.4200], device='cuda:6'), covar=tensor([0.0694, 0.0801, 0.0832, 0.1288, 0.0661, 0.0508, 0.0760, 0.0149], device='cuda:6'), in_proj_covar=tensor([0.0040, 0.0041, 0.0042, 0.0046, 0.0042, 0.0041, 0.0041, 0.0065], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0013, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0017], device='cuda:6') 2023-04-26 12:36:32,446 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=9669.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:36:38,400 INFO [finetune.py:976] (6/7) Epoch 2, batch 3950, loss[loss=0.2237, simple_loss=0.2781, pruned_loss=0.0847, over 4778.00 frames. ], tot_loss[loss=0.2541, simple_loss=0.3036, pruned_loss=0.1024, over 955743.09 frames. ], batch size: 28, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 12:36:55,195 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=9690.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:37:29,751 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=9717.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:37:36,035 INFO [optim.py:369] (6/7) 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 12:37:36,691 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=9719.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:37:47,964 INFO [finetune.py:976] (6/7) Epoch 2, batch 4000, loss[loss=0.2341, simple_loss=0.2899, pruned_loss=0.08916, over 4929.00 frames. ], tot_loss[loss=0.2545, simple_loss=0.3037, pruned_loss=0.1026, over 954583.36 frames. ], batch size: 33, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 12:37:49,999 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.64 vs. limit=2.0 2023-04-26 12:37:58,323 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=9743.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:38:15,582 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=9770.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:38:20,830 INFO [finetune.py:976] (6/7) Epoch 2, batch 4050, loss[loss=0.2415, simple_loss=0.2794, pruned_loss=0.1018, over 4262.00 frames. ], tot_loss[loss=0.2582, simple_loss=0.3073, pruned_loss=0.1046, over 953656.05 frames. ], batch size: 18, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 12:38:26,819 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5865, 2.1154, 1.4176, 1.2524, 1.1744, 1.1931, 1.4033, 1.1411], device='cuda:6'), covar=tensor([0.2385, 0.1850, 0.2824, 0.3123, 0.3881, 0.2861, 0.2173, 0.3170], device='cuda:6'), in_proj_covar=tensor([0.0200, 0.0227, 0.0196, 0.0218, 0.0235, 0.0197, 0.0192, 0.0211], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-26 12:38:38,420 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=9804.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:38:47,819 INFO [optim.py:369] (6/7) 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 12:38:54,757 INFO [finetune.py:976] (6/7) Epoch 2, batch 4100, loss[loss=0.2803, simple_loss=0.3306, pruned_loss=0.1149, over 4887.00 frames. ], tot_loss[loss=0.2592, simple_loss=0.3086, pruned_loss=0.1048, over 952951.43 frames. ], batch size: 43, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 12:39:03,066 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.3479, 1.3846, 3.8799, 3.6221, 3.4748, 3.7251, 3.7480, 3.4025], device='cuda:6'), covar=tensor([0.6733, 0.5558, 0.1115, 0.1736, 0.1116, 0.1622, 0.1229, 0.1574], device='cuda:6'), in_proj_covar=tensor([0.0333, 0.0313, 0.0448, 0.0451, 0.0378, 0.0432, 0.0341, 0.0397], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0002], device='cuda:6') 2023-04-26 12:39:03,507 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.68 vs. limit=2.0 2023-04-26 12:39:30,757 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-04-26 12:39:34,300 INFO [finetune.py:976] (6/7) Epoch 2, batch 4150, loss[loss=0.2191, simple_loss=0.2865, pruned_loss=0.07581, over 4793.00 frames. ], tot_loss[loss=0.2612, simple_loss=0.3096, pruned_loss=0.1063, over 952735.65 frames. ], batch size: 29, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 12:40:18,084 INFO [optim.py:369] (6/7) 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 12:40:29,028 INFO [finetune.py:976] (6/7) Epoch 2, batch 4200, loss[loss=0.2411, simple_loss=0.3022, pruned_loss=0.09001, over 4918.00 frames. ], tot_loss[loss=0.2609, simple_loss=0.3105, pruned_loss=0.1056, over 954447.33 frames. ], batch size: 33, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 12:40:42,860 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.62 vs. limit=2.0 2023-04-26 12:40:49,802 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=9940.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:41:04,586 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-04-26 12:41:13,624 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 2023-04-26 12:41:36,132 INFO [finetune.py:976] (6/7) Epoch 2, batch 4250, loss[loss=0.2435, simple_loss=0.2879, pruned_loss=0.09954, over 4824.00 frames. ], tot_loss[loss=0.2585, simple_loss=0.3081, pruned_loss=0.1044, over 955004.01 frames. ], batch size: 33, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 12:42:30,340 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=10014.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:42:33,327 INFO [optim.py:369] (6/7) 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] (6/7) Epoch 2, batch 4300, loss[loss=0.2655, simple_loss=0.3046, pruned_loss=0.1132, over 4811.00 frames. ], tot_loss[loss=0.2554, simple_loss=0.3049, pruned_loss=0.103, over 955013.95 frames. ], batch size: 45, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 12:43:02,741 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=10039.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:43:18,153 INFO [zipformer.py:1188] (6/7) 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] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=10070.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:43:27,738 INFO [finetune.py:976] (6/7) Epoch 2, batch 4350, loss[loss=0.1976, simple_loss=0.243, pruned_loss=0.07609, over 4285.00 frames. ], tot_loss[loss=0.252, simple_loss=0.3011, pruned_loss=0.1014, over 955958.62 frames. ], batch size: 18, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 12:43:33,750 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.0785, 2.8607, 1.1266, 1.4103, 1.4240, 2.0763, 1.8346, 1.0733], device='cuda:6'), covar=tensor([0.1939, 0.1574, 0.2043, 0.1948, 0.1511, 0.1322, 0.1542, 0.1959], device='cuda:6'), in_proj_covar=tensor([0.0125, 0.0272, 0.0152, 0.0132, 0.0144, 0.0166, 0.0129, 0.0134], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-26 12:43:41,995 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=10099.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:43:43,157 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=10100.0, num_to_drop=1, layers_to_drop={2} 2023-04-26 12:43:54,966 INFO [zipformer.py:1188] (6/7) 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] (6/7) 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 12:43:58,063 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=10123.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 12:44:00,983 INFO [finetune.py:976] (6/7) Epoch 2, batch 4400, loss[loss=0.3071, simple_loss=0.3543, pruned_loss=0.13, over 4916.00 frames. ], tot_loss[loss=0.2557, simple_loss=0.3041, pruned_loss=0.1036, over 955408.25 frames. ], batch size: 36, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 12:44:34,794 INFO [finetune.py:976] (6/7) Epoch 2, batch 4450, loss[loss=0.2824, simple_loss=0.3343, pruned_loss=0.1153, over 4845.00 frames. ], tot_loss[loss=0.2596, simple_loss=0.3087, pruned_loss=0.1053, over 955181.66 frames. ], batch size: 44, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 12:44:39,410 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.76 vs. limit=2.0 2023-04-26 12:45:03,112 INFO [optim.py:369] (6/7) 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 12:45:08,636 INFO [finetune.py:976] (6/7) Epoch 2, batch 4500, loss[loss=0.2435, simple_loss=0.3019, pruned_loss=0.09253, over 4900.00 frames. ], tot_loss[loss=0.2613, simple_loss=0.3104, pruned_loss=0.1061, over 954727.99 frames. ], batch size: 37, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 12:45:16,001 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 2, batch 4550, loss[loss=0.2439, simple_loss=0.3019, pruned_loss=0.0929, over 4931.00 frames. ], tot_loss[loss=0.2606, simple_loss=0.3103, pruned_loss=0.1055, over 955673.56 frames. ], batch size: 42, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 12:45:48,417 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=10288.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:46:21,452 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=10314.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:46:29,831 INFO [optim.py:369] (6/7) 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 12:46:41,916 INFO [finetune.py:976] (6/7) Epoch 2, batch 4600, loss[loss=0.2683, simple_loss=0.3204, pruned_loss=0.1081, over 4796.00 frames. ], tot_loss[loss=0.2587, simple_loss=0.309, pruned_loss=0.1042, over 955764.00 frames. ], batch size: 51, lr: 3.99e-03, grad_scale: 64.0 2023-04-26 12:46:56,433 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-04-26 12:47:09,971 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=10362.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:47:10,003 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6234, 3.8585, 0.7813, 2.1233, 2.0381, 2.5984, 2.4217, 1.0278], device='cuda:6'), covar=tensor([0.1325, 0.0856, 0.2280, 0.1227, 0.1116, 0.1106, 0.1307, 0.2099], device='cuda:6'), in_proj_covar=tensor([0.0126, 0.0275, 0.0153, 0.0133, 0.0146, 0.0168, 0.0131, 0.0135], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-26 12:47:29,154 INFO [finetune.py:976] (6/7) Epoch 2, batch 4650, loss[loss=0.2139, simple_loss=0.2743, pruned_loss=0.07674, over 4759.00 frames. ], tot_loss[loss=0.2559, simple_loss=0.3059, pruned_loss=0.1029, over 957235.09 frames. ], batch size: 28, lr: 3.99e-03, grad_scale: 64.0 2023-04-26 12:47:49,975 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=2.13 vs. limit=2.0 2023-04-26 12:47:50,969 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=10395.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 12:47:58,953 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=10399.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:48:03,073 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0822, 1.8065, 2.0251, 2.4553, 2.3743, 1.9090, 1.5535, 2.1623], device='cuda:6'), covar=tensor([0.1133, 0.1330, 0.0779, 0.0684, 0.0728, 0.1119, 0.1288, 0.0738], device='cuda:6'), in_proj_covar=tensor([0.0216, 0.0215, 0.0193, 0.0187, 0.0184, 0.0204, 0.0181, 0.0199], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-26 12:48:09,560 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.3862, 1.6288, 1.2081, 0.9889, 1.1159, 1.0892, 1.1823, 1.0576], device='cuda:6'), covar=tensor([0.2401, 0.1843, 0.2599, 0.2795, 0.3628, 0.2861, 0.1990, 0.3041], device='cuda:6'), in_proj_covar=tensor([0.0200, 0.0227, 0.0195, 0.0218, 0.0234, 0.0197, 0.0191, 0.0210], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-26 12:48:12,453 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([3.8370, 3.7616, 2.8184, 4.4212, 3.8443, 3.8320, 1.6373, 3.6350], device='cuda:6'), covar=tensor([0.1587, 0.1183, 0.2811, 0.1496, 0.2561, 0.1875, 0.5829, 0.2316], device='cuda:6'), in_proj_covar=tensor([0.0254, 0.0229, 0.0271, 0.0320, 0.0317, 0.0266, 0.0280, 0.0283], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-26 12:48:12,486 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=10409.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:48:21,121 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9961, 2.2915, 0.9390, 1.2546, 1.5770, 1.2750, 2.5183, 1.4812], device='cuda:6'), covar=tensor([0.0693, 0.0616, 0.0757, 0.1298, 0.0504, 0.1004, 0.0335, 0.0708], device='cuda:6'), in_proj_covar=tensor([0.0056, 0.0072, 0.0054, 0.0050, 0.0055, 0.0055, 0.0085, 0.0054], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0008, 0.0008, 0.0008, 0.0008, 0.0011, 0.0008], device='cuda:6') 2023-04-26 12:48:21,156 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.1666, 0.9766, 1.2508, 1.1900, 1.0679, 0.8876, 0.9561, 0.5577], device='cuda:6'), covar=tensor([0.0842, 0.1043, 0.0888, 0.0985, 0.1300, 0.1830, 0.0859, 0.1618], device='cuda:6'), in_proj_covar=tensor([0.0071, 0.0080, 0.0076, 0.0073, 0.0087, 0.0096, 0.0090, 0.0081], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003], device='cuda:6') 2023-04-26 12:48:23,999 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=10418.0, num_to_drop=1, layers_to_drop={3} 2023-04-26 12:48:25,138 INFO [optim.py:369] (6/7) 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 12:48:35,918 INFO [finetune.py:976] (6/7) Epoch 2, batch 4700, loss[loss=0.2384, simple_loss=0.2783, pruned_loss=0.09924, over 4892.00 frames. ], tot_loss[loss=0.2532, simple_loss=0.3023, pruned_loss=0.102, over 957623.18 frames. ], batch size: 32, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 12:48:47,933 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.21 vs. limit=5.0 2023-04-26 12:48:53,257 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=10447.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:48:57,649 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=10454.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:49:06,105 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=5.04 vs. limit=5.0 2023-04-26 12:49:10,032 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 2, batch 4750, loss[loss=0.2841, simple_loss=0.332, pruned_loss=0.1181, over 4850.00 frames. ], tot_loss[loss=0.2503, simple_loss=0.2992, pruned_loss=0.1008, over 955971.24 frames. ], batch size: 44, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 12:49:25,323 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.45 vs. limit=5.0 2023-04-26 12:49:32,950 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0824, 1.2564, 1.2358, 1.3311, 1.3370, 1.5075, 1.3498, 1.3534], device='cuda:6'), covar=tensor([1.8110, 3.2739, 2.6327, 2.2303, 2.6312, 4.1170, 3.3510, 2.6805], device='cuda:6'), in_proj_covar=tensor([0.0288, 0.0388, 0.0309, 0.0313, 0.0341, 0.0383, 0.0373, 0.0338], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-26 12:49:37,014 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6216, 1.1262, 1.3762, 1.4249, 1.2305, 1.0988, 0.5692, 1.0268], device='cuda:6'), covar=tensor([0.4886, 0.6402, 0.2769, 0.4849, 0.5536, 0.4690, 0.8583, 0.5747], device='cuda:6'), in_proj_covar=tensor([0.0270, 0.0277, 0.0223, 0.0347, 0.0235, 0.0235, 0.0269, 0.0216], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-26 12:49:40,440 INFO [zipformer.py:1188] (6/7) 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] (6/7) 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] (6/7) Epoch 2, batch 4800, loss[loss=0.2579, simple_loss=0.3035, pruned_loss=0.1061, over 4760.00 frames. ], tot_loss[loss=0.251, simple_loss=0.3008, pruned_loss=0.1006, over 955541.01 frames. ], batch size: 27, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 12:50:23,189 INFO [finetune.py:976] (6/7) Epoch 2, batch 4850, loss[loss=0.2934, simple_loss=0.3452, pruned_loss=0.1208, over 4816.00 frames. ], tot_loss[loss=0.2546, simple_loss=0.3053, pruned_loss=0.102, over 955245.16 frames. ], batch size: 39, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 12:50:49,757 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.61 vs. limit=2.0 2023-04-26 12:50:50,786 INFO [optim.py:369] (6/7) 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 12:50:56,081 INFO [finetune.py:976] (6/7) Epoch 2, batch 4900, loss[loss=0.2871, simple_loss=0.3295, pruned_loss=0.1223, over 4808.00 frames. ], tot_loss[loss=0.2557, simple_loss=0.3068, pruned_loss=0.1023, over 954022.16 frames. ], batch size: 33, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 12:51:39,526 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.75 vs. limit=2.0 2023-04-26 12:51:46,562 INFO [finetune.py:976] (6/7) Epoch 2, batch 4950, loss[loss=0.3287, simple_loss=0.3429, pruned_loss=0.1572, over 4231.00 frames. ], tot_loss[loss=0.2575, simple_loss=0.3081, pruned_loss=0.1034, over 953360.23 frames. ], batch size: 65, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 12:51:57,999 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.3972, 3.2668, 0.9663, 1.9524, 1.8255, 2.4881, 2.0659, 1.0790], device='cuda:6'), covar=tensor([0.1301, 0.0843, 0.1872, 0.1247, 0.1017, 0.0910, 0.1305, 0.2106], device='cuda:6'), in_proj_covar=tensor([0.0125, 0.0272, 0.0152, 0.0132, 0.0144, 0.0166, 0.0129, 0.0134], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-26 12:52:10,656 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=10695.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:52:42,670 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=10718.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 12:52:43,791 INFO [optim.py:369] (6/7) 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] (6/7) Epoch 2, batch 5000, loss[loss=0.2233, simple_loss=0.2727, pruned_loss=0.08691, over 4825.00 frames. ], tot_loss[loss=0.2552, simple_loss=0.3057, pruned_loss=0.1023, over 954205.81 frames. ], batch size: 38, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 12:53:16,591 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=10743.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:53:17,275 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=10744.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:53:29,936 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=10765.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:53:30,549 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=10766.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 12:53:42,294 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 2, batch 5050, loss[loss=0.1986, simple_loss=0.2652, pruned_loss=0.06598, over 4859.00 frames. ], tot_loss[loss=0.2507, simple_loss=0.3013, pruned_loss=0.1001, over 952809.44 frames. ], batch size: 34, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 12:53:52,810 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-04-26 12:54:25,690 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=10805.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:54:28,749 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=10810.0, num_to_drop=1, layers_to_drop={2} 2023-04-26 12:54:40,506 INFO [optim.py:369] (6/7) 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 12:54:51,192 INFO [finetune.py:976] (6/7) Epoch 2, batch 5100, loss[loss=0.2038, simple_loss=0.2559, pruned_loss=0.07583, over 4829.00 frames. ], tot_loss[loss=0.2463, simple_loss=0.297, pruned_loss=0.09783, over 955246.64 frames. ], batch size: 33, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 12:55:02,000 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=10835.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:55:45,707 INFO [finetune.py:976] (6/7) Epoch 2, batch 5150, loss[loss=0.2469, simple_loss=0.3056, pruned_loss=0.09406, over 4890.00 frames. ], tot_loss[loss=0.2459, simple_loss=0.2963, pruned_loss=0.09778, over 955142.25 frames. ], batch size: 35, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 12:55:50,632 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=10885.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 12:55:50,702 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.75 vs. limit=2.0 2023-04-26 12:56:13,137 INFO [optim.py:369] (6/7) 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 12:56:18,493 INFO [finetune.py:976] (6/7) Epoch 2, batch 5200, loss[loss=0.2647, simple_loss=0.3108, pruned_loss=0.1093, over 4823.00 frames. ], tot_loss[loss=0.2507, simple_loss=0.3014, pruned_loss=0.1, over 953620.78 frames. ], batch size: 30, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 12:56:32,004 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=10946.0, num_to_drop=1, layers_to_drop={3} 2023-04-26 12:56:57,990 INFO [finetune.py:976] (6/7) Epoch 2, batch 5250, loss[loss=0.3015, simple_loss=0.348, pruned_loss=0.1275, over 4809.00 frames. ], tot_loss[loss=0.2537, simple_loss=0.3047, pruned_loss=0.1013, over 954313.27 frames. ], batch size: 40, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 12:57:20,376 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-04-26 12:57:34,877 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.1084, 1.3534, 1.2492, 1.3560, 1.3590, 1.4944, 1.3875, 1.3913], device='cuda:6'), covar=tensor([1.8458, 2.9384, 2.4791, 2.1096, 2.4911, 4.2868, 3.0592, 2.5415], device='cuda:6'), in_proj_covar=tensor([0.0292, 0.0393, 0.0312, 0.0316, 0.0345, 0.0389, 0.0378, 0.0342], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-26 12:57:40,211 INFO [optim.py:369] (6/7) 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 12:57:49,682 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.4898, 1.2714, 4.1256, 3.8319, 3.6802, 3.9283, 3.8966, 3.6650], device='cuda:6'), covar=tensor([0.6629, 0.5488, 0.1036, 0.1618, 0.1093, 0.1180, 0.1398, 0.1494], device='cuda:6'), in_proj_covar=tensor([0.0326, 0.0307, 0.0438, 0.0442, 0.0372, 0.0424, 0.0334, 0.0392], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-26 12:57:51,421 INFO [finetune.py:976] (6/7) Epoch 2, batch 5300, loss[loss=0.2615, simple_loss=0.3264, pruned_loss=0.09828, over 4925.00 frames. ], tot_loss[loss=0.2572, simple_loss=0.3075, pruned_loss=0.1035, over 953269.98 frames. ], batch size: 42, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 12:58:45,135 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=11065.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:58:46,362 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([0.9944, 1.3584, 3.2827, 3.0160, 2.9476, 3.2144, 3.2026, 2.9021], device='cuda:6'), covar=tensor([0.6788, 0.5033, 0.1472, 0.2101, 0.1416, 0.1731, 0.1623, 0.1697], device='cuda:6'), in_proj_covar=tensor([0.0328, 0.0308, 0.0439, 0.0444, 0.0373, 0.0426, 0.0335, 0.0393], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0002], device='cuda:6') 2023-04-26 12:58:57,154 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.5188, 1.7686, 1.5979, 2.1393, 2.4330, 2.0790, 1.9282, 1.8566], device='cuda:6'), covar=tensor([0.1618, 0.1982, 0.2324, 0.2699, 0.1395, 0.2392, 0.2658, 0.2077], device='cuda:6'), in_proj_covar=tensor([0.0324, 0.0340, 0.0348, 0.0317, 0.0351, 0.0367, 0.0322, 0.0355], device='cuda:6'), out_proj_covar=tensor([7.0766e-05, 7.3100e-05, 7.5387e-05, 6.6629e-05, 7.4881e-05, 8.0324e-05, 7.0107e-05, 7.6846e-05], device='cuda:6') 2023-04-26 12:58:57,626 INFO [finetune.py:976] (6/7) Epoch 2, batch 5350, loss[loss=0.1693, simple_loss=0.2324, pruned_loss=0.05309, over 4758.00 frames. ], tot_loss[loss=0.2566, simple_loss=0.3078, pruned_loss=0.1027, over 955945.42 frames. ], batch size: 26, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 12:59:29,511 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=11100.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 12:59:47,709 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=11110.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 12:59:50,026 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=11113.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:00:00,284 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 2023-04-26 13:00:00,705 INFO [optim.py:369] (6/7) 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 13:00:11,121 INFO [finetune.py:976] (6/7) Epoch 2, batch 5400, loss[loss=0.2949, simple_loss=0.3257, pruned_loss=0.1321, over 4248.00 frames. ], tot_loss[loss=0.2543, simple_loss=0.3051, pruned_loss=0.1018, over 955132.32 frames. ], batch size: 65, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 13:00:12,411 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=11130.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:00:27,892 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.1758, 2.6988, 1.0889, 1.3646, 1.8935, 1.3195, 3.2573, 1.7082], device='cuda:6'), covar=tensor([0.0652, 0.0593, 0.0883, 0.1264, 0.0561, 0.1000, 0.0213, 0.0670], device='cuda:6'), in_proj_covar=tensor([0.0056, 0.0073, 0.0054, 0.0050, 0.0055, 0.0056, 0.0085, 0.0054], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0008, 0.0008, 0.0008, 0.0008, 0.0011, 0.0008], device='cuda:6') 2023-04-26 13:00:32,457 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=11158.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:00:42,738 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=11174.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:00:45,043 INFO [finetune.py:976] (6/7) Epoch 2, batch 5450, loss[loss=0.2173, simple_loss=0.2717, pruned_loss=0.08148, over 4903.00 frames. ], tot_loss[loss=0.249, simple_loss=0.3, pruned_loss=0.09896, over 954770.67 frames. ], batch size: 32, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 13:01:09,023 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-04-26 13:01:18,384 INFO [optim.py:369] (6/7) 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] (6/7) Epoch 2, batch 5500, loss[loss=0.1853, simple_loss=0.2518, pruned_loss=0.05939, over 4870.00 frames. ], tot_loss[loss=0.245, simple_loss=0.2961, pruned_loss=0.09689, over 954779.97 frames. ], batch size: 31, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 13:01:27,467 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8901, 1.8571, 2.0572, 2.2490, 2.2422, 1.7868, 1.4722, 1.9903], device='cuda:6'), covar=tensor([0.1052, 0.1133, 0.0648, 0.0721, 0.0665, 0.1053, 0.1273, 0.0684], device='cuda:6'), in_proj_covar=tensor([0.0215, 0.0215, 0.0193, 0.0189, 0.0185, 0.0204, 0.0181, 0.0200], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-26 13:01:28,093 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=11235.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:01:29,930 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7484, 0.8522, 1.1338, 1.3059, 1.3700, 1.4792, 1.1947, 1.2473], device='cuda:6'), covar=tensor([1.3139, 2.3008, 1.9983, 1.6575, 1.9944, 3.2667, 2.4331, 2.0584], device='cuda:6'), in_proj_covar=tensor([0.0290, 0.0391, 0.0309, 0.0313, 0.0342, 0.0387, 0.0375, 0.0338], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-26 13:01:31,702 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=11241.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 13:01:57,530 INFO [finetune.py:976] (6/7) Epoch 2, batch 5550, loss[loss=0.2947, simple_loss=0.3408, pruned_loss=0.1243, over 4729.00 frames. ], tot_loss[loss=0.246, simple_loss=0.2973, pruned_loss=0.09732, over 954935.85 frames. ], batch size: 59, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 13:02:35,256 INFO [optim.py:369] (6/7) 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] (6/7) Epoch 2, batch 5600, loss[loss=0.2372, simple_loss=0.2864, pruned_loss=0.09401, over 4832.00 frames. ], tot_loss[loss=0.2476, simple_loss=0.2999, pruned_loss=0.09765, over 954017.62 frames. ], batch size: 30, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 13:03:10,814 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.3832, 3.0288, 1.0282, 1.5464, 2.4230, 1.3077, 4.2329, 1.9106], device='cuda:6'), covar=tensor([0.0671, 0.0710, 0.0955, 0.1338, 0.0549, 0.1080, 0.0231, 0.0665], device='cuda:6'), in_proj_covar=tensor([0.0056, 0.0073, 0.0054, 0.0050, 0.0055, 0.0056, 0.0086, 0.0054], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0008, 0.0008, 0.0008, 0.0008, 0.0011, 0.0008], device='cuda:6') 2023-04-26 13:03:42,987 INFO [finetune.py:976] (6/7) Epoch 2, batch 5650, loss[loss=0.2077, simple_loss=0.2549, pruned_loss=0.08023, over 4021.00 frames. ], tot_loss[loss=0.2504, simple_loss=0.3029, pruned_loss=0.09894, over 954318.03 frames. ], batch size: 17, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 13:04:13,895 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=11400.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:04:25,342 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8070, 1.8964, 2.1730, 2.0663, 1.9483, 1.7740, 1.9015, 1.8139], device='cuda:6'), covar=tensor([2.7098, 3.7093, 4.2796, 4.6985, 3.0737, 4.8400, 4.9098, 3.5562], device='cuda:6'), in_proj_covar=tensor([0.0449, 0.0505, 0.0601, 0.0600, 0.0482, 0.0521, 0.0533, 0.0544], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-26 13:04:31,190 INFO [optim.py:369] (6/7) 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 13:04:35,907 INFO [finetune.py:976] (6/7) Epoch 2, batch 5700, loss[loss=0.2294, simple_loss=0.2579, pruned_loss=0.1004, over 3958.00 frames. ], tot_loss[loss=0.2476, simple_loss=0.2987, pruned_loss=0.09826, over 937811.47 frames. ], batch size: 17, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 13:04:35,973 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8104, 3.7143, 1.1877, 1.9345, 2.1293, 2.6578, 2.2541, 1.3793], device='cuda:6'), covar=tensor([0.1346, 0.1309, 0.2123, 0.1553, 0.1148, 0.1059, 0.1523, 0.1728], device='cuda:6'), in_proj_covar=tensor([0.0125, 0.0272, 0.0153, 0.0133, 0.0144, 0.0165, 0.0129, 0.0134], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-26 13:04:37,208 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=11430.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:04:47,962 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=11448.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:05:07,858 INFO [finetune.py:976] (6/7) Epoch 3, batch 0, loss[loss=0.2796, simple_loss=0.3229, pruned_loss=0.1181, over 4753.00 frames. ], tot_loss[loss=0.2796, simple_loss=0.3229, pruned_loss=0.1181, over 4753.00 frames. ], batch size: 26, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 13:05:07,858 INFO [finetune.py:1001] (6/7) Computing validation loss 2023-04-26 13:05:10,464 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.4134, 0.8962, 1.1586, 1.6563, 1.5904, 1.2633, 1.2392, 1.2358], device='cuda:6'), covar=tensor([2.6388, 3.7269, 4.2649, 4.6405, 2.7835, 4.4190, 4.6395, 3.2594], device='cuda:6'), in_proj_covar=tensor([0.0449, 0.0506, 0.0601, 0.0600, 0.0482, 0.0521, 0.0533, 0.0544], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-26 13:05:21,830 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.4625, 3.0575, 0.9676, 1.6967, 1.8575, 2.2503, 2.0112, 1.0252], device='cuda:6'), covar=tensor([0.1275, 0.0966, 0.1909, 0.1421, 0.1001, 0.0929, 0.1439, 0.1623], device='cuda:6'), in_proj_covar=tensor([0.0125, 0.0271, 0.0152, 0.0132, 0.0144, 0.0165, 0.0129, 0.0134], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-26 13:05:24,857 INFO [finetune.py:1010] (6/7) 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] (6/7) Maximum memory allocated so far is 6326MB 2023-04-26 13:05:35,803 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.74 vs. limit=5.0 2023-04-26 13:05:42,225 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=11478.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:06:01,721 INFO [finetune.py:976] (6/7) Epoch 3, batch 50, loss[loss=0.2412, simple_loss=0.2753, pruned_loss=0.1036, over 4427.00 frames. ], tot_loss[loss=0.2537, simple_loss=0.3059, pruned_loss=0.1008, over 215300.95 frames. ], batch size: 19, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 13:06:11,126 INFO [optim.py:369] (6/7) 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 13:06:17,275 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=11530.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:06:18,550 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=11532.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:06:24,378 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=11541.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 13:06:34,901 INFO [finetune.py:976] (6/7) Epoch 3, batch 100, loss[loss=0.2212, simple_loss=0.2757, pruned_loss=0.08333, over 4850.00 frames. ], tot_loss[loss=0.2442, simple_loss=0.2957, pruned_loss=0.09637, over 379688.64 frames. ], batch size: 49, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 13:06:55,968 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=11589.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 13:06:58,892 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=11593.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:07:08,542 INFO [finetune.py:976] (6/7) Epoch 3, batch 150, loss[loss=0.2337, simple_loss=0.2817, pruned_loss=0.09283, over 4868.00 frames. ], tot_loss[loss=0.2393, simple_loss=0.2899, pruned_loss=0.09436, over 508411.45 frames. ], batch size: 34, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 13:07:18,015 INFO [optim.py:369] (6/7) 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 13:07:27,328 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=11635.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:07:29,847 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9396, 1.3011, 1.7559, 2.2000, 1.6180, 1.2968, 1.0518, 1.4173], device='cuda:6'), covar=tensor([0.6232, 0.7455, 0.3535, 0.6343, 0.6984, 0.5633, 0.9750, 0.7288], device='cuda:6'), in_proj_covar=tensor([0.0271, 0.0278, 0.0225, 0.0347, 0.0234, 0.0237, 0.0268, 0.0215], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-26 13:07:42,007 INFO [finetune.py:976] (6/7) Epoch 3, batch 200, loss[loss=0.2175, simple_loss=0.2757, pruned_loss=0.0796, over 4894.00 frames. ], tot_loss[loss=0.237, simple_loss=0.2879, pruned_loss=0.09309, over 607924.92 frames. ], batch size: 32, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 13:08:29,903 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=11696.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:08:31,699 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=11699.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:08:42,576 INFO [finetune.py:976] (6/7) Epoch 3, batch 250, loss[loss=0.3239, simple_loss=0.3625, pruned_loss=0.1426, over 4827.00 frames. ], tot_loss[loss=0.2432, simple_loss=0.2936, pruned_loss=0.09639, over 684721.53 frames. ], batch size: 30, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 13:09:02,115 INFO [optim.py:369] (6/7) 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 13:09:23,965 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0251, 2.5348, 1.1045, 1.2944, 2.1457, 1.2039, 3.4970, 1.7182], device='cuda:6'), covar=tensor([0.0751, 0.0817, 0.0861, 0.1321, 0.0550, 0.1109, 0.0234, 0.0675], device='cuda:6'), in_proj_covar=tensor([0.0056, 0.0073, 0.0054, 0.0051, 0.0056, 0.0057, 0.0086, 0.0054], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0008, 0.0008, 0.0008, 0.0008, 0.0011, 0.0008], device='cuda:6') 2023-04-26 13:09:28,235 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.1329, 2.6124, 0.9762, 1.2700, 2.1603, 1.2335, 3.6769, 1.6106], device='cuda:6'), covar=tensor([0.0752, 0.0759, 0.0975, 0.1535, 0.0629, 0.1197, 0.0295, 0.0830], device='cuda:6'), in_proj_covar=tensor([0.0056, 0.0074, 0.0054, 0.0051, 0.0056, 0.0057, 0.0086, 0.0054], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0008, 0.0008, 0.0008, 0.0008, 0.0011, 0.0008], device='cuda:6') 2023-04-26 13:09:36,498 INFO [finetune.py:976] (6/7) Epoch 3, batch 300, loss[loss=0.2232, simple_loss=0.2981, pruned_loss=0.07415, over 4865.00 frames. ], tot_loss[loss=0.246, simple_loss=0.2977, pruned_loss=0.09713, over 745613.12 frames. ], batch size: 34, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 13:09:38,439 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.82 vs. limit=5.0 2023-04-26 13:09:40,557 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=11760.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:09:52,921 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6398, 1.9037, 0.9934, 1.2992, 2.1636, 1.5147, 1.4660, 1.6107], device='cuda:6'), covar=tensor([0.0578, 0.0422, 0.0402, 0.0622, 0.0281, 0.0566, 0.0548, 0.0654], device='cuda:6'), in_proj_covar=tensor([0.0032, 0.0027, 0.0025, 0.0031, 0.0022, 0.0031, 0.0030, 0.0032], device='cuda:6'), out_proj_covar=tensor([0.0049, 0.0044, 0.0038, 0.0049, 0.0038, 0.0048, 0.0047, 0.0050], device='cuda:6') 2023-04-26 13:10:10,601 INFO [finetune.py:976] (6/7) Epoch 3, batch 350, loss[loss=0.1988, simple_loss=0.2667, pruned_loss=0.0655, over 4760.00 frames. ], tot_loss[loss=0.2471, simple_loss=0.2993, pruned_loss=0.09742, over 792289.32 frames. ], batch size: 28, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 13:10:18,388 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=11816.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:10:21,169 INFO [optim.py:369] (6/7) 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 13:10:28,338 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=11830.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:10:38,185 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 3, batch 400, loss[loss=0.2242, simple_loss=0.2986, pruned_loss=0.07484, over 4781.00 frames. ], tot_loss[loss=0.2464, simple_loss=0.2998, pruned_loss=0.09651, over 830828.67 frames. ], batch size: 45, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 13:11:20,870 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=11877.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:11:21,408 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=11878.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:11:33,336 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=11888.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:11:41,019 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-04-26 13:11:44,733 INFO [finetune.py:976] (6/7) Epoch 3, batch 450, loss[loss=0.2191, simple_loss=0.2854, pruned_loss=0.07646, over 4771.00 frames. ], tot_loss[loss=0.2445, simple_loss=0.2983, pruned_loss=0.09535, over 856706.74 frames. ], batch size: 28, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 13:11:45,508 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=11907.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:11:54,742 INFO [optim.py:369] (6/7) 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 13:11:54,854 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=11920.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:11:59,426 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.4890, 1.3481, 1.7192, 1.7732, 1.6828, 1.4225, 1.5355, 1.5978], device='cuda:6'), covar=tensor([2.8600, 3.8152, 4.5066, 4.7539, 2.9456, 5.2090, 4.8799, 3.7880], device='cuda:6'), in_proj_covar=tensor([0.0449, 0.0505, 0.0599, 0.0601, 0.0481, 0.0520, 0.0531, 0.0543], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-26 13:12:00,292 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 2023-04-26 13:12:13,323 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=11948.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:12:18,148 INFO [finetune.py:976] (6/7) Epoch 3, batch 500, loss[loss=0.232, simple_loss=0.2851, pruned_loss=0.08946, over 4818.00 frames. ], tot_loss[loss=0.2427, simple_loss=0.2955, pruned_loss=0.09494, over 876051.83 frames. ], batch size: 25, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 13:12:36,384 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=11981.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:12:38,143 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=11984.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:12:42,330 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=11991.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:12:52,573 INFO [finetune.py:976] (6/7) Epoch 3, batch 550, loss[loss=0.213, simple_loss=0.2763, pruned_loss=0.07484, over 4783.00 frames. ], tot_loss[loss=0.2404, simple_loss=0.2927, pruned_loss=0.09403, over 894229.46 frames. ], batch size: 28, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 13:12:55,028 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=12009.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:12:57,156 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-04-26 13:13:02,613 INFO [optim.py:369] (6/7) 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 13:13:13,230 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([0.6082, 1.6861, 1.6040, 1.2986, 1.7470, 1.5002, 2.2189, 1.4266], device='cuda:6'), covar=tensor([0.4564, 0.1875, 0.5624, 0.3267, 0.1869, 0.2366, 0.1523, 0.4739], device='cuda:6'), in_proj_covar=tensor([0.0350, 0.0354, 0.0436, 0.0371, 0.0404, 0.0378, 0.0400, 0.0417], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-26 13:13:19,334 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=12045.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:13:19,885 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0409, 1.4401, 5.3766, 5.0328, 4.6548, 5.0067, 4.7186, 4.7525], device='cuda:6'), covar=tensor([0.6026, 0.5716, 0.0853, 0.1662, 0.1049, 0.1326, 0.1024, 0.1342], device='cuda:6'), in_proj_covar=tensor([0.0327, 0.0308, 0.0438, 0.0445, 0.0373, 0.0426, 0.0337, 0.0392], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-26 13:13:30,905 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=12055.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:13:31,439 INFO [finetune.py:976] (6/7) Epoch 3, batch 600, loss[loss=0.2304, simple_loss=0.2924, pruned_loss=0.08413, over 4935.00 frames. ], tot_loss[loss=0.2411, simple_loss=0.2934, pruned_loss=0.09442, over 908238.52 frames. ], batch size: 38, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 13:14:01,454 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=12076.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:14:38,094 INFO [finetune.py:976] (6/7) Epoch 3, batch 650, loss[loss=0.2491, simple_loss=0.3051, pruned_loss=0.0965, over 4751.00 frames. ], tot_loss[loss=0.2451, simple_loss=0.2976, pruned_loss=0.0963, over 918694.07 frames. ], batch size: 28, lr: 3.99e-03, grad_scale: 32.0 2023-04-26 13:14:55,287 INFO [optim.py:369] (6/7) 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 13:15:18,803 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=12137.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:15:30,238 INFO [finetune.py:976] (6/7) Epoch 3, batch 700, loss[loss=0.2503, simple_loss=0.3137, pruned_loss=0.0935, over 4818.00 frames. ], tot_loss[loss=0.2478, simple_loss=0.3009, pruned_loss=0.09731, over 926719.90 frames. ], batch size: 39, lr: 3.98e-03, grad_scale: 32.0 2023-04-26 13:15:31,887 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-04-26 13:15:35,861 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.1289, 1.3154, 1.3324, 1.4918, 1.4761, 1.5824, 1.4431, 1.4656], device='cuda:6'), covar=tensor([1.5540, 2.7695, 2.3153, 1.9591, 2.2128, 3.7067, 2.8552, 2.3562], device='cuda:6'), in_proj_covar=tensor([0.0296, 0.0396, 0.0313, 0.0318, 0.0346, 0.0395, 0.0380, 0.0343], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-26 13:15:40,476 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=12172.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:15:52,186 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=12188.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:16:00,823 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 3, batch 750, loss[loss=0.2427, simple_loss=0.3062, pruned_loss=0.08963, over 4848.00 frames. ], tot_loss[loss=0.2476, simple_loss=0.3016, pruned_loss=0.09681, over 934960.37 frames. ], batch size: 49, lr: 3.98e-03, grad_scale: 32.0 2023-04-26 13:16:11,582 INFO [optim.py:369] (6/7) 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 13:16:23,273 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=12236.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:16:26,663 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([3.6086, 3.5239, 2.6998, 4.1539, 3.5537, 3.6327, 1.5596, 3.5216], device='cuda:6'), covar=tensor([0.1832, 0.1303, 0.3280, 0.1936, 0.2793, 0.1934, 0.5870, 0.2582], device='cuda:6'), in_proj_covar=tensor([0.0253, 0.0228, 0.0268, 0.0320, 0.0313, 0.0264, 0.0280, 0.0281], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-26 13:16:42,544 INFO [finetune.py:976] (6/7) Epoch 3, batch 800, loss[loss=0.2639, simple_loss=0.3109, pruned_loss=0.1085, over 4810.00 frames. ], tot_loss[loss=0.2464, simple_loss=0.3006, pruned_loss=0.09607, over 938826.65 frames. ], batch size: 40, lr: 3.98e-03, grad_scale: 32.0 2023-04-26 13:17:00,572 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=12276.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:17:11,636 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=12291.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:17:14,571 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6263, 2.1483, 1.7451, 1.9574, 1.5480, 1.6422, 1.9799, 1.3475], device='cuda:6'), covar=tensor([0.2840, 0.2242, 0.1792, 0.2193, 0.3380, 0.2260, 0.2302, 0.3431], device='cuda:6'), in_proj_covar=tensor([0.0319, 0.0341, 0.0249, 0.0313, 0.0318, 0.0291, 0.0280, 0.0302], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-26 13:17:16,635 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-04-26 13:17:19,308 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.0031, 1.3555, 1.2682, 1.6298, 1.4722, 1.4894, 1.3157, 2.5079], device='cuda:6'), covar=tensor([0.0697, 0.0792, 0.0829, 0.1290, 0.0675, 0.0569, 0.0768, 0.0240], device='cuda:6'), in_proj_covar=tensor([0.0041, 0.0041, 0.0042, 0.0047, 0.0042, 0.0042, 0.0041, 0.0065], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0013, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0017], device='cuda:6') 2023-04-26 13:17:19,897 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=12304.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:17:21,022 INFO [finetune.py:976] (6/7) Epoch 3, batch 850, loss[loss=0.2171, simple_loss=0.2666, pruned_loss=0.0838, over 4781.00 frames. ], tot_loss[loss=0.2452, simple_loss=0.2985, pruned_loss=0.09598, over 942721.94 frames. ], batch size: 29, lr: 3.98e-03, grad_scale: 32.0 2023-04-26 13:17:29,562 INFO [optim.py:369] (6/7) 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 13:17:43,093 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=12339.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:17:43,734 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=12340.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:17:54,292 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=12355.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:17:54,807 INFO [finetune.py:976] (6/7) Epoch 3, batch 900, loss[loss=0.2013, simple_loss=0.2605, pruned_loss=0.07109, over 4903.00 frames. ], tot_loss[loss=0.2418, simple_loss=0.2949, pruned_loss=0.09431, over 946919.81 frames. ], batch size: 35, lr: 3.98e-03, grad_scale: 32.0 2023-04-26 13:18:02,529 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.70 vs. limit=5.0 2023-04-26 13:18:26,967 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=12403.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:18:26,980 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([3.3672, 3.2508, 2.5809, 3.8786, 3.3544, 3.3730, 1.4570, 3.2397], device='cuda:6'), covar=tensor([0.1868, 0.1418, 0.2862, 0.2206, 0.3420, 0.1909, 0.6235, 0.2608], device='cuda:6'), in_proj_covar=tensor([0.0254, 0.0228, 0.0270, 0.0322, 0.0315, 0.0266, 0.0281, 0.0282], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-26 13:18:28,775 INFO [finetune.py:976] (6/7) Epoch 3, batch 950, loss[loss=0.2456, simple_loss=0.2951, pruned_loss=0.0981, over 4856.00 frames. ], tot_loss[loss=0.2398, simple_loss=0.2932, pruned_loss=0.09322, over 950292.72 frames. ], batch size: 31, lr: 3.98e-03, grad_scale: 64.0 2023-04-26 13:18:37,307 INFO [optim.py:369] (6/7) 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 13:18:50,757 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=12432.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:19:03,159 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0804, 1.6558, 2.3265, 2.3594, 1.6756, 1.3266, 1.8547, 1.0183], device='cuda:6'), covar=tensor([0.1077, 0.1479, 0.0710, 0.1344, 0.1596, 0.1986, 0.1561, 0.1893], device='cuda:6'), in_proj_covar=tensor([0.0070, 0.0079, 0.0076, 0.0072, 0.0085, 0.0096, 0.0088, 0.0080], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003], device='cuda:6') 2023-04-26 13:19:13,294 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=12446.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:19:20,768 INFO [finetune.py:976] (6/7) Epoch 3, batch 1000, loss[loss=0.2394, simple_loss=0.287, pruned_loss=0.09588, over 4784.00 frames. ], tot_loss[loss=0.2433, simple_loss=0.2958, pruned_loss=0.09534, over 951555.21 frames. ], batch size: 25, lr: 3.98e-03, grad_scale: 64.0 2023-04-26 13:19:30,610 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=12472.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:19:43,127 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9809, 1.2455, 1.2507, 1.3513, 1.3264, 1.4907, 1.3734, 1.3823], device='cuda:6'), covar=tensor([1.4794, 2.5589, 2.0811, 1.8907, 2.0961, 3.4845, 2.5538, 2.2562], device='cuda:6'), in_proj_covar=tensor([0.0295, 0.0395, 0.0313, 0.0318, 0.0346, 0.0396, 0.0380, 0.0342], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-26 13:19:56,509 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2023-04-26 13:20:03,631 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=12502.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:20:06,940 INFO [finetune.py:976] (6/7) Epoch 3, batch 1050, loss[loss=0.2525, simple_loss=0.3164, pruned_loss=0.09434, over 4907.00 frames. ], tot_loss[loss=0.245, simple_loss=0.2981, pruned_loss=0.09595, over 951888.45 frames. ], batch size: 43, lr: 3.98e-03, grad_scale: 64.0 2023-04-26 13:20:07,659 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=12507.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:20:25,977 INFO [optim.py:369] (6/7) 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 13:20:26,055 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=12520.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:20:26,385 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 2023-04-26 13:20:59,187 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.76 vs. limit=5.0 2023-04-26 13:21:02,686 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 3, batch 1100, loss[loss=0.2216, simple_loss=0.2755, pruned_loss=0.08387, over 4752.00 frames. ], tot_loss[loss=0.2454, simple_loss=0.299, pruned_loss=0.09587, over 952659.10 frames. ], batch size: 59, lr: 3.98e-03, grad_scale: 64.0 2023-04-26 13:21:25,870 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=12576.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:21:44,448 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=12604.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:21:46,056 INFO [finetune.py:976] (6/7) Epoch 3, batch 1150, loss[loss=0.2342, simple_loss=0.2953, pruned_loss=0.08654, over 4908.00 frames. ], tot_loss[loss=0.2453, simple_loss=0.2999, pruned_loss=0.09531, over 955541.48 frames. ], batch size: 38, lr: 3.98e-03, grad_scale: 64.0 2023-04-26 13:21:55,572 INFO [optim.py:369] (6/7) 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 13:21:58,110 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=12624.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:22:09,599 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.4525, 1.5357, 1.7200, 1.8564, 1.7598, 2.1310, 1.7557, 1.8247], device='cuda:6'), covar=tensor([1.0071, 1.9486, 1.6299, 1.3849, 1.8089, 2.3854, 1.7814, 1.6322], device='cuda:6'), in_proj_covar=tensor([0.0297, 0.0398, 0.0315, 0.0321, 0.0348, 0.0399, 0.0382, 0.0344], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-26 13:22:19,232 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=12640.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:22:32,488 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.19 vs. limit=5.0 2023-04-26 13:22:34,210 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=12652.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:22:36,637 INFO [finetune.py:976] (6/7) Epoch 3, batch 1200, loss[loss=0.249, simple_loss=0.2929, pruned_loss=0.1025, over 4076.00 frames. ], tot_loss[loss=0.2447, simple_loss=0.2987, pruned_loss=0.09535, over 952250.04 frames. ], batch size: 17, lr: 3.98e-03, grad_scale: 64.0 2023-04-26 13:22:50,303 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.1786, 1.6473, 1.5476, 1.8628, 1.7416, 2.0387, 1.4947, 3.6765], device='cuda:6'), covar=tensor([0.0692, 0.0765, 0.0807, 0.1259, 0.0654, 0.0517, 0.0737, 0.0150], device='cuda:6'), in_proj_covar=tensor([0.0041, 0.0041, 0.0042, 0.0047, 0.0042, 0.0042, 0.0041, 0.0065], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0013, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0017], device='cuda:6') 2023-04-26 13:22:55,199 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0356, 1.1901, 1.3480, 1.5239, 1.4724, 1.7070, 1.4450, 1.5057], device='cuda:6'), covar=tensor([1.4214, 2.2062, 1.8829, 1.6572, 1.8959, 3.0997, 2.2881, 1.8944], device='cuda:6'), in_proj_covar=tensor([0.0297, 0.0398, 0.0315, 0.0321, 0.0348, 0.0400, 0.0383, 0.0345], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-26 13:22:57,554 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=12688.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:23:09,931 INFO [finetune.py:976] (6/7) Epoch 3, batch 1250, loss[loss=0.2674, simple_loss=0.3194, pruned_loss=0.1077, over 4899.00 frames. ], tot_loss[loss=0.2435, simple_loss=0.2969, pruned_loss=0.09511, over 954950.96 frames. ], batch size: 43, lr: 3.98e-03, grad_scale: 64.0 2023-04-26 13:23:19,410 INFO [optim.py:369] (6/7) 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 13:23:25,000 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6454, 1.3517, 1.2555, 1.2341, 1.8780, 1.5396, 1.1758, 1.2399], device='cuda:6'), covar=tensor([0.1532, 0.1354, 0.1970, 0.1482, 0.0745, 0.1563, 0.2169, 0.1961], device='cuda:6'), in_proj_covar=tensor([0.0319, 0.0334, 0.0345, 0.0310, 0.0345, 0.0359, 0.0317, 0.0350], device='cuda:6'), out_proj_covar=tensor([6.9436e-05, 7.1711e-05, 7.4640e-05, 6.4968e-05, 7.3481e-05, 7.8575e-05, 6.9036e-05, 7.5566e-05], device='cuda:6') 2023-04-26 13:23:26,756 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=12732.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:23:32,266 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.5590, 1.7267, 1.6333, 2.2226, 2.4424, 2.1712, 1.9667, 1.8295], device='cuda:6'), covar=tensor([0.2167, 0.2206, 0.2305, 0.1878, 0.1636, 0.2141, 0.2830, 0.2367], device='cuda:6'), in_proj_covar=tensor([0.0319, 0.0334, 0.0345, 0.0310, 0.0346, 0.0360, 0.0318, 0.0350], device='cuda:6'), out_proj_covar=tensor([6.9525e-05, 7.1745e-05, 7.4670e-05, 6.5033e-05, 7.3559e-05, 7.8651e-05, 6.9151e-05, 7.5671e-05], device='cuda:6') 2023-04-26 13:23:42,759 INFO [finetune.py:976] (6/7) Epoch 3, batch 1300, loss[loss=0.171, simple_loss=0.2352, pruned_loss=0.05345, over 4777.00 frames. ], tot_loss[loss=0.239, simple_loss=0.2925, pruned_loss=0.09274, over 955791.55 frames. ], batch size: 28, lr: 3.98e-03, grad_scale: 64.0 2023-04-26 13:23:45,286 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.2003, 2.7433, 1.1323, 1.3605, 2.1416, 1.2686, 3.4853, 1.7663], device='cuda:6'), covar=tensor([0.0665, 0.0634, 0.0951, 0.1261, 0.0533, 0.1026, 0.0237, 0.0655], device='cuda:6'), in_proj_covar=tensor([0.0055, 0.0073, 0.0054, 0.0050, 0.0055, 0.0056, 0.0085, 0.0054], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0008, 0.0008, 0.0008, 0.0008, 0.0011, 0.0008], device='cuda:6') 2023-04-26 13:23:47,598 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=12763.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:24:09,819 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=12780.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:24:31,986 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-04-26 13:24:35,999 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=12802.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:24:39,389 INFO [finetune.py:976] (6/7) Epoch 3, batch 1350, loss[loss=0.2637, simple_loss=0.3093, pruned_loss=0.109, over 4146.00 frames. ], tot_loss[loss=0.239, simple_loss=0.2925, pruned_loss=0.09271, over 955692.82 frames. ], batch size: 65, lr: 3.98e-03, grad_scale: 64.0 2023-04-26 13:24:48,916 INFO [optim.py:369] (6/7) 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 13:24:51,949 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=12824.0, num_to_drop=1, layers_to_drop={2} 2023-04-26 13:25:06,170 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.77 vs. limit=5.0 2023-04-26 13:25:12,232 INFO [finetune.py:976] (6/7) Epoch 3, batch 1400, loss[loss=0.2634, simple_loss=0.3286, pruned_loss=0.09909, over 4904.00 frames. ], tot_loss[loss=0.2423, simple_loss=0.2962, pruned_loss=0.09419, over 954466.62 frames. ], batch size: 37, lr: 3.98e-03, grad_scale: 64.0 2023-04-26 13:25:12,910 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-04-26 13:25:15,965 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-04-26 13:25:20,017 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.3885, 0.9705, 0.2859, 1.1033, 1.0481, 1.3136, 1.1796, 1.1907], device='cuda:6'), covar=tensor([0.0578, 0.0503, 0.0569, 0.0616, 0.0366, 0.0575, 0.0605, 0.0682], device='cuda:6'), in_proj_covar=tensor([0.0032, 0.0027, 0.0025, 0.0031, 0.0022, 0.0031, 0.0030, 0.0032], device='cuda:6'), out_proj_covar=tensor([0.0048, 0.0044, 0.0038, 0.0049, 0.0038, 0.0048, 0.0047, 0.0050], device='cuda:6') 2023-04-26 13:25:56,158 INFO [finetune.py:976] (6/7) Epoch 3, batch 1450, loss[loss=0.2732, simple_loss=0.3193, pruned_loss=0.1136, over 4929.00 frames. ], tot_loss[loss=0.2442, simple_loss=0.2985, pruned_loss=0.09498, over 955035.51 frames. ], batch size: 33, lr: 3.98e-03, grad_scale: 64.0 2023-04-26 13:26:17,005 INFO [optim.py:369] (6/7) 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 13:26:48,944 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.98 vs. limit=2.0 2023-04-26 13:27:02,920 INFO [finetune.py:976] (6/7) Epoch 3, batch 1500, loss[loss=0.2485, simple_loss=0.2996, pruned_loss=0.09873, over 4712.00 frames. ], tot_loss[loss=0.246, simple_loss=0.2999, pruned_loss=0.09607, over 954834.83 frames. ], batch size: 54, lr: 3.98e-03, grad_scale: 64.0 2023-04-26 13:27:41,571 INFO [finetune.py:976] (6/7) Epoch 3, batch 1550, loss[loss=0.2525, simple_loss=0.2992, pruned_loss=0.1029, over 4847.00 frames. ], tot_loss[loss=0.2431, simple_loss=0.2978, pruned_loss=0.09423, over 954811.84 frames. ], batch size: 49, lr: 3.98e-03, grad_scale: 64.0 2023-04-26 13:28:02,730 INFO [optim.py:369] (6/7) 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 13:28:47,304 INFO [finetune.py:976] (6/7) Epoch 3, batch 1600, loss[loss=0.1871, simple_loss=0.2556, pruned_loss=0.05931, over 4932.00 frames. ], tot_loss[loss=0.2403, simple_loss=0.2945, pruned_loss=0.09306, over 956784.99 frames. ], batch size: 38, lr: 3.98e-03, grad_scale: 64.0 2023-04-26 13:28:55,171 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-04-26 13:29:22,133 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-04-26 13:29:23,820 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=13102.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:29:26,217 INFO [finetune.py:976] (6/7) Epoch 3, batch 1650, loss[loss=0.2104, simple_loss=0.2673, pruned_loss=0.0767, over 4910.00 frames. ], tot_loss[loss=0.2367, simple_loss=0.2907, pruned_loss=0.09138, over 957440.96 frames. ], batch size: 36, lr: 3.98e-03, grad_scale: 64.0 2023-04-26 13:29:31,051 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.3327, 1.5654, 1.3263, 1.5210, 1.3235, 1.2756, 1.3561, 1.1331], device='cuda:6'), covar=tensor([0.1476, 0.0965, 0.0945, 0.0979, 0.2376, 0.1114, 0.1455, 0.2062], device='cuda:6'), in_proj_covar=tensor([0.0320, 0.0341, 0.0249, 0.0312, 0.0319, 0.0292, 0.0281, 0.0304], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-26 13:29:34,728 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=13119.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 13:29:35,233 INFO [optim.py:369] (6/7) 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 13:29:44,661 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.3387, 3.5201, 0.8507, 1.7135, 1.9633, 2.3732, 2.0464, 1.0170], device='cuda:6'), covar=tensor([0.1564, 0.0858, 0.2339, 0.1544, 0.1090, 0.1261, 0.1604, 0.2104], device='cuda:6'), in_proj_covar=tensor([0.0125, 0.0271, 0.0152, 0.0132, 0.0143, 0.0166, 0.0129, 0.0134], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-26 13:29:53,534 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.3449, 1.1294, 3.7510, 3.2749, 3.3624, 3.4581, 3.5367, 3.1531], device='cuda:6'), covar=tensor([0.8746, 0.7851, 0.2096, 0.3267, 0.2221, 0.3590, 0.2657, 0.3410], device='cuda:6'), in_proj_covar=tensor([0.0329, 0.0313, 0.0441, 0.0446, 0.0376, 0.0429, 0.0339, 0.0395], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0002], device='cuda:6') 2023-04-26 13:29:55,919 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 3, batch 1700, loss[loss=0.2457, simple_loss=0.289, pruned_loss=0.1013, over 4901.00 frames. ], tot_loss[loss=0.2364, simple_loss=0.2896, pruned_loss=0.0916, over 957222.11 frames. ], batch size: 35, lr: 3.98e-03, grad_scale: 64.0 2023-04-26 13:30:00,213 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0153, 2.6061, 2.1804, 2.3866, 1.8696, 2.0806, 2.0957, 1.8374], device='cuda:6'), covar=tensor([0.2439, 0.1659, 0.1083, 0.1574, 0.3384, 0.1679, 0.2469, 0.3102], device='cuda:6'), in_proj_covar=tensor([0.0320, 0.0341, 0.0249, 0.0312, 0.0319, 0.0292, 0.0281, 0.0304], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-26 13:30:33,506 INFO [finetune.py:976] (6/7) Epoch 3, batch 1750, loss[loss=0.2452, simple_loss=0.3042, pruned_loss=0.09306, over 4818.00 frames. ], tot_loss[loss=0.2399, simple_loss=0.293, pruned_loss=0.09342, over 954838.11 frames. ], batch size: 38, lr: 3.98e-03, grad_scale: 64.0 2023-04-26 13:30:43,132 INFO [optim.py:369] (6/7) 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 13:31:07,405 INFO [finetune.py:976] (6/7) Epoch 3, batch 1800, loss[loss=0.2878, simple_loss=0.3545, pruned_loss=0.1105, over 4867.00 frames. ], tot_loss[loss=0.242, simple_loss=0.2961, pruned_loss=0.0939, over 955518.28 frames. ], batch size: 44, lr: 3.98e-03, grad_scale: 32.0 2023-04-26 13:31:20,363 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 2023-04-26 13:31:47,120 INFO [finetune.py:976] (6/7) Epoch 3, batch 1850, loss[loss=0.2653, simple_loss=0.3103, pruned_loss=0.1102, over 4924.00 frames. ], tot_loss[loss=0.2415, simple_loss=0.296, pruned_loss=0.09348, over 954397.87 frames. ], batch size: 37, lr: 3.98e-03, grad_scale: 16.0 2023-04-26 13:31:56,814 INFO [optim.py:369] (6/7) 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 13:32:30,330 INFO [finetune.py:976] (6/7) Epoch 3, batch 1900, loss[loss=0.2968, simple_loss=0.3484, pruned_loss=0.1226, over 4847.00 frames. ], tot_loss[loss=0.2422, simple_loss=0.2977, pruned_loss=0.09335, over 957377.04 frames. ], batch size: 47, lr: 3.98e-03, grad_scale: 16.0 2023-04-26 13:32:38,732 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6804, 2.2937, 1.6075, 1.4865, 1.2513, 1.3262, 1.6467, 1.2237], device='cuda:6'), covar=tensor([0.2191, 0.1860, 0.2318, 0.2722, 0.3309, 0.2499, 0.1795, 0.2762], device='cuda:6'), in_proj_covar=tensor([0.0198, 0.0225, 0.0191, 0.0216, 0.0229, 0.0195, 0.0186, 0.0206], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-26 13:33:26,629 INFO [finetune.py:976] (6/7) Epoch 3, batch 1950, loss[loss=0.2185, simple_loss=0.276, pruned_loss=0.08052, over 4827.00 frames. ], tot_loss[loss=0.2407, simple_loss=0.2956, pruned_loss=0.09293, over 958192.98 frames. ], batch size: 30, lr: 3.98e-03, grad_scale: 16.0 2023-04-26 13:33:28,625 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.1521, 1.5579, 1.3704, 1.8866, 1.6374, 2.0064, 1.4398, 3.5338], device='cuda:6'), covar=tensor([0.0701, 0.0778, 0.0855, 0.1228, 0.0687, 0.0552, 0.0785, 0.0134], device='cuda:6'), in_proj_covar=tensor([0.0040, 0.0041, 0.0042, 0.0046, 0.0042, 0.0041, 0.0041, 0.0065], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0013, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0017], device='cuda:6') 2023-04-26 13:33:39,974 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=13419.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:33:41,756 INFO [optim.py:369] (6/7) 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] (6/7) Epoch 3, batch 2000, loss[loss=0.2201, simple_loss=0.2693, pruned_loss=0.08542, over 4766.00 frames. ], tot_loss[loss=0.2389, simple_loss=0.2928, pruned_loss=0.09253, over 957318.18 frames. ], batch size: 23, lr: 3.98e-03, grad_scale: 16.0 2023-04-26 13:34:24,601 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=13467.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:34:25,247 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.1660, 1.4804, 1.2841, 1.8156, 1.5873, 1.9968, 1.3803, 3.2529], device='cuda:6'), covar=tensor([0.0742, 0.0775, 0.0872, 0.1242, 0.0681, 0.0520, 0.0782, 0.0184], device='cuda:6'), in_proj_covar=tensor([0.0040, 0.0041, 0.0042, 0.0046, 0.0042, 0.0041, 0.0041, 0.0065], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0013, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0017], device='cuda:6') 2023-04-26 13:34:30,329 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 2023-04-26 13:34:51,089 INFO [finetune.py:976] (6/7) Epoch 3, batch 2050, loss[loss=0.1976, simple_loss=0.2549, pruned_loss=0.0701, over 4814.00 frames. ], tot_loss[loss=0.2367, simple_loss=0.2902, pruned_loss=0.09162, over 958827.00 frames. ], batch size: 38, lr: 3.98e-03, grad_scale: 16.0 2023-04-26 13:34:51,860 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.2452, 1.5770, 1.4560, 1.5574, 1.4875, 1.6360, 1.6020, 1.5525], device='cuda:6'), covar=tensor([1.3473, 2.2832, 1.9345, 1.6651, 1.9448, 3.0565, 2.2681, 2.0443], device='cuda:6'), in_proj_covar=tensor([0.0297, 0.0397, 0.0313, 0.0318, 0.0346, 0.0398, 0.0381, 0.0341], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-26 13:35:01,246 INFO [optim.py:369] (6/7) 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 13:35:01,948 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9285, 1.5959, 2.0831, 2.0637, 1.6647, 1.3416, 1.8473, 1.2415], device='cuda:6'), covar=tensor([0.0800, 0.0905, 0.0657, 0.0710, 0.0891, 0.1591, 0.0732, 0.1236], device='cuda:6'), in_proj_covar=tensor([0.0071, 0.0080, 0.0077, 0.0072, 0.0085, 0.0097, 0.0088, 0.0081], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003], device='cuda:6') 2023-04-26 13:35:24,226 INFO [finetune.py:976] (6/7) Epoch 3, batch 2100, loss[loss=0.2707, simple_loss=0.3259, pruned_loss=0.1078, over 4844.00 frames. ], tot_loss[loss=0.2361, simple_loss=0.2897, pruned_loss=0.09129, over 957961.71 frames. ], batch size: 44, lr: 3.98e-03, grad_scale: 16.0 2023-04-26 13:35:42,572 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.4976, 1.5155, 1.6880, 1.7477, 1.7297, 1.4574, 1.5746, 1.5781], device='cuda:6'), covar=tensor([2.3939, 3.0125, 3.7916, 3.9830, 2.4628, 4.0170, 4.2037, 3.0261], device='cuda:6'), in_proj_covar=tensor([0.0448, 0.0499, 0.0594, 0.0602, 0.0480, 0.0515, 0.0526, 0.0539], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-26 13:35:57,670 INFO [finetune.py:976] (6/7) Epoch 3, batch 2150, loss[loss=0.2472, simple_loss=0.3053, pruned_loss=0.09459, over 4835.00 frames. ], tot_loss[loss=0.2403, simple_loss=0.2944, pruned_loss=0.09307, over 958821.28 frames. ], batch size: 47, lr: 3.98e-03, grad_scale: 16.0 2023-04-26 13:36:01,500 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-04-26 13:36:04,100 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=5.28 vs. limit=5.0 2023-04-26 13:36:07,902 INFO [optim.py:369] (6/7) 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 13:36:30,857 INFO [finetune.py:976] (6/7) Epoch 3, batch 2200, loss[loss=0.2279, simple_loss=0.2969, pruned_loss=0.07948, over 4883.00 frames. ], tot_loss[loss=0.2414, simple_loss=0.2961, pruned_loss=0.09336, over 958984.41 frames. ], batch size: 35, lr: 3.98e-03, grad_scale: 16.0 2023-04-26 13:37:16,057 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-04-26 13:37:16,584 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0646, 1.8456, 2.1294, 2.3678, 2.4409, 1.8264, 1.5148, 1.9877], device='cuda:6'), covar=tensor([0.0952, 0.1096, 0.0635, 0.0677, 0.0538, 0.1144, 0.1055, 0.0712], device='cuda:6'), in_proj_covar=tensor([0.0209, 0.0209, 0.0188, 0.0184, 0.0182, 0.0199, 0.0175, 0.0195], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-26 13:37:21,694 INFO [finetune.py:976] (6/7) Epoch 3, batch 2250, loss[loss=0.2862, simple_loss=0.3411, pruned_loss=0.1156, over 4903.00 frames. ], tot_loss[loss=0.2429, simple_loss=0.298, pruned_loss=0.09395, over 958393.25 frames. ], batch size: 36, lr: 3.98e-03, grad_scale: 16.0 2023-04-26 13:37:31,818 INFO [optim.py:369] (6/7) 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 13:37:43,774 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=13737.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:37:55,384 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-04-26 13:38:02,150 INFO [finetune.py:976] (6/7) Epoch 3, batch 2300, loss[loss=0.2467, simple_loss=0.2961, pruned_loss=0.09869, over 4828.00 frames. ], tot_loss[loss=0.2417, simple_loss=0.2971, pruned_loss=0.09316, over 956608.81 frames. ], batch size: 30, lr: 3.98e-03, grad_scale: 16.0 2023-04-26 13:38:44,312 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.3809, 3.1907, 0.9090, 1.6372, 1.6939, 2.2413, 1.9506, 1.0653], device='cuda:6'), covar=tensor([0.1672, 0.1667, 0.2506, 0.1736, 0.1348, 0.1387, 0.1617, 0.2195], device='cuda:6'), in_proj_covar=tensor([0.0125, 0.0271, 0.0151, 0.0132, 0.0143, 0.0166, 0.0129, 0.0133], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-26 13:38:57,786 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=13798.0, num_to_drop=1, layers_to_drop={3} 2023-04-26 13:39:09,164 INFO [finetune.py:976] (6/7) Epoch 3, batch 2350, loss[loss=0.1984, simple_loss=0.2589, pruned_loss=0.0689, over 4868.00 frames. ], tot_loss[loss=0.241, simple_loss=0.2953, pruned_loss=0.0934, over 956569.01 frames. ], batch size: 31, lr: 3.98e-03, grad_scale: 16.0 2023-04-26 13:39:37,981 INFO [optim.py:369] (6/7) 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 13:40:20,276 INFO [finetune.py:976] (6/7) Epoch 3, batch 2400, loss[loss=0.2303, simple_loss=0.2755, pruned_loss=0.09255, over 4918.00 frames. ], tot_loss[loss=0.237, simple_loss=0.2914, pruned_loss=0.09128, over 957748.86 frames. ], batch size: 43, lr: 3.98e-03, grad_scale: 16.0 2023-04-26 13:40:38,079 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.52 vs. limit=5.0 2023-04-26 13:40:51,164 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.4376, 1.1660, 1.6127, 1.5168, 1.2756, 1.0944, 1.2911, 0.9202], device='cuda:6'), covar=tensor([0.0807, 0.1065, 0.0636, 0.1131, 0.1088, 0.1680, 0.0858, 0.1187], device='cuda:6'), in_proj_covar=tensor([0.0070, 0.0079, 0.0076, 0.0071, 0.0084, 0.0097, 0.0088, 0.0081], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003], device='cuda:6') 2023-04-26 13:40:57,682 INFO [finetune.py:976] (6/7) Epoch 3, batch 2450, loss[loss=0.2646, simple_loss=0.3114, pruned_loss=0.1089, over 4869.00 frames. ], tot_loss[loss=0.236, simple_loss=0.2893, pruned_loss=0.09137, over 958575.06 frames. ], batch size: 34, lr: 3.98e-03, grad_scale: 16.0 2023-04-26 13:41:07,648 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([0.4602, 1.3486, 1.2755, 1.0192, 1.3338, 1.1143, 1.5753, 1.1228], device='cuda:6'), covar=tensor([0.3035, 0.1332, 0.4292, 0.2229, 0.1353, 0.1804, 0.1630, 0.3538], device='cuda:6'), in_proj_covar=tensor([0.0353, 0.0355, 0.0438, 0.0373, 0.0405, 0.0383, 0.0400, 0.0418], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-26 13:41:09,355 INFO [optim.py:369] (6/7) 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 13:41:31,084 INFO [finetune.py:976] (6/7) Epoch 3, batch 2500, loss[loss=0.2134, simple_loss=0.2736, pruned_loss=0.07655, over 4778.00 frames. ], tot_loss[loss=0.2385, simple_loss=0.2914, pruned_loss=0.09285, over 956187.97 frames. ], batch size: 28, lr: 3.98e-03, grad_scale: 16.0 2023-04-26 13:41:31,201 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.3187, 2.1751, 2.4423, 2.7805, 2.7268, 2.0946, 1.8155, 2.3142], device='cuda:6'), covar=tensor([0.1191, 0.1131, 0.0758, 0.0770, 0.0643, 0.1245, 0.1161, 0.0812], device='cuda:6'), in_proj_covar=tensor([0.0212, 0.0212, 0.0191, 0.0185, 0.0185, 0.0202, 0.0177, 0.0198], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-26 13:41:44,404 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.56 vs. limit=2.0 2023-04-26 13:42:05,882 INFO [finetune.py:976] (6/7) Epoch 3, batch 2550, loss[loss=0.2944, simple_loss=0.3217, pruned_loss=0.1335, over 4863.00 frames. ], tot_loss[loss=0.2411, simple_loss=0.2952, pruned_loss=0.09349, over 955898.88 frames. ], batch size: 31, lr: 3.98e-03, grad_scale: 16.0 2023-04-26 13:42:28,109 INFO [optim.py:369] (6/7) 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 13:42:46,670 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9410, 1.3977, 4.6862, 4.3266, 4.1253, 4.4058, 4.1367, 4.1325], device='cuda:6'), covar=tensor([0.6755, 0.5938, 0.1030, 0.1756, 0.1000, 0.1089, 0.1940, 0.1574], device='cuda:6'), in_proj_covar=tensor([0.0325, 0.0309, 0.0434, 0.0442, 0.0370, 0.0421, 0.0333, 0.0390], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-26 13:43:13,250 INFO [finetune.py:976] (6/7) Epoch 3, batch 2600, loss[loss=0.2245, simple_loss=0.288, pruned_loss=0.08044, over 4752.00 frames. ], tot_loss[loss=0.2414, simple_loss=0.2961, pruned_loss=0.09333, over 955572.13 frames. ], batch size: 54, lr: 3.98e-03, grad_scale: 16.0 2023-04-26 13:43:34,516 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0979, 2.6423, 1.0807, 1.3130, 2.0269, 1.2515, 3.3767, 1.5920], device='cuda:6'), covar=tensor([0.0674, 0.0879, 0.1019, 0.1164, 0.0503, 0.0955, 0.0203, 0.0609], device='cuda:6'), in_proj_covar=tensor([0.0055, 0.0073, 0.0054, 0.0050, 0.0055, 0.0056, 0.0085, 0.0053], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0008, 0.0008, 0.0008, 0.0008, 0.0011, 0.0008], device='cuda:6') 2023-04-26 13:43:41,627 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=14072.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:44:06,892 INFO [zipformer.py:1188] (6/7) 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] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=14103.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:44:25,128 INFO [finetune.py:976] (6/7) Epoch 3, batch 2650, loss[loss=0.2179, simple_loss=0.2759, pruned_loss=0.07998, over 4754.00 frames. ], tot_loss[loss=0.2426, simple_loss=0.2974, pruned_loss=0.09396, over 954486.64 frames. ], batch size: 27, lr: 3.98e-03, grad_scale: 16.0 2023-04-26 13:44:35,826 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6393, 1.2006, 1.4664, 1.5056, 1.3171, 1.0881, 0.5302, 1.0780], device='cuda:6'), covar=tensor([0.5282, 0.5857, 0.2840, 0.4316, 0.5241, 0.4507, 0.7785, 0.5020], device='cuda:6'), in_proj_covar=tensor([0.0270, 0.0272, 0.0224, 0.0342, 0.0231, 0.0235, 0.0262, 0.0209], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-26 13:44:40,403 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=14120.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 13:44:47,127 INFO [optim.py:369] (6/7) 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 13:45:00,934 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 3, batch 2700, loss[loss=0.3381, simple_loss=0.3625, pruned_loss=0.1569, over 4206.00 frames. ], tot_loss[loss=0.2412, simple_loss=0.2957, pruned_loss=0.09333, over 951409.49 frames. ], batch size: 65, lr: 3.98e-03, grad_scale: 16.0 2023-04-26 13:45:41,243 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=14164.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:45:54,126 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.54 vs. limit=2.0 2023-04-26 13:46:04,154 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=14181.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 13:46:25,132 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 2023-04-26 13:46:30,715 INFO [finetune.py:976] (6/7) Epoch 3, batch 2750, loss[loss=0.2003, simple_loss=0.2539, pruned_loss=0.07342, over 4774.00 frames. ], tot_loss[loss=0.2373, simple_loss=0.2914, pruned_loss=0.09158, over 954035.96 frames. ], batch size: 28, lr: 3.98e-03, grad_scale: 16.0 2023-04-26 13:46:38,093 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7222, 1.1362, 1.5132, 1.3140, 1.8632, 1.6289, 1.2724, 1.3639], device='cuda:6'), covar=tensor([0.1994, 0.2160, 0.2632, 0.1936, 0.1285, 0.1712, 0.2915, 0.2766], device='cuda:6'), in_proj_covar=tensor([0.0320, 0.0338, 0.0348, 0.0312, 0.0350, 0.0360, 0.0319, 0.0351], device='cuda:6'), out_proj_covar=tensor([6.9693e-05, 7.2548e-05, 7.5441e-05, 6.5436e-05, 7.4382e-05, 7.8707e-05, 6.9375e-05, 7.5860e-05], device='cuda:6') 2023-04-26 13:46:40,382 INFO [optim.py:369] (6/7) 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 13:46:57,084 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.4805, 3.5105, 1.0208, 1.7341, 1.9404, 2.3816, 2.0857, 0.9966], device='cuda:6'), covar=tensor([0.1430, 0.0865, 0.2067, 0.1532, 0.1119, 0.1197, 0.1475, 0.2091], device='cuda:6'), in_proj_covar=tensor([0.0123, 0.0267, 0.0149, 0.0130, 0.0141, 0.0164, 0.0127, 0.0130], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-26 13:47:03,044 INFO [finetune.py:976] (6/7) Epoch 3, batch 2800, loss[loss=0.2339, simple_loss=0.2699, pruned_loss=0.09896, over 4539.00 frames. ], tot_loss[loss=0.233, simple_loss=0.287, pruned_loss=0.08953, over 953224.85 frames. ], batch size: 19, lr: 3.98e-03, grad_scale: 16.0 2023-04-26 13:47:29,451 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.4692, 3.4809, 0.8387, 1.9110, 1.9837, 2.3503, 2.1177, 0.8893], device='cuda:6'), covar=tensor([0.1437, 0.0896, 0.2261, 0.1407, 0.1075, 0.1207, 0.1406, 0.2129], device='cuda:6'), in_proj_covar=tensor([0.0123, 0.0267, 0.0149, 0.0131, 0.0141, 0.0164, 0.0127, 0.0131], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-26 13:47:35,422 INFO [finetune.py:976] (6/7) Epoch 3, batch 2850, loss[loss=0.2608, simple_loss=0.3131, pruned_loss=0.1042, over 4928.00 frames. ], tot_loss[loss=0.235, simple_loss=0.2881, pruned_loss=0.09096, over 952601.61 frames. ], batch size: 33, lr: 3.98e-03, grad_scale: 16.0 2023-04-26 13:47:41,535 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2023-04-26 13:47:45,433 INFO [optim.py:369] (6/7) 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 13:48:08,416 INFO [finetune.py:976] (6/7) Epoch 3, batch 2900, loss[loss=0.2632, simple_loss=0.3053, pruned_loss=0.1105, over 4767.00 frames. ], tot_loss[loss=0.2402, simple_loss=0.2932, pruned_loss=0.09366, over 952254.13 frames. ], batch size: 28, lr: 3.98e-03, grad_scale: 16.0 2023-04-26 13:48:33,706 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=14393.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:48:41,653 INFO [finetune.py:976] (6/7) Epoch 3, batch 2950, loss[loss=0.2211, simple_loss=0.2686, pruned_loss=0.08681, over 4721.00 frames. ], tot_loss[loss=0.2437, simple_loss=0.2974, pruned_loss=0.09501, over 951562.64 frames. ], batch size: 23, lr: 3.98e-03, grad_scale: 16.0 2023-04-26 13:48:46,058 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=14413.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:48:51,910 INFO [optim.py:369] (6/7) 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] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=14428.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:49:04,517 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=14441.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:49:14,494 INFO [finetune.py:976] (6/7) Epoch 3, batch 3000, loss[loss=0.2074, simple_loss=0.2535, pruned_loss=0.08061, over 4742.00 frames. ], tot_loss[loss=0.2436, simple_loss=0.2978, pruned_loss=0.09469, over 952927.33 frames. ], batch size: 23, lr: 3.98e-03, grad_scale: 16.0 2023-04-26 13:49:14,494 INFO [finetune.py:1001] (6/7) Computing validation loss 2023-04-26 13:49:20,541 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6237, 2.0437, 1.6763, 1.9422, 1.6133, 1.5884, 1.7039, 1.4007], device='cuda:6'), covar=tensor([0.2175, 0.1388, 0.1227, 0.1301, 0.3824, 0.1606, 0.2041, 0.3175], device='cuda:6'), in_proj_covar=tensor([0.0319, 0.0339, 0.0248, 0.0310, 0.0322, 0.0291, 0.0280, 0.0302], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-26 13:49:25,021 INFO [finetune.py:1010] (6/7) Epoch 3, validation: loss=0.1699, simple_loss=0.2433, pruned_loss=0.04821, over 2265189.00 frames. 2023-04-26 13:49:25,022 INFO [finetune.py:1011] (6/7) Maximum memory allocated so far is 6326MB 2023-04-26 13:49:26,928 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=14459.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:49:36,031 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=14474.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:49:37,155 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=14476.0, num_to_drop=1, layers_to_drop={2} 2023-04-26 13:49:44,323 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=14487.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:50:05,567 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-04-26 13:50:13,197 INFO [finetune.py:976] (6/7) Epoch 3, batch 3050, loss[loss=0.2001, simple_loss=0.2579, pruned_loss=0.07114, over 4729.00 frames. ], tot_loss[loss=0.2447, simple_loss=0.2988, pruned_loss=0.09529, over 952281.64 frames. ], batch size: 23, lr: 3.98e-03, grad_scale: 16.0 2023-04-26 13:50:24,171 INFO [optim.py:369] (6/7) 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 13:50:31,579 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.40 vs. limit=5.0 2023-04-26 13:50:52,347 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=14548.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:51:03,213 INFO [finetune.py:976] (6/7) Epoch 3, batch 3100, loss[loss=0.2064, simple_loss=0.2704, pruned_loss=0.07122, over 4830.00 frames. ], tot_loss[loss=0.2404, simple_loss=0.295, pruned_loss=0.09287, over 954987.42 frames. ], batch size: 38, lr: 3.98e-03, grad_scale: 16.0 2023-04-26 13:52:09,676 INFO [finetune.py:976] (6/7) Epoch 3, batch 3150, loss[loss=0.2609, simple_loss=0.3035, pruned_loss=0.1092, over 4844.00 frames. ], tot_loss[loss=0.2385, simple_loss=0.2926, pruned_loss=0.09219, over 956003.79 frames. ], batch size: 44, lr: 3.98e-03, grad_scale: 16.0 2023-04-26 13:52:30,356 INFO [optim.py:369] (6/7) 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] (6/7) Epoch 3, batch 3200, loss[loss=0.2218, simple_loss=0.2872, pruned_loss=0.07822, over 4808.00 frames. ], tot_loss[loss=0.2342, simple_loss=0.288, pruned_loss=0.09022, over 957524.95 frames. ], batch size: 45, lr: 3.98e-03, grad_scale: 16.0 2023-04-26 13:53:25,510 INFO [finetune.py:976] (6/7) Epoch 3, batch 3250, loss[loss=0.2421, simple_loss=0.2997, pruned_loss=0.0922, over 4772.00 frames. ], tot_loss[loss=0.235, simple_loss=0.2885, pruned_loss=0.09079, over 954109.42 frames. ], batch size: 54, lr: 3.98e-03, grad_scale: 16.0 2023-04-26 13:53:25,623 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5586, 1.2366, 0.7118, 1.1369, 1.4365, 1.4426, 1.2727, 1.3086], device='cuda:6'), covar=tensor([0.0596, 0.0452, 0.0478, 0.0641, 0.0330, 0.0571, 0.0580, 0.0669], device='cuda:6'), in_proj_covar=tensor([0.0031, 0.0027, 0.0025, 0.0031, 0.0022, 0.0031, 0.0030, 0.0032], device='cuda:6'), out_proj_covar=tensor([0.0048, 0.0044, 0.0038, 0.0049, 0.0037, 0.0048, 0.0047, 0.0050], device='cuda:6') 2023-04-26 13:53:37,682 INFO [optim.py:369] (6/7) 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 13:53:41,983 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=14728.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:53:46,118 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-04-26 13:53:49,358 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=14740.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:53:51,799 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6760, 2.2998, 1.6154, 1.5466, 1.2325, 1.2579, 1.6895, 1.2073], device='cuda:6'), covar=tensor([0.2267, 0.1919, 0.2204, 0.2697, 0.3524, 0.2564, 0.1659, 0.2767], device='cuda:6'), in_proj_covar=tensor([0.0195, 0.0223, 0.0188, 0.0212, 0.0226, 0.0192, 0.0182, 0.0202], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-26 13:53:53,702 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-04-26 13:53:59,444 INFO [finetune.py:976] (6/7) Epoch 3, batch 3300, loss[loss=0.2187, simple_loss=0.2912, pruned_loss=0.07309, over 4749.00 frames. ], tot_loss[loss=0.2377, simple_loss=0.2918, pruned_loss=0.09179, over 953849.68 frames. ], batch size: 27, lr: 3.98e-03, grad_scale: 16.0 2023-04-26 13:54:01,388 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=14759.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:54:09,917 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=14769.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:54:14,106 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7174, 1.4948, 0.8233, 1.3545, 1.4494, 1.6092, 1.4827, 1.4972], device='cuda:6'), covar=tensor([0.0540, 0.0405, 0.0442, 0.0537, 0.0332, 0.0500, 0.0538, 0.0592], device='cuda:6'), in_proj_covar=tensor([0.0031, 0.0027, 0.0025, 0.0031, 0.0022, 0.0031, 0.0030, 0.0032], device='cuda:6'), out_proj_covar=tensor([0.0049, 0.0044, 0.0038, 0.0049, 0.0037, 0.0048, 0.0047, 0.0050], device='cuda:6') 2023-04-26 13:54:14,672 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=14776.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:54:14,699 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=14776.0, num_to_drop=1, layers_to_drop={2} 2023-04-26 13:54:30,613 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 3, batch 3350, loss[loss=0.2479, simple_loss=0.3081, pruned_loss=0.09385, over 4893.00 frames. ], tot_loss[loss=0.2385, simple_loss=0.2932, pruned_loss=0.09186, over 953883.93 frames. ], batch size: 32, lr: 3.98e-03, grad_scale: 16.0 2023-04-26 13:54:34,179 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=14807.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:54:44,577 INFO [optim.py:369] (6/7) 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 13:54:46,822 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=14824.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 13:54:48,683 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.1235, 1.5449, 1.9207, 2.4920, 1.8200, 1.4587, 1.2839, 1.7772], device='cuda:6'), covar=tensor([0.4873, 0.5598, 0.2827, 0.4364, 0.5065, 0.4641, 0.7294, 0.4586], device='cuda:6'), in_proj_covar=tensor([0.0272, 0.0273, 0.0224, 0.0343, 0.0231, 0.0235, 0.0262, 0.0209], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-26 13:54:58,946 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=14843.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:55:13,046 INFO [finetune.py:976] (6/7) Epoch 3, batch 3400, loss[loss=0.2479, simple_loss=0.3087, pruned_loss=0.09352, over 4893.00 frames. ], tot_loss[loss=0.2393, simple_loss=0.2943, pruned_loss=0.0921, over 954416.11 frames. ], batch size: 32, lr: 3.98e-03, grad_scale: 16.0 2023-04-26 13:56:08,499 INFO [finetune.py:976] (6/7) Epoch 3, batch 3450, loss[loss=0.196, simple_loss=0.2512, pruned_loss=0.07038, over 4794.00 frames. ], tot_loss[loss=0.2388, simple_loss=0.2939, pruned_loss=0.09182, over 952624.49 frames. ], batch size: 25, lr: 3.98e-03, grad_scale: 16.0 2023-04-26 13:56:11,643 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0466, 1.2612, 1.3579, 1.4758, 1.4759, 1.6038, 1.4115, 1.4287], device='cuda:6'), covar=tensor([1.3066, 1.9831, 1.7145, 1.4774, 1.7240, 2.7232, 2.1827, 1.8240], device='cuda:6'), in_proj_covar=tensor([0.0303, 0.0401, 0.0317, 0.0324, 0.0351, 0.0406, 0.0387, 0.0344], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-26 13:56:18,852 INFO [optim.py:369] (6/7) 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 13:56:28,491 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5437, 1.3049, 4.3483, 4.0427, 3.8099, 4.1141, 4.0604, 3.9148], device='cuda:6'), covar=tensor([0.6838, 0.5823, 0.0984, 0.1545, 0.1047, 0.1516, 0.1217, 0.1384], device='cuda:6'), in_proj_covar=tensor([0.0330, 0.0314, 0.0438, 0.0445, 0.0375, 0.0428, 0.0336, 0.0396], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0002], device='cuda:6') 2023-04-26 13:56:42,394 INFO [finetune.py:976] (6/7) Epoch 3, batch 3500, loss[loss=0.2496, simple_loss=0.3063, pruned_loss=0.09642, over 4729.00 frames. ], tot_loss[loss=0.2377, simple_loss=0.2921, pruned_loss=0.0916, over 953220.43 frames. ], batch size: 54, lr: 3.98e-03, grad_scale: 16.0 2023-04-26 13:56:44,333 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9442, 2.6584, 2.2463, 2.4332, 2.0293, 2.2758, 2.2844, 1.9124], device='cuda:6'), covar=tensor([0.2173, 0.1251, 0.0890, 0.1197, 0.2611, 0.1219, 0.1701, 0.2356], device='cuda:6'), in_proj_covar=tensor([0.0317, 0.0335, 0.0245, 0.0308, 0.0319, 0.0288, 0.0278, 0.0300], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-26 13:57:16,926 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0239, 1.4475, 1.8833, 2.3972, 1.7494, 1.3815, 1.1599, 1.6403], device='cuda:6'), covar=tensor([0.5152, 0.5975, 0.2963, 0.4708, 0.5294, 0.4575, 0.7558, 0.5209], device='cuda:6'), in_proj_covar=tensor([0.0271, 0.0271, 0.0223, 0.0342, 0.0230, 0.0234, 0.0261, 0.0208], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-26 13:57:38,374 INFO [finetune.py:976] (6/7) Epoch 3, batch 3550, loss[loss=0.2558, simple_loss=0.2998, pruned_loss=0.106, over 4911.00 frames. ], tot_loss[loss=0.2352, simple_loss=0.2892, pruned_loss=0.09064, over 954541.12 frames. ], batch size: 37, lr: 3.98e-03, grad_scale: 16.0 2023-04-26 13:57:41,509 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([3.9596, 3.9199, 2.8753, 4.5010, 4.0274, 3.9387, 1.8803, 3.8286], device='cuda:6'), covar=tensor([0.1662, 0.1178, 0.3110, 0.1459, 0.2826, 0.1703, 0.5448, 0.2094], device='cuda:6'), in_proj_covar=tensor([0.0255, 0.0226, 0.0267, 0.0319, 0.0312, 0.0263, 0.0280, 0.0280], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-26 13:57:42,112 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([3.5850, 3.4582, 2.8356, 4.0941, 3.3504, 3.5851, 1.8617, 3.4988], device='cuda:6'), covar=tensor([0.1549, 0.1194, 0.3804, 0.1350, 0.2593, 0.1656, 0.4488, 0.2065], device='cuda:6'), in_proj_covar=tensor([0.0255, 0.0226, 0.0267, 0.0319, 0.0312, 0.0263, 0.0280, 0.0280], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-26 13:57:54,036 INFO [optim.py:369] (6/7) 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 13:58:28,221 INFO [finetune.py:976] (6/7) Epoch 3, batch 3600, loss[loss=0.2491, simple_loss=0.2979, pruned_loss=0.1002, over 4904.00 frames. ], tot_loss[loss=0.2321, simple_loss=0.2861, pruned_loss=0.08903, over 956290.38 frames. ], batch size: 35, lr: 3.98e-03, grad_scale: 16.0 2023-04-26 13:58:36,851 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=15069.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:58:40,571 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.42 vs. limit=5.0 2023-04-26 13:58:50,385 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-04-26 13:58:56,114 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=15096.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:59:02,216 INFO [finetune.py:976] (6/7) Epoch 3, batch 3650, loss[loss=0.293, simple_loss=0.3482, pruned_loss=0.1189, over 4822.00 frames. ], tot_loss[loss=0.2332, simple_loss=0.2874, pruned_loss=0.08947, over 955834.64 frames. ], batch size: 45, lr: 3.98e-03, grad_scale: 16.0 2023-04-26 13:59:09,047 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=15117.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:59:12,567 INFO [optim.py:369] (6/7) 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 13:59:28,507 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=15143.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 13:59:36,235 INFO [finetune.py:976] (6/7) Epoch 3, batch 3700, loss[loss=0.2829, simple_loss=0.3219, pruned_loss=0.122, over 4117.00 frames. ], tot_loss[loss=0.2374, simple_loss=0.292, pruned_loss=0.09143, over 953497.99 frames. ], batch size: 65, lr: 3.98e-03, grad_scale: 16.0 2023-04-26 13:59:59,862 INFO [zipformer.py:1188] (6/7) 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] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=15195.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 14:00:10,005 INFO [finetune.py:976] (6/7) Epoch 3, batch 3750, loss[loss=0.2685, simple_loss=0.328, pruned_loss=0.1044, over 4919.00 frames. ], tot_loss[loss=0.2381, simple_loss=0.2931, pruned_loss=0.09153, over 955170.30 frames. ], batch size: 33, lr: 3.98e-03, grad_scale: 16.0 2023-04-26 14:00:21,351 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9147, 1.8365, 1.9886, 2.1816, 2.2053, 1.7428, 1.4242, 1.8705], device='cuda:6'), covar=tensor([0.1011, 0.1034, 0.0752, 0.0718, 0.0681, 0.1085, 0.1145, 0.0725], device='cuda:6'), in_proj_covar=tensor([0.0208, 0.0209, 0.0188, 0.0183, 0.0182, 0.0199, 0.0174, 0.0194], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-26 14:00:24,836 INFO [optim.py:369] (6/7) 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 14:00:42,264 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.48 vs. limit=2.0 2023-04-26 14:00:55,185 INFO [finetune.py:976] (6/7) Epoch 3, batch 3800, loss[loss=0.2184, simple_loss=0.2679, pruned_loss=0.08443, over 4745.00 frames. ], tot_loss[loss=0.238, simple_loss=0.2934, pruned_loss=0.09132, over 954672.31 frames. ], batch size: 23, lr: 3.98e-03, grad_scale: 16.0 2023-04-26 14:00:55,294 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=15256.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 14:01:02,925 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.74 vs. limit=5.0 2023-04-26 14:01:06,860 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6717, 4.2117, 0.8249, 2.2235, 2.3918, 2.5863, 2.5026, 0.9951], device='cuda:6'), covar=tensor([0.1500, 0.0947, 0.2303, 0.1417, 0.1086, 0.1323, 0.1534, 0.2223], device='cuda:6'), in_proj_covar=tensor([0.0125, 0.0270, 0.0150, 0.0132, 0.0142, 0.0166, 0.0129, 0.0132], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-26 14:01:13,042 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-04-26 14:01:16,080 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=5.18 vs. limit=5.0 2023-04-26 14:01:27,295 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.1116, 2.3455, 1.1976, 1.3829, 1.8934, 1.3912, 2.8600, 1.6437], device='cuda:6'), covar=tensor([0.0569, 0.0896, 0.0838, 0.1027, 0.0465, 0.0828, 0.0263, 0.0552], device='cuda:6'), in_proj_covar=tensor([0.0055, 0.0072, 0.0053, 0.0050, 0.0055, 0.0056, 0.0084, 0.0053], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0008, 0.0008, 0.0008, 0.0008, 0.0011, 0.0008], device='cuda:6') 2023-04-26 14:01:29,011 INFO [finetune.py:976] (6/7) Epoch 3, batch 3850, loss[loss=0.2564, simple_loss=0.3053, pruned_loss=0.1037, over 4918.00 frames. ], tot_loss[loss=0.2369, simple_loss=0.292, pruned_loss=0.09089, over 954571.05 frames. ], batch size: 46, lr: 3.98e-03, grad_scale: 32.0 2023-04-26 14:01:38,747 INFO [optim.py:369] (6/7) 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 14:02:01,733 INFO [finetune.py:976] (6/7) Epoch 3, batch 3900, loss[loss=0.2062, simple_loss=0.2673, pruned_loss=0.07258, over 4826.00 frames. ], tot_loss[loss=0.2329, simple_loss=0.2879, pruned_loss=0.08895, over 954647.85 frames. ], batch size: 33, lr: 3.98e-03, grad_scale: 32.0 2023-04-26 14:02:02,935 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.4257, 2.4223, 2.6085, 2.8068, 2.7232, 2.0697, 1.8129, 2.4093], device='cuda:6'), covar=tensor([0.1010, 0.0881, 0.0571, 0.0739, 0.0670, 0.1064, 0.1124, 0.0630], device='cuda:6'), in_proj_covar=tensor([0.0209, 0.0210, 0.0188, 0.0184, 0.0182, 0.0199, 0.0174, 0.0194], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-26 14:02:05,385 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7070, 1.6958, 1.8203, 1.9576, 2.0186, 1.5966, 1.2630, 1.8112], device='cuda:6'), covar=tensor([0.1026, 0.1117, 0.0740, 0.0750, 0.0688, 0.1123, 0.1157, 0.0671], device='cuda:6'), in_proj_covar=tensor([0.0209, 0.0210, 0.0188, 0.0184, 0.0182, 0.0199, 0.0174, 0.0194], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-26 14:02:30,465 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.6093, 1.6559, 1.8055, 2.3223, 2.6351, 2.2311, 2.0056, 1.9806], device='cuda:6'), covar=tensor([0.1847, 0.2123, 0.2306, 0.2196, 0.1351, 0.2278, 0.2935, 0.2098], device='cuda:6'), in_proj_covar=tensor([0.0321, 0.0339, 0.0352, 0.0312, 0.0351, 0.0361, 0.0319, 0.0353], device='cuda:6'), out_proj_covar=tensor([6.9667e-05, 7.2770e-05, 7.6291e-05, 6.5382e-05, 7.4641e-05, 7.8789e-05, 6.9519e-05, 7.6171e-05], device='cuda:6') 2023-04-26 14:02:48,482 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.80 vs. limit=2.0 2023-04-26 14:02:49,653 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.93 vs. limit=5.0 2023-04-26 14:02:50,618 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=15396.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 14:03:03,039 INFO [finetune.py:976] (6/7) Epoch 3, batch 3950, loss[loss=0.2007, simple_loss=0.2623, pruned_loss=0.06951, over 4851.00 frames. ], tot_loss[loss=0.2289, simple_loss=0.2839, pruned_loss=0.08693, over 953662.87 frames. ], batch size: 44, lr: 3.98e-03, grad_scale: 32.0 2023-04-26 14:03:24,417 INFO [optim.py:369] (6/7) 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 14:03:54,150 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=15444.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 14:04:08,302 INFO [finetune.py:976] (6/7) Epoch 3, batch 4000, loss[loss=0.2188, simple_loss=0.2666, pruned_loss=0.08556, over 4808.00 frames. ], tot_loss[loss=0.2273, simple_loss=0.2821, pruned_loss=0.0863, over 954963.63 frames. ], batch size: 25, lr: 3.98e-03, grad_scale: 32.0 2023-04-26 14:05:04,590 INFO [finetune.py:976] (6/7) Epoch 3, batch 4050, loss[loss=0.2539, simple_loss=0.3184, pruned_loss=0.09469, over 4916.00 frames. ], tot_loss[loss=0.231, simple_loss=0.2866, pruned_loss=0.08775, over 957599.91 frames. ], batch size: 35, lr: 3.98e-03, grad_scale: 32.0 2023-04-26 14:05:15,731 INFO [optim.py:369] (6/7) 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 14:05:33,870 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=15551.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 14:05:36,854 INFO [finetune.py:976] (6/7) Epoch 3, batch 4100, loss[loss=0.3296, simple_loss=0.3509, pruned_loss=0.1542, over 4885.00 frames. ], tot_loss[loss=0.2347, simple_loss=0.2909, pruned_loss=0.08926, over 957276.97 frames. ], batch size: 36, lr: 3.98e-03, grad_scale: 32.0 2023-04-26 14:05:51,898 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.0944, 0.6622, 0.9458, 0.6842, 1.2251, 0.9612, 0.7615, 0.9701], device='cuda:6'), covar=tensor([0.1756, 0.1771, 0.1990, 0.1772, 0.1057, 0.1652, 0.2044, 0.2032], device='cuda:6'), in_proj_covar=tensor([0.0321, 0.0338, 0.0350, 0.0313, 0.0351, 0.0361, 0.0319, 0.0353], device='cuda:6'), out_proj_covar=tensor([6.9676e-05, 7.2507e-05, 7.5960e-05, 6.5605e-05, 7.4763e-05, 7.8876e-05, 6.9530e-05, 7.6212e-05], device='cuda:6') 2023-04-26 14:06:10,675 INFO [finetune.py:976] (6/7) Epoch 3, batch 4150, loss[loss=0.2446, simple_loss=0.3158, pruned_loss=0.08666, over 4920.00 frames. ], tot_loss[loss=0.2351, simple_loss=0.2918, pruned_loss=0.08926, over 957307.88 frames. ], batch size: 42, lr: 3.98e-03, grad_scale: 32.0 2023-04-26 14:06:23,083 INFO [optim.py:369] (6/7) 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 14:06:44,250 INFO [finetune.py:976] (6/7) Epoch 3, batch 4200, loss[loss=0.2589, simple_loss=0.3127, pruned_loss=0.1025, over 4760.00 frames. ], tot_loss[loss=0.2365, simple_loss=0.2931, pruned_loss=0.0899, over 953925.77 frames. ], batch size: 27, lr: 3.98e-03, grad_scale: 32.0 2023-04-26 14:07:07,370 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5199, 1.1938, 1.5517, 1.8192, 1.6678, 1.4830, 1.5471, 1.5083], device='cuda:6'), covar=tensor([1.8245, 2.4458, 2.5734, 2.8498, 1.9907, 2.8510, 2.9339, 2.3019], device='cuda:6'), in_proj_covar=tensor([0.0445, 0.0495, 0.0585, 0.0596, 0.0477, 0.0508, 0.0520, 0.0530], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-26 14:07:17,985 INFO [finetune.py:976] (6/7) Epoch 3, batch 4250, loss[loss=0.1805, simple_loss=0.2406, pruned_loss=0.06024, over 4726.00 frames. ], tot_loss[loss=0.2339, simple_loss=0.2901, pruned_loss=0.08885, over 955457.29 frames. ], batch size: 23, lr: 3.98e-03, grad_scale: 32.0 2023-04-26 14:07:29,539 INFO [optim.py:369] (6/7) 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 14:07:50,583 INFO [finetune.py:976] (6/7) Epoch 3, batch 4300, loss[loss=0.2062, simple_loss=0.2658, pruned_loss=0.0733, over 4898.00 frames. ], tot_loss[loss=0.2313, simple_loss=0.2872, pruned_loss=0.08766, over 956734.27 frames. ], batch size: 32, lr: 3.98e-03, grad_scale: 32.0 2023-04-26 14:07:53,028 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0195, 1.1288, 1.4012, 1.5188, 1.5402, 1.7060, 1.3860, 1.4460], device='cuda:6'), covar=tensor([1.2523, 2.0479, 1.6145, 1.4467, 1.6721, 2.5884, 2.0861, 1.8202], device='cuda:6'), in_proj_covar=tensor([0.0307, 0.0405, 0.0319, 0.0325, 0.0353, 0.0409, 0.0389, 0.0346], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-26 14:08:59,710 INFO [finetune.py:976] (6/7) Epoch 3, batch 4350, loss[loss=0.1808, simple_loss=0.2375, pruned_loss=0.06201, over 4810.00 frames. ], tot_loss[loss=0.2267, simple_loss=0.2826, pruned_loss=0.08544, over 958227.60 frames. ], batch size: 25, lr: 3.98e-03, grad_scale: 32.0 2023-04-26 14:09:07,875 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8586, 1.2952, 1.5872, 1.5729, 1.4634, 1.2635, 0.5981, 1.1327], device='cuda:6'), covar=tensor([0.4717, 0.5744, 0.2706, 0.4531, 0.5207, 0.4268, 0.7213, 0.4947], device='cuda:6'), in_proj_covar=tensor([0.0271, 0.0269, 0.0221, 0.0339, 0.0227, 0.0233, 0.0257, 0.0207], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-26 14:09:09,062 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.3362, 2.6124, 1.0859, 1.3613, 2.1657, 1.2448, 3.3814, 1.7189], device='cuda:6'), covar=tensor([0.0620, 0.0579, 0.0876, 0.1489, 0.0527, 0.1074, 0.0294, 0.0677], device='cuda:6'), in_proj_covar=tensor([0.0055, 0.0072, 0.0053, 0.0050, 0.0055, 0.0055, 0.0084, 0.0053], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0008, 0.0008, 0.0008, 0.0008, 0.0011, 0.0008], device='cuda:6') 2023-04-26 14:09:21,816 INFO [optim.py:369] (6/7) 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 14:09:54,631 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.85 vs. limit=5.0 2023-04-26 14:10:01,744 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=15851.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 14:10:04,715 INFO [finetune.py:976] (6/7) Epoch 3, batch 4400, loss[loss=0.2533, simple_loss=0.313, pruned_loss=0.09686, over 4857.00 frames. ], tot_loss[loss=0.2286, simple_loss=0.284, pruned_loss=0.08667, over 956977.22 frames. ], batch size: 31, lr: 3.97e-03, grad_scale: 32.0 2023-04-26 14:10:13,521 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([0.8249, 1.8850, 1.8169, 1.4846, 2.0309, 1.5688, 2.6619, 1.6187], device='cuda:6'), covar=tensor([0.4951, 0.1902, 0.5359, 0.3919, 0.1925, 0.2806, 0.1494, 0.4920], device='cuda:6'), in_proj_covar=tensor([0.0357, 0.0359, 0.0442, 0.0377, 0.0410, 0.0386, 0.0404, 0.0422], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-26 14:10:39,612 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=15883.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 14:10:49,612 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=15899.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 14:10:53,841 INFO [finetune.py:976] (6/7) Epoch 3, batch 4450, loss[loss=0.239, simple_loss=0.3127, pruned_loss=0.08268, over 4817.00 frames. ], tot_loss[loss=0.2328, simple_loss=0.2884, pruned_loss=0.08861, over 957605.08 frames. ], batch size: 38, lr: 3.97e-03, grad_scale: 32.0 2023-04-26 14:10:54,017 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.44 vs. limit=5.0 2023-04-26 14:11:04,062 INFO [optim.py:369] (6/7) 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 14:11:20,074 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=15944.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 14:11:27,287 INFO [finetune.py:976] (6/7) Epoch 3, batch 4500, loss[loss=0.2373, simple_loss=0.3042, pruned_loss=0.08522, over 4823.00 frames. ], tot_loss[loss=0.2324, simple_loss=0.2886, pruned_loss=0.08804, over 957122.88 frames. ], batch size: 33, lr: 3.97e-03, grad_scale: 32.0 2023-04-26 14:11:27,674 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.10 vs. limit=5.0 2023-04-26 14:11:59,083 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 3, batch 4550, loss[loss=0.2691, simple_loss=0.322, pruned_loss=0.1081, over 4848.00 frames. ], tot_loss[loss=0.2336, simple_loss=0.2899, pruned_loss=0.08859, over 956684.92 frames. ], batch size: 44, lr: 3.97e-03, grad_scale: 32.0 2023-04-26 14:12:06,036 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.56 vs. limit=2.0 2023-04-26 14:12:11,704 INFO [optim.py:369] (6/7) 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 14:12:34,648 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0348, 1.7223, 4.5865, 4.2682, 4.1378, 4.3599, 4.2851, 4.1155], device='cuda:6'), covar=tensor([0.6430, 0.5691, 0.0990, 0.1534, 0.1003, 0.1739, 0.1032, 0.1268], device='cuda:6'), in_proj_covar=tensor([0.0326, 0.0313, 0.0437, 0.0440, 0.0373, 0.0425, 0.0335, 0.0392], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-26 14:12:35,166 INFO [finetune.py:976] (6/7) Epoch 3, batch 4600, loss[loss=0.1594, simple_loss=0.213, pruned_loss=0.05286, over 4125.00 frames. ], tot_loss[loss=0.2326, simple_loss=0.2892, pruned_loss=0.08798, over 956662.48 frames. ], batch size: 18, lr: 3.97e-03, grad_scale: 32.0 2023-04-26 14:12:38,926 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=16062.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 14:12:53,917 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=16085.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 14:13:08,934 INFO [finetune.py:976] (6/7) Epoch 3, batch 4650, loss[loss=0.1816, simple_loss=0.2423, pruned_loss=0.06045, over 4759.00 frames. ], tot_loss[loss=0.2308, simple_loss=0.2869, pruned_loss=0.08735, over 956956.30 frames. ], batch size: 27, lr: 3.97e-03, grad_scale: 32.0 2023-04-26 14:13:15,730 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.1381, 2.4399, 1.0375, 1.2324, 1.7484, 1.1435, 3.0795, 1.4636], device='cuda:6'), covar=tensor([0.0650, 0.0552, 0.0803, 0.1348, 0.0526, 0.1055, 0.0271, 0.0667], device='cuda:6'), in_proj_covar=tensor([0.0055, 0.0072, 0.0054, 0.0050, 0.0055, 0.0056, 0.0084, 0.0053], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0008, 0.0008, 0.0008, 0.0008, 0.0011, 0.0008], device='cuda:6') 2023-04-26 14:13:18,702 INFO [optim.py:369] (6/7) 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 14:13:46,734 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-04-26 14:13:47,583 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=16146.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 14:13:49,907 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 3, batch 4700, loss[loss=0.233, simple_loss=0.2771, pruned_loss=0.0944, over 4725.00 frames. ], tot_loss[loss=0.229, simple_loss=0.2848, pruned_loss=0.08661, over 957806.86 frames. ], batch size: 54, lr: 3.97e-03, grad_scale: 32.0 2023-04-26 14:14:01,110 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-04-26 14:14:06,915 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.28 vs. limit=5.0 2023-04-26 14:14:38,784 INFO [finetune.py:976] (6/7) Epoch 3, batch 4750, loss[loss=0.23, simple_loss=0.2929, pruned_loss=0.08352, over 4924.00 frames. ], tot_loss[loss=0.2265, simple_loss=0.2819, pruned_loss=0.08555, over 957680.54 frames. ], batch size: 38, lr: 3.97e-03, grad_scale: 32.0 2023-04-26 14:14:47,177 INFO [zipformer.py:1188] (6/7) 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] (6/7) 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 14:15:10,829 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=16239.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 14:15:35,034 INFO [finetune.py:976] (6/7) Epoch 3, batch 4800, loss[loss=0.259, simple_loss=0.3128, pruned_loss=0.1026, over 4739.00 frames. ], tot_loss[loss=0.2291, simple_loss=0.2849, pruned_loss=0.08669, over 956917.05 frames. ], batch size: 54, lr: 3.97e-03, grad_scale: 32.0 2023-04-26 14:16:42,029 INFO [finetune.py:976] (6/7) Epoch 3, batch 4850, loss[loss=0.1984, simple_loss=0.2585, pruned_loss=0.06914, over 4763.00 frames. ], tot_loss[loss=0.2322, simple_loss=0.2886, pruned_loss=0.08789, over 955826.95 frames. ], batch size: 26, lr: 3.97e-03, grad_scale: 32.0 2023-04-26 14:17:03,615 INFO [optim.py:369] (6/7) 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 14:17:31,159 INFO [finetune.py:976] (6/7) Epoch 3, batch 4900, loss[loss=0.272, simple_loss=0.335, pruned_loss=0.1046, over 4792.00 frames. ], tot_loss[loss=0.2344, simple_loss=0.2908, pruned_loss=0.08902, over 955838.12 frames. ], batch size: 51, lr: 3.97e-03, grad_scale: 32.0 2023-04-26 14:17:32,810 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=16357.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 14:17:42,345 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=16371.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 14:18:04,093 INFO [finetune.py:976] (6/7) Epoch 3, batch 4950, loss[loss=0.2409, simple_loss=0.2907, pruned_loss=0.09559, over 4777.00 frames. ], tot_loss[loss=0.2369, simple_loss=0.2933, pruned_loss=0.09024, over 954538.72 frames. ], batch size: 25, lr: 3.97e-03, grad_scale: 16.0 2023-04-26 14:18:08,208 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.1549, 1.5538, 2.0814, 2.4326, 1.9260, 1.5438, 1.2710, 1.7953], device='cuda:6'), covar=tensor([0.5068, 0.5741, 0.2527, 0.4800, 0.5440, 0.4322, 0.6774, 0.4590], device='cuda:6'), in_proj_covar=tensor([0.0271, 0.0269, 0.0223, 0.0339, 0.0227, 0.0233, 0.0257, 0.0204], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-26 14:18:16,362 INFO [optim.py:369] (6/7) 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 14:18:21,890 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=16432.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 14:18:27,289 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 3, batch 5000, loss[loss=0.2399, simple_loss=0.2885, pruned_loss=0.09569, over 4878.00 frames. ], tot_loss[loss=0.2341, simple_loss=0.2903, pruned_loss=0.08891, over 953838.45 frames. ], batch size: 32, lr: 3.97e-03, grad_scale: 16.0 2023-04-26 14:18:40,318 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.3888, 1.6900, 1.6426, 2.1731, 2.4963, 2.0665, 1.9715, 1.7585], device='cuda:6'), covar=tensor([0.2159, 0.2063, 0.2284, 0.2128, 0.1440, 0.2015, 0.2695, 0.1913], device='cuda:6'), in_proj_covar=tensor([0.0317, 0.0337, 0.0350, 0.0312, 0.0349, 0.0359, 0.0318, 0.0353], device='cuda:6'), out_proj_covar=tensor([6.8829e-05, 7.2248e-05, 7.5892e-05, 6.5267e-05, 7.4149e-05, 7.8297e-05, 6.9207e-05, 7.6136e-05], device='cuda:6') 2023-04-26 14:19:09,415 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-04-26 14:19:09,891 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=16505.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 14:19:10,440 INFO [finetune.py:976] (6/7) Epoch 3, batch 5050, loss[loss=0.1789, simple_loss=0.253, pruned_loss=0.05244, over 4759.00 frames. ], tot_loss[loss=0.2313, simple_loss=0.2871, pruned_loss=0.08774, over 954664.20 frames. ], batch size: 28, lr: 3.97e-03, grad_scale: 16.0 2023-04-26 14:19:23,710 INFO [optim.py:369] (6/7) 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 14:19:33,536 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=16539.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 14:19:43,699 INFO [finetune.py:976] (6/7) Epoch 3, batch 5100, loss[loss=0.235, simple_loss=0.2734, pruned_loss=0.09829, over 4036.00 frames. ], tot_loss[loss=0.2279, simple_loss=0.2834, pruned_loss=0.08622, over 955849.52 frames. ], batch size: 17, lr: 3.97e-03, grad_scale: 16.0 2023-04-26 14:19:57,677 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0095, 1.4966, 1.8168, 2.1985, 1.7803, 1.4627, 1.3449, 1.6105], device='cuda:6'), covar=tensor([0.3720, 0.4686, 0.2077, 0.3416, 0.4659, 0.3534, 0.6507, 0.4308], device='cuda:6'), in_proj_covar=tensor([0.0272, 0.0269, 0.0224, 0.0341, 0.0228, 0.0234, 0.0258, 0.0205], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-26 14:20:00,060 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9878, 2.4643, 2.0918, 2.2261, 1.6808, 1.9894, 2.2138, 1.6576], device='cuda:6'), covar=tensor([0.2428, 0.1356, 0.1053, 0.1653, 0.3670, 0.1494, 0.2100, 0.2855], device='cuda:6'), in_proj_covar=tensor([0.0317, 0.0334, 0.0245, 0.0308, 0.0320, 0.0287, 0.0278, 0.0300], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-26 14:20:10,610 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=16587.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 14:20:33,132 INFO [finetune.py:976] (6/7) Epoch 3, batch 5150, loss[loss=0.2315, simple_loss=0.2913, pruned_loss=0.0858, over 4746.00 frames. ], tot_loss[loss=0.2261, simple_loss=0.2818, pruned_loss=0.08526, over 955990.77 frames. ], batch size: 54, lr: 3.97e-03, grad_scale: 16.0 2023-04-26 14:20:56,492 INFO [optim.py:369] (6/7) 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 14:21:30,488 INFO [finetune.py:976] (6/7) Epoch 3, batch 5200, loss[loss=0.1666, simple_loss=0.229, pruned_loss=0.05206, over 4781.00 frames. ], tot_loss[loss=0.2299, simple_loss=0.2856, pruned_loss=0.08711, over 953129.63 frames. ], batch size: 26, lr: 3.97e-03, grad_scale: 16.0 2023-04-26 14:21:31,188 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=16657.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 14:22:02,897 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=16705.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 14:22:02,957 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9421, 2.4838, 2.0858, 2.3049, 1.7595, 1.9537, 2.0859, 1.7505], device='cuda:6'), covar=tensor([0.2403, 0.1101, 0.1001, 0.1389, 0.3187, 0.1377, 0.2009, 0.2686], device='cuda:6'), in_proj_covar=tensor([0.0319, 0.0335, 0.0247, 0.0309, 0.0322, 0.0288, 0.0279, 0.0301], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-26 14:22:03,452 INFO [finetune.py:976] (6/7) Epoch 3, batch 5250, loss[loss=0.2077, simple_loss=0.2817, pruned_loss=0.06691, over 4859.00 frames. ], tot_loss[loss=0.2313, simple_loss=0.2877, pruned_loss=0.08746, over 955509.20 frames. ], batch size: 44, lr: 3.97e-03, grad_scale: 16.0 2023-04-26 14:22:14,742 INFO [optim.py:369] (6/7) 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 14:22:18,694 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=16727.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 14:22:33,188 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=16741.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 14:22:47,372 INFO [finetune.py:976] (6/7) Epoch 3, batch 5300, loss[loss=0.2245, simple_loss=0.278, pruned_loss=0.08545, over 4887.00 frames. ], tot_loss[loss=0.2321, simple_loss=0.2891, pruned_loss=0.08749, over 957037.13 frames. ], batch size: 43, lr: 3.97e-03, grad_scale: 16.0 2023-04-26 14:22:47,528 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.2836, 1.5274, 1.5360, 1.6486, 1.5234, 1.6758, 1.6586, 1.6510], device='cuda:6'), covar=tensor([1.1274, 1.8357, 1.5411, 1.3046, 1.5821, 2.5398, 1.9746, 1.5987], device='cuda:6'), in_proj_covar=tensor([0.0308, 0.0402, 0.0321, 0.0326, 0.0352, 0.0412, 0.0389, 0.0346], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-26 14:22:49,379 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.2731, 2.5794, 1.0442, 1.3067, 1.9380, 1.2363, 3.5185, 1.6913], device='cuda:6'), covar=tensor([0.0633, 0.0662, 0.0859, 0.1308, 0.0566, 0.1041, 0.0306, 0.0671], device='cuda:6'), in_proj_covar=tensor([0.0054, 0.0071, 0.0053, 0.0050, 0.0055, 0.0055, 0.0083, 0.0053], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0008, 0.0008, 0.0008, 0.0008, 0.0011, 0.0008], device='cuda:6') 2023-04-26 14:23:10,431 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=16789.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 14:23:20,222 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=16805.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 14:23:20,768 INFO [finetune.py:976] (6/7) Epoch 3, batch 5350, loss[loss=0.2096, simple_loss=0.2451, pruned_loss=0.08707, over 4311.00 frames. ], tot_loss[loss=0.2319, simple_loss=0.2891, pruned_loss=0.08733, over 955249.10 frames. ], batch size: 18, lr: 3.97e-03, grad_scale: 16.0 2023-04-26 14:23:27,658 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.54 vs. limit=2.0 2023-04-26 14:23:31,444 INFO [optim.py:369] (6/7) 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 14:23:52,048 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=16853.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 14:23:53,794 INFO [finetune.py:976] (6/7) Epoch 3, batch 5400, loss[loss=0.2288, simple_loss=0.2768, pruned_loss=0.09037, over 4904.00 frames. ], tot_loss[loss=0.2294, simple_loss=0.2861, pruned_loss=0.08631, over 954224.63 frames. ], batch size: 32, lr: 3.97e-03, grad_scale: 16.0 2023-04-26 14:24:01,196 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=16868.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 14:24:27,336 INFO [finetune.py:976] (6/7) Epoch 3, batch 5450, loss[loss=0.1985, simple_loss=0.256, pruned_loss=0.07052, over 4917.00 frames. ], tot_loss[loss=0.2272, simple_loss=0.2834, pruned_loss=0.08553, over 955925.18 frames. ], batch size: 46, lr: 3.97e-03, grad_scale: 16.0 2023-04-26 14:24:27,439 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=16906.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 14:24:37,667 INFO [optim.py:369] (6/7) 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 14:24:41,865 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=16929.0, num_to_drop=1, layers_to_drop={3} 2023-04-26 14:25:00,559 INFO [finetune.py:976] (6/7) Epoch 3, batch 5500, loss[loss=0.2191, simple_loss=0.2619, pruned_loss=0.08811, over 4771.00 frames. ], tot_loss[loss=0.2245, simple_loss=0.2802, pruned_loss=0.08445, over 953162.50 frames. ], batch size: 26, lr: 3.97e-03, grad_scale: 16.0 2023-04-26 14:25:06,141 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.1584, 2.5565, 1.0789, 1.3658, 1.9758, 1.2093, 3.5021, 1.6629], device='cuda:6'), covar=tensor([0.0640, 0.0796, 0.0870, 0.1263, 0.0525, 0.0998, 0.0284, 0.0626], device='cuda:6'), in_proj_covar=tensor([0.0054, 0.0071, 0.0053, 0.0050, 0.0055, 0.0055, 0.0084, 0.0053], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0008, 0.0008, 0.0008, 0.0008, 0.0011, 0.0008], device='cuda:6') 2023-04-26 14:25:07,393 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=16967.0, num_to_drop=1, layers_to_drop={3} 2023-04-26 14:25:19,947 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8635, 1.2998, 1.7275, 2.0279, 1.6369, 1.2829, 0.9707, 1.4964], device='cuda:6'), covar=tensor([0.4944, 0.5785, 0.2723, 0.4555, 0.5233, 0.4296, 0.7373, 0.4246], device='cuda:6'), in_proj_covar=tensor([0.0274, 0.0271, 0.0225, 0.0341, 0.0229, 0.0235, 0.0259, 0.0205], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-26 14:25:33,766 INFO [finetune.py:976] (6/7) Epoch 3, batch 5550, loss[loss=0.2339, simple_loss=0.2874, pruned_loss=0.09022, over 4906.00 frames. ], tot_loss[loss=0.2278, simple_loss=0.2832, pruned_loss=0.08621, over 954903.28 frames. ], batch size: 37, lr: 3.97e-03, grad_scale: 16.0 2023-04-26 14:25:43,998 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.0992, 0.5904, 0.9203, 0.7243, 1.2386, 0.9565, 0.7587, 0.9354], device='cuda:6'), covar=tensor([0.1750, 0.1915, 0.2350, 0.1770, 0.1026, 0.1629, 0.2080, 0.2154], device='cuda:6'), in_proj_covar=tensor([0.0317, 0.0337, 0.0352, 0.0312, 0.0348, 0.0358, 0.0317, 0.0354], device='cuda:6'), out_proj_covar=tensor([6.8742e-05, 7.2337e-05, 7.6218e-05, 6.5410e-05, 7.3943e-05, 7.7961e-05, 6.8994e-05, 7.6593e-05], device='cuda:6') 2023-04-26 14:25:54,589 INFO [optim.py:369] (6/7) 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 14:26:02,729 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=17027.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 14:26:03,666 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-04-26 14:26:26,208 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.77 vs. limit=5.0 2023-04-26 14:26:35,600 INFO [finetune.py:976] (6/7) Epoch 3, batch 5600, loss[loss=0.2365, simple_loss=0.2887, pruned_loss=0.09218, over 4879.00 frames. ], tot_loss[loss=0.2334, simple_loss=0.2894, pruned_loss=0.08876, over 955654.11 frames. ], batch size: 32, lr: 3.97e-03, grad_scale: 16.0 2023-04-26 14:26:59,014 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=17075.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 14:27:04,989 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.87 vs. limit=5.0 2023-04-26 14:27:17,601 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.77 vs. limit=2.0 2023-04-26 14:27:38,805 INFO [finetune.py:976] (6/7) Epoch 3, batch 5650, loss[loss=0.2768, simple_loss=0.3196, pruned_loss=0.117, over 4901.00 frames. ], tot_loss[loss=0.2345, simple_loss=0.2913, pruned_loss=0.08882, over 955253.62 frames. ], batch size: 35, lr: 3.97e-03, grad_scale: 16.0 2023-04-26 14:27:55,218 INFO [optim.py:369] (6/7) 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 14:28:26,730 INFO [finetune.py:976] (6/7) Epoch 3, batch 5700, loss[loss=0.2192, simple_loss=0.249, pruned_loss=0.09466, over 4238.00 frames. ], tot_loss[loss=0.2316, simple_loss=0.2867, pruned_loss=0.08829, over 936612.85 frames. ], batch size: 18, lr: 3.97e-03, grad_scale: 16.0 2023-04-26 14:29:04,043 INFO [finetune.py:976] (6/7) Epoch 4, batch 0, loss[loss=0.2637, simple_loss=0.3168, pruned_loss=0.1053, over 4912.00 frames. ], tot_loss[loss=0.2637, simple_loss=0.3168, pruned_loss=0.1053, over 4912.00 frames. ], batch size: 33, lr: 3.97e-03, grad_scale: 16.0 2023-04-26 14:29:04,043 INFO [finetune.py:1001] (6/7) Computing validation loss 2023-04-26 14:29:11,445 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5150, 1.1349, 1.2958, 1.1822, 1.7504, 1.4040, 1.1118, 1.2894], device='cuda:6'), covar=tensor([0.2114, 0.1820, 0.2751, 0.1945, 0.1094, 0.1981, 0.2828, 0.2662], device='cuda:6'), in_proj_covar=tensor([0.0313, 0.0334, 0.0348, 0.0309, 0.0345, 0.0353, 0.0314, 0.0351], device='cuda:6'), 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:6') 2023-04-26 14:29:26,719 INFO [finetune.py:1010] (6/7) Epoch 4, validation: loss=0.1686, simple_loss=0.2415, pruned_loss=0.04788, over 2265189.00 frames. 2023-04-26 14:29:26,720 INFO [finetune.py:1011] (6/7) Maximum memory allocated so far is 6326MB 2023-04-26 14:29:31,437 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=17189.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 14:29:33,679 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-04-26 14:29:52,777 INFO [optim.py:369] (6/7) 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 14:29:53,449 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=17224.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 14:29:56,557 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7558, 1.7357, 1.8835, 2.1342, 2.2034, 1.7216, 1.4157, 1.7466], device='cuda:6'), covar=tensor([0.1079, 0.1143, 0.0838, 0.0719, 0.0671, 0.1087, 0.1162, 0.0814], device='cuda:6'), in_proj_covar=tensor([0.0209, 0.0211, 0.0188, 0.0184, 0.0183, 0.0199, 0.0174, 0.0193], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-26 14:29:58,839 INFO [finetune.py:976] (6/7) Epoch 4, batch 50, loss[loss=0.247, simple_loss=0.3072, pruned_loss=0.0934, over 4903.00 frames. ], tot_loss[loss=0.2346, simple_loss=0.2906, pruned_loss=0.0893, over 215350.17 frames. ], batch size: 37, lr: 3.97e-03, grad_scale: 16.0 2023-04-26 14:30:11,313 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=17250.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 14:30:15,023 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6917, 2.4429, 1.8116, 1.5004, 1.3047, 1.3301, 1.8527, 1.3011], device='cuda:6'), covar=tensor([0.2155, 0.1816, 0.2090, 0.2687, 0.3337, 0.2523, 0.1557, 0.2683], device='cuda:6'), in_proj_covar=tensor([0.0194, 0.0220, 0.0184, 0.0209, 0.0222, 0.0190, 0.0178, 0.0200], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-26 14:30:18,613 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=17262.0, num_to_drop=1, layers_to_drop={2} 2023-04-26 14:30:31,863 INFO [finetune.py:976] (6/7) Epoch 4, batch 100, loss[loss=0.233, simple_loss=0.2762, pruned_loss=0.09485, over 4939.00 frames. ], tot_loss[loss=0.2265, simple_loss=0.283, pruned_loss=0.08496, over 381668.05 frames. ], batch size: 38, lr: 3.97e-03, grad_scale: 16.0 2023-04-26 14:30:49,901 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.3753, 1.7310, 1.6244, 2.1759, 1.9119, 2.2279, 1.5707, 4.5772], device='cuda:6'), covar=tensor([0.0642, 0.0795, 0.0837, 0.1171, 0.0653, 0.0565, 0.0732, 0.0128], device='cuda:6'), in_proj_covar=tensor([0.0040, 0.0040, 0.0042, 0.0046, 0.0041, 0.0041, 0.0040, 0.0063], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0017], device='cuda:6') 2023-04-26 14:30:58,842 INFO [optim.py:369] (6/7) 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 14:31:02,641 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.3018, 3.1591, 1.0187, 1.4974, 2.1359, 1.3524, 4.2740, 2.0883], device='cuda:6'), covar=tensor([0.0690, 0.0691, 0.0951, 0.1357, 0.0605, 0.1126, 0.0199, 0.0664], device='cuda:6'), in_proj_covar=tensor([0.0054, 0.0071, 0.0054, 0.0050, 0.0055, 0.0055, 0.0084, 0.0053], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0008, 0.0008, 0.0008, 0.0008, 0.0011, 0.0008], device='cuda:6') 2023-04-26 14:31:03,901 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([0.9416, 1.8558, 1.6483, 1.4519, 1.8893, 1.5243, 2.1257, 1.4169], device='cuda:6'), covar=tensor([0.2973, 0.1193, 0.4106, 0.2347, 0.1325, 0.1767, 0.1832, 0.3772], device='cuda:6'), in_proj_covar=tensor([0.0351, 0.0353, 0.0437, 0.0367, 0.0404, 0.0380, 0.0397, 0.0418], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-26 14:31:04,974 INFO [finetune.py:976] (6/7) Epoch 4, batch 150, loss[loss=0.2283, simple_loss=0.2876, pruned_loss=0.08446, over 4907.00 frames. ], tot_loss[loss=0.2221, simple_loss=0.2778, pruned_loss=0.0832, over 510022.42 frames. ], batch size: 35, lr: 3.97e-03, grad_scale: 16.0 2023-04-26 14:31:24,774 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=17362.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 14:31:38,043 INFO [finetune.py:976] (6/7) Epoch 4, batch 200, loss[loss=0.1624, simple_loss=0.2329, pruned_loss=0.04594, over 4761.00 frames. ], tot_loss[loss=0.2207, simple_loss=0.2762, pruned_loss=0.08255, over 608576.44 frames. ], batch size: 26, lr: 3.97e-03, grad_scale: 16.0 2023-04-26 14:31:38,179 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=17383.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 14:32:05,047 INFO [optim.py:369] (6/7) 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] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=17423.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 14:32:11,146 INFO [finetune.py:976] (6/7) Epoch 4, batch 250, loss[loss=0.2349, simple_loss=0.305, pruned_loss=0.08242, over 4891.00 frames. ], tot_loss[loss=0.2239, simple_loss=0.2805, pruned_loss=0.08367, over 685262.78 frames. ], batch size: 35, lr: 3.97e-03, grad_scale: 16.0 2023-04-26 14:32:25,197 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=17444.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 14:33:16,611 INFO [finetune.py:976] (6/7) Epoch 4, batch 300, loss[loss=0.2821, simple_loss=0.332, pruned_loss=0.1161, over 4810.00 frames. ], tot_loss[loss=0.2271, simple_loss=0.2838, pruned_loss=0.08517, over 744911.59 frames. ], batch size: 38, lr: 3.97e-03, grad_scale: 16.0 2023-04-26 14:34:04,407 INFO [optim.py:369] (6/7) 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 14:34:10,648 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=17524.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 14:34:22,591 INFO [finetune.py:976] (6/7) Epoch 4, batch 350, loss[loss=0.2312, simple_loss=0.2915, pruned_loss=0.08549, over 4730.00 frames. ], tot_loss[loss=0.2293, simple_loss=0.2862, pruned_loss=0.08616, over 791989.74 frames. ], batch size: 59, lr: 3.97e-03, grad_scale: 16.0 2023-04-26 14:34:36,260 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=17545.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 14:34:58,570 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7415, 0.8683, 1.1821, 1.3466, 1.3661, 1.5321, 1.2188, 1.2567], device='cuda:6'), covar=tensor([0.9222, 1.4941, 1.3256, 1.1704, 1.3299, 2.0275, 1.5623, 1.3986], device='cuda:6'), in_proj_covar=tensor([0.0305, 0.0400, 0.0319, 0.0324, 0.0350, 0.0411, 0.0386, 0.0341], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-26 14:34:59,135 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=17562.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 14:35:10,346 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=17572.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 14:35:17,456 INFO [finetune.py:976] (6/7) Epoch 4, batch 400, loss[loss=0.2032, simple_loss=0.2731, pruned_loss=0.06665, over 4737.00 frames. ], tot_loss[loss=0.2309, simple_loss=0.2879, pruned_loss=0.08689, over 828548.74 frames. ], batch size: 54, lr: 3.97e-03, grad_scale: 16.0 2023-04-26 14:35:24,709 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-04-26 14:35:36,593 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5969, 1.8883, 1.5153, 1.1422, 1.2478, 1.2261, 1.4695, 1.2002], device='cuda:6'), covar=tensor([0.2127, 0.1665, 0.2078, 0.2428, 0.3103, 0.2501, 0.1523, 0.2471], device='cuda:6'), in_proj_covar=tensor([0.0195, 0.0221, 0.0185, 0.0210, 0.0222, 0.0191, 0.0179, 0.0201], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-26 14:35:37,144 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=17610.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 14:35:45,052 INFO [optim.py:369] (6/7) 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 14:35:51,212 INFO [finetune.py:976] (6/7) Epoch 4, batch 450, loss[loss=0.2048, simple_loss=0.265, pruned_loss=0.07225, over 4825.00 frames. ], tot_loss[loss=0.2287, simple_loss=0.2857, pruned_loss=0.08585, over 856901.17 frames. ], batch size: 25, lr: 3.97e-03, grad_scale: 16.0 2023-04-26 14:36:00,450 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.4145, 3.1730, 0.9280, 1.8670, 1.7789, 2.5034, 1.9443, 1.1207], device='cuda:6'), covar=tensor([0.1356, 0.0892, 0.2014, 0.1229, 0.1044, 0.0976, 0.1423, 0.2031], device='cuda:6'), in_proj_covar=tensor([0.0123, 0.0266, 0.0149, 0.0129, 0.0140, 0.0162, 0.0127, 0.0130], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-26 14:36:20,897 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5883, 3.4639, 0.8086, 2.0231, 1.9628, 2.5602, 2.0811, 1.0311], device='cuda:6'), covar=tensor([0.1327, 0.1043, 0.2268, 0.1338, 0.1034, 0.1050, 0.1430, 0.1902], device='cuda:6'), in_proj_covar=tensor([0.0123, 0.0266, 0.0149, 0.0129, 0.0141, 0.0162, 0.0127, 0.0130], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-26 14:36:25,011 INFO [finetune.py:976] (6/7) Epoch 4, batch 500, loss[loss=0.2117, simple_loss=0.2685, pruned_loss=0.07742, over 4765.00 frames. ], tot_loss[loss=0.2268, simple_loss=0.2833, pruned_loss=0.08514, over 878664.49 frames. ], batch size: 27, lr: 3.97e-03, grad_scale: 16.0 2023-04-26 14:36:35,732 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=17700.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 14:36:45,691 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-04-26 14:36:49,493 INFO [zipformer.py:1188] (6/7) 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] (6/7) 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 14:36:56,857 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=17730.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 14:36:58,591 INFO [finetune.py:976] (6/7) Epoch 4, batch 550, loss[loss=0.2314, simple_loss=0.2747, pruned_loss=0.09409, over 4909.00 frames. ], tot_loss[loss=0.2247, simple_loss=0.2808, pruned_loss=0.08432, over 896855.48 frames. ], batch size: 36, lr: 3.97e-03, grad_scale: 16.0 2023-04-26 14:37:02,831 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=17739.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 14:37:18,655 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=17761.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 14:37:43,912 INFO [finetune.py:976] (6/7) Epoch 4, batch 600, loss[loss=0.2448, simple_loss=0.282, pruned_loss=0.1038, over 3967.00 frames. ], tot_loss[loss=0.2266, simple_loss=0.283, pruned_loss=0.08515, over 910484.10 frames. ], batch size: 17, lr: 3.97e-03, grad_scale: 16.0 2023-04-26 14:37:49,346 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=17791.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 14:37:50,874 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-04-26 14:37:51,775 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0041, 1.0811, 1.2688, 1.4299, 1.3716, 1.5854, 1.3576, 1.3801], device='cuda:6'), covar=tensor([0.8998, 1.5059, 1.2513, 1.1820, 1.3202, 2.2038, 1.4504, 1.3258], device='cuda:6'), in_proj_covar=tensor([0.0307, 0.0402, 0.0320, 0.0325, 0.0351, 0.0413, 0.0387, 0.0343], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-26 14:37:58,954 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.47 vs. limit=5.0 2023-04-26 14:38:23,100 INFO [optim.py:369] (6/7) 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 14:38:34,867 INFO [finetune.py:976] (6/7) Epoch 4, batch 650, loss[loss=0.2291, simple_loss=0.2786, pruned_loss=0.08978, over 4757.00 frames. ], tot_loss[loss=0.2293, simple_loss=0.2861, pruned_loss=0.08624, over 922570.00 frames. ], batch size: 27, lr: 3.97e-03, grad_scale: 16.0 2023-04-26 14:38:52,806 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 4, batch 700, loss[loss=0.2317, simple_loss=0.2835, pruned_loss=0.08993, over 4803.00 frames. ], tot_loss[loss=0.23, simple_loss=0.2873, pruned_loss=0.08635, over 930860.82 frames. ], batch size: 51, lr: 3.97e-03, grad_scale: 16.0 2023-04-26 14:39:28,940 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=17893.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 14:39:56,950 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([0.4055, 1.3677, 1.3812, 1.0489, 1.4477, 1.2266, 1.7904, 1.2791], device='cuda:6'), covar=tensor([0.3963, 0.1848, 0.5918, 0.2765, 0.1645, 0.2168, 0.1897, 0.4987], device='cuda:6'), in_proj_covar=tensor([0.0355, 0.0357, 0.0441, 0.0370, 0.0406, 0.0381, 0.0399, 0.0422], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-26 14:40:07,832 INFO [optim.py:369] (6/7) 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 14:40:08,942 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.76 vs. limit=2.0 2023-04-26 14:40:19,659 INFO [finetune.py:976] (6/7) Epoch 4, batch 750, loss[loss=0.2522, simple_loss=0.3076, pruned_loss=0.09836, over 4807.00 frames. ], tot_loss[loss=0.2309, simple_loss=0.2884, pruned_loss=0.08671, over 937503.61 frames. ], batch size: 45, lr: 3.97e-03, grad_scale: 16.0 2023-04-26 14:40:42,524 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.70 vs. limit=2.0 2023-04-26 14:40:53,855 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=17960.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 14:41:24,410 INFO [finetune.py:976] (6/7) Epoch 4, batch 800, loss[loss=0.2154, simple_loss=0.2761, pruned_loss=0.07735, over 4836.00 frames. ], tot_loss[loss=0.2292, simple_loss=0.2868, pruned_loss=0.08574, over 939825.85 frames. ], batch size: 49, lr: 3.97e-03, grad_scale: 16.0 2023-04-26 14:41:48,958 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=18018.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 14:41:51,295 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=18021.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 14:41:52,409 INFO [optim.py:369] (6/7) 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] (6/7) Epoch 4, batch 850, loss[loss=0.2214, simple_loss=0.2737, pruned_loss=0.08452, over 4873.00 frames. ], tot_loss[loss=0.2269, simple_loss=0.2839, pruned_loss=0.08499, over 943651.34 frames. ], batch size: 34, lr: 3.97e-03, grad_scale: 16.0 2023-04-26 14:42:02,668 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=18039.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 14:42:13,540 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=18056.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 14:42:20,557 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=18066.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 14:42:32,277 INFO [finetune.py:976] (6/7) Epoch 4, batch 900, loss[loss=0.206, simple_loss=0.2604, pruned_loss=0.07584, over 4830.00 frames. ], tot_loss[loss=0.2234, simple_loss=0.2798, pruned_loss=0.0835, over 944700.59 frames. ], batch size: 33, lr: 3.97e-03, grad_scale: 16.0 2023-04-26 14:42:34,141 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=18086.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 14:42:34,729 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=18087.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 14:42:53,422 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-04-26 14:42:58,117 INFO [optim.py:369] (6/7) 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] (6/7) Epoch 4, batch 950, loss[loss=0.2227, simple_loss=0.2891, pruned_loss=0.0782, over 4892.00 frames. ], tot_loss[loss=0.2232, simple_loss=0.279, pruned_loss=0.08371, over 948763.33 frames. ], batch size: 43, lr: 3.97e-03, grad_scale: 16.0 2023-04-26 14:43:13,945 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7144, 1.5835, 1.9911, 1.9531, 1.5659, 1.2993, 1.6919, 1.1373], device='cuda:6'), covar=tensor([0.0789, 0.1068, 0.0560, 0.0929, 0.0984, 0.1659, 0.0915, 0.1114], device='cuda:6'), in_proj_covar=tensor([0.0070, 0.0078, 0.0076, 0.0071, 0.0083, 0.0097, 0.0087, 0.0080], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:6') 2023-04-26 14:43:44,170 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=18174.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 14:43:45,269 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=18175.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 14:43:53,141 INFO [finetune.py:976] (6/7) Epoch 4, batch 1000, loss[loss=0.243, simple_loss=0.2909, pruned_loss=0.09753, over 4821.00 frames. ], tot_loss[loss=0.2258, simple_loss=0.2818, pruned_loss=0.0849, over 950148.33 frames. ], batch size: 25, lr: 3.97e-03, grad_scale: 16.0 2023-04-26 14:44:03,857 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.4064, 2.4662, 2.2573, 2.3193, 2.6001, 2.0890, 3.6348, 1.9267], device='cuda:6'), covar=tensor([0.5162, 0.2425, 0.5148, 0.3536, 0.2370, 0.3216, 0.1687, 0.4940], device='cuda:6'), in_proj_covar=tensor([0.0353, 0.0357, 0.0441, 0.0369, 0.0406, 0.0382, 0.0399, 0.0421], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-26 14:44:14,202 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9099, 1.1349, 1.2530, 1.4218, 1.3820, 1.5517, 1.3239, 1.3658], device='cuda:6'), covar=tensor([1.1026, 1.6054, 1.3399, 1.1425, 1.4081, 2.1635, 1.6802, 1.3934], device='cuda:6'), in_proj_covar=tensor([0.0308, 0.0403, 0.0320, 0.0326, 0.0351, 0.0413, 0.0387, 0.0342], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-26 14:44:30,497 INFO [optim.py:369] (6/7) 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] (6/7) Epoch 4, batch 1050, loss[loss=0.1484, simple_loss=0.2247, pruned_loss=0.03604, over 4761.00 frames. ], tot_loss[loss=0.2261, simple_loss=0.2828, pruned_loss=0.08464, over 949558.62 frames. ], batch size: 27, lr: 3.97e-03, grad_scale: 16.0 2023-04-26 14:44:39,861 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=18235.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 14:44:40,462 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=18236.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 14:44:43,459 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8679, 1.9181, 2.1912, 2.2221, 1.8510, 1.5266, 2.0097, 1.0427], device='cuda:6'), covar=tensor([0.1063, 0.1090, 0.0963, 0.1386, 0.1092, 0.1534, 0.1080, 0.1519], device='cuda:6'), in_proj_covar=tensor([0.0070, 0.0079, 0.0077, 0.0071, 0.0083, 0.0098, 0.0087, 0.0080], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003], device='cuda:6') 2023-04-26 14:44:44,047 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0921, 1.8757, 2.0965, 2.5310, 2.3757, 1.8839, 1.5318, 2.0849], device='cuda:6'), covar=tensor([0.0988, 0.1265, 0.0749, 0.0663, 0.0673, 0.1113, 0.1141, 0.0737], device='cuda:6'), in_proj_covar=tensor([0.0208, 0.0209, 0.0187, 0.0182, 0.0182, 0.0199, 0.0172, 0.0192], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-26 14:44:57,334 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.8595, 2.8956, 2.2943, 3.2754, 2.9159, 2.8898, 1.1396, 2.7125], device='cuda:6'), covar=tensor([0.1915, 0.1539, 0.3174, 0.2762, 0.3402, 0.1980, 0.5783, 0.2741], device='cuda:6'), in_proj_covar=tensor([0.0252, 0.0225, 0.0263, 0.0318, 0.0311, 0.0261, 0.0277, 0.0280], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-26 14:45:12,454 INFO [finetune.py:976] (6/7) Epoch 4, batch 1100, loss[loss=0.2164, simple_loss=0.2694, pruned_loss=0.08173, over 4019.00 frames. ], tot_loss[loss=0.2274, simple_loss=0.2841, pruned_loss=0.08529, over 951781.50 frames. ], batch size: 17, lr: 3.97e-03, grad_scale: 16.0 2023-04-26 14:45:18,362 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.0346, 1.5923, 1.3645, 1.8519, 1.6590, 2.0900, 1.4616, 3.6835], device='cuda:6'), covar=tensor([0.0738, 0.0762, 0.0879, 0.1212, 0.0680, 0.0501, 0.0751, 0.0165], device='cuda:6'), in_proj_covar=tensor([0.0040, 0.0040, 0.0042, 0.0046, 0.0041, 0.0041, 0.0040, 0.0063], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0017], device='cuda:6') 2023-04-26 14:45:30,410 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-04-26 14:45:50,513 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=18316.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 14:46:00,504 INFO [optim.py:369] (6/7) 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] (6/7) Epoch 4, batch 1150, loss[loss=0.2599, simple_loss=0.2917, pruned_loss=0.1141, over 4356.00 frames. ], tot_loss[loss=0.2298, simple_loss=0.2869, pruned_loss=0.08635, over 952157.76 frames. ], batch size: 19, lr: 3.97e-03, grad_scale: 16.0 2023-04-26 14:46:23,846 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=18356.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 14:46:25,693 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6295, 1.5265, 1.9428, 1.7998, 1.4973, 1.2155, 1.6644, 1.0340], device='cuda:6'), covar=tensor([0.0815, 0.0960, 0.0528, 0.1055, 0.1192, 0.1692, 0.0799, 0.1176], device='cuda:6'), in_proj_covar=tensor([0.0070, 0.0078, 0.0077, 0.0071, 0.0083, 0.0098, 0.0087, 0.0080], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003], device='cuda:6') 2023-04-26 14:46:52,607 INFO [finetune.py:976] (6/7) Epoch 4, batch 1200, loss[loss=0.2216, simple_loss=0.2714, pruned_loss=0.08593, over 4908.00 frames. ], tot_loss[loss=0.2271, simple_loss=0.2842, pruned_loss=0.08497, over 954045.05 frames. ], batch size: 36, lr: 3.97e-03, grad_scale: 16.0 2023-04-26 14:46:54,997 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-04-26 14:46:55,531 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=18386.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 14:47:23,929 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=18404.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 14:47:36,010 INFO [optim.py:369] (6/7) 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 14:47:42,651 INFO [finetune.py:976] (6/7) Epoch 4, batch 1250, loss[loss=0.178, simple_loss=0.2424, pruned_loss=0.05686, over 4909.00 frames. ], tot_loss[loss=0.2233, simple_loss=0.281, pruned_loss=0.08286, over 956099.88 frames. ], batch size: 35, lr: 3.97e-03, grad_scale: 32.0 2023-04-26 14:47:43,824 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=18434.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 14:48:07,601 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6450, 1.6638, 1.8474, 2.0618, 1.9891, 1.5560, 1.2410, 1.7369], device='cuda:6'), covar=tensor([0.0910, 0.1108, 0.0637, 0.0595, 0.0649, 0.1043, 0.1092, 0.0626], device='cuda:6'), in_proj_covar=tensor([0.0209, 0.0210, 0.0187, 0.0183, 0.0183, 0.0200, 0.0173, 0.0193], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-26 14:48:15,942 INFO [finetune.py:976] (6/7) Epoch 4, batch 1300, loss[loss=0.1618, simple_loss=0.2401, pruned_loss=0.04173, over 4772.00 frames. ], tot_loss[loss=0.2193, simple_loss=0.277, pruned_loss=0.0808, over 957116.97 frames. ], batch size: 29, lr: 3.97e-03, grad_scale: 32.0 2023-04-26 14:48:21,734 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=18490.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 14:48:34,009 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=18508.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 14:48:36,214 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-04-26 14:48:43,001 INFO [optim.py:369] (6/7) 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] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=18530.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 14:48:48,433 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 4, batch 1350, loss[loss=0.2073, simple_loss=0.2829, pruned_loss=0.0659, over 4864.00 frames. ], tot_loss[loss=0.2216, simple_loss=0.2789, pruned_loss=0.08219, over 956709.30 frames. ], batch size: 44, lr: 3.97e-03, grad_scale: 32.0 2023-04-26 14:49:08,079 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=18551.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 14:49:32,004 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=18569.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 14:49:45,982 INFO [finetune.py:976] (6/7) Epoch 4, batch 1400, loss[loss=0.2361, simple_loss=0.296, pruned_loss=0.08812, over 4915.00 frames. ], tot_loss[loss=0.2263, simple_loss=0.2837, pruned_loss=0.08444, over 956129.24 frames. ], batch size: 36, lr: 3.97e-03, grad_scale: 32.0 2023-04-26 14:50:09,111 INFO [zipformer.py:1188] (6/7) 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] (6/7) 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 14:50:19,723 INFO [finetune.py:976] (6/7) Epoch 4, batch 1450, loss[loss=0.173, simple_loss=0.238, pruned_loss=0.05404, over 4776.00 frames. ], tot_loss[loss=0.2261, simple_loss=0.2841, pruned_loss=0.08408, over 954890.60 frames. ], batch size: 29, lr: 3.97e-03, grad_scale: 32.0 2023-04-26 14:50:40,793 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=18664.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 14:50:42,908 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.14 vs. limit=2.0 2023-04-26 14:50:52,632 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=18674.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 14:51:03,066 INFO [finetune.py:976] (6/7) Epoch 4, batch 1500, loss[loss=0.2535, simple_loss=0.29, pruned_loss=0.1085, over 4149.00 frames. ], tot_loss[loss=0.2274, simple_loss=0.2852, pruned_loss=0.08479, over 954211.63 frames. ], batch size: 65, lr: 3.96e-03, grad_scale: 32.0 2023-04-26 14:51:57,997 INFO [optim.py:369] (6/7) 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 14:52:09,749 INFO [finetune.py:976] (6/7) Epoch 4, batch 1550, loss[loss=0.2265, simple_loss=0.278, pruned_loss=0.08745, over 4694.00 frames. ], tot_loss[loss=0.2251, simple_loss=0.2833, pruned_loss=0.0834, over 954213.54 frames. ], batch size: 59, lr: 3.96e-03, grad_scale: 32.0 2023-04-26 14:52:11,101 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=18735.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 14:52:57,999 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5494, 1.3586, 4.3817, 4.0972, 3.9292, 4.1493, 4.0239, 3.8502], device='cuda:6'), covar=tensor([0.6545, 0.5618, 0.0946, 0.1496, 0.1059, 0.1518, 0.1478, 0.1289], device='cuda:6'), in_proj_covar=tensor([0.0319, 0.0306, 0.0425, 0.0428, 0.0364, 0.0415, 0.0328, 0.0386], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-26 14:53:12,328 INFO [finetune.py:976] (6/7) Epoch 4, batch 1600, loss[loss=0.2041, simple_loss=0.2646, pruned_loss=0.07184, over 4852.00 frames. ], tot_loss[loss=0.2242, simple_loss=0.2818, pruned_loss=0.08331, over 953891.67 frames. ], batch size: 44, lr: 3.96e-03, grad_scale: 32.0 2023-04-26 14:53:19,912 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7483, 1.9811, 0.9405, 1.4748, 2.1626, 1.6590, 1.5623, 1.6293], device='cuda:6'), covar=tensor([0.0546, 0.0404, 0.0376, 0.0588, 0.0249, 0.0544, 0.0536, 0.0630], device='cuda:6'), in_proj_covar=tensor([0.0031, 0.0027, 0.0025, 0.0031, 0.0022, 0.0030, 0.0030, 0.0032], device='cuda:6'), out_proj_covar=tensor([0.0049, 0.0044, 0.0038, 0.0049, 0.0038, 0.0048, 0.0047, 0.0050], device='cuda:6') 2023-04-26 14:53:44,152 INFO [optim.py:369] (6/7) 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 14:53:48,489 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=18830.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 14:53:49,110 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 4, batch 1650, loss[loss=0.2045, simple_loss=0.2564, pruned_loss=0.07629, over 4874.00 frames. ], tot_loss[loss=0.2226, simple_loss=0.2795, pruned_loss=0.08282, over 954240.59 frames. ], batch size: 31, lr: 3.96e-03, grad_scale: 32.0 2023-04-26 14:53:58,585 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=18846.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 14:54:11,509 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=18864.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 14:54:14,592 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7959, 2.1947, 1.0716, 1.5052, 2.1891, 1.7240, 1.6192, 1.7367], device='cuda:6'), covar=tensor([0.0532, 0.0392, 0.0377, 0.0575, 0.0267, 0.0582, 0.0577, 0.0623], device='cuda:6'), in_proj_covar=tensor([0.0031, 0.0027, 0.0025, 0.0031, 0.0022, 0.0030, 0.0030, 0.0031], device='cuda:6'), out_proj_covar=tensor([0.0049, 0.0044, 0.0038, 0.0049, 0.0038, 0.0048, 0.0047, 0.0050], device='cuda:6') 2023-04-26 14:54:20,473 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=18878.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 14:54:21,074 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 4, batch 1700, loss[loss=0.1692, simple_loss=0.2308, pruned_loss=0.05376, over 4792.00 frames. ], tot_loss[loss=0.2205, simple_loss=0.2775, pruned_loss=0.08175, over 956611.72 frames. ], batch size: 25, lr: 3.96e-03, grad_scale: 32.0 2023-04-26 14:54:33,510 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.65 vs. limit=2.0 2023-04-26 14:55:17,078 INFO [optim.py:369] (6/7) 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 14:55:21,955 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 4, batch 1750, loss[loss=0.2426, simple_loss=0.3106, pruned_loss=0.08731, over 4849.00 frames. ], tot_loss[loss=0.2231, simple_loss=0.2803, pruned_loss=0.08293, over 954846.42 frames. ], batch size: 47, lr: 3.96e-03, grad_scale: 32.0 2023-04-26 14:55:25,633 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.2069, 1.6127, 2.1253, 2.8113, 2.0338, 1.5678, 1.4272, 1.9725], device='cuda:6'), covar=tensor([0.4728, 0.5381, 0.2381, 0.4084, 0.5031, 0.3979, 0.6668, 0.4436], device='cuda:6'), in_proj_covar=tensor([0.0269, 0.0264, 0.0220, 0.0333, 0.0223, 0.0230, 0.0251, 0.0200], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-26 14:55:37,655 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([3.7112, 3.6068, 2.7932, 4.3151, 3.6726, 3.6871, 1.7625, 3.6922], device='cuda:6'), covar=tensor([0.1531, 0.1212, 0.3861, 0.1222, 0.2597, 0.1967, 0.5121, 0.2056], device='cuda:6'), in_proj_covar=tensor([0.0254, 0.0226, 0.0265, 0.0320, 0.0312, 0.0263, 0.0279, 0.0283], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-26 14:55:55,038 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([3.4826, 3.4618, 2.5845, 4.0395, 3.5240, 3.4279, 1.6934, 3.4454], device='cuda:6'), covar=tensor([0.1702, 0.1267, 0.3544, 0.2075, 0.2752, 0.2212, 0.5172, 0.2268], device='cuda:6'), in_proj_covar=tensor([0.0253, 0.0225, 0.0264, 0.0318, 0.0311, 0.0262, 0.0277, 0.0281], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-26 14:55:56,819 INFO [finetune.py:976] (6/7) Epoch 4, batch 1800, loss[loss=0.2214, simple_loss=0.2847, pruned_loss=0.07905, over 4792.00 frames. ], tot_loss[loss=0.2255, simple_loss=0.2829, pruned_loss=0.084, over 954615.32 frames. ], batch size: 51, lr: 3.96e-03, grad_scale: 32.0 2023-04-26 14:56:08,204 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=18991.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 14:56:44,253 INFO [optim.py:369] (6/7) 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 14:56:44,389 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5389, 1.7982, 1.7646, 2.0882, 1.9507, 2.1951, 1.6687, 3.6698], device='cuda:6'), covar=tensor([0.0630, 0.0678, 0.0729, 0.1044, 0.0570, 0.0524, 0.0688, 0.0212], device='cuda:6'), in_proj_covar=tensor([0.0040, 0.0040, 0.0042, 0.0046, 0.0041, 0.0041, 0.0040, 0.0063], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0017], device='cuda:6') 2023-04-26 14:56:46,718 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.4486, 3.9154, 0.9079, 2.0666, 2.0528, 2.5887, 2.2785, 0.8734], device='cuda:6'), covar=tensor([0.1357, 0.0858, 0.1874, 0.1264, 0.1057, 0.1034, 0.1386, 0.2193], device='cuda:6'), in_proj_covar=tensor([0.0122, 0.0265, 0.0149, 0.0129, 0.0140, 0.0161, 0.0126, 0.0129], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-26 14:56:49,158 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 4, batch 1850, loss[loss=0.2703, simple_loss=0.3157, pruned_loss=0.1124, over 4864.00 frames. ], tot_loss[loss=0.2277, simple_loss=0.2852, pruned_loss=0.0851, over 952259.29 frames. ], batch size: 31, lr: 3.96e-03, grad_scale: 32.0 2023-04-26 14:57:24,082 INFO [finetune.py:976] (6/7) Epoch 4, batch 1900, loss[loss=0.2568, simple_loss=0.3265, pruned_loss=0.09351, over 4847.00 frames. ], tot_loss[loss=0.2276, simple_loss=0.2856, pruned_loss=0.08481, over 951874.12 frames. ], batch size: 44, lr: 3.96e-03, grad_scale: 32.0 2023-04-26 14:57:50,669 INFO [optim.py:369] (6/7) 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 14:57:54,901 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=19129.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 14:57:57,177 INFO [finetune.py:976] (6/7) Epoch 4, batch 1950, loss[loss=0.2111, simple_loss=0.2642, pruned_loss=0.07899, over 4819.00 frames. ], tot_loss[loss=0.2248, simple_loss=0.2831, pruned_loss=0.0833, over 952741.30 frames. ], batch size: 25, lr: 3.96e-03, grad_scale: 32.0 2023-04-26 14:58:05,516 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=19146.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 14:58:17,046 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=19164.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 14:58:27,763 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.57 vs. limit=5.0 2023-04-26 14:58:30,351 INFO [finetune.py:976] (6/7) Epoch 4, batch 2000, loss[loss=0.2139, simple_loss=0.2647, pruned_loss=0.08153, over 4724.00 frames. ], tot_loss[loss=0.223, simple_loss=0.2807, pruned_loss=0.08262, over 954168.29 frames. ], batch size: 54, lr: 3.96e-03, grad_scale: 32.0 2023-04-26 14:58:40,466 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=19190.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 14:58:48,412 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=19194.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 14:59:11,325 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=19212.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 14:59:24,206 INFO [optim.py:369] (6/7) 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 14:59:36,725 INFO [finetune.py:976] (6/7) Epoch 4, batch 2050, loss[loss=0.2061, simple_loss=0.2632, pruned_loss=0.07451, over 4850.00 frames. ], tot_loss[loss=0.2189, simple_loss=0.2761, pruned_loss=0.08081, over 954380.00 frames. ], batch size: 44, lr: 3.96e-03, grad_scale: 32.0 2023-04-26 15:00:15,261 INFO [finetune.py:976] (6/7) Epoch 4, batch 2100, loss[loss=0.2225, simple_loss=0.2825, pruned_loss=0.08127, over 4817.00 frames. ], tot_loss[loss=0.219, simple_loss=0.2758, pruned_loss=0.08107, over 954165.29 frames. ], batch size: 38, lr: 3.96e-03, grad_scale: 32.0 2023-04-26 15:00:17,153 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=19286.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:01:01,577 INFO [optim.py:369] (6/7) 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 15:01:12,239 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=19330.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:01:14,464 INFO [finetune.py:976] (6/7) Epoch 4, batch 2150, loss[loss=0.2232, simple_loss=0.296, pruned_loss=0.0752, over 4893.00 frames. ], tot_loss[loss=0.2227, simple_loss=0.2796, pruned_loss=0.08284, over 950771.19 frames. ], batch size: 32, lr: 3.96e-03, grad_scale: 32.0 2023-04-26 15:02:07,705 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=19378.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:02:16,425 INFO [finetune.py:976] (6/7) Epoch 4, batch 2200, loss[loss=0.1876, simple_loss=0.2521, pruned_loss=0.06158, over 4768.00 frames. ], tot_loss[loss=0.2238, simple_loss=0.2819, pruned_loss=0.08289, over 951821.89 frames. ], batch size: 27, lr: 3.96e-03, grad_scale: 32.0 2023-04-26 15:02:30,495 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=19394.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:03:06,441 INFO [optim.py:369] (6/7) 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 15:03:18,529 INFO [finetune.py:976] (6/7) Epoch 4, batch 2250, loss[loss=0.2251, simple_loss=0.2853, pruned_loss=0.08245, over 4821.00 frames. ], tot_loss[loss=0.2259, simple_loss=0.284, pruned_loss=0.08385, over 953428.68 frames. ], batch size: 33, lr: 3.96e-03, grad_scale: 32.0 2023-04-26 15:03:51,633 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=19455.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:03:55,225 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.2923, 1.2070, 1.5264, 1.4498, 1.2225, 1.0089, 1.2721, 0.9282], device='cuda:6'), covar=tensor([0.0705, 0.0659, 0.0624, 0.0606, 0.0806, 0.1161, 0.0584, 0.0921], device='cuda:6'), in_proj_covar=tensor([0.0069, 0.0078, 0.0077, 0.0071, 0.0083, 0.0098, 0.0087, 0.0080], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:6') 2023-04-26 15:04:06,839 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([3.5961, 3.5512, 2.7587, 4.1463, 3.6070, 3.5799, 1.8986, 3.4935], device='cuda:6'), covar=tensor([0.1614, 0.1169, 0.3350, 0.1518, 0.2471, 0.1744, 0.4925, 0.2450], device='cuda:6'), in_proj_covar=tensor([0.0254, 0.0226, 0.0263, 0.0318, 0.0311, 0.0261, 0.0278, 0.0283], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-26 15:04:21,505 INFO [finetune.py:976] (6/7) Epoch 4, batch 2300, loss[loss=0.2601, simple_loss=0.3028, pruned_loss=0.1087, over 4730.00 frames. ], tot_loss[loss=0.225, simple_loss=0.2837, pruned_loss=0.08314, over 954500.26 frames. ], batch size: 54, lr: 3.96e-03, grad_scale: 32.0 2023-04-26 15:04:23,304 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=19485.0, num_to_drop=1, layers_to_drop={3} 2023-04-26 15:04:57,552 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.71 vs. limit=5.0 2023-04-26 15:04:58,954 INFO [optim.py:369] (6/7) 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 15:05:05,526 INFO [finetune.py:976] (6/7) Epoch 4, batch 2350, loss[loss=0.2107, simple_loss=0.2638, pruned_loss=0.0788, over 4909.00 frames. ], tot_loss[loss=0.2229, simple_loss=0.2815, pruned_loss=0.08218, over 955579.78 frames. ], batch size: 36, lr: 3.96e-03, grad_scale: 32.0 2023-04-26 15:05:39,105 INFO [finetune.py:976] (6/7) Epoch 4, batch 2400, loss[loss=0.1579, simple_loss=0.2301, pruned_loss=0.04278, over 4759.00 frames. ], tot_loss[loss=0.2214, simple_loss=0.2792, pruned_loss=0.08179, over 956456.99 frames. ], batch size: 27, lr: 3.96e-03, grad_scale: 32.0 2023-04-26 15:05:41,049 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=19586.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:05:41,253 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 2023-04-26 15:06:00,702 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([3.8939, 3.7775, 2.8830, 4.5157, 3.8988, 3.9379, 1.7832, 3.7053], device='cuda:6'), covar=tensor([0.1809, 0.1211, 0.3037, 0.1438, 0.3505, 0.1696, 0.5812, 0.2503], device='cuda:6'), in_proj_covar=tensor([0.0254, 0.0226, 0.0264, 0.0318, 0.0312, 0.0262, 0.0279, 0.0283], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-26 15:06:03,646 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=19618.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:06:06,614 INFO [optim.py:369] (6/7) 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 15:06:12,780 INFO [finetune.py:976] (6/7) Epoch 4, batch 2450, loss[loss=0.2341, simple_loss=0.2918, pruned_loss=0.0882, over 4923.00 frames. ], tot_loss[loss=0.218, simple_loss=0.2757, pruned_loss=0.08015, over 956663.16 frames. ], batch size: 38, lr: 3.96e-03, grad_scale: 32.0 2023-04-26 15:06:13,453 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=19634.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:06:17,038 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.3650, 3.2082, 0.8967, 1.7735, 1.7082, 2.2938, 1.8893, 0.9968], device='cuda:6'), covar=tensor([0.1410, 0.0972, 0.1987, 0.1396, 0.1154, 0.1084, 0.1531, 0.2006], device='cuda:6'), in_proj_covar=tensor([0.0122, 0.0264, 0.0148, 0.0129, 0.0139, 0.0161, 0.0125, 0.0129], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-26 15:06:41,816 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.1279, 1.5288, 2.0571, 2.6248, 1.8720, 1.5145, 1.2721, 1.7515], device='cuda:6'), covar=tensor([0.4771, 0.6006, 0.2607, 0.3996, 0.5231, 0.4318, 0.6517, 0.4817], device='cuda:6'), in_proj_covar=tensor([0.0272, 0.0266, 0.0221, 0.0336, 0.0224, 0.0231, 0.0251, 0.0201], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-26 15:06:44,219 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 4, batch 2500, loss[loss=0.1712, simple_loss=0.2492, pruned_loss=0.04657, over 4773.00 frames. ], tot_loss[loss=0.221, simple_loss=0.2786, pruned_loss=0.08174, over 956420.63 frames. ], batch size: 27, lr: 3.96e-03, grad_scale: 32.0 2023-04-26 15:06:57,030 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 2023-04-26 15:07:31,980 INFO [optim.py:369] (6/7) 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 15:07:41,038 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5898, 2.1945, 1.6156, 1.5986, 1.2336, 1.2898, 1.6777, 1.1819], device='cuda:6'), covar=tensor([0.2083, 0.1824, 0.1967, 0.2327, 0.3184, 0.2183, 0.1473, 0.2494], device='cuda:6'), in_proj_covar=tensor([0.0194, 0.0222, 0.0183, 0.0211, 0.0222, 0.0189, 0.0177, 0.0200], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-26 15:07:43,942 INFO [finetune.py:976] (6/7) Epoch 4, batch 2550, loss[loss=0.235, simple_loss=0.2947, pruned_loss=0.0877, over 4885.00 frames. ], tot_loss[loss=0.2235, simple_loss=0.2823, pruned_loss=0.08237, over 956668.58 frames. ], batch size: 32, lr: 3.96e-03, grad_scale: 32.0 2023-04-26 15:08:07,138 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 4, batch 2600, loss[loss=0.2346, simple_loss=0.3028, pruned_loss=0.08319, over 4919.00 frames. ], tot_loss[loss=0.2248, simple_loss=0.2837, pruned_loss=0.08291, over 956585.67 frames. ], batch size: 38, lr: 3.96e-03, grad_scale: 32.0 2023-04-26 15:08:51,173 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.4438, 1.0626, 0.3557, 1.1224, 1.1585, 1.3482, 1.2126, 1.1831], device='cuda:6'), covar=tensor([0.0542, 0.0418, 0.0522, 0.0578, 0.0340, 0.0549, 0.0519, 0.0621], device='cuda:6'), in_proj_covar=tensor([0.0031, 0.0027, 0.0025, 0.0031, 0.0022, 0.0031, 0.0030, 0.0032], device='cuda:6'), out_proj_covar=tensor([0.0049, 0.0045, 0.0038, 0.0049, 0.0038, 0.0048, 0.0047, 0.0050], device='cuda:6') 2023-04-26 15:08:57,302 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=19785.0, num_to_drop=1, layers_to_drop={2} 2023-04-26 15:09:06,407 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.79 vs. limit=2.0 2023-04-26 15:09:23,390 INFO [optim.py:369] (6/7) 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 15:09:29,906 INFO [finetune.py:976] (6/7) Epoch 4, batch 2650, loss[loss=0.2226, simple_loss=0.2611, pruned_loss=0.09209, over 4105.00 frames. ], tot_loss[loss=0.2264, simple_loss=0.285, pruned_loss=0.08385, over 955878.30 frames. ], batch size: 17, lr: 3.96e-03, grad_scale: 32.0 2023-04-26 15:09:29,969 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=19833.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 15:10:12,159 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.4430, 1.2405, 4.1313, 3.8524, 3.7084, 3.8836, 3.8732, 3.6822], device='cuda:6'), covar=tensor([0.6806, 0.5604, 0.1056, 0.1738, 0.1193, 0.1765, 0.1611, 0.1582], device='cuda:6'), in_proj_covar=tensor([0.0320, 0.0310, 0.0428, 0.0434, 0.0367, 0.0420, 0.0329, 0.0387], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-26 15:10:35,942 INFO [finetune.py:976] (6/7) Epoch 4, batch 2700, loss[loss=0.2684, simple_loss=0.3121, pruned_loss=0.1124, over 4169.00 frames. ], tot_loss[loss=0.2262, simple_loss=0.2843, pruned_loss=0.08402, over 952080.92 frames. ], batch size: 65, lr: 3.96e-03, grad_scale: 32.0 2023-04-26 15:11:15,313 INFO [optim.py:369] (6/7) 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 15:11:21,337 INFO [finetune.py:976] (6/7) Epoch 4, batch 2750, loss[loss=0.2389, simple_loss=0.2794, pruned_loss=0.09916, over 4052.00 frames. ], tot_loss[loss=0.2218, simple_loss=0.2798, pruned_loss=0.0819, over 952412.93 frames. ], batch size: 17, lr: 3.96e-03, grad_scale: 32.0 2023-04-26 15:11:49,851 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=19974.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:11:55,249 INFO [finetune.py:976] (6/7) Epoch 4, batch 2800, loss[loss=0.2083, simple_loss=0.2482, pruned_loss=0.08417, over 4387.00 frames. ], tot_loss[loss=0.219, simple_loss=0.2765, pruned_loss=0.08075, over 953860.34 frames. ], batch size: 19, lr: 3.96e-03, grad_scale: 32.0 2023-04-26 15:12:00,732 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.44 vs. limit=5.0 2023-04-26 15:12:21,309 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=20020.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:12:23,466 INFO [optim.py:369] (6/7) 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 15:12:30,049 INFO [finetune.py:976] (6/7) Epoch 4, batch 2850, loss[loss=0.2648, simple_loss=0.3133, pruned_loss=0.1081, over 4745.00 frames. ], tot_loss[loss=0.2172, simple_loss=0.2749, pruned_loss=0.07981, over 954559.47 frames. ], batch size: 54, lr: 3.96e-03, grad_scale: 32.0 2023-04-26 15:12:40,892 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=20050.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:13:02,631 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=20081.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:13:03,757 INFO [finetune.py:976] (6/7) Epoch 4, batch 2900, loss[loss=0.256, simple_loss=0.3155, pruned_loss=0.09828, over 4863.00 frames. ], tot_loss[loss=0.2205, simple_loss=0.2784, pruned_loss=0.0813, over 953807.95 frames. ], batch size: 44, lr: 3.96e-03, grad_scale: 32.0 2023-04-26 15:13:07,526 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=20089.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:13:13,384 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=20098.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:13:22,359 INFO [zipformer.py:1188] (6/7) 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] (6/7) 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 15:13:46,968 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([3.3080, 3.2883, 2.4754, 3.8487, 3.3404, 3.2716, 1.5186, 3.2102], device='cuda:6'), covar=tensor([0.1933, 0.1261, 0.3538, 0.2435, 0.2717, 0.2162, 0.5468, 0.2647], device='cuda:6'), in_proj_covar=tensor([0.0250, 0.0221, 0.0260, 0.0312, 0.0308, 0.0258, 0.0275, 0.0279], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-26 15:13:47,511 INFO [finetune.py:976] (6/7) Epoch 4, batch 2950, loss[loss=0.253, simple_loss=0.3246, pruned_loss=0.09069, over 4917.00 frames. ], tot_loss[loss=0.224, simple_loss=0.2825, pruned_loss=0.0828, over 952460.25 frames. ], batch size: 42, lr: 3.96e-03, grad_scale: 32.0 2023-04-26 15:14:09,928 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=20150.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:14:35,360 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=20173.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:14:42,462 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5186, 1.6291, 0.7831, 1.2189, 1.8452, 1.4417, 1.3047, 1.3778], device='cuda:6'), covar=tensor([0.0602, 0.0429, 0.0464, 0.0617, 0.0321, 0.0603, 0.0591, 0.0663], device='cuda:6'), in_proj_covar=tensor([0.0031, 0.0027, 0.0025, 0.0031, 0.0022, 0.0031, 0.0030, 0.0032], device='cuda:6'), out_proj_covar=tensor([0.0049, 0.0045, 0.0038, 0.0049, 0.0038, 0.0048, 0.0047, 0.0050], device='cuda:6') 2023-04-26 15:14:54,003 INFO [finetune.py:976] (6/7) Epoch 4, batch 3000, loss[loss=0.2128, simple_loss=0.2775, pruned_loss=0.07403, over 4924.00 frames. ], tot_loss[loss=0.2242, simple_loss=0.2827, pruned_loss=0.08282, over 953589.39 frames. ], batch size: 33, lr: 3.96e-03, grad_scale: 32.0 2023-04-26 15:14:54,003 INFO [finetune.py:1001] (6/7) Computing validation loss 2023-04-26 15:15:02,450 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.3925, 1.3102, 3.8408, 3.5352, 3.4582, 3.7131, 3.7864, 3.3978], device='cuda:6'), covar=tensor([0.7592, 0.5843, 0.1237, 0.2247, 0.1384, 0.1519, 0.0819, 0.1682], device='cuda:6'), in_proj_covar=tensor([0.0322, 0.0312, 0.0430, 0.0437, 0.0369, 0.0422, 0.0331, 0.0391], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-26 15:15:03,613 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([0.7218, 1.8123, 1.8950, 1.4480, 1.9435, 1.5921, 2.3395, 1.6838], device='cuda:6'), covar=tensor([0.3573, 0.1530, 0.4299, 0.2554, 0.1361, 0.2253, 0.1567, 0.4108], device='cuda:6'), in_proj_covar=tensor([0.0354, 0.0359, 0.0444, 0.0373, 0.0405, 0.0382, 0.0398, 0.0423], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-26 15:15:15,690 INFO [finetune.py:1010] (6/7) Epoch 4, validation: loss=0.1632, simple_loss=0.2363, pruned_loss=0.04509, over 2265189.00 frames. 2023-04-26 15:15:15,690 INFO [finetune.py:1011] (6/7) Maximum memory allocated so far is 6326MB 2023-04-26 15:15:35,616 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=20204.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:15:47,602 INFO [optim.py:369] (6/7) 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 15:16:05,214 INFO [finetune.py:976] (6/7) Epoch 4, batch 3050, loss[loss=0.1912, simple_loss=0.254, pruned_loss=0.0642, over 4767.00 frames. ], tot_loss[loss=0.2248, simple_loss=0.2841, pruned_loss=0.08272, over 955151.79 frames. ], batch size: 26, lr: 3.96e-03, grad_scale: 32.0 2023-04-26 15:16:43,924 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=20265.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:16:49,439 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 4, batch 3100, loss[loss=0.1978, simple_loss=0.26, pruned_loss=0.06778, over 4814.00 frames. ], tot_loss[loss=0.2237, simple_loss=0.2831, pruned_loss=0.08216, over 957103.29 frames. ], batch size: 38, lr: 3.96e-03, grad_scale: 32.0 2023-04-26 15:17:21,575 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=20322.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:17:22,124 INFO [optim.py:369] (6/7) 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 15:17:23,028 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 2023-04-26 15:17:28,233 INFO [finetune.py:976] (6/7) Epoch 4, batch 3150, loss[loss=0.2013, simple_loss=0.2613, pruned_loss=0.07066, over 4896.00 frames. ], tot_loss[loss=0.22, simple_loss=0.2787, pruned_loss=0.08062, over 956946.01 frames. ], batch size: 43, lr: 3.96e-03, grad_scale: 32.0 2023-04-26 15:17:36,676 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.60 vs. limit=5.0 2023-04-26 15:17:57,221 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=20376.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:18:01,416 INFO [finetune.py:976] (6/7) Epoch 4, batch 3200, loss[loss=0.1684, simple_loss=0.2245, pruned_loss=0.05611, over 4727.00 frames. ], tot_loss[loss=0.2174, simple_loss=0.2757, pruned_loss=0.07954, over 956575.56 frames. ], batch size: 23, lr: 3.96e-03, grad_scale: 32.0 2023-04-26 15:18:04,546 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=20388.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:18:25,073 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=20418.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:18:28,512 INFO [optim.py:369] (6/7) 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 15:18:34,646 INFO [finetune.py:976] (6/7) Epoch 4, batch 3250, loss[loss=0.2156, simple_loss=0.2833, pruned_loss=0.07397, over 4906.00 frames. ], tot_loss[loss=0.2189, simple_loss=0.2765, pruned_loss=0.08068, over 954457.20 frames. ], batch size: 37, lr: 3.96e-03, grad_scale: 64.0 2023-04-26 15:18:42,469 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=20445.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:18:45,903 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=20449.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:19:04,243 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=20468.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:19:09,074 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.71 vs. limit=2.0 2023-04-26 15:19:11,485 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=20479.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:19:13,810 INFO [finetune.py:976] (6/7) Epoch 4, batch 3300, loss[loss=0.3453, simple_loss=0.3818, pruned_loss=0.1544, over 4926.00 frames. ], tot_loss[loss=0.2223, simple_loss=0.2805, pruned_loss=0.08201, over 954536.55 frames. ], batch size: 42, lr: 3.96e-03, grad_scale: 32.0 2023-04-26 15:19:47,327 INFO [optim.py:369] (6/7) 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 15:19:58,665 INFO [finetune.py:976] (6/7) Epoch 4, batch 3350, loss[loss=0.2235, simple_loss=0.2857, pruned_loss=0.08064, over 4839.00 frames. ], tot_loss[loss=0.2249, simple_loss=0.2842, pruned_loss=0.08279, over 957199.07 frames. ], batch size: 47, lr: 3.96e-03, grad_scale: 16.0 2023-04-26 15:20:16,085 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6029, 2.2184, 1.6068, 1.4734, 1.2457, 1.2687, 1.6023, 1.2028], device='cuda:6'), covar=tensor([0.2017, 0.1722, 0.2070, 0.2438, 0.3200, 0.2389, 0.1542, 0.2461], device='cuda:6'), in_proj_covar=tensor([0.0194, 0.0222, 0.0182, 0.0211, 0.0221, 0.0189, 0.0176, 0.0200], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-26 15:20:39,486 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=20560.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:20:55,813 INFO [finetune.py:976] (6/7) Epoch 4, batch 3400, loss[loss=0.2112, simple_loss=0.2775, pruned_loss=0.07243, over 4747.00 frames. ], tot_loss[loss=0.2248, simple_loss=0.2838, pruned_loss=0.08289, over 955816.32 frames. ], batch size: 26, lr: 3.96e-03, grad_scale: 16.0 2023-04-26 15:21:04,549 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=20588.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:21:23,518 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.4292, 2.3880, 2.6528, 2.9528, 2.6726, 2.1484, 1.8759, 2.3499], device='cuda:6'), covar=tensor([0.1138, 0.0959, 0.0529, 0.0578, 0.0668, 0.1233, 0.1112, 0.0714], device='cuda:6'), in_proj_covar=tensor([0.0210, 0.0211, 0.0188, 0.0184, 0.0184, 0.0201, 0.0173, 0.0195], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-26 15:21:25,952 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5674, 1.5457, 0.7877, 1.2928, 1.8078, 1.4222, 1.3700, 1.3316], device='cuda:6'), covar=tensor([0.0554, 0.0417, 0.0469, 0.0611, 0.0332, 0.0583, 0.0573, 0.0681], device='cuda:6'), in_proj_covar=tensor([0.0031, 0.0027, 0.0025, 0.0031, 0.0022, 0.0031, 0.0030, 0.0032], device='cuda:6'), out_proj_covar=tensor([0.0049, 0.0044, 0.0038, 0.0049, 0.0038, 0.0048, 0.0047, 0.0050], device='cuda:6') 2023-04-26 15:21:29,471 INFO [optim.py:369] (6/7) 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 15:21:34,361 INFO [finetune.py:976] (6/7) Epoch 4, batch 3450, loss[loss=0.1494, simple_loss=0.2103, pruned_loss=0.04425, over 4029.00 frames. ], tot_loss[loss=0.2242, simple_loss=0.2833, pruned_loss=0.08257, over 956651.07 frames. ], batch size: 17, lr: 3.96e-03, grad_scale: 16.0 2023-04-26 15:21:44,654 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=20649.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:22:19,755 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 4, batch 3500, loss[loss=0.2214, simple_loss=0.2706, pruned_loss=0.08616, over 4857.00 frames. ], tot_loss[loss=0.223, simple_loss=0.2811, pruned_loss=0.0825, over 954968.72 frames. ], batch size: 44, lr: 3.96e-03, grad_scale: 16.0 2023-04-26 15:22:24,629 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5856, 1.4330, 0.6004, 1.2763, 1.5635, 1.4254, 1.3436, 1.3223], device='cuda:6'), covar=tensor([0.0644, 0.0402, 0.0454, 0.0642, 0.0296, 0.0718, 0.0688, 0.0686], device='cuda:6'), in_proj_covar=tensor([0.0031, 0.0027, 0.0025, 0.0031, 0.0022, 0.0031, 0.0030, 0.0032], device='cuda:6'), out_proj_covar=tensor([0.0049, 0.0044, 0.0038, 0.0049, 0.0038, 0.0048, 0.0047, 0.0050], device='cuda:6') 2023-04-26 15:22:55,338 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.70 vs. limit=2.0 2023-04-26 15:22:58,021 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=20724.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:22:58,401 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-04-26 15:22:58,549 INFO [optim.py:369] (6/7) 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] (6/7) Epoch 4, batch 3550, loss[loss=0.2301, simple_loss=0.2891, pruned_loss=0.08556, over 4816.00 frames. ], tot_loss[loss=0.2218, simple_loss=0.2792, pruned_loss=0.08225, over 956030.58 frames. ], batch size: 39, lr: 3.96e-03, grad_scale: 16.0 2023-04-26 15:23:15,951 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=20744.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:23:22,236 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=20745.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:23:48,532 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=20768.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:23:58,637 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=20774.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:24:01,792 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.55 vs. limit=2.0 2023-04-26 15:24:04,023 INFO [finetune.py:976] (6/7) Epoch 4, batch 3600, loss[loss=0.2557, simple_loss=0.3106, pruned_loss=0.1004, over 4911.00 frames. ], tot_loss[loss=0.2204, simple_loss=0.2773, pruned_loss=0.08174, over 956706.76 frames. ], batch size: 43, lr: 3.96e-03, grad_scale: 16.0 2023-04-26 15:24:10,253 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=20793.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:24:27,170 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5003, 2.3983, 2.1862, 2.3321, 2.6012, 2.0790, 3.4838, 2.0634], device='cuda:6'), covar=tensor([0.5437, 0.2871, 0.5407, 0.4363, 0.2661, 0.3477, 0.2268, 0.4659], device='cuda:6'), in_proj_covar=tensor([0.0354, 0.0356, 0.0441, 0.0371, 0.0404, 0.0382, 0.0398, 0.0423], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-26 15:24:31,212 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=20816.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:24:38,616 INFO [optim.py:369] (6/7) 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] (6/7) Epoch 4, batch 3650, loss[loss=0.2974, simple_loss=0.3488, pruned_loss=0.123, over 4720.00 frames. ], tot_loss[loss=0.2246, simple_loss=0.2811, pruned_loss=0.08406, over 955688.53 frames. ], batch size: 59, lr: 3.96e-03, grad_scale: 16.0 2023-04-26 15:25:00,516 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=20860.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:25:06,383 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.3534, 1.6181, 1.5292, 1.7334, 1.6097, 1.7577, 1.6941, 1.6754], device='cuda:6'), covar=tensor([0.9164, 1.5132, 1.2744, 1.1210, 1.3650, 2.1688, 1.5823, 1.3201], device='cuda:6'), in_proj_covar=tensor([0.0311, 0.0400, 0.0321, 0.0327, 0.0350, 0.0413, 0.0386, 0.0340], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-26 15:25:16,799 INFO [finetune.py:976] (6/7) Epoch 4, batch 3700, loss[loss=0.2177, simple_loss=0.2829, pruned_loss=0.07621, over 4834.00 frames. ], tot_loss[loss=0.2244, simple_loss=0.2823, pruned_loss=0.08324, over 955624.64 frames. ], batch size: 47, lr: 3.96e-03, grad_scale: 16.0 2023-04-26 15:25:19,617 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=5.24 vs. limit=5.0 2023-04-26 15:25:32,433 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=20908.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:25:32,495 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.1395, 2.2997, 1.9069, 1.8829, 2.1870, 1.7341, 2.8109, 1.4695], device='cuda:6'), covar=tensor([0.4211, 0.1645, 0.3769, 0.3255, 0.2178, 0.3047, 0.1298, 0.4737], device='cuda:6'), in_proj_covar=tensor([0.0354, 0.0356, 0.0440, 0.0371, 0.0402, 0.0381, 0.0397, 0.0423], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-26 15:25:50,202 INFO [optim.py:369] (6/7) 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 15:26:01,985 INFO [finetune.py:976] (6/7) Epoch 4, batch 3750, loss[loss=0.2708, simple_loss=0.3234, pruned_loss=0.1091, over 4868.00 frames. ], tot_loss[loss=0.2249, simple_loss=0.2829, pruned_loss=0.08347, over 954058.39 frames. ], batch size: 34, lr: 3.96e-03, grad_scale: 16.0 2023-04-26 15:26:14,218 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=20944.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:26:24,369 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 2023-04-26 15:26:36,176 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6553, 3.1751, 1.2388, 1.9625, 1.9657, 2.4988, 2.0227, 1.2976], device='cuda:6'), covar=tensor([0.1107, 0.0791, 0.1692, 0.1156, 0.0971, 0.0918, 0.1258, 0.2040], device='cuda:6'), in_proj_covar=tensor([0.0120, 0.0261, 0.0146, 0.0127, 0.0138, 0.0160, 0.0124, 0.0127], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-26 15:27:02,462 INFO [finetune.py:976] (6/7) Epoch 4, batch 3800, loss[loss=0.2216, simple_loss=0.2897, pruned_loss=0.07671, over 4777.00 frames. ], tot_loss[loss=0.2247, simple_loss=0.2838, pruned_loss=0.08286, over 954152.87 frames. ], batch size: 26, lr: 3.96e-03, grad_scale: 16.0 2023-04-26 15:27:46,872 INFO [optim.py:369] (6/7) 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 15:28:05,802 INFO [finetune.py:976] (6/7) Epoch 4, batch 3850, loss[loss=0.2227, simple_loss=0.2721, pruned_loss=0.08671, over 4821.00 frames. ], tot_loss[loss=0.223, simple_loss=0.282, pruned_loss=0.08204, over 953496.38 frames. ], batch size: 38, lr: 3.96e-03, grad_scale: 16.0 2023-04-26 15:28:18,886 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=21044.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:28:20,156 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5196, 1.1079, 1.4939, 1.8774, 1.6982, 1.4665, 1.5401, 1.5941], device='cuda:6'), covar=tensor([1.2417, 1.6017, 1.7476, 1.8572, 1.2886, 1.9067, 1.9180, 1.5184], device='cuda:6'), in_proj_covar=tensor([0.0430, 0.0472, 0.0563, 0.0581, 0.0463, 0.0488, 0.0501, 0.0506], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-26 15:28:29,659 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=21052.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:28:54,718 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=21074.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:29:11,477 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7158, 1.7104, 2.0231, 2.0952, 1.6998, 1.3563, 1.8773, 1.2123], device='cuda:6'), covar=tensor([0.0927, 0.0957, 0.0697, 0.1023, 0.1027, 0.1462, 0.0852, 0.1195], device='cuda:6'), in_proj_covar=tensor([0.0069, 0.0078, 0.0077, 0.0070, 0.0082, 0.0098, 0.0086, 0.0079], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:6') 2023-04-26 15:29:11,948 INFO [finetune.py:976] (6/7) Epoch 4, batch 3900, loss[loss=0.2143, simple_loss=0.2601, pruned_loss=0.08419, over 4745.00 frames. ], tot_loss[loss=0.222, simple_loss=0.2795, pruned_loss=0.08228, over 953978.17 frames. ], batch size: 54, lr: 3.96e-03, grad_scale: 16.0 2023-04-26 15:29:17,321 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.3167, 3.2375, 0.9618, 1.6783, 1.8243, 2.3181, 1.9153, 1.0372], device='cuda:6'), covar=tensor([0.1393, 0.1003, 0.2047, 0.1391, 0.1146, 0.1049, 0.1402, 0.2011], device='cuda:6'), in_proj_covar=tensor([0.0122, 0.0264, 0.0148, 0.0129, 0.0140, 0.0162, 0.0126, 0.0129], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-26 15:29:18,474 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=21092.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:29:31,299 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=21113.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:29:31,347 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.53 vs. limit=5.0 2023-04-26 15:29:34,791 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=21118.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:29:37,131 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=21122.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:29:38,885 INFO [optim.py:369] (6/7) 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 15:29:43,720 INFO [finetune.py:976] (6/7) Epoch 4, batch 3950, loss[loss=0.2333, simple_loss=0.2982, pruned_loss=0.08426, over 4158.00 frames. ], tot_loss[loss=0.2177, simple_loss=0.2756, pruned_loss=0.07991, over 953771.38 frames. ], batch size: 65, lr: 3.96e-03, grad_scale: 16.0 2023-04-26 15:29:56,070 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6453, 1.9601, 1.6044, 1.9003, 1.5252, 1.5480, 1.7112, 1.2606], device='cuda:6'), covar=tensor([0.1921, 0.1432, 0.1009, 0.1279, 0.3178, 0.1383, 0.1802, 0.2598], device='cuda:6'), in_proj_covar=tensor([0.0312, 0.0333, 0.0242, 0.0306, 0.0321, 0.0285, 0.0274, 0.0297], device='cuda:6'), out_proj_covar=tensor([1.2726e-04, 1.3577e-04, 9.8797e-05, 1.2349e-04, 1.3236e-04, 1.1519e-04, 1.1284e-04, 1.2000e-04], device='cuda:6') 2023-04-26 15:29:57,385 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-04-26 15:30:14,127 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=21179.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:30:16,454 INFO [finetune.py:976] (6/7) Epoch 4, batch 4000, loss[loss=0.2209, simple_loss=0.276, pruned_loss=0.08291, over 4906.00 frames. ], tot_loss[loss=0.2167, simple_loss=0.2742, pruned_loss=0.07958, over 955406.03 frames. ], batch size: 35, lr: 3.96e-03, grad_scale: 16.0 2023-04-26 15:30:31,089 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.2214, 1.8344, 2.1295, 2.3882, 2.3190, 1.8692, 1.6326, 2.0418], device='cuda:6'), covar=tensor([0.0990, 0.1216, 0.0715, 0.0723, 0.0729, 0.1114, 0.1160, 0.0743], device='cuda:6'), in_proj_covar=tensor([0.0207, 0.0207, 0.0185, 0.0181, 0.0180, 0.0197, 0.0170, 0.0192], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-26 15:30:38,673 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.78 vs. limit=5.0 2023-04-26 15:30:45,242 INFO [optim.py:369] (6/7) 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] (6/7) Epoch 4, batch 4050, loss[loss=0.2148, simple_loss=0.2738, pruned_loss=0.07788, over 4836.00 frames. ], tot_loss[loss=0.2211, simple_loss=0.2784, pruned_loss=0.08193, over 952920.73 frames. ], batch size: 30, lr: 3.96e-03, grad_scale: 16.0 2023-04-26 15:30:58,879 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=21244.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:31:07,857 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.0011, 1.2738, 3.3148, 3.0714, 2.9590, 3.2111, 3.2202, 2.9249], device='cuda:6'), covar=tensor([0.6725, 0.5384, 0.1396, 0.2104, 0.1422, 0.1853, 0.1367, 0.1655], device='cuda:6'), in_proj_covar=tensor([0.0313, 0.0307, 0.0419, 0.0429, 0.0362, 0.0415, 0.0326, 0.0384], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-26 15:31:10,944 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5017, 1.4559, 0.7438, 1.2455, 1.6908, 1.3831, 1.2936, 1.3862], device='cuda:6'), covar=tensor([0.0585, 0.0455, 0.0455, 0.0636, 0.0302, 0.0627, 0.0599, 0.0692], device='cuda:6'), in_proj_covar=tensor([0.0031, 0.0027, 0.0025, 0.0031, 0.0022, 0.0030, 0.0030, 0.0031], device='cuda:6'), out_proj_covar=tensor([0.0049, 0.0044, 0.0038, 0.0049, 0.0038, 0.0048, 0.0047, 0.0050], device='cuda:6') 2023-04-26 15:31:17,051 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.0843, 1.4445, 1.3177, 1.7568, 1.5584, 1.7823, 1.4092, 3.1251], device='cuda:6'), covar=tensor([0.0759, 0.0834, 0.0882, 0.1300, 0.0689, 0.0530, 0.0810, 0.0218], device='cuda:6'), in_proj_covar=tensor([0.0040, 0.0040, 0.0042, 0.0046, 0.0041, 0.0041, 0.0040, 0.0062], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0017], device='cuda:6') 2023-04-26 15:31:23,491 INFO [finetune.py:976] (6/7) Epoch 4, batch 4100, loss[loss=0.2029, simple_loss=0.2759, pruned_loss=0.065, over 4207.00 frames. ], tot_loss[loss=0.224, simple_loss=0.2815, pruned_loss=0.0832, over 952380.09 frames. ], batch size: 66, lr: 3.96e-03, grad_scale: 16.0 2023-04-26 15:31:30,024 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=21292.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:31:41,622 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-04-26 15:31:43,174 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=21312.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:31:49,243 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.4394, 0.9690, 0.4724, 1.1239, 1.2081, 1.3397, 1.2056, 1.1924], device='cuda:6'), covar=tensor([0.0599, 0.0469, 0.0477, 0.0641, 0.0322, 0.0618, 0.0597, 0.0674], device='cuda:6'), in_proj_covar=tensor([0.0031, 0.0027, 0.0025, 0.0031, 0.0022, 0.0030, 0.0030, 0.0031], device='cuda:6'), out_proj_covar=tensor([0.0049, 0.0044, 0.0038, 0.0049, 0.0038, 0.0048, 0.0047, 0.0050], device='cuda:6') 2023-04-26 15:31:50,918 INFO [optim.py:369] (6/7) 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 15:31:52,875 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=21328.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:31:54,049 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.52 vs. limit=2.0 2023-04-26 15:31:56,305 INFO [finetune.py:976] (6/7) Epoch 4, batch 4150, loss[loss=0.208, simple_loss=0.2728, pruned_loss=0.07164, over 4822.00 frames. ], tot_loss[loss=0.2245, simple_loss=0.2822, pruned_loss=0.08339, over 951496.04 frames. ], batch size: 30, lr: 3.95e-03, grad_scale: 16.0 2023-04-26 15:32:40,012 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=21373.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:32:46,018 INFO [finetune.py:976] (6/7) Epoch 4, batch 4200, loss[loss=0.2314, simple_loss=0.2946, pruned_loss=0.08407, over 4791.00 frames. ], tot_loss[loss=0.2239, simple_loss=0.2822, pruned_loss=0.08286, over 951642.89 frames. ], batch size: 51, lr: 3.95e-03, grad_scale: 16.0 2023-04-26 15:32:50,252 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=21389.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:33:21,607 INFO [zipformer.py:1188] (6/7) 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] (6/7) 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 15:33:52,193 INFO [finetune.py:976] (6/7) Epoch 4, batch 4250, loss[loss=0.1921, simple_loss=0.2424, pruned_loss=0.07091, over 4714.00 frames. ], tot_loss[loss=0.2206, simple_loss=0.2787, pruned_loss=0.08129, over 952628.48 frames. ], batch size: 59, lr: 3.95e-03, grad_scale: 16.0 2023-04-26 15:34:30,959 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=21474.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:34:41,948 INFO [finetune.py:976] (6/7) Epoch 4, batch 4300, loss[loss=0.1657, simple_loss=0.2268, pruned_loss=0.05232, over 4939.00 frames. ], tot_loss[loss=0.2181, simple_loss=0.2761, pruned_loss=0.08004, over 954226.07 frames. ], batch size: 38, lr: 3.95e-03, grad_scale: 16.0 2023-04-26 15:35:27,096 INFO [optim.py:369] (6/7) 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 15:35:31,902 INFO [finetune.py:976] (6/7) Epoch 4, batch 4350, loss[loss=0.2554, simple_loss=0.295, pruned_loss=0.1079, over 4910.00 frames. ], tot_loss[loss=0.2149, simple_loss=0.2725, pruned_loss=0.07868, over 952656.34 frames. ], batch size: 37, lr: 3.95e-03, grad_scale: 16.0 2023-04-26 15:35:51,819 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.21 vs. limit=5.0 2023-04-26 15:36:05,674 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([3.2962, 3.2421, 2.4030, 3.8350, 3.2875, 3.3001, 1.4318, 3.2584], device='cuda:6'), covar=tensor([0.1935, 0.1454, 0.3563, 0.2268, 0.3514, 0.2048, 0.5606, 0.2585], device='cuda:6'), in_proj_covar=tensor([0.0252, 0.0224, 0.0262, 0.0316, 0.0311, 0.0259, 0.0277, 0.0281], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-26 15:36:37,729 INFO [finetune.py:976] (6/7) Epoch 4, batch 4400, loss[loss=0.2613, simple_loss=0.3246, pruned_loss=0.09898, over 4855.00 frames. ], tot_loss[loss=0.2165, simple_loss=0.2744, pruned_loss=0.07934, over 953592.30 frames. ], batch size: 44, lr: 3.95e-03, grad_scale: 16.0 2023-04-26 15:36:56,218 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=21595.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:36:56,267 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.23 vs. limit=5.0 2023-04-26 15:37:09,344 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=21613.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:37:17,114 INFO [optim.py:369] (6/7) 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 15:37:22,024 INFO [finetune.py:976] (6/7) Epoch 4, batch 4450, loss[loss=0.1911, simple_loss=0.2463, pruned_loss=0.0679, over 4093.00 frames. ], tot_loss[loss=0.2201, simple_loss=0.2788, pruned_loss=0.08072, over 951428.55 frames. ], batch size: 18, lr: 3.95e-03, grad_scale: 16.0 2023-04-26 15:37:36,742 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=21656.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:37:46,936 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=21668.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:37:48,525 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-04-26 15:37:50,694 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 4, batch 4500, loss[loss=0.1826, simple_loss=0.2515, pruned_loss=0.05681, over 4780.00 frames. ], tot_loss[loss=0.22, simple_loss=0.2792, pruned_loss=0.08035, over 952616.29 frames. ], batch size: 29, lr: 3.95e-03, grad_scale: 16.0 2023-04-26 15:37:56,777 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=21684.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:38:11,965 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=21708.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:38:24,950 INFO [optim.py:369] (6/7) 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 15:38:29,906 INFO [finetune.py:976] (6/7) Epoch 4, batch 4550, loss[loss=0.2509, simple_loss=0.3183, pruned_loss=0.09176, over 4732.00 frames. ], tot_loss[loss=0.2219, simple_loss=0.281, pruned_loss=0.08142, over 952429.43 frames. ], batch size: 54, lr: 3.95e-03, grad_scale: 16.0 2023-04-26 15:38:33,251 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 2023-04-26 15:38:44,655 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=21756.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:38:49,067 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=21763.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:38:58,636 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=21774.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:39:04,078 INFO [finetune.py:976] (6/7) Epoch 4, batch 4600, loss[loss=0.2215, simple_loss=0.2797, pruned_loss=0.0816, over 4889.00 frames. ], tot_loss[loss=0.2218, simple_loss=0.2811, pruned_loss=0.08127, over 953529.84 frames. ], batch size: 35, lr: 3.95e-03, grad_scale: 16.0 2023-04-26 15:39:30,456 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=21822.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:39:36,953 INFO [zipformer.py:1188] (6/7) 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] (6/7) 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 15:39:39,877 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6530, 2.0867, 1.5759, 1.2329, 1.2556, 1.2539, 1.5704, 1.2663], device='cuda:6'), covar=tensor([0.2112, 0.1672, 0.1983, 0.2370, 0.3079, 0.2345, 0.1563, 0.2443], device='cuda:6'), in_proj_covar=tensor([0.0192, 0.0219, 0.0179, 0.0208, 0.0218, 0.0187, 0.0173, 0.0197], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-26 15:39:48,033 INFO [finetune.py:976] (6/7) Epoch 4, batch 4650, loss[loss=0.218, simple_loss=0.2639, pruned_loss=0.08605, over 4936.00 frames. ], tot_loss[loss=0.2187, simple_loss=0.2775, pruned_loss=0.07992, over 952331.03 frames. ], batch size: 33, lr: 3.95e-03, grad_scale: 16.0 2023-04-26 15:40:02,250 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7145, 1.2567, 1.5254, 1.5513, 1.4324, 1.1815, 0.6573, 1.1671], device='cuda:6'), covar=tensor([0.4225, 0.4775, 0.2223, 0.3160, 0.3986, 0.3638, 0.5917, 0.3615], device='cuda:6'), in_proj_covar=tensor([0.0272, 0.0263, 0.0221, 0.0333, 0.0222, 0.0230, 0.0248, 0.0199], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-26 15:40:09,220 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=21850.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:40:31,221 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6727, 2.0362, 1.5261, 1.2535, 1.2549, 1.2783, 1.5256, 1.2047], device='cuda:6'), covar=tensor([0.2030, 0.1681, 0.2036, 0.2402, 0.3117, 0.2314, 0.1525, 0.2445], device='cuda:6'), in_proj_covar=tensor([0.0193, 0.0220, 0.0180, 0.0209, 0.0218, 0.0187, 0.0173, 0.0197], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-26 15:40:49,528 INFO [finetune.py:976] (6/7) Epoch 4, batch 4700, loss[loss=0.2186, simple_loss=0.2708, pruned_loss=0.0832, over 4926.00 frames. ], tot_loss[loss=0.2166, simple_loss=0.2749, pruned_loss=0.07918, over 953383.61 frames. ], batch size: 37, lr: 3.95e-03, grad_scale: 16.0 2023-04-26 15:41:12,775 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=21911.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:41:21,741 INFO [optim.py:369] (6/7) 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 15:41:28,679 INFO [finetune.py:976] (6/7) Epoch 4, batch 4750, loss[loss=0.2243, simple_loss=0.2861, pruned_loss=0.08122, over 4922.00 frames. ], tot_loss[loss=0.2148, simple_loss=0.2728, pruned_loss=0.07841, over 953729.39 frames. ], batch size: 38, lr: 3.95e-03, grad_scale: 16.0 2023-04-26 15:41:34,938 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-04-26 15:41:40,252 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=21951.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:41:48,206 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6680, 1.7286, 4.1742, 3.8673, 3.7413, 3.9089, 3.9911, 3.6755], device='cuda:6'), covar=tensor([0.6615, 0.5172, 0.1135, 0.1977, 0.1180, 0.1489, 0.1225, 0.1692], device='cuda:6'), in_proj_covar=tensor([0.0316, 0.0306, 0.0422, 0.0426, 0.0363, 0.0415, 0.0325, 0.0382], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-26 15:41:50,385 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-04-26 15:41:50,662 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=21968.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:41:51,757 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=21969.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:41:52,440 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=21970.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:42:02,274 INFO [finetune.py:976] (6/7) Epoch 4, batch 4800, loss[loss=0.234, simple_loss=0.2982, pruned_loss=0.08487, over 4831.00 frames. ], tot_loss[loss=0.2179, simple_loss=0.276, pruned_loss=0.07987, over 953818.76 frames. ], batch size: 33, lr: 3.95e-03, grad_scale: 16.0 2023-04-26 15:42:02,999 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=21984.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:42:39,326 INFO [zipformer.py:1188] (6/7) 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] (6/7) 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 15:42:54,749 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=22031.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:42:55,828 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=22032.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:42:56,368 INFO [finetune.py:976] (6/7) Epoch 4, batch 4850, loss[loss=0.2605, simple_loss=0.3203, pruned_loss=0.1003, over 4781.00 frames. ], tot_loss[loss=0.2201, simple_loss=0.2793, pruned_loss=0.08042, over 953697.86 frames. ], batch size: 29, lr: 3.95e-03, grad_scale: 16.0 2023-04-26 15:43:45,552 INFO [finetune.py:976] (6/7) Epoch 4, batch 4900, loss[loss=0.2595, simple_loss=0.3189, pruned_loss=0.1001, over 4834.00 frames. ], tot_loss[loss=0.2213, simple_loss=0.281, pruned_loss=0.08085, over 955618.05 frames. ], batch size: 49, lr: 3.95e-03, grad_scale: 16.0 2023-04-26 15:44:01,943 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.40 vs. limit=5.0 2023-04-26 15:44:15,735 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=22119.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:44:19,288 INFO [optim.py:369] (6/7) 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 15:44:22,990 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=22130.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:44:24,745 INFO [finetune.py:976] (6/7) Epoch 4, batch 4950, loss[loss=0.2873, simple_loss=0.339, pruned_loss=0.1178, over 4799.00 frames. ], tot_loss[loss=0.2239, simple_loss=0.2834, pruned_loss=0.08225, over 955146.27 frames. ], batch size: 45, lr: 3.95e-03, grad_scale: 16.0 2023-04-26 15:44:53,668 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=22176.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:44:58,359 INFO [finetune.py:976] (6/7) Epoch 4, batch 5000, loss[loss=0.2443, simple_loss=0.2963, pruned_loss=0.09616, over 4711.00 frames. ], tot_loss[loss=0.2197, simple_loss=0.2801, pruned_loss=0.07965, over 957496.20 frames. ], batch size: 59, lr: 3.95e-03, grad_scale: 16.0 2023-04-26 15:45:04,355 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=22191.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:45:15,022 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=22206.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:45:20,504 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=22215.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:45:31,709 INFO [optim.py:369] (6/7) 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 15:45:42,551 INFO [finetune.py:976] (6/7) Epoch 4, batch 5050, loss[loss=0.2031, simple_loss=0.2686, pruned_loss=0.06879, over 4791.00 frames. ], tot_loss[loss=0.2182, simple_loss=0.2775, pruned_loss=0.07943, over 953632.65 frames. ], batch size: 29, lr: 3.95e-03, grad_scale: 16.0 2023-04-26 15:45:45,110 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=22237.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:45:45,124 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=22237.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:46:06,564 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.4817, 3.6576, 0.9220, 1.8211, 2.0428, 2.4444, 2.0366, 1.0621], device='cuda:6'), covar=tensor([0.1550, 0.0950, 0.2306, 0.1456, 0.1138, 0.1190, 0.1599, 0.2060], device='cuda:6'), in_proj_covar=tensor([0.0122, 0.0264, 0.0149, 0.0129, 0.0139, 0.0162, 0.0126, 0.0128], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-26 15:46:06,579 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=22251.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:46:18,118 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-04-26 15:46:28,143 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=22269.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:46:38,178 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=22276.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:46:42,369 INFO [finetune.py:976] (6/7) Epoch 4, batch 5100, loss[loss=0.2306, simple_loss=0.2756, pruned_loss=0.09282, over 4916.00 frames. ], tot_loss[loss=0.2158, simple_loss=0.2744, pruned_loss=0.07861, over 951700.95 frames. ], batch size: 37, lr: 3.95e-03, grad_scale: 16.0 2023-04-26 15:46:52,580 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=22298.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:46:53,598 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=22299.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:47:06,010 INFO [zipformer.py:1188] (6/7) 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] (6/7) 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 15:47:11,501 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=22326.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:47:14,320 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2023-04-26 15:47:15,715 INFO [finetune.py:976] (6/7) Epoch 4, batch 5150, loss[loss=0.2013, simple_loss=0.2674, pruned_loss=0.06764, over 4739.00 frames. ], tot_loss[loss=0.2151, simple_loss=0.2738, pruned_loss=0.07827, over 953895.54 frames. ], batch size: 27, lr: 3.95e-03, grad_scale: 16.0 2023-04-26 15:47:22,408 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5663, 1.6916, 0.7935, 1.2606, 1.8533, 1.4769, 1.4081, 1.4241], device='cuda:6'), covar=tensor([0.0537, 0.0392, 0.0432, 0.0594, 0.0307, 0.0587, 0.0549, 0.0616], device='cuda:6'), in_proj_covar=tensor([0.0031, 0.0027, 0.0025, 0.0031, 0.0022, 0.0031, 0.0030, 0.0031], device='cuda:6'), out_proj_covar=tensor([0.0049, 0.0045, 0.0038, 0.0049, 0.0038, 0.0048, 0.0047, 0.0050], device='cuda:6') 2023-04-26 15:47:24,804 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.1695, 1.5756, 1.4917, 1.9141, 1.6665, 2.1735, 1.5220, 4.0411], device='cuda:6'), covar=tensor([0.0638, 0.0746, 0.0775, 0.1260, 0.0664, 0.0579, 0.0767, 0.0125], device='cuda:6'), in_proj_covar=tensor([0.0040, 0.0040, 0.0042, 0.0046, 0.0041, 0.0041, 0.0040, 0.0063], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0017], device='cuda:6') 2023-04-26 15:47:54,339 INFO [finetune.py:976] (6/7) Epoch 4, batch 5200, loss[loss=0.1713, simple_loss=0.244, pruned_loss=0.04933, over 4711.00 frames. ], tot_loss[loss=0.22, simple_loss=0.2789, pruned_loss=0.08058, over 954026.01 frames. ], batch size: 23, lr: 3.95e-03, grad_scale: 16.0 2023-04-26 15:48:01,254 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=22385.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:48:13,798 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.4263, 1.5377, 1.5507, 2.0558, 1.6929, 2.1958, 1.5436, 4.2856], device='cuda:6'), covar=tensor([0.0643, 0.0816, 0.0847, 0.1273, 0.0749, 0.0576, 0.0839, 0.0124], device='cuda:6'), in_proj_covar=tensor([0.0040, 0.0040, 0.0042, 0.0046, 0.0041, 0.0041, 0.0040, 0.0063], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0017], device='cuda:6') 2023-04-26 15:48:30,912 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=22419.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:48:32,103 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9213, 2.5569, 1.0264, 1.2580, 2.0517, 1.1294, 3.3195, 1.6124], device='cuda:6'), covar=tensor([0.0726, 0.0811, 0.0960, 0.1324, 0.0514, 0.1054, 0.0226, 0.0669], device='cuda:6'), in_proj_covar=tensor([0.0054, 0.0070, 0.0052, 0.0049, 0.0054, 0.0055, 0.0083, 0.0053], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0011, 0.0008], device='cuda:6') 2023-04-26 15:48:34,446 INFO [optim.py:369] (6/7) 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 15:48:39,402 INFO [finetune.py:976] (6/7) Epoch 4, batch 5250, loss[loss=0.2096, simple_loss=0.2765, pruned_loss=0.07137, over 4731.00 frames. ], tot_loss[loss=0.2216, simple_loss=0.2814, pruned_loss=0.08089, over 955890.24 frames. ], batch size: 54, lr: 3.95e-03, grad_scale: 16.0 2023-04-26 15:48:44,048 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.25 vs. limit=5.0 2023-04-26 15:48:47,969 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=22446.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:49:09,425 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 4, batch 5300, loss[loss=0.2606, simple_loss=0.3124, pruned_loss=0.1044, over 4886.00 frames. ], tot_loss[loss=0.2214, simple_loss=0.2812, pruned_loss=0.0808, over 955274.86 frames. ], batch size: 32, lr: 3.95e-03, grad_scale: 16.0 2023-04-26 15:49:32,901 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=22486.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:49:56,904 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=22506.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:50:15,640 INFO [optim.py:369] (6/7) 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 15:50:19,966 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=22532.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:50:20,501 INFO [finetune.py:976] (6/7) Epoch 4, batch 5350, loss[loss=0.1725, simple_loss=0.2352, pruned_loss=0.05489, over 4894.00 frames. ], tot_loss[loss=0.2194, simple_loss=0.2799, pruned_loss=0.0795, over 954971.84 frames. ], batch size: 32, lr: 3.95e-03, grad_scale: 32.0 2023-04-26 15:50:33,853 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=22554.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:50:46,634 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=22571.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:50:53,884 INFO [finetune.py:976] (6/7) Epoch 4, batch 5400, loss[loss=0.1713, simple_loss=0.2406, pruned_loss=0.05103, over 4749.00 frames. ], tot_loss[loss=0.2161, simple_loss=0.2765, pruned_loss=0.07785, over 956005.83 frames. ], batch size: 28, lr: 3.95e-03, grad_scale: 32.0 2023-04-26 15:51:00,003 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=22593.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:51:15,273 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.1922, 1.5603, 1.4113, 1.8518, 1.6293, 2.0080, 1.4234, 3.4367], device='cuda:6'), covar=tensor([0.0704, 0.0707, 0.0781, 0.1146, 0.0636, 0.0605, 0.0727, 0.0162], device='cuda:6'), in_proj_covar=tensor([0.0040, 0.0040, 0.0042, 0.0046, 0.0041, 0.0041, 0.0040, 0.0063], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0017], device='cuda:6') 2023-04-26 15:51:38,544 INFO [optim.py:369] (6/7) 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 15:51:39,239 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=22626.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:51:47,648 INFO [finetune.py:976] (6/7) Epoch 4, batch 5450, loss[loss=0.1989, simple_loss=0.2581, pruned_loss=0.06986, over 4897.00 frames. ], tot_loss[loss=0.2138, simple_loss=0.2734, pruned_loss=0.07708, over 956864.71 frames. ], batch size: 35, lr: 3.95e-03, grad_scale: 32.0 2023-04-26 15:51:51,435 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=22639.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:52:10,965 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.7290, 1.9486, 1.8083, 1.9810, 1.8057, 2.0558, 1.9281, 1.8455], device='cuda:6'), covar=tensor([0.7553, 1.3840, 1.1481, 0.9377, 1.1698, 1.7442, 1.5376, 1.3340], device='cuda:6'), in_proj_covar=tensor([0.0311, 0.0398, 0.0320, 0.0326, 0.0351, 0.0414, 0.0385, 0.0340], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-26 15:52:13,174 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0970, 2.5099, 1.0549, 1.5364, 1.9350, 1.2266, 3.4925, 1.7912], device='cuda:6'), covar=tensor([0.0696, 0.0686, 0.0807, 0.1288, 0.0562, 0.1028, 0.0352, 0.0658], device='cuda:6'), in_proj_covar=tensor([0.0054, 0.0070, 0.0052, 0.0049, 0.0054, 0.0055, 0.0082, 0.0053], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0011, 0.0008], device='cuda:6') 2023-04-26 15:52:18,662 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-04-26 15:52:39,606 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 4, batch 5500, loss[loss=0.2078, simple_loss=0.2787, pruned_loss=0.06843, over 4765.00 frames. ], tot_loss[loss=0.2118, simple_loss=0.2711, pruned_loss=0.07626, over 957946.95 frames. ], batch size: 27, lr: 3.95e-03, grad_scale: 32.0 2023-04-26 15:53:05,092 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.5169, 1.8216, 1.7577, 1.9073, 1.7827, 1.9835, 1.9211, 1.8371], device='cuda:6'), covar=tensor([0.7475, 1.2927, 1.1280, 0.9568, 1.1850, 1.6904, 1.2796, 1.1601], device='cuda:6'), in_proj_covar=tensor([0.0311, 0.0398, 0.0320, 0.0325, 0.0350, 0.0413, 0.0385, 0.0339], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-26 15:53:12,635 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=22700.0, num_to_drop=1, layers_to_drop={2} 2023-04-26 15:53:34,590 INFO [optim.py:369] (6/7) 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 15:53:40,510 INFO [finetune.py:976] (6/7) Epoch 4, batch 5550, loss[loss=0.2307, simple_loss=0.2838, pruned_loss=0.08881, over 4782.00 frames. ], tot_loss[loss=0.2151, simple_loss=0.2743, pruned_loss=0.07792, over 956823.27 frames. ], batch size: 25, lr: 3.95e-03, grad_scale: 32.0 2023-04-26 15:53:45,466 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=22741.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:54:11,075 INFO [finetune.py:976] (6/7) Epoch 4, batch 5600, loss[loss=0.2255, simple_loss=0.2918, pruned_loss=0.07963, over 4844.00 frames. ], tot_loss[loss=0.2159, simple_loss=0.2758, pruned_loss=0.07803, over 955768.72 frames. ], batch size: 49, lr: 3.95e-03, grad_scale: 32.0 2023-04-26 15:54:12,895 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=22786.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:54:18,059 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=22794.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:54:18,713 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-04-26 15:54:37,297 INFO [optim.py:369] (6/7) 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 15:54:37,380 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=22825.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:54:38,922 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-04-26 15:54:41,490 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=22832.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:54:42,025 INFO [finetune.py:976] (6/7) Epoch 4, batch 5650, loss[loss=0.2074, simple_loss=0.2696, pruned_loss=0.07262, over 4220.00 frames. ], tot_loss[loss=0.2169, simple_loss=0.2779, pruned_loss=0.07798, over 954544.77 frames. ], batch size: 65, lr: 3.95e-03, grad_scale: 32.0 2023-04-26 15:54:42,646 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=22834.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:54:55,166 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=22855.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:55:05,063 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-04-26 15:55:15,953 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=22871.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:55:27,437 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=22880.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:55:29,217 INFO [finetune.py:976] (6/7) Epoch 4, batch 5700, loss[loss=0.2499, simple_loss=0.2732, pruned_loss=0.1133, over 4435.00 frames. ], tot_loss[loss=0.2158, simple_loss=0.2754, pruned_loss=0.07814, over 939326.88 frames. ], batch size: 19, lr: 3.95e-03, grad_scale: 32.0 2023-04-26 15:55:31,126 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=22886.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:55:40,628 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=22893.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:56:19,322 INFO [finetune.py:976] (6/7) Epoch 5, batch 0, loss[loss=0.2522, simple_loss=0.2998, pruned_loss=0.1023, over 4840.00 frames. ], tot_loss[loss=0.2522, simple_loss=0.2998, pruned_loss=0.1023, over 4840.00 frames. ], batch size: 49, lr: 3.95e-03, grad_scale: 32.0 2023-04-26 15:56:19,322 INFO [finetune.py:1001] (6/7) Computing validation loss 2023-04-26 15:56:29,096 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.4527, 1.2569, 1.7056, 1.5720, 1.3389, 1.1853, 1.4131, 0.9050], device='cuda:6'), covar=tensor([0.0755, 0.0848, 0.0550, 0.0895, 0.0992, 0.1594, 0.0662, 0.1108], device='cuda:6'), in_proj_covar=tensor([0.0068, 0.0077, 0.0075, 0.0069, 0.0080, 0.0096, 0.0083, 0.0078], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-26 15:56:30,075 INFO [finetune.py:1010] (6/7) 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] (6/7) Maximum memory allocated so far is 6326MB 2023-04-26 15:56:35,031 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=22919.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:56:38,534 INFO [optim.py:369] (6/7) 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 15:56:48,362 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=22941.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:57:02,115 INFO [finetune.py:976] (6/7) Epoch 5, batch 50, loss[loss=0.2302, simple_loss=0.2863, pruned_loss=0.08707, over 4835.00 frames. ], tot_loss[loss=0.2278, simple_loss=0.2851, pruned_loss=0.08522, over 216093.52 frames. ], batch size: 47, lr: 3.95e-03, grad_scale: 32.0 2023-04-26 15:57:30,162 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=22995.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 15:57:52,330 INFO [finetune.py:976] (6/7) Epoch 5, batch 100, loss[loss=0.1875, simple_loss=0.2529, pruned_loss=0.06107, over 4902.00 frames. ], tot_loss[loss=0.2177, simple_loss=0.2747, pruned_loss=0.08033, over 378454.79 frames. ], batch size: 36, lr: 3.95e-03, grad_scale: 32.0 2023-04-26 15:58:02,669 INFO [optim.py:369] (6/7) 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 15:58:10,122 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-04-26 15:58:12,553 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=23041.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:58:20,371 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.3844, 1.6643, 1.6359, 1.8780, 1.7321, 1.9855, 1.5815, 3.0935], device='cuda:6'), covar=tensor([0.0624, 0.0628, 0.0655, 0.0968, 0.0519, 0.0624, 0.0638, 0.0197], device='cuda:6'), in_proj_covar=tensor([0.0040, 0.0040, 0.0042, 0.0046, 0.0041, 0.0041, 0.0040, 0.0063], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0017], device='cuda:6') 2023-04-26 15:58:25,549 INFO [finetune.py:976] (6/7) Epoch 5, batch 150, loss[loss=0.2099, simple_loss=0.2649, pruned_loss=0.0775, over 4763.00 frames. ], tot_loss[loss=0.2128, simple_loss=0.2699, pruned_loss=0.07786, over 505682.14 frames. ], batch size: 27, lr: 3.95e-03, grad_scale: 32.0 2023-04-26 15:58:37,398 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.0429, 1.3627, 1.2545, 1.6693, 1.5227, 1.4465, 1.3413, 2.4505], device='cuda:6'), covar=tensor([0.0665, 0.0836, 0.0886, 0.1316, 0.0694, 0.0559, 0.0787, 0.0237], device='cuda:6'), in_proj_covar=tensor([0.0040, 0.0040, 0.0042, 0.0046, 0.0041, 0.0041, 0.0040, 0.0063], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0017], device='cuda:6') 2023-04-26 15:58:44,620 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=23089.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:58:57,342 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=23108.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:58:59,553 INFO [finetune.py:976] (6/7) Epoch 5, batch 200, loss[loss=0.2006, simple_loss=0.2619, pruned_loss=0.06963, over 4910.00 frames. ], tot_loss[loss=0.2133, simple_loss=0.2703, pruned_loss=0.07817, over 606335.16 frames. ], batch size: 35, lr: 3.95e-03, grad_scale: 32.0 2023-04-26 15:59:09,618 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0286, 1.2417, 1.3994, 1.5467, 1.4778, 1.6207, 1.4690, 1.5210], device='cuda:6'), covar=tensor([0.8039, 1.1111, 0.9936, 0.8861, 1.0586, 1.6848, 1.1088, 1.0133], device='cuda:6'), in_proj_covar=tensor([0.0311, 0.0397, 0.0320, 0.0326, 0.0349, 0.0413, 0.0384, 0.0339], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-26 15:59:10,067 INFO [optim.py:369] (6/7) 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 15:59:25,390 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 5, batch 250, loss[loss=0.2403, simple_loss=0.3044, pruned_loss=0.08812, over 4818.00 frames. ], tot_loss[loss=0.2169, simple_loss=0.2749, pruned_loss=0.07944, over 685022.07 frames. ], batch size: 38, lr: 3.95e-03, grad_scale: 32.0 2023-04-26 15:59:34,856 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5700, 1.7975, 1.4062, 0.9882, 1.2361, 1.1647, 1.4271, 1.1662], device='cuda:6'), covar=tensor([0.2093, 0.1675, 0.2084, 0.2347, 0.2968, 0.2453, 0.1497, 0.2459], device='cuda:6'), in_proj_covar=tensor([0.0196, 0.0223, 0.0181, 0.0211, 0.0221, 0.0190, 0.0175, 0.0198], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-26 15:59:38,541 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=23169.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 15:59:47,773 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 5, batch 300, loss[loss=0.2405, simple_loss=0.2685, pruned_loss=0.1062, over 4216.00 frames. ], tot_loss[loss=0.2185, simple_loss=0.2773, pruned_loss=0.0799, over 743899.97 frames. ], batch size: 18, lr: 3.95e-03, grad_scale: 32.0 2023-04-26 16:00:10,754 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=23217.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 16:00:15,564 INFO [optim.py:369] (6/7) 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 16:00:18,575 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=23229.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 16:00:34,703 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([0.9133, 1.9410, 1.8250, 1.6536, 2.2482, 1.7399, 2.7695, 1.5960], device='cuda:6'), covar=tensor([0.4307, 0.2092, 0.5228, 0.3396, 0.1759, 0.2563, 0.1332, 0.4916], device='cuda:6'), in_proj_covar=tensor([0.0358, 0.0360, 0.0444, 0.0374, 0.0405, 0.0388, 0.0401, 0.0425], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-26 16:00:35,401 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.8846, 2.0648, 1.9167, 2.7449, 2.8059, 2.5661, 2.3929, 2.1870], device='cuda:6'), covar=tensor([0.2021, 0.1830, 0.2220, 0.1778, 0.1231, 0.1931, 0.2606, 0.2068], device='cuda:6'), in_proj_covar=tensor([0.0313, 0.0332, 0.0349, 0.0306, 0.0346, 0.0347, 0.0311, 0.0353], device='cuda:6'), out_proj_covar=tensor([6.7439e-05, 7.1263e-05, 7.5591e-05, 6.3988e-05, 7.3276e-05, 7.5474e-05, 6.7589e-05, 7.6138e-05], device='cuda:6') 2023-04-26 16:00:55,702 INFO [finetune.py:976] (6/7) Epoch 5, batch 350, loss[loss=0.2673, simple_loss=0.3111, pruned_loss=0.1118, over 4814.00 frames. ], tot_loss[loss=0.2206, simple_loss=0.2789, pruned_loss=0.08112, over 789129.04 frames. ], batch size: 33, lr: 3.95e-03, grad_scale: 32.0 2023-04-26 16:01:18,166 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=23278.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 16:01:28,479 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=2.02 vs. limit=2.0 2023-04-26 16:01:31,695 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=23290.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 16:01:41,013 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=23295.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 16:01:55,883 INFO [finetune.py:976] (6/7) Epoch 5, batch 400, loss[loss=0.2386, simple_loss=0.2917, pruned_loss=0.09279, over 4773.00 frames. ], tot_loss[loss=0.2213, simple_loss=0.2803, pruned_loss=0.08113, over 824316.80 frames. ], batch size: 51, lr: 3.95e-03, grad_scale: 32.0 2023-04-26 16:02:05,330 INFO [optim.py:369] (6/7) 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 16:02:18,849 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=23343.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 16:02:29,821 INFO [finetune.py:976] (6/7) Epoch 5, batch 450, loss[loss=0.2158, simple_loss=0.2898, pruned_loss=0.07087, over 4898.00 frames. ], tot_loss[loss=0.2196, simple_loss=0.2786, pruned_loss=0.0803, over 855469.93 frames. ], batch size: 36, lr: 3.95e-03, grad_scale: 32.0 2023-04-26 16:02:30,539 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7321, 1.2776, 1.3113, 1.3763, 1.9294, 1.6072, 1.2240, 1.2613], device='cuda:6'), covar=tensor([0.1370, 0.1356, 0.1746, 0.1353, 0.0719, 0.1363, 0.1968, 0.1739], device='cuda:6'), in_proj_covar=tensor([0.0311, 0.0330, 0.0347, 0.0305, 0.0343, 0.0344, 0.0309, 0.0351], device='cuda:6'), out_proj_covar=tensor([6.6882e-05, 7.0639e-05, 7.5248e-05, 6.3657e-05, 7.2570e-05, 7.4834e-05, 6.7142e-05, 7.5675e-05], device='cuda:6') 2023-04-26 16:02:37,856 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 2023-04-26 16:03:14,683 INFO [finetune.py:976] (6/7) Epoch 5, batch 500, loss[loss=0.3092, simple_loss=0.3517, pruned_loss=0.1334, over 4254.00 frames. ], tot_loss[loss=0.2174, simple_loss=0.276, pruned_loss=0.07939, over 876992.19 frames. ], batch size: 66, lr: 3.95e-03, grad_scale: 32.0 2023-04-26 16:03:29,842 INFO [optim.py:369] (6/7) 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 16:03:36,491 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.2232, 2.8162, 2.4605, 2.5382, 2.1325, 2.2833, 2.5935, 1.9328], device='cuda:6'), covar=tensor([0.3190, 0.1846, 0.1176, 0.1976, 0.3664, 0.1858, 0.2620, 0.3580], device='cuda:6'), in_proj_covar=tensor([0.0311, 0.0331, 0.0240, 0.0304, 0.0323, 0.0287, 0.0273, 0.0296], device='cuda:6'), 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:6') 2023-04-26 16:03:47,511 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=23450.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 16:03:54,143 INFO [finetune.py:976] (6/7) Epoch 5, batch 550, loss[loss=0.2181, simple_loss=0.2752, pruned_loss=0.08052, over 4767.00 frames. ], tot_loss[loss=0.2143, simple_loss=0.2727, pruned_loss=0.07798, over 894616.74 frames. ], batch size: 27, lr: 3.95e-03, grad_scale: 32.0 2023-04-26 16:03:56,065 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=23464.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 16:04:07,428 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=23481.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 16:04:12,220 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=23488.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 16:04:19,705 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=23498.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 16:04:27,580 INFO [finetune.py:976] (6/7) Epoch 5, batch 600, loss[loss=0.2321, simple_loss=0.2909, pruned_loss=0.0867, over 4909.00 frames. ], tot_loss[loss=0.2132, simple_loss=0.2722, pruned_loss=0.07707, over 910127.87 frames. ], batch size: 37, lr: 3.94e-03, grad_scale: 32.0 2023-04-26 16:04:36,133 INFO [optim.py:369] (6/7) 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 16:04:39,614 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=23529.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 16:04:53,891 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=23549.0, num_to_drop=1, layers_to_drop={3} 2023-04-26 16:05:01,024 INFO [finetune.py:976] (6/7) Epoch 5, batch 650, loss[loss=0.2099, simple_loss=0.2788, pruned_loss=0.07048, over 4864.00 frames. ], tot_loss[loss=0.2158, simple_loss=0.2756, pruned_loss=0.07798, over 921597.18 frames. ], batch size: 44, lr: 3.94e-03, grad_scale: 32.0 2023-04-26 16:05:08,399 INFO [zipformer.py:1188] (6/7) 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] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=23585.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 16:05:25,507 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([0.6671, 1.7345, 1.8637, 1.3604, 1.8525, 1.4355, 2.3066, 1.6308], device='cuda:6'), covar=tensor([0.3927, 0.1626, 0.3939, 0.3015, 0.1484, 0.2448, 0.1654, 0.4089], device='cuda:6'), in_proj_covar=tensor([0.0353, 0.0355, 0.0439, 0.0370, 0.0400, 0.0386, 0.0397, 0.0421], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-26 16:05:34,424 INFO [finetune.py:976] (6/7) Epoch 5, batch 700, loss[loss=0.2185, simple_loss=0.2871, pruned_loss=0.07495, over 4740.00 frames. ], tot_loss[loss=0.2179, simple_loss=0.2774, pruned_loss=0.07922, over 926538.79 frames. ], batch size: 27, lr: 3.94e-03, grad_scale: 32.0 2023-04-26 16:05:42,879 INFO [optim.py:369] (6/7) 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 16:05:56,485 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9910, 1.3975, 1.4552, 1.5980, 2.1347, 1.8511, 1.3562, 1.4693], device='cuda:6'), covar=tensor([0.1654, 0.1866, 0.2261, 0.1432, 0.1001, 0.1777, 0.2972, 0.2162], device='cuda:6'), in_proj_covar=tensor([0.0314, 0.0334, 0.0351, 0.0308, 0.0344, 0.0347, 0.0313, 0.0353], device='cuda:6'), out_proj_covar=tensor([6.7718e-05, 7.1741e-05, 7.5998e-05, 6.4314e-05, 7.2923e-05, 7.5475e-05, 6.7995e-05, 7.6287e-05], device='cuda:6') 2023-04-26 16:06:19,708 INFO [finetune.py:976] (6/7) Epoch 5, batch 750, loss[loss=0.1795, simple_loss=0.258, pruned_loss=0.05053, over 4834.00 frames. ], tot_loss[loss=0.22, simple_loss=0.2796, pruned_loss=0.08027, over 931067.22 frames. ], batch size: 47, lr: 3.94e-03, grad_scale: 32.0 2023-04-26 16:07:02,613 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-04-26 16:07:26,346 INFO [finetune.py:976] (6/7) Epoch 5, batch 800, loss[loss=0.193, simple_loss=0.2615, pruned_loss=0.06226, over 4892.00 frames. ], tot_loss[loss=0.2182, simple_loss=0.2781, pruned_loss=0.07918, over 935605.68 frames. ], batch size: 35, lr: 3.94e-03, grad_scale: 32.0 2023-04-26 16:07:31,880 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.1058, 2.7467, 1.2937, 1.5138, 2.0843, 1.4449, 3.1660, 1.7979], device='cuda:6'), covar=tensor([0.0630, 0.0733, 0.0822, 0.1093, 0.0450, 0.0874, 0.0251, 0.0580], device='cuda:6'), in_proj_covar=tensor([0.0054, 0.0071, 0.0053, 0.0050, 0.0054, 0.0055, 0.0083, 0.0053], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0008, 0.0008, 0.0008, 0.0008, 0.0011, 0.0008], device='cuda:6') 2023-04-26 16:07:34,819 INFO [optim.py:369] (6/7) 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 16:08:00,117 INFO [finetune.py:976] (6/7) Epoch 5, batch 850, loss[loss=0.1909, simple_loss=0.2522, pruned_loss=0.0648, over 4815.00 frames. ], tot_loss[loss=0.2165, simple_loss=0.2759, pruned_loss=0.07856, over 938827.78 frames. ], batch size: 40, lr: 3.94e-03, grad_scale: 32.0 2023-04-26 16:08:02,001 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=23764.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 16:08:39,275 INFO [finetune.py:976] (6/7) Epoch 5, batch 900, loss[loss=0.2269, simple_loss=0.2858, pruned_loss=0.08401, over 4908.00 frames. ], tot_loss[loss=0.2138, simple_loss=0.2729, pruned_loss=0.07732, over 942432.52 frames. ], batch size: 36, lr: 3.94e-03, grad_scale: 32.0 2023-04-26 16:08:40,388 INFO [zipformer.py:1188] (6/7) 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] (6/7) 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 16:09:23,844 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=23844.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 16:09:46,930 INFO [finetune.py:976] (6/7) Epoch 5, batch 950, loss[loss=0.1741, simple_loss=0.2427, pruned_loss=0.05275, over 4752.00 frames. ], tot_loss[loss=0.212, simple_loss=0.2709, pruned_loss=0.07652, over 946065.66 frames. ], batch size: 28, lr: 3.94e-03, grad_scale: 32.0 2023-04-26 16:09:52,700 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.1219, 2.1695, 1.7568, 1.7278, 2.2773, 1.6192, 2.8132, 1.5827], device='cuda:6'), covar=tensor([0.4620, 0.1973, 0.5632, 0.3762, 0.1965, 0.3119, 0.1367, 0.4836], device='cuda:6'), in_proj_covar=tensor([0.0351, 0.0355, 0.0438, 0.0371, 0.0400, 0.0383, 0.0398, 0.0420], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-26 16:09:54,505 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.6263, 2.5494, 2.7827, 3.2000, 2.8569, 2.5400, 2.0047, 2.7413], device='cuda:6'), covar=tensor([0.0999, 0.1074, 0.0641, 0.0575, 0.0671, 0.1036, 0.1097, 0.0653], device='cuda:6'), in_proj_covar=tensor([0.0207, 0.0210, 0.0186, 0.0182, 0.0183, 0.0199, 0.0171, 0.0194], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-26 16:09:56,955 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=23873.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 16:10:14,850 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=23885.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 16:10:14,926 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-04-26 16:10:44,369 INFO [finetune.py:976] (6/7) Epoch 5, batch 1000, loss[loss=0.2233, simple_loss=0.2727, pruned_loss=0.08694, over 4798.00 frames. ], tot_loss[loss=0.2135, simple_loss=0.2731, pruned_loss=0.07696, over 950768.87 frames. ], batch size: 25, lr: 3.94e-03, grad_scale: 32.0 2023-04-26 16:10:52,028 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=23921.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 16:10:54,328 INFO [optim.py:369] (6/7) 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 16:10:59,243 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=23933.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 16:11:05,272 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.4558, 3.4051, 0.8022, 1.6962, 1.8056, 2.1685, 1.8879, 1.1117], device='cuda:6'), covar=tensor([0.1776, 0.1522, 0.2613, 0.1884, 0.1313, 0.1512, 0.1728, 0.2241], device='cuda:6'), in_proj_covar=tensor([0.0123, 0.0263, 0.0146, 0.0128, 0.0138, 0.0160, 0.0126, 0.0128], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-26 16:11:17,666 INFO [finetune.py:976] (6/7) Epoch 5, batch 1050, loss[loss=0.2457, simple_loss=0.3003, pruned_loss=0.09555, over 4835.00 frames. ], tot_loss[loss=0.2174, simple_loss=0.277, pruned_loss=0.07889, over 950508.73 frames. ], batch size: 47, lr: 3.94e-03, grad_scale: 32.0 2023-04-26 16:12:31,172 INFO [finetune.py:976] (6/7) Epoch 5, batch 1100, loss[loss=0.2403, simple_loss=0.2981, pruned_loss=0.09122, over 4830.00 frames. ], tot_loss[loss=0.2184, simple_loss=0.2783, pruned_loss=0.07925, over 949171.14 frames. ], batch size: 30, lr: 3.94e-03, grad_scale: 32.0 2023-04-26 16:12:45,518 INFO [optim.py:369] (6/7) 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 16:12:53,352 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=24037.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 16:13:08,679 INFO [finetune.py:976] (6/7) Epoch 5, batch 1150, loss[loss=0.2099, simple_loss=0.2674, pruned_loss=0.07619, over 4775.00 frames. ], tot_loss[loss=0.2181, simple_loss=0.2781, pruned_loss=0.07903, over 950858.92 frames. ], batch size: 25, lr: 3.94e-03, grad_scale: 32.0 2023-04-26 16:13:33,504 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=24098.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 16:13:42,163 INFO [finetune.py:976] (6/7) Epoch 5, batch 1200, loss[loss=0.2033, simple_loss=0.2567, pruned_loss=0.07498, over 4808.00 frames. ], tot_loss[loss=0.2165, simple_loss=0.2763, pruned_loss=0.07831, over 950753.53 frames. ], batch size: 39, lr: 3.94e-03, grad_scale: 32.0 2023-04-26 16:13:43,883 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.80 vs. limit=5.0 2023-04-26 16:13:52,168 INFO [optim.py:369] (6/7) 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 16:14:04,705 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=24144.0, num_to_drop=1, layers_to_drop={2} 2023-04-26 16:14:07,175 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.6235, 1.8833, 1.6858, 1.8852, 1.7291, 1.9374, 1.8376, 1.8118], device='cuda:6'), covar=tensor([0.8509, 1.4239, 1.1760, 1.1203, 1.2354, 1.7609, 1.5433, 1.3098], device='cuda:6'), in_proj_covar=tensor([0.0311, 0.0396, 0.0316, 0.0326, 0.0347, 0.0411, 0.0381, 0.0336], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-26 16:14:07,826 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.72 vs. limit=2.0 2023-04-26 16:14:15,796 INFO [finetune.py:976] (6/7) Epoch 5, batch 1250, loss[loss=0.155, simple_loss=0.2268, pruned_loss=0.04163, over 4825.00 frames. ], tot_loss[loss=0.2147, simple_loss=0.2744, pruned_loss=0.07748, over 951838.42 frames. ], batch size: 25, lr: 3.94e-03, grad_scale: 32.0 2023-04-26 16:14:23,085 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9269, 2.4491, 2.0658, 2.3117, 1.7852, 2.0561, 2.1956, 1.6730], device='cuda:6'), covar=tensor([0.2264, 0.1256, 0.0945, 0.1315, 0.3068, 0.1295, 0.1656, 0.2664], device='cuda:6'), in_proj_covar=tensor([0.0313, 0.0332, 0.0242, 0.0306, 0.0325, 0.0287, 0.0275, 0.0299], device='cuda:6'), out_proj_covar=tensor([1.2757e-04, 1.3488e-04, 9.8366e-05, 1.2305e-04, 1.3395e-04, 1.1602e-04, 1.1312e-04, 1.2091e-04], device='cuda:6') 2023-04-26 16:14:42,785 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=24192.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 16:14:49,046 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.1660, 2.8147, 0.9733, 1.3648, 1.9019, 1.3403, 3.5358, 1.9391], device='cuda:6'), covar=tensor([0.0674, 0.0662, 0.0944, 0.1392, 0.0568, 0.1043, 0.0268, 0.0621], device='cuda:6'), in_proj_covar=tensor([0.0055, 0.0071, 0.0053, 0.0050, 0.0054, 0.0055, 0.0083, 0.0053], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0008, 0.0008, 0.0008, 0.0008, 0.0011, 0.0008], device='cuda:6') 2023-04-26 16:15:00,605 INFO [finetune.py:976] (6/7) Epoch 5, batch 1300, loss[loss=0.1936, simple_loss=0.255, pruned_loss=0.06607, over 4832.00 frames. ], tot_loss[loss=0.2122, simple_loss=0.2714, pruned_loss=0.07649, over 954523.06 frames. ], batch size: 49, lr: 3.94e-03, grad_scale: 32.0 2023-04-26 16:15:07,865 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.0662, 1.4804, 1.3217, 1.7322, 1.6008, 1.9859, 1.3693, 3.4383], device='cuda:6'), covar=tensor([0.0791, 0.0837, 0.0896, 0.1357, 0.0716, 0.0587, 0.0834, 0.0168], device='cuda:6'), in_proj_covar=tensor([0.0040, 0.0040, 0.0041, 0.0046, 0.0041, 0.0041, 0.0040, 0.0062], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0017], device='cuda:6') 2023-04-26 16:15:10,680 INFO [optim.py:369] (6/7) 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 16:15:49,984 INFO [finetune.py:976] (6/7) Epoch 5, batch 1350, loss[loss=0.1977, simple_loss=0.2586, pruned_loss=0.06841, over 4865.00 frames. ], tot_loss[loss=0.2113, simple_loss=0.2708, pruned_loss=0.07594, over 955742.59 frames. ], batch size: 31, lr: 3.94e-03, grad_scale: 32.0 2023-04-26 16:16:09,272 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9766, 2.2937, 2.0654, 2.2536, 1.7610, 2.0052, 2.1860, 1.7861], device='cuda:6'), covar=tensor([0.1873, 0.0999, 0.0802, 0.1074, 0.2749, 0.1038, 0.1531, 0.2113], device='cuda:6'), in_proj_covar=tensor([0.0313, 0.0332, 0.0242, 0.0306, 0.0326, 0.0287, 0.0275, 0.0299], device='cuda:6'), out_proj_covar=tensor([1.2748e-04, 1.3492e-04, 9.8396e-05, 1.2313e-04, 1.3425e-04, 1.1594e-04, 1.1312e-04, 1.2100e-04], device='cuda:6') 2023-04-26 16:16:29,536 INFO [finetune.py:976] (6/7) Epoch 5, batch 1400, loss[loss=0.2332, simple_loss=0.3007, pruned_loss=0.08286, over 4807.00 frames. ], tot_loss[loss=0.2139, simple_loss=0.2739, pruned_loss=0.07694, over 956606.26 frames. ], batch size: 38, lr: 3.94e-03, grad_scale: 32.0 2023-04-26 16:16:33,109 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5681, 1.7038, 0.8410, 1.2894, 1.8866, 1.4617, 1.3995, 1.4277], device='cuda:6'), covar=tensor([0.0531, 0.0396, 0.0412, 0.0579, 0.0305, 0.0553, 0.0543, 0.0610], device='cuda:6'), in_proj_covar=tensor([0.0031, 0.0027, 0.0024, 0.0030, 0.0022, 0.0030, 0.0029, 0.0031], device='cuda:6'), out_proj_covar=tensor([0.0049, 0.0044, 0.0038, 0.0049, 0.0038, 0.0048, 0.0047, 0.0050], device='cuda:6') 2023-04-26 16:16:38,503 INFO [optim.py:369] (6/7) 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 16:17:19,158 INFO [finetune.py:976] (6/7) Epoch 5, batch 1450, loss[loss=0.1845, simple_loss=0.2487, pruned_loss=0.06019, over 3180.00 frames. ], tot_loss[loss=0.2176, simple_loss=0.2775, pruned_loss=0.07883, over 952445.12 frames. ], batch size: 13, lr: 3.94e-03, grad_scale: 32.0 2023-04-26 16:17:19,743 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6787, 3.7930, 0.9059, 1.8374, 2.1391, 2.5594, 2.1389, 1.0164], device='cuda:6'), covar=tensor([0.1404, 0.1020, 0.2166, 0.1425, 0.1042, 0.1132, 0.1504, 0.2072], device='cuda:6'), in_proj_covar=tensor([0.0122, 0.0264, 0.0146, 0.0128, 0.0138, 0.0160, 0.0125, 0.0128], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-26 16:18:00,683 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=24393.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 16:18:07,427 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-04-26 16:18:13,335 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.4838, 1.8885, 2.2663, 2.9380, 2.2364, 1.6903, 1.5307, 2.2754], device='cuda:6'), covar=tensor([0.4633, 0.4813, 0.2222, 0.4419, 0.4288, 0.3999, 0.6179, 0.3611], device='cuda:6'), in_proj_covar=tensor([0.0275, 0.0262, 0.0221, 0.0332, 0.0221, 0.0230, 0.0248, 0.0197], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-26 16:18:14,546 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.1845, 2.1410, 2.3437, 2.5326, 2.4426, 1.9539, 1.7355, 2.1072], device='cuda:6'), covar=tensor([0.1016, 0.0971, 0.0606, 0.0672, 0.0711, 0.1099, 0.1045, 0.0713], device='cuda:6'), in_proj_covar=tensor([0.0206, 0.0208, 0.0184, 0.0181, 0.0182, 0.0197, 0.0168, 0.0193], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-26 16:18:16,238 INFO [finetune.py:976] (6/7) Epoch 5, batch 1500, loss[loss=0.2043, simple_loss=0.2704, pruned_loss=0.06914, over 4931.00 frames. ], tot_loss[loss=0.2181, simple_loss=0.2781, pruned_loss=0.07902, over 951993.39 frames. ], batch size: 33, lr: 3.94e-03, grad_scale: 32.0 2023-04-26 16:18:25,732 INFO [optim.py:369] (6/7) 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 16:18:49,533 INFO [finetune.py:976] (6/7) Epoch 5, batch 1550, loss[loss=0.2211, simple_loss=0.2897, pruned_loss=0.07627, over 4842.00 frames. ], tot_loss[loss=0.2182, simple_loss=0.2783, pruned_loss=0.07906, over 953099.80 frames. ], batch size: 44, lr: 3.94e-03, grad_scale: 32.0 2023-04-26 16:18:59,801 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.1566, 1.9784, 1.6636, 1.6318, 2.0181, 1.6536, 2.3362, 1.4680], device='cuda:6'), covar=tensor([0.3364, 0.1383, 0.4086, 0.2600, 0.1589, 0.2084, 0.1567, 0.3922], device='cuda:6'), in_proj_covar=tensor([0.0352, 0.0357, 0.0439, 0.0371, 0.0399, 0.0386, 0.0395, 0.0421], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-26 16:19:22,932 INFO [finetune.py:976] (6/7) Epoch 5, batch 1600, loss[loss=0.1886, simple_loss=0.239, pruned_loss=0.06908, over 4829.00 frames. ], tot_loss[loss=0.2164, simple_loss=0.2763, pruned_loss=0.07822, over 955247.31 frames. ], batch size: 30, lr: 3.94e-03, grad_scale: 32.0 2023-04-26 16:19:32,012 INFO [optim.py:369] (6/7) 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 16:19:48,163 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=24547.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 16:19:56,585 INFO [finetune.py:976] (6/7) Epoch 5, batch 1650, loss[loss=0.2389, simple_loss=0.2952, pruned_loss=0.09132, over 4908.00 frames. ], tot_loss[loss=0.2134, simple_loss=0.2733, pruned_loss=0.07677, over 954630.72 frames. ], batch size: 32, lr: 3.94e-03, grad_scale: 64.0 2023-04-26 16:19:59,168 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6269, 1.4264, 1.8835, 1.8779, 1.5318, 1.2573, 1.6936, 1.0424], device='cuda:6'), covar=tensor([0.0885, 0.0959, 0.0534, 0.0817, 0.0958, 0.1351, 0.0729, 0.0998], device='cuda:6'), in_proj_covar=tensor([0.0068, 0.0077, 0.0076, 0.0070, 0.0080, 0.0096, 0.0083, 0.0079], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-26 16:20:01,610 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=24569.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 16:20:27,221 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.2620, 2.6364, 1.0691, 1.4349, 2.1166, 1.2788, 3.5961, 1.8668], device='cuda:6'), covar=tensor([0.0611, 0.0634, 0.0808, 0.1330, 0.0522, 0.1044, 0.0238, 0.0654], device='cuda:6'), in_proj_covar=tensor([0.0054, 0.0071, 0.0053, 0.0050, 0.0055, 0.0055, 0.0083, 0.0053], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0008, 0.0008, 0.0008, 0.0008, 0.0011, 0.0008], device='cuda:6') 2023-04-26 16:20:28,443 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.0891, 1.4670, 1.2294, 1.8134, 1.6222, 1.7679, 1.3507, 3.3726], device='cuda:6'), covar=tensor([0.0681, 0.0798, 0.0867, 0.1171, 0.0628, 0.0506, 0.0778, 0.0158], device='cuda:6'), in_proj_covar=tensor([0.0040, 0.0040, 0.0041, 0.0046, 0.0041, 0.0041, 0.0040, 0.0062], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0017], device='cuda:6') 2023-04-26 16:20:28,463 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=24608.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 16:20:30,138 INFO [finetune.py:976] (6/7) Epoch 5, batch 1700, loss[loss=0.1743, simple_loss=0.2429, pruned_loss=0.05288, over 4753.00 frames. ], tot_loss[loss=0.2099, simple_loss=0.2695, pruned_loss=0.07516, over 955450.01 frames. ], batch size: 26, lr: 3.94e-03, grad_scale: 64.0 2023-04-26 16:20:38,588 INFO [optim.py:369] (6/7) 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 16:20:42,277 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 5, batch 1750, loss[loss=0.2602, simple_loss=0.3132, pruned_loss=0.1036, over 4905.00 frames. ], tot_loss[loss=0.211, simple_loss=0.271, pruned_loss=0.0755, over 955319.61 frames. ], batch size: 37, lr: 3.94e-03, grad_scale: 64.0 2023-04-26 16:21:35,457 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=24693.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 16:21:48,679 INFO [finetune.py:976] (6/7) Epoch 5, batch 1800, loss[loss=0.2571, simple_loss=0.3138, pruned_loss=0.1002, over 4758.00 frames. ], tot_loss[loss=0.2138, simple_loss=0.2742, pruned_loss=0.07666, over 956048.57 frames. ], batch size: 54, lr: 3.94e-03, grad_scale: 64.0 2023-04-26 16:21:52,469 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7395, 2.1602, 0.9458, 1.1291, 1.4974, 1.0598, 2.4552, 1.3267], device='cuda:6'), covar=tensor([0.0721, 0.0588, 0.0670, 0.1344, 0.0474, 0.1072, 0.0327, 0.0724], device='cuda:6'), in_proj_covar=tensor([0.0055, 0.0071, 0.0053, 0.0050, 0.0055, 0.0055, 0.0083, 0.0053], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0008, 0.0008, 0.0008, 0.0008, 0.0011, 0.0008], device='cuda:6') 2023-04-26 16:21:57,279 INFO [optim.py:369] (6/7) 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] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=24741.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 16:22:27,728 INFO [finetune.py:976] (6/7) Epoch 5, batch 1850, loss[loss=0.2169, simple_loss=0.2931, pruned_loss=0.07037, over 4889.00 frames. ], tot_loss[loss=0.2153, simple_loss=0.2763, pruned_loss=0.07713, over 955308.16 frames. ], batch size: 43, lr: 3.94e-03, grad_scale: 64.0 2023-04-26 16:22:30,279 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=24765.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 16:23:29,633 INFO [finetune.py:976] (6/7) Epoch 5, batch 1900, loss[loss=0.2198, simple_loss=0.2852, pruned_loss=0.07721, over 4820.00 frames. ], tot_loss[loss=0.2163, simple_loss=0.2777, pruned_loss=0.07748, over 955219.86 frames. ], batch size: 47, lr: 3.94e-03, grad_scale: 64.0 2023-04-26 16:23:44,193 INFO [optim.py:369] (6/7) 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 16:23:50,618 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=24826.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 16:24:16,056 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=24848.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 16:24:31,092 INFO [finetune.py:976] (6/7) Epoch 5, batch 1950, loss[loss=0.1952, simple_loss=0.2591, pruned_loss=0.06567, over 4763.00 frames. ], tot_loss[loss=0.2149, simple_loss=0.2766, pruned_loss=0.07666, over 954597.67 frames. ], batch size: 28, lr: 3.94e-03, grad_scale: 64.0 2023-04-26 16:24:52,711 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7933, 1.7609, 1.9815, 2.2597, 2.1856, 1.7780, 1.3396, 1.8620], device='cuda:6'), covar=tensor([0.1012, 0.1210, 0.0735, 0.0615, 0.0658, 0.0970, 0.1044, 0.0682], device='cuda:6'), in_proj_covar=tensor([0.0208, 0.0210, 0.0186, 0.0182, 0.0183, 0.0198, 0.0169, 0.0195], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-26 16:24:55,130 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.2614, 1.7226, 1.5647, 2.0966, 1.7933, 2.1004, 1.4537, 4.2698], device='cuda:6'), covar=tensor([0.0689, 0.0753, 0.0791, 0.1181, 0.0643, 0.0569, 0.0771, 0.0107], device='cuda:6'), in_proj_covar=tensor([0.0040, 0.0040, 0.0041, 0.0046, 0.0041, 0.0041, 0.0040, 0.0062], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0017], device='cuda:6') 2023-04-26 16:24:58,637 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=24903.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 16:25:03,335 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=24909.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 16:25:04,919 INFO [finetune.py:976] (6/7) Epoch 5, batch 2000, loss[loss=0.1899, simple_loss=0.2536, pruned_loss=0.06309, over 4785.00 frames. ], tot_loss[loss=0.2124, simple_loss=0.2737, pruned_loss=0.07558, over 956845.24 frames. ], batch size: 29, lr: 3.94e-03, grad_scale: 64.0 2023-04-26 16:25:13,837 INFO [optim.py:369] (6/7) 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 16:25:13,933 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=24925.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 16:25:37,720 INFO [finetune.py:976] (6/7) Epoch 5, batch 2050, loss[loss=0.2006, simple_loss=0.2584, pruned_loss=0.07136, over 4899.00 frames. ], tot_loss[loss=0.2116, simple_loss=0.272, pruned_loss=0.07557, over 957840.87 frames. ], batch size: 32, lr: 3.94e-03, grad_scale: 64.0 2023-04-26 16:25:59,903 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=24995.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 16:26:10,620 INFO [finetune.py:976] (6/7) Epoch 5, batch 2100, loss[loss=0.1871, simple_loss=0.2593, pruned_loss=0.05747, over 4811.00 frames. ], tot_loss[loss=0.2094, simple_loss=0.2696, pruned_loss=0.07466, over 956456.45 frames. ], batch size: 41, lr: 3.94e-03, grad_scale: 64.0 2023-04-26 16:26:21,075 INFO [optim.py:369] (6/7) 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 16:26:40,942 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=25056.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 16:26:43,816 INFO [finetune.py:976] (6/7) Epoch 5, batch 2150, loss[loss=0.1682, simple_loss=0.2245, pruned_loss=0.05601, over 4303.00 frames. ], tot_loss[loss=0.2123, simple_loss=0.2727, pruned_loss=0.07589, over 956241.97 frames. ], batch size: 19, lr: 3.94e-03, grad_scale: 64.0 2023-04-26 16:26:46,242 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([0.2880, 1.2283, 1.3311, 0.9798, 1.2784, 1.0533, 1.5902, 1.3043], device='cuda:6'), covar=tensor([0.3166, 0.1490, 0.4254, 0.1980, 0.1257, 0.1945, 0.1394, 0.3726], device='cuda:6'), in_proj_covar=tensor([0.0348, 0.0353, 0.0434, 0.0364, 0.0394, 0.0381, 0.0393, 0.0415], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-26 16:27:04,260 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.0421, 0.9966, 1.2055, 1.1661, 1.0075, 0.8734, 0.9712, 0.5426], device='cuda:6'), covar=tensor([0.0709, 0.0822, 0.0591, 0.0805, 0.0917, 0.1347, 0.0520, 0.1066], device='cuda:6'), in_proj_covar=tensor([0.0068, 0.0077, 0.0075, 0.0069, 0.0080, 0.0095, 0.0083, 0.0078], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-26 16:27:17,133 INFO [finetune.py:976] (6/7) Epoch 5, batch 2200, loss[loss=0.2824, simple_loss=0.338, pruned_loss=0.1134, over 4249.00 frames. ], tot_loss[loss=0.2171, simple_loss=0.2774, pruned_loss=0.0784, over 955220.26 frames. ], batch size: 66, lr: 3.94e-03, grad_scale: 64.0 2023-04-26 16:27:24,281 INFO [zipformer.py:1188] (6/7) 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] (6/7) 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] (6/7) Epoch 5, batch 2250, loss[loss=0.2241, simple_loss=0.2918, pruned_loss=0.0782, over 4809.00 frames. ], tot_loss[loss=0.2201, simple_loss=0.2804, pruned_loss=0.0799, over 954760.79 frames. ], batch size: 38, lr: 3.94e-03, grad_scale: 64.0 2023-04-26 16:28:23,430 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.02 vs. limit=5.0 2023-04-26 16:28:26,336 INFO [zipformer.py:1188] (6/7) 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] (6/7) 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] (6/7) Epoch 5, batch 2300, loss[loss=0.2189, simple_loss=0.2748, pruned_loss=0.08152, over 4813.00 frames. ], tot_loss[loss=0.2183, simple_loss=0.2792, pruned_loss=0.07872, over 953543.24 frames. ], batch size: 33, lr: 3.94e-03, grad_scale: 64.0 2023-04-26 16:28:52,197 INFO [optim.py:369] (6/7) 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] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=25225.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 16:29:11,901 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5768, 1.2693, 0.6492, 1.2623, 1.4316, 1.4613, 1.3381, 1.3659], device='cuda:6'), covar=tensor([0.0557, 0.0461, 0.0455, 0.0607, 0.0326, 0.0555, 0.0553, 0.0636], device='cuda:6'), in_proj_covar=tensor([0.0031, 0.0027, 0.0024, 0.0030, 0.0022, 0.0030, 0.0030, 0.0031], device='cuda:6'), out_proj_covar=tensor([0.0049, 0.0044, 0.0038, 0.0049, 0.0038, 0.0048, 0.0047, 0.0050], device='cuda:6') 2023-04-26 16:29:26,418 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=25251.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 16:29:38,267 INFO [finetune.py:976] (6/7) Epoch 5, batch 2350, loss[loss=0.2106, simple_loss=0.2618, pruned_loss=0.07965, over 4815.00 frames. ], tot_loss[loss=0.2166, simple_loss=0.277, pruned_loss=0.07811, over 953404.80 frames. ], batch size: 51, lr: 3.94e-03, grad_scale: 64.0 2023-04-26 16:29:56,884 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.64 vs. limit=2.0 2023-04-26 16:29:57,357 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=25273.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 16:30:39,274 INFO [finetune.py:976] (6/7) Epoch 5, batch 2400, loss[loss=0.1952, simple_loss=0.2597, pruned_loss=0.0653, over 4827.00 frames. ], tot_loss[loss=0.2134, simple_loss=0.2734, pruned_loss=0.07664, over 954974.80 frames. ], batch size: 38, lr: 3.94e-03, grad_scale: 64.0 2023-04-26 16:30:48,373 INFO [optim.py:369] (6/7) 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 16:31:07,233 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-04-26 16:31:12,395 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=25351.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 16:31:18,437 INFO [finetune.py:976] (6/7) Epoch 5, batch 2450, loss[loss=0.1927, simple_loss=0.259, pruned_loss=0.06315, over 4921.00 frames. ], tot_loss[loss=0.2105, simple_loss=0.27, pruned_loss=0.07547, over 955527.81 frames. ], batch size: 36, lr: 3.94e-03, grad_scale: 64.0 2023-04-26 16:31:20,397 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=3.87 vs. limit=5.0 2023-04-26 16:32:08,174 INFO [finetune.py:976] (6/7) Epoch 5, batch 2500, loss[loss=0.2224, simple_loss=0.2938, pruned_loss=0.07543, over 4854.00 frames. ], tot_loss[loss=0.2131, simple_loss=0.2723, pruned_loss=0.07691, over 955440.43 frames. ], batch size: 44, lr: 3.94e-03, grad_scale: 64.0 2023-04-26 16:32:08,340 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-04-26 16:32:09,179 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-04-26 16:32:15,250 INFO [zipformer.py:1188] (6/7) 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] (6/7) 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 16:32:19,535 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=25428.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 16:32:41,938 INFO [finetune.py:976] (6/7) Epoch 5, batch 2550, loss[loss=0.2038, simple_loss=0.2685, pruned_loss=0.06959, over 4820.00 frames. ], tot_loss[loss=0.2153, simple_loss=0.275, pruned_loss=0.07779, over 955025.50 frames. ], batch size: 33, lr: 3.94e-03, grad_scale: 32.0 2023-04-26 16:32:47,380 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=25469.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 16:33:01,977 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=25489.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 16:33:11,439 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=25504.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 16:33:13,506 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-04-26 16:33:15,642 INFO [finetune.py:976] (6/7) Epoch 5, batch 2600, loss[loss=0.2864, simple_loss=0.3251, pruned_loss=0.1238, over 4862.00 frames. ], tot_loss[loss=0.2162, simple_loss=0.2765, pruned_loss=0.07799, over 955611.28 frames. ], batch size: 44, lr: 3.94e-03, grad_scale: 32.0 2023-04-26 16:33:22,987 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.3503, 3.3806, 0.7193, 1.8605, 1.7863, 2.3264, 1.8654, 0.9930], device='cuda:6'), covar=tensor([0.1396, 0.0989, 0.2164, 0.1263, 0.1087, 0.1096, 0.1589, 0.2026], device='cuda:6'), in_proj_covar=tensor([0.0122, 0.0264, 0.0147, 0.0129, 0.0138, 0.0161, 0.0126, 0.0129], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-26 16:33:25,301 INFO [optim.py:369] (6/7) 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 16:33:43,927 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 5, batch 2650, loss[loss=0.2677, simple_loss=0.3186, pruned_loss=0.1084, over 4810.00 frames. ], tot_loss[loss=0.2166, simple_loss=0.2773, pruned_loss=0.07798, over 956796.59 frames. ], batch size: 45, lr: 3.94e-03, grad_scale: 32.0 2023-04-26 16:34:28,766 INFO [finetune.py:976] (6/7) Epoch 5, batch 2700, loss[loss=0.2039, simple_loss=0.2616, pruned_loss=0.0731, over 4808.00 frames. ], tot_loss[loss=0.2166, simple_loss=0.2777, pruned_loss=0.07775, over 959398.36 frames. ], batch size: 40, lr: 3.94e-03, grad_scale: 32.0 2023-04-26 16:34:48,614 INFO [optim.py:369] (6/7) 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 16:35:11,085 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5942, 1.1599, 1.2673, 1.3232, 1.8471, 1.5148, 1.1958, 1.2155], device='cuda:6'), covar=tensor([0.1810, 0.1687, 0.2073, 0.1439, 0.0797, 0.1370, 0.2243, 0.2062], device='cuda:6'), in_proj_covar=tensor([0.0314, 0.0332, 0.0350, 0.0307, 0.0346, 0.0343, 0.0309, 0.0352], device='cuda:6'), out_proj_covar=tensor([6.7581e-05, 7.1227e-05, 7.5858e-05, 6.4245e-05, 7.3300e-05, 7.4635e-05, 6.6931e-05, 7.5897e-05], device='cuda:6') 2023-04-26 16:35:20,267 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=25651.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 16:35:37,125 INFO [finetune.py:976] (6/7) Epoch 5, batch 2750, loss[loss=0.2066, simple_loss=0.2754, pruned_loss=0.06892, over 4818.00 frames. ], tot_loss[loss=0.2134, simple_loss=0.2741, pruned_loss=0.07641, over 960011.23 frames. ], batch size: 40, lr: 3.94e-03, grad_scale: 32.0 2023-04-26 16:36:24,102 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=25699.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 16:36:32,333 INFO [finetune.py:976] (6/7) Epoch 5, batch 2800, loss[loss=0.1563, simple_loss=0.2244, pruned_loss=0.04408, over 4901.00 frames. ], tot_loss[loss=0.21, simple_loss=0.2702, pruned_loss=0.0749, over 960340.97 frames. ], batch size: 35, lr: 3.93e-03, grad_scale: 32.0 2023-04-26 16:36:47,385 INFO [optim.py:369] (6/7) 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 16:37:32,911 INFO [finetune.py:976] (6/7) Epoch 5, batch 2850, loss[loss=0.2394, simple_loss=0.2983, pruned_loss=0.09022, over 4899.00 frames. ], tot_loss[loss=0.2077, simple_loss=0.2678, pruned_loss=0.07376, over 961026.64 frames. ], batch size: 36, lr: 3.93e-03, grad_scale: 32.0 2023-04-26 16:37:47,001 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=25784.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 16:38:06,696 INFO [finetune.py:976] (6/7) Epoch 5, batch 2900, loss[loss=0.1704, simple_loss=0.242, pruned_loss=0.04943, over 4755.00 frames. ], tot_loss[loss=0.211, simple_loss=0.2713, pruned_loss=0.07533, over 958327.23 frames. ], batch size: 28, lr: 3.93e-03, grad_scale: 32.0 2023-04-26 16:38:15,791 INFO [optim.py:369] (6/7) 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 16:38:37,808 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=25858.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 16:38:39,945 INFO [finetune.py:976] (6/7) Epoch 5, batch 2950, loss[loss=0.2057, simple_loss=0.2724, pruned_loss=0.06951, over 4928.00 frames. ], tot_loss[loss=0.2129, simple_loss=0.2738, pruned_loss=0.07597, over 957360.81 frames. ], batch size: 38, lr: 3.93e-03, grad_scale: 32.0 2023-04-26 16:39:12,518 INFO [finetune.py:976] (6/7) Epoch 5, batch 3000, loss[loss=0.2081, simple_loss=0.2749, pruned_loss=0.07063, over 4916.00 frames. ], tot_loss[loss=0.2145, simple_loss=0.2756, pruned_loss=0.07673, over 956255.27 frames. ], batch size: 37, lr: 3.93e-03, grad_scale: 32.0 2023-04-26 16:39:12,518 INFO [finetune.py:1001] (6/7) Computing validation loss 2023-04-26 16:39:29,056 INFO [finetune.py:1010] (6/7) Epoch 5, validation: loss=0.1595, simple_loss=0.233, pruned_loss=0.04303, over 2265189.00 frames. 2023-04-26 16:39:29,057 INFO [finetune.py:1011] (6/7) Maximum memory allocated so far is 6326MB 2023-04-26 16:39:41,229 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=25919.0, num_to_drop=1, layers_to_drop={2} 2023-04-26 16:39:51,736 INFO [optim.py:369] (6/7) 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 16:39:58,631 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=3.87 vs. limit=5.0 2023-04-26 16:40:00,815 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9939, 2.4521, 0.9464, 1.1406, 1.7131, 1.1423, 3.0258, 1.3864], device='cuda:6'), covar=tensor([0.0722, 0.0566, 0.0785, 0.1405, 0.0563, 0.1089, 0.0342, 0.0780], device='cuda:6'), in_proj_covar=tensor([0.0055, 0.0071, 0.0053, 0.0050, 0.0055, 0.0055, 0.0083, 0.0053], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0008, 0.0008, 0.0008, 0.0008, 0.0011, 0.0008], device='cuda:6') 2023-04-26 16:40:35,901 INFO [finetune.py:976] (6/7) Epoch 5, batch 3050, loss[loss=0.1792, simple_loss=0.2509, pruned_loss=0.05375, over 4766.00 frames. ], tot_loss[loss=0.2138, simple_loss=0.2753, pruned_loss=0.07615, over 955585.29 frames. ], batch size: 28, lr: 3.93e-03, grad_scale: 32.0 2023-04-26 16:40:43,937 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.3732, 0.6094, 1.1600, 1.7127, 1.5118, 1.2530, 1.2265, 1.3236], device='cuda:6'), covar=tensor([0.9910, 1.3751, 1.4099, 1.5886, 1.1970, 1.5504, 1.6420, 1.2999], device='cuda:6'), in_proj_covar=tensor([0.0423, 0.0458, 0.0544, 0.0564, 0.0453, 0.0477, 0.0488, 0.0488], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-26 16:41:22,558 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2023-04-26 16:41:32,850 INFO [finetune.py:976] (6/7) Epoch 5, batch 3100, loss[loss=0.2388, simple_loss=0.2558, pruned_loss=0.1109, over 3963.00 frames. ], tot_loss[loss=0.2118, simple_loss=0.2732, pruned_loss=0.07523, over 954321.60 frames. ], batch size: 17, lr: 3.93e-03, grad_scale: 32.0 2023-04-26 16:41:43,947 INFO [optim.py:369] (6/7) 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 16:41:54,534 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6264, 1.2522, 1.6728, 1.9704, 1.6890, 1.6010, 1.6516, 1.7063], device='cuda:6'), covar=tensor([1.2201, 1.5231, 1.5724, 1.8779, 1.3383, 1.9162, 1.9450, 1.4986], device='cuda:6'), in_proj_covar=tensor([0.0421, 0.0456, 0.0542, 0.0561, 0.0452, 0.0474, 0.0486, 0.0486], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-26 16:42:03,386 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8058, 1.7164, 1.9699, 2.1462, 2.1749, 1.7418, 1.4242, 1.8022], device='cuda:6'), covar=tensor([0.0828, 0.0992, 0.0556, 0.0589, 0.0541, 0.0904, 0.0878, 0.0628], device='cuda:6'), in_proj_covar=tensor([0.0207, 0.0209, 0.0184, 0.0181, 0.0181, 0.0198, 0.0169, 0.0193], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-26 16:42:06,248 INFO [finetune.py:976] (6/7) Epoch 5, batch 3150, loss[loss=0.2104, simple_loss=0.2684, pruned_loss=0.07618, over 4914.00 frames. ], tot_loss[loss=0.2106, simple_loss=0.2713, pruned_loss=0.07498, over 955630.23 frames. ], batch size: 46, lr: 3.93e-03, grad_scale: 32.0 2023-04-26 16:42:27,114 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=26084.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 16:42:34,576 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-04-26 16:42:46,435 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=26102.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 16:42:47,049 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.1218, 1.4552, 1.3602, 1.8059, 1.5704, 1.7150, 1.3664, 3.0607], device='cuda:6'), covar=tensor([0.0687, 0.0781, 0.0839, 0.1170, 0.0682, 0.0558, 0.0741, 0.0188], device='cuda:6'), in_proj_covar=tensor([0.0039, 0.0040, 0.0041, 0.0046, 0.0041, 0.0041, 0.0040, 0.0062], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0017], device='cuda:6') 2023-04-26 16:42:54,362 INFO [finetune.py:976] (6/7) Epoch 5, batch 3200, loss[loss=0.2386, simple_loss=0.2893, pruned_loss=0.09389, over 4919.00 frames. ], tot_loss[loss=0.2083, simple_loss=0.2685, pruned_loss=0.07404, over 956041.77 frames. ], batch size: 37, lr: 3.93e-03, grad_scale: 32.0 2023-04-26 16:43:09,265 INFO [optim.py:369] (6/7) 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 16:43:14,416 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=26132.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 16:43:29,113 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7615, 2.5805, 1.8596, 1.8526, 1.3641, 1.3703, 2.0268, 1.3748], device='cuda:6'), covar=tensor([0.1977, 0.1853, 0.1746, 0.2312, 0.2992, 0.2267, 0.1354, 0.2426], device='cuda:6'), in_proj_covar=tensor([0.0196, 0.0222, 0.0180, 0.0209, 0.0216, 0.0189, 0.0172, 0.0196], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-26 16:43:31,947 INFO [finetune.py:976] (6/7) Epoch 5, batch 3250, loss[loss=0.2403, simple_loss=0.3045, pruned_loss=0.08808, over 4867.00 frames. ], tot_loss[loss=0.2096, simple_loss=0.2697, pruned_loss=0.07479, over 956312.57 frames. ], batch size: 44, lr: 3.93e-03, grad_scale: 16.0 2023-04-26 16:43:33,754 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=26163.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 16:43:37,345 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=26168.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 16:43:52,342 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2023-04-26 16:43:53,881 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.0606, 0.9965, 1.2513, 1.1918, 1.0111, 0.9059, 0.9975, 0.5308], device='cuda:6'), covar=tensor([0.0663, 0.0731, 0.0618, 0.0701, 0.0837, 0.1422, 0.0585, 0.1127], device='cuda:6'), in_proj_covar=tensor([0.0068, 0.0077, 0.0076, 0.0069, 0.0081, 0.0097, 0.0083, 0.0079], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-26 16:43:58,135 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.2162, 1.5661, 1.4564, 1.7370, 1.6050, 1.7411, 1.4540, 2.8309], device='cuda:6'), covar=tensor([0.0620, 0.0679, 0.0703, 0.1033, 0.0571, 0.0593, 0.0688, 0.0244], device='cuda:6'), in_proj_covar=tensor([0.0039, 0.0040, 0.0041, 0.0046, 0.0041, 0.0041, 0.0040, 0.0062], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0017], device='cuda:6') 2023-04-26 16:44:03,122 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.4069, 4.0790, 0.6361, 1.8601, 1.9576, 2.5936, 2.0511, 0.9654], device='cuda:6'), covar=tensor([0.1944, 0.1958, 0.3058, 0.2203, 0.1464, 0.1560, 0.2163, 0.2506], device='cuda:6'), in_proj_covar=tensor([0.0122, 0.0262, 0.0147, 0.0129, 0.0138, 0.0160, 0.0125, 0.0128], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-26 16:44:05,485 INFO [finetune.py:976] (6/7) Epoch 5, batch 3300, loss[loss=0.2138, simple_loss=0.2873, pruned_loss=0.07019, over 4818.00 frames. ], tot_loss[loss=0.2131, simple_loss=0.2734, pruned_loss=0.0764, over 954438.29 frames. ], batch size: 51, lr: 3.93e-03, grad_scale: 16.0 2023-04-26 16:44:07,401 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=26214.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 16:44:15,061 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.5635, 1.9132, 1.6839, 1.8953, 1.7724, 1.9702, 1.7693, 1.7416], device='cuda:6'), covar=tensor([0.7181, 1.1637, 1.0281, 0.9037, 1.0053, 1.2928, 1.2939, 1.1318], device='cuda:6'), in_proj_covar=tensor([0.0314, 0.0396, 0.0317, 0.0327, 0.0347, 0.0411, 0.0379, 0.0336], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-26 16:44:16,102 INFO [optim.py:369] (6/7) 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 16:44:17,420 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=26229.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 16:44:32,051 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=26250.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 16:44:38,664 INFO [finetune.py:976] (6/7) Epoch 5, batch 3350, loss[loss=0.1831, simple_loss=0.2473, pruned_loss=0.0594, over 4840.00 frames. ], tot_loss[loss=0.2156, simple_loss=0.2764, pruned_loss=0.07733, over 956557.01 frames. ], batch size: 44, lr: 3.93e-03, grad_scale: 16.0 2023-04-26 16:45:50,831 INFO [finetune.py:976] (6/7) Epoch 5, batch 3400, loss[loss=0.1881, simple_loss=0.2561, pruned_loss=0.06008, over 4810.00 frames. ], tot_loss[loss=0.2168, simple_loss=0.2782, pruned_loss=0.07767, over 958279.55 frames. ], batch size: 39, lr: 3.93e-03, grad_scale: 16.0 2023-04-26 16:45:50,957 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=26311.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 16:45:53,848 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-04-26 16:46:13,228 INFO [optim.py:369] (6/7) 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 16:46:45,741 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=3.90 vs. limit=5.0 2023-04-26 16:46:56,944 INFO [finetune.py:976] (6/7) Epoch 5, batch 3450, loss[loss=0.2163, simple_loss=0.2759, pruned_loss=0.07835, over 4908.00 frames. ], tot_loss[loss=0.2149, simple_loss=0.2765, pruned_loss=0.07665, over 956387.32 frames. ], batch size: 46, lr: 3.93e-03, grad_scale: 16.0 2023-04-26 16:46:58,949 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7538, 2.4959, 1.7474, 1.6018, 1.2727, 1.3819, 1.8497, 1.3166], device='cuda:6'), covar=tensor([0.1979, 0.1526, 0.1757, 0.2296, 0.2760, 0.2236, 0.1284, 0.2311], device='cuda:6'), in_proj_covar=tensor([0.0196, 0.0222, 0.0180, 0.0209, 0.0216, 0.0189, 0.0172, 0.0197], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-26 16:48:03,180 INFO [finetune.py:976] (6/7) Epoch 5, batch 3500, loss[loss=0.2025, simple_loss=0.2621, pruned_loss=0.07143, over 4911.00 frames. ], tot_loss[loss=0.2118, simple_loss=0.273, pruned_loss=0.07528, over 957828.87 frames. ], batch size: 36, lr: 3.93e-03, grad_scale: 16.0 2023-04-26 16:48:03,908 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6932, 1.2549, 1.5394, 1.5856, 1.4505, 1.1942, 0.6267, 1.1801], device='cuda:6'), covar=tensor([0.4527, 0.5097, 0.2367, 0.3361, 0.4053, 0.3682, 0.6484, 0.3818], device='cuda:6'), in_proj_covar=tensor([0.0276, 0.0263, 0.0222, 0.0332, 0.0221, 0.0231, 0.0247, 0.0197], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-26 16:48:05,092 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=26414.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 16:48:25,047 INFO [optim.py:369] (6/7) 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 16:49:08,577 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=26458.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 16:49:08,658 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.4708, 1.6567, 1.7548, 1.9061, 1.7741, 1.8579, 1.8505, 1.8276], device='cuda:6'), covar=tensor([0.7801, 1.1956, 0.9382, 0.8829, 1.0048, 1.5204, 1.1942, 1.0184], device='cuda:6'), in_proj_covar=tensor([0.0312, 0.0394, 0.0316, 0.0324, 0.0345, 0.0409, 0.0378, 0.0334], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-26 16:49:10,354 INFO [finetune.py:976] (6/7) Epoch 5, batch 3550, loss[loss=0.1662, simple_loss=0.2343, pruned_loss=0.04904, over 4792.00 frames. ], tot_loss[loss=0.2088, simple_loss=0.2694, pruned_loss=0.07413, over 954461.67 frames. ], batch size: 29, lr: 3.93e-03, grad_scale: 16.0 2023-04-26 16:49:24,644 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=26475.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 16:49:38,863 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([4.4477, 4.3807, 2.9801, 5.1431, 4.4507, 4.4743, 1.8402, 4.4179], device='cuda:6'), covar=tensor([0.1717, 0.1257, 0.3399, 0.1087, 0.4912, 0.1767, 0.6050, 0.2358], device='cuda:6'), in_proj_covar=tensor([0.0249, 0.0223, 0.0257, 0.0313, 0.0308, 0.0258, 0.0277, 0.0281], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-26 16:49:49,838 INFO [finetune.py:976] (6/7) Epoch 5, batch 3600, loss[loss=0.2278, simple_loss=0.2843, pruned_loss=0.08565, over 4863.00 frames. ], tot_loss[loss=0.2075, simple_loss=0.2674, pruned_loss=0.0738, over 954694.62 frames. ], batch size: 34, lr: 3.93e-03, grad_scale: 16.0 2023-04-26 16:49:51,755 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=26514.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 16:49:57,836 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=26524.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 16:49:57,936 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.55 vs. limit=5.0 2023-04-26 16:49:59,590 INFO [optim.py:369] (6/7) 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 16:50:09,812 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6872, 2.1028, 1.0671, 1.6009, 2.3581, 1.7136, 1.6076, 1.7492], device='cuda:6'), covar=tensor([0.0543, 0.0384, 0.0370, 0.0601, 0.0238, 0.0563, 0.0544, 0.0600], device='cuda:6'), in_proj_covar=tensor([0.0031, 0.0027, 0.0024, 0.0030, 0.0021, 0.0030, 0.0029, 0.0031], device='cuda:6'), out_proj_covar=tensor([0.0049, 0.0044, 0.0038, 0.0049, 0.0038, 0.0048, 0.0047, 0.0050], device='cuda:6') 2023-04-26 16:50:23,344 INFO [finetune.py:976] (6/7) Epoch 5, batch 3650, loss[loss=0.2327, simple_loss=0.291, pruned_loss=0.08717, over 4842.00 frames. ], tot_loss[loss=0.2093, simple_loss=0.2693, pruned_loss=0.07465, over 954111.27 frames. ], batch size: 30, lr: 3.93e-03, grad_scale: 16.0 2023-04-26 16:50:24,021 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=26562.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 16:50:53,164 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 5, batch 3700, loss[loss=0.2228, simple_loss=0.2995, pruned_loss=0.07305, over 4805.00 frames. ], tot_loss[loss=0.2126, simple_loss=0.2731, pruned_loss=0.07602, over 953984.88 frames. ], batch size: 45, lr: 3.93e-03, grad_scale: 16.0 2023-04-26 16:51:06,781 INFO [optim.py:369] (6/7) 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 16:51:29,840 INFO [finetune.py:976] (6/7) Epoch 5, batch 3750, loss[loss=0.2476, simple_loss=0.3017, pruned_loss=0.09679, over 4699.00 frames. ], tot_loss[loss=0.2149, simple_loss=0.2755, pruned_loss=0.07717, over 953389.23 frames. ], batch size: 23, lr: 3.93e-03, grad_scale: 16.0 2023-04-26 16:52:12,296 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.3289, 3.4859, 0.9722, 1.8794, 2.0248, 2.4035, 1.9223, 1.0318], device='cuda:6'), covar=tensor([0.1514, 0.0902, 0.2079, 0.1339, 0.1019, 0.1115, 0.1564, 0.2060], device='cuda:6'), in_proj_covar=tensor([0.0121, 0.0258, 0.0145, 0.0127, 0.0137, 0.0158, 0.0123, 0.0127], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-26 16:52:20,497 INFO [finetune.py:976] (6/7) Epoch 5, batch 3800, loss[loss=0.1964, simple_loss=0.2689, pruned_loss=0.06193, over 4922.00 frames. ], tot_loss[loss=0.2165, simple_loss=0.2772, pruned_loss=0.07786, over 953421.02 frames. ], batch size: 42, lr: 3.93e-03, grad_scale: 16.0 2023-04-26 16:52:31,669 INFO [optim.py:369] (6/7) 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 16:52:36,566 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.7713, 1.9999, 5.5816, 5.2008, 4.8282, 5.2058, 4.8796, 5.0526], device='cuda:6'), covar=tensor([0.5188, 0.5302, 0.0840, 0.1593, 0.0960, 0.0954, 0.0918, 0.1403], device='cuda:6'), in_proj_covar=tensor([0.0314, 0.0306, 0.0417, 0.0420, 0.0356, 0.0409, 0.0321, 0.0377], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-26 16:52:43,163 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.0734, 1.3599, 1.2420, 1.6267, 1.4641, 1.6151, 1.3474, 2.3990], device='cuda:6'), covar=tensor([0.0644, 0.0741, 0.0808, 0.1119, 0.0600, 0.0565, 0.0683, 0.0236], device='cuda:6'), in_proj_covar=tensor([0.0039, 0.0040, 0.0041, 0.0045, 0.0041, 0.0041, 0.0040, 0.0061], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0017], device='cuda:6') 2023-04-26 16:52:52,596 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=26758.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 16:52:54,315 INFO [finetune.py:976] (6/7) Epoch 5, batch 3850, loss[loss=0.227, simple_loss=0.2755, pruned_loss=0.08923, over 4902.00 frames. ], tot_loss[loss=0.2148, simple_loss=0.2756, pruned_loss=0.07699, over 953011.53 frames. ], batch size: 35, lr: 3.93e-03, grad_scale: 16.0 2023-04-26 16:53:00,826 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=26770.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 16:53:41,437 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=26806.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 16:53:44,002 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-04-26 16:53:44,436 INFO [finetune.py:976] (6/7) Epoch 5, batch 3900, loss[loss=0.185, simple_loss=0.2433, pruned_loss=0.06329, over 4902.00 frames. ], tot_loss[loss=0.2117, simple_loss=0.2721, pruned_loss=0.07568, over 953905.52 frames. ], batch size: 43, lr: 3.93e-03, grad_scale: 16.0 2023-04-26 16:54:03,741 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=26824.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 16:54:05,469 INFO [optim.py:369] (6/7) 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 16:54:12,377 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7085, 1.3136, 4.2166, 3.9183, 3.7568, 3.9182, 3.9212, 3.7746], device='cuda:6'), covar=tensor([0.6631, 0.6101, 0.0976, 0.1654, 0.1150, 0.1509, 0.2162, 0.1318], device='cuda:6'), in_proj_covar=tensor([0.0313, 0.0306, 0.0415, 0.0419, 0.0354, 0.0407, 0.0321, 0.0376], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-26 16:54:50,053 INFO [finetune.py:976] (6/7) Epoch 5, batch 3950, loss[loss=0.1904, simple_loss=0.2475, pruned_loss=0.06667, over 4747.00 frames. ], tot_loss[loss=0.208, simple_loss=0.2682, pruned_loss=0.07394, over 954179.13 frames. ], batch size: 27, lr: 3.93e-03, grad_scale: 16.0 2023-04-26 16:54:58,573 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.38 vs. limit=5.0 2023-04-26 16:55:08,486 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=26872.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 16:55:20,207 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.1734, 2.7508, 1.0662, 1.3920, 2.1124, 1.3367, 3.7835, 1.7286], device='cuda:6'), covar=tensor([0.0714, 0.1238, 0.1022, 0.1340, 0.0573, 0.1040, 0.0268, 0.0690], device='cuda:6'), in_proj_covar=tensor([0.0054, 0.0070, 0.0052, 0.0049, 0.0054, 0.0055, 0.0082, 0.0053], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0011, 0.0008], device='cuda:6') 2023-04-26 16:55:40,394 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8570, 2.6850, 1.7517, 2.0422, 1.6160, 1.5491, 1.9093, 1.5035], device='cuda:6'), covar=tensor([0.1791, 0.1494, 0.1939, 0.1826, 0.2683, 0.2427, 0.1212, 0.2111], device='cuda:6'), in_proj_covar=tensor([0.0194, 0.0220, 0.0178, 0.0208, 0.0214, 0.0188, 0.0170, 0.0195], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-26 16:55:47,044 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=26904.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 16:55:48,206 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 5, batch 4000, loss[loss=0.2498, simple_loss=0.3076, pruned_loss=0.09598, over 4760.00 frames. ], tot_loss[loss=0.2085, simple_loss=0.2683, pruned_loss=0.07441, over 953621.00 frames. ], batch size: 59, lr: 3.93e-03, grad_scale: 16.0 2023-04-26 16:55:54,779 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7243, 1.5592, 1.8531, 1.9946, 1.9922, 1.5146, 1.2808, 1.7948], device='cuda:6'), covar=tensor([0.0980, 0.1240, 0.0648, 0.0667, 0.0714, 0.1153, 0.0997, 0.0701], device='cuda:6'), in_proj_covar=tensor([0.0203, 0.0206, 0.0182, 0.0178, 0.0177, 0.0194, 0.0165, 0.0190], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-26 16:56:02,927 INFO [optim.py:369] (6/7) 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 16:56:17,502 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=26950.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 16:56:20,384 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=26954.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 16:56:25,067 INFO [finetune.py:976] (6/7) Epoch 5, batch 4050, loss[loss=0.2198, simple_loss=0.266, pruned_loss=0.08682, over 4283.00 frames. ], tot_loss[loss=0.2126, simple_loss=0.2721, pruned_loss=0.07655, over 953149.89 frames. ], batch size: 65, lr: 3.93e-03, grad_scale: 16.0 2023-04-26 16:56:28,124 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=26965.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 16:56:31,669 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5743, 2.4331, 1.4352, 1.5305, 1.2533, 1.2375, 1.4667, 1.1357], device='cuda:6'), covar=tensor([0.2303, 0.1692, 0.2288, 0.2441, 0.3185, 0.2818, 0.1571, 0.2540], device='cuda:6'), in_proj_covar=tensor([0.0196, 0.0221, 0.0179, 0.0210, 0.0216, 0.0189, 0.0172, 0.0196], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-26 16:56:58,986 INFO [finetune.py:976] (6/7) Epoch 5, batch 4100, loss[loss=0.2554, simple_loss=0.308, pruned_loss=0.1014, over 4917.00 frames. ], tot_loss[loss=0.2163, simple_loss=0.276, pruned_loss=0.07825, over 954150.19 frames. ], batch size: 33, lr: 3.93e-03, grad_scale: 16.0 2023-04-26 16:56:59,092 INFO [zipformer.py:1188] (6/7) 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] (6/7) 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 16:58:04,490 INFO [finetune.py:976] (6/7) Epoch 5, batch 4150, loss[loss=0.2791, simple_loss=0.334, pruned_loss=0.1121, over 4805.00 frames. ], tot_loss[loss=0.2165, simple_loss=0.2761, pruned_loss=0.07841, over 952620.69 frames. ], batch size: 39, lr: 3.93e-03, grad_scale: 16.0 2023-04-26 16:58:16,102 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=27070.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 16:58:47,720 INFO [finetune.py:976] (6/7) Epoch 5, batch 4200, loss[loss=0.2208, simple_loss=0.2955, pruned_loss=0.07307, over 4899.00 frames. ], tot_loss[loss=0.2158, simple_loss=0.2766, pruned_loss=0.07754, over 954163.23 frames. ], batch size: 37, lr: 3.93e-03, grad_scale: 16.0 2023-04-26 16:58:53,015 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=27118.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 16:58:58,900 INFO [optim.py:369] (6/7) 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 16:59:07,616 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.1186, 2.5505, 1.0305, 1.4157, 1.9607, 1.3627, 3.6542, 1.8964], device='cuda:6'), covar=tensor([0.0686, 0.0754, 0.0841, 0.1305, 0.0556, 0.0962, 0.0215, 0.0637], device='cuda:6'), in_proj_covar=tensor([0.0054, 0.0070, 0.0052, 0.0049, 0.0054, 0.0055, 0.0082, 0.0053], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0011, 0.0008], device='cuda:6') 2023-04-26 16:59:13,601 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=27148.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 16:59:21,418 INFO [finetune.py:976] (6/7) Epoch 5, batch 4250, loss[loss=0.1994, simple_loss=0.2604, pruned_loss=0.06918, over 4772.00 frames. ], tot_loss[loss=0.2125, simple_loss=0.2728, pruned_loss=0.07609, over 953734.21 frames. ], batch size: 54, lr: 3.93e-03, grad_scale: 16.0 2023-04-26 16:59:54,268 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=27209.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 16:59:55,346 INFO [finetune.py:976] (6/7) Epoch 5, batch 4300, loss[loss=0.1863, simple_loss=0.2513, pruned_loss=0.06061, over 4892.00 frames. ], tot_loss[loss=0.2096, simple_loss=0.2698, pruned_loss=0.07473, over 955347.63 frames. ], batch size: 32, lr: 3.93e-03, grad_scale: 16.0 2023-04-26 17:00:15,527 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([0.5383, 1.7629, 1.7446, 1.2993, 1.7823, 1.4554, 2.3091, 1.5242], device='cuda:6'), covar=tensor([0.4517, 0.1792, 0.5322, 0.3064, 0.1877, 0.2526, 0.1639, 0.4962], device='cuda:6'), in_proj_covar=tensor([0.0350, 0.0353, 0.0435, 0.0367, 0.0394, 0.0383, 0.0390, 0.0418], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-26 17:00:17,881 INFO [optim.py:369] (6/7) 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 17:01:04,150 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=27260.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 17:01:04,677 INFO [finetune.py:976] (6/7) Epoch 5, batch 4350, loss[loss=0.2532, simple_loss=0.2988, pruned_loss=0.1039, over 4794.00 frames. ], tot_loss[loss=0.2074, simple_loss=0.2674, pruned_loss=0.07368, over 955866.52 frames. ], batch size: 51, lr: 3.93e-03, grad_scale: 16.0 2023-04-26 17:01:46,636 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=27306.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 17:01:49,557 INFO [finetune.py:976] (6/7) Epoch 5, batch 4400, loss[loss=0.2058, simple_loss=0.2676, pruned_loss=0.072, over 4774.00 frames. ], tot_loss[loss=0.2084, simple_loss=0.2684, pruned_loss=0.07418, over 956858.31 frames. ], batch size: 26, lr: 3.93e-03, grad_scale: 16.0 2023-04-26 17:02:00,136 INFO [optim.py:369] (6/7) 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 17:02:06,729 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([3.7309, 3.5768, 2.7038, 4.2772, 3.6960, 3.7087, 1.7560, 3.6644], device='cuda:6'), covar=tensor([0.1739, 0.1334, 0.3286, 0.1896, 0.3419, 0.1867, 0.5972, 0.2538], device='cuda:6'), in_proj_covar=tensor([0.0249, 0.0222, 0.0255, 0.0312, 0.0305, 0.0257, 0.0277, 0.0278], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-26 17:02:23,093 INFO [finetune.py:976] (6/7) Epoch 5, batch 4450, loss[loss=0.2797, simple_loss=0.3337, pruned_loss=0.1128, over 4848.00 frames. ], tot_loss[loss=0.2128, simple_loss=0.2737, pruned_loss=0.07597, over 956854.88 frames. ], batch size: 49, lr: 3.93e-03, grad_scale: 16.0 2023-04-26 17:02:26,981 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-04-26 17:02:40,133 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 2023-04-26 17:02:50,043 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.3370, 1.6420, 2.1861, 2.8263, 2.1506, 1.6652, 1.4793, 2.0802], device='cuda:6'), covar=tensor([0.4088, 0.4673, 0.2145, 0.3444, 0.3965, 0.3627, 0.5485, 0.3574], device='cuda:6'), in_proj_covar=tensor([0.0277, 0.0260, 0.0222, 0.0331, 0.0220, 0.0230, 0.0245, 0.0196], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-26 17:02:56,797 INFO [finetune.py:976] (6/7) Epoch 5, batch 4500, loss[loss=0.1641, simple_loss=0.2291, pruned_loss=0.04956, over 4750.00 frames. ], tot_loss[loss=0.214, simple_loss=0.2752, pruned_loss=0.0764, over 956537.11 frames. ], batch size: 28, lr: 3.93e-03, grad_scale: 16.0 2023-04-26 17:03:04,650 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([0.7838, 1.7494, 1.5260, 1.3253, 1.7484, 1.4387, 2.0530, 1.2569], device='cuda:6'), covar=tensor([0.3943, 0.1582, 0.4777, 0.3168, 0.1842, 0.2215, 0.1873, 0.4657], device='cuda:6'), in_proj_covar=tensor([0.0352, 0.0354, 0.0437, 0.0367, 0.0396, 0.0384, 0.0392, 0.0420], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-26 17:03:17,442 INFO [optim.py:369] (6/7) 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 17:04:02,849 INFO [finetune.py:976] (6/7) Epoch 5, batch 4550, loss[loss=0.2598, simple_loss=0.3042, pruned_loss=0.1078, over 4130.00 frames. ], tot_loss[loss=0.2147, simple_loss=0.2762, pruned_loss=0.07663, over 954594.79 frames. ], batch size: 66, lr: 3.93e-03, grad_scale: 16.0 2023-04-26 17:04:11,971 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.2655, 1.2252, 3.8154, 3.5402, 3.4259, 3.6559, 3.6826, 3.3815], device='cuda:6'), covar=tensor([0.7132, 0.5772, 0.1171, 0.1945, 0.1105, 0.1693, 0.1369, 0.1442], device='cuda:6'), in_proj_covar=tensor([0.0315, 0.0310, 0.0418, 0.0424, 0.0355, 0.0410, 0.0321, 0.0379], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-26 17:04:31,479 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=27504.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 17:04:36,155 INFO [finetune.py:976] (6/7) Epoch 5, batch 4600, loss[loss=0.2518, simple_loss=0.2908, pruned_loss=0.1065, over 4340.00 frames. ], tot_loss[loss=0.2132, simple_loss=0.2746, pruned_loss=0.07592, over 955205.24 frames. ], batch size: 19, lr: 3.93e-03, grad_scale: 16.0 2023-04-26 17:04:46,281 INFO [optim.py:369] (6/7) 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 17:04:59,748 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-04-26 17:05:08,957 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=27560.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 17:05:09,444 INFO [finetune.py:976] (6/7) Epoch 5, batch 4650, loss[loss=0.2372, simple_loss=0.2917, pruned_loss=0.09134, over 4838.00 frames. ], tot_loss[loss=0.2091, simple_loss=0.2705, pruned_loss=0.07383, over 955232.80 frames. ], batch size: 47, lr: 3.93e-03, grad_scale: 16.0 2023-04-26 17:05:40,068 INFO [zipformer.py:1188] (6/7) 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] (6/7) 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] (6/7) Epoch 5, batch 4700, loss[loss=0.165, simple_loss=0.2283, pruned_loss=0.05091, over 4874.00 frames. ], tot_loss[loss=0.2075, simple_loss=0.2681, pruned_loss=0.07343, over 954798.52 frames. ], batch size: 31, lr: 3.93e-03, grad_scale: 16.0 2023-04-26 17:05:54,153 INFO [optim.py:369] (6/7) 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 17:06:23,335 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=27654.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 17:06:34,204 INFO [finetune.py:976] (6/7) Epoch 5, batch 4750, loss[loss=0.1981, simple_loss=0.26, pruned_loss=0.06814, over 4871.00 frames. ], tot_loss[loss=0.2074, simple_loss=0.2673, pruned_loss=0.0738, over 952687.52 frames. ], batch size: 31, lr: 3.93e-03, grad_scale: 16.0 2023-04-26 17:06:54,222 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=3.61 vs. limit=5.0 2023-04-26 17:07:04,531 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([3.5839, 3.5801, 2.6942, 4.1908, 3.6259, 3.6183, 1.6479, 3.6411], device='cuda:6'), covar=tensor([0.1850, 0.1260, 0.3596, 0.1592, 0.4177, 0.1753, 0.5466, 0.2197], device='cuda:6'), in_proj_covar=tensor([0.0248, 0.0222, 0.0255, 0.0314, 0.0306, 0.0257, 0.0278, 0.0278], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-26 17:07:40,896 INFO [finetune.py:976] (6/7) Epoch 5, batch 4800, loss[loss=0.2267, simple_loss=0.2854, pruned_loss=0.084, over 4861.00 frames. ], tot_loss[loss=0.2098, simple_loss=0.2702, pruned_loss=0.07476, over 955160.14 frames. ], batch size: 34, lr: 3.93e-03, grad_scale: 16.0 2023-04-26 17:07:49,090 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=27714.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 17:08:02,309 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0180, 2.7726, 2.1722, 2.5532, 1.9520, 2.3474, 2.5209, 1.8850], device='cuda:6'), covar=tensor([0.2734, 0.1387, 0.1077, 0.1772, 0.3630, 0.1440, 0.2415, 0.3331], device='cuda:6'), in_proj_covar=tensor([0.0307, 0.0324, 0.0235, 0.0300, 0.0319, 0.0279, 0.0268, 0.0291], device='cuda:6'), out_proj_covar=tensor([1.2491e-04, 1.3161e-04, 9.5849e-05, 1.2045e-04, 1.3107e-04, 1.1302e-04, 1.1017e-04, 1.1755e-04], device='cuda:6') 2023-04-26 17:08:02,772 INFO [optim.py:369] (6/7) 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 17:08:41,288 INFO [finetune.py:976] (6/7) Epoch 5, batch 4850, loss[loss=0.2334, simple_loss=0.3067, pruned_loss=0.08011, over 4819.00 frames. ], tot_loss[loss=0.2117, simple_loss=0.2731, pruned_loss=0.07512, over 953926.70 frames. ], batch size: 39, lr: 3.92e-03, grad_scale: 16.0 2023-04-26 17:08:50,922 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=27775.0, num_to_drop=1, layers_to_drop={2} 2023-04-26 17:09:01,191 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([0.5939, 1.7261, 1.5743, 1.2108, 1.7218, 1.4018, 2.2026, 1.3570], device='cuda:6'), covar=tensor([0.4229, 0.1712, 0.5129, 0.3336, 0.1869, 0.2556, 0.1605, 0.4615], device='cuda:6'), in_proj_covar=tensor([0.0349, 0.0353, 0.0434, 0.0366, 0.0395, 0.0383, 0.0392, 0.0419], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-26 17:09:14,405 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=27804.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 17:09:18,622 INFO [finetune.py:976] (6/7) Epoch 5, batch 4900, loss[loss=0.2154, simple_loss=0.2722, pruned_loss=0.07927, over 4778.00 frames. ], tot_loss[loss=0.2127, simple_loss=0.2742, pruned_loss=0.07563, over 952905.19 frames. ], batch size: 26, lr: 3.92e-03, grad_scale: 16.0 2023-04-26 17:09:30,228 INFO [optim.py:369] (6/7) 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 17:09:46,108 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=27852.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 17:09:52,537 INFO [finetune.py:976] (6/7) Epoch 5, batch 4950, loss[loss=0.2542, simple_loss=0.2998, pruned_loss=0.1043, over 4889.00 frames. ], tot_loss[loss=0.2126, simple_loss=0.2745, pruned_loss=0.07538, over 952222.52 frames. ], batch size: 32, lr: 3.92e-03, grad_scale: 16.0 2023-04-26 17:10:08,784 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6752, 1.6556, 1.0334, 1.3881, 1.8051, 1.5347, 1.4353, 1.5050], device='cuda:6'), covar=tensor([0.0547, 0.0438, 0.0390, 0.0617, 0.0304, 0.0580, 0.0573, 0.0640], device='cuda:6'), in_proj_covar=tensor([0.0031, 0.0027, 0.0024, 0.0030, 0.0021, 0.0030, 0.0030, 0.0031], device='cuda:6'), out_proj_covar=tensor([0.0049, 0.0044, 0.0038, 0.0049, 0.0038, 0.0048, 0.0047, 0.0050], device='cuda:6') 2023-04-26 17:10:31,677 INFO [finetune.py:976] (6/7) Epoch 5, batch 5000, loss[loss=0.1704, simple_loss=0.2328, pruned_loss=0.05398, over 4733.00 frames. ], tot_loss[loss=0.2105, simple_loss=0.272, pruned_loss=0.07445, over 953112.64 frames. ], batch size: 23, lr: 3.92e-03, grad_scale: 16.0 2023-04-26 17:10:42,356 INFO [optim.py:369] (6/7) 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 17:10:56,710 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8778, 1.4020, 1.3495, 1.6404, 2.0776, 1.6930, 1.3687, 1.3270], device='cuda:6'), covar=tensor([0.1317, 0.1801, 0.2163, 0.1399, 0.1019, 0.1869, 0.2615, 0.2297], device='cuda:6'), in_proj_covar=tensor([0.0311, 0.0333, 0.0350, 0.0308, 0.0343, 0.0337, 0.0308, 0.0351], device='cuda:6'), out_proj_covar=tensor([6.6866e-05, 7.1307e-05, 7.5853e-05, 6.4257e-05, 7.2551e-05, 7.3274e-05, 6.6852e-05, 7.5589e-05], device='cuda:6') 2023-04-26 17:10:57,900 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0311, 2.3491, 2.3462, 2.7590, 2.5409, 2.7414, 2.1508, 4.8270], device='cuda:6'), covar=tensor([0.0562, 0.0635, 0.0696, 0.0969, 0.0538, 0.0497, 0.0645, 0.0140], device='cuda:6'), in_proj_covar=tensor([0.0039, 0.0040, 0.0041, 0.0045, 0.0041, 0.0041, 0.0040, 0.0062], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0017], device='cuda:6') 2023-04-26 17:11:04,346 INFO [finetune.py:976] (6/7) Epoch 5, batch 5050, loss[loss=0.1824, simple_loss=0.238, pruned_loss=0.06343, over 4789.00 frames. ], tot_loss[loss=0.2084, simple_loss=0.2691, pruned_loss=0.0738, over 954804.16 frames. ], batch size: 26, lr: 3.92e-03, grad_scale: 16.0 2023-04-26 17:11:04,940 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5871, 1.4690, 0.7477, 1.2966, 1.6909, 1.4451, 1.3674, 1.4242], device='cuda:6'), covar=tensor([0.0533, 0.0440, 0.0429, 0.0599, 0.0296, 0.0566, 0.0548, 0.0657], device='cuda:6'), in_proj_covar=tensor([0.0031, 0.0027, 0.0024, 0.0030, 0.0021, 0.0030, 0.0029, 0.0031], device='cuda:6'), out_proj_covar=tensor([0.0049, 0.0044, 0.0038, 0.0049, 0.0038, 0.0048, 0.0047, 0.0050], device='cuda:6') 2023-04-26 17:12:05,263 INFO [finetune.py:976] (6/7) Epoch 5, batch 5100, loss[loss=0.1812, simple_loss=0.2512, pruned_loss=0.05563, over 4819.00 frames. ], tot_loss[loss=0.2058, simple_loss=0.2662, pruned_loss=0.07274, over 955771.41 frames. ], batch size: 40, lr: 3.92e-03, grad_scale: 16.0 2023-04-26 17:12:27,655 INFO [optim.py:369] (6/7) 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 17:12:38,246 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.2649, 1.4614, 1.3308, 1.4832, 1.2459, 1.2798, 1.3405, 1.1352], device='cuda:6'), covar=tensor([0.1395, 0.1195, 0.0815, 0.0968, 0.2485, 0.1149, 0.1417, 0.1760], device='cuda:6'), in_proj_covar=tensor([0.0306, 0.0325, 0.0236, 0.0299, 0.0319, 0.0279, 0.0268, 0.0291], device='cuda:6'), out_proj_covar=tensor([1.2455e-04, 1.3184e-04, 9.6096e-05, 1.2014e-04, 1.3125e-04, 1.1279e-04, 1.1042e-04, 1.1733e-04], device='cuda:6') 2023-04-26 17:12:38,927 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 2023-04-26 17:13:12,920 INFO [finetune.py:976] (6/7) Epoch 5, batch 5150, loss[loss=0.2394, simple_loss=0.3014, pruned_loss=0.08867, over 4720.00 frames. ], tot_loss[loss=0.2066, simple_loss=0.2671, pruned_loss=0.07306, over 954519.43 frames. ], batch size: 54, lr: 3.92e-03, grad_scale: 16.0 2023-04-26 17:13:31,479 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=28070.0, num_to_drop=1, layers_to_drop={3} 2023-04-26 17:13:31,513 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.1533, 1.4549, 1.2621, 1.7212, 1.5280, 1.9550, 1.3177, 3.3764], device='cuda:6'), covar=tensor([0.0711, 0.0809, 0.0845, 0.1214, 0.0678, 0.0615, 0.0772, 0.0169], device='cuda:6'), in_proj_covar=tensor([0.0039, 0.0040, 0.0041, 0.0045, 0.0041, 0.0040, 0.0040, 0.0061], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0017], device='cuda:6') 2023-04-26 17:14:09,043 INFO [finetune.py:976] (6/7) Epoch 5, batch 5200, loss[loss=0.1888, simple_loss=0.2399, pruned_loss=0.06887, over 3977.00 frames. ], tot_loss[loss=0.211, simple_loss=0.2716, pruned_loss=0.07519, over 953189.12 frames. ], batch size: 17, lr: 3.92e-03, grad_scale: 16.0 2023-04-26 17:14:19,885 INFO [optim.py:369] (6/7) 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 17:14:42,336 INFO [finetune.py:976] (6/7) Epoch 5, batch 5250, loss[loss=0.1833, simple_loss=0.2329, pruned_loss=0.06683, over 4746.00 frames. ], tot_loss[loss=0.2123, simple_loss=0.2735, pruned_loss=0.07562, over 953849.86 frames. ], batch size: 23, lr: 3.92e-03, grad_scale: 32.0 2023-04-26 17:14:52,008 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8266, 1.2528, 1.3896, 1.4226, 1.9831, 1.5801, 1.2354, 1.3111], device='cuda:6'), covar=tensor([0.1748, 0.1903, 0.2249, 0.1483, 0.0963, 0.1703, 0.2772, 0.2166], device='cuda:6'), in_proj_covar=tensor([0.0313, 0.0334, 0.0351, 0.0307, 0.0344, 0.0338, 0.0309, 0.0353], device='cuda:6'), out_proj_covar=tensor([6.7181e-05, 7.1483e-05, 7.6028e-05, 6.4033e-05, 7.2745e-05, 7.3308e-05, 6.7032e-05, 7.6021e-05], device='cuda:6') 2023-04-26 17:14:52,582 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.1454, 1.5772, 1.3088, 1.7104, 1.5936, 1.9875, 1.3400, 3.3570], device='cuda:6'), covar=tensor([0.0710, 0.0805, 0.0850, 0.1301, 0.0682, 0.0527, 0.0802, 0.0154], device='cuda:6'), in_proj_covar=tensor([0.0040, 0.0040, 0.0041, 0.0046, 0.0041, 0.0041, 0.0040, 0.0062], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0017], device='cuda:6') 2023-04-26 17:15:16,323 INFO [finetune.py:976] (6/7) Epoch 5, batch 5300, loss[loss=0.2673, simple_loss=0.3239, pruned_loss=0.1054, over 4891.00 frames. ], tot_loss[loss=0.214, simple_loss=0.2754, pruned_loss=0.0763, over 956038.62 frames. ], batch size: 35, lr: 3.92e-03, grad_scale: 32.0 2023-04-26 17:15:27,004 INFO [optim.py:369] (6/7) 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 17:15:28,352 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=28229.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 17:15:31,465 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7001, 2.1255, 1.1288, 1.5202, 2.1418, 1.6600, 1.6095, 1.7345], device='cuda:6'), covar=tensor([0.0537, 0.0375, 0.0344, 0.0589, 0.0255, 0.0546, 0.0534, 0.0593], device='cuda:6'), in_proj_covar=tensor([0.0031, 0.0026, 0.0024, 0.0030, 0.0021, 0.0030, 0.0030, 0.0031], device='cuda:6'), out_proj_covar=tensor([0.0049, 0.0044, 0.0038, 0.0049, 0.0038, 0.0048, 0.0047, 0.0050], device='cuda:6') 2023-04-26 17:15:46,195 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=28255.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 17:15:49,730 INFO [finetune.py:976] (6/7) Epoch 5, batch 5350, loss[loss=0.237, simple_loss=0.2944, pruned_loss=0.08982, over 4894.00 frames. ], tot_loss[loss=0.2134, simple_loss=0.2754, pruned_loss=0.07573, over 956807.70 frames. ], batch size: 37, lr: 3.92e-03, grad_scale: 32.0 2023-04-26 17:16:07,726 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=28287.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 17:16:10,014 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=28290.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 17:16:23,690 INFO [finetune.py:976] (6/7) Epoch 5, batch 5400, loss[loss=0.2084, simple_loss=0.2575, pruned_loss=0.07971, over 4749.00 frames. ], tot_loss[loss=0.2104, simple_loss=0.2719, pruned_loss=0.07439, over 956963.56 frames. ], batch size: 54, lr: 3.92e-03, grad_scale: 32.0 2023-04-26 17:16:27,309 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=28316.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 17:16:34,349 INFO [optim.py:369] (6/7) 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 17:16:49,208 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=28348.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 17:16:56,142 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2023-04-26 17:16:57,452 INFO [finetune.py:976] (6/7) Epoch 5, batch 5450, loss[loss=0.1504, simple_loss=0.2223, pruned_loss=0.03924, over 4921.00 frames. ], tot_loss[loss=0.2081, simple_loss=0.2691, pruned_loss=0.07358, over 953482.84 frames. ], batch size: 36, lr: 3.92e-03, grad_scale: 32.0 2023-04-26 17:17:03,566 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=28370.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 17:17:08,490 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.1339, 0.7514, 0.9280, 0.7145, 1.2972, 0.9847, 0.8269, 0.9722], device='cuda:6'), covar=tensor([0.1728, 0.1795, 0.2217, 0.1865, 0.1108, 0.1677, 0.2140, 0.2214], device='cuda:6'), in_proj_covar=tensor([0.0310, 0.0332, 0.0348, 0.0305, 0.0340, 0.0337, 0.0306, 0.0350], device='cuda:6'), out_proj_covar=tensor([6.6573e-05, 7.0966e-05, 7.5416e-05, 6.3549e-05, 7.2025e-05, 7.3054e-05, 6.6421e-05, 7.5319e-05], device='cuda:6') 2023-04-26 17:17:37,438 INFO [finetune.py:976] (6/7) Epoch 5, batch 5500, loss[loss=0.2661, simple_loss=0.3202, pruned_loss=0.106, over 4198.00 frames. ], tot_loss[loss=0.2038, simple_loss=0.2647, pruned_loss=0.07143, over 953523.98 frames. ], batch size: 65, lr: 3.92e-03, grad_scale: 32.0 2023-04-26 17:17:47,037 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=28418.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 17:17:59,139 INFO [optim.py:369] (6/7) 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 17:18:49,582 INFO [finetune.py:976] (6/7) Epoch 5, batch 5550, loss[loss=0.2206, simple_loss=0.2926, pruned_loss=0.07426, over 4810.00 frames. ], tot_loss[loss=0.2053, simple_loss=0.2666, pruned_loss=0.07204, over 954194.69 frames. ], batch size: 38, lr: 3.92e-03, grad_scale: 32.0 2023-04-26 17:19:10,291 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=28486.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 17:19:14,448 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=28492.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 17:19:25,621 INFO [finetune.py:976] (6/7) Epoch 5, batch 5600, loss[loss=0.2327, simple_loss=0.3002, pruned_loss=0.08263, over 4809.00 frames. ], tot_loss[loss=0.2081, simple_loss=0.2696, pruned_loss=0.07327, over 951262.06 frames. ], batch size: 41, lr: 3.92e-03, grad_scale: 32.0 2023-04-26 17:19:34,905 INFO [optim.py:369] (6/7) 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 17:19:47,089 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=28547.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 17:19:51,015 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 5, batch 5650, loss[loss=0.223, simple_loss=0.2933, pruned_loss=0.07636, over 4934.00 frames. ], tot_loss[loss=0.2095, simple_loss=0.2719, pruned_loss=0.07355, over 952869.17 frames. ], batch size: 33, lr: 3.92e-03, grad_scale: 32.0 2023-04-26 17:20:09,690 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=28585.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 17:20:25,135 INFO [finetune.py:976] (6/7) Epoch 5, batch 5700, loss[loss=0.2083, simple_loss=0.2511, pruned_loss=0.08272, over 4612.00 frames. ], tot_loss[loss=0.2073, simple_loss=0.2689, pruned_loss=0.07289, over 935958.41 frames. ], batch size: 20, lr: 3.92e-03, grad_scale: 32.0 2023-04-26 17:20:25,193 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=28611.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 17:20:34,597 INFO [optim.py:369] (6/7) 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 17:20:58,936 INFO [finetune.py:976] (6/7) Epoch 6, batch 0, loss[loss=0.1534, simple_loss=0.2278, pruned_loss=0.03953, over 4770.00 frames. ], tot_loss[loss=0.1534, simple_loss=0.2278, pruned_loss=0.03953, over 4770.00 frames. ], batch size: 26, lr: 3.92e-03, grad_scale: 32.0 2023-04-26 17:20:58,936 INFO [finetune.py:1001] (6/7) Computing validation loss 2023-04-26 17:21:06,854 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.2041, 1.6665, 1.4392, 1.9589, 1.6617, 1.9731, 1.4550, 3.1510], device='cuda:6'), covar=tensor([0.0682, 0.0745, 0.0788, 0.1153, 0.0657, 0.0448, 0.0747, 0.0233], device='cuda:6'), in_proj_covar=tensor([0.0039, 0.0040, 0.0041, 0.0045, 0.0041, 0.0040, 0.0040, 0.0061], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0017], device='cuda:6') 2023-04-26 17:21:13,391 INFO [finetune.py:1010] (6/7) 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] (6/7) Maximum memory allocated so far is 6326MB 2023-04-26 17:21:19,203 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=28643.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 17:21:19,251 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0349, 1.9712, 2.2444, 2.4348, 2.4467, 1.9299, 1.4656, 2.1440], device='cuda:6'), covar=tensor([0.1002, 0.1058, 0.0588, 0.0739, 0.0631, 0.1077, 0.1050, 0.0664], device='cuda:6'), in_proj_covar=tensor([0.0202, 0.0204, 0.0181, 0.0177, 0.0177, 0.0192, 0.0165, 0.0188], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-26 17:21:21,112 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.4341, 2.3443, 2.6933, 2.9226, 2.8355, 2.2332, 1.8368, 2.4608], device='cuda:6'), covar=tensor([0.0950, 0.0990, 0.0546, 0.0639, 0.0593, 0.0990, 0.0994, 0.0635], device='cuda:6'), in_proj_covar=tensor([0.0202, 0.0204, 0.0181, 0.0177, 0.0177, 0.0192, 0.0165, 0.0188], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-26 17:21:29,876 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0590, 1.5254, 1.3966, 1.9516, 2.1961, 1.8174, 1.7735, 1.4437], device='cuda:6'), covar=tensor([0.2143, 0.2013, 0.2199, 0.1776, 0.1553, 0.1950, 0.2018, 0.2080], device='cuda:6'), in_proj_covar=tensor([0.0310, 0.0331, 0.0349, 0.0305, 0.0341, 0.0336, 0.0305, 0.0351], device='cuda:6'), out_proj_covar=tensor([6.6652e-05, 7.0850e-05, 7.5417e-05, 6.3753e-05, 7.2213e-05, 7.2886e-05, 6.6233e-05, 7.5610e-05], device='cuda:6') 2023-04-26 17:21:59,786 INFO [finetune.py:976] (6/7) Epoch 6, batch 50, loss[loss=0.2086, simple_loss=0.2709, pruned_loss=0.0731, over 4904.00 frames. ], tot_loss[loss=0.2114, simple_loss=0.2732, pruned_loss=0.07483, over 217666.21 frames. ], batch size: 33, lr: 3.92e-03, grad_scale: 32.0 2023-04-26 17:22:14,751 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.8257, 2.4686, 2.7095, 3.1869, 2.6787, 2.2398, 2.7837, 1.9102], device='cuda:6'), covar=tensor([0.0428, 0.0683, 0.0445, 0.0530, 0.0628, 0.0982, 0.0503, 0.0815], device='cuda:6'), in_proj_covar=tensor([0.0067, 0.0075, 0.0074, 0.0068, 0.0079, 0.0096, 0.0082, 0.0077], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-26 17:22:24,841 INFO [optim.py:369] (6/7) 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 17:22:33,640 INFO [finetune.py:976] (6/7) Epoch 6, batch 100, loss[loss=0.1748, simple_loss=0.2366, pruned_loss=0.05651, over 4755.00 frames. ], tot_loss[loss=0.2082, simple_loss=0.2674, pruned_loss=0.07452, over 380450.09 frames. ], batch size: 27, lr: 3.92e-03, grad_scale: 32.0 2023-04-26 17:22:43,763 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7193, 2.0229, 1.2322, 1.4734, 2.0814, 1.6652, 1.5732, 1.6351], device='cuda:6'), covar=tensor([0.0537, 0.0357, 0.0333, 0.0550, 0.0251, 0.0530, 0.0519, 0.0594], device='cuda:6'), in_proj_covar=tensor([0.0031, 0.0026, 0.0024, 0.0030, 0.0021, 0.0030, 0.0029, 0.0031], device='cuda:6'), out_proj_covar=tensor([0.0049, 0.0044, 0.0038, 0.0049, 0.0037, 0.0048, 0.0047, 0.0049], device='cuda:6') 2023-04-26 17:22:46,821 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-04-26 17:23:06,831 INFO [finetune.py:976] (6/7) Epoch 6, batch 150, loss[loss=0.1981, simple_loss=0.241, pruned_loss=0.07758, over 4315.00 frames. ], tot_loss[loss=0.2018, simple_loss=0.2614, pruned_loss=0.07112, over 508358.22 frames. ], batch size: 65, lr: 3.92e-03, grad_scale: 32.0 2023-04-26 17:23:36,888 INFO [optim.py:369] (6/7) 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 17:23:57,270 INFO [finetune.py:976] (6/7) Epoch 6, batch 200, loss[loss=0.2115, simple_loss=0.2805, pruned_loss=0.07126, over 4824.00 frames. ], tot_loss[loss=0.2026, simple_loss=0.2614, pruned_loss=0.07187, over 606910.55 frames. ], batch size: 40, lr: 3.92e-03, grad_scale: 32.0 2023-04-26 17:23:59,203 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=28842.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 17:24:08,170 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=28848.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 17:24:22,300 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=28861.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 17:24:55,330 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=28885.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 17:25:03,861 INFO [finetune.py:976] (6/7) Epoch 6, batch 250, loss[loss=0.226, simple_loss=0.29, pruned_loss=0.08096, over 4817.00 frames. ], tot_loss[loss=0.205, simple_loss=0.2648, pruned_loss=0.07267, over 685456.91 frames. ], batch size: 33, lr: 3.92e-03, grad_scale: 16.0 2023-04-26 17:25:29,049 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=28911.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 17:25:40,844 INFO [zipformer.py:1188] (6/7) 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] (6/7) 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 17:25:55,384 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=28933.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 17:25:59,499 INFO [finetune.py:976] (6/7) Epoch 6, batch 300, loss[loss=0.2268, simple_loss=0.2854, pruned_loss=0.08407, over 4921.00 frames. ], tot_loss[loss=0.2067, simple_loss=0.2677, pruned_loss=0.07283, over 744491.36 frames. ], batch size: 33, lr: 3.92e-03, grad_scale: 16.0 2023-04-26 17:26:08,042 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=28943.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 17:26:28,687 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=28959.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 17:27:01,683 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.2131, 3.1691, 1.0126, 1.6406, 1.6768, 2.2857, 1.8524, 0.9561], device='cuda:6'), covar=tensor([0.1600, 0.0956, 0.1889, 0.1461, 0.1238, 0.1063, 0.1425, 0.2121], device='cuda:6'), in_proj_covar=tensor([0.0121, 0.0261, 0.0146, 0.0128, 0.0138, 0.0160, 0.0123, 0.0127], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-26 17:27:05,113 INFO [finetune.py:976] (6/7) Epoch 6, batch 350, loss[loss=0.17, simple_loss=0.2536, pruned_loss=0.04321, over 4801.00 frames. ], tot_loss[loss=0.2094, simple_loss=0.2711, pruned_loss=0.07384, over 791082.64 frames. ], batch size: 29, lr: 3.92e-03, grad_scale: 16.0 2023-04-26 17:27:11,880 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=28991.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 17:27:22,002 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2023-04-26 17:27:24,965 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=29002.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 17:27:53,127 INFO [optim.py:369] (6/7) 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 17:27:59,810 INFO [finetune.py:976] (6/7) Epoch 6, batch 400, loss[loss=0.1783, simple_loss=0.255, pruned_loss=0.05075, over 4764.00 frames. ], tot_loss[loss=0.2087, simple_loss=0.2714, pruned_loss=0.07297, over 828702.44 frames. ], batch size: 28, lr: 3.92e-03, grad_scale: 16.0 2023-04-26 17:28:16,908 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=29063.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 17:28:21,652 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.2959, 1.4564, 2.0731, 2.6759, 2.0683, 1.5505, 1.4552, 1.8942], device='cuda:6'), covar=tensor([0.4496, 0.5060, 0.2325, 0.3848, 0.4368, 0.3637, 0.5930, 0.3606], device='cuda:6'), in_proj_covar=tensor([0.0277, 0.0259, 0.0221, 0.0330, 0.0219, 0.0230, 0.0244, 0.0193], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-26 17:28:33,174 INFO [finetune.py:976] (6/7) Epoch 6, batch 450, loss[loss=0.2302, simple_loss=0.2941, pruned_loss=0.08314, over 4897.00 frames. ], tot_loss[loss=0.2077, simple_loss=0.2702, pruned_loss=0.07261, over 857176.29 frames. ], batch size: 32, lr: 3.92e-03, grad_scale: 16.0 2023-04-26 17:28:44,383 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7321, 2.4142, 1.7604, 1.5219, 1.2836, 1.3866, 1.8156, 1.2446], device='cuda:6'), covar=tensor([0.2113, 0.1738, 0.2027, 0.2507, 0.3201, 0.2672, 0.1508, 0.2714], device='cuda:6'), in_proj_covar=tensor([0.0195, 0.0220, 0.0177, 0.0208, 0.0213, 0.0187, 0.0170, 0.0194], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-26 17:28:59,677 INFO [optim.py:369] (6/7) 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 17:29:06,387 INFO [finetune.py:976] (6/7) Epoch 6, batch 500, loss[loss=0.1902, simple_loss=0.2545, pruned_loss=0.06298, over 4875.00 frames. ], tot_loss[loss=0.205, simple_loss=0.2671, pruned_loss=0.07146, over 879062.06 frames. ], batch size: 31, lr: 3.92e-03, grad_scale: 16.0 2023-04-26 17:29:08,778 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=29142.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 17:29:13,386 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=29148.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 17:29:39,681 INFO [finetune.py:976] (6/7) Epoch 6, batch 550, loss[loss=0.2021, simple_loss=0.2564, pruned_loss=0.07384, over 4818.00 frames. ], tot_loss[loss=0.2032, simple_loss=0.2645, pruned_loss=0.07091, over 897313.76 frames. ], batch size: 38, lr: 3.92e-03, grad_scale: 16.0 2023-04-26 17:29:40,346 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=29190.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 17:29:44,463 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=29196.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 17:29:59,122 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=29217.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 17:30:06,232 INFO [optim.py:369] (6/7) 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 17:30:11,163 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=29236.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 17:30:12,890 INFO [finetune.py:976] (6/7) Epoch 6, batch 600, loss[loss=0.233, simple_loss=0.2983, pruned_loss=0.08386, over 4904.00 frames. ], tot_loss[loss=0.2043, simple_loss=0.2653, pruned_loss=0.07161, over 909270.06 frames. ], batch size: 36, lr: 3.92e-03, grad_scale: 16.0 2023-04-26 17:30:47,032 INFO [finetune.py:976] (6/7) Epoch 6, batch 650, loss[loss=0.2723, simple_loss=0.3384, pruned_loss=0.103, over 4791.00 frames. ], tot_loss[loss=0.2079, simple_loss=0.2694, pruned_loss=0.07327, over 917535.95 frames. ], batch size: 59, lr: 3.92e-03, grad_scale: 16.0 2023-04-26 17:30:57,629 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=29297.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 17:31:42,432 INFO [optim.py:369] (6/7) 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 17:32:01,094 INFO [finetune.py:976] (6/7) Epoch 6, batch 700, loss[loss=0.1733, simple_loss=0.2462, pruned_loss=0.05018, over 4742.00 frames. ], tot_loss[loss=0.2086, simple_loss=0.2707, pruned_loss=0.07328, over 924865.30 frames. ], batch size: 27, lr: 3.92e-03, grad_scale: 16.0 2023-04-26 17:32:03,442 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 2023-04-26 17:32:25,508 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=29358.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 17:33:07,829 INFO [finetune.py:976] (6/7) Epoch 6, batch 750, loss[loss=0.2174, simple_loss=0.2826, pruned_loss=0.07611, over 4933.00 frames. ], tot_loss[loss=0.2096, simple_loss=0.2718, pruned_loss=0.07369, over 931313.86 frames. ], batch size: 38, lr: 3.92e-03, grad_scale: 16.0 2023-04-26 17:33:23,038 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8193, 1.3757, 1.3326, 1.5941, 2.0757, 1.6135, 1.3694, 1.2993], device='cuda:6'), covar=tensor([0.2121, 0.1878, 0.2520, 0.1615, 0.1119, 0.2123, 0.2779, 0.2337], device='cuda:6'), in_proj_covar=tensor([0.0311, 0.0332, 0.0351, 0.0308, 0.0342, 0.0337, 0.0307, 0.0352], device='cuda:6'), 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:6') 2023-04-26 17:33:42,632 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.8398, 2.8581, 2.2841, 3.2773, 2.8447, 2.8800, 1.1385, 2.8004], device='cuda:6'), covar=tensor([0.2200, 0.1573, 0.3510, 0.3005, 0.3228, 0.2193, 0.5701, 0.2963], device='cuda:6'), in_proj_covar=tensor([0.0246, 0.0218, 0.0254, 0.0312, 0.0303, 0.0257, 0.0275, 0.0276], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-26 17:34:03,540 INFO [optim.py:369] (6/7) 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 17:34:14,520 INFO [finetune.py:976] (6/7) Epoch 6, batch 800, loss[loss=0.1953, simple_loss=0.2663, pruned_loss=0.06215, over 4825.00 frames. ], tot_loss[loss=0.2081, simple_loss=0.2708, pruned_loss=0.07276, over 937284.22 frames. ], batch size: 47, lr: 3.92e-03, grad_scale: 16.0 2023-04-26 17:34:48,533 INFO [finetune.py:976] (6/7) Epoch 6, batch 850, loss[loss=0.2041, simple_loss=0.2596, pruned_loss=0.07434, over 4759.00 frames. ], tot_loss[loss=0.2059, simple_loss=0.2685, pruned_loss=0.07169, over 940686.35 frames. ], batch size: 27, lr: 3.92e-03, grad_scale: 16.0 2023-04-26 17:35:08,000 INFO [zipformer.py:1188] (6/7) 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] (6/7) 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 17:35:22,138 INFO [finetune.py:976] (6/7) Epoch 6, batch 900, loss[loss=0.2333, simple_loss=0.2862, pruned_loss=0.09024, over 4765.00 frames. ], tot_loss[loss=0.2046, simple_loss=0.2663, pruned_loss=0.07145, over 941489.17 frames. ], batch size: 26, lr: 3.91e-03, grad_scale: 16.0 2023-04-26 17:35:30,028 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=29551.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 17:35:38,987 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=29565.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 17:35:56,022 INFO [finetune.py:976] (6/7) Epoch 6, batch 950, loss[loss=0.2106, simple_loss=0.2813, pruned_loss=0.06995, over 4796.00 frames. ], tot_loss[loss=0.2027, simple_loss=0.2644, pruned_loss=0.07048, over 943437.02 frames. ], batch size: 29, lr: 3.91e-03, grad_scale: 16.0 2023-04-26 17:35:57,880 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=29592.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 17:36:10,609 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=29612.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 17:36:15,233 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-04-26 17:36:27,227 INFO [optim.py:369] (6/7) 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 17:36:39,504 INFO [finetune.py:976] (6/7) Epoch 6, batch 1000, loss[loss=0.2964, simple_loss=0.3622, pruned_loss=0.1153, over 4842.00 frames. ], tot_loss[loss=0.207, simple_loss=0.2687, pruned_loss=0.07264, over 945181.26 frames. ], batch size: 49, lr: 3.91e-03, grad_scale: 16.0 2023-04-26 17:37:09,272 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=29658.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 17:37:11,715 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=29662.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 17:37:52,138 INFO [finetune.py:976] (6/7) Epoch 6, batch 1050, loss[loss=0.2233, simple_loss=0.2965, pruned_loss=0.07509, over 4918.00 frames. ], tot_loss[loss=0.208, simple_loss=0.2704, pruned_loss=0.07279, over 948143.78 frames. ], batch size: 36, lr: 3.91e-03, grad_scale: 16.0 2023-04-26 17:38:06,521 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7514, 2.1371, 1.1186, 1.6570, 2.2767, 1.7694, 1.7101, 1.7976], device='cuda:6'), covar=tensor([0.0534, 0.0401, 0.0346, 0.0572, 0.0241, 0.0599, 0.0570, 0.0621], device='cuda:6'), in_proj_covar=tensor([0.0031, 0.0026, 0.0024, 0.0030, 0.0021, 0.0030, 0.0029, 0.0031], device='cuda:6'), out_proj_covar=tensor([0.0049, 0.0045, 0.0038, 0.0049, 0.0038, 0.0048, 0.0048, 0.0050], device='cuda:6') 2023-04-26 17:38:13,531 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=29706.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 17:38:34,982 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=29723.0, num_to_drop=1, layers_to_drop={3} 2023-04-26 17:38:39,283 INFO [optim.py:369] (6/7) 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] (6/7) Epoch 6, batch 1100, loss[loss=0.2095, simple_loss=0.2697, pruned_loss=0.07468, over 4867.00 frames. ], tot_loss[loss=0.2109, simple_loss=0.2735, pruned_loss=0.07421, over 952102.21 frames. ], batch size: 31, lr: 3.91e-03, grad_scale: 16.0 2023-04-26 17:39:11,115 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5279, 1.8398, 1.6003, 2.1356, 1.7444, 2.1997, 1.5836, 4.3400], device='cuda:6'), covar=tensor([0.0588, 0.0763, 0.0737, 0.1090, 0.0633, 0.0557, 0.0739, 0.0101], device='cuda:6'), in_proj_covar=tensor([0.0039, 0.0040, 0.0041, 0.0045, 0.0041, 0.0040, 0.0040, 0.0061], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0017], device='cuda:6') 2023-04-26 17:39:58,702 INFO [finetune.py:976] (6/7) Epoch 6, batch 1150, loss[loss=0.1804, simple_loss=0.2275, pruned_loss=0.06659, over 4716.00 frames. ], tot_loss[loss=0.2112, simple_loss=0.2735, pruned_loss=0.07448, over 953137.32 frames. ], batch size: 23, lr: 3.91e-03, grad_scale: 16.0 2023-04-26 17:40:04,444 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 2023-04-26 17:40:07,765 INFO [zipformer.py:1188] (6/7) 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] (6/7) 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] (6/7) Epoch 6, batch 1200, loss[loss=0.2229, simple_loss=0.2942, pruned_loss=0.07579, over 4815.00 frames. ], tot_loss[loss=0.2103, simple_loss=0.2727, pruned_loss=0.07398, over 953284.51 frames. ], batch size: 41, lr: 3.91e-03, grad_scale: 16.0 2023-04-26 17:40:48,170 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=29863.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 17:41:05,314 INFO [finetune.py:976] (6/7) Epoch 6, batch 1250, loss[loss=0.2615, simple_loss=0.3084, pruned_loss=0.1073, over 4820.00 frames. ], tot_loss[loss=0.2088, simple_loss=0.2701, pruned_loss=0.07378, over 951698.53 frames. ], batch size: 40, lr: 3.91e-03, grad_scale: 16.0 2023-04-26 17:41:07,217 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=29892.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 17:41:17,261 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=29907.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 17:41:30,521 INFO [optim.py:369] (6/7) 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] (6/7) Epoch 6, batch 1300, loss[loss=0.1837, simple_loss=0.2544, pruned_loss=0.05646, over 4931.00 frames. ], tot_loss[loss=0.2045, simple_loss=0.266, pruned_loss=0.07153, over 953887.40 frames. ], batch size: 33, lr: 3.91e-03, grad_scale: 16.0 2023-04-26 17:41:39,345 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=29940.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 17:41:56,315 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-04-26 17:42:12,123 INFO [finetune.py:976] (6/7) Epoch 6, batch 1350, loss[loss=0.1806, simple_loss=0.2516, pruned_loss=0.05486, over 4778.00 frames. ], tot_loss[loss=0.2048, simple_loss=0.2668, pruned_loss=0.0714, over 955706.29 frames. ], batch size: 26, lr: 3.91e-03, grad_scale: 16.0 2023-04-26 17:42:28,158 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-04-26 17:42:32,677 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=30018.0, num_to_drop=1, layers_to_drop={3} 2023-04-26 17:42:39,144 INFO [optim.py:369] (6/7) 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 17:42:52,096 INFO [finetune.py:976] (6/7) Epoch 6, batch 1400, loss[loss=0.2212, simple_loss=0.2898, pruned_loss=0.07635, over 4911.00 frames. ], tot_loss[loss=0.2082, simple_loss=0.2706, pruned_loss=0.07293, over 956325.77 frames. ], batch size: 35, lr: 3.91e-03, grad_scale: 16.0 2023-04-26 17:42:58,480 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8119, 2.6552, 1.7826, 1.9671, 1.3564, 1.3466, 1.9945, 1.2099], device='cuda:6'), covar=tensor([0.1859, 0.1621, 0.1775, 0.1974, 0.2645, 0.2069, 0.1257, 0.2318], device='cuda:6'), in_proj_covar=tensor([0.0193, 0.0218, 0.0175, 0.0205, 0.0211, 0.0184, 0.0167, 0.0191], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-26 17:43:01,580 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-04-26 17:43:54,431 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.2580, 1.6284, 2.0649, 2.5684, 2.0943, 1.5674, 1.6490, 1.8532], device='cuda:6'), covar=tensor([0.4118, 0.4403, 0.2041, 0.3900, 0.3914, 0.3310, 0.5300, 0.3617], device='cuda:6'), in_proj_covar=tensor([0.0274, 0.0257, 0.0219, 0.0326, 0.0217, 0.0228, 0.0241, 0.0192], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-26 17:43:56,754 INFO [finetune.py:976] (6/7) Epoch 6, batch 1450, loss[loss=0.201, simple_loss=0.2589, pruned_loss=0.07159, over 4770.00 frames. ], tot_loss[loss=0.2097, simple_loss=0.2723, pruned_loss=0.07355, over 956327.80 frames. ], batch size: 26, lr: 3.91e-03, grad_scale: 16.0 2023-04-26 17:44:42,749 INFO [zipformer.py:1188] (6/7) 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] (6/7) 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 17:44:57,143 INFO [finetune.py:976] (6/7) Epoch 6, batch 1500, loss[loss=0.206, simple_loss=0.2645, pruned_loss=0.07377, over 4784.00 frames. ], tot_loss[loss=0.2114, simple_loss=0.274, pruned_loss=0.07442, over 956244.40 frames. ], batch size: 25, lr: 3.91e-03, grad_scale: 16.0 2023-04-26 17:45:25,819 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=30158.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 17:45:59,233 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 6, batch 1550, loss[loss=0.1672, simple_loss=0.2303, pruned_loss=0.05206, over 4778.00 frames. ], tot_loss[loss=0.212, simple_loss=0.2743, pruned_loss=0.07482, over 955324.78 frames. ], batch size: 27, lr: 3.91e-03, grad_scale: 16.0 2023-04-26 17:46:30,718 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=30207.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 17:46:31,935 INFO [zipformer.py:1188] (6/7) 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] (6/7) 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] (6/7) Epoch 6, batch 1600, loss[loss=0.1628, simple_loss=0.2388, pruned_loss=0.04343, over 4755.00 frames. ], tot_loss[loss=0.2093, simple_loss=0.2712, pruned_loss=0.07366, over 954531.45 frames. ], batch size: 28, lr: 3.91e-03, grad_scale: 16.0 2023-04-26 17:47:31,235 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=30255.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 17:47:38,021 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.00 vs. limit=5.0 2023-04-26 17:47:41,432 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=30270.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 17:47:53,341 INFO [finetune.py:976] (6/7) Epoch 6, batch 1650, loss[loss=0.1846, simple_loss=0.257, pruned_loss=0.05612, over 4824.00 frames. ], tot_loss[loss=0.205, simple_loss=0.2664, pruned_loss=0.07175, over 954240.52 frames. ], batch size: 39, lr: 3.91e-03, grad_scale: 16.0 2023-04-26 17:48:11,494 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-04-26 17:48:13,379 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=30318.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 17:48:19,318 INFO [optim.py:369] (6/7) 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 17:48:26,039 INFO [finetune.py:976] (6/7) Epoch 6, batch 1700, loss[loss=0.2351, simple_loss=0.2822, pruned_loss=0.09398, over 4891.00 frames. ], tot_loss[loss=0.2022, simple_loss=0.2638, pruned_loss=0.07031, over 955489.41 frames. ], batch size: 32, lr: 3.91e-03, grad_scale: 16.0 2023-04-26 17:48:51,275 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=30360.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 17:49:00,654 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=30366.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 17:49:26,052 INFO [finetune.py:976] (6/7) Epoch 6, batch 1750, loss[loss=0.2558, simple_loss=0.3096, pruned_loss=0.1011, over 4075.00 frames. ], tot_loss[loss=0.2053, simple_loss=0.2669, pruned_loss=0.07178, over 954006.38 frames. ], batch size: 65, lr: 3.91e-03, grad_scale: 16.0 2023-04-26 17:50:03,368 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=30421.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 17:50:08,042 INFO [optim.py:369] (6/7) 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 17:50:20,189 INFO [finetune.py:976] (6/7) Epoch 6, batch 1800, loss[loss=0.2239, simple_loss=0.2824, pruned_loss=0.0827, over 4884.00 frames. ], tot_loss[loss=0.2088, simple_loss=0.2709, pruned_loss=0.07335, over 953798.35 frames. ], batch size: 32, lr: 3.91e-03, grad_scale: 16.0 2023-04-26 17:50:43,034 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=30458.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 17:51:09,063 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=30481.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 17:51:13,896 INFO [finetune.py:976] (6/7) Epoch 6, batch 1850, loss[loss=0.2231, simple_loss=0.2835, pruned_loss=0.08132, over 4119.00 frames. ], tot_loss[loss=0.2105, simple_loss=0.2722, pruned_loss=0.07436, over 952735.78 frames. ], batch size: 65, lr: 3.91e-03, grad_scale: 16.0 2023-04-26 17:51:15,215 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.2343, 1.8005, 1.4641, 2.0586, 1.7585, 2.1092, 1.5529, 4.3987], device='cuda:6'), covar=tensor([0.0643, 0.0763, 0.0809, 0.1194, 0.0649, 0.0568, 0.0756, 0.0102], device='cuda:6'), in_proj_covar=tensor([0.0039, 0.0039, 0.0041, 0.0045, 0.0041, 0.0040, 0.0040, 0.0061], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0017], device='cuda:6') 2023-04-26 17:51:24,866 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=30506.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 17:51:39,579 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.0104, 2.0125, 1.7450, 1.6751, 2.1887, 1.7579, 2.5855, 1.5517], device='cuda:6'), covar=tensor([0.4505, 0.1905, 0.5561, 0.3454, 0.1735, 0.2673, 0.1535, 0.5083], device='cuda:6'), in_proj_covar=tensor([0.0353, 0.0359, 0.0443, 0.0370, 0.0398, 0.0391, 0.0394, 0.0425], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-26 17:51:40,049 INFO [optim.py:369] (6/7) 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 17:51:47,206 INFO [finetune.py:976] (6/7) Epoch 6, batch 1900, loss[loss=0.1896, simple_loss=0.2497, pruned_loss=0.06469, over 4903.00 frames. ], tot_loss[loss=0.2109, simple_loss=0.2731, pruned_loss=0.07434, over 952668.08 frames. ], batch size: 36, lr: 3.91e-03, grad_scale: 16.0 2023-04-26 17:52:06,454 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.09 vs. limit=5.0 2023-04-26 17:52:15,910 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8746, 2.3819, 1.9842, 2.3034, 1.7623, 1.9876, 2.0776, 1.7012], device='cuda:6'), covar=tensor([0.2311, 0.1567, 0.0953, 0.1509, 0.3394, 0.1477, 0.2073, 0.2639], device='cuda:6'), in_proj_covar=tensor([0.0303, 0.0324, 0.0234, 0.0297, 0.0316, 0.0278, 0.0264, 0.0290], device='cuda:6'), out_proj_covar=tensor([1.2311e-04, 1.3154e-04, 9.5217e-05, 1.1906e-04, 1.3025e-04, 1.1251e-04, 1.0844e-04, 1.1668e-04], device='cuda:6') 2023-04-26 17:52:17,582 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=30565.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 17:52:25,200 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.8935, 2.8363, 2.2131, 3.2861, 2.9191, 2.8758, 1.0862, 2.8390], device='cuda:6'), covar=tensor([0.1698, 0.1561, 0.3501, 0.2741, 0.2846, 0.1980, 0.5822, 0.2528], device='cuda:6'), in_proj_covar=tensor([0.0246, 0.0220, 0.0256, 0.0313, 0.0304, 0.0257, 0.0276, 0.0277], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-26 17:52:34,611 INFO [finetune.py:976] (6/7) Epoch 6, batch 1950, loss[loss=0.183, simple_loss=0.2477, pruned_loss=0.05922, over 4771.00 frames. ], tot_loss[loss=0.2103, simple_loss=0.2724, pruned_loss=0.07408, over 952839.55 frames. ], batch size: 27, lr: 3.91e-03, grad_scale: 16.0 2023-04-26 17:53:00,333 INFO [optim.py:369] (6/7) 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 17:53:02,307 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7563, 2.3375, 1.8188, 1.5150, 1.2808, 1.3629, 1.7871, 1.2711], device='cuda:6'), covar=tensor([0.1963, 0.1861, 0.1941, 0.2550, 0.3077, 0.2480, 0.1395, 0.2499], device='cuda:6'), in_proj_covar=tensor([0.0194, 0.0220, 0.0176, 0.0206, 0.0212, 0.0186, 0.0168, 0.0192], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-26 17:53:08,042 INFO [finetune.py:976] (6/7) Epoch 6, batch 2000, loss[loss=0.1703, simple_loss=0.2226, pruned_loss=0.05898, over 3952.00 frames. ], tot_loss[loss=0.2076, simple_loss=0.2691, pruned_loss=0.07307, over 952679.53 frames. ], batch size: 17, lr: 3.91e-03, grad_scale: 16.0 2023-04-26 17:53:17,890 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=30655.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 17:53:24,495 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0519, 2.8942, 2.0778, 2.0155, 1.5127, 1.5155, 2.2337, 1.4533], device='cuda:6'), covar=tensor([0.2121, 0.1773, 0.1953, 0.2433, 0.3087, 0.2378, 0.1372, 0.2635], device='cuda:6'), in_proj_covar=tensor([0.0194, 0.0220, 0.0175, 0.0206, 0.0211, 0.0186, 0.0167, 0.0192], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-26 17:53:25,880 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-04-26 17:53:41,351 INFO [finetune.py:976] (6/7) Epoch 6, batch 2050, loss[loss=0.1419, simple_loss=0.2124, pruned_loss=0.03568, over 4778.00 frames. ], tot_loss[loss=0.2055, simple_loss=0.2663, pruned_loss=0.07232, over 951689.28 frames. ], batch size: 27, lr: 3.91e-03, grad_scale: 16.0 2023-04-26 17:53:58,930 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=30716.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 17:53:58,997 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=30716.0, num_to_drop=1, layers_to_drop={2} 2023-04-26 17:54:12,372 INFO [optim.py:369] (6/7) 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] (6/7) Epoch 6, batch 2100, loss[loss=0.1983, simple_loss=0.2668, pruned_loss=0.06491, over 4761.00 frames. ], tot_loss[loss=0.207, simple_loss=0.2675, pruned_loss=0.07322, over 950302.03 frames. ], batch size: 54, lr: 3.91e-03, grad_scale: 16.0 2023-04-26 17:54:39,197 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7540, 1.0172, 1.3342, 1.4732, 1.4475, 1.6089, 1.3827, 1.3930], device='cuda:6'), covar=tensor([0.6221, 0.8481, 0.7155, 0.7146, 0.8298, 1.2289, 0.8449, 0.7598], device='cuda:6'), in_proj_covar=tensor([0.0316, 0.0395, 0.0316, 0.0327, 0.0345, 0.0410, 0.0375, 0.0332], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-26 17:54:44,728 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-04-26 17:54:54,252 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=30781.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 17:54:59,005 INFO [finetune.py:976] (6/7) Epoch 6, batch 2150, loss[loss=0.1599, simple_loss=0.2174, pruned_loss=0.05118, over 4386.00 frames. ], tot_loss[loss=0.2073, simple_loss=0.2685, pruned_loss=0.07308, over 951413.40 frames. ], batch size: 19, lr: 3.91e-03, grad_scale: 16.0 2023-04-26 17:55:40,195 INFO [optim.py:369] (6/7) 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 17:55:40,286 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.1762, 3.1346, 0.7763, 1.6115, 1.5413, 2.2221, 1.7810, 0.9965], device='cuda:6'), covar=tensor([0.1744, 0.1279, 0.2489, 0.1582, 0.1372, 0.1293, 0.1747, 0.2284], device='cuda:6'), in_proj_covar=tensor([0.0122, 0.0261, 0.0147, 0.0128, 0.0138, 0.0160, 0.0123, 0.0127], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-26 17:55:40,856 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=30829.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 17:55:54,382 INFO [finetune.py:976] (6/7) Epoch 6, batch 2200, loss[loss=0.1535, simple_loss=0.2274, pruned_loss=0.03979, over 4840.00 frames. ], tot_loss[loss=0.2112, simple_loss=0.2725, pruned_loss=0.0749, over 950459.87 frames. ], batch size: 30, lr: 3.91e-03, grad_scale: 16.0 2023-04-26 17:56:29,019 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=30865.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 17:56:51,897 INFO [finetune.py:976] (6/7) Epoch 6, batch 2250, loss[loss=0.2071, simple_loss=0.2764, pruned_loss=0.06885, over 4728.00 frames. ], tot_loss[loss=0.2111, simple_loss=0.2733, pruned_loss=0.07451, over 953045.77 frames. ], batch size: 59, lr: 3.91e-03, grad_scale: 32.0 2023-04-26 17:56:58,590 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([3.3302, 3.2503, 2.3651, 3.8328, 3.3067, 3.3750, 1.2554, 3.1912], device='cuda:6'), covar=tensor([0.1713, 0.1277, 0.3299, 0.2539, 0.2248, 0.1854, 0.5559, 0.2773], device='cuda:6'), in_proj_covar=tensor([0.0244, 0.0219, 0.0254, 0.0313, 0.0303, 0.0255, 0.0275, 0.0275], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-26 17:57:11,925 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=30903.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 17:57:22,864 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=30913.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 17:57:40,821 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=30925.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 17:57:42,532 INFO [optim.py:369] (6/7) 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 17:57:43,770 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8025, 2.4465, 2.0084, 2.2663, 1.6288, 2.0565, 2.1505, 1.6790], device='cuda:6'), covar=tensor([0.2554, 0.1276, 0.0965, 0.1512, 0.3488, 0.1341, 0.2194, 0.2998], device='cuda:6'), in_proj_covar=tensor([0.0302, 0.0322, 0.0233, 0.0295, 0.0315, 0.0277, 0.0263, 0.0288], device='cuda:6'), out_proj_covar=tensor([1.2260e-04, 1.3069e-04, 9.4875e-05, 1.1823e-04, 1.2951e-04, 1.1190e-04, 1.0799e-04, 1.1584e-04], device='cuda:6') 2023-04-26 17:57:46,800 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 6, batch 2300, loss[loss=0.1975, simple_loss=0.2589, pruned_loss=0.06808, over 4900.00 frames. ], tot_loss[loss=0.2101, simple_loss=0.2726, pruned_loss=0.07385, over 953165.22 frames. ], batch size: 35, lr: 3.91e-03, grad_scale: 32.0 2023-04-26 17:58:07,554 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=30964.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 17:58:21,371 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 6, batch 2350, loss[loss=0.2585, simple_loss=0.3011, pruned_loss=0.1079, over 4906.00 frames. ], tot_loss[loss=0.208, simple_loss=0.2702, pruned_loss=0.07284, over 951549.41 frames. ], batch size: 36, lr: 3.91e-03, grad_scale: 32.0 2023-04-26 17:58:28,713 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=30995.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 17:58:39,470 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=31011.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 17:58:39,498 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=31011.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 17:58:42,543 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=31016.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 17:58:49,763 INFO [optim.py:369] (6/7) 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 17:58:56,918 INFO [finetune.py:976] (6/7) Epoch 6, batch 2400, loss[loss=0.1992, simple_loss=0.2604, pruned_loss=0.06893, over 4892.00 frames. ], tot_loss[loss=0.2049, simple_loss=0.2665, pruned_loss=0.07161, over 952832.89 frames. ], batch size: 35, lr: 3.91e-03, grad_scale: 32.0 2023-04-26 17:59:04,058 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-04-26 17:59:08,429 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.1514, 1.1932, 3.0119, 2.6372, 2.7119, 2.7351, 2.7924, 2.6044], device='cuda:6'), covar=tensor([0.8094, 0.6265, 0.2198, 0.3482, 0.2527, 0.4157, 0.6075, 0.3477], device='cuda:6'), in_proj_covar=tensor([0.0312, 0.0304, 0.0413, 0.0416, 0.0353, 0.0408, 0.0316, 0.0372], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-26 17:59:14,977 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=31064.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 17:59:19,955 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=31072.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 17:59:25,500 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.3538, 2.8071, 1.0157, 1.4442, 2.2192, 1.4115, 3.8676, 1.8257], device='cuda:6'), covar=tensor([0.0650, 0.0898, 0.0877, 0.1259, 0.0472, 0.1016, 0.0251, 0.0628], device='cuda:6'), in_proj_covar=tensor([0.0054, 0.0070, 0.0052, 0.0049, 0.0054, 0.0054, 0.0082, 0.0052], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0011, 0.0008], device='cuda:6') 2023-04-26 17:59:27,771 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.67 vs. limit=5.0 2023-04-26 17:59:30,875 INFO [finetune.py:976] (6/7) Epoch 6, batch 2450, loss[loss=0.2398, simple_loss=0.2941, pruned_loss=0.09272, over 4926.00 frames. ], tot_loss[loss=0.2029, simple_loss=0.2639, pruned_loss=0.07094, over 951171.28 frames. ], batch size: 38, lr: 3.91e-03, grad_scale: 32.0 2023-04-26 17:59:31,651 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=3.71 vs. limit=5.0 2023-04-26 17:59:42,061 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.4963, 0.9680, 1.4155, 1.9193, 1.6781, 1.4074, 1.4063, 1.4897], device='cuda:6'), covar=tensor([0.7200, 1.0275, 1.0262, 1.1628, 0.8316, 1.2008, 1.2274, 1.0200], device='cuda:6'), in_proj_covar=tensor([0.0412, 0.0441, 0.0524, 0.0545, 0.0441, 0.0463, 0.0475, 0.0473], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-26 17:59:57,659 INFO [optim.py:369] (6/7) 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 18:00:04,431 INFO [finetune.py:976] (6/7) Epoch 6, batch 2500, loss[loss=0.1584, simple_loss=0.2303, pruned_loss=0.0432, over 4754.00 frames. ], tot_loss[loss=0.2054, simple_loss=0.266, pruned_loss=0.07242, over 950227.23 frames. ], batch size: 27, lr: 3.91e-03, grad_scale: 32.0 2023-04-26 18:00:13,180 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7178, 1.5854, 1.9419, 2.0064, 1.8914, 1.6466, 1.7513, 1.8433], device='cuda:6'), covar=tensor([0.9192, 1.2899, 1.4337, 1.4740, 1.0654, 1.6864, 1.7301, 1.4313], device='cuda:6'), in_proj_covar=tensor([0.0412, 0.0441, 0.0524, 0.0545, 0.0441, 0.0463, 0.0476, 0.0473], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-26 18:00:25,672 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.1153, 0.6764, 0.8785, 0.7066, 1.1953, 0.9103, 0.6705, 0.9380], device='cuda:6'), covar=tensor([0.1769, 0.1600, 0.2005, 0.1632, 0.0935, 0.1525, 0.1940, 0.2064], device='cuda:6'), in_proj_covar=tensor([0.0309, 0.0331, 0.0350, 0.0306, 0.0341, 0.0335, 0.0308, 0.0351], device='cuda:6'), out_proj_covar=tensor([6.6315e-05, 7.0645e-05, 7.5835e-05, 6.3756e-05, 7.1977e-05, 7.2500e-05, 6.6723e-05, 7.5725e-05], device='cuda:6') 2023-04-26 18:00:31,765 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=3.75 vs. limit=5.0 2023-04-26 18:00:37,629 INFO [finetune.py:976] (6/7) Epoch 6, batch 2550, loss[loss=0.2072, simple_loss=0.2679, pruned_loss=0.07321, over 4900.00 frames. ], tot_loss[loss=0.2092, simple_loss=0.2705, pruned_loss=0.07397, over 952246.59 frames. ], batch size: 32, lr: 3.91e-03, grad_scale: 32.0 2023-04-26 18:00:43,542 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([3.7939, 3.6806, 2.8171, 4.4253, 3.8001, 3.8553, 1.6337, 3.7834], device='cuda:6'), covar=tensor([0.1533, 0.1333, 0.3333, 0.1581, 0.3317, 0.1861, 0.5665, 0.2229], device='cuda:6'), in_proj_covar=tensor([0.0243, 0.0219, 0.0253, 0.0312, 0.0300, 0.0254, 0.0273, 0.0273], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-26 18:00:50,234 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9668, 2.7049, 1.9656, 1.8627, 1.4763, 1.4816, 2.1106, 1.3789], device='cuda:6'), covar=tensor([0.1921, 0.1855, 0.1776, 0.2169, 0.2699, 0.2197, 0.1292, 0.2362], device='cuda:6'), in_proj_covar=tensor([0.0195, 0.0220, 0.0177, 0.0207, 0.0212, 0.0186, 0.0168, 0.0193], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-26 18:01:09,724 INFO [optim.py:369] (6/7) 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 18:01:22,019 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.0581, 1.3322, 1.2185, 1.6704, 1.3706, 1.6639, 1.2774, 3.0033], device='cuda:6'), covar=tensor([0.0669, 0.0786, 0.0798, 0.1151, 0.0684, 0.0553, 0.0763, 0.0181], device='cuda:6'), in_proj_covar=tensor([0.0039, 0.0039, 0.0041, 0.0045, 0.0041, 0.0040, 0.0040, 0.0061], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0017], device='cuda:6') 2023-04-26 18:01:22,515 INFO [finetune.py:976] (6/7) Epoch 6, batch 2600, loss[loss=0.2374, simple_loss=0.2988, pruned_loss=0.08801, over 4849.00 frames. ], tot_loss[loss=0.2086, simple_loss=0.2707, pruned_loss=0.07325, over 951905.12 frames. ], batch size: 44, lr: 3.91e-03, grad_scale: 32.0 2023-04-26 18:01:23,253 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.2400, 2.1266, 2.0635, 1.7933, 2.3220, 2.0258, 2.8420, 1.8160], device='cuda:6'), covar=tensor([0.3548, 0.1710, 0.4313, 0.2800, 0.1508, 0.2243, 0.1383, 0.3952], device='cuda:6'), in_proj_covar=tensor([0.0348, 0.0354, 0.0439, 0.0368, 0.0392, 0.0384, 0.0389, 0.0420], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-26 18:01:51,864 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=31259.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:02:10,474 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0071, 2.7507, 2.2602, 2.5643, 1.9301, 2.3909, 2.5527, 1.9233], device='cuda:6'), covar=tensor([0.2316, 0.1152, 0.0866, 0.1234, 0.3257, 0.1081, 0.1995, 0.2800], device='cuda:6'), in_proj_covar=tensor([0.0301, 0.0323, 0.0233, 0.0294, 0.0314, 0.0275, 0.0262, 0.0286], device='cuda:6'), out_proj_covar=tensor([1.2237e-04, 1.3112e-04, 9.4516e-05, 1.1816e-04, 1.2936e-04, 1.1140e-04, 1.0770e-04, 1.1530e-04], device='cuda:6') 2023-04-26 18:02:14,140 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=31281.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:02:24,320 INFO [finetune.py:976] (6/7) Epoch 6, batch 2650, loss[loss=0.2041, simple_loss=0.2673, pruned_loss=0.07046, over 4828.00 frames. ], tot_loss[loss=0.2081, simple_loss=0.2705, pruned_loss=0.07286, over 952332.06 frames. ], batch size: 30, lr: 3.91e-03, grad_scale: 32.0 2023-04-26 18:02:24,976 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=31290.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:02:55,889 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=31311.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:03:18,340 INFO [optim.py:369] (6/7) 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] (6/7) Epoch 6, batch 2700, loss[loss=0.1562, simple_loss=0.2198, pruned_loss=0.0463, over 4723.00 frames. ], tot_loss[loss=0.2073, simple_loss=0.2699, pruned_loss=0.07238, over 952720.32 frames. ], batch size: 23, lr: 3.90e-03, grad_scale: 32.0 2023-04-26 18:04:01,833 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=31359.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:04:14,066 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=31367.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 18:04:23,834 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0680, 2.9400, 2.3305, 2.6985, 2.1461, 2.3831, 2.6327, 1.9349], device='cuda:6'), covar=tensor([0.2494, 0.1575, 0.1028, 0.1471, 0.2976, 0.1432, 0.1886, 0.3041], device='cuda:6'), in_proj_covar=tensor([0.0303, 0.0324, 0.0234, 0.0295, 0.0316, 0.0277, 0.0263, 0.0288], device='cuda:6'), out_proj_covar=tensor([1.2307e-04, 1.3179e-04, 9.5085e-05, 1.1827e-04, 1.2997e-04, 1.1188e-04, 1.0802e-04, 1.1573e-04], device='cuda:6') 2023-04-26 18:04:44,997 INFO [finetune.py:976] (6/7) Epoch 6, batch 2750, loss[loss=0.1941, simple_loss=0.2572, pruned_loss=0.06547, over 4909.00 frames. ], tot_loss[loss=0.2052, simple_loss=0.2674, pruned_loss=0.07151, over 953569.90 frames. ], batch size: 43, lr: 3.90e-03, grad_scale: 32.0 2023-04-26 18:05:10,772 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=31409.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:05:28,795 INFO [optim.py:369] (6/7) 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 18:05:34,290 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8669, 0.9770, 1.3212, 1.4947, 1.5104, 1.6920, 1.3778, 1.3909], device='cuda:6'), covar=tensor([0.5429, 0.7397, 0.6885, 0.6047, 0.7266, 1.0961, 0.7782, 0.7137], device='cuda:6'), in_proj_covar=tensor([0.0316, 0.0392, 0.0316, 0.0325, 0.0344, 0.0409, 0.0374, 0.0331], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-26 18:05:35,957 INFO [finetune.py:976] (6/7) Epoch 6, batch 2800, loss[loss=0.1977, simple_loss=0.2613, pruned_loss=0.06707, over 4705.00 frames. ], tot_loss[loss=0.2019, simple_loss=0.2637, pruned_loss=0.07003, over 953644.87 frames. ], batch size: 23, lr: 3.90e-03, grad_scale: 32.0 2023-04-26 18:05:37,362 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.80 vs. limit=2.0 2023-04-26 18:05:47,047 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7004, 1.3561, 1.7644, 2.0751, 1.8183, 1.6306, 1.6753, 1.7392], device='cuda:6'), covar=tensor([0.9217, 1.2249, 1.2611, 1.3005, 1.0155, 1.5235, 1.4804, 1.3420], device='cuda:6'), in_proj_covar=tensor([0.0415, 0.0444, 0.0527, 0.0548, 0.0443, 0.0465, 0.0479, 0.0477], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-26 18:05:57,416 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 6, batch 2850, loss[loss=0.2302, simple_loss=0.298, pruned_loss=0.08116, over 4812.00 frames. ], tot_loss[loss=0.2026, simple_loss=0.2639, pruned_loss=0.07061, over 952899.67 frames. ], batch size: 39, lr: 3.90e-03, grad_scale: 32.0 2023-04-26 18:06:10,536 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-04-26 18:06:11,333 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=31491.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 18:06:18,083 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.0227, 1.4062, 1.2892, 1.6461, 1.4549, 1.7787, 1.3551, 3.0372], device='cuda:6'), covar=tensor([0.0730, 0.0841, 0.0853, 0.1250, 0.0701, 0.0546, 0.0791, 0.0184], device='cuda:6'), in_proj_covar=tensor([0.0039, 0.0040, 0.0041, 0.0045, 0.0041, 0.0040, 0.0040, 0.0061], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0017], device='cuda:6') 2023-04-26 18:06:19,311 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.8276, 3.0258, 1.0768, 1.7556, 2.4188, 1.7561, 4.5219, 2.3089], device='cuda:6'), covar=tensor([0.0569, 0.0769, 0.0917, 0.1405, 0.0556, 0.1048, 0.0296, 0.0646], device='cuda:6'), in_proj_covar=tensor([0.0054, 0.0070, 0.0052, 0.0049, 0.0054, 0.0054, 0.0082, 0.0053], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0011, 0.0008], device='cuda:6') 2023-04-26 18:06:32,724 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 2023-04-26 18:06:35,987 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.0031, 0.9206, 1.1482, 1.0859, 0.9600, 0.7733, 0.9162, 0.5843], device='cuda:6'), covar=tensor([0.0505, 0.0754, 0.0725, 0.0690, 0.0761, 0.1432, 0.0600, 0.0953], device='cuda:6'), in_proj_covar=tensor([0.0067, 0.0075, 0.0074, 0.0068, 0.0078, 0.0095, 0.0081, 0.0077], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-26 18:06:36,455 INFO [optim.py:369] (6/7) 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 18:06:37,902 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.23 vs. limit=5.0 2023-04-26 18:06:43,785 INFO [finetune.py:976] (6/7) Epoch 6, batch 2900, loss[loss=0.2129, simple_loss=0.2786, pruned_loss=0.07362, over 4801.00 frames. ], tot_loss[loss=0.2032, simple_loss=0.265, pruned_loss=0.07067, over 952656.49 frames. ], batch size: 51, lr: 3.90e-03, grad_scale: 32.0 2023-04-26 18:06:47,518 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.1203, 1.9293, 1.6903, 1.6143, 1.9839, 1.7184, 2.3389, 1.4979], device='cuda:6'), covar=tensor([0.3306, 0.1679, 0.4322, 0.2906, 0.1519, 0.2195, 0.1433, 0.4475], device='cuda:6'), in_proj_covar=tensor([0.0350, 0.0357, 0.0439, 0.0368, 0.0393, 0.0385, 0.0389, 0.0422], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-26 18:06:51,763 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=31552.0, num_to_drop=1, layers_to_drop={2} 2023-04-26 18:06:56,018 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=31559.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:07:11,690 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=31581.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:07:16,959 INFO [finetune.py:976] (6/7) Epoch 6, batch 2950, loss[loss=0.2082, simple_loss=0.2762, pruned_loss=0.07012, over 4772.00 frames. ], tot_loss[loss=0.2073, simple_loss=0.2698, pruned_loss=0.07236, over 950478.51 frames. ], batch size: 54, lr: 3.90e-03, grad_scale: 32.0 2023-04-26 18:07:17,676 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=31590.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:07:28,072 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=31607.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:07:42,552 INFO [optim.py:369] (6/7) 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] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=31629.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:07:49,142 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 6, batch 3000, loss[loss=0.2359, simple_loss=0.3012, pruned_loss=0.08525, over 4837.00 frames. ], tot_loss[loss=0.2076, simple_loss=0.2704, pruned_loss=0.07237, over 950161.43 frames. ], batch size: 49, lr: 3.90e-03, grad_scale: 32.0 2023-04-26 18:07:49,695 INFO [finetune.py:1001] (6/7) Computing validation loss 2023-04-26 18:07:54,250 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.3855, 1.1939, 1.5610, 1.5081, 1.2676, 1.0811, 1.2736, 0.9442], device='cuda:6'), covar=tensor([0.0625, 0.0927, 0.0626, 0.0676, 0.0925, 0.1253, 0.0757, 0.0815], device='cuda:6'), in_proj_covar=tensor([0.0067, 0.0075, 0.0074, 0.0067, 0.0078, 0.0095, 0.0081, 0.0077], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-26 18:08:00,214 INFO [finetune.py:1010] (6/7) 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] (6/7) Maximum memory allocated so far is 6326MB 2023-04-26 18:08:17,859 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 6, batch 3050, loss[loss=0.1937, simple_loss=0.2584, pruned_loss=0.0645, over 4748.00 frames. ], tot_loss[loss=0.2089, simple_loss=0.2718, pruned_loss=0.073, over 952657.78 frames. ], batch size: 26, lr: 3.90e-03, grad_scale: 32.0 2023-04-26 18:09:09,618 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5916, 1.9242, 1.6198, 1.0929, 1.2413, 1.2490, 1.6450, 1.2021], device='cuda:6'), covar=tensor([0.2069, 0.1748, 0.1899, 0.2549, 0.2908, 0.2348, 0.1281, 0.2515], device='cuda:6'), in_proj_covar=tensor([0.0195, 0.0220, 0.0176, 0.0206, 0.0211, 0.0187, 0.0168, 0.0193], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-26 18:09:10,801 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=31704.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:09:20,318 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.1169, 1.4949, 1.9633, 2.2317, 1.8710, 1.4569, 1.1222, 1.6214], device='cuda:6'), covar=tensor([0.3404, 0.4014, 0.1801, 0.2948, 0.3536, 0.3049, 0.5240, 0.2830], device='cuda:6'), in_proj_covar=tensor([0.0277, 0.0256, 0.0220, 0.0327, 0.0216, 0.0230, 0.0242, 0.0192], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-26 18:09:23,347 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=31715.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 18:09:40,854 INFO [optim.py:369] (6/7) 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] (6/7) Epoch 6, batch 3100, loss[loss=0.1755, simple_loss=0.2463, pruned_loss=0.05235, over 4899.00 frames. ], tot_loss[loss=0.2058, simple_loss=0.2689, pruned_loss=0.07134, over 953501.35 frames. ], batch size: 36, lr: 3.90e-03, grad_scale: 32.0 2023-04-26 18:10:25,845 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=31765.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:10:25,889 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=31765.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:10:57,472 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=31788.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:10:57,968 INFO [finetune.py:976] (6/7) Epoch 6, batch 3150, loss[loss=0.2383, simple_loss=0.2863, pruned_loss=0.0952, over 4840.00 frames. ], tot_loss[loss=0.2041, simple_loss=0.2662, pruned_loss=0.07101, over 953399.13 frames. ], batch size: 30, lr: 3.90e-03, grad_scale: 32.0 2023-04-26 18:11:21,865 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.2429, 1.3464, 5.3233, 4.7673, 4.7592, 4.9824, 4.5803, 4.4422], device='cuda:6'), covar=tensor([0.7752, 0.8283, 0.0976, 0.2346, 0.1301, 0.2480, 0.1945, 0.2061], device='cuda:6'), in_proj_covar=tensor([0.0313, 0.0305, 0.0413, 0.0417, 0.0351, 0.0408, 0.0316, 0.0372], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-26 18:11:38,719 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.78 vs. limit=2.0 2023-04-26 18:11:42,889 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=31823.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:11:51,414 INFO [optim.py:369] (6/7) 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 18:12:02,482 INFO [finetune.py:976] (6/7) Epoch 6, batch 3200, loss[loss=0.2255, simple_loss=0.2829, pruned_loss=0.08402, over 4916.00 frames. ], tot_loss[loss=0.2021, simple_loss=0.2633, pruned_loss=0.07045, over 953809.44 frames. ], batch size: 36, lr: 3.90e-03, grad_scale: 32.0 2023-04-26 18:12:14,984 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=31847.0, num_to_drop=1, layers_to_drop={2} 2023-04-26 18:12:16,344 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.03 vs. limit=5.0 2023-04-26 18:12:22,015 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=31849.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:12:36,968 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6646, 1.2176, 1.2592, 1.3890, 1.8613, 1.4823, 1.1490, 1.2523], device='cuda:6'), covar=tensor([0.1926, 0.1646, 0.2261, 0.1573, 0.0918, 0.1554, 0.2530, 0.2043], device='cuda:6'), in_proj_covar=tensor([0.0315, 0.0338, 0.0356, 0.0310, 0.0346, 0.0340, 0.0314, 0.0357], device='cuda:6'), 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:6') 2023-04-26 18:12:55,832 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7247, 3.6260, 1.1237, 1.9394, 2.0595, 2.5416, 2.1447, 1.0319], device='cuda:6'), covar=tensor([0.1236, 0.0926, 0.1882, 0.1261, 0.1060, 0.1140, 0.1423, 0.2003], device='cuda:6'), in_proj_covar=tensor([0.0122, 0.0260, 0.0146, 0.0127, 0.0137, 0.0160, 0.0123, 0.0127], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-26 18:12:57,062 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=31878.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:12:58,864 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.1554, 2.3663, 1.0429, 1.4948, 1.4533, 1.8710, 1.5974, 0.9162], device='cuda:6'), covar=tensor([0.1154, 0.1130, 0.1351, 0.1119, 0.1006, 0.0798, 0.1239, 0.1788], device='cuda:6'), in_proj_covar=tensor([0.0122, 0.0260, 0.0146, 0.0127, 0.0137, 0.0160, 0.0123, 0.0127], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-26 18:13:06,263 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=31884.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:13:09,139 INFO [finetune.py:976] (6/7) Epoch 6, batch 3250, loss[loss=0.2544, simple_loss=0.3086, pruned_loss=0.1001, over 4908.00 frames. ], tot_loss[loss=0.2037, simple_loss=0.2651, pruned_loss=0.07119, over 952773.06 frames. ], batch size: 37, lr: 3.90e-03, grad_scale: 32.0 2023-04-26 18:13:09,381 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=6.15 vs. limit=5.0 2023-04-26 18:13:29,348 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.3377, 1.4125, 1.3771, 2.0694, 1.6192, 2.0422, 1.5156, 4.2940], device='cuda:6'), covar=tensor([0.0689, 0.0864, 0.0981, 0.1254, 0.0722, 0.0557, 0.0836, 0.0174], device='cuda:6'), in_proj_covar=tensor([0.0039, 0.0040, 0.0041, 0.0045, 0.0041, 0.0040, 0.0040, 0.0061], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0017], device='cuda:6') 2023-04-26 18:13:39,528 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=31915.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:13:51,193 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7867, 1.2484, 1.5903, 1.5600, 1.5172, 1.2190, 0.6299, 1.2300], device='cuda:6'), covar=tensor([0.4196, 0.4429, 0.2070, 0.3037, 0.3679, 0.3433, 0.5482, 0.3077], device='cuda:6'), in_proj_covar=tensor([0.0278, 0.0257, 0.0220, 0.0328, 0.0217, 0.0230, 0.0243, 0.0193], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-26 18:13:57,722 INFO [optim.py:369] (6/7) 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] (6/7) Epoch 6, batch 3300, loss[loss=0.2119, simple_loss=0.2795, pruned_loss=0.07212, over 4798.00 frames. ], tot_loss[loss=0.2052, simple_loss=0.2677, pruned_loss=0.07142, over 952494.58 frames. ], batch size: 45, lr: 3.90e-03, grad_scale: 32.0 2023-04-26 18:14:10,321 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=31939.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:14:18,850 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=31950.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:14:35,916 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=31976.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:14:43,707 INFO [finetune.py:976] (6/7) Epoch 6, batch 3350, loss[loss=0.1888, simple_loss=0.2579, pruned_loss=0.05987, over 4792.00 frames. ], tot_loss[loss=0.207, simple_loss=0.2702, pruned_loss=0.07192, over 953839.82 frames. ], batch size: 51, lr: 3.90e-03, grad_scale: 32.0 2023-04-26 18:15:00,880 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=32011.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:15:06,047 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.2827, 1.7399, 1.4730, 1.8920, 1.7777, 2.1870, 1.5115, 3.6437], device='cuda:6'), covar=tensor([0.0615, 0.0717, 0.0813, 0.1136, 0.0598, 0.0429, 0.0721, 0.0127], device='cuda:6'), in_proj_covar=tensor([0.0039, 0.0039, 0.0041, 0.0045, 0.0041, 0.0040, 0.0040, 0.0061], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0017], device='cuda:6') 2023-04-26 18:15:12,031 INFO [optim.py:369] (6/7) 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] (6/7) Epoch 6, batch 3400, loss[loss=0.2222, simple_loss=0.2802, pruned_loss=0.08206, over 4828.00 frames. ], tot_loss[loss=0.2083, simple_loss=0.271, pruned_loss=0.07276, over 951343.47 frames. ], batch size: 47, lr: 3.90e-03, grad_scale: 32.0 2023-04-26 18:15:33,362 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9722, 1.7900, 1.9407, 2.3072, 1.7180, 1.4858, 1.8037, 1.0787], device='cuda:6'), covar=tensor([0.0843, 0.1076, 0.0853, 0.0888, 0.1081, 0.1434, 0.0965, 0.1319], device='cuda:6'), in_proj_covar=tensor([0.0067, 0.0075, 0.0074, 0.0068, 0.0078, 0.0095, 0.0081, 0.0077], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-26 18:15:55,961 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=32060.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:16:04,740 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=32065.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:16:20,376 INFO [finetune.py:976] (6/7) Epoch 6, batch 3450, loss[loss=0.2, simple_loss=0.2549, pruned_loss=0.07255, over 4836.00 frames. ], tot_loss[loss=0.2061, simple_loss=0.2695, pruned_loss=0.07134, over 951769.99 frames. ], batch size: 30, lr: 3.90e-03, grad_scale: 32.0 2023-04-26 18:16:30,202 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.97 vs. limit=2.0 2023-04-26 18:16:37,416 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=32113.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:16:47,433 INFO [optim.py:369] (6/7) 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 18:16:47,575 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([0.2883, 1.3567, 1.4404, 0.9219, 1.3809, 1.1969, 1.7282, 1.4257], device='cuda:6'), covar=tensor([0.4068, 0.1810, 0.4910, 0.2882, 0.1659, 0.2375, 0.1732, 0.4666], device='cuda:6'), in_proj_covar=tensor([0.0348, 0.0355, 0.0435, 0.0366, 0.0392, 0.0383, 0.0388, 0.0416], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-26 18:16:54,109 INFO [finetune.py:976] (6/7) Epoch 6, batch 3500, loss[loss=0.2259, simple_loss=0.2596, pruned_loss=0.09607, over 4728.00 frames. ], tot_loss[loss=0.2035, simple_loss=0.2661, pruned_loss=0.07039, over 952026.71 frames. ], batch size: 23, lr: 3.90e-03, grad_scale: 32.0 2023-04-26 18:16:57,315 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=32144.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:16:59,136 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=32147.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 18:17:21,487 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 6, batch 3550, loss[loss=0.2068, simple_loss=0.2651, pruned_loss=0.07419, over 4837.00 frames. ], tot_loss[loss=0.2025, simple_loss=0.2642, pruned_loss=0.07036, over 953215.98 frames. ], batch size: 30, lr: 3.90e-03, grad_scale: 32.0 2023-04-26 18:17:43,433 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=32195.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 18:18:29,258 INFO [optim.py:369] (6/7) 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 18:18:38,662 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5073, 1.3606, 1.7897, 1.7546, 1.4315, 1.1169, 1.5825, 1.0171], device='cuda:6'), covar=tensor([0.0884, 0.0943, 0.0551, 0.0889, 0.0886, 0.1309, 0.0779, 0.1034], device='cuda:6'), in_proj_covar=tensor([0.0067, 0.0074, 0.0073, 0.0067, 0.0078, 0.0094, 0.0081, 0.0076], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-26 18:18:39,230 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=32234.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:18:47,837 INFO [finetune.py:976] (6/7) Epoch 6, batch 3600, loss[loss=0.1897, simple_loss=0.2516, pruned_loss=0.06388, over 4813.00 frames. ], tot_loss[loss=0.2007, simple_loss=0.2621, pruned_loss=0.06964, over 955593.58 frames. ], batch size: 38, lr: 3.90e-03, grad_scale: 32.0 2023-04-26 18:18:51,337 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-04-26 18:19:10,106 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-04-26 18:19:22,891 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6619, 1.6423, 1.8869, 2.2918, 2.1308, 1.6896, 1.4908, 1.7901], device='cuda:6'), covar=tensor([0.1032, 0.1163, 0.0698, 0.0629, 0.0734, 0.0956, 0.0940, 0.0750], device='cuda:6'), in_proj_covar=tensor([0.0202, 0.0207, 0.0181, 0.0179, 0.0180, 0.0195, 0.0166, 0.0189], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-26 18:19:31,473 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=32271.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:19:55,461 INFO [finetune.py:976] (6/7) Epoch 6, batch 3650, loss[loss=0.2279, simple_loss=0.294, pruned_loss=0.0809, over 4903.00 frames. ], tot_loss[loss=0.2037, simple_loss=0.2653, pruned_loss=0.07103, over 955399.70 frames. ], batch size: 35, lr: 3.90e-03, grad_scale: 32.0 2023-04-26 18:20:16,843 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=32306.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:20:27,140 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 2023-04-26 18:20:43,392 INFO [optim.py:369] (6/7) 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 18:21:00,977 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.1162, 1.9747, 2.1586, 2.6281, 2.4531, 2.0525, 1.7727, 2.1956], device='cuda:6'), covar=tensor([0.0895, 0.0979, 0.0701, 0.0569, 0.0636, 0.0901, 0.0873, 0.0617], device='cuda:6'), in_proj_covar=tensor([0.0203, 0.0207, 0.0182, 0.0180, 0.0181, 0.0196, 0.0166, 0.0190], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-26 18:21:03,305 INFO [finetune.py:976] (6/7) Epoch 6, batch 3700, loss[loss=0.2321, simple_loss=0.2915, pruned_loss=0.08635, over 4803.00 frames. ], tot_loss[loss=0.2082, simple_loss=0.2703, pruned_loss=0.07303, over 956536.60 frames. ], batch size: 45, lr: 3.90e-03, grad_scale: 32.0 2023-04-26 18:21:26,776 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=32360.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:21:34,056 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.4854, 1.3298, 1.7143, 1.7519, 1.3813, 0.9559, 1.4442, 0.9396], device='cuda:6'), covar=tensor([0.0799, 0.0610, 0.0558, 0.0639, 0.0820, 0.1874, 0.0751, 0.1113], device='cuda:6'), in_proj_covar=tensor([0.0067, 0.0075, 0.0074, 0.0068, 0.0078, 0.0095, 0.0081, 0.0077], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-26 18:22:04,006 INFO [finetune.py:976] (6/7) Epoch 6, batch 3750, loss[loss=0.2336, simple_loss=0.2866, pruned_loss=0.09028, over 4825.00 frames. ], tot_loss[loss=0.2071, simple_loss=0.2693, pruned_loss=0.07242, over 953354.95 frames. ], batch size: 47, lr: 3.90e-03, grad_scale: 32.0 2023-04-26 18:22:21,628 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=32408.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:22:35,218 INFO [optim.py:369] (6/7) 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] (6/7) Epoch 6, batch 3800, loss[loss=0.2231, simple_loss=0.2912, pruned_loss=0.07752, over 4915.00 frames. ], tot_loss[loss=0.208, simple_loss=0.2704, pruned_loss=0.07281, over 952304.27 frames. ], batch size: 33, lr: 3.90e-03, grad_scale: 32.0 2023-04-26 18:22:53,228 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=32444.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:23:21,565 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=32479.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:23:33,780 INFO [finetune.py:976] (6/7) Epoch 6, batch 3850, loss[loss=0.1879, simple_loss=0.2639, pruned_loss=0.05591, over 4784.00 frames. ], tot_loss[loss=0.2066, simple_loss=0.2692, pruned_loss=0.07197, over 953518.07 frames. ], batch size: 29, lr: 3.90e-03, grad_scale: 32.0 2023-04-26 18:23:41,607 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=32492.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:24:04,928 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6043, 1.4529, 1.8843, 1.9025, 1.4726, 1.2813, 1.5860, 1.0214], device='cuda:6'), covar=tensor([0.0709, 0.0874, 0.0543, 0.0851, 0.0940, 0.1324, 0.0788, 0.0992], device='cuda:6'), in_proj_covar=tensor([0.0067, 0.0075, 0.0074, 0.0067, 0.0078, 0.0095, 0.0081, 0.0076], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-26 18:24:18,831 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=32527.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:24:19,368 INFO [optim.py:369] (6/7) 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 18:24:24,632 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=32534.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:24:28,027 INFO [finetune.py:976] (6/7) Epoch 6, batch 3900, loss[loss=0.2283, simple_loss=0.2768, pruned_loss=0.0899, over 4927.00 frames. ], tot_loss[loss=0.2044, simple_loss=0.2668, pruned_loss=0.07107, over 953964.50 frames. ], batch size: 33, lr: 3.90e-03, grad_scale: 32.0 2023-04-26 18:24:41,085 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0913, 1.5767, 1.9286, 2.1768, 1.8980, 1.5258, 1.1817, 1.6350], device='cuda:6'), covar=tensor([0.3830, 0.4015, 0.1797, 0.3142, 0.3243, 0.3121, 0.5213, 0.2904], device='cuda:6'), in_proj_covar=tensor([0.0277, 0.0255, 0.0219, 0.0327, 0.0216, 0.0228, 0.0241, 0.0192], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-26 18:25:09,812 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=32571.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:25:22,755 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 6, batch 3950, loss[loss=0.1719, simple_loss=0.2369, pruned_loss=0.0534, over 4748.00 frames. ], tot_loss[loss=0.2014, simple_loss=0.2636, pruned_loss=0.06961, over 955373.63 frames. ], batch size: 59, lr: 3.90e-03, grad_scale: 16.0 2023-04-26 18:25:50,958 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=32606.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:25:58,795 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=32619.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:26:10,338 INFO [optim.py:369] (6/7) 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 18:26:22,978 INFO [finetune.py:976] (6/7) Epoch 6, batch 4000, loss[loss=0.207, simple_loss=0.2655, pruned_loss=0.0743, over 4133.00 frames. ], tot_loss[loss=0.202, simple_loss=0.2634, pruned_loss=0.07035, over 952189.16 frames. ], batch size: 65, lr: 3.90e-03, grad_scale: 16.0 2023-04-26 18:26:41,150 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2023-04-26 18:26:51,941 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=32654.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:26:52,661 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.6747, 1.8884, 1.7313, 1.9574, 1.7535, 2.0344, 1.8666, 1.7931], device='cuda:6'), covar=tensor([0.5568, 1.0101, 0.9149, 0.8012, 0.8751, 1.1952, 1.0890, 0.9863], device='cuda:6'), in_proj_covar=tensor([0.0318, 0.0396, 0.0319, 0.0328, 0.0346, 0.0412, 0.0375, 0.0333], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-26 18:27:19,405 INFO [finetune.py:976] (6/7) Epoch 6, batch 4050, loss[loss=0.2622, simple_loss=0.3186, pruned_loss=0.1029, over 4813.00 frames. ], tot_loss[loss=0.2048, simple_loss=0.2667, pruned_loss=0.0715, over 952176.90 frames. ], batch size: 39, lr: 3.90e-03, grad_scale: 16.0 2023-04-26 18:27:46,364 INFO [optim.py:369] (6/7) 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 18:27:52,926 INFO [finetune.py:976] (6/7) Epoch 6, batch 4100, loss[loss=0.2025, simple_loss=0.2704, pruned_loss=0.06733, over 4754.00 frames. ], tot_loss[loss=0.2082, simple_loss=0.2702, pruned_loss=0.07314, over 950507.15 frames. ], batch size: 28, lr: 3.90e-03, grad_scale: 16.0 2023-04-26 18:28:19,296 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.50 vs. limit=2.0 2023-04-26 18:28:26,742 INFO [finetune.py:976] (6/7) Epoch 6, batch 4150, loss[loss=0.2392, simple_loss=0.3089, pruned_loss=0.08471, over 4906.00 frames. ], tot_loss[loss=0.2098, simple_loss=0.2717, pruned_loss=0.07393, over 949785.80 frames. ], batch size: 36, lr: 3.90e-03, grad_scale: 16.0 2023-04-26 18:28:31,197 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.19 vs. limit=5.0 2023-04-26 18:29:01,021 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.13 vs. limit=2.0 2023-04-26 18:29:05,517 INFO [optim.py:369] (6/7) 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 18:29:12,124 INFO [finetune.py:976] (6/7) Epoch 6, batch 4200, loss[loss=0.2317, simple_loss=0.2834, pruned_loss=0.08996, over 4894.00 frames. ], tot_loss[loss=0.2083, simple_loss=0.2709, pruned_loss=0.07287, over 949625.83 frames. ], batch size: 43, lr: 3.90e-03, grad_scale: 16.0 2023-04-26 18:29:15,282 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([0.9085, 1.3219, 3.2456, 3.0097, 2.9464, 3.1442, 3.0985, 2.8451], device='cuda:6'), covar=tensor([0.7253, 0.5204, 0.1531, 0.2147, 0.1405, 0.2218, 0.2578, 0.1748], device='cuda:6'), in_proj_covar=tensor([0.0314, 0.0307, 0.0415, 0.0420, 0.0354, 0.0411, 0.0319, 0.0372], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-26 18:29:45,431 INFO [finetune.py:976] (6/7) Epoch 6, batch 4250, loss[loss=0.2309, simple_loss=0.2809, pruned_loss=0.09045, over 4889.00 frames. ], tot_loss[loss=0.2067, simple_loss=0.2689, pruned_loss=0.0723, over 950450.18 frames. ], batch size: 35, lr: 3.90e-03, grad_scale: 16.0 2023-04-26 18:30:13,034 INFO [optim.py:369] (6/7) 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 18:30:19,130 INFO [finetune.py:976] (6/7) Epoch 6, batch 4300, loss[loss=0.1636, simple_loss=0.2299, pruned_loss=0.04865, over 4755.00 frames. ], tot_loss[loss=0.203, simple_loss=0.2647, pruned_loss=0.07062, over 949625.29 frames. ], batch size: 23, lr: 3.90e-03, grad_scale: 16.0 2023-04-26 18:30:47,977 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=32966.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:31:19,600 INFO [finetune.py:976] (6/7) Epoch 6, batch 4350, loss[loss=0.2148, simple_loss=0.2914, pruned_loss=0.06912, over 4805.00 frames. ], tot_loss[loss=0.2001, simple_loss=0.2616, pruned_loss=0.06927, over 949609.44 frames. ], batch size: 39, lr: 3.90e-03, grad_scale: 16.0 2023-04-26 18:32:12,349 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=33027.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:32:13,433 INFO [optim.py:369] (6/7) 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 18:32:24,805 INFO [finetune.py:976] (6/7) Epoch 6, batch 4400, loss[loss=0.2289, simple_loss=0.2994, pruned_loss=0.07923, over 4177.00 frames. ], tot_loss[loss=0.2013, simple_loss=0.2627, pruned_loss=0.06995, over 950618.93 frames. ], batch size: 66, lr: 3.90e-03, grad_scale: 16.0 2023-04-26 18:33:06,917 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=33069.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:33:29,663 INFO [finetune.py:976] (6/7) Epoch 6, batch 4450, loss[loss=0.2024, simple_loss=0.2683, pruned_loss=0.06825, over 4855.00 frames. ], tot_loss[loss=0.2044, simple_loss=0.2665, pruned_loss=0.07117, over 950527.68 frames. ], batch size: 31, lr: 3.90e-03, grad_scale: 16.0 2023-04-26 18:34:11,452 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.2982, 3.0286, 0.9739, 1.6566, 1.7636, 2.2094, 1.7533, 0.9694], device='cuda:6'), covar=tensor([0.1470, 0.0958, 0.1869, 0.1378, 0.1142, 0.1026, 0.1697, 0.1880], device='cuda:6'), in_proj_covar=tensor([0.0122, 0.0259, 0.0144, 0.0127, 0.0138, 0.0159, 0.0123, 0.0126], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-26 18:34:12,570 INFO [optim.py:369] (6/7) 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 18:34:13,330 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 6, batch 4500, loss[loss=0.2199, simple_loss=0.2743, pruned_loss=0.08274, over 4827.00 frames. ], tot_loss[loss=0.2055, simple_loss=0.2683, pruned_loss=0.07134, over 953096.89 frames. ], batch size: 30, lr: 3.89e-03, grad_scale: 16.0 2023-04-26 18:34:29,019 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=33155.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:34:31,469 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5018, 1.7298, 1.3709, 0.9590, 1.2316, 1.1864, 1.3570, 1.1001], device='cuda:6'), covar=tensor([0.1813, 0.1334, 0.1637, 0.1995, 0.2495, 0.1950, 0.1169, 0.2163], device='cuda:6'), in_proj_covar=tensor([0.0193, 0.0217, 0.0174, 0.0204, 0.0209, 0.0185, 0.0166, 0.0191], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:6') 2023-04-26 18:34:43,549 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-04-26 18:34:52,374 INFO [finetune.py:976] (6/7) Epoch 6, batch 4550, loss[loss=0.3469, simple_loss=0.3748, pruned_loss=0.1595, over 4167.00 frames. ], tot_loss[loss=0.2085, simple_loss=0.2705, pruned_loss=0.07327, over 950890.90 frames. ], batch size: 66, lr: 3.89e-03, grad_scale: 16.0 2023-04-26 18:35:09,486 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=33216.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:35:18,795 INFO [optim.py:369] (6/7) 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 18:35:25,831 INFO [finetune.py:976] (6/7) Epoch 6, batch 4600, loss[loss=0.1681, simple_loss=0.2271, pruned_loss=0.05452, over 4706.00 frames. ], tot_loss[loss=0.2072, simple_loss=0.2697, pruned_loss=0.07239, over 951098.57 frames. ], batch size: 23, lr: 3.89e-03, grad_scale: 16.0 2023-04-26 18:35:52,354 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-04-26 18:35:59,783 INFO [finetune.py:976] (6/7) Epoch 6, batch 4650, loss[loss=0.2175, simple_loss=0.2773, pruned_loss=0.07881, over 4932.00 frames. ], tot_loss[loss=0.2051, simple_loss=0.2672, pruned_loss=0.07146, over 951151.36 frames. ], batch size: 46, lr: 3.89e-03, grad_scale: 16.0 2023-04-26 18:36:06,228 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.14 vs. limit=2.0 2023-04-26 18:36:20,282 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=33322.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:36:25,472 INFO [optim.py:369] (6/7) 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] (6/7) Epoch 6, batch 4700, loss[loss=0.1619, simple_loss=0.2326, pruned_loss=0.04561, over 4881.00 frames. ], tot_loss[loss=0.202, simple_loss=0.2636, pruned_loss=0.07018, over 953094.07 frames. ], batch size: 31, lr: 3.89e-03, grad_scale: 16.0 2023-04-26 18:37:20,898 INFO [finetune.py:976] (6/7) Epoch 6, batch 4750, loss[loss=0.2176, simple_loss=0.2546, pruned_loss=0.09029, over 4793.00 frames. ], tot_loss[loss=0.2026, simple_loss=0.2633, pruned_loss=0.07095, over 953369.91 frames. ], batch size: 25, lr: 3.89e-03, grad_scale: 16.0 2023-04-26 18:37:42,302 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.1659, 1.5404, 1.3876, 1.6527, 1.5556, 1.8798, 1.3632, 3.5070], device='cuda:6'), covar=tensor([0.0657, 0.0781, 0.0813, 0.1228, 0.0656, 0.0572, 0.0778, 0.0175], device='cuda:6'), in_proj_covar=tensor([0.0039, 0.0040, 0.0041, 0.0045, 0.0041, 0.0040, 0.0040, 0.0062], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0017], device='cuda:6') 2023-04-26 18:38:05,448 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([3.6513, 3.6120, 2.6244, 4.2346, 3.7062, 3.6103, 1.5900, 3.5830], device='cuda:6'), covar=tensor([0.1932, 0.1321, 0.3699, 0.1698, 0.3521, 0.1895, 0.5580, 0.2528], device='cuda:6'), in_proj_covar=tensor([0.0244, 0.0219, 0.0253, 0.0312, 0.0303, 0.0255, 0.0275, 0.0274], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-26 18:38:06,073 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=33425.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:38:14,207 INFO [optim.py:369] (6/7) 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 18:38:27,281 INFO [finetune.py:976] (6/7) Epoch 6, batch 4800, loss[loss=0.2274, simple_loss=0.3038, pruned_loss=0.07548, over 4924.00 frames. ], tot_loss[loss=0.2061, simple_loss=0.2673, pruned_loss=0.07247, over 954208.09 frames. ], batch size: 36, lr: 3.89e-03, grad_scale: 16.0 2023-04-26 18:38:27,411 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0743, 1.7602, 1.9829, 2.4204, 2.3871, 1.8279, 1.5936, 2.1994], device='cuda:6'), covar=tensor([0.0713, 0.1042, 0.0632, 0.0509, 0.0491, 0.0882, 0.0825, 0.0465], device='cuda:6'), in_proj_covar=tensor([0.0202, 0.0206, 0.0179, 0.0178, 0.0179, 0.0195, 0.0165, 0.0187], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-26 18:39:01,137 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=33472.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:39:12,307 INFO [finetune.py:976] (6/7) Epoch 6, batch 4850, loss[loss=0.2062, simple_loss=0.2767, pruned_loss=0.06781, over 4809.00 frames. ], tot_loss[loss=0.2079, simple_loss=0.2701, pruned_loss=0.07288, over 955914.75 frames. ], batch size: 41, lr: 3.89e-03, grad_scale: 16.0 2023-04-26 18:39:15,359 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5969, 1.0748, 1.1826, 1.2649, 1.8231, 1.3765, 1.0949, 1.1419], device='cuda:6'), covar=tensor([0.1686, 0.1667, 0.2248, 0.1588, 0.0775, 0.1862, 0.2106, 0.2187], device='cuda:6'), in_proj_covar=tensor([0.0311, 0.0333, 0.0353, 0.0305, 0.0342, 0.0334, 0.0310, 0.0355], device='cuda:6'), out_proj_covar=tensor([6.6410e-05, 7.1010e-05, 7.6467e-05, 6.3353e-05, 7.2036e-05, 7.2120e-05, 6.7111e-05, 7.6499e-05], device='cuda:6') 2023-04-26 18:39:17,187 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8363, 1.5634, 1.8042, 2.1915, 2.2132, 1.6669, 1.3641, 1.9390], device='cuda:6'), covar=tensor([0.0909, 0.1175, 0.0667, 0.0610, 0.0593, 0.1017, 0.1001, 0.0613], device='cuda:6'), in_proj_covar=tensor([0.0202, 0.0206, 0.0180, 0.0179, 0.0180, 0.0196, 0.0165, 0.0188], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-26 18:39:27,161 INFO [zipformer.py:1188] (6/7) 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] (6/7) 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 18:39:41,513 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=33533.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:39:44,986 INFO [finetune.py:976] (6/7) Epoch 6, batch 4900, loss[loss=0.2052, simple_loss=0.2638, pruned_loss=0.07332, over 4862.00 frames. ], tot_loss[loss=0.21, simple_loss=0.2726, pruned_loss=0.07374, over 957463.41 frames. ], batch size: 31, lr: 3.89e-03, grad_scale: 16.0 2023-04-26 18:40:14,912 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.2922, 2.9937, 0.8953, 1.6903, 1.7418, 2.1465, 1.7644, 0.9425], device='cuda:6'), covar=tensor([0.1486, 0.0932, 0.1975, 0.1301, 0.1131, 0.1013, 0.1607, 0.1881], device='cuda:6'), in_proj_covar=tensor([0.0121, 0.0259, 0.0144, 0.0126, 0.0137, 0.0159, 0.0122, 0.0125], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-26 18:40:18,320 INFO [finetune.py:976] (6/7) Epoch 6, batch 4950, loss[loss=0.2126, simple_loss=0.2848, pruned_loss=0.07015, over 4776.00 frames. ], tot_loss[loss=0.2097, simple_loss=0.2732, pruned_loss=0.07314, over 956269.59 frames. ], batch size: 51, lr: 3.89e-03, grad_scale: 16.0 2023-04-26 18:40:40,490 INFO [zipformer.py:1188] (6/7) 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] (6/7) 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 18:40:51,674 INFO [finetune.py:976] (6/7) Epoch 6, batch 5000, loss[loss=0.1909, simple_loss=0.2559, pruned_loss=0.06291, over 4787.00 frames. ], tot_loss[loss=0.2066, simple_loss=0.2701, pruned_loss=0.07156, over 956702.24 frames. ], batch size: 45, lr: 3.89e-03, grad_scale: 16.0 2023-04-26 18:41:24,497 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=33670.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:41:28,794 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=33677.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:41:36,417 INFO [finetune.py:976] (6/7) Epoch 6, batch 5050, loss[loss=0.2165, simple_loss=0.2763, pruned_loss=0.07836, over 4285.00 frames. ], tot_loss[loss=0.2056, simple_loss=0.2683, pruned_loss=0.07141, over 956115.16 frames. ], batch size: 65, lr: 3.89e-03, grad_scale: 16.0 2023-04-26 18:41:55,782 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.75 vs. limit=2.0 2023-04-26 18:42:01,184 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=33725.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:42:03,505 INFO [optim.py:369] (6/7) 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 18:42:14,949 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=33738.0, num_to_drop=1, layers_to_drop={2} 2023-04-26 18:42:15,447 INFO [finetune.py:976] (6/7) Epoch 6, batch 5100, loss[loss=0.2247, simple_loss=0.2786, pruned_loss=0.08539, over 4825.00 frames. ], tot_loss[loss=0.2023, simple_loss=0.2647, pruned_loss=0.06996, over 956835.59 frames. ], batch size: 40, lr: 3.89e-03, grad_scale: 16.0 2023-04-26 18:42:27,892 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([0.9587, 2.0786, 1.8282, 1.7193, 2.1700, 1.7908, 2.8417, 1.5644], device='cuda:6'), covar=tensor([0.4839, 0.2077, 0.5728, 0.3566, 0.2266, 0.2858, 0.1450, 0.5077], device='cuda:6'), in_proj_covar=tensor([0.0346, 0.0353, 0.0436, 0.0362, 0.0393, 0.0383, 0.0384, 0.0417], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-26 18:42:38,208 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-04-26 18:43:00,677 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=33773.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:43:21,824 INFO [finetune.py:976] (6/7) Epoch 6, batch 5150, loss[loss=0.2379, simple_loss=0.2964, pruned_loss=0.08974, over 4833.00 frames. ], tot_loss[loss=0.2047, simple_loss=0.2668, pruned_loss=0.07132, over 954730.71 frames. ], batch size: 33, lr: 3.89e-03, grad_scale: 16.0 2023-04-26 18:43:42,896 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=33811.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:43:54,702 INFO [zipformer.py:1188] (6/7) 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] (6/7) 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] (6/7) Epoch 6, batch 5200, loss[loss=0.284, simple_loss=0.3361, pruned_loss=0.1159, over 4798.00 frames. ], tot_loss[loss=0.2078, simple_loss=0.2705, pruned_loss=0.07253, over 955455.49 frames. ], batch size: 41, lr: 3.89e-03, grad_scale: 16.0 2023-04-26 18:44:38,054 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=33859.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:44:48,328 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0782, 1.4440, 1.9063, 2.4675, 1.8960, 1.4822, 1.2905, 1.8076], device='cuda:6'), covar=tensor([0.3769, 0.4434, 0.2154, 0.3296, 0.3688, 0.3541, 0.5639, 0.3256], device='cuda:6'), in_proj_covar=tensor([0.0278, 0.0256, 0.0220, 0.0325, 0.0216, 0.0230, 0.0241, 0.0191], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-26 18:45:14,786 INFO [finetune.py:976] (6/7) Epoch 6, batch 5250, loss[loss=0.1964, simple_loss=0.256, pruned_loss=0.06838, over 4691.00 frames. ], tot_loss[loss=0.2084, simple_loss=0.2723, pruned_loss=0.07224, over 958255.61 frames. ], batch size: 23, lr: 3.89e-03, grad_scale: 16.0 2023-04-26 18:45:58,333 INFO [optim.py:369] (6/7) 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 18:46:04,414 INFO [finetune.py:976] (6/7) Epoch 6, batch 5300, loss[loss=0.2795, simple_loss=0.3239, pruned_loss=0.1176, over 4167.00 frames. ], tot_loss[loss=0.209, simple_loss=0.273, pruned_loss=0.07253, over 957687.27 frames. ], batch size: 66, lr: 3.89e-03, grad_scale: 16.0 2023-04-26 18:46:04,527 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 6, batch 5350, loss[loss=0.2462, simple_loss=0.2886, pruned_loss=0.1019, over 4909.00 frames. ], tot_loss[loss=0.2098, simple_loss=0.2736, pruned_loss=0.07298, over 955545.92 frames. ], batch size: 36, lr: 3.89e-03, grad_scale: 16.0 2023-04-26 18:46:46,572 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=34000.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:46:52,264 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.81 vs. limit=2.0 2023-04-26 18:47:06,984 INFO [optim.py:369] (6/7) 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 18:47:09,498 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=34033.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 18:47:13,056 INFO [finetune.py:976] (6/7) Epoch 6, batch 5400, loss[loss=0.2047, simple_loss=0.2644, pruned_loss=0.07254, over 4818.00 frames. ], tot_loss[loss=0.2072, simple_loss=0.2701, pruned_loss=0.07213, over 955878.64 frames. ], batch size: 39, lr: 3.89e-03, grad_scale: 16.0 2023-04-26 18:47:16,186 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=34044.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:47:22,728 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.16 vs. limit=5.0 2023-04-26 18:47:25,622 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6735, 1.2081, 1.6547, 2.0636, 1.7678, 1.5927, 1.6095, 1.6522], device='cuda:6'), covar=tensor([0.7906, 1.1780, 1.2153, 1.1943, 0.9507, 1.3081, 1.3992, 1.2086], device='cuda:6'), in_proj_covar=tensor([0.0411, 0.0437, 0.0522, 0.0543, 0.0440, 0.0459, 0.0473, 0.0471], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-26 18:47:30,686 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 2023-04-26 18:47:44,528 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=5.54 vs. limit=5.0 2023-04-26 18:47:46,754 INFO [finetune.py:976] (6/7) Epoch 6, batch 5450, loss[loss=0.1571, simple_loss=0.2341, pruned_loss=0.04001, over 4793.00 frames. ], tot_loss[loss=0.2036, simple_loss=0.2663, pruned_loss=0.07049, over 954174.81 frames. ], batch size: 29, lr: 3.89e-03, grad_scale: 16.0 2023-04-26 18:47:54,789 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5737, 0.9262, 1.5131, 1.8750, 1.6083, 1.4695, 1.5111, 1.5770], device='cuda:6'), covar=tensor([1.1125, 1.4406, 1.6104, 1.8984, 1.2816, 1.8227, 1.8209, 1.4166], device='cuda:6'), in_proj_covar=tensor([0.0412, 0.0438, 0.0523, 0.0544, 0.0441, 0.0461, 0.0475, 0.0472], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-26 18:47:57,071 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=34105.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:48:12,449 INFO [zipformer.py:1188] (6/7) 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] (6/7) 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 18:48:25,399 INFO [finetune.py:976] (6/7) Epoch 6, batch 5500, loss[loss=0.2005, simple_loss=0.2625, pruned_loss=0.06924, over 4875.00 frames. ], tot_loss[loss=0.201, simple_loss=0.2632, pruned_loss=0.06943, over 955325.12 frames. ], batch size: 34, lr: 3.89e-03, grad_scale: 16.0 2023-04-26 18:48:29,210 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=34145.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:48:48,536 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7652, 1.3458, 2.0067, 1.9520, 1.4922, 1.3607, 1.6596, 1.1770], device='cuda:6'), covar=tensor([0.0572, 0.1591, 0.0660, 0.0889, 0.0980, 0.1451, 0.0801, 0.0894], device='cuda:6'), in_proj_covar=tensor([0.0068, 0.0076, 0.0074, 0.0068, 0.0079, 0.0096, 0.0081, 0.0078], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-26 18:49:10,834 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=34176.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:49:30,904 INFO [finetune.py:976] (6/7) Epoch 6, batch 5550, loss[loss=0.2047, simple_loss=0.2517, pruned_loss=0.07881, over 4247.00 frames. ], tot_loss[loss=0.2035, simple_loss=0.2652, pruned_loss=0.07086, over 954509.21 frames. ], batch size: 18, lr: 3.89e-03, grad_scale: 16.0 2023-04-26 18:49:52,182 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=34206.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:50:02,836 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-04-26 18:50:24,791 INFO [optim.py:369] (6/7) 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 18:50:36,988 INFO [finetune.py:976] (6/7) Epoch 6, batch 5600, loss[loss=0.2023, simple_loss=0.2793, pruned_loss=0.0626, over 4899.00 frames. ], tot_loss[loss=0.2084, simple_loss=0.271, pruned_loss=0.07287, over 956238.80 frames. ], batch size: 37, lr: 3.89e-03, grad_scale: 16.0 2023-04-26 18:51:28,269 INFO [finetune.py:976] (6/7) Epoch 6, batch 5650, loss[loss=0.1692, simple_loss=0.2462, pruned_loss=0.04609, over 4914.00 frames. ], tot_loss[loss=0.2107, simple_loss=0.2738, pruned_loss=0.07385, over 954205.93 frames. ], batch size: 37, lr: 3.89e-03, grad_scale: 16.0 2023-04-26 18:51:31,886 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=34295.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:51:49,212 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5278, 2.4186, 2.2505, 2.1673, 2.5753, 2.2889, 3.1190, 1.9616], device='cuda:6'), covar=tensor([0.3289, 0.1541, 0.3730, 0.2409, 0.1461, 0.1856, 0.1038, 0.3769], device='cuda:6'), in_proj_covar=tensor([0.0347, 0.0355, 0.0436, 0.0364, 0.0393, 0.0384, 0.0385, 0.0418], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-26 18:52:03,000 INFO [optim.py:369] (6/7) 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 18:52:05,462 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=34333.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:52:09,024 INFO [finetune.py:976] (6/7) Epoch 6, batch 5700, loss[loss=0.1872, simple_loss=0.2318, pruned_loss=0.0713, over 4142.00 frames. ], tot_loss[loss=0.207, simple_loss=0.2683, pruned_loss=0.07288, over 932822.22 frames. ], batch size: 17, lr: 3.89e-03, grad_scale: 16.0 2023-04-26 18:52:23,910 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8008, 1.4893, 2.0168, 2.0457, 1.5196, 1.2981, 1.5877, 1.1050], device='cuda:6'), covar=tensor([0.0737, 0.1020, 0.0643, 0.0738, 0.0871, 0.1461, 0.0774, 0.1094], device='cuda:6'), in_proj_covar=tensor([0.0068, 0.0075, 0.0073, 0.0068, 0.0078, 0.0096, 0.0081, 0.0077], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-26 18:52:39,658 INFO [finetune.py:976] (6/7) Epoch 7, batch 0, loss[loss=0.163, simple_loss=0.2439, pruned_loss=0.041, over 4898.00 frames. ], tot_loss[loss=0.163, simple_loss=0.2439, pruned_loss=0.041, over 4898.00 frames. ], batch size: 43, lr: 3.89e-03, grad_scale: 16.0 2023-04-26 18:52:39,658 INFO [finetune.py:1001] (6/7) Computing validation loss 2023-04-26 18:52:50,219 INFO [finetune.py:1010] (6/7) Epoch 7, validation: loss=0.1579, simple_loss=0.2317, pruned_loss=0.04207, over 2265189.00 frames. 2023-04-26 18:52:50,219 INFO [finetune.py:1011] (6/7) Maximum memory allocated so far is 6326MB 2023-04-26 18:52:59,341 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=34381.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:53:11,520 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=34400.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:53:23,181 INFO [finetune.py:976] (6/7) Epoch 7, batch 50, loss[loss=0.1564, simple_loss=0.2228, pruned_loss=0.04497, over 4760.00 frames. ], tot_loss[loss=0.202, simple_loss=0.2659, pruned_loss=0.06909, over 217105.11 frames. ], batch size: 27, lr: 3.89e-03, grad_scale: 16.0 2023-04-26 18:53:23,795 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.1592, 1.5670, 1.3309, 1.8004, 1.5252, 1.8862, 1.3757, 3.7089], device='cuda:6'), covar=tensor([0.0705, 0.0798, 0.0882, 0.1235, 0.0706, 0.0579, 0.0771, 0.0179], device='cuda:6'), in_proj_covar=tensor([0.0039, 0.0039, 0.0041, 0.0045, 0.0041, 0.0040, 0.0039, 0.0061], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0017], device='cuda:6') 2023-04-26 18:53:32,046 INFO [optim.py:369] (6/7) 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 18:53:34,635 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-04-26 18:53:56,442 INFO [finetune.py:976] (6/7) Epoch 7, batch 100, loss[loss=0.1696, simple_loss=0.2326, pruned_loss=0.05327, over 4820.00 frames. ], tot_loss[loss=0.1965, simple_loss=0.2589, pruned_loss=0.06702, over 380260.82 frames. ], batch size: 39, lr: 3.89e-03, grad_scale: 16.0 2023-04-26 18:54:12,170 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.2057, 1.5335, 1.3356, 1.7313, 1.5448, 1.9854, 1.2984, 3.5674], device='cuda:6'), covar=tensor([0.0657, 0.0776, 0.0847, 0.1215, 0.0664, 0.0587, 0.0789, 0.0185], device='cuda:6'), in_proj_covar=tensor([0.0039, 0.0039, 0.0041, 0.0044, 0.0040, 0.0040, 0.0039, 0.0060], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0014, 0.0013, 0.0013, 0.0013, 0.0017], device='cuda:6') 2023-04-26 18:54:19,342 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=34501.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:54:29,758 INFO [finetune.py:976] (6/7) Epoch 7, batch 150, loss[loss=0.2141, simple_loss=0.2732, pruned_loss=0.07748, over 4864.00 frames. ], tot_loss[loss=0.1943, simple_loss=0.2557, pruned_loss=0.0664, over 508377.77 frames. ], batch size: 31, lr: 3.89e-03, grad_scale: 16.0 2023-04-26 18:54:39,165 INFO [optim.py:369] (6/7) 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 18:54:40,569 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.3061, 1.7547, 2.2422, 2.7474, 2.1192, 1.6464, 1.4693, 2.0863], device='cuda:6'), covar=tensor([0.3920, 0.4343, 0.1945, 0.2838, 0.3693, 0.3599, 0.5236, 0.2864], device='cuda:6'), in_proj_covar=tensor([0.0275, 0.0252, 0.0216, 0.0320, 0.0213, 0.0227, 0.0237, 0.0188], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:6') 2023-04-26 18:55:02,021 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5094, 1.3082, 1.7379, 1.9596, 1.6650, 1.4956, 1.5852, 1.5729], device='cuda:6'), covar=tensor([0.8065, 1.0643, 1.1079, 1.1485, 0.9546, 1.2855, 1.3216, 1.1758], device='cuda:6'), in_proj_covar=tensor([0.0412, 0.0435, 0.0523, 0.0541, 0.0441, 0.0459, 0.0473, 0.0472], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-26 18:55:03,724 INFO [finetune.py:976] (6/7) Epoch 7, batch 200, loss[loss=0.227, simple_loss=0.2852, pruned_loss=0.08436, over 4820.00 frames. ], tot_loss[loss=0.1974, simple_loss=0.2578, pruned_loss=0.06849, over 609290.10 frames. ], batch size: 51, lr: 3.89e-03, grad_scale: 32.0 2023-04-26 18:55:32,449 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=34593.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:55:33,665 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=34595.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:55:46,311 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5553, 2.0758, 1.5574, 1.3131, 1.1942, 1.2329, 1.5826, 1.0991], device='cuda:6'), covar=tensor([0.1616, 0.1355, 0.1524, 0.1981, 0.2451, 0.1929, 0.1058, 0.2103], device='cuda:6'), in_proj_covar=tensor([0.0195, 0.0220, 0.0175, 0.0206, 0.0210, 0.0186, 0.0167, 0.0190], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-26 18:56:04,708 INFO [finetune.py:976] (6/7) Epoch 7, batch 250, loss[loss=0.2097, simple_loss=0.2637, pruned_loss=0.07786, over 4886.00 frames. ], tot_loss[loss=0.2036, simple_loss=0.265, pruned_loss=0.0711, over 685203.68 frames. ], batch size: 32, lr: 3.88e-03, grad_scale: 32.0 2023-04-26 18:56:09,043 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=34622.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:56:17,366 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=34626.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:56:19,022 INFO [optim.py:369] (6/7) 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] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=34643.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:56:51,405 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 7, batch 300, loss[loss=0.1784, simple_loss=0.2518, pruned_loss=0.05254, over 4909.00 frames. ], tot_loss[loss=0.206, simple_loss=0.2687, pruned_loss=0.07162, over 742808.66 frames. ], batch size: 43, lr: 3.88e-03, grad_scale: 32.0 2023-04-26 18:57:14,794 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5655, 1.3215, 1.7738, 1.7642, 1.3881, 1.2751, 1.4506, 1.0032], device='cuda:6'), covar=tensor([0.0716, 0.0992, 0.0597, 0.0823, 0.1012, 0.1355, 0.0904, 0.0893], device='cuda:6'), in_proj_covar=tensor([0.0067, 0.0075, 0.0073, 0.0068, 0.0078, 0.0095, 0.0081, 0.0077], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-26 18:57:35,527 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=34683.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 18:57:37,956 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=34687.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:57:42,154 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([3.4304, 3.3633, 2.5467, 4.0084, 3.4479, 3.4698, 1.6410, 3.4202], device='cuda:6'), covar=tensor([0.1801, 0.1310, 0.3883, 0.1871, 0.2851, 0.1938, 0.5382, 0.2386], device='cuda:6'), in_proj_covar=tensor([0.0244, 0.0218, 0.0253, 0.0310, 0.0301, 0.0253, 0.0274, 0.0275], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-26 18:57:51,401 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=34700.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:57:52,051 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0152, 1.4668, 1.8915, 2.3739, 1.9012, 1.4334, 1.2498, 1.7802], device='cuda:6'), covar=tensor([0.3757, 0.4383, 0.1882, 0.3150, 0.3365, 0.3228, 0.5489, 0.2997], device='cuda:6'), in_proj_covar=tensor([0.0277, 0.0254, 0.0218, 0.0323, 0.0214, 0.0228, 0.0239, 0.0190], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:6') 2023-04-26 18:57:52,625 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.0328, 1.5747, 1.3734, 1.7088, 1.4891, 1.9087, 1.3274, 3.3313], device='cuda:6'), covar=tensor([0.0660, 0.0736, 0.0750, 0.1187, 0.0666, 0.0490, 0.0746, 0.0172], device='cuda:6'), in_proj_covar=tensor([0.0039, 0.0039, 0.0041, 0.0044, 0.0040, 0.0040, 0.0039, 0.0060], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0014, 0.0013, 0.0013, 0.0013, 0.0017], device='cuda:6') 2023-04-26 18:57:53,211 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0178, 1.4306, 1.4029, 1.9370, 2.1766, 1.8532, 1.8122, 1.4638], device='cuda:6'), covar=tensor([0.1793, 0.1813, 0.1804, 0.1941, 0.1244, 0.1938, 0.1867, 0.1733], device='cuda:6'), in_proj_covar=tensor([0.0310, 0.0332, 0.0352, 0.0305, 0.0341, 0.0332, 0.0310, 0.0354], device='cuda:6'), out_proj_covar=tensor([6.6279e-05, 7.0770e-05, 7.6223e-05, 6.3548e-05, 7.1802e-05, 7.1618e-05, 6.6991e-05, 7.6313e-05], device='cuda:6') 2023-04-26 18:58:03,735 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=34710.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 18:58:12,744 INFO [finetune.py:976] (6/7) Epoch 7, batch 350, loss[loss=0.1969, simple_loss=0.2565, pruned_loss=0.06866, over 4295.00 frames. ], tot_loss[loss=0.208, simple_loss=0.2706, pruned_loss=0.07265, over 787618.75 frames. ], batch size: 65, lr: 3.88e-03, grad_scale: 32.0 2023-04-26 18:58:28,028 INFO [optim.py:369] (6/7) 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 18:58:35,695 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.47 vs. limit=2.0 2023-04-26 18:58:48,336 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.60 vs. limit=5.0 2023-04-26 18:58:50,334 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=34748.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:59:02,248 INFO [finetune.py:976] (6/7) Epoch 7, batch 400, loss[loss=0.1906, simple_loss=0.2533, pruned_loss=0.06396, over 4772.00 frames. ], tot_loss[loss=0.2079, simple_loss=0.2707, pruned_loss=0.07257, over 824283.70 frames. ], batch size: 28, lr: 3.88e-03, grad_scale: 32.0 2023-04-26 18:59:02,363 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.2745, 2.1706, 2.5489, 2.9540, 2.7490, 2.2784, 1.8477, 2.4137], device='cuda:6'), covar=tensor([0.0927, 0.0968, 0.0581, 0.0597, 0.0595, 0.0966, 0.0947, 0.0594], device='cuda:6'), in_proj_covar=tensor([0.0204, 0.0209, 0.0183, 0.0180, 0.0182, 0.0197, 0.0167, 0.0191], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-26 18:59:05,940 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=34771.0, num_to_drop=1, layers_to_drop={3} 2023-04-26 18:59:26,216 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=34801.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 18:59:36,186 INFO [finetune.py:976] (6/7) Epoch 7, batch 450, loss[loss=0.1632, simple_loss=0.2264, pruned_loss=0.04999, over 4219.00 frames. ], tot_loss[loss=0.2069, simple_loss=0.2694, pruned_loss=0.07224, over 852507.87 frames. ], batch size: 65, lr: 3.88e-03, grad_scale: 32.0 2023-04-26 18:59:36,318 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.4425, 1.2766, 1.7617, 1.7063, 1.3039, 1.1779, 1.4339, 0.9086], device='cuda:6'), covar=tensor([0.0801, 0.0935, 0.0547, 0.0889, 0.1066, 0.1460, 0.0757, 0.0944], device='cuda:6'), in_proj_covar=tensor([0.0068, 0.0075, 0.0073, 0.0068, 0.0079, 0.0096, 0.0081, 0.0077], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-26 18:59:45,537 INFO [optim.py:369] (6/7) 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 18:59:58,680 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 7, batch 500, loss[loss=0.2056, simple_loss=0.2647, pruned_loss=0.07322, over 4908.00 frames. ], tot_loss[loss=0.2039, simple_loss=0.2657, pruned_loss=0.07104, over 873787.30 frames. ], batch size: 36, lr: 3.88e-03, grad_scale: 32.0 2023-04-26 19:00:22,648 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.76 vs. limit=2.0 2023-04-26 19:00:42,231 INFO [finetune.py:976] (6/7) Epoch 7, batch 550, loss[loss=0.1719, simple_loss=0.2484, pruned_loss=0.04768, over 4771.00 frames. ], tot_loss[loss=0.2, simple_loss=0.262, pruned_loss=0.06901, over 892695.44 frames. ], batch size: 26, lr: 3.88e-03, grad_scale: 32.0 2023-04-26 19:00:51,085 INFO [optim.py:369] (6/7) 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 19:01:10,916 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=34949.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:01:32,183 INFO [finetune.py:976] (6/7) Epoch 7, batch 600, loss[loss=0.1837, simple_loss=0.2392, pruned_loss=0.06413, over 4696.00 frames. ], tot_loss[loss=0.2022, simple_loss=0.2639, pruned_loss=0.07025, over 908302.53 frames. ], batch size: 23, lr: 3.88e-03, grad_scale: 32.0 2023-04-26 19:01:32,923 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7491, 1.6426, 1.8790, 2.0894, 2.1145, 1.7170, 1.3185, 1.9172], device='cuda:6'), covar=tensor([0.0892, 0.1190, 0.0644, 0.0643, 0.0611, 0.0889, 0.0966, 0.0563], device='cuda:6'), in_proj_covar=tensor([0.0204, 0.0209, 0.0183, 0.0180, 0.0182, 0.0197, 0.0167, 0.0191], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-26 19:01:45,342 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=34978.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 19:01:54,261 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=34982.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:02:05,677 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9446, 2.5498, 1.9764, 1.7733, 1.4629, 1.4863, 2.1156, 1.4306], device='cuda:6'), covar=tensor([0.1782, 0.1517, 0.1645, 0.2010, 0.2783, 0.2200, 0.1138, 0.2178], device='cuda:6'), in_proj_covar=tensor([0.0195, 0.0219, 0.0174, 0.0206, 0.0210, 0.0186, 0.0166, 0.0191], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:6') 2023-04-26 19:02:23,577 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.54 vs. limit=2.0 2023-04-26 19:02:27,945 INFO [finetune.py:976] (6/7) Epoch 7, batch 650, loss[loss=0.2056, simple_loss=0.2766, pruned_loss=0.06729, over 4927.00 frames. ], tot_loss[loss=0.2045, simple_loss=0.2671, pruned_loss=0.07091, over 920937.94 frames. ], batch size: 38, lr: 3.88e-03, grad_scale: 32.0 2023-04-26 19:02:42,343 INFO [optim.py:369] (6/7) 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 19:03:19,050 INFO [finetune.py:976] (6/7) Epoch 7, batch 700, loss[loss=0.2196, simple_loss=0.2848, pruned_loss=0.07718, over 4885.00 frames. ], tot_loss[loss=0.2068, simple_loss=0.2697, pruned_loss=0.07194, over 929635.50 frames. ], batch size: 32, lr: 3.88e-03, grad_scale: 32.0 2023-04-26 19:03:19,119 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=35066.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 19:03:47,928 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=35092.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:03:58,684 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5595, 1.9625, 1.7213, 1.8827, 1.4956, 1.5426, 1.6330, 1.4090], device='cuda:6'), covar=tensor([0.2343, 0.1718, 0.1026, 0.1385, 0.3625, 0.1582, 0.2140, 0.2830], device='cuda:6'), in_proj_covar=tensor([0.0308, 0.0326, 0.0238, 0.0299, 0.0318, 0.0279, 0.0266, 0.0289], device='cuda:6'), out_proj_covar=tensor([1.2523e-04, 1.3196e-04, 9.6434e-05, 1.2005e-04, 1.3112e-04, 1.1273e-04, 1.0928e-04, 1.1652e-04], device='cuda:6') 2023-04-26 19:04:19,901 INFO [finetune.py:976] (6/7) Epoch 7, batch 750, loss[loss=0.2207, simple_loss=0.282, pruned_loss=0.07973, over 4823.00 frames. ], tot_loss[loss=0.208, simple_loss=0.2712, pruned_loss=0.07246, over 937107.68 frames. ], batch size: 47, lr: 3.88e-03, grad_scale: 32.0 2023-04-26 19:04:33,722 INFO [optim.py:369] (6/7) 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 19:05:05,770 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=35153.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:05:25,188 INFO [finetune.py:976] (6/7) Epoch 7, batch 800, loss[loss=0.2539, simple_loss=0.2973, pruned_loss=0.1052, over 4811.00 frames. ], tot_loss[loss=0.2082, simple_loss=0.2713, pruned_loss=0.0726, over 941887.28 frames. ], batch size: 25, lr: 3.88e-03, grad_scale: 32.0 2023-04-26 19:06:16,643 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.71 vs. limit=2.0 2023-04-26 19:06:20,320 INFO [finetune.py:976] (6/7) Epoch 7, batch 850, loss[loss=0.1896, simple_loss=0.2585, pruned_loss=0.06037, over 4741.00 frames. ], tot_loss[loss=0.2066, simple_loss=0.269, pruned_loss=0.07212, over 944248.95 frames. ], batch size: 54, lr: 3.88e-03, grad_scale: 32.0 2023-04-26 19:06:32,702 INFO [optim.py:369] (6/7) 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 19:06:35,187 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.0689, 2.1878, 1.7905, 1.8506, 2.1338, 1.5467, 2.6706, 1.4370], device='cuda:6'), covar=tensor([0.3543, 0.1589, 0.4144, 0.2595, 0.1740, 0.2684, 0.1292, 0.4565], device='cuda:6'), in_proj_covar=tensor([0.0346, 0.0352, 0.0435, 0.0364, 0.0390, 0.0381, 0.0385, 0.0417], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-26 19:06:58,280 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 7, batch 900, loss[loss=0.2111, simple_loss=0.2623, pruned_loss=0.08, over 4162.00 frames. ], tot_loss[loss=0.2036, simple_loss=0.2657, pruned_loss=0.07068, over 947190.23 frames. ], batch size: 65, lr: 3.88e-03, grad_scale: 32.0 2023-04-26 19:07:38,516 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=35278.0, num_to_drop=1, layers_to_drop={2} 2023-04-26 19:07:40,995 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=35282.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:08:02,085 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=35297.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:08:09,342 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.47 vs. limit=2.0 2023-04-26 19:08:12,143 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6716, 2.0351, 1.0165, 1.4681, 2.4473, 1.6412, 1.5530, 1.6662], device='cuda:6'), covar=tensor([0.0505, 0.0363, 0.0345, 0.0556, 0.0219, 0.0534, 0.0513, 0.0548], device='cuda:6'), in_proj_covar=tensor([0.0030, 0.0026, 0.0024, 0.0030, 0.0021, 0.0030, 0.0029, 0.0031], device='cuda:6'), out_proj_covar=tensor([0.0049, 0.0044, 0.0038, 0.0050, 0.0037, 0.0048, 0.0048, 0.0050], device='cuda:6') 2023-04-26 19:08:14,477 INFO [finetune.py:976] (6/7) Epoch 7, batch 950, loss[loss=0.1974, simple_loss=0.2636, pruned_loss=0.06554, over 4768.00 frames. ], tot_loss[loss=0.203, simple_loss=0.2649, pruned_loss=0.07053, over 948662.96 frames. ], batch size: 27, lr: 3.88e-03, grad_scale: 32.0 2023-04-26 19:08:14,657 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.80 vs. limit=2.0 2023-04-26 19:08:27,254 INFO [zipformer.py:1188] (6/7) 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] (6/7) 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 19:08:29,701 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=35330.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:08:50,431 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=35345.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 19:09:20,445 INFO [finetune.py:976] (6/7) Epoch 7, batch 1000, loss[loss=0.2139, simple_loss=0.2844, pruned_loss=0.07168, over 4912.00 frames. ], tot_loss[loss=0.2045, simple_loss=0.2667, pruned_loss=0.0712, over 950192.52 frames. ], batch size: 37, lr: 3.88e-03, grad_scale: 32.0 2023-04-26 19:09:20,555 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=35366.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 19:09:29,998 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=35374.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:09:34,273 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9419, 0.9608, 1.3533, 1.5152, 1.4475, 1.7274, 1.3790, 1.3918], device='cuda:6'), covar=tensor([0.5747, 0.8897, 0.7571, 0.7038, 0.8358, 1.2106, 0.8040, 0.8039], device='cuda:6'), in_proj_covar=tensor([0.0320, 0.0392, 0.0316, 0.0326, 0.0343, 0.0409, 0.0372, 0.0329], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-26 19:09:52,134 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5808, 2.3688, 1.1716, 1.3404, 2.5394, 1.5061, 1.4396, 1.6475], device='cuda:6'), covar=tensor([0.0698, 0.0317, 0.0352, 0.0660, 0.0227, 0.0719, 0.0723, 0.0681], device='cuda:6'), in_proj_covar=tensor([0.0031, 0.0026, 0.0024, 0.0030, 0.0021, 0.0030, 0.0029, 0.0031], device='cuda:6'), out_proj_covar=tensor([0.0049, 0.0044, 0.0038, 0.0050, 0.0037, 0.0048, 0.0048, 0.0050], device='cuda:6') 2023-04-26 19:10:06,695 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=35406.0, num_to_drop=1, layers_to_drop={2} 2023-04-26 19:10:24,157 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=35414.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 19:10:25,352 INFO [finetune.py:976] (6/7) Epoch 7, batch 1050, loss[loss=0.2293, simple_loss=0.2773, pruned_loss=0.09061, over 4711.00 frames. ], tot_loss[loss=0.2051, simple_loss=0.2679, pruned_loss=0.0712, over 951414.34 frames. ], batch size: 23, lr: 3.88e-03, grad_scale: 32.0 2023-04-26 19:10:39,105 INFO [optim.py:369] (6/7) 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 19:10:48,442 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=35435.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:11:06,932 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=35448.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:11:09,309 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([3.6028, 3.5826, 2.6134, 4.1457, 3.6284, 3.6263, 1.4956, 3.6142], device='cuda:6'), covar=tensor([0.1811, 0.1169, 0.3131, 0.1815, 0.2535, 0.1843, 0.5585, 0.2215], device='cuda:6'), in_proj_covar=tensor([0.0242, 0.0218, 0.0251, 0.0308, 0.0300, 0.0252, 0.0272, 0.0272], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-26 19:11:31,124 INFO [finetune.py:976] (6/7) Epoch 7, batch 1100, loss[loss=0.2542, simple_loss=0.3135, pruned_loss=0.09742, over 4925.00 frames. ], tot_loss[loss=0.2064, simple_loss=0.2696, pruned_loss=0.07162, over 953993.23 frames. ], batch size: 42, lr: 3.88e-03, grad_scale: 32.0 2023-04-26 19:11:56,826 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-04-26 19:12:09,917 INFO [finetune.py:976] (6/7) Epoch 7, batch 1150, loss[loss=0.1874, simple_loss=0.2403, pruned_loss=0.06729, over 4167.00 frames. ], tot_loss[loss=0.2062, simple_loss=0.2694, pruned_loss=0.07152, over 950405.94 frames. ], batch size: 18, lr: 3.88e-03, grad_scale: 32.0 2023-04-26 19:12:18,334 INFO [optim.py:369] (6/7) 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 19:12:19,086 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=35530.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:12:30,575 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6349, 1.8298, 0.7207, 1.3095, 2.0590, 1.5113, 1.4161, 1.4438], device='cuda:6'), covar=tensor([0.0580, 0.0410, 0.0443, 0.0649, 0.0277, 0.0607, 0.0577, 0.0685], device='cuda:6'), in_proj_covar=tensor([0.0031, 0.0026, 0.0024, 0.0031, 0.0021, 0.0030, 0.0029, 0.0031], device='cuda:6'), out_proj_covar=tensor([0.0049, 0.0045, 0.0038, 0.0050, 0.0038, 0.0048, 0.0048, 0.0050], device='cuda:6') 2023-04-26 19:12:30,621 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6784, 1.4051, 1.8166, 2.0742, 1.8149, 1.6580, 1.7313, 1.7677], device='cuda:6'), covar=tensor([0.8304, 1.1827, 1.2332, 1.2170, 1.0591, 1.4258, 1.4645, 1.2007], device='cuda:6'), in_proj_covar=tensor([0.0412, 0.0435, 0.0519, 0.0540, 0.0439, 0.0459, 0.0471, 0.0470], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-26 19:12:43,336 INFO [finetune.py:976] (6/7) Epoch 7, batch 1200, loss[loss=0.1607, simple_loss=0.2377, pruned_loss=0.04178, over 4794.00 frames. ], tot_loss[loss=0.204, simple_loss=0.2674, pruned_loss=0.07033, over 952506.76 frames. ], batch size: 51, lr: 3.88e-03, grad_scale: 32.0 2023-04-26 19:13:00,096 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=35591.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:13:17,331 INFO [finetune.py:976] (6/7) Epoch 7, batch 1250, loss[loss=0.1805, simple_loss=0.2426, pruned_loss=0.05926, over 4833.00 frames. ], tot_loss[loss=0.2028, simple_loss=0.2656, pruned_loss=0.07, over 954143.51 frames. ], batch size: 33, lr: 3.88e-03, grad_scale: 32.0 2023-04-26 19:13:26,243 INFO [optim.py:369] (6/7) 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 19:13:51,308 INFO [finetune.py:976] (6/7) Epoch 7, batch 1300, loss[loss=0.1347, simple_loss=0.208, pruned_loss=0.03073, over 4755.00 frames. ], tot_loss[loss=0.1998, simple_loss=0.2626, pruned_loss=0.06851, over 954691.36 frames. ], batch size: 28, lr: 3.88e-03, grad_scale: 32.0 2023-04-26 19:14:14,092 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=35701.0, num_to_drop=1, layers_to_drop={3} 2023-04-26 19:14:35,192 INFO [finetune.py:976] (6/7) Epoch 7, batch 1350, loss[loss=0.1877, simple_loss=0.2541, pruned_loss=0.06063, over 4829.00 frames. ], tot_loss[loss=0.1996, simple_loss=0.2627, pruned_loss=0.0682, over 955986.02 frames. ], batch size: 40, lr: 3.88e-03, grad_scale: 32.0 2023-04-26 19:14:54,293 INFO [optim.py:369] (6/7) 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 19:14:55,001 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=35730.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:15:17,740 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=35748.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:15:40,833 INFO [finetune.py:976] (6/7) Epoch 7, batch 1400, loss[loss=0.1856, simple_loss=0.2672, pruned_loss=0.05197, over 4760.00 frames. ], tot_loss[loss=0.203, simple_loss=0.2667, pruned_loss=0.06964, over 956538.38 frames. ], batch size: 27, lr: 3.88e-03, grad_scale: 32.0 2023-04-26 19:15:42,239 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-04-26 19:15:44,445 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=35771.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:16:06,192 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=35796.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:16:26,251 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=35811.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:16:30,144 INFO [finetune.py:976] (6/7) Epoch 7, batch 1450, loss[loss=0.1797, simple_loss=0.2532, pruned_loss=0.05314, over 4798.00 frames. ], tot_loss[loss=0.2041, simple_loss=0.2686, pruned_loss=0.06974, over 955949.26 frames. ], batch size: 51, lr: 3.88e-03, grad_scale: 32.0 2023-04-26 19:16:50,311 INFO [optim.py:369] (6/7) 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 19:16:51,200 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-04-26 19:16:58,477 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=35832.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:17:41,152 INFO [finetune.py:976] (6/7) Epoch 7, batch 1500, loss[loss=0.2606, simple_loss=0.3141, pruned_loss=0.1035, over 4884.00 frames. ], tot_loss[loss=0.2051, simple_loss=0.2695, pruned_loss=0.07036, over 957724.22 frames. ], batch size: 35, lr: 3.88e-03, grad_scale: 32.0 2023-04-26 19:17:46,499 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=35872.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:17:56,340 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=35886.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:18:20,581 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8083, 2.3877, 1.7934, 1.5707, 1.2731, 1.3273, 1.9864, 1.2683], device='cuda:6'), covar=tensor([0.1887, 0.1568, 0.1704, 0.2237, 0.2657, 0.2282, 0.1166, 0.2303], device='cuda:6'), in_proj_covar=tensor([0.0196, 0.0219, 0.0174, 0.0206, 0.0211, 0.0188, 0.0167, 0.0191], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-26 19:18:31,248 INFO [finetune.py:976] (6/7) Epoch 7, batch 1550, loss[loss=0.2183, simple_loss=0.271, pruned_loss=0.08282, over 4865.00 frames. ], tot_loss[loss=0.2042, simple_loss=0.2684, pruned_loss=0.07006, over 955069.40 frames. ], batch size: 31, lr: 3.88e-03, grad_scale: 32.0 2023-04-26 19:18:51,454 INFO [optim.py:369] (6/7) 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 19:19:38,740 INFO [finetune.py:976] (6/7) Epoch 7, batch 1600, loss[loss=0.2125, simple_loss=0.2692, pruned_loss=0.07789, over 4933.00 frames. ], tot_loss[loss=0.2037, simple_loss=0.2668, pruned_loss=0.07031, over 955279.27 frames. ], batch size: 38, lr: 3.88e-03, grad_scale: 32.0 2023-04-26 19:20:30,758 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=36001.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 19:20:45,229 INFO [finetune.py:976] (6/7) Epoch 7, batch 1650, loss[loss=0.2174, simple_loss=0.2676, pruned_loss=0.08356, over 4833.00 frames. ], tot_loss[loss=0.2021, simple_loss=0.2644, pruned_loss=0.06988, over 954504.00 frames. ], batch size: 40, lr: 3.88e-03, grad_scale: 32.0 2023-04-26 19:20:59,791 INFO [optim.py:369] (6/7) 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 19:21:00,970 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=36030.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:21:26,282 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=36049.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 19:21:48,193 INFO [finetune.py:976] (6/7) Epoch 7, batch 1700, loss[loss=0.2386, simple_loss=0.2936, pruned_loss=0.0918, over 4820.00 frames. ], tot_loss[loss=0.1997, simple_loss=0.2617, pruned_loss=0.0689, over 956071.59 frames. ], batch size: 39, lr: 3.88e-03, grad_scale: 32.0 2023-04-26 19:21:49,546 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.4588, 1.5728, 1.2886, 0.9150, 1.1223, 1.1192, 1.3013, 1.0391], device='cuda:6'), covar=tensor([0.1601, 0.1574, 0.1657, 0.2013, 0.2354, 0.2026, 0.1123, 0.2029], device='cuda:6'), in_proj_covar=tensor([0.0194, 0.0217, 0.0173, 0.0204, 0.0208, 0.0185, 0.0165, 0.0189], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:6') 2023-04-26 19:21:56,093 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 7, batch 1750, loss[loss=0.2114, simple_loss=0.2826, pruned_loss=0.07016, over 4928.00 frames. ], tot_loss[loss=0.2031, simple_loss=0.265, pruned_loss=0.07056, over 955545.90 frames. ], batch size: 38, lr: 3.88e-03, grad_scale: 32.0 2023-04-26 19:22:28,797 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=36127.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:22:29,946 INFO [optim.py:369] (6/7) 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 19:22:54,679 INFO [finetune.py:976] (6/7) Epoch 7, batch 1800, loss[loss=0.1833, simple_loss=0.2487, pruned_loss=0.05893, over 4762.00 frames. ], tot_loss[loss=0.2057, simple_loss=0.2691, pruned_loss=0.0711, over 955942.01 frames. ], batch size: 28, lr: 3.88e-03, grad_scale: 32.0 2023-04-26 19:22:55,369 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=36167.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:23:05,459 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=36181.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:23:08,949 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=36186.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:23:21,038 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.2888, 1.2945, 1.3911, 1.6318, 1.6432, 1.1935, 0.8809, 1.3631], device='cuda:6'), covar=tensor([0.1052, 0.1398, 0.0839, 0.0672, 0.0689, 0.1066, 0.1032, 0.0774], device='cuda:6'), in_proj_covar=tensor([0.0201, 0.0207, 0.0182, 0.0178, 0.0181, 0.0195, 0.0163, 0.0189], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-26 19:23:44,255 INFO [finetune.py:976] (6/7) Epoch 7, batch 1850, loss[loss=0.1543, simple_loss=0.2212, pruned_loss=0.04368, over 4706.00 frames. ], tot_loss[loss=0.2053, simple_loss=0.2693, pruned_loss=0.07064, over 954307.84 frames. ], batch size: 23, lr: 3.88e-03, grad_scale: 32.0 2023-04-26 19:24:03,618 INFO [optim.py:369] (6/7) 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] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=36234.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:24:19,072 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=36242.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:24:40,911 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.66 vs. limit=5.0 2023-04-26 19:24:43,638 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0934, 1.5047, 1.9372, 2.3350, 1.8712, 1.4610, 1.1548, 1.7055], device='cuda:6'), covar=tensor([0.4220, 0.4554, 0.2077, 0.3525, 0.3610, 0.3648, 0.5261, 0.3175], device='cuda:6'), in_proj_covar=tensor([0.0278, 0.0252, 0.0216, 0.0320, 0.0214, 0.0228, 0.0236, 0.0189], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:6') 2023-04-26 19:24:45,929 INFO [finetune.py:976] (6/7) Epoch 7, batch 1900, loss[loss=0.1911, simple_loss=0.2683, pruned_loss=0.05701, over 4921.00 frames. ], tot_loss[loss=0.2036, simple_loss=0.2685, pruned_loss=0.06932, over 955004.53 frames. ], batch size: 41, lr: 3.87e-03, grad_scale: 32.0 2023-04-26 19:25:19,585 INFO [finetune.py:976] (6/7) Epoch 7, batch 1950, loss[loss=0.1781, simple_loss=0.2403, pruned_loss=0.05794, over 4892.00 frames. ], tot_loss[loss=0.2023, simple_loss=0.2669, pruned_loss=0.06886, over 953241.15 frames. ], batch size: 32, lr: 3.87e-03, grad_scale: 32.0 2023-04-26 19:25:27,367 INFO [optim.py:369] (6/7) 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 19:25:32,104 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([0.6692, 1.7095, 1.7088, 1.3486, 1.8249, 1.4996, 2.3279, 1.4426], device='cuda:6'), covar=tensor([0.4198, 0.1694, 0.5279, 0.2953, 0.1625, 0.2456, 0.1509, 0.5140], device='cuda:6'), in_proj_covar=tensor([0.0346, 0.0352, 0.0434, 0.0364, 0.0388, 0.0385, 0.0387, 0.0420], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-26 19:25:38,017 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-04-26 19:26:09,641 INFO [finetune.py:976] (6/7) Epoch 7, batch 2000, loss[loss=0.2211, simple_loss=0.2703, pruned_loss=0.08594, over 4868.00 frames. ], tot_loss[loss=0.2012, simple_loss=0.2649, pruned_loss=0.06873, over 951480.97 frames. ], batch size: 31, lr: 3.87e-03, grad_scale: 32.0 2023-04-26 19:26:51,963 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0928, 1.6240, 5.2331, 4.8417, 4.5347, 4.9885, 4.6571, 4.5688], device='cuda:6'), covar=tensor([0.6723, 0.5905, 0.0856, 0.1544, 0.0993, 0.1147, 0.1189, 0.1344], device='cuda:6'), in_proj_covar=tensor([0.0308, 0.0302, 0.0407, 0.0411, 0.0349, 0.0406, 0.0315, 0.0364], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-26 19:27:00,246 INFO [finetune.py:976] (6/7) Epoch 7, batch 2050, loss[loss=0.2198, simple_loss=0.2742, pruned_loss=0.0827, over 4906.00 frames. ], tot_loss[loss=0.199, simple_loss=0.2624, pruned_loss=0.06783, over 951674.16 frames. ], batch size: 36, lr: 3.87e-03, grad_scale: 32.0 2023-04-26 19:27:05,282 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.1104, 1.8074, 2.0790, 2.3664, 2.3203, 2.0327, 1.7434, 2.0176], device='cuda:6'), covar=tensor([0.0714, 0.0909, 0.0465, 0.0478, 0.0585, 0.0734, 0.0809, 0.0560], device='cuda:6'), in_proj_covar=tensor([0.0200, 0.0206, 0.0181, 0.0177, 0.0180, 0.0194, 0.0162, 0.0188], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-26 19:27:07,134 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=36427.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:27:08,256 INFO [optim.py:369] (6/7) 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 19:27:43,885 INFO [finetune.py:976] (6/7) Epoch 7, batch 2100, loss[loss=0.1855, simple_loss=0.2426, pruned_loss=0.06423, over 4780.00 frames. ], tot_loss[loss=0.1982, simple_loss=0.2612, pruned_loss=0.06761, over 952351.02 frames. ], batch size: 26, lr: 3.87e-03, grad_scale: 32.0 2023-04-26 19:27:45,092 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=36467.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:27:49,909 INFO [zipformer.py:1188] (6/7) 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] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=36515.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:28:17,707 INFO [finetune.py:976] (6/7) Epoch 7, batch 2150, loss[loss=0.2356, simple_loss=0.3024, pruned_loss=0.08441, over 4917.00 frames. ], tot_loss[loss=0.2004, simple_loss=0.2641, pruned_loss=0.06835, over 953374.45 frames. ], batch size: 38, lr: 3.87e-03, grad_scale: 32.0 2023-04-26 19:28:25,935 INFO [optim.py:369] (6/7) 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 19:28:28,232 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-04-26 19:28:30,877 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=36537.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:28:33,386 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9456, 1.4052, 1.8551, 2.0497, 1.7924, 1.4037, 1.0847, 1.6400], device='cuda:6'), covar=tensor([0.4171, 0.4164, 0.1967, 0.3312, 0.3558, 0.3403, 0.5575, 0.3063], device='cuda:6'), in_proj_covar=tensor([0.0278, 0.0253, 0.0217, 0.0321, 0.0214, 0.0228, 0.0237, 0.0189], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:6') 2023-04-26 19:28:51,288 INFO [finetune.py:976] (6/7) Epoch 7, batch 2200, loss[loss=0.2155, simple_loss=0.2667, pruned_loss=0.08217, over 4829.00 frames. ], tot_loss[loss=0.2022, simple_loss=0.266, pruned_loss=0.06921, over 952299.15 frames. ], batch size: 25, lr: 3.87e-03, grad_scale: 64.0 2023-04-26 19:29:28,432 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=36605.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:29:30,361 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=36608.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:29:38,705 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=36612.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:29:41,091 INFO [finetune.py:976] (6/7) Epoch 7, batch 2250, loss[loss=0.205, simple_loss=0.2777, pruned_loss=0.06614, over 4899.00 frames. ], tot_loss[loss=0.2037, simple_loss=0.2679, pruned_loss=0.06975, over 952260.13 frames. ], batch size: 43, lr: 3.87e-03, grad_scale: 64.0 2023-04-26 19:29:42,334 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6036, 1.2990, 1.9269, 1.9105, 1.4838, 1.2829, 1.5854, 1.0050], device='cuda:6'), covar=tensor([0.0671, 0.1057, 0.0601, 0.0653, 0.0840, 0.1521, 0.0731, 0.1260], device='cuda:6'), in_proj_covar=tensor([0.0066, 0.0074, 0.0072, 0.0067, 0.0076, 0.0094, 0.0080, 0.0076], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-26 19:30:00,404 INFO [optim.py:369] (6/7) 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 19:30:40,443 INFO [finetune.py:976] (6/7) Epoch 7, batch 2300, loss[loss=0.2166, simple_loss=0.2835, pruned_loss=0.07485, over 4855.00 frames. ], tot_loss[loss=0.2043, simple_loss=0.2687, pruned_loss=0.06998, over 953437.49 frames. ], batch size: 31, lr: 3.87e-03, grad_scale: 64.0 2023-04-26 19:30:40,561 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=36666.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:30:42,404 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=36669.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:30:45,354 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=36673.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:31:02,258 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=36700.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 19:31:23,678 INFO [finetune.py:976] (6/7) Epoch 7, batch 2350, loss[loss=0.1951, simple_loss=0.2423, pruned_loss=0.07392, over 4816.00 frames. ], tot_loss[loss=0.2025, simple_loss=0.2663, pruned_loss=0.06931, over 954455.55 frames. ], batch size: 30, lr: 3.87e-03, grad_scale: 64.0 2023-04-26 19:31:27,174 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-04-26 19:31:38,222 INFO [optim.py:369] (6/7) 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 19:32:21,179 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=36761.0, num_to_drop=1, layers_to_drop={3} 2023-04-26 19:32:30,355 INFO [finetune.py:976] (6/7) Epoch 7, batch 2400, loss[loss=0.1784, simple_loss=0.2459, pruned_loss=0.05539, over 4937.00 frames. ], tot_loss[loss=0.2013, simple_loss=0.2644, pruned_loss=0.06907, over 956307.84 frames. ], batch size: 33, lr: 3.87e-03, grad_scale: 64.0 2023-04-26 19:32:53,303 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0148, 1.8425, 2.2152, 2.3709, 2.0835, 1.9349, 2.0407, 2.1441], device='cuda:6'), covar=tensor([0.8773, 1.2467, 1.4010, 1.2855, 1.0805, 1.6328, 1.5487, 1.2701], device='cuda:6'), in_proj_covar=tensor([0.0415, 0.0437, 0.0521, 0.0543, 0.0443, 0.0463, 0.0474, 0.0473], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-26 19:32:56,367 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=36805.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 19:33:03,828 INFO [finetune.py:976] (6/7) Epoch 7, batch 2450, loss[loss=0.1892, simple_loss=0.2574, pruned_loss=0.06047, over 4752.00 frames. ], tot_loss[loss=0.1986, simple_loss=0.261, pruned_loss=0.06814, over 956371.92 frames. ], batch size: 27, lr: 3.87e-03, grad_scale: 64.0 2023-04-26 19:33:12,661 INFO [optim.py:369] (6/7) 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 19:33:18,545 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=36837.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:33:37,048 INFO [finetune.py:976] (6/7) Epoch 7, batch 2500, loss[loss=0.2315, simple_loss=0.293, pruned_loss=0.08504, over 4808.00 frames. ], tot_loss[loss=0.2013, simple_loss=0.2635, pruned_loss=0.06954, over 954073.67 frames. ], batch size: 41, lr: 3.87e-03, grad_scale: 64.0 2023-04-26 19:33:37,689 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=36866.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 19:33:51,064 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=36885.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:34:07,402 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=36911.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:34:10,363 INFO [finetune.py:976] (6/7) Epoch 7, batch 2550, loss[loss=0.1831, simple_loss=0.2623, pruned_loss=0.05195, over 4818.00 frames. ], tot_loss[loss=0.2031, simple_loss=0.2665, pruned_loss=0.06991, over 954325.78 frames. ], batch size: 38, lr: 3.87e-03, grad_scale: 64.0 2023-04-26 19:34:20,134 INFO [optim.py:369] (6/7) 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 19:34:50,682 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=36958.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:34:52,520 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=36961.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:34:54,347 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 7, batch 2600, loss[loss=0.2373, simple_loss=0.2911, pruned_loss=0.09178, over 4898.00 frames. ], tot_loss[loss=0.2046, simple_loss=0.2684, pruned_loss=0.07041, over 955290.40 frames. ], batch size: 36, lr: 3.87e-03, grad_scale: 64.0 2023-04-26 19:35:00,463 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.47 vs. limit=2.0 2023-04-26 19:35:00,981 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=36968.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:35:04,510 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=36972.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:35:48,149 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=37004.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:36:01,124 INFO [finetune.py:976] (6/7) Epoch 7, batch 2650, loss[loss=0.2384, simple_loss=0.2902, pruned_loss=0.09333, over 4790.00 frames. ], tot_loss[loss=0.2062, simple_loss=0.2699, pruned_loss=0.07127, over 955522.93 frames. ], batch size: 25, lr: 3.87e-03, grad_scale: 64.0 2023-04-26 19:36:03,054 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=37019.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:36:09,956 INFO [optim.py:369] (6/7) 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] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=37056.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 19:36:34,318 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=37065.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:36:34,795 INFO [finetune.py:976] (6/7) Epoch 7, batch 2700, loss[loss=0.1843, simple_loss=0.2449, pruned_loss=0.0618, over 4321.00 frames. ], tot_loss[loss=0.2046, simple_loss=0.2689, pruned_loss=0.07019, over 956534.64 frames. ], batch size: 18, lr: 3.87e-03, grad_scale: 64.0 2023-04-26 19:37:19,208 INFO [finetune.py:976] (6/7) Epoch 7, batch 2750, loss[loss=0.2159, simple_loss=0.2647, pruned_loss=0.08352, over 4812.00 frames. ], tot_loss[loss=0.2022, simple_loss=0.2659, pruned_loss=0.06924, over 955932.39 frames. ], batch size: 40, lr: 3.87e-03, grad_scale: 64.0 2023-04-26 19:37:32,275 INFO [optim.py:369] (6/7) 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 19:37:58,365 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-04-26 19:38:17,424 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=37161.0, num_to_drop=1, layers_to_drop={2} 2023-04-26 19:38:20,450 INFO [finetune.py:976] (6/7) Epoch 7, batch 2800, loss[loss=0.1651, simple_loss=0.2355, pruned_loss=0.04741, over 4794.00 frames. ], tot_loss[loss=0.1995, simple_loss=0.2624, pruned_loss=0.06829, over 953600.14 frames. ], batch size: 29, lr: 3.87e-03, grad_scale: 64.0 2023-04-26 19:38:30,462 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0106, 1.8667, 2.2268, 2.4966, 2.5009, 2.1005, 1.6240, 2.1078], device='cuda:6'), covar=tensor([0.0976, 0.1177, 0.0646, 0.0653, 0.0579, 0.0928, 0.0940, 0.0658], device='cuda:6'), in_proj_covar=tensor([0.0202, 0.0208, 0.0183, 0.0179, 0.0181, 0.0195, 0.0165, 0.0191], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-26 19:38:51,092 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0671, 2.4758, 1.2004, 1.4924, 1.8832, 1.3619, 2.8243, 1.6258], device='cuda:6'), covar=tensor([0.0576, 0.0750, 0.0703, 0.0992, 0.0422, 0.0832, 0.0224, 0.0569], device='cuda:6'), in_proj_covar=tensor([0.0053, 0.0068, 0.0051, 0.0048, 0.0053, 0.0053, 0.0080, 0.0052], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0011, 0.0008], device='cuda:6') 2023-04-26 19:39:00,022 INFO [finetune.py:976] (6/7) Epoch 7, batch 2850, loss[loss=0.1601, simple_loss=0.2364, pruned_loss=0.04191, over 4831.00 frames. ], tot_loss[loss=0.1984, simple_loss=0.2614, pruned_loss=0.06771, over 954261.44 frames. ], batch size: 25, lr: 3.87e-03, grad_scale: 32.0 2023-04-26 19:39:08,538 INFO [optim.py:369] (6/7) 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 19:39:23,793 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9587, 1.9152, 2.2475, 2.5539, 2.5388, 2.0683, 1.6514, 2.0159], device='cuda:6'), covar=tensor([0.1053, 0.1123, 0.0653, 0.0612, 0.0654, 0.1016, 0.1023, 0.0718], device='cuda:6'), in_proj_covar=tensor([0.0202, 0.0207, 0.0183, 0.0179, 0.0181, 0.0195, 0.0164, 0.0190], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-26 19:39:30,801 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=37261.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:39:32,612 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=37264.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:39:33,741 INFO [finetune.py:976] (6/7) Epoch 7, batch 2900, loss[loss=0.272, simple_loss=0.333, pruned_loss=0.1055, over 4728.00 frames. ], tot_loss[loss=0.2013, simple_loss=0.2641, pruned_loss=0.06919, over 951659.27 frames. ], batch size: 59, lr: 3.87e-03, grad_scale: 32.0 2023-04-26 19:39:34,409 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=37267.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:39:35,008 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=37268.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:39:35,025 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=37268.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:40:03,050 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=37309.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:40:10,642 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=37312.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:40:11,880 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 7, batch 2950, loss[loss=0.1931, simple_loss=0.2555, pruned_loss=0.06531, over 4711.00 frames. ], tot_loss[loss=0.2022, simple_loss=0.2657, pruned_loss=0.06936, over 951391.29 frames. ], batch size: 23, lr: 3.87e-03, grad_scale: 32.0 2023-04-26 19:40:13,133 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=37316.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:40:32,643 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=37329.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:40:33,129 INFO [optim.py:369] (6/7) 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 19:41:05,678 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=37356.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 19:41:08,609 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=37360.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:41:17,428 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=37364.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:41:18,547 INFO [finetune.py:976] (6/7) Epoch 7, batch 3000, loss[loss=0.2046, simple_loss=0.2608, pruned_loss=0.07423, over 4722.00 frames. ], tot_loss[loss=0.2044, simple_loss=0.2675, pruned_loss=0.07067, over 949578.63 frames. ], batch size: 54, lr: 3.87e-03, grad_scale: 32.0 2023-04-26 19:41:18,547 INFO [finetune.py:1001] (6/7) Computing validation loss 2023-04-26 19:41:40,583 INFO [finetune.py:1010] (6/7) Epoch 7, validation: loss=0.1559, simple_loss=0.2289, pruned_loss=0.04148, over 2265189.00 frames. 2023-04-26 19:41:40,599 INFO [finetune.py:1011] (6/7) Maximum memory allocated so far is 6326MB 2023-04-26 19:42:13,773 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.3570, 1.5977, 1.5276, 1.7399, 1.5273, 1.9077, 1.3205, 3.7348], device='cuda:6'), covar=tensor([0.0596, 0.0781, 0.0769, 0.1194, 0.0671, 0.0527, 0.0759, 0.0138], device='cuda:6'), in_proj_covar=tensor([0.0039, 0.0039, 0.0041, 0.0044, 0.0040, 0.0040, 0.0039, 0.0060], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0014, 0.0013, 0.0013, 0.0013, 0.0017], device='cuda:6') 2023-04-26 19:42:15,402 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=37404.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 19:42:23,695 INFO [finetune.py:976] (6/7) Epoch 7, batch 3050, loss[loss=0.1841, simple_loss=0.2438, pruned_loss=0.06219, over 4813.00 frames. ], tot_loss[loss=0.2042, simple_loss=0.2674, pruned_loss=0.07045, over 949656.04 frames. ], batch size: 33, lr: 3.87e-03, grad_scale: 32.0 2023-04-26 19:42:30,744 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=37425.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:42:34,151 INFO [optim.py:369] (6/7) 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 19:42:43,917 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([0.4413, 1.3746, 1.3669, 1.0921, 1.3554, 1.1415, 1.7551, 1.1553], device='cuda:6'), covar=tensor([0.3361, 0.1614, 0.5182, 0.2413, 0.1583, 0.2040, 0.1625, 0.5028], device='cuda:6'), in_proj_covar=tensor([0.0347, 0.0354, 0.0436, 0.0365, 0.0391, 0.0385, 0.0385, 0.0421], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-26 19:42:54,546 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=37461.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 19:42:57,501 INFO [finetune.py:976] (6/7) Epoch 7, batch 3100, loss[loss=0.1476, simple_loss=0.2215, pruned_loss=0.03683, over 4765.00 frames. ], tot_loss[loss=0.202, simple_loss=0.2653, pruned_loss=0.06935, over 950099.04 frames. ], batch size: 28, lr: 3.87e-03, grad_scale: 32.0 2023-04-26 19:42:57,656 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6831, 2.2567, 1.6604, 1.4031, 1.3122, 1.3422, 1.7063, 1.2774], device='cuda:6'), covar=tensor([0.1826, 0.1506, 0.1626, 0.2168, 0.2605, 0.2175, 0.1238, 0.2237], device='cuda:6'), in_proj_covar=tensor([0.0193, 0.0218, 0.0173, 0.0205, 0.0208, 0.0186, 0.0165, 0.0190], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:6') 2023-04-26 19:43:26,082 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=37509.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 19:43:31,232 INFO [finetune.py:976] (6/7) Epoch 7, batch 3150, loss[loss=0.192, simple_loss=0.2452, pruned_loss=0.0694, over 4824.00 frames. ], tot_loss[loss=0.2004, simple_loss=0.2633, pruned_loss=0.06877, over 951930.99 frames. ], batch size: 30, lr: 3.87e-03, grad_scale: 32.0 2023-04-26 19:43:32,052 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=3.86 vs. limit=5.0 2023-04-26 19:43:52,226 INFO [optim.py:369] (6/7) 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 19:44:36,863 INFO [finetune.py:976] (6/7) Epoch 7, batch 3200, loss[loss=0.2347, simple_loss=0.2855, pruned_loss=0.09193, over 4750.00 frames. ], tot_loss[loss=0.1976, simple_loss=0.26, pruned_loss=0.06761, over 952058.32 frames. ], batch size: 54, lr: 3.87e-03, grad_scale: 32.0 2023-04-26 19:44:37,565 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=37567.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:45:31,842 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6056, 3.1940, 1.3127, 2.0616, 2.0316, 2.5385, 2.1280, 1.3553], device='cuda:6'), covar=tensor([0.1071, 0.0746, 0.1338, 0.1041, 0.0898, 0.0788, 0.1146, 0.1859], device='cuda:6'), in_proj_covar=tensor([0.0118, 0.0255, 0.0142, 0.0125, 0.0135, 0.0155, 0.0121, 0.0123], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-26 19:45:41,724 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.94 vs. limit=5.0 2023-04-26 19:45:42,716 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=37614.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:45:43,285 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=37615.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:45:43,828 INFO [finetune.py:976] (6/7) Epoch 7, batch 3250, loss[loss=0.1672, simple_loss=0.2335, pruned_loss=0.05049, over 4789.00 frames. ], tot_loss[loss=0.1985, simple_loss=0.2612, pruned_loss=0.06789, over 952064.70 frames. ], batch size: 26, lr: 3.87e-03, grad_scale: 32.0 2023-04-26 19:45:55,389 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=37624.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:46:03,767 INFO [optim.py:369] (6/7) 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 19:46:41,416 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=37660.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:46:42,586 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=37662.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:46:45,456 INFO [finetune.py:976] (6/7) Epoch 7, batch 3300, loss[loss=0.1994, simple_loss=0.2685, pruned_loss=0.06514, over 4892.00 frames. ], tot_loss[loss=0.2033, simple_loss=0.2664, pruned_loss=0.0701, over 954070.65 frames. ], batch size: 35, lr: 3.87e-03, grad_scale: 32.0 2023-04-26 19:47:14,163 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([3.6994, 3.6967, 2.8609, 4.3193, 3.6948, 3.7467, 1.7648, 3.6609], device='cuda:6'), covar=tensor([0.1554, 0.1149, 0.3215, 0.1206, 0.2626, 0.1609, 0.5079, 0.2139], device='cuda:6'), in_proj_covar=tensor([0.0242, 0.0218, 0.0251, 0.0308, 0.0302, 0.0251, 0.0272, 0.0272], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-26 19:47:25,802 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8028, 2.5098, 1.7042, 1.5318, 1.3491, 1.3706, 1.8231, 1.2586], device='cuda:6'), covar=tensor([0.1828, 0.1401, 0.1665, 0.2133, 0.2623, 0.2186, 0.1304, 0.2179], device='cuda:6'), in_proj_covar=tensor([0.0193, 0.0218, 0.0173, 0.0206, 0.0208, 0.0186, 0.0165, 0.0189], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:6') 2023-04-26 19:47:28,167 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.3604, 1.6553, 2.1430, 2.9482, 2.1700, 1.7300, 1.5950, 2.0358], device='cuda:6'), covar=tensor([0.4144, 0.4295, 0.2042, 0.3050, 0.3723, 0.3288, 0.5026, 0.3213], device='cuda:6'), in_proj_covar=tensor([0.0280, 0.0255, 0.0218, 0.0323, 0.0216, 0.0230, 0.0238, 0.0190], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-26 19:47:48,146 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=37708.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:47:57,366 INFO [finetune.py:976] (6/7) Epoch 7, batch 3350, loss[loss=0.2073, simple_loss=0.2669, pruned_loss=0.07384, over 4761.00 frames. ], tot_loss[loss=0.2041, simple_loss=0.268, pruned_loss=0.07014, over 955639.99 frames. ], batch size: 27, lr: 3.87e-03, grad_scale: 32.0 2023-04-26 19:48:00,367 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=37720.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:48:12,882 INFO [optim.py:369] (6/7) 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 19:48:15,349 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0545, 3.0792, 1.8553, 2.1707, 1.4989, 1.5010, 2.0925, 1.4392], device='cuda:6'), covar=tensor([0.2058, 0.1581, 0.2013, 0.2103, 0.3127, 0.2668, 0.1417, 0.2455], device='cuda:6'), in_proj_covar=tensor([0.0194, 0.0219, 0.0173, 0.0206, 0.0209, 0.0186, 0.0165, 0.0190], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:6') 2023-04-26 19:48:21,804 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5982, 1.8049, 0.9258, 1.3303, 2.1246, 1.4995, 1.4168, 1.4644], device='cuda:6'), covar=tensor([0.0527, 0.0394, 0.0371, 0.0584, 0.0267, 0.0534, 0.0522, 0.0614], device='cuda:6'), in_proj_covar=tensor([0.0030, 0.0026, 0.0024, 0.0030, 0.0021, 0.0030, 0.0029, 0.0030], device='cuda:6'), out_proj_covar=tensor([0.0049, 0.0044, 0.0038, 0.0050, 0.0037, 0.0048, 0.0048, 0.0050], device='cuda:6') 2023-04-26 19:48:26,095 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.73 vs. limit=2.0 2023-04-26 19:48:36,823 INFO [finetune.py:976] (6/7) Epoch 7, batch 3400, loss[loss=0.2533, simple_loss=0.3197, pruned_loss=0.09348, over 4923.00 frames. ], tot_loss[loss=0.2057, simple_loss=0.2696, pruned_loss=0.07086, over 956307.73 frames. ], batch size: 38, lr: 3.87e-03, grad_scale: 32.0 2023-04-26 19:49:13,969 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.8460, 2.8199, 2.1709, 3.2649, 2.8748, 2.7708, 1.1719, 2.7338], device='cuda:6'), covar=tensor([0.1919, 0.1677, 0.3245, 0.2626, 0.3019, 0.2366, 0.5718, 0.2876], device='cuda:6'), in_proj_covar=tensor([0.0242, 0.0217, 0.0250, 0.0307, 0.0301, 0.0251, 0.0272, 0.0271], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-26 19:49:36,099 INFO [finetune.py:976] (6/7) Epoch 7, batch 3450, loss[loss=0.2113, simple_loss=0.2714, pruned_loss=0.07561, over 4906.00 frames. ], tot_loss[loss=0.2048, simple_loss=0.2689, pruned_loss=0.07038, over 955838.21 frames. ], batch size: 37, lr: 3.87e-03, grad_scale: 32.0 2023-04-26 19:49:41,903 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7862, 4.0593, 0.9542, 2.1339, 2.2845, 2.8986, 2.3336, 1.1275], device='cuda:6'), covar=tensor([0.1290, 0.0968, 0.2075, 0.1399, 0.0999, 0.0940, 0.1501, 0.2111], device='cuda:6'), in_proj_covar=tensor([0.0119, 0.0255, 0.0143, 0.0125, 0.0136, 0.0155, 0.0121, 0.0124], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-26 19:49:55,704 INFO [optim.py:369] (6/7) 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 19:50:37,108 INFO [finetune.py:976] (6/7) Epoch 7, batch 3500, loss[loss=0.238, simple_loss=0.2892, pruned_loss=0.09339, over 4819.00 frames. ], tot_loss[loss=0.2025, simple_loss=0.2665, pruned_loss=0.06926, over 956367.24 frames. ], batch size: 39, lr: 3.86e-03, grad_scale: 32.0 2023-04-26 19:51:17,011 INFO [finetune.py:976] (6/7) Epoch 7, batch 3550, loss[loss=0.1741, simple_loss=0.2448, pruned_loss=0.05175, over 4760.00 frames. ], tot_loss[loss=0.2004, simple_loss=0.2638, pruned_loss=0.06852, over 955314.94 frames. ], batch size: 27, lr: 3.86e-03, grad_scale: 32.0 2023-04-26 19:51:27,523 INFO [zipformer.py:1188] (6/7) 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] (6/7) 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 19:52:01,352 INFO [finetune.py:976] (6/7) Epoch 7, batch 3600, loss[loss=0.2592, simple_loss=0.3225, pruned_loss=0.09798, over 4733.00 frames. ], tot_loss[loss=0.1985, simple_loss=0.2613, pruned_loss=0.06787, over 954541.52 frames. ], batch size: 59, lr: 3.86e-03, grad_scale: 32.0 2023-04-26 19:52:05,030 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=37972.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:52:09,393 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=37979.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:52:32,117 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.14 vs. limit=5.0 2023-04-26 19:52:35,340 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([3.4857, 3.4545, 2.7977, 4.1281, 3.4705, 3.5165, 1.5373, 3.5081], device='cuda:6'), covar=tensor([0.1898, 0.1241, 0.3711, 0.1616, 0.3098, 0.2019, 0.5981, 0.2504], device='cuda:6'), in_proj_covar=tensor([0.0245, 0.0219, 0.0252, 0.0310, 0.0304, 0.0254, 0.0275, 0.0274], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-26 19:52:35,874 INFO [finetune.py:976] (6/7) Epoch 7, batch 3650, loss[loss=0.2543, simple_loss=0.3212, pruned_loss=0.09373, over 4722.00 frames. ], tot_loss[loss=0.2017, simple_loss=0.2647, pruned_loss=0.06935, over 956287.89 frames. ], batch size: 54, lr: 3.86e-03, grad_scale: 32.0 2023-04-26 19:52:38,368 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=38020.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:52:42,672 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5720, 1.5035, 1.8488, 1.9453, 1.4987, 1.2572, 1.6068, 1.0013], device='cuda:6'), covar=tensor([0.0795, 0.0993, 0.0544, 0.0883, 0.1014, 0.1516, 0.0949, 0.1068], device='cuda:6'), in_proj_covar=tensor([0.0065, 0.0074, 0.0072, 0.0067, 0.0076, 0.0095, 0.0080, 0.0076], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-26 19:52:44,353 INFO [optim.py:369] (6/7) 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 19:52:51,038 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=38040.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:53:08,982 INFO [finetune.py:976] (6/7) Epoch 7, batch 3700, loss[loss=0.2551, simple_loss=0.3048, pruned_loss=0.1027, over 4888.00 frames. ], tot_loss[loss=0.2038, simple_loss=0.2675, pruned_loss=0.07001, over 956369.16 frames. ], batch size: 32, lr: 3.86e-03, grad_scale: 32.0 2023-04-26 19:53:10,247 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=38068.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:53:13,407 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5094, 0.9344, 1.4544, 1.9149, 1.6216, 1.4496, 1.5159, 1.5644], device='cuda:6'), covar=tensor([0.7228, 0.9565, 0.9723, 0.9818, 0.8391, 1.1485, 1.1319, 1.0048], device='cuda:6'), in_proj_covar=tensor([0.0412, 0.0432, 0.0518, 0.0540, 0.0442, 0.0463, 0.0473, 0.0470], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-26 19:53:30,014 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6664, 1.0892, 1.5376, 2.0434, 1.7578, 1.5914, 1.6461, 1.6965], device='cuda:6'), covar=tensor([0.8109, 1.0440, 1.0265, 1.1617, 0.9693, 1.1974, 1.2566, 0.9936], device='cuda:6'), in_proj_covar=tensor([0.0413, 0.0433, 0.0519, 0.0541, 0.0443, 0.0463, 0.0474, 0.0471], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-26 19:53:41,670 INFO [finetune.py:976] (6/7) Epoch 7, batch 3750, loss[loss=0.2511, simple_loss=0.319, pruned_loss=0.09161, over 4903.00 frames. ], tot_loss[loss=0.2059, simple_loss=0.2696, pruned_loss=0.07112, over 956711.38 frames. ], batch size: 36, lr: 3.86e-03, grad_scale: 32.0 2023-04-26 19:53:50,669 INFO [optim.py:369] (6/7) 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 19:53:56,370 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=3.83 vs. limit=5.0 2023-04-26 19:54:31,493 INFO [finetune.py:976] (6/7) Epoch 7, batch 3800, loss[loss=0.1797, simple_loss=0.2511, pruned_loss=0.05412, over 4813.00 frames. ], tot_loss[loss=0.2064, simple_loss=0.2703, pruned_loss=0.07127, over 956358.74 frames. ], batch size: 30, lr: 3.86e-03, grad_scale: 32.0 2023-04-26 19:54:54,438 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9223, 2.6927, 2.2122, 2.4390, 1.9659, 2.0627, 2.1676, 1.7354], device='cuda:6'), covar=tensor([0.2503, 0.1611, 0.0944, 0.1352, 0.3524, 0.1578, 0.2345, 0.3152], device='cuda:6'), in_proj_covar=tensor([0.0304, 0.0323, 0.0235, 0.0295, 0.0316, 0.0275, 0.0264, 0.0288], device='cuda:6'), out_proj_covar=tensor([1.2352e-04, 1.3080e-04, 9.5151e-05, 1.1843e-04, 1.3021e-04, 1.1124e-04, 1.0819e-04, 1.1593e-04], device='cuda:6') 2023-04-26 19:55:03,093 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6340, 1.3307, 1.7449, 2.0716, 1.7658, 1.5666, 1.6776, 1.7010], device='cuda:6'), covar=tensor([0.7479, 1.0496, 1.0504, 1.0307, 0.8940, 1.3031, 1.3070, 1.1020], device='cuda:6'), in_proj_covar=tensor([0.0411, 0.0431, 0.0517, 0.0538, 0.0442, 0.0462, 0.0472, 0.0470], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-26 19:55:36,932 INFO [finetune.py:976] (6/7) Epoch 7, batch 3850, loss[loss=0.21, simple_loss=0.2732, pruned_loss=0.07346, over 4924.00 frames. ], tot_loss[loss=0.2027, simple_loss=0.2668, pruned_loss=0.06927, over 954717.21 frames. ], batch size: 37, lr: 3.86e-03, grad_scale: 32.0 2023-04-26 19:55:56,086 INFO [optim.py:369] (6/7) 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 19:56:29,008 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-04-26 19:56:41,863 INFO [finetune.py:976] (6/7) Epoch 7, batch 3900, loss[loss=0.1349, simple_loss=0.2076, pruned_loss=0.03111, over 4782.00 frames. ], tot_loss[loss=0.2001, simple_loss=0.2638, pruned_loss=0.0682, over 954551.22 frames. ], batch size: 29, lr: 3.86e-03, grad_scale: 32.0 2023-04-26 19:57:47,849 INFO [finetune.py:976] (6/7) Epoch 7, batch 3950, loss[loss=0.1974, simple_loss=0.2556, pruned_loss=0.06957, over 4850.00 frames. ], tot_loss[loss=0.1961, simple_loss=0.2595, pruned_loss=0.06631, over 955425.24 frames. ], batch size: 49, lr: 3.86e-03, grad_scale: 32.0 2023-04-26 19:58:09,819 INFO [optim.py:369] (6/7) 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 19:58:18,664 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=38335.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:58:32,608 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 2023-04-26 19:58:49,624 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=38360.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 19:58:53,667 INFO [finetune.py:976] (6/7) Epoch 7, batch 4000, loss[loss=0.2617, simple_loss=0.3121, pruned_loss=0.1057, over 4167.00 frames. ], tot_loss[loss=0.1964, simple_loss=0.2591, pruned_loss=0.06688, over 954428.50 frames. ], batch size: 65, lr: 3.86e-03, grad_scale: 32.0 2023-04-26 19:59:13,413 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8807, 1.5492, 1.3692, 1.6082, 2.1821, 1.6626, 1.4046, 1.3672], device='cuda:6'), covar=tensor([0.1642, 0.1516, 0.2402, 0.1280, 0.0848, 0.1660, 0.2101, 0.2152], device='cuda:6'), in_proj_covar=tensor([0.0308, 0.0328, 0.0356, 0.0303, 0.0341, 0.0327, 0.0308, 0.0354], device='cuda:6'), out_proj_covar=tensor([6.5598e-05, 6.9790e-05, 7.7144e-05, 6.2991e-05, 7.1724e-05, 7.0413e-05, 6.6619e-05, 7.6166e-05], device='cuda:6') 2023-04-26 19:59:34,617 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.64 vs. limit=2.0 2023-04-26 20:00:00,680 INFO [finetune.py:976] (6/7) Epoch 7, batch 4050, loss[loss=0.19, simple_loss=0.2704, pruned_loss=0.05475, over 4909.00 frames. ], tot_loss[loss=0.1996, simple_loss=0.2628, pruned_loss=0.06817, over 955725.11 frames. ], batch size: 37, lr: 3.86e-03, grad_scale: 32.0 2023-04-26 20:00:10,091 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=38421.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:00:10,150 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-04-26 20:00:21,445 INFO [optim.py:369] (6/7) 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] (6/7) Epoch 7, batch 4100, loss[loss=0.1866, simple_loss=0.26, pruned_loss=0.05657, over 4848.00 frames. ], tot_loss[loss=0.201, simple_loss=0.2652, pruned_loss=0.06843, over 956638.99 frames. ], batch size: 44, lr: 3.86e-03, grad_scale: 32.0 2023-04-26 20:01:29,886 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0846, 2.6989, 1.6567, 2.0328, 2.6679, 1.9996, 1.9479, 2.1368], device='cuda:6'), covar=tensor([0.0496, 0.0326, 0.0293, 0.0528, 0.0228, 0.0529, 0.0516, 0.0571], device='cuda:6'), in_proj_covar=tensor([0.0030, 0.0026, 0.0024, 0.0030, 0.0021, 0.0030, 0.0029, 0.0030], device='cuda:6'), out_proj_covar=tensor([0.0049, 0.0044, 0.0038, 0.0050, 0.0038, 0.0048, 0.0048, 0.0050], device='cuda:6') 2023-04-26 20:01:32,868 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6482, 1.9664, 1.1381, 1.4839, 2.1505, 1.5328, 1.4849, 1.5508], device='cuda:6'), covar=tensor([0.0522, 0.0373, 0.0351, 0.0573, 0.0282, 0.0564, 0.0515, 0.0621], device='cuda:6'), in_proj_covar=tensor([0.0030, 0.0026, 0.0024, 0.0030, 0.0021, 0.0030, 0.0029, 0.0030], device='cuda:6'), out_proj_covar=tensor([0.0049, 0.0044, 0.0038, 0.0050, 0.0038, 0.0048, 0.0048, 0.0050], device='cuda:6') 2023-04-26 20:02:13,352 INFO [finetune.py:976] (6/7) Epoch 7, batch 4150, loss[loss=0.201, simple_loss=0.2793, pruned_loss=0.06133, over 4921.00 frames. ], tot_loss[loss=0.2019, simple_loss=0.2662, pruned_loss=0.06887, over 955011.31 frames. ], batch size: 42, lr: 3.86e-03, grad_scale: 32.0 2023-04-26 20:02:14,137 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9085, 1.3899, 1.8414, 2.2352, 1.9662, 1.7705, 1.8110, 1.8491], device='cuda:6'), covar=tensor([0.7251, 1.0091, 1.0394, 1.0702, 0.8923, 1.2057, 1.2734, 0.9793], device='cuda:6'), in_proj_covar=tensor([0.0409, 0.0428, 0.0511, 0.0533, 0.0439, 0.0457, 0.0469, 0.0465], device='cuda:6'), out_proj_covar=tensor([9.9471e-05, 1.0590e-04, 1.1530e-04, 1.2623e-04, 1.0679e-04, 1.1073e-04, 1.1280e-04, 1.1332e-04], device='cuda:6') 2023-04-26 20:02:28,366 INFO [optim.py:369] (6/7) 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 20:02:57,365 INFO [finetune.py:976] (6/7) Epoch 7, batch 4200, loss[loss=0.1857, simple_loss=0.2333, pruned_loss=0.06905, over 4714.00 frames. ], tot_loss[loss=0.2021, simple_loss=0.2663, pruned_loss=0.06893, over 955018.79 frames. ], batch size: 23, lr: 3.86e-03, grad_scale: 32.0 2023-04-26 20:03:04,186 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=3.76 vs. limit=5.0 2023-04-26 20:03:29,905 INFO [finetune.py:976] (6/7) Epoch 7, batch 4250, loss[loss=0.1938, simple_loss=0.2662, pruned_loss=0.06063, over 4759.00 frames. ], tot_loss[loss=0.2006, simple_loss=0.2646, pruned_loss=0.0683, over 955033.12 frames. ], batch size: 28, lr: 3.86e-03, grad_scale: 32.0 2023-04-26 20:03:40,422 INFO [optim.py:369] (6/7) 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 20:03:44,051 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=38635.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:04:03,668 INFO [finetune.py:976] (6/7) Epoch 7, batch 4300, loss[loss=0.1993, simple_loss=0.2623, pruned_loss=0.06818, over 4940.00 frames. ], tot_loss[loss=0.1984, simple_loss=0.2616, pruned_loss=0.06759, over 955788.92 frames. ], batch size: 33, lr: 3.86e-03, grad_scale: 32.0 2023-04-26 20:04:26,432 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=38683.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:05:01,639 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-04-26 20:05:03,901 INFO [finetune.py:976] (6/7) Epoch 7, batch 4350, loss[loss=0.1792, simple_loss=0.2467, pruned_loss=0.05585, over 4821.00 frames. ], tot_loss[loss=0.195, simple_loss=0.2577, pruned_loss=0.06609, over 957064.53 frames. ], batch size: 30, lr: 3.86e-03, grad_scale: 32.0 2023-04-26 20:05:08,720 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=38716.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:05:18,154 INFO [optim.py:369] (6/7) 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 20:05:52,726 INFO [finetune.py:976] (6/7) Epoch 7, batch 4400, loss[loss=0.198, simple_loss=0.2774, pruned_loss=0.05932, over 4827.00 frames. ], tot_loss[loss=0.1952, simple_loss=0.2583, pruned_loss=0.06605, over 957054.62 frames. ], batch size: 40, lr: 3.86e-03, grad_scale: 32.0 2023-04-26 20:06:26,657 INFO [finetune.py:976] (6/7) Epoch 7, batch 4450, loss[loss=0.1806, simple_loss=0.2583, pruned_loss=0.05144, over 4927.00 frames. ], tot_loss[loss=0.1988, simple_loss=0.2627, pruned_loss=0.06743, over 956106.82 frames. ], batch size: 38, lr: 3.86e-03, grad_scale: 32.0 2023-04-26 20:06:36,492 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=38823.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:06:45,523 INFO [optim.py:369] (6/7) 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 20:07:37,013 INFO [finetune.py:976] (6/7) Epoch 7, batch 4500, loss[loss=0.2268, simple_loss=0.284, pruned_loss=0.08485, over 4900.00 frames. ], tot_loss[loss=0.2011, simple_loss=0.2649, pruned_loss=0.06866, over 953577.85 frames. ], batch size: 36, lr: 3.86e-03, grad_scale: 32.0 2023-04-26 20:08:00,068 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=38884.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:08:24,197 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-04-26 20:08:44,166 INFO [finetune.py:976] (6/7) Epoch 7, batch 4550, loss[loss=0.2157, simple_loss=0.2923, pruned_loss=0.06955, over 4797.00 frames. ], tot_loss[loss=0.2039, simple_loss=0.2679, pruned_loss=0.06992, over 954262.73 frames. ], batch size: 29, lr: 3.86e-03, grad_scale: 32.0 2023-04-26 20:08:58,071 INFO [optim.py:369] (6/7) 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 20:09:15,961 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0429, 2.4748, 0.8868, 1.5348, 1.7956, 1.3280, 3.3555, 1.6579], device='cuda:6'), covar=tensor([0.0666, 0.0637, 0.0814, 0.1205, 0.0539, 0.0946, 0.0214, 0.0645], device='cuda:6'), in_proj_covar=tensor([0.0052, 0.0068, 0.0051, 0.0048, 0.0052, 0.0053, 0.0079, 0.0051], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0011, 0.0007], device='cuda:6') 2023-04-26 20:09:28,418 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.4662, 1.2758, 4.2844, 3.9959, 3.8040, 4.0486, 4.0371, 3.7500], device='cuda:6'), covar=tensor([0.6761, 0.5695, 0.1045, 0.1630, 0.1059, 0.1478, 0.1301, 0.1399], device='cuda:6'), in_proj_covar=tensor([0.0308, 0.0304, 0.0406, 0.0410, 0.0346, 0.0403, 0.0315, 0.0364], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-26 20:09:50,214 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-04-26 20:09:50,306 INFO [finetune.py:976] (6/7) Epoch 7, batch 4600, loss[loss=0.1851, simple_loss=0.262, pruned_loss=0.0541, over 4787.00 frames. ], tot_loss[loss=0.2018, simple_loss=0.2662, pruned_loss=0.06869, over 954266.86 frames. ], batch size: 29, lr: 3.86e-03, grad_scale: 32.0 2023-04-26 20:09:51,047 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.1245, 1.4629, 1.2797, 1.7878, 1.5885, 1.6570, 1.3601, 3.0876], device='cuda:6'), covar=tensor([0.0642, 0.0799, 0.0829, 0.1143, 0.0631, 0.0528, 0.0727, 0.0171], device='cuda:6'), in_proj_covar=tensor([0.0039, 0.0039, 0.0041, 0.0044, 0.0040, 0.0040, 0.0039, 0.0060], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0017], device='cuda:6') 2023-04-26 20:10:30,976 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6328, 1.5664, 4.5383, 4.2097, 3.9950, 4.2417, 4.2371, 3.9660], device='cuda:6'), covar=tensor([0.6777, 0.5568, 0.0934, 0.1753, 0.1213, 0.1623, 0.1330, 0.1450], device='cuda:6'), in_proj_covar=tensor([0.0307, 0.0304, 0.0405, 0.0409, 0.0346, 0.0402, 0.0315, 0.0364], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-26 20:10:56,082 INFO [finetune.py:976] (6/7) Epoch 7, batch 4650, loss[loss=0.2283, simple_loss=0.2906, pruned_loss=0.083, over 4793.00 frames. ], tot_loss[loss=0.2005, simple_loss=0.264, pruned_loss=0.06844, over 953749.69 frames. ], batch size: 51, lr: 3.86e-03, grad_scale: 32.0 2023-04-26 20:10:56,176 INFO [zipformer.py:1188] (6/7) 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] (6/7) 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 20:11:27,551 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6467, 1.6690, 0.7610, 1.3535, 1.9095, 1.5434, 1.3998, 1.4838], device='cuda:6'), covar=tensor([0.0550, 0.0395, 0.0406, 0.0596, 0.0269, 0.0571, 0.0544, 0.0607], device='cuda:6'), in_proj_covar=tensor([0.0031, 0.0026, 0.0024, 0.0030, 0.0021, 0.0030, 0.0029, 0.0030], device='cuda:6'), out_proj_covar=tensor([0.0049, 0.0045, 0.0038, 0.0050, 0.0038, 0.0048, 0.0048, 0.0050], device='cuda:6') 2023-04-26 20:12:02,798 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=39064.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:12:03,962 INFO [finetune.py:976] (6/7) Epoch 7, batch 4700, loss[loss=0.1936, simple_loss=0.2562, pruned_loss=0.06551, over 4819.00 frames. ], tot_loss[loss=0.1981, simple_loss=0.2612, pruned_loss=0.06747, over 954173.77 frames. ], batch size: 25, lr: 3.86e-03, grad_scale: 32.0 2023-04-26 20:12:23,467 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-04-26 20:13:04,271 INFO [finetune.py:976] (6/7) Epoch 7, batch 4750, loss[loss=0.2084, simple_loss=0.2696, pruned_loss=0.0736, over 4796.00 frames. ], tot_loss[loss=0.1967, simple_loss=0.2592, pruned_loss=0.06705, over 952276.40 frames. ], batch size: 51, lr: 3.86e-03, grad_scale: 32.0 2023-04-26 20:13:24,484 INFO [optim.py:369] (6/7) 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 20:14:16,833 INFO [finetune.py:976] (6/7) Epoch 7, batch 4800, loss[loss=0.228, simple_loss=0.2842, pruned_loss=0.08587, over 4884.00 frames. ], tot_loss[loss=0.2, simple_loss=0.2624, pruned_loss=0.06878, over 951717.54 frames. ], batch size: 31, lr: 3.86e-03, grad_scale: 32.0 2023-04-26 20:14:31,378 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=39179.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:14:39,885 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.4628, 1.2897, 1.8358, 1.7554, 1.3364, 1.1435, 1.5064, 1.0008], device='cuda:6'), covar=tensor([0.0850, 0.0956, 0.0550, 0.0862, 0.1024, 0.1556, 0.0896, 0.1101], device='cuda:6'), in_proj_covar=tensor([0.0066, 0.0074, 0.0073, 0.0067, 0.0077, 0.0095, 0.0080, 0.0075], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-26 20:14:40,486 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=39185.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:15:09,473 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.4808, 2.9362, 0.8635, 1.6829, 1.7710, 2.1266, 1.7790, 0.9050], device='cuda:6'), covar=tensor([0.1323, 0.1013, 0.1887, 0.1316, 0.1022, 0.1027, 0.1448, 0.1989], device='cuda:6'), in_proj_covar=tensor([0.0119, 0.0255, 0.0143, 0.0125, 0.0135, 0.0155, 0.0121, 0.0123], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-26 20:15:12,274 INFO [finetune.py:976] (6/7) Epoch 7, batch 4850, loss[loss=0.2018, simple_loss=0.2651, pruned_loss=0.06923, over 4887.00 frames. ], tot_loss[loss=0.2023, simple_loss=0.266, pruned_loss=0.06929, over 950882.61 frames. ], batch size: 35, lr: 3.86e-03, grad_scale: 64.0 2023-04-26 20:15:21,782 INFO [optim.py:369] (6/7) 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 20:15:31,639 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=39246.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:15:45,477 INFO [finetune.py:976] (6/7) Epoch 7, batch 4900, loss[loss=0.1853, simple_loss=0.269, pruned_loss=0.05079, over 4821.00 frames. ], tot_loss[loss=0.2043, simple_loss=0.2683, pruned_loss=0.07021, over 953814.45 frames. ], batch size: 33, lr: 3.86e-03, grad_scale: 64.0 2023-04-26 20:16:09,168 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=39284.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:16:50,963 INFO [finetune.py:976] (6/7) Epoch 7, batch 4950, loss[loss=0.1759, simple_loss=0.2466, pruned_loss=0.05256, over 4877.00 frames. ], tot_loss[loss=0.2051, simple_loss=0.2692, pruned_loss=0.07047, over 952170.09 frames. ], batch size: 34, lr: 3.86e-03, grad_scale: 64.0 2023-04-26 20:17:06,089 INFO [optim.py:369] (6/7) 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 20:17:25,360 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=39345.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:17:36,094 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.6154, 3.3241, 3.1521, 3.2379, 3.6513, 3.2547, 4.1469, 3.0285], device='cuda:6'), covar=tensor([0.2470, 0.1364, 0.2021, 0.2439, 0.1002, 0.1983, 0.0898, 0.2497], device='cuda:6'), in_proj_covar=tensor([0.0346, 0.0354, 0.0433, 0.0364, 0.0388, 0.0382, 0.0386, 0.0421], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-26 20:17:49,761 INFO [finetune.py:976] (6/7) Epoch 7, batch 5000, loss[loss=0.1659, simple_loss=0.2237, pruned_loss=0.05403, over 4741.00 frames. ], tot_loss[loss=0.2023, simple_loss=0.2666, pruned_loss=0.06904, over 951746.70 frames. ], batch size: 59, lr: 3.86e-03, grad_scale: 64.0 2023-04-26 20:18:23,433 INFO [finetune.py:976] (6/7) Epoch 7, batch 5050, loss[loss=0.2032, simple_loss=0.2754, pruned_loss=0.06549, over 4937.00 frames. ], tot_loss[loss=0.2005, simple_loss=0.2642, pruned_loss=0.06841, over 953026.80 frames. ], batch size: 33, lr: 3.85e-03, grad_scale: 64.0 2023-04-26 20:18:33,400 INFO [optim.py:369] (6/7) 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 20:18:47,153 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.1615, 1.6727, 1.4482, 1.8170, 1.7420, 2.0124, 1.4522, 3.6785], device='cuda:6'), covar=tensor([0.0634, 0.0753, 0.0792, 0.1170, 0.0610, 0.0512, 0.0717, 0.0145], device='cuda:6'), in_proj_covar=tensor([0.0039, 0.0039, 0.0041, 0.0044, 0.0040, 0.0040, 0.0039, 0.0060], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0014, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:6') 2023-04-26 20:18:54,477 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=39463.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:18:55,791 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.10 vs. limit=5.0 2023-04-26 20:18:56,692 INFO [finetune.py:976] (6/7) Epoch 7, batch 5100, loss[loss=0.1862, simple_loss=0.2552, pruned_loss=0.05865, over 4788.00 frames. ], tot_loss[loss=0.1978, simple_loss=0.2607, pruned_loss=0.06742, over 952926.72 frames. ], batch size: 29, lr: 3.85e-03, grad_scale: 64.0 2023-04-26 20:18:59,470 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-04-26 20:19:02,315 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.81 vs. limit=2.0 2023-04-26 20:19:05,636 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=39479.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:19:22,696 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.4410, 0.6413, 1.2841, 1.8074, 1.5410, 1.3261, 1.3262, 1.4232], device='cuda:6'), covar=tensor([0.6460, 0.9393, 0.9288, 0.9673, 0.8152, 1.0790, 1.1223, 0.9330], device='cuda:6'), in_proj_covar=tensor([0.0412, 0.0430, 0.0514, 0.0536, 0.0441, 0.0459, 0.0472, 0.0468], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-26 20:19:23,912 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5917, 1.5451, 0.7237, 1.3026, 1.8632, 1.4713, 1.3792, 1.4360], device='cuda:6'), covar=tensor([0.0532, 0.0406, 0.0400, 0.0597, 0.0285, 0.0576, 0.0545, 0.0614], device='cuda:6'), in_proj_covar=tensor([0.0030, 0.0026, 0.0024, 0.0030, 0.0021, 0.0030, 0.0029, 0.0030], device='cuda:6'), out_proj_covar=tensor([0.0049, 0.0044, 0.0038, 0.0050, 0.0037, 0.0048, 0.0048, 0.0049], device='cuda:6') 2023-04-26 20:19:29,891 INFO [finetune.py:976] (6/7) Epoch 7, batch 5150, loss[loss=0.1642, simple_loss=0.2479, pruned_loss=0.04022, over 4908.00 frames. ], tot_loss[loss=0.199, simple_loss=0.2618, pruned_loss=0.06815, over 952764.42 frames. ], batch size: 37, lr: 3.85e-03, grad_scale: 32.0 2023-04-26 20:19:35,825 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=39524.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:19:38,071 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=39527.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:19:40,456 INFO [optim.py:369] (6/7) 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 20:19:47,591 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=39541.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:20:03,314 INFO [finetune.py:976] (6/7) Epoch 7, batch 5200, loss[loss=0.2679, simple_loss=0.3189, pruned_loss=0.1084, over 4809.00 frames. ], tot_loss[loss=0.2005, simple_loss=0.2646, pruned_loss=0.06824, over 953326.52 frames. ], batch size: 51, lr: 3.85e-03, grad_scale: 32.0 2023-04-26 20:20:06,368 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.3413, 3.3905, 2.7055, 2.8985, 2.4668, 2.7675, 2.7786, 2.2616], device='cuda:6'), covar=tensor([0.2544, 0.1142, 0.0895, 0.1431, 0.2925, 0.1244, 0.2017, 0.3083], device='cuda:6'), in_proj_covar=tensor([0.0303, 0.0323, 0.0234, 0.0295, 0.0318, 0.0275, 0.0264, 0.0288], device='cuda:6'), out_proj_covar=tensor([1.2292e-04, 1.3065e-04, 9.4845e-05, 1.1809e-04, 1.3098e-04, 1.1106e-04, 1.0805e-04, 1.1575e-04], device='cuda:6') 2023-04-26 20:20:37,216 INFO [finetune.py:976] (6/7) Epoch 7, batch 5250, loss[loss=0.2316, simple_loss=0.2862, pruned_loss=0.08852, over 4197.00 frames. ], tot_loss[loss=0.2022, simple_loss=0.2662, pruned_loss=0.06904, over 952119.29 frames. ], batch size: 65, lr: 3.85e-03, grad_scale: 32.0 2023-04-26 20:20:47,796 INFO [optim.py:369] (6/7) 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 20:20:53,211 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=39639.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 20:20:53,790 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=39640.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:21:10,369 INFO [finetune.py:976] (6/7) Epoch 7, batch 5300, loss[loss=0.1743, simple_loss=0.2398, pruned_loss=0.05443, over 4838.00 frames. ], tot_loss[loss=0.2022, simple_loss=0.2665, pruned_loss=0.06895, over 952711.80 frames. ], batch size: 49, lr: 3.85e-03, grad_scale: 32.0 2023-04-26 20:21:50,881 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=39700.0, num_to_drop=1, layers_to_drop={2} 2023-04-26 20:22:11,713 INFO [finetune.py:976] (6/7) Epoch 7, batch 5350, loss[loss=0.2183, simple_loss=0.2729, pruned_loss=0.0818, over 4716.00 frames. ], tot_loss[loss=0.2025, simple_loss=0.2673, pruned_loss=0.06884, over 954155.80 frames. ], batch size: 59, lr: 3.85e-03, grad_scale: 32.0 2023-04-26 20:22:31,528 INFO [optim.py:369] (6/7) 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 20:23:03,547 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=2.05 vs. limit=2.0 2023-04-26 20:23:17,037 INFO [finetune.py:976] (6/7) Epoch 7, batch 5400, loss[loss=0.1884, simple_loss=0.25, pruned_loss=0.06338, over 4783.00 frames. ], tot_loss[loss=0.1999, simple_loss=0.2646, pruned_loss=0.0676, over 956531.10 frames. ], batch size: 29, lr: 3.85e-03, grad_scale: 32.0 2023-04-26 20:23:17,753 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=39767.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:23:30,068 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=3.94 vs. limit=5.0 2023-04-26 20:24:11,502 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.3689, 1.6745, 1.5949, 2.0203, 1.8836, 1.9734, 1.4997, 4.2660], device='cuda:6'), covar=tensor([0.0590, 0.0737, 0.0743, 0.1138, 0.0595, 0.0580, 0.0719, 0.0107], device='cuda:6'), in_proj_covar=tensor([0.0039, 0.0039, 0.0040, 0.0044, 0.0040, 0.0040, 0.0039, 0.0059], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0014, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:6') 2023-04-26 20:24:18,122 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=39815.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:24:18,616 INFO [finetune.py:976] (6/7) Epoch 7, batch 5450, loss[loss=0.2197, simple_loss=0.2844, pruned_loss=0.07747, over 4846.00 frames. ], tot_loss[loss=0.1992, simple_loss=0.2628, pruned_loss=0.06779, over 956105.71 frames. ], batch size: 47, lr: 3.85e-03, grad_scale: 32.0 2023-04-26 20:24:20,494 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=39819.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:24:20,556 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.1145, 1.3013, 1.8975, 2.5315, 2.0346, 1.4630, 1.3109, 1.8863], device='cuda:6'), covar=tensor([0.4289, 0.4689, 0.2231, 0.3290, 0.3358, 0.3357, 0.5224, 0.2909], device='cuda:6'), in_proj_covar=tensor([0.0280, 0.0253, 0.0219, 0.0323, 0.0215, 0.0229, 0.0237, 0.0188], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:6') 2023-04-26 20:24:31,589 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=39828.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:24:33,272 INFO [optim.py:369] (6/7) 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 20:24:36,996 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-04-26 20:24:39,283 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6947, 1.5025, 1.6584, 2.0732, 2.0687, 1.6751, 1.3484, 1.7418], device='cuda:6'), covar=tensor([0.0864, 0.1267, 0.0813, 0.0599, 0.0567, 0.0911, 0.0855, 0.0732], device='cuda:6'), in_proj_covar=tensor([0.0199, 0.0204, 0.0181, 0.0179, 0.0179, 0.0193, 0.0162, 0.0188], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-26 20:24:40,439 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=39841.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:24:57,926 INFO [finetune.py:976] (6/7) Epoch 7, batch 5500, loss[loss=0.1719, simple_loss=0.2411, pruned_loss=0.0514, over 4757.00 frames. ], tot_loss[loss=0.1965, simple_loss=0.2596, pruned_loss=0.06675, over 954300.16 frames. ], batch size: 26, lr: 3.85e-03, grad_scale: 32.0 2023-04-26 20:25:04,136 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=39876.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:25:12,057 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=39889.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:25:31,743 INFO [finetune.py:976] (6/7) Epoch 7, batch 5550, loss[loss=0.2848, simple_loss=0.3273, pruned_loss=0.1212, over 4826.00 frames. ], tot_loss[loss=0.1987, simple_loss=0.2618, pruned_loss=0.06784, over 954960.14 frames. ], batch size: 47, lr: 3.85e-03, grad_scale: 32.0 2023-04-26 20:25:37,429 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8676, 1.2338, 5.2460, 4.9361, 4.5779, 5.0249, 4.6780, 4.5933], device='cuda:6'), covar=tensor([0.6805, 0.7133, 0.1142, 0.1907, 0.1074, 0.1666, 0.1118, 0.1580], device='cuda:6'), in_proj_covar=tensor([0.0311, 0.0310, 0.0411, 0.0417, 0.0351, 0.0409, 0.0321, 0.0371], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-26 20:25:40,907 INFO [optim.py:369] (6/7) 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 20:25:46,472 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=39940.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:25:48,624 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.3950, 2.0628, 1.6382, 1.8218, 2.1734, 1.7296, 2.4432, 1.4671], device='cuda:6'), covar=tensor([0.3251, 0.1643, 0.4658, 0.2627, 0.1689, 0.2323, 0.1527, 0.4260], device='cuda:6'), in_proj_covar=tensor([0.0345, 0.0350, 0.0431, 0.0362, 0.0386, 0.0382, 0.0385, 0.0418], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-26 20:26:02,971 INFO [finetune.py:976] (6/7) Epoch 7, batch 5600, loss[loss=0.2457, simple_loss=0.3069, pruned_loss=0.09228, over 4818.00 frames. ], tot_loss[loss=0.2019, simple_loss=0.2653, pruned_loss=0.06927, over 953091.33 frames. ], batch size: 40, lr: 3.85e-03, grad_scale: 32.0 2023-04-26 20:26:15,865 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=39988.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:26:18,806 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6668, 1.4323, 4.5172, 4.2310, 3.9507, 4.2781, 4.2331, 4.0011], device='cuda:6'), covar=tensor([0.6495, 0.5789, 0.1043, 0.1675, 0.1165, 0.1538, 0.1045, 0.1402], device='cuda:6'), in_proj_covar=tensor([0.0311, 0.0309, 0.0409, 0.0415, 0.0350, 0.0408, 0.0320, 0.0370], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-26 20:26:19,989 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=39995.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 20:26:38,933 INFO [finetune.py:976] (6/7) Epoch 7, batch 5650, loss[loss=0.1801, simple_loss=0.2517, pruned_loss=0.0543, over 4851.00 frames. ], tot_loss[loss=0.2028, simple_loss=0.267, pruned_loss=0.06931, over 954325.62 frames. ], batch size: 31, lr: 3.85e-03, grad_scale: 32.0 2023-04-26 20:26:53,488 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.1498, 2.5024, 0.9463, 1.3476, 1.8642, 1.2395, 3.3532, 1.7187], device='cuda:6'), covar=tensor([0.0702, 0.0630, 0.0823, 0.1385, 0.0536, 0.1080, 0.0265, 0.0684], device='cuda:6'), in_proj_covar=tensor([0.0054, 0.0069, 0.0051, 0.0048, 0.0053, 0.0054, 0.0080, 0.0052], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0011, 0.0008], device='cuda:6') 2023-04-26 20:26:53,978 INFO [optim.py:369] (6/7) 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 20:27:27,068 INFO [finetune.py:976] (6/7) Epoch 7, batch 5700, loss[loss=0.1831, simple_loss=0.23, pruned_loss=0.06808, over 3604.00 frames. ], tot_loss[loss=0.2014, simple_loss=0.2641, pruned_loss=0.06937, over 938307.12 frames. ], batch size: 15, lr: 3.85e-03, grad_scale: 32.0 2023-04-26 20:28:14,051 INFO [finetune.py:976] (6/7) Epoch 8, batch 0, loss[loss=0.2217, simple_loss=0.2874, pruned_loss=0.07802, over 4752.00 frames. ], tot_loss[loss=0.2217, simple_loss=0.2874, pruned_loss=0.07802, over 4752.00 frames. ], batch size: 26, lr: 3.85e-03, grad_scale: 32.0 2023-04-26 20:28:14,051 INFO [finetune.py:1001] (6/7) Computing validation loss 2023-04-26 20:28:25,119 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.4363, 1.2542, 1.6632, 1.5995, 1.3144, 1.1591, 1.3399, 0.8034], device='cuda:6'), covar=tensor([0.0645, 0.0854, 0.0542, 0.0731, 0.0843, 0.1341, 0.0709, 0.0948], device='cuda:6'), in_proj_covar=tensor([0.0066, 0.0074, 0.0073, 0.0067, 0.0077, 0.0096, 0.0080, 0.0075], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-26 20:28:30,561 INFO [finetune.py:1010] (6/7) Epoch 8, validation: loss=0.1574, simple_loss=0.2299, pruned_loss=0.04247, over 2265189.00 frames. 2023-04-26 20:28:30,562 INFO [finetune.py:1011] (6/7) Maximum memory allocated so far is 6326MB 2023-04-26 20:29:03,021 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=40119.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:29:03,424 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-04-26 20:29:05,434 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=40123.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:29:10,786 INFO [optim.py:369] (6/7) 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 20:29:20,982 INFO [finetune.py:976] (6/7) Epoch 8, batch 50, loss[loss=0.237, simple_loss=0.2895, pruned_loss=0.09227, over 4920.00 frames. ], tot_loss[loss=0.2085, simple_loss=0.2722, pruned_loss=0.07238, over 216373.25 frames. ], batch size: 41, lr: 3.85e-03, grad_scale: 32.0 2023-04-26 20:29:24,051 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6018, 1.5465, 0.6792, 1.3663, 1.8644, 1.4584, 1.3928, 1.4844], device='cuda:6'), covar=tensor([0.0519, 0.0396, 0.0403, 0.0575, 0.0288, 0.0566, 0.0517, 0.0598], device='cuda:6'), in_proj_covar=tensor([0.0030, 0.0026, 0.0024, 0.0030, 0.0021, 0.0030, 0.0029, 0.0030], device='cuda:6'), out_proj_covar=tensor([0.0049, 0.0044, 0.0038, 0.0050, 0.0038, 0.0048, 0.0048, 0.0050], device='cuda:6') 2023-04-26 20:29:36,059 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=40167.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:29:38,547 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=40171.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:29:41,449 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2023-04-26 20:29:42,173 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([3.9217, 3.8359, 2.8234, 4.5615, 4.0602, 3.9746, 1.7842, 3.9035], device='cuda:6'), covar=tensor([0.2033, 0.1265, 0.3769, 0.1411, 0.3116, 0.1953, 0.6123, 0.2484], device='cuda:6'), in_proj_covar=tensor([0.0244, 0.0218, 0.0253, 0.0310, 0.0303, 0.0253, 0.0274, 0.0272], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-26 20:29:54,377 INFO [finetune.py:976] (6/7) Epoch 8, batch 100, loss[loss=0.1776, simple_loss=0.2384, pruned_loss=0.0584, over 4923.00 frames. ], tot_loss[loss=0.1993, simple_loss=0.2614, pruned_loss=0.06861, over 377265.16 frames. ], batch size: 38, lr: 3.85e-03, grad_scale: 32.0 2023-04-26 20:29:55,589 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=40195.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:30:01,249 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5815, 1.1758, 1.4092, 1.2127, 1.7869, 1.4577, 1.2176, 1.3281], device='cuda:6'), covar=tensor([0.1639, 0.1701, 0.2019, 0.1493, 0.1021, 0.1552, 0.2137, 0.2271], device='cuda:6'), in_proj_covar=tensor([0.0308, 0.0328, 0.0357, 0.0303, 0.0341, 0.0328, 0.0309, 0.0354], device='cuda:6'), out_proj_covar=tensor([6.5548e-05, 6.9670e-05, 7.7417e-05, 6.2802e-05, 7.1811e-05, 7.0622e-05, 6.6724e-05, 7.5959e-05], device='cuda:6') 2023-04-26 20:30:10,523 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=40218.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:30:18,317 INFO [optim.py:369] (6/7) 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 20:30:28,097 INFO [finetune.py:976] (6/7) Epoch 8, batch 150, loss[loss=0.1767, simple_loss=0.2281, pruned_loss=0.06269, over 4817.00 frames. ], tot_loss[loss=0.1951, simple_loss=0.2565, pruned_loss=0.06692, over 504618.38 frames. ], batch size: 40, lr: 3.85e-03, grad_scale: 32.0 2023-04-26 20:30:36,432 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=40256.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:30:46,196 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([0.3895, 1.4061, 1.3723, 1.1035, 1.4879, 1.1206, 1.6857, 1.3019], device='cuda:6'), covar=tensor([0.3750, 0.1539, 0.5038, 0.2456, 0.1468, 0.2216, 0.1741, 0.4653], device='cuda:6'), in_proj_covar=tensor([0.0345, 0.0350, 0.0432, 0.0361, 0.0386, 0.0380, 0.0382, 0.0417], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-26 20:30:46,199 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=40272.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:30:50,429 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=40279.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:31:01,727 INFO [finetune.py:976] (6/7) Epoch 8, batch 200, loss[loss=0.267, simple_loss=0.3247, pruned_loss=0.1047, over 4742.00 frames. ], tot_loss[loss=0.1987, simple_loss=0.2591, pruned_loss=0.06908, over 606180.56 frames. ], batch size: 54, lr: 3.85e-03, grad_scale: 32.0 2023-04-26 20:31:02,443 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=40295.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 20:31:24,763 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=40330.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:31:25,242 INFO [optim.py:369] (6/7) 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 20:31:26,585 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=40333.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:31:34,058 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=40343.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 20:31:35,038 INFO [finetune.py:976] (6/7) Epoch 8, batch 250, loss[loss=0.1794, simple_loss=0.2232, pruned_loss=0.06783, over 4124.00 frames. ], tot_loss[loss=0.2037, simple_loss=0.2651, pruned_loss=0.07112, over 680698.82 frames. ], batch size: 17, lr: 3.85e-03, grad_scale: 32.0 2023-04-26 20:32:06,232 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 8, batch 300, loss[loss=0.1956, simple_loss=0.2451, pruned_loss=0.07301, over 4775.00 frames. ], tot_loss[loss=0.2021, simple_loss=0.2652, pruned_loss=0.06945, over 742847.90 frames. ], batch size: 26, lr: 3.85e-03, grad_scale: 32.0 2023-04-26 20:32:10,371 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.9970, 2.2326, 2.0195, 2.2923, 1.9645, 2.2016, 2.2424, 2.1535], device='cuda:6'), covar=tensor([0.5541, 0.8929, 0.7833, 0.6700, 0.8140, 1.0986, 0.8365, 0.7898], device='cuda:6'), in_proj_covar=tensor([0.0323, 0.0392, 0.0318, 0.0327, 0.0344, 0.0407, 0.0370, 0.0331], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-26 20:32:27,222 INFO [zipformer.py:1188] (6/7) 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] (6/7) 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 20:32:45,857 INFO [finetune.py:976] (6/7) Epoch 8, batch 350, loss[loss=0.2012, simple_loss=0.2711, pruned_loss=0.06568, over 4811.00 frames. ], tot_loss[loss=0.2024, simple_loss=0.2661, pruned_loss=0.0693, over 789782.54 frames. ], batch size: 25, lr: 3.85e-03, grad_scale: 32.0 2023-04-26 20:33:27,708 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=40471.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:33:27,756 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=40471.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:33:52,005 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=40492.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:33:53,112 INFO [finetune.py:976] (6/7) Epoch 8, batch 400, loss[loss=0.2024, simple_loss=0.2734, pruned_loss=0.06577, over 4713.00 frames. ], tot_loss[loss=0.2047, simple_loss=0.2684, pruned_loss=0.07053, over 828891.09 frames. ], batch size: 54, lr: 3.85e-03, grad_scale: 32.0 2023-04-26 20:33:58,315 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5943, 0.7032, 1.2840, 1.8949, 1.6604, 1.4699, 1.4018, 1.5038], device='cuda:6'), covar=tensor([0.6615, 0.8785, 0.9565, 0.9919, 0.7850, 1.0619, 1.0877, 0.8765], device='cuda:6'), in_proj_covar=tensor([0.0407, 0.0425, 0.0510, 0.0530, 0.0438, 0.0457, 0.0468, 0.0467], device='cuda:6'), out_proj_covar=tensor([9.9040e-05, 1.0525e-04, 1.1517e-04, 1.2578e-04, 1.0650e-04, 1.1073e-04, 1.1266e-04, 1.1357e-04], device='cuda:6') 2023-04-26 20:34:32,787 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=40519.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:34:45,209 INFO [optim.py:369] (6/7) 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 20:34:55,787 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-04-26 20:35:05,025 INFO [finetune.py:976] (6/7) Epoch 8, batch 450, loss[loss=0.179, simple_loss=0.255, pruned_loss=0.05153, over 4911.00 frames. ], tot_loss[loss=0.204, simple_loss=0.2677, pruned_loss=0.07019, over 857095.76 frames. ], batch size: 37, lr: 3.85e-03, grad_scale: 32.0 2023-04-26 20:35:05,794 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8694, 1.5983, 2.0972, 2.2908, 1.9560, 1.7661, 1.8673, 1.9479], device='cuda:6'), covar=tensor([0.7045, 1.0043, 1.0414, 1.0128, 0.8771, 1.3123, 1.3931, 1.1203], device='cuda:6'), in_proj_covar=tensor([0.0409, 0.0426, 0.0510, 0.0531, 0.0439, 0.0458, 0.0469, 0.0468], device='cuda:6'), out_proj_covar=tensor([9.9273e-05, 1.0554e-04, 1.1514e-04, 1.2591e-04, 1.0666e-04, 1.1091e-04, 1.1285e-04, 1.1361e-04], device='cuda:6') 2023-04-26 20:35:17,324 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=40551.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:35:18,626 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=40553.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:35:38,797 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=40574.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:35:43,764 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7281, 2.4031, 1.7443, 1.5382, 1.2915, 1.3270, 1.8056, 1.2736], device='cuda:6'), covar=tensor([0.1831, 0.1582, 0.1620, 0.2197, 0.2642, 0.2093, 0.1173, 0.2247], device='cuda:6'), in_proj_covar=tensor([0.0195, 0.0218, 0.0173, 0.0206, 0.0207, 0.0186, 0.0163, 0.0190], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:6') 2023-04-26 20:35:55,307 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.1032, 1.9599, 2.2741, 2.5625, 2.5427, 2.1005, 1.7461, 2.1406], device='cuda:6'), covar=tensor([0.0804, 0.1044, 0.0561, 0.0576, 0.0544, 0.0803, 0.0829, 0.0626], device='cuda:6'), in_proj_covar=tensor([0.0202, 0.0209, 0.0185, 0.0182, 0.0182, 0.0196, 0.0165, 0.0191], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-26 20:35:56,422 INFO [finetune.py:976] (6/7) Epoch 8, batch 500, loss[loss=0.1931, simple_loss=0.2491, pruned_loss=0.06852, over 4804.00 frames. ], tot_loss[loss=0.2016, simple_loss=0.2646, pruned_loss=0.06931, over 878382.48 frames. ], batch size: 29, lr: 3.85e-03, grad_scale: 32.0 2023-04-26 20:36:37,615 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=40628.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:36:39,369 INFO [optim.py:369] (6/7) 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 20:36:43,773 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6426, 1.2868, 1.2998, 1.3375, 1.9325, 1.5049, 1.1651, 1.2929], device='cuda:6'), covar=tensor([0.1477, 0.1436, 0.2037, 0.1111, 0.0694, 0.1443, 0.2033, 0.1986], device='cuda:6'), in_proj_covar=tensor([0.0307, 0.0326, 0.0355, 0.0300, 0.0340, 0.0326, 0.0308, 0.0352], device='cuda:6'), out_proj_covar=tensor([6.5369e-05, 6.9272e-05, 7.7050e-05, 6.2342e-05, 7.1520e-05, 7.0304e-05, 6.6552e-05, 7.5605e-05], device='cuda:6') 2023-04-26 20:36:52,502 INFO [finetune.py:976] (6/7) Epoch 8, batch 550, loss[loss=0.1329, simple_loss=0.1835, pruned_loss=0.04112, over 2809.00 frames. ], tot_loss[loss=0.1978, simple_loss=0.2609, pruned_loss=0.06733, over 894699.66 frames. ], batch size: 11, lr: 3.85e-03, grad_scale: 32.0 2023-04-26 20:36:54,507 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.6238, 2.6998, 3.0255, 3.3225, 2.9649, 2.6572, 2.2629, 2.6164], device='cuda:6'), covar=tensor([0.0928, 0.0901, 0.0486, 0.0600, 0.0635, 0.0927, 0.0843, 0.0712], device='cuda:6'), in_proj_covar=tensor([0.0201, 0.0208, 0.0184, 0.0182, 0.0181, 0.0195, 0.0165, 0.0190], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-26 20:36:56,972 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=40651.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:37:25,620 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-04-26 20:37:29,169 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.14 vs. limit=2.0 2023-04-26 20:37:47,196 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=40686.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:37:52,036 INFO [finetune.py:976] (6/7) Epoch 8, batch 600, loss[loss=0.2284, simple_loss=0.2945, pruned_loss=0.08116, over 4870.00 frames. ], tot_loss[loss=0.1993, simple_loss=0.2622, pruned_loss=0.0682, over 905261.05 frames. ], batch size: 34, lr: 3.84e-03, grad_scale: 32.0 2023-04-26 20:37:57,818 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([3.5928, 3.6055, 2.7683, 4.2023, 3.6796, 3.5811, 1.4933, 3.5321], device='cuda:6'), covar=tensor([0.1906, 0.1237, 0.2919, 0.1943, 0.3015, 0.2031, 0.6244, 0.2533], device='cuda:6'), in_proj_covar=tensor([0.0245, 0.0219, 0.0253, 0.0309, 0.0303, 0.0254, 0.0274, 0.0273], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-26 20:38:12,520 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5994, 1.5705, 0.9987, 1.3257, 1.8046, 1.4813, 1.4351, 1.4080], device='cuda:6'), covar=tensor([0.0504, 0.0373, 0.0396, 0.0548, 0.0297, 0.0526, 0.0463, 0.0574], device='cuda:6'), in_proj_covar=tensor([0.0030, 0.0026, 0.0024, 0.0030, 0.0021, 0.0030, 0.0029, 0.0030], device='cuda:6'), out_proj_covar=tensor([0.0049, 0.0044, 0.0038, 0.0050, 0.0037, 0.0048, 0.0048, 0.0050], device='cuda:6') 2023-04-26 20:38:21,415 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=40712.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:38:44,483 INFO [optim.py:369] (6/7) 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 20:39:02,721 INFO [finetune.py:976] (6/7) Epoch 8, batch 650, loss[loss=0.1685, simple_loss=0.2451, pruned_loss=0.04595, over 4916.00 frames. ], tot_loss[loss=0.2011, simple_loss=0.2646, pruned_loss=0.06874, over 914709.01 frames. ], batch size: 38, lr: 3.84e-03, grad_scale: 32.0 2023-04-26 20:40:08,309 INFO [finetune.py:976] (6/7) Epoch 8, batch 700, loss[loss=0.1546, simple_loss=0.2278, pruned_loss=0.04068, over 4809.00 frames. ], tot_loss[loss=0.2016, simple_loss=0.2655, pruned_loss=0.06888, over 921827.48 frames. ], batch size: 45, lr: 3.84e-03, grad_scale: 32.0 2023-04-26 20:40:10,617 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-04-26 20:40:21,219 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9975, 1.4298, 1.8406, 2.2889, 1.7880, 1.3734, 1.1741, 1.5954], device='cuda:6'), covar=tensor([0.4126, 0.4222, 0.2007, 0.3041, 0.3500, 0.3480, 0.5687, 0.3197], device='cuda:6'), in_proj_covar=tensor([0.0279, 0.0252, 0.0218, 0.0321, 0.0214, 0.0228, 0.0237, 0.0188], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:6') 2023-04-26 20:40:29,116 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.4636, 1.8114, 2.2043, 2.8569, 2.2316, 1.7441, 1.6211, 2.2357], device='cuda:6'), covar=tensor([0.4171, 0.3960, 0.1894, 0.3421, 0.3854, 0.3365, 0.5338, 0.2934], device='cuda:6'), in_proj_covar=tensor([0.0279, 0.0252, 0.0218, 0.0321, 0.0214, 0.0228, 0.0237, 0.0188], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:6') 2023-04-26 20:40:30,803 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8573, 2.2868, 1.8519, 2.1264, 1.6046, 1.8086, 1.8730, 1.5767], device='cuda:6'), covar=tensor([0.2205, 0.1326, 0.0978, 0.1244, 0.3454, 0.1374, 0.2028, 0.2576], device='cuda:6'), in_proj_covar=tensor([0.0301, 0.0320, 0.0232, 0.0294, 0.0317, 0.0274, 0.0262, 0.0285], device='cuda:6'), out_proj_covar=tensor([1.2220e-04, 1.2946e-04, 9.4036e-05, 1.1768e-04, 1.3031e-04, 1.1063e-04, 1.0745e-04, 1.1464e-04], device='cuda:6') 2023-04-26 20:40:41,311 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([0.8765, 1.2754, 3.2604, 2.9943, 2.9259, 3.2032, 3.1806, 2.9036], device='cuda:6'), covar=tensor([0.7183, 0.4989, 0.1462, 0.2261, 0.1387, 0.1580, 0.1630, 0.1488], device='cuda:6'), in_proj_covar=tensor([0.0310, 0.0306, 0.0407, 0.0413, 0.0348, 0.0403, 0.0316, 0.0367], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-26 20:40:55,217 INFO [optim.py:369] (6/7) 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 20:41:02,237 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7318, 2.2838, 1.6687, 1.4960, 1.2606, 1.3247, 1.6032, 1.1960], device='cuda:6'), covar=tensor([0.1944, 0.1440, 0.1820, 0.2052, 0.2821, 0.2270, 0.1330, 0.2332], device='cuda:6'), in_proj_covar=tensor([0.0194, 0.0216, 0.0172, 0.0204, 0.0206, 0.0185, 0.0163, 0.0189], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:6') 2023-04-26 20:41:14,342 INFO [finetune.py:976] (6/7) Epoch 8, batch 750, loss[loss=0.2299, simple_loss=0.2975, pruned_loss=0.08111, over 4797.00 frames. ], tot_loss[loss=0.2034, simple_loss=0.2679, pruned_loss=0.06943, over 931351.30 frames. ], batch size: 45, lr: 3.84e-03, grad_scale: 32.0 2023-04-26 20:41:16,863 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=40848.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:41:24,344 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=40851.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:41:38,683 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=40865.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:41:42,868 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.0759, 2.2322, 1.9568, 1.9765, 2.4230, 1.8073, 2.9556, 1.7586], device='cuda:6'), covar=tensor([0.4368, 0.2014, 0.5084, 0.3420, 0.1903, 0.2828, 0.1276, 0.4276], device='cuda:6'), in_proj_covar=tensor([0.0343, 0.0347, 0.0427, 0.0360, 0.0384, 0.0378, 0.0379, 0.0414], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-26 20:41:44,669 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=40874.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:41:57,695 INFO [finetune.py:976] (6/7) Epoch 8, batch 800, loss[loss=0.225, simple_loss=0.2918, pruned_loss=0.07913, over 4907.00 frames. ], tot_loss[loss=0.2026, simple_loss=0.2675, pruned_loss=0.06882, over 936773.38 frames. ], batch size: 36, lr: 3.84e-03, grad_scale: 32.0 2023-04-26 20:42:00,809 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=40899.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:42:01,464 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=40900.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:42:16,690 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=40922.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:42:19,198 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=40926.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:42:20,406 INFO [zipformer.py:1188] (6/7) 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] (6/7) 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 20:42:31,000 INFO [finetune.py:976] (6/7) Epoch 8, batch 850, loss[loss=0.2155, simple_loss=0.2716, pruned_loss=0.07965, over 4786.00 frames. ], tot_loss[loss=0.1998, simple_loss=0.2641, pruned_loss=0.06776, over 941806.87 frames. ], batch size: 28, lr: 3.84e-03, grad_scale: 32.0 2023-04-26 20:42:41,344 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=40961.0, num_to_drop=1, layers_to_drop={3} 2023-04-26 20:42:47,651 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6391, 2.1704, 1.6540, 1.4157, 1.2303, 1.2900, 1.6043, 1.1313], device='cuda:6'), covar=tensor([0.1620, 0.1262, 0.1553, 0.1896, 0.2499, 0.1986, 0.1099, 0.2198], device='cuda:6'), in_proj_covar=tensor([0.0195, 0.0217, 0.0173, 0.0205, 0.0207, 0.0185, 0.0163, 0.0190], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:6') 2023-04-26 20:42:52,275 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=40976.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:42:58,819 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=40986.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:43:03,442 INFO [finetune.py:976] (6/7) Epoch 8, batch 900, loss[loss=0.1899, simple_loss=0.2552, pruned_loss=0.06233, over 4778.00 frames. ], tot_loss[loss=0.1984, simple_loss=0.2622, pruned_loss=0.06733, over 942677.68 frames. ], batch size: 28, lr: 3.84e-03, grad_scale: 32.0 2023-04-26 20:43:05,778 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=40997.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:43:12,089 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=41007.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:43:28,963 INFO [optim.py:369] (6/7) 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 20:43:30,304 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7000, 2.0744, 1.6850, 1.8556, 1.5580, 1.6726, 1.7610, 1.3923], device='cuda:6'), covar=tensor([0.1575, 0.1030, 0.0911, 0.1117, 0.2759, 0.1037, 0.1500, 0.1967], device='cuda:6'), in_proj_covar=tensor([0.0302, 0.0322, 0.0234, 0.0296, 0.0319, 0.0276, 0.0264, 0.0286], device='cuda:6'), out_proj_covar=tensor([1.2262e-04, 1.3032e-04, 9.4818e-05, 1.1882e-04, 1.3128e-04, 1.1127e-04, 1.0813e-04, 1.1512e-04], device='cuda:6') 2023-04-26 20:43:30,876 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=41034.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:43:37,318 INFO [finetune.py:976] (6/7) Epoch 8, batch 950, loss[loss=0.2214, simple_loss=0.2875, pruned_loss=0.07765, over 4812.00 frames. ], tot_loss[loss=0.1972, simple_loss=0.2606, pruned_loss=0.06689, over 946461.44 frames. ], batch size: 45, lr: 3.84e-03, grad_scale: 32.0 2023-04-26 20:43:38,047 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=41045.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:43:46,385 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=41058.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:43:53,558 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9321, 2.4666, 1.9547, 2.1074, 1.6088, 1.9024, 2.0506, 1.5563], device='cuda:6'), covar=tensor([0.2186, 0.1293, 0.0967, 0.1448, 0.3287, 0.1311, 0.2105, 0.2707], device='cuda:6'), in_proj_covar=tensor([0.0304, 0.0324, 0.0235, 0.0297, 0.0321, 0.0276, 0.0265, 0.0288], device='cuda:6'), out_proj_covar=tensor([1.2317e-04, 1.3103e-04, 9.5031e-05, 1.1923e-04, 1.3190e-04, 1.1156e-04, 1.0857e-04, 1.1562e-04], device='cuda:6') 2023-04-26 20:44:02,311 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.3975, 1.6585, 1.6978, 1.7602, 1.6800, 1.8626, 1.8648, 1.7313], device='cuda:6'), covar=tensor([0.5331, 0.8372, 0.6927, 0.6553, 0.7771, 1.1148, 0.8347, 0.7334], device='cuda:6'), in_proj_covar=tensor([0.0324, 0.0389, 0.0317, 0.0329, 0.0343, 0.0408, 0.0369, 0.0331], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-26 20:44:04,051 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9582, 2.3857, 1.0056, 1.2202, 1.7711, 1.1155, 3.2059, 1.6246], device='cuda:6'), covar=tensor([0.0687, 0.0691, 0.0840, 0.1346, 0.0528, 0.1077, 0.0245, 0.0701], device='cuda:6'), in_proj_covar=tensor([0.0053, 0.0069, 0.0051, 0.0048, 0.0052, 0.0053, 0.0079, 0.0051], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0011, 0.0008], device='cuda:6') 2023-04-26 20:44:10,578 INFO [finetune.py:976] (6/7) Epoch 8, batch 1000, loss[loss=0.2133, simple_loss=0.2796, pruned_loss=0.07346, over 4133.00 frames. ], tot_loss[loss=0.1996, simple_loss=0.2634, pruned_loss=0.06793, over 948759.02 frames. ], batch size: 65, lr: 3.84e-03, grad_scale: 32.0 2023-04-26 20:44:16,006 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=41102.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:44:18,393 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=41106.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:44:35,826 INFO [optim.py:369] (6/7) 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 20:44:37,754 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.1227, 2.0945, 1.6852, 1.7954, 2.1306, 1.6241, 2.4822, 1.4922], device='cuda:6'), covar=tensor([0.3994, 0.1650, 0.5106, 0.3130, 0.1965, 0.2753, 0.1680, 0.4524], device='cuda:6'), in_proj_covar=tensor([0.0345, 0.0349, 0.0431, 0.0363, 0.0387, 0.0380, 0.0382, 0.0416], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-26 20:44:44,111 INFO [finetune.py:976] (6/7) Epoch 8, batch 1050, loss[loss=0.1645, simple_loss=0.2468, pruned_loss=0.04112, over 4765.00 frames. ], tot_loss[loss=0.2008, simple_loss=0.2651, pruned_loss=0.06825, over 950331.60 frames. ], batch size: 28, lr: 3.84e-03, grad_scale: 32.0 2023-04-26 20:44:46,607 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=41148.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:44:56,171 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=41163.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:45:27,984 INFO [finetune.py:976] (6/7) Epoch 8, batch 1100, loss[loss=0.2114, simple_loss=0.2734, pruned_loss=0.07474, over 4896.00 frames. ], tot_loss[loss=0.2015, simple_loss=0.2661, pruned_loss=0.06846, over 951814.19 frames. ], batch size: 43, lr: 3.84e-03, grad_scale: 32.0 2023-04-26 20:45:29,278 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=41196.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:45:31,129 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.3089, 2.6720, 1.2141, 1.5336, 2.3382, 1.4680, 3.6591, 2.1613], device='cuda:6'), covar=tensor([0.0623, 0.0649, 0.0717, 0.1260, 0.0407, 0.0947, 0.0261, 0.0531], device='cuda:6'), in_proj_covar=tensor([0.0052, 0.0068, 0.0050, 0.0048, 0.0052, 0.0053, 0.0079, 0.0051], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0011, 0.0007], device='cuda:6') 2023-04-26 20:45:50,740 INFO [zipformer.py:1188] (6/7) 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] (6/7) 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 20:46:17,819 INFO [finetune.py:976] (6/7) Epoch 8, batch 1150, loss[loss=0.2103, simple_loss=0.2738, pruned_loss=0.07343, over 4814.00 frames. ], tot_loss[loss=0.1999, simple_loss=0.2652, pruned_loss=0.0673, over 953070.65 frames. ], batch size: 38, lr: 3.84e-03, grad_scale: 32.0 2023-04-26 20:46:36,333 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=41256.0, num_to_drop=1, layers_to_drop={3} 2023-04-26 20:46:38,825 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7079, 1.3809, 1.7957, 2.1662, 1.9055, 1.6469, 1.7361, 1.7413], device='cuda:6'), covar=tensor([0.6980, 1.0142, 0.9664, 0.8974, 0.8017, 1.1353, 1.1748, 1.1337], device='cuda:6'), in_proj_covar=tensor([0.0410, 0.0427, 0.0509, 0.0531, 0.0438, 0.0457, 0.0470, 0.0467], device='cuda:6'), out_proj_covar=tensor([9.9515e-05, 1.0600e-04, 1.1501e-04, 1.2588e-04, 1.0660e-04, 1.1080e-04, 1.1322e-04, 1.1355e-04], device='cuda:6') 2023-04-26 20:46:41,509 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.86 vs. limit=5.0 2023-04-26 20:47:24,548 INFO [finetune.py:976] (6/7) Epoch 8, batch 1200, loss[loss=0.1574, simple_loss=0.2268, pruned_loss=0.04399, over 4755.00 frames. ], tot_loss[loss=0.2, simple_loss=0.2647, pruned_loss=0.06766, over 951232.81 frames. ], batch size: 27, lr: 3.84e-03, grad_scale: 32.0 2023-04-26 20:47:44,084 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=41307.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:47:59,151 INFO [optim.py:369] (6/7) 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 20:48:08,618 INFO [finetune.py:976] (6/7) Epoch 8, batch 1250, loss[loss=0.1744, simple_loss=0.2529, pruned_loss=0.04794, over 4817.00 frames. ], tot_loss[loss=0.1982, simple_loss=0.2626, pruned_loss=0.06689, over 953035.49 frames. ], batch size: 30, lr: 3.84e-03, grad_scale: 32.0 2023-04-26 20:48:14,608 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=41353.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:48:15,820 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=41355.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:48:27,976 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2023-04-26 20:48:42,247 INFO [finetune.py:976] (6/7) Epoch 8, batch 1300, loss[loss=0.2231, simple_loss=0.2779, pruned_loss=0.08415, over 4174.00 frames. ], tot_loss[loss=0.1951, simple_loss=0.259, pruned_loss=0.06558, over 955122.09 frames. ], batch size: 65, lr: 3.84e-03, grad_scale: 32.0 2023-04-26 20:48:46,733 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=41401.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:49:03,337 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-04-26 20:49:05,847 INFO [optim.py:369] (6/7) 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 20:49:15,124 INFO [finetune.py:976] (6/7) Epoch 8, batch 1350, loss[loss=0.2043, simple_loss=0.2753, pruned_loss=0.06668, over 4900.00 frames. ], tot_loss[loss=0.1957, simple_loss=0.26, pruned_loss=0.06575, over 956695.82 frames. ], batch size: 37, lr: 3.84e-03, grad_scale: 32.0 2023-04-26 20:49:25,037 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=41458.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:49:48,138 INFO [finetune.py:976] (6/7) Epoch 8, batch 1400, loss[loss=0.1767, simple_loss=0.2503, pruned_loss=0.05158, over 4915.00 frames. ], tot_loss[loss=0.1976, simple_loss=0.2625, pruned_loss=0.06638, over 957566.90 frames. ], batch size: 36, lr: 3.84e-03, grad_scale: 64.0 2023-04-26 20:49:48,733 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8696, 4.3596, 0.8188, 2.0881, 2.3731, 2.8314, 2.3803, 0.8094], device='cuda:6'), covar=tensor([0.1420, 0.0927, 0.2293, 0.1428, 0.1092, 0.1154, 0.1580, 0.2458], device='cuda:6'), in_proj_covar=tensor([0.0119, 0.0255, 0.0143, 0.0125, 0.0136, 0.0156, 0.0122, 0.0124], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-26 20:50:06,403 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=41521.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:50:12,932 INFO [optim.py:369] (6/7) 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] (6/7) Epoch 8, batch 1450, loss[loss=0.2798, simple_loss=0.2931, pruned_loss=0.1332, over 4111.00 frames. ], tot_loss[loss=0.1995, simple_loss=0.2645, pruned_loss=0.06721, over 955424.77 frames. ], batch size: 18, lr: 3.84e-03, grad_scale: 64.0 2023-04-26 20:50:30,555 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=41556.0, num_to_drop=1, layers_to_drop={2} 2023-04-26 20:50:38,926 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=41569.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:50:38,970 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6266, 1.8697, 1.0963, 1.3454, 2.1839, 1.4857, 1.4155, 1.5518], device='cuda:6'), covar=tensor([0.0565, 0.0380, 0.0362, 0.0610, 0.0266, 0.0577, 0.0559, 0.0587], device='cuda:6'), in_proj_covar=tensor([0.0030, 0.0026, 0.0024, 0.0030, 0.0021, 0.0030, 0.0029, 0.0030], device='cuda:6'), out_proj_covar=tensor([0.0050, 0.0045, 0.0038, 0.0050, 0.0038, 0.0049, 0.0048, 0.0050], device='cuda:6') 2023-04-26 20:50:54,401 INFO [finetune.py:976] (6/7) Epoch 8, batch 1500, loss[loss=0.208, simple_loss=0.2909, pruned_loss=0.06255, over 4922.00 frames. ], tot_loss[loss=0.2014, simple_loss=0.2663, pruned_loss=0.06822, over 955496.70 frames. ], batch size: 42, lr: 3.84e-03, grad_scale: 64.0 2023-04-26 20:51:01,803 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.1876, 1.4470, 1.3295, 1.6632, 1.5011, 1.7527, 1.3487, 3.0853], device='cuda:6'), covar=tensor([0.0680, 0.0829, 0.0860, 0.1272, 0.0700, 0.0525, 0.0792, 0.0167], device='cuda:6'), in_proj_covar=tensor([0.0039, 0.0039, 0.0041, 0.0044, 0.0041, 0.0040, 0.0039, 0.0060], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0014, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:6') 2023-04-26 20:51:02,378 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=41604.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:51:25,118 INFO [optim.py:369] (6/7) 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 20:51:38,883 INFO [finetune.py:976] (6/7) Epoch 8, batch 1550, loss[loss=0.1721, simple_loss=0.2393, pruned_loss=0.05248, over 4827.00 frames. ], tot_loss[loss=0.2004, simple_loss=0.2662, pruned_loss=0.06728, over 956819.67 frames. ], batch size: 30, lr: 3.84e-03, grad_scale: 64.0 2023-04-26 20:51:57,800 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=41653.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:51:58,453 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=41654.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:52:02,029 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6669, 1.2700, 1.7240, 2.0735, 1.7531, 1.6658, 1.7206, 1.7661], device='cuda:6'), covar=tensor([0.6583, 0.8896, 0.9046, 0.9508, 0.8331, 1.1231, 1.0923, 0.9870], device='cuda:6'), in_proj_covar=tensor([0.0406, 0.0424, 0.0504, 0.0526, 0.0435, 0.0454, 0.0465, 0.0463], device='cuda:6'), out_proj_covar=tensor([9.8659e-05, 1.0498e-04, 1.1405e-04, 1.2486e-04, 1.0577e-04, 1.0996e-04, 1.1194e-04, 1.1260e-04], device='cuda:6') 2023-04-26 20:52:31,820 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9137, 1.8562, 2.2524, 2.3534, 1.8130, 1.5410, 1.9457, 1.1751], device='cuda:6'), covar=tensor([0.0706, 0.0960, 0.0699, 0.0870, 0.0953, 0.1268, 0.0935, 0.0925], device='cuda:6'), in_proj_covar=tensor([0.0065, 0.0073, 0.0072, 0.0067, 0.0076, 0.0095, 0.0079, 0.0074], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-26 20:52:47,106 INFO [finetune.py:976] (6/7) Epoch 8, batch 1600, loss[loss=0.2021, simple_loss=0.2645, pruned_loss=0.06981, over 4922.00 frames. ], tot_loss[loss=0.1995, simple_loss=0.2643, pruned_loss=0.06731, over 957819.32 frames. ], batch size: 38, lr: 3.84e-03, grad_scale: 64.0 2023-04-26 20:52:56,990 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=41701.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:52:57,017 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=41701.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:53:12,211 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=41715.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 20:53:22,288 INFO [optim.py:369] (6/7) 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 20:53:30,673 INFO [finetune.py:976] (6/7) Epoch 8, batch 1650, loss[loss=0.1867, simple_loss=0.2489, pruned_loss=0.06221, over 4914.00 frames. ], tot_loss[loss=0.1963, simple_loss=0.2607, pruned_loss=0.06588, over 958318.21 frames. ], batch size: 37, lr: 3.84e-03, grad_scale: 64.0 2023-04-26 20:53:33,799 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=41749.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:53:35,636 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([3.7279, 3.6173, 2.8843, 4.3446, 3.6430, 3.7333, 1.7752, 3.7030], device='cuda:6'), covar=tensor([0.1782, 0.1287, 0.3315, 0.1274, 0.3142, 0.1845, 0.5588, 0.2257], device='cuda:6'), in_proj_covar=tensor([0.0247, 0.0220, 0.0252, 0.0310, 0.0303, 0.0253, 0.0275, 0.0273], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-26 20:53:40,312 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=41758.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:54:03,920 INFO [finetune.py:976] (6/7) Epoch 8, batch 1700, loss[loss=0.1812, simple_loss=0.2436, pruned_loss=0.05945, over 3993.00 frames. ], tot_loss[loss=0.1944, simple_loss=0.2584, pruned_loss=0.06519, over 957348.16 frames. ], batch size: 17, lr: 3.84e-03, grad_scale: 64.0 2023-04-26 20:54:11,564 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=41806.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:54:24,087 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8731, 1.6425, 1.9616, 2.1994, 1.9600, 1.7877, 1.8931, 1.8714], device='cuda:6'), covar=tensor([0.7153, 0.9835, 1.0854, 0.9899, 0.7966, 1.2627, 1.3642, 1.2049], device='cuda:6'), in_proj_covar=tensor([0.0409, 0.0427, 0.0509, 0.0530, 0.0439, 0.0457, 0.0470, 0.0467], device='cuda:6'), out_proj_covar=tensor([9.9408e-05, 1.0587e-04, 1.1501e-04, 1.2577e-04, 1.0668e-04, 1.1062e-04, 1.1308e-04, 1.1353e-04], device='cuda:6') 2023-04-26 20:54:29,250 INFO [optim.py:369] (6/7) 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 20:54:31,876 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-04-26 20:54:37,045 INFO [finetune.py:976] (6/7) Epoch 8, batch 1750, loss[loss=0.2069, simple_loss=0.2794, pruned_loss=0.06723, over 4809.00 frames. ], tot_loss[loss=0.1955, simple_loss=0.2597, pruned_loss=0.0657, over 957430.06 frames. ], batch size: 51, lr: 3.84e-03, grad_scale: 64.0 2023-04-26 20:54:54,151 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6577, 1.7993, 1.2043, 1.3338, 2.0477, 1.5396, 1.4370, 1.5375], device='cuda:6'), covar=tensor([0.0523, 0.0377, 0.0331, 0.0556, 0.0271, 0.0544, 0.0516, 0.0597], device='cuda:6'), in_proj_covar=tensor([0.0030, 0.0026, 0.0024, 0.0030, 0.0021, 0.0030, 0.0029, 0.0030], device='cuda:6'), out_proj_covar=tensor([0.0049, 0.0045, 0.0038, 0.0050, 0.0038, 0.0049, 0.0048, 0.0050], device='cuda:6') 2023-04-26 20:55:00,123 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.4055, 1.7257, 2.2366, 3.0138, 2.1699, 1.7369, 1.5437, 2.2056], device='cuda:6'), covar=tensor([0.3856, 0.4319, 0.1938, 0.3056, 0.3721, 0.3351, 0.5106, 0.2824], device='cuda:6'), in_proj_covar=tensor([0.0282, 0.0253, 0.0219, 0.0323, 0.0215, 0.0230, 0.0237, 0.0188], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:6') 2023-04-26 20:55:03,141 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.5281, 3.0071, 1.0067, 1.6379, 2.4265, 1.5417, 4.3006, 2.1520], device='cuda:6'), covar=tensor([0.0604, 0.0794, 0.0875, 0.1418, 0.0493, 0.0998, 0.0188, 0.0604], device='cuda:6'), in_proj_covar=tensor([0.0052, 0.0068, 0.0050, 0.0048, 0.0052, 0.0052, 0.0079, 0.0051], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0007], device='cuda:6') 2023-04-26 20:55:10,266 INFO [finetune.py:976] (6/7) Epoch 8, batch 1800, loss[loss=0.2147, simple_loss=0.2749, pruned_loss=0.07722, over 4876.00 frames. ], tot_loss[loss=0.1979, simple_loss=0.2628, pruned_loss=0.06653, over 958115.83 frames. ], batch size: 34, lr: 3.84e-03, grad_scale: 64.0 2023-04-26 20:55:18,162 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=41906.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:55:35,583 INFO [optim.py:369] (6/7) 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 20:55:43,855 INFO [finetune.py:976] (6/7) Epoch 8, batch 1850, loss[loss=0.195, simple_loss=0.2674, pruned_loss=0.06125, over 4814.00 frames. ], tot_loss[loss=0.2004, simple_loss=0.2656, pruned_loss=0.06757, over 958587.30 frames. ], batch size: 40, lr: 3.84e-03, grad_scale: 64.0 2023-04-26 20:55:46,381 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=41948.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 20:55:52,280 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.3458, 2.1933, 2.2064, 2.0026, 2.5172, 1.9765, 2.9973, 1.9224], device='cuda:6'), covar=tensor([0.3902, 0.2019, 0.4146, 0.3156, 0.1492, 0.2473, 0.1439, 0.4100], device='cuda:6'), in_proj_covar=tensor([0.0350, 0.0354, 0.0437, 0.0366, 0.0392, 0.0386, 0.0385, 0.0422], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-26 20:55:58,806 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=41967.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 20:56:02,847 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7291, 2.7609, 1.6724, 1.9873, 1.3507, 1.3020, 1.9664, 1.3476], device='cuda:6'), covar=tensor([0.2017, 0.1629, 0.1945, 0.2014, 0.2769, 0.2412, 0.1247, 0.2252], device='cuda:6'), in_proj_covar=tensor([0.0196, 0.0219, 0.0174, 0.0206, 0.0208, 0.0186, 0.0164, 0.0190], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:6') 2023-04-26 20:56:17,000 INFO [finetune.py:976] (6/7) Epoch 8, batch 1900, loss[loss=0.1929, simple_loss=0.2753, pruned_loss=0.05522, over 4800.00 frames. ], tot_loss[loss=0.201, simple_loss=0.2667, pruned_loss=0.06761, over 957717.38 frames. ], batch size: 51, lr: 3.84e-03, grad_scale: 32.0 2023-04-26 20:56:21,980 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9789, 2.4684, 1.0474, 1.1826, 1.9280, 1.0773, 3.3774, 1.5365], device='cuda:6'), covar=tensor([0.0780, 0.0761, 0.0871, 0.1589, 0.0603, 0.1253, 0.0347, 0.0825], device='cuda:6'), in_proj_covar=tensor([0.0052, 0.0068, 0.0050, 0.0048, 0.0052, 0.0053, 0.0080, 0.0052], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0011, 0.0008], device='cuda:6') 2023-04-26 20:56:27,975 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=42009.0, num_to_drop=1, layers_to_drop={3} 2023-04-26 20:56:28,520 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=42010.0, num_to_drop=1, layers_to_drop={3} 2023-04-26 20:56:49,877 INFO [optim.py:369] (6/7) 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 20:57:08,689 INFO [finetune.py:976] (6/7) Epoch 8, batch 1950, loss[loss=0.1568, simple_loss=0.2297, pruned_loss=0.04197, over 4801.00 frames. ], tot_loss[loss=0.1988, simple_loss=0.2651, pruned_loss=0.06624, over 958793.55 frames. ], batch size: 29, lr: 3.84e-03, grad_scale: 32.0 2023-04-26 20:57:10,617 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0248, 1.7403, 1.9969, 2.4834, 2.3471, 1.9780, 1.7552, 2.0560], device='cuda:6'), covar=tensor([0.0862, 0.1195, 0.0656, 0.0552, 0.0592, 0.0868, 0.0786, 0.0661], device='cuda:6'), in_proj_covar=tensor([0.0200, 0.0207, 0.0183, 0.0180, 0.0180, 0.0193, 0.0162, 0.0189], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-26 20:58:13,787 INFO [finetune.py:976] (6/7) Epoch 8, batch 2000, loss[loss=0.1825, simple_loss=0.241, pruned_loss=0.06203, over 4913.00 frames. ], tot_loss[loss=0.1966, simple_loss=0.2618, pruned_loss=0.06565, over 959696.65 frames. ], batch size: 36, lr: 3.84e-03, grad_scale: 32.0 2023-04-26 20:59:06,779 INFO [optim.py:369] (6/7) 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 20:59:17,304 INFO [finetune.py:976] (6/7) Epoch 8, batch 2050, loss[loss=0.134, simple_loss=0.2166, pruned_loss=0.02573, over 4755.00 frames. ], tot_loss[loss=0.1939, simple_loss=0.2581, pruned_loss=0.0648, over 959429.78 frames. ], batch size: 26, lr: 3.84e-03, grad_scale: 32.0 2023-04-26 20:59:28,157 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9847, 1.3128, 4.9188, 4.5612, 4.2457, 4.6813, 4.3732, 4.3147], device='cuda:6'), covar=tensor([0.7117, 0.6452, 0.1170, 0.2044, 0.1183, 0.1770, 0.1374, 0.1364], device='cuda:6'), in_proj_covar=tensor([0.0312, 0.0308, 0.0409, 0.0415, 0.0351, 0.0408, 0.0318, 0.0370], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-26 21:00:12,508 INFO [finetune.py:976] (6/7) Epoch 8, batch 2100, loss[loss=0.2304, simple_loss=0.3031, pruned_loss=0.07881, over 4808.00 frames. ], tot_loss[loss=0.1933, simple_loss=0.2574, pruned_loss=0.06457, over 957918.55 frames. ], batch size: 41, lr: 3.83e-03, grad_scale: 32.0 2023-04-26 21:00:24,494 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.48 vs. limit=2.0 2023-04-26 21:00:27,906 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.3284, 1.5622, 1.6884, 1.8307, 1.6457, 1.8270, 1.8062, 1.7547], device='cuda:6'), covar=tensor([0.5922, 0.8036, 0.6504, 0.5822, 0.7409, 1.0413, 0.7532, 0.6658], device='cuda:6'), in_proj_covar=tensor([0.0321, 0.0385, 0.0316, 0.0327, 0.0341, 0.0405, 0.0368, 0.0328], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-26 21:00:38,850 INFO [optim.py:369] (6/7) 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 21:00:41,473 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8739, 1.1117, 1.4144, 1.5275, 1.4338, 1.6508, 1.4553, 1.4071], device='cuda:6'), covar=tensor([0.4648, 0.5883, 0.5580, 0.4894, 0.6207, 0.8806, 0.5752, 0.6016], device='cuda:6'), in_proj_covar=tensor([0.0322, 0.0385, 0.0316, 0.0327, 0.0342, 0.0405, 0.0368, 0.0329], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-26 21:00:46,719 INFO [finetune.py:976] (6/7) Epoch 8, batch 2150, loss[loss=0.1585, simple_loss=0.2355, pruned_loss=0.04077, over 4792.00 frames. ], tot_loss[loss=0.1988, simple_loss=0.263, pruned_loss=0.0673, over 953824.76 frames. ], batch size: 29, lr: 3.83e-03, grad_scale: 32.0 2023-04-26 21:00:58,115 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=42262.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 21:01:20,323 INFO [finetune.py:976] (6/7) Epoch 8, batch 2200, loss[loss=0.1506, simple_loss=0.2309, pruned_loss=0.03518, over 4792.00 frames. ], tot_loss[loss=0.2008, simple_loss=0.2656, pruned_loss=0.06802, over 955294.41 frames. ], batch size: 29, lr: 3.83e-03, grad_scale: 32.0 2023-04-26 21:01:26,906 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=42304.0, num_to_drop=1, layers_to_drop={2} 2023-04-26 21:01:29,107 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-04-26 21:01:30,518 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=42310.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 21:01:44,809 INFO [optim.py:369] (6/7) 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 21:01:52,473 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6445, 1.6944, 0.8316, 1.2907, 1.8171, 1.5307, 1.4272, 1.4815], device='cuda:6'), covar=tensor([0.0540, 0.0385, 0.0371, 0.0593, 0.0275, 0.0555, 0.0522, 0.0604], device='cuda:6'), in_proj_covar=tensor([0.0030, 0.0026, 0.0024, 0.0030, 0.0021, 0.0030, 0.0029, 0.0030], device='cuda:6'), out_proj_covar=tensor([0.0049, 0.0044, 0.0038, 0.0050, 0.0038, 0.0048, 0.0048, 0.0050], device='cuda:6') 2023-04-26 21:01:52,974 INFO [finetune.py:976] (6/7) Epoch 8, batch 2250, loss[loss=0.1973, simple_loss=0.2696, pruned_loss=0.06245, over 4866.00 frames. ], tot_loss[loss=0.2026, simple_loss=0.2673, pruned_loss=0.069, over 954660.27 frames. ], batch size: 34, lr: 3.83e-03, grad_scale: 32.0 2023-04-26 21:01:55,424 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7853, 1.7006, 1.9756, 2.1490, 1.7646, 1.4074, 1.7579, 1.0428], device='cuda:6'), covar=tensor([0.0877, 0.0889, 0.0633, 0.0924, 0.0863, 0.1224, 0.0920, 0.1048], device='cuda:6'), in_proj_covar=tensor([0.0066, 0.0073, 0.0072, 0.0067, 0.0076, 0.0096, 0.0080, 0.0075], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-26 21:02:02,515 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=42358.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 21:02:11,125 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.0999, 2.0387, 1.7633, 1.8198, 2.2621, 1.7441, 2.7007, 1.6284], device='cuda:6'), covar=tensor([0.4606, 0.2312, 0.5332, 0.3592, 0.2013, 0.3034, 0.1771, 0.4915], device='cuda:6'), in_proj_covar=tensor([0.0347, 0.0352, 0.0434, 0.0363, 0.0389, 0.0384, 0.0384, 0.0422], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-26 21:02:15,805 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=42379.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 21:02:24,545 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-04-26 21:02:26,180 INFO [finetune.py:976] (6/7) Epoch 8, batch 2300, loss[loss=0.1726, simple_loss=0.2475, pruned_loss=0.0489, over 4919.00 frames. ], tot_loss[loss=0.2007, simple_loss=0.2662, pruned_loss=0.06762, over 953537.10 frames. ], batch size: 38, lr: 3.83e-03, grad_scale: 32.0 2023-04-26 21:02:50,915 INFO [optim.py:369] (6/7) 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 21:02:51,770 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.77 vs. limit=2.0 2023-04-26 21:02:56,814 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=42440.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 21:02:58,875 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-04-26 21:02:59,687 INFO [finetune.py:976] (6/7) Epoch 8, batch 2350, loss[loss=0.2164, simple_loss=0.2725, pruned_loss=0.08013, over 4815.00 frames. ], tot_loss[loss=0.1977, simple_loss=0.2628, pruned_loss=0.06628, over 952946.49 frames. ], batch size: 40, lr: 3.83e-03, grad_scale: 32.0 2023-04-26 21:03:07,739 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5901, 1.1164, 1.6169, 1.9702, 1.6653, 1.5184, 1.6133, 1.6048], device='cuda:6'), covar=tensor([0.7189, 0.9537, 0.9317, 1.0536, 0.8699, 1.2226, 1.1807, 1.0382], device='cuda:6'), in_proj_covar=tensor([0.0407, 0.0424, 0.0506, 0.0529, 0.0437, 0.0456, 0.0468, 0.0466], device='cuda:6'), out_proj_covar=tensor([9.9000e-05, 1.0518e-04, 1.1445e-04, 1.2544e-04, 1.0629e-04, 1.1029e-04, 1.1270e-04, 1.1307e-04], device='cuda:6') 2023-04-26 21:03:20,564 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-04-26 21:03:48,471 INFO [finetune.py:976] (6/7) Epoch 8, batch 2400, loss[loss=0.1354, simple_loss=0.2089, pruned_loss=0.03093, over 4763.00 frames. ], tot_loss[loss=0.1945, simple_loss=0.2591, pruned_loss=0.06498, over 949465.89 frames. ], batch size: 28, lr: 3.83e-03, grad_scale: 32.0 2023-04-26 21:04:08,561 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=42506.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 21:04:40,671 INFO [optim.py:369] (6/7) 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 21:04:42,591 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.35 vs. limit=5.0 2023-04-26 21:04:43,771 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.4539, 1.0206, 0.3412, 1.1312, 1.0692, 1.3563, 1.2414, 1.2121], device='cuda:6'), covar=tensor([0.0568, 0.0410, 0.0455, 0.0583, 0.0307, 0.0524, 0.0500, 0.0609], device='cuda:6'), in_proj_covar=tensor([0.0030, 0.0026, 0.0024, 0.0030, 0.0021, 0.0030, 0.0029, 0.0030], device='cuda:6'), out_proj_covar=tensor([0.0049, 0.0044, 0.0038, 0.0050, 0.0038, 0.0049, 0.0048, 0.0050], device='cuda:6') 2023-04-26 21:04:54,585 INFO [finetune.py:976] (6/7) Epoch 8, batch 2450, loss[loss=0.2335, simple_loss=0.3039, pruned_loss=0.08154, over 4739.00 frames. ], tot_loss[loss=0.1926, simple_loss=0.2565, pruned_loss=0.06431, over 949374.27 frames. ], batch size: 59, lr: 3.83e-03, grad_scale: 32.0 2023-04-26 21:05:24,366 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=42562.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 21:05:27,957 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=42567.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 21:06:00,482 INFO [finetune.py:976] (6/7) Epoch 8, batch 2500, loss[loss=0.2928, simple_loss=0.3552, pruned_loss=0.1152, over 4855.00 frames. ], tot_loss[loss=0.1933, simple_loss=0.2573, pruned_loss=0.06468, over 951097.31 frames. ], batch size: 44, lr: 3.83e-03, grad_scale: 32.0 2023-04-26 21:06:20,313 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=42604.0, num_to_drop=1, layers_to_drop={2} 2023-04-26 21:06:22,066 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6448, 2.3511, 1.9183, 2.2318, 2.4065, 1.9981, 2.8790, 1.7548], device='cuda:6'), covar=tensor([0.3797, 0.1589, 0.4439, 0.2986, 0.1709, 0.2486, 0.1936, 0.4404], device='cuda:6'), in_proj_covar=tensor([0.0352, 0.0355, 0.0438, 0.0369, 0.0393, 0.0388, 0.0387, 0.0425], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-26 21:06:29,257 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=42610.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 21:06:54,509 INFO [optim.py:369] (6/7) 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 21:07:06,465 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.1767, 2.5202, 0.8483, 1.4464, 1.5171, 1.8237, 1.6078, 0.8223], device='cuda:6'), covar=tensor([0.1337, 0.1203, 0.1654, 0.1347, 0.1093, 0.0957, 0.1413, 0.1657], device='cuda:6'), in_proj_covar=tensor([0.0117, 0.0253, 0.0142, 0.0124, 0.0136, 0.0156, 0.0121, 0.0122], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-26 21:07:08,190 INFO [finetune.py:976] (6/7) Epoch 8, batch 2550, loss[loss=0.2021, simple_loss=0.2683, pruned_loss=0.068, over 4904.00 frames. ], tot_loss[loss=0.1975, simple_loss=0.2621, pruned_loss=0.06651, over 950153.61 frames. ], batch size: 43, lr: 3.83e-03, grad_scale: 32.0 2023-04-26 21:07:11,911 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=42649.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 21:07:14,684 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=42652.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 21:07:41,702 INFO [finetune.py:976] (6/7) Epoch 8, batch 2600, loss[loss=0.1938, simple_loss=0.2726, pruned_loss=0.05755, over 4901.00 frames. ], tot_loss[loss=0.2, simple_loss=0.2649, pruned_loss=0.06757, over 949706.81 frames. ], batch size: 43, lr: 3.83e-03, grad_scale: 32.0 2023-04-26 21:07:46,939 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.84 vs. limit=5.0 2023-04-26 21:07:53,064 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=42710.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 21:08:07,495 INFO [optim.py:369] (6/7) 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 21:08:09,420 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 8, batch 2650, loss[loss=0.2184, simple_loss=0.2603, pruned_loss=0.08829, over 3992.00 frames. ], tot_loss[loss=0.2014, simple_loss=0.2668, pruned_loss=0.06801, over 951563.06 frames. ], batch size: 17, lr: 3.83e-03, grad_scale: 32.0 2023-04-26 21:08:15,535 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9890, 1.9395, 2.2516, 2.5337, 1.8328, 1.5486, 1.9707, 1.1387], device='cuda:6'), covar=tensor([0.0728, 0.0822, 0.0724, 0.0746, 0.0891, 0.1407, 0.0884, 0.1185], device='cuda:6'), in_proj_covar=tensor([0.0066, 0.0073, 0.0072, 0.0067, 0.0076, 0.0095, 0.0079, 0.0074], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-26 21:08:34,456 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.53 vs. limit=2.0 2023-04-26 21:08:43,599 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.3202, 1.6808, 1.7120, 1.8781, 1.7531, 1.9438, 1.8169, 1.7810], device='cuda:6'), covar=tensor([0.5595, 0.6669, 0.6349, 0.5165, 0.6560, 0.8827, 0.6705, 0.6305], device='cuda:6'), in_proj_covar=tensor([0.0322, 0.0384, 0.0316, 0.0326, 0.0342, 0.0405, 0.0367, 0.0328], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-26 21:08:48,362 INFO [finetune.py:976] (6/7) Epoch 8, batch 2700, loss[loss=0.2097, simple_loss=0.2654, pruned_loss=0.07701, over 4752.00 frames. ], tot_loss[loss=0.1997, simple_loss=0.2653, pruned_loss=0.067, over 952112.69 frames. ], batch size: 54, lr: 3.83e-03, grad_scale: 32.0 2023-04-26 21:09:04,036 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7158, 1.1822, 1.6974, 2.1031, 1.8079, 1.6246, 1.6644, 1.6923], device='cuda:6'), covar=tensor([0.7006, 0.9826, 0.9386, 1.0006, 0.8478, 1.1147, 1.2157, 1.1422], device='cuda:6'), in_proj_covar=tensor([0.0409, 0.0428, 0.0509, 0.0532, 0.0439, 0.0460, 0.0470, 0.0469], device='cuda:6'), out_proj_covar=tensor([9.9527e-05, 1.0610e-04, 1.1501e-04, 1.2623e-04, 1.0679e-04, 1.1109e-04, 1.1327e-04, 1.1369e-04], device='cuda:6') 2023-04-26 21:09:06,249 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.1300, 1.6425, 1.3693, 1.8058, 1.6242, 1.8955, 1.4485, 3.4563], device='cuda:6'), covar=tensor([0.0628, 0.0727, 0.0767, 0.1120, 0.0662, 0.0493, 0.0704, 0.0133], device='cuda:6'), in_proj_covar=tensor([0.0038, 0.0039, 0.0040, 0.0044, 0.0040, 0.0039, 0.0039, 0.0059], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0014, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:6') 2023-04-26 21:09:12,231 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([3.6810, 3.5320, 2.6315, 4.2098, 3.6515, 3.6275, 1.7216, 3.6077], device='cuda:6'), covar=tensor([0.1784, 0.1347, 0.3268, 0.1955, 0.2551, 0.1980, 0.5324, 0.2409], device='cuda:6'), in_proj_covar=tensor([0.0244, 0.0217, 0.0249, 0.0308, 0.0299, 0.0250, 0.0272, 0.0270], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-26 21:09:14,553 INFO [optim.py:369] (6/7) 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 21:09:16,606 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-04-26 21:09:21,928 INFO [finetune.py:976] (6/7) Epoch 8, batch 2750, loss[loss=0.1855, simple_loss=0.2472, pruned_loss=0.06193, over 4853.00 frames. ], tot_loss[loss=0.1986, simple_loss=0.2633, pruned_loss=0.06694, over 954019.19 frames. ], batch size: 47, lr: 3.83e-03, grad_scale: 32.0 2023-04-26 21:09:32,066 INFO [zipformer.py:1188] (6/7) 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] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=42862.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 21:09:57,305 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=42879.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 21:10:17,378 INFO [finetune.py:976] (6/7) Epoch 8, batch 2800, loss[loss=0.1865, simple_loss=0.247, pruned_loss=0.06298, over 4815.00 frames. ], tot_loss[loss=0.1956, simple_loss=0.2595, pruned_loss=0.06586, over 951691.46 frames. ], batch size: 39, lr: 3.83e-03, grad_scale: 32.0 2023-04-26 21:10:51,628 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=42920.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 21:11:02,372 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.81 vs. limit=2.0 2023-04-26 21:11:10,643 INFO [optim.py:369] (6/7) 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 21:11:16,208 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=42940.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 21:11:18,536 INFO [finetune.py:976] (6/7) Epoch 8, batch 2850, loss[loss=0.1395, simple_loss=0.1991, pruned_loss=0.03996, over 3939.00 frames. ], tot_loss[loss=0.194, simple_loss=0.2576, pruned_loss=0.06519, over 952616.79 frames. ], batch size: 17, lr: 3.83e-03, grad_scale: 32.0 2023-04-26 21:11:30,127 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5575, 1.2549, 0.7147, 1.2725, 1.4487, 1.4704, 1.3600, 1.3948], device='cuda:6'), covar=tensor([0.0568, 0.0433, 0.0429, 0.0604, 0.0311, 0.0564, 0.0569, 0.0619], device='cuda:6'), in_proj_covar=tensor([0.0030, 0.0026, 0.0024, 0.0030, 0.0021, 0.0030, 0.0029, 0.0030], device='cuda:6'), out_proj_covar=tensor([0.0049, 0.0044, 0.0038, 0.0050, 0.0038, 0.0049, 0.0048, 0.0050], device='cuda:6') 2023-04-26 21:11:56,701 INFO [finetune.py:976] (6/7) Epoch 8, batch 2900, loss[loss=0.2377, simple_loss=0.3116, pruned_loss=0.08188, over 4813.00 frames. ], tot_loss[loss=0.1974, simple_loss=0.2614, pruned_loss=0.06672, over 950875.70 frames. ], batch size: 45, lr: 3.83e-03, grad_scale: 32.0 2023-04-26 21:12:15,327 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=43005.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 21:12:24,777 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5419, 1.4768, 1.8592, 1.8011, 1.4087, 1.2289, 1.5078, 1.1186], device='cuda:6'), covar=tensor([0.0716, 0.0931, 0.0507, 0.0706, 0.0895, 0.1306, 0.0816, 0.0834], device='cuda:6'), in_proj_covar=tensor([0.0065, 0.0073, 0.0072, 0.0067, 0.0076, 0.0095, 0.0079, 0.0074], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-26 21:12:51,490 INFO [optim.py:369] (6/7) 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] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=43035.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 21:13:10,706 INFO [finetune.py:976] (6/7) Epoch 8, batch 2950, loss[loss=0.2189, simple_loss=0.2722, pruned_loss=0.08282, over 4839.00 frames. ], tot_loss[loss=0.1997, simple_loss=0.264, pruned_loss=0.06766, over 951731.09 frames. ], batch size: 49, lr: 3.83e-03, grad_scale: 32.0 2023-04-26 21:13:35,143 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.2487, 1.5713, 1.5215, 1.8746, 1.6767, 1.8733, 1.4581, 3.5069], device='cuda:6'), covar=tensor([0.0618, 0.0758, 0.0742, 0.1141, 0.0620, 0.0509, 0.0742, 0.0168], device='cuda:6'), in_proj_covar=tensor([0.0039, 0.0039, 0.0040, 0.0044, 0.0040, 0.0039, 0.0039, 0.0059], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0014, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:6') 2023-04-26 21:13:37,394 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=43083.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 21:13:40,524 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.1355, 2.8200, 2.4920, 2.5785, 2.0277, 2.3920, 2.5189, 2.0062], device='cuda:6'), covar=tensor([0.2269, 0.1105, 0.0674, 0.1153, 0.3275, 0.1194, 0.1942, 0.2640], device='cuda:6'), in_proj_covar=tensor([0.0302, 0.0322, 0.0233, 0.0296, 0.0320, 0.0276, 0.0262, 0.0287], device='cuda:6'), out_proj_covar=tensor([1.2259e-04, 1.3021e-04, 9.4235e-05, 1.1843e-04, 1.3162e-04, 1.1156e-04, 1.0743e-04, 1.1539e-04], device='cuda:6') 2023-04-26 21:13:44,490 INFO [finetune.py:976] (6/7) Epoch 8, batch 3000, loss[loss=0.2143, simple_loss=0.2623, pruned_loss=0.08317, over 3920.00 frames. ], tot_loss[loss=0.1989, simple_loss=0.2636, pruned_loss=0.06706, over 949229.70 frames. ], batch size: 17, lr: 3.83e-03, grad_scale: 32.0 2023-04-26 21:13:44,490 INFO [finetune.py:1001] (6/7) Computing validation loss 2023-04-26 21:13:49,753 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.3151, 1.3472, 3.8660, 3.5722, 3.4623, 3.7511, 3.8174, 3.3811], device='cuda:6'), covar=tensor([0.7130, 0.5099, 0.1123, 0.1764, 0.1172, 0.1361, 0.0693, 0.1502], device='cuda:6'), in_proj_covar=tensor([0.0311, 0.0309, 0.0411, 0.0415, 0.0351, 0.0407, 0.0318, 0.0372], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-26 21:13:50,775 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7700, 2.1223, 1.8587, 2.0258, 1.6565, 1.7393, 1.7956, 1.4754], device='cuda:6'), covar=tensor([0.2012, 0.1302, 0.0890, 0.1304, 0.3651, 0.1361, 0.1899, 0.2671], device='cuda:6'), in_proj_covar=tensor([0.0302, 0.0322, 0.0233, 0.0296, 0.0320, 0.0276, 0.0262, 0.0287], device='cuda:6'), out_proj_covar=tensor([1.2254e-04, 1.3010e-04, 9.4147e-05, 1.1835e-04, 1.3156e-04, 1.1149e-04, 1.0732e-04, 1.1528e-04], device='cuda:6') 2023-04-26 21:13:54,962 INFO [finetune.py:1010] (6/7) Epoch 8, validation: loss=0.1551, simple_loss=0.2273, pruned_loss=0.04149, over 2265189.00 frames. 2023-04-26 21:13:54,962 INFO [finetune.py:1011] (6/7) Maximum memory allocated so far is 6345MB 2023-04-26 21:14:16,869 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.3154, 2.4213, 2.0119, 2.1962, 2.5513, 1.9813, 3.3817, 1.7658], device='cuda:6'), covar=tensor([0.4533, 0.2355, 0.5086, 0.3514, 0.2147, 0.3276, 0.2012, 0.5001], device='cuda:6'), in_proj_covar=tensor([0.0349, 0.0350, 0.0433, 0.0364, 0.0390, 0.0385, 0.0385, 0.0421], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-26 21:14:18,552 INFO [optim.py:369] (6/7) 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 21:14:27,799 INFO [finetune.py:976] (6/7) Epoch 8, batch 3050, loss[loss=0.2114, simple_loss=0.2715, pruned_loss=0.07565, over 4690.00 frames. ], tot_loss[loss=0.1987, simple_loss=0.2639, pruned_loss=0.06678, over 950906.15 frames. ], batch size: 23, lr: 3.83e-03, grad_scale: 32.0 2023-04-26 21:14:39,672 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=43162.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 21:14:46,250 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.73 vs. limit=5.0 2023-04-26 21:14:58,146 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([3.4910, 3.3098, 2.5037, 4.0412, 3.4412, 3.4488, 1.5275, 3.4685], device='cuda:6'), covar=tensor([0.1914, 0.1494, 0.3687, 0.2055, 0.3036, 0.2131, 0.5706, 0.2630], device='cuda:6'), in_proj_covar=tensor([0.0242, 0.0215, 0.0248, 0.0307, 0.0297, 0.0248, 0.0269, 0.0270], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-26 21:15:00,986 INFO [finetune.py:976] (6/7) Epoch 8, batch 3100, loss[loss=0.1705, simple_loss=0.234, pruned_loss=0.05347, over 4735.00 frames. ], tot_loss[loss=0.1979, simple_loss=0.2626, pruned_loss=0.06658, over 951359.09 frames. ], batch size: 23, lr: 3.83e-03, grad_scale: 32.0 2023-04-26 21:15:11,988 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=43210.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 21:15:15,105 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=43215.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 21:15:25,431 INFO [optim.py:369] (6/7) 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 21:15:27,886 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 8, batch 3150, loss[loss=0.1898, simple_loss=0.2622, pruned_loss=0.05871, over 4868.00 frames. ], tot_loss[loss=0.1952, simple_loss=0.2595, pruned_loss=0.06539, over 950112.92 frames. ], batch size: 34, lr: 3.83e-03, grad_scale: 32.0 2023-04-26 21:16:05,533 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-04-26 21:16:18,043 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-04-26 21:16:18,113 INFO [finetune.py:976] (6/7) Epoch 8, batch 3200, loss[loss=0.1759, simple_loss=0.2479, pruned_loss=0.05194, over 4789.00 frames. ], tot_loss[loss=0.1923, simple_loss=0.2563, pruned_loss=0.06415, over 949563.96 frames. ], batch size: 29, lr: 3.83e-03, grad_scale: 32.0 2023-04-26 21:16:18,260 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.1727, 2.7226, 2.3159, 2.1871, 1.7369, 1.8514, 2.4468, 1.7674], device='cuda:6'), covar=tensor([0.1667, 0.1582, 0.1352, 0.1787, 0.2314, 0.1945, 0.0941, 0.1923], device='cuda:6'), in_proj_covar=tensor([0.0196, 0.0220, 0.0173, 0.0208, 0.0208, 0.0186, 0.0164, 0.0191], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:6') 2023-04-26 21:16:30,934 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=43305.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 21:17:03,561 INFO [optim.py:369] (6/7) 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 21:17:20,612 INFO [finetune.py:976] (6/7) Epoch 8, batch 3250, loss[loss=0.2257, simple_loss=0.2849, pruned_loss=0.08329, over 4745.00 frames. ], tot_loss[loss=0.1958, simple_loss=0.2589, pruned_loss=0.06637, over 948543.56 frames. ], batch size: 54, lr: 3.83e-03, grad_scale: 32.0 2023-04-26 21:17:33,416 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=43353.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 21:17:36,442 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7412, 1.3102, 1.2473, 1.6061, 1.9867, 1.6568, 1.4361, 1.2050], device='cuda:6'), covar=tensor([0.1576, 0.1600, 0.1839, 0.1313, 0.0949, 0.1390, 0.1886, 0.1927], device='cuda:6'), in_proj_covar=tensor([0.0312, 0.0328, 0.0357, 0.0305, 0.0344, 0.0329, 0.0313, 0.0358], device='cuda:6'), out_proj_covar=tensor([6.6148e-05, 6.9721e-05, 7.7235e-05, 6.3253e-05, 7.2488e-05, 7.0747e-05, 6.7542e-05, 7.6867e-05], device='cuda:6') 2023-04-26 21:18:26,603 INFO [finetune.py:976] (6/7) Epoch 8, batch 3300, loss[loss=0.1568, simple_loss=0.2342, pruned_loss=0.03971, over 4891.00 frames. ], tot_loss[loss=0.1966, simple_loss=0.2608, pruned_loss=0.06623, over 948053.39 frames. ], batch size: 32, lr: 3.83e-03, grad_scale: 32.0 2023-04-26 21:18:59,194 INFO [optim.py:369] (6/7) 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 21:19:04,195 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.2033, 1.6121, 1.4219, 1.7760, 1.6798, 1.9389, 1.4221, 3.8041], device='cuda:6'), covar=tensor([0.0644, 0.0732, 0.0806, 0.1175, 0.0650, 0.0486, 0.0760, 0.0142], device='cuda:6'), in_proj_covar=tensor([0.0039, 0.0039, 0.0040, 0.0044, 0.0041, 0.0040, 0.0040, 0.0059], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:6') 2023-04-26 21:19:06,531 INFO [finetune.py:976] (6/7) Epoch 8, batch 3350, loss[loss=0.2337, simple_loss=0.3058, pruned_loss=0.08079, over 4810.00 frames. ], tot_loss[loss=0.1987, simple_loss=0.2634, pruned_loss=0.06694, over 949741.04 frames. ], batch size: 45, lr: 3.83e-03, grad_scale: 32.0 2023-04-26 21:19:10,785 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6564, 2.5385, 1.6027, 1.6496, 1.1641, 1.2270, 1.7237, 1.2276], device='cuda:6'), covar=tensor([0.2001, 0.1501, 0.1895, 0.2247, 0.3068, 0.2530, 0.1270, 0.2397], device='cuda:6'), in_proj_covar=tensor([0.0195, 0.0219, 0.0173, 0.0207, 0.0207, 0.0185, 0.0164, 0.0190], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:6') 2023-04-26 21:19:35,011 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6152, 1.4132, 1.7392, 1.8587, 1.4840, 1.2010, 1.3751, 0.9475], device='cuda:6'), covar=tensor([0.0584, 0.0720, 0.0498, 0.0570, 0.0845, 0.1687, 0.0815, 0.0908], device='cuda:6'), in_proj_covar=tensor([0.0066, 0.0074, 0.0072, 0.0067, 0.0077, 0.0096, 0.0079, 0.0075], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-26 21:19:38,049 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6030, 1.2951, 4.2259, 3.9272, 3.6877, 3.9283, 3.8202, 3.7306], device='cuda:6'), covar=tensor([0.6956, 0.6073, 0.0950, 0.1646, 0.1184, 0.1609, 0.2668, 0.1305], device='cuda:6'), in_proj_covar=tensor([0.0311, 0.0309, 0.0411, 0.0414, 0.0353, 0.0406, 0.0319, 0.0372], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-26 21:19:40,385 INFO [finetune.py:976] (6/7) Epoch 8, batch 3400, loss[loss=0.2257, simple_loss=0.2811, pruned_loss=0.08516, over 4136.00 frames. ], tot_loss[loss=0.1994, simple_loss=0.2641, pruned_loss=0.06734, over 951704.75 frames. ], batch size: 18, lr: 3.83e-03, grad_scale: 32.0 2023-04-26 21:19:42,957 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([0.3531, 1.3436, 1.3979, 1.1082, 1.4250, 1.0950, 1.8304, 1.3434], device='cuda:6'), covar=tensor([0.4009, 0.1834, 0.5297, 0.2681, 0.1508, 0.2420, 0.1630, 0.4916], device='cuda:6'), in_proj_covar=tensor([0.0350, 0.0355, 0.0436, 0.0367, 0.0391, 0.0388, 0.0388, 0.0423], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-26 21:19:49,345 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6860, 1.7625, 0.9074, 1.3436, 1.7067, 1.5495, 1.4448, 1.4582], device='cuda:6'), covar=tensor([0.0514, 0.0388, 0.0402, 0.0595, 0.0314, 0.0578, 0.0545, 0.0615], device='cuda:6'), in_proj_covar=tensor([0.0030, 0.0026, 0.0024, 0.0030, 0.0021, 0.0030, 0.0029, 0.0030], device='cuda:6'), out_proj_covar=tensor([0.0049, 0.0044, 0.0038, 0.0050, 0.0038, 0.0048, 0.0048, 0.0050], device='cuda:6') 2023-04-26 21:19:54,929 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=43514.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 21:19:55,544 INFO [zipformer.py:1188] (6/7) 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] (6/7) 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 21:20:08,561 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=43535.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 21:20:13,907 INFO [finetune.py:976] (6/7) Epoch 8, batch 3450, loss[loss=0.2231, simple_loss=0.2747, pruned_loss=0.08575, over 4823.00 frames. ], tot_loss[loss=0.1992, simple_loss=0.264, pruned_loss=0.06721, over 951814.79 frames. ], batch size: 38, lr: 3.83e-03, grad_scale: 32.0 2023-04-26 21:20:17,039 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([0.3872, 1.3849, 1.4167, 1.0890, 1.3823, 1.1285, 1.8344, 1.2743], device='cuda:6'), covar=tensor([0.3780, 0.1809, 0.5039, 0.2704, 0.1523, 0.2329, 0.1732, 0.4981], device='cuda:6'), in_proj_covar=tensor([0.0349, 0.0353, 0.0433, 0.0366, 0.0389, 0.0386, 0.0386, 0.0421], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-26 21:20:27,395 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=43563.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 21:20:33,770 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.4939, 1.0833, 0.3424, 1.2204, 1.1768, 1.4104, 1.3360, 1.3040], device='cuda:6'), covar=tensor([0.0539, 0.0417, 0.0442, 0.0591, 0.0312, 0.0534, 0.0511, 0.0603], device='cuda:6'), in_proj_covar=tensor([0.0030, 0.0026, 0.0024, 0.0030, 0.0021, 0.0030, 0.0029, 0.0030], device='cuda:6'), out_proj_covar=tensor([0.0049, 0.0044, 0.0038, 0.0050, 0.0038, 0.0048, 0.0048, 0.0050], device='cuda:6') 2023-04-26 21:20:35,649 INFO [zipformer.py:1188] (6/7) 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] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=43583.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 21:20:47,608 INFO [finetune.py:976] (6/7) Epoch 8, batch 3500, loss[loss=0.2624, simple_loss=0.3061, pruned_loss=0.1093, over 4801.00 frames. ], tot_loss[loss=0.1965, simple_loss=0.2612, pruned_loss=0.06591, over 951302.44 frames. ], batch size: 39, lr: 3.83e-03, grad_scale: 32.0 2023-04-26 21:20:49,553 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6479, 1.3848, 1.7365, 1.8270, 1.5001, 1.3224, 1.5033, 1.0910], device='cuda:6'), covar=tensor([0.0489, 0.1047, 0.0517, 0.0815, 0.0799, 0.1229, 0.0656, 0.0732], device='cuda:6'), in_proj_covar=tensor([0.0065, 0.0073, 0.0072, 0.0067, 0.0076, 0.0095, 0.0078, 0.0075], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-26 21:21:13,732 INFO [optim.py:369] (6/7) 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 21:21:21,527 INFO [finetune.py:976] (6/7) Epoch 8, batch 3550, loss[loss=0.1534, simple_loss=0.2214, pruned_loss=0.04268, over 4731.00 frames. ], tot_loss[loss=0.1937, simple_loss=0.2579, pruned_loss=0.0648, over 953769.80 frames. ], batch size: 23, lr: 3.82e-03, grad_scale: 32.0 2023-04-26 21:21:38,679 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([0.9702, 1.9426, 1.7448, 1.6237, 2.0871, 1.6885, 2.4050, 1.5108], device='cuda:6'), covar=tensor([0.3558, 0.1485, 0.3876, 0.2683, 0.1365, 0.2324, 0.1413, 0.4385], device='cuda:6'), in_proj_covar=tensor([0.0344, 0.0349, 0.0428, 0.0362, 0.0385, 0.0382, 0.0381, 0.0417], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-26 21:22:22,682 INFO [finetune.py:976] (6/7) Epoch 8, batch 3600, loss[loss=0.2213, simple_loss=0.2688, pruned_loss=0.08689, over 4804.00 frames. ], tot_loss[loss=0.1922, simple_loss=0.2558, pruned_loss=0.06428, over 954369.20 frames. ], batch size: 51, lr: 3.82e-03, grad_scale: 32.0 2023-04-26 21:23:15,351 INFO [optim.py:369] (6/7) 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 21:23:34,070 INFO [finetune.py:976] (6/7) Epoch 8, batch 3650, loss[loss=0.2006, simple_loss=0.2715, pruned_loss=0.06486, over 4921.00 frames. ], tot_loss[loss=0.1948, simple_loss=0.2585, pruned_loss=0.06551, over 955604.28 frames. ], batch size: 42, lr: 3.82e-03, grad_scale: 32.0 2023-04-26 21:24:29,561 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=43787.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 21:24:32,037 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6030, 1.2592, 1.7543, 2.0833, 1.7868, 1.6247, 1.6572, 1.6637], device='cuda:6'), covar=tensor([0.7363, 0.9706, 0.9484, 0.9906, 0.8618, 1.1175, 1.1438, 1.0633], device='cuda:6'), in_proj_covar=tensor([0.0408, 0.0422, 0.0507, 0.0528, 0.0437, 0.0455, 0.0468, 0.0466], device='cuda:6'), out_proj_covar=tensor([9.9233e-05, 1.0475e-04, 1.1460e-04, 1.2529e-04, 1.0624e-04, 1.1004e-04, 1.1266e-04, 1.1301e-04], device='cuda:6') 2023-04-26 21:24:33,740 INFO [finetune.py:976] (6/7) Epoch 8, batch 3700, loss[loss=0.2148, simple_loss=0.2797, pruned_loss=0.07496, over 4882.00 frames. ], tot_loss[loss=0.198, simple_loss=0.2624, pruned_loss=0.0668, over 954407.50 frames. ], batch size: 32, lr: 3.82e-03, grad_scale: 32.0 2023-04-26 21:24:43,200 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.3433, 2.2384, 2.5058, 3.0412, 2.3735, 2.0258, 2.4183, 1.5715], device='cuda:6'), covar=tensor([0.0683, 0.0878, 0.0529, 0.0534, 0.0687, 0.1076, 0.0726, 0.0941], device='cuda:6'), in_proj_covar=tensor([0.0066, 0.0073, 0.0072, 0.0067, 0.0077, 0.0096, 0.0079, 0.0075], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-26 21:25:03,742 INFO [optim.py:369] (6/7) 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 21:25:11,091 INFO [finetune.py:976] (6/7) Epoch 8, batch 3750, loss[loss=0.1857, simple_loss=0.2611, pruned_loss=0.05517, over 4877.00 frames. ], tot_loss[loss=0.1999, simple_loss=0.2647, pruned_loss=0.06758, over 953285.03 frames. ], batch size: 35, lr: 3.82e-03, grad_scale: 32.0 2023-04-26 21:25:14,034 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=43848.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 21:25:27,345 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=43870.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 21:25:36,632 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5425, 1.0875, 1.6202, 1.9626, 1.6126, 1.4951, 1.5363, 1.5754], device='cuda:6'), covar=tensor([0.7704, 1.0416, 1.0719, 1.0732, 0.8944, 1.2854, 1.3342, 1.1552], device='cuda:6'), in_proj_covar=tensor([0.0409, 0.0424, 0.0509, 0.0531, 0.0439, 0.0457, 0.0470, 0.0468], device='cuda:6'), out_proj_covar=tensor([9.9547e-05, 1.0512e-04, 1.1509e-04, 1.2576e-04, 1.0669e-04, 1.1045e-04, 1.1306e-04, 1.1342e-04], device='cuda:6') 2023-04-26 21:25:44,748 INFO [finetune.py:976] (6/7) Epoch 8, batch 3800, loss[loss=0.1723, simple_loss=0.2379, pruned_loss=0.05328, over 4772.00 frames. ], tot_loss[loss=0.199, simple_loss=0.2645, pruned_loss=0.06669, over 953519.73 frames. ], batch size: 26, lr: 3.82e-03, grad_scale: 32.0 2023-04-26 21:25:54,477 INFO [zipformer.py:1188] (6/7) 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] (6/7) 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 21:26:18,576 INFO [finetune.py:976] (6/7) Epoch 8, batch 3850, loss[loss=0.1849, simple_loss=0.2531, pruned_loss=0.05836, over 4796.00 frames. ], tot_loss[loss=0.1971, simple_loss=0.2629, pruned_loss=0.06562, over 954041.20 frames. ], batch size: 29, lr: 3.82e-03, grad_scale: 32.0 2023-04-26 21:26:28,339 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=43959.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 21:26:35,133 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=43970.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 21:27:02,341 INFO [finetune.py:976] (6/7) Epoch 8, batch 3900, loss[loss=0.2397, simple_loss=0.29, pruned_loss=0.09465, over 4818.00 frames. ], tot_loss[loss=0.1955, simple_loss=0.2605, pruned_loss=0.06522, over 955319.64 frames. ], batch size: 39, lr: 3.82e-03, grad_scale: 64.0 2023-04-26 21:27:28,228 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.53 vs. limit=2.0 2023-04-26 21:27:35,512 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=44020.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 21:27:48,407 INFO [optim.py:369] (6/7) 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 21:28:08,678 INFO [finetune.py:976] (6/7) Epoch 8, batch 3950, loss[loss=0.1949, simple_loss=0.2568, pruned_loss=0.06653, over 4820.00 frames. ], tot_loss[loss=0.1925, simple_loss=0.257, pruned_loss=0.06393, over 957328.13 frames. ], batch size: 41, lr: 3.82e-03, grad_scale: 64.0 2023-04-26 21:28:42,923 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9932, 1.4598, 1.5727, 1.6261, 2.1681, 1.8224, 1.4977, 1.4589], device='cuda:6'), covar=tensor([0.1438, 0.1615, 0.1977, 0.1377, 0.0780, 0.1464, 0.2096, 0.2068], device='cuda:6'), in_proj_covar=tensor([0.0306, 0.0321, 0.0350, 0.0300, 0.0335, 0.0322, 0.0308, 0.0351], device='cuda:6'), out_proj_covar=tensor([6.4882e-05, 6.8114e-05, 7.5765e-05, 6.2152e-05, 7.0180e-05, 6.9179e-05, 6.6306e-05, 7.5464e-05], device='cuda:6') 2023-04-26 21:28:58,847 INFO [finetune.py:976] (6/7) Epoch 8, batch 4000, loss[loss=0.1799, simple_loss=0.2533, pruned_loss=0.05324, over 4834.00 frames. ], tot_loss[loss=0.1942, simple_loss=0.2584, pruned_loss=0.06499, over 955070.49 frames. ], batch size: 33, lr: 3.82e-03, grad_scale: 64.0 2023-04-26 21:29:37,623 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.7672, 1.9129, 1.8235, 2.0792, 1.8439, 2.0378, 1.9305, 1.8617], device='cuda:6'), covar=tensor([0.5780, 0.9724, 0.7360, 0.6130, 0.8368, 1.0778, 0.9702, 0.8312], device='cuda:6'), in_proj_covar=tensor([0.0324, 0.0388, 0.0318, 0.0328, 0.0343, 0.0407, 0.0369, 0.0329], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-26 21:29:49,240 INFO [optim.py:369] (6/7) 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 21:30:02,130 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 8, batch 4050, loss[loss=0.2343, simple_loss=0.3077, pruned_loss=0.08044, over 4816.00 frames. ], tot_loss[loss=0.1999, simple_loss=0.2644, pruned_loss=0.06769, over 953474.03 frames. ], batch size: 38, lr: 3.82e-03, grad_scale: 64.0 2023-04-26 21:30:35,464 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=44170.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 21:30:45,646 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=3.67 vs. limit=5.0 2023-04-26 21:31:07,890 INFO [finetune.py:976] (6/7) Epoch 8, batch 4100, loss[loss=0.1934, simple_loss=0.2607, pruned_loss=0.06308, over 4900.00 frames. ], tot_loss[loss=0.2004, simple_loss=0.2658, pruned_loss=0.06747, over 953216.69 frames. ], batch size: 37, lr: 3.82e-03, grad_scale: 64.0 2023-04-26 21:31:26,570 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.23 vs. limit=5.0 2023-04-26 21:31:41,630 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=44218.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 21:32:00,831 INFO [optim.py:369] (6/7) 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 21:32:13,839 INFO [finetune.py:976] (6/7) Epoch 8, batch 4150, loss[loss=0.1869, simple_loss=0.2562, pruned_loss=0.05878, over 4812.00 frames. ], tot_loss[loss=0.2011, simple_loss=0.2667, pruned_loss=0.06775, over 952434.22 frames. ], batch size: 33, lr: 3.82e-03, grad_scale: 64.0 2023-04-26 21:32:33,491 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([0.6370, 1.7531, 1.6168, 1.3667, 1.8374, 1.4716, 2.3311, 1.4029], device='cuda:6'), covar=tensor([0.4183, 0.1807, 0.5303, 0.2837, 0.1597, 0.2437, 0.1425, 0.5147], device='cuda:6'), in_proj_covar=tensor([0.0345, 0.0349, 0.0429, 0.0363, 0.0385, 0.0381, 0.0381, 0.0419], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-26 21:32:34,619 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=44265.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 21:32:40,622 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7760, 1.6780, 1.9360, 2.1342, 2.3263, 1.7376, 1.4453, 1.8543], device='cuda:6'), covar=tensor([0.0911, 0.1154, 0.0694, 0.0668, 0.0501, 0.0912, 0.0905, 0.0682], device='cuda:6'), in_proj_covar=tensor([0.0196, 0.0202, 0.0179, 0.0175, 0.0175, 0.0188, 0.0160, 0.0184], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-26 21:32:52,669 INFO [finetune.py:976] (6/7) Epoch 8, batch 4200, loss[loss=0.1885, simple_loss=0.2557, pruned_loss=0.06066, over 4855.00 frames. ], tot_loss[loss=0.2005, simple_loss=0.2665, pruned_loss=0.0672, over 953587.67 frames. ], batch size: 49, lr: 3.82e-03, grad_scale: 64.0 2023-04-26 21:33:08,265 INFO [zipformer.py:1188] (6/7) 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] (6/7) 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] (6/7) Epoch 8, batch 4250, loss[loss=0.2183, simple_loss=0.2789, pruned_loss=0.0788, over 4822.00 frames. ], tot_loss[loss=0.1981, simple_loss=0.2634, pruned_loss=0.06642, over 953648.58 frames. ], batch size: 41, lr: 3.82e-03, grad_scale: 64.0 2023-04-26 21:33:27,081 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6328, 2.0205, 1.6876, 1.8981, 1.6395, 1.6080, 1.7163, 1.4162], device='cuda:6'), covar=tensor([0.1653, 0.1158, 0.0894, 0.1058, 0.2914, 0.1258, 0.1655, 0.2342], device='cuda:6'), in_proj_covar=tensor([0.0302, 0.0322, 0.0231, 0.0294, 0.0320, 0.0276, 0.0262, 0.0286], device='cuda:6'), out_proj_covar=tensor([1.2252e-04, 1.2992e-04, 9.3363e-05, 1.1757e-04, 1.3167e-04, 1.1133e-04, 1.0728e-04, 1.1476e-04], device='cuda:6') 2023-04-26 21:34:00,167 INFO [finetune.py:976] (6/7) Epoch 8, batch 4300, loss[loss=0.1891, simple_loss=0.2595, pruned_loss=0.05935, over 4869.00 frames. ], tot_loss[loss=0.1941, simple_loss=0.2588, pruned_loss=0.06475, over 953740.23 frames. ], batch size: 31, lr: 3.82e-03, grad_scale: 64.0 2023-04-26 21:34:15,557 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-04-26 21:34:26,198 INFO [optim.py:369] (6/7) 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 21:34:44,409 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=44443.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 21:34:44,912 INFO [finetune.py:976] (6/7) Epoch 8, batch 4350, loss[loss=0.1797, simple_loss=0.2485, pruned_loss=0.05546, over 4826.00 frames. ], tot_loss[loss=0.1924, simple_loss=0.2566, pruned_loss=0.06407, over 953585.50 frames. ], batch size: 33, lr: 3.82e-03, grad_scale: 64.0 2023-04-26 21:35:40,513 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.4789, 1.9499, 2.3661, 3.0758, 2.3577, 1.8917, 1.8990, 2.2574], device='cuda:6'), covar=tensor([0.4285, 0.4173, 0.2088, 0.3152, 0.3465, 0.3320, 0.4673, 0.3017], device='cuda:6'), in_proj_covar=tensor([0.0281, 0.0251, 0.0219, 0.0321, 0.0213, 0.0229, 0.0236, 0.0188], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:6') 2023-04-26 21:35:42,280 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.3481, 1.3086, 1.7757, 1.6341, 1.3252, 1.0776, 1.4426, 0.9673], device='cuda:6'), covar=tensor([0.0719, 0.0941, 0.0524, 0.0918, 0.0937, 0.1323, 0.0834, 0.0904], device='cuda:6'), in_proj_covar=tensor([0.0066, 0.0073, 0.0073, 0.0067, 0.0077, 0.0096, 0.0079, 0.0075], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-26 21:35:45,312 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5226, 3.4448, 0.7426, 1.7947, 1.9647, 2.4593, 1.9770, 0.9252], device='cuda:6'), covar=tensor([0.1381, 0.1035, 0.2104, 0.1349, 0.1049, 0.0976, 0.1443, 0.2105], device='cuda:6'), in_proj_covar=tensor([0.0117, 0.0253, 0.0142, 0.0124, 0.0135, 0.0155, 0.0120, 0.0121], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-26 21:35:46,506 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 8, batch 4400, loss[loss=0.2656, simple_loss=0.3135, pruned_loss=0.1089, over 4173.00 frames. ], tot_loss[loss=0.1952, simple_loss=0.2591, pruned_loss=0.06566, over 954678.26 frames. ], batch size: 65, lr: 3.82e-03, grad_scale: 64.0 2023-04-26 21:36:11,294 INFO [zipformer.py:1188] (6/7) 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] (6/7) 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 21:36:33,286 INFO [finetune.py:976] (6/7) Epoch 8, batch 4450, loss[loss=0.211, simple_loss=0.2806, pruned_loss=0.07067, over 4918.00 frames. ], tot_loss[loss=0.1988, simple_loss=0.2633, pruned_loss=0.06719, over 952585.44 frames. ], batch size: 36, lr: 3.82e-03, grad_scale: 64.0 2023-04-26 21:36:42,758 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.3458, 2.9997, 0.9386, 1.4633, 1.9974, 1.2496, 3.9531, 1.6870], device='cuda:6'), covar=tensor([0.0649, 0.0722, 0.0947, 0.1312, 0.0528, 0.1082, 0.0237, 0.0681], device='cuda:6'), in_proj_covar=tensor([0.0052, 0.0068, 0.0050, 0.0048, 0.0052, 0.0053, 0.0079, 0.0052], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0011, 0.0008], device='cuda:6') 2023-04-26 21:36:42,784 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=44550.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 21:36:58,024 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=44565.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 21:37:30,515 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=44588.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 21:37:39,768 INFO [finetune.py:976] (6/7) Epoch 8, batch 4500, loss[loss=0.2471, simple_loss=0.2953, pruned_loss=0.09947, over 4808.00 frames. ], tot_loss[loss=0.1983, simple_loss=0.263, pruned_loss=0.06678, over 952413.87 frames. ], batch size: 39, lr: 3.82e-03, grad_scale: 64.0 2023-04-26 21:37:42,336 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7068, 1.3312, 1.3276, 1.4953, 1.9816, 1.5593, 1.3073, 1.3000], device='cuda:6'), covar=tensor([0.2014, 0.1551, 0.2251, 0.1509, 0.0897, 0.1790, 0.2623, 0.2169], device='cuda:6'), in_proj_covar=tensor([0.0306, 0.0322, 0.0350, 0.0300, 0.0336, 0.0321, 0.0307, 0.0353], device='cuda:6'), out_proj_covar=tensor([6.4723e-05, 6.8315e-05, 7.5673e-05, 6.2222e-05, 7.0477e-05, 6.9054e-05, 6.6193e-05, 7.5773e-05], device='cuda:6') 2023-04-26 21:37:57,813 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=44611.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 21:37:58,983 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=44613.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 21:38:00,235 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=44615.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 21:38:30,721 INFO [optim.py:369] (6/7) 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 21:38:43,668 INFO [finetune.py:976] (6/7) Epoch 8, batch 4550, loss[loss=0.1887, simple_loss=0.2609, pruned_loss=0.05823, over 4867.00 frames. ], tot_loss[loss=0.1992, simple_loss=0.2643, pruned_loss=0.06703, over 952640.89 frames. ], batch size: 31, lr: 3.82e-03, grad_scale: 32.0 2023-04-26 21:38:48,711 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9697, 1.7869, 2.1323, 2.3874, 2.5045, 1.8264, 1.4758, 2.0972], device='cuda:6'), covar=tensor([0.0992, 0.1154, 0.0622, 0.0634, 0.0507, 0.0927, 0.0991, 0.0635], device='cuda:6'), in_proj_covar=tensor([0.0197, 0.0203, 0.0179, 0.0175, 0.0176, 0.0189, 0.0160, 0.0184], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-26 21:38:55,369 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=44663.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 21:39:13,544 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.1692, 2.0382, 2.3829, 2.5983, 2.6734, 2.1038, 1.6672, 2.2304], device='cuda:6'), covar=tensor([0.0919, 0.1060, 0.0591, 0.0582, 0.0508, 0.0822, 0.0923, 0.0629], device='cuda:6'), in_proj_covar=tensor([0.0196, 0.0202, 0.0178, 0.0175, 0.0175, 0.0189, 0.0159, 0.0183], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-26 21:39:17,128 INFO [finetune.py:976] (6/7) Epoch 8, batch 4600, loss[loss=0.2492, simple_loss=0.3129, pruned_loss=0.09276, over 4775.00 frames. ], tot_loss[loss=0.1982, simple_loss=0.2635, pruned_loss=0.06643, over 953010.38 frames. ], batch size: 51, lr: 3.82e-03, grad_scale: 32.0 2023-04-26 21:39:39,945 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-04-26 21:39:43,237 INFO [optim.py:369] (6/7) 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 21:39:55,902 INFO [finetune.py:976] (6/7) Epoch 8, batch 4650, loss[loss=0.2281, simple_loss=0.2712, pruned_loss=0.09246, over 4587.00 frames. ], tot_loss[loss=0.1959, simple_loss=0.2609, pruned_loss=0.06548, over 953468.78 frames. ], batch size: 20, lr: 3.82e-03, grad_scale: 32.0 2023-04-26 21:40:50,916 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.63 vs. limit=5.0 2023-04-26 21:41:01,772 INFO [finetune.py:976] (6/7) Epoch 8, batch 4700, loss[loss=0.1868, simple_loss=0.2534, pruned_loss=0.06009, over 4865.00 frames. ], tot_loss[loss=0.1927, simple_loss=0.2577, pruned_loss=0.06388, over 955681.01 frames. ], batch size: 31, lr: 3.82e-03, grad_scale: 32.0 2023-04-26 21:41:26,394 INFO [optim.py:369] (6/7) 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 21:41:30,828 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-04-26 21:41:35,057 INFO [finetune.py:976] (6/7) Epoch 8, batch 4750, loss[loss=0.2239, simple_loss=0.2862, pruned_loss=0.08085, over 4826.00 frames. ], tot_loss[loss=0.1927, simple_loss=0.2568, pruned_loss=0.06429, over 955804.23 frames. ], batch size: 39, lr: 3.82e-03, grad_scale: 32.0 2023-04-26 21:41:38,003 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7276, 3.5572, 0.8496, 1.9247, 2.1915, 2.5069, 2.0342, 0.9825], device='cuda:6'), covar=tensor([0.1294, 0.1057, 0.2274, 0.1366, 0.0976, 0.1078, 0.1592, 0.2195], device='cuda:6'), in_proj_covar=tensor([0.0116, 0.0251, 0.0141, 0.0123, 0.0135, 0.0154, 0.0119, 0.0121], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-26 21:41:41,405 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.71 vs. limit=5.0 2023-04-26 21:41:59,218 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 8, batch 4800, loss[loss=0.2056, simple_loss=0.2742, pruned_loss=0.06847, over 4812.00 frames. ], tot_loss[loss=0.1948, simple_loss=0.2589, pruned_loss=0.06537, over 955625.77 frames. ], batch size: 51, lr: 3.82e-03, grad_scale: 32.0 2023-04-26 21:42:16,079 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=44906.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 21:42:17,064 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-04-26 21:42:32,274 INFO [optim.py:369] (6/7) 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 21:42:40,900 INFO [finetune.py:976] (6/7) Epoch 8, batch 4850, loss[loss=0.2373, simple_loss=0.3047, pruned_loss=0.08491, over 4911.00 frames. ], tot_loss[loss=0.197, simple_loss=0.2616, pruned_loss=0.06621, over 955199.14 frames. ], batch size: 38, lr: 3.82e-03, grad_scale: 32.0 2023-04-26 21:43:40,818 INFO [finetune.py:976] (6/7) Epoch 8, batch 4900, loss[loss=0.2056, simple_loss=0.2843, pruned_loss=0.06344, over 4803.00 frames. ], tot_loss[loss=0.1974, simple_loss=0.2626, pruned_loss=0.06611, over 954972.25 frames. ], batch size: 40, lr: 3.82e-03, grad_scale: 32.0 2023-04-26 21:44:25,591 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.2443, 2.9681, 0.8609, 1.5357, 1.6622, 2.1909, 1.6697, 0.9348], device='cuda:6'), covar=tensor([0.1517, 0.0898, 0.1999, 0.1405, 0.1153, 0.0942, 0.1579, 0.1955], device='cuda:6'), in_proj_covar=tensor([0.0117, 0.0251, 0.0141, 0.0124, 0.0135, 0.0155, 0.0119, 0.0122], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-26 21:44:34,743 INFO [optim.py:369] (6/7) 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 21:44:46,777 INFO [finetune.py:976] (6/7) Epoch 8, batch 4950, loss[loss=0.2405, simple_loss=0.3086, pruned_loss=0.08626, over 4175.00 frames. ], tot_loss[loss=0.1992, simple_loss=0.2646, pruned_loss=0.06689, over 953180.06 frames. ], batch size: 65, lr: 3.81e-03, grad_scale: 32.0 2023-04-26 21:45:08,570 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.1345, 1.4689, 1.2484, 1.6563, 1.4590, 1.5796, 1.2884, 2.9601], device='cuda:6'), covar=tensor([0.0675, 0.0767, 0.0777, 0.1230, 0.0642, 0.0564, 0.0769, 0.0185], device='cuda:6'), in_proj_covar=tensor([0.0039, 0.0039, 0.0040, 0.0044, 0.0040, 0.0039, 0.0039, 0.0059], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0014, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:6') 2023-04-26 21:45:08,586 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=45056.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 21:45:11,649 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6288, 1.6941, 0.9382, 1.3446, 1.7951, 1.5007, 1.4096, 1.4940], device='cuda:6'), covar=tensor([0.0533, 0.0398, 0.0370, 0.0582, 0.0296, 0.0589, 0.0535, 0.0636], device='cuda:6'), in_proj_covar=tensor([0.0030, 0.0026, 0.0024, 0.0030, 0.0021, 0.0030, 0.0029, 0.0030], device='cuda:6'), out_proj_covar=tensor([0.0049, 0.0044, 0.0038, 0.0050, 0.0038, 0.0048, 0.0048, 0.0050], device='cuda:6') 2023-04-26 21:45:21,008 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([3.7668, 3.7287, 2.5562, 4.3459, 3.7991, 3.7792, 1.5727, 3.6110], device='cuda:6'), covar=tensor([0.1677, 0.1372, 0.3303, 0.1677, 0.2425, 0.1648, 0.5718, 0.2637], device='cuda:6'), in_proj_covar=tensor([0.0246, 0.0218, 0.0252, 0.0309, 0.0302, 0.0252, 0.0274, 0.0274], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-26 21:45:39,759 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=45081.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 21:45:47,621 INFO [finetune.py:976] (6/7) Epoch 8, batch 5000, loss[loss=0.2071, simple_loss=0.265, pruned_loss=0.07458, over 4903.00 frames. ], tot_loss[loss=0.1988, simple_loss=0.2641, pruned_loss=0.06673, over 953449.55 frames. ], batch size: 36, lr: 3.81e-03, grad_scale: 32.0 2023-04-26 21:45:51,218 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0812, 1.5942, 1.9223, 2.0808, 1.8345, 1.5040, 0.9326, 1.5760], device='cuda:6'), covar=tensor([0.3604, 0.3456, 0.1711, 0.2399, 0.2662, 0.2684, 0.4856, 0.2291], device='cuda:6'), in_proj_covar=tensor([0.0281, 0.0252, 0.0220, 0.0320, 0.0214, 0.0229, 0.0236, 0.0188], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:6') 2023-04-26 21:46:01,718 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([0.8099, 1.8961, 1.6951, 1.4236, 1.9059, 1.5830, 2.3818, 1.4965], device='cuda:6'), covar=tensor([0.3812, 0.1565, 0.4368, 0.3126, 0.1744, 0.2259, 0.1498, 0.4605], device='cuda:6'), in_proj_covar=tensor([0.0349, 0.0353, 0.0434, 0.0367, 0.0390, 0.0385, 0.0385, 0.0422], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-26 21:46:04,159 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=45117.0, num_to_drop=1, layers_to_drop={2} 2023-04-26 21:46:13,681 INFO [optim.py:369] (6/7) 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 21:46:15,046 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0826, 1.5460, 1.9272, 2.2688, 1.8520, 1.5061, 1.1322, 1.7375], device='cuda:6'), covar=tensor([0.3530, 0.3759, 0.1799, 0.2694, 0.3070, 0.2989, 0.4690, 0.2531], device='cuda:6'), in_proj_covar=tensor([0.0281, 0.0252, 0.0219, 0.0319, 0.0214, 0.0228, 0.0235, 0.0187], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:6') 2023-04-26 21:46:19,282 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=45142.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 21:46:20,394 INFO [finetune.py:976] (6/7) Epoch 8, batch 5050, loss[loss=0.1632, simple_loss=0.2307, pruned_loss=0.04785, over 4921.00 frames. ], tot_loss[loss=0.1964, simple_loss=0.2614, pruned_loss=0.06571, over 955305.19 frames. ], batch size: 33, lr: 3.81e-03, grad_scale: 32.0 2023-04-26 21:46:47,003 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 8, batch 5100, loss[loss=0.1349, simple_loss=0.213, pruned_loss=0.02838, over 4817.00 frames. ], tot_loss[loss=0.1932, simple_loss=0.2578, pruned_loss=0.06426, over 956070.96 frames. ], batch size: 25, lr: 3.81e-03, grad_scale: 32.0 2023-04-26 21:47:02,073 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=45206.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 21:47:18,728 INFO [zipformer.py:1188] (6/7) 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] (6/7) 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 21:47:26,587 INFO [finetune.py:976] (6/7) Epoch 8, batch 5150, loss[loss=0.229, simple_loss=0.2866, pruned_loss=0.08564, over 4837.00 frames. ], tot_loss[loss=0.1935, simple_loss=0.2578, pruned_loss=0.06462, over 954696.63 frames. ], batch size: 33, lr: 3.81e-03, grad_scale: 32.0 2023-04-26 21:47:32,729 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=45254.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 21:47:46,400 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.12 vs. limit=2.0 2023-04-26 21:47:58,372 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-04-26 21:48:00,051 INFO [finetune.py:976] (6/7) Epoch 8, batch 5200, loss[loss=0.1583, simple_loss=0.2294, pruned_loss=0.0436, over 4767.00 frames. ], tot_loss[loss=0.1956, simple_loss=0.2612, pruned_loss=0.06499, over 955173.93 frames. ], batch size: 26, lr: 3.81e-03, grad_scale: 32.0 2023-04-26 21:48:01,044 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.91 vs. limit=5.0 2023-04-26 21:48:14,204 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=3.30 vs. limit=5.0 2023-04-26 21:48:44,627 INFO [optim.py:369] (6/7) 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 21:48:56,373 INFO [finetune.py:976] (6/7) Epoch 8, batch 5250, loss[loss=0.2132, simple_loss=0.298, pruned_loss=0.06425, over 4896.00 frames. ], tot_loss[loss=0.1967, simple_loss=0.2625, pruned_loss=0.0655, over 953231.43 frames. ], batch size: 43, lr: 3.81e-03, grad_scale: 32.0 2023-04-26 21:48:58,955 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=45348.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 21:50:01,145 INFO [finetune.py:976] (6/7) Epoch 8, batch 5300, loss[loss=0.2821, simple_loss=0.3285, pruned_loss=0.1179, over 4816.00 frames. ], tot_loss[loss=0.1984, simple_loss=0.2644, pruned_loss=0.06617, over 955634.87 frames. ], batch size: 39, lr: 3.81e-03, grad_scale: 32.0 2023-04-26 21:50:21,684 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=45409.0, num_to_drop=1, layers_to_drop={3} 2023-04-26 21:50:23,980 INFO [zipformer.py:1188] (6/7) 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] (6/7) 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 21:50:57,915 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=45437.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 21:51:07,773 INFO [finetune.py:976] (6/7) Epoch 8, batch 5350, loss[loss=0.1725, simple_loss=0.2379, pruned_loss=0.05357, over 4918.00 frames. ], tot_loss[loss=0.199, simple_loss=0.2651, pruned_loss=0.06649, over 955795.43 frames. ], batch size: 43, lr: 3.81e-03, grad_scale: 32.0 2023-04-26 21:51:45,339 INFO [finetune.py:976] (6/7) Epoch 8, batch 5400, loss[loss=0.1443, simple_loss=0.2066, pruned_loss=0.04104, over 4764.00 frames. ], tot_loss[loss=0.1957, simple_loss=0.2612, pruned_loss=0.06506, over 955481.93 frames. ], batch size: 28, lr: 3.81e-03, grad_scale: 32.0 2023-04-26 21:51:51,474 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.1695, 1.8563, 2.1814, 2.5283, 2.5527, 2.0683, 1.6350, 2.0960], device='cuda:6'), covar=tensor([0.0833, 0.1188, 0.0669, 0.0579, 0.0578, 0.0891, 0.0895, 0.0669], device='cuda:6'), in_proj_covar=tensor([0.0196, 0.0204, 0.0179, 0.0176, 0.0177, 0.0189, 0.0160, 0.0185], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-26 21:52:11,225 INFO [optim.py:369] (6/7) 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 21:52:18,383 INFO [finetune.py:976] (6/7) Epoch 8, batch 5450, loss[loss=0.1754, simple_loss=0.2334, pruned_loss=0.05876, over 4722.00 frames. ], tot_loss[loss=0.1949, simple_loss=0.2595, pruned_loss=0.06509, over 956802.36 frames. ], batch size: 23, lr: 3.81e-03, grad_scale: 32.0 2023-04-26 21:52:46,344 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8009, 2.2423, 1.9360, 2.0566, 1.6211, 1.8309, 1.9067, 1.4553], device='cuda:6'), covar=tensor([0.2002, 0.1143, 0.0804, 0.1242, 0.3176, 0.1064, 0.1925, 0.2547], device='cuda:6'), in_proj_covar=tensor([0.0298, 0.0319, 0.0229, 0.0291, 0.0317, 0.0272, 0.0259, 0.0281], device='cuda:6'), out_proj_covar=tensor([1.2066e-04, 1.2863e-04, 9.2353e-05, 1.1657e-04, 1.3018e-04, 1.0989e-04, 1.0588e-04, 1.1255e-04], device='cuda:6') 2023-04-26 21:52:51,571 INFO [finetune.py:976] (6/7) Epoch 8, batch 5500, loss[loss=0.1777, simple_loss=0.2388, pruned_loss=0.05837, over 4751.00 frames. ], tot_loss[loss=0.1932, simple_loss=0.2571, pruned_loss=0.06459, over 956087.68 frames. ], batch size: 59, lr: 3.81e-03, grad_scale: 32.0 2023-04-26 21:53:12,477 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-04-26 21:53:16,426 INFO [optim.py:369] (6/7) 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 21:53:23,182 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-04-26 21:53:24,570 INFO [finetune.py:976] (6/7) Epoch 8, batch 5550, loss[loss=0.181, simple_loss=0.2627, pruned_loss=0.04967, over 4894.00 frames. ], tot_loss[loss=0.1948, simple_loss=0.2592, pruned_loss=0.06521, over 956544.21 frames. ], batch size: 43, lr: 3.81e-03, grad_scale: 32.0 2023-04-26 21:54:06,992 INFO [finetune.py:976] (6/7) Epoch 8, batch 5600, loss[loss=0.2023, simple_loss=0.2686, pruned_loss=0.06803, over 4904.00 frames. ], tot_loss[loss=0.1979, simple_loss=0.2629, pruned_loss=0.0665, over 955943.46 frames. ], batch size: 37, lr: 3.81e-03, grad_scale: 32.0 2023-04-26 21:54:12,153 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=45702.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 21:54:13,289 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=45704.0, num_to_drop=1, layers_to_drop={3} 2023-04-26 21:54:13,972 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-04-26 21:54:17,972 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=45712.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 21:54:30,165 INFO [optim.py:369] (6/7) 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 21:54:32,594 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=45737.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 21:54:36,956 INFO [finetune.py:976] (6/7) Epoch 8, batch 5650, loss[loss=0.2143, simple_loss=0.2828, pruned_loss=0.07292, over 4821.00 frames. ], tot_loss[loss=0.1983, simple_loss=0.2646, pruned_loss=0.06604, over 956656.45 frames. ], batch size: 33, lr: 3.81e-03, grad_scale: 32.0 2023-04-26 21:54:46,839 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=45760.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 21:54:48,653 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=45763.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 21:54:54,976 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=45774.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 21:55:07,633 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=45785.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 21:55:18,512 INFO [finetune.py:976] (6/7) Epoch 8, batch 5700, loss[loss=0.1632, simple_loss=0.2172, pruned_loss=0.05459, over 4252.00 frames. ], tot_loss[loss=0.1964, simple_loss=0.261, pruned_loss=0.06586, over 939469.00 frames. ], batch size: 18, lr: 3.81e-03, grad_scale: 32.0 2023-04-26 21:55:25,091 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.2614, 1.5898, 5.5582, 5.2050, 4.8156, 5.3588, 4.7899, 4.9497], device='cuda:6'), covar=tensor([0.6257, 0.6034, 0.0870, 0.1494, 0.0938, 0.1307, 0.0923, 0.1322], device='cuda:6'), in_proj_covar=tensor([0.0306, 0.0303, 0.0403, 0.0406, 0.0343, 0.0400, 0.0312, 0.0364], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-26 21:55:39,445 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([0.6420, 1.6700, 1.7284, 1.3227, 1.7752, 1.4039, 2.2590, 1.5070], device='cuda:6'), covar=tensor([0.3533, 0.1619, 0.4060, 0.2804, 0.1446, 0.2223, 0.1592, 0.4361], device='cuda:6'), in_proj_covar=tensor([0.0348, 0.0352, 0.0433, 0.0366, 0.0391, 0.0385, 0.0384, 0.0423], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-26 21:55:59,017 INFO [finetune.py:976] (6/7) Epoch 9, batch 0, loss[loss=0.1735, simple_loss=0.2424, pruned_loss=0.05234, over 4728.00 frames. ], tot_loss[loss=0.1735, simple_loss=0.2424, pruned_loss=0.05234, over 4728.00 frames. ], batch size: 59, lr: 3.81e-03, grad_scale: 32.0 2023-04-26 21:55:59,017 INFO [finetune.py:1001] (6/7) Computing validation loss 2023-04-26 21:56:05,201 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.4189, 3.0616, 0.9145, 1.7826, 1.9228, 2.2517, 1.8937, 1.0491], device='cuda:6'), covar=tensor([0.1340, 0.1135, 0.1987, 0.1285, 0.0935, 0.0913, 0.1538, 0.1741], device='cuda:6'), in_proj_covar=tensor([0.0118, 0.0253, 0.0142, 0.0124, 0.0135, 0.0156, 0.0120, 0.0122], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-26 21:56:14,947 INFO [finetune.py:1010] (6/7) 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] (6/7) Maximum memory allocated so far is 6345MB 2023-04-26 21:56:33,615 INFO [optim.py:369] (6/7) 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 21:56:34,965 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=45835.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 21:56:46,621 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.2130, 2.8776, 0.9553, 1.3804, 2.0203, 1.2921, 3.7148, 1.5498], device='cuda:6'), covar=tensor([0.0641, 0.1009, 0.0886, 0.1205, 0.0514, 0.0915, 0.0179, 0.0653], device='cuda:6'), in_proj_covar=tensor([0.0052, 0.0068, 0.0050, 0.0048, 0.0052, 0.0053, 0.0079, 0.0052], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0011, 0.0008], device='cuda:6') 2023-04-26 21:57:05,574 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=45859.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 21:57:19,425 INFO [finetune.py:976] (6/7) Epoch 9, batch 50, loss[loss=0.2129, simple_loss=0.2741, pruned_loss=0.07588, over 4902.00 frames. ], tot_loss[loss=0.2001, simple_loss=0.267, pruned_loss=0.06665, over 217366.65 frames. ], batch size: 35, lr: 3.81e-03, grad_scale: 32.0 2023-04-26 21:57:38,685 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8402, 1.5258, 1.4241, 1.6808, 2.1435, 1.7678, 1.5168, 1.3598], device='cuda:6'), covar=tensor([0.1906, 0.1540, 0.1851, 0.1504, 0.0893, 0.1517, 0.2085, 0.2179], device='cuda:6'), in_proj_covar=tensor([0.0312, 0.0326, 0.0354, 0.0303, 0.0341, 0.0327, 0.0311, 0.0357], device='cuda:6'), 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:6') 2023-04-26 21:57:51,863 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=45920.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 21:57:52,937 INFO [finetune.py:976] (6/7) Epoch 9, batch 100, loss[loss=0.1832, simple_loss=0.2344, pruned_loss=0.06599, over 4763.00 frames. ], tot_loss[loss=0.1958, simple_loss=0.2607, pruned_loss=0.06548, over 383273.69 frames. ], batch size: 26, lr: 3.81e-03, grad_scale: 32.0 2023-04-26 21:58:01,517 INFO [optim.py:369] (6/7) 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 21:58:24,854 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-04-26 21:58:26,130 INFO [finetune.py:976] (6/7) Epoch 9, batch 150, loss[loss=0.1496, simple_loss=0.2147, pruned_loss=0.04226, over 4802.00 frames. ], tot_loss[loss=0.1904, simple_loss=0.2539, pruned_loss=0.0635, over 510232.92 frames. ], batch size: 51, lr: 3.81e-03, grad_scale: 32.0 2023-04-26 21:58:47,801 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=46004.0, num_to_drop=1, layers_to_drop={2} 2023-04-26 21:58:59,030 INFO [finetune.py:976] (6/7) Epoch 9, batch 200, loss[loss=0.181, simple_loss=0.2444, pruned_loss=0.05883, over 4916.00 frames. ], tot_loss[loss=0.1939, simple_loss=0.2557, pruned_loss=0.06605, over 607091.77 frames. ], batch size: 36, lr: 3.81e-03, grad_scale: 32.0 2023-04-26 21:59:07,991 INFO [optim.py:369] (6/7) 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 21:59:17,373 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7983, 2.6186, 1.7722, 1.8763, 1.3181, 1.3647, 1.9980, 1.2219], device='cuda:6'), covar=tensor([0.1792, 0.1653, 0.1572, 0.2034, 0.2617, 0.2087, 0.1097, 0.2224], device='cuda:6'), in_proj_covar=tensor([0.0195, 0.0217, 0.0172, 0.0205, 0.0206, 0.0185, 0.0162, 0.0188], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:6') 2023-04-26 21:59:19,779 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=46052.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 21:59:23,461 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=46058.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 21:59:32,395 INFO [finetune.py:976] (6/7) Epoch 9, batch 250, loss[loss=0.1741, simple_loss=0.2404, pruned_loss=0.05389, over 4805.00 frames. ], tot_loss[loss=0.1994, simple_loss=0.2623, pruned_loss=0.06826, over 685080.99 frames. ], batch size: 25, lr: 3.81e-03, grad_scale: 32.0 2023-04-26 21:59:49,123 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.76 vs. limit=5.0 2023-04-26 21:59:51,856 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.14 vs. limit=2.0 2023-04-26 21:59:53,244 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.14 vs. limit=2.0 2023-04-26 22:00:05,478 INFO [finetune.py:976] (6/7) Epoch 9, batch 300, loss[loss=0.1602, simple_loss=0.2269, pruned_loss=0.0467, over 4071.00 frames. ], tot_loss[loss=0.2, simple_loss=0.2642, pruned_loss=0.06792, over 744450.72 frames. ], batch size: 17, lr: 3.81e-03, grad_scale: 32.0 2023-04-26 22:00:10,885 INFO [zipformer.py:1188] (6/7) 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] (6/7) 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 22:00:24,313 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.3232, 1.4341, 1.6687, 1.8126, 1.6542, 1.7680, 1.8055, 1.7683], device='cuda:6'), covar=tensor([0.5241, 0.7071, 0.5978, 0.5570, 0.6855, 0.9635, 0.6992, 0.6774], device='cuda:6'), in_proj_covar=tensor([0.0320, 0.0381, 0.0314, 0.0324, 0.0339, 0.0400, 0.0361, 0.0324], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-26 22:00:42,788 INFO [finetune.py:976] (6/7) Epoch 9, batch 350, loss[loss=0.2258, simple_loss=0.2834, pruned_loss=0.08414, over 4919.00 frames. ], tot_loss[loss=0.2015, simple_loss=0.2659, pruned_loss=0.06857, over 792325.35 frames. ], batch size: 33, lr: 3.81e-03, grad_scale: 32.0 2023-04-26 22:01:16,908 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5275, 1.2511, 1.2516, 1.3769, 1.8201, 1.5094, 1.2472, 1.1949], device='cuda:6'), covar=tensor([0.1587, 0.1450, 0.1725, 0.1417, 0.0726, 0.1442, 0.2043, 0.1956], device='cuda:6'), in_proj_covar=tensor([0.0309, 0.0324, 0.0351, 0.0300, 0.0340, 0.0324, 0.0308, 0.0353], device='cuda:6'), out_proj_covar=tensor([6.5508e-05, 6.8713e-05, 7.5864e-05, 6.2209e-05, 7.1447e-05, 6.9531e-05, 6.6312e-05, 7.5842e-05], device='cuda:6') 2023-04-26 22:01:38,766 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=46215.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:01:48,012 INFO [finetune.py:976] (6/7) Epoch 9, batch 400, loss[loss=0.2033, simple_loss=0.2718, pruned_loss=0.06743, over 4290.00 frames. ], tot_loss[loss=0.2013, simple_loss=0.2657, pruned_loss=0.06839, over 826966.17 frames. ], batch size: 66, lr: 3.80e-03, grad_scale: 32.0 2023-04-26 22:02:01,098 INFO [optim.py:369] (6/7) 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 22:02:14,229 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5391, 0.9915, 1.5384, 1.9297, 1.6645, 1.5299, 1.5297, 1.5723], device='cuda:6'), covar=tensor([0.6104, 0.8553, 0.8164, 0.9012, 0.7892, 1.0031, 0.9666, 0.9305], device='cuda:6'), in_proj_covar=tensor([0.0408, 0.0421, 0.0504, 0.0523, 0.0437, 0.0455, 0.0467, 0.0465], device='cuda:6'), 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:6') 2023-04-26 22:02:26,905 INFO [finetune.py:976] (6/7) Epoch 9, batch 450, loss[loss=0.2078, simple_loss=0.2697, pruned_loss=0.07297, over 4815.00 frames. ], tot_loss[loss=0.1987, simple_loss=0.2636, pruned_loss=0.06683, over 855270.73 frames. ], batch size: 39, lr: 3.80e-03, grad_scale: 32.0 2023-04-26 22:02:28,859 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9933, 2.0212, 2.1439, 2.4328, 2.4575, 1.9646, 1.5225, 2.0216], device='cuda:6'), covar=tensor([0.0930, 0.0979, 0.0667, 0.0639, 0.0543, 0.0842, 0.0904, 0.0693], device='cuda:6'), in_proj_covar=tensor([0.0197, 0.0204, 0.0180, 0.0177, 0.0178, 0.0190, 0.0159, 0.0185], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-26 22:02:29,438 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=46276.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:02:57,754 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0989, 1.9384, 2.3997, 2.6394, 2.0091, 1.7853, 2.1324, 1.2055], device='cuda:6'), covar=tensor([0.0740, 0.0954, 0.0682, 0.0870, 0.0871, 0.1341, 0.0872, 0.1251], device='cuda:6'), in_proj_covar=tensor([0.0066, 0.0073, 0.0072, 0.0067, 0.0076, 0.0095, 0.0078, 0.0074], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-26 22:03:00,648 INFO [finetune.py:976] (6/7) Epoch 9, batch 500, loss[loss=0.158, simple_loss=0.2352, pruned_loss=0.04035, over 4785.00 frames. ], tot_loss[loss=0.196, simple_loss=0.2607, pruned_loss=0.06565, over 877117.96 frames. ], batch size: 29, lr: 3.80e-03, grad_scale: 32.0 2023-04-26 22:03:07,759 INFO [optim.py:369] (6/7) 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 22:03:10,753 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=46337.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:03:26,023 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=46358.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:03:34,350 INFO [finetune.py:976] (6/7) Epoch 9, batch 550, loss[loss=0.2254, simple_loss=0.2746, pruned_loss=0.08809, over 4910.00 frames. ], tot_loss[loss=0.1921, simple_loss=0.2567, pruned_loss=0.0637, over 895292.27 frames. ], batch size: 32, lr: 3.80e-03, grad_scale: 32.0 2023-04-26 22:03:39,356 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.1688, 1.6190, 2.0809, 2.3755, 1.9817, 1.6445, 1.2791, 1.6822], device='cuda:6'), covar=tensor([0.3915, 0.3892, 0.1933, 0.2707, 0.3088, 0.3046, 0.4877, 0.2636], device='cuda:6'), in_proj_covar=tensor([0.0281, 0.0251, 0.0219, 0.0318, 0.0212, 0.0228, 0.0235, 0.0187], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:6') 2023-04-26 22:03:57,939 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 9, batch 600, loss[loss=0.2129, simple_loss=0.2719, pruned_loss=0.07698, over 4894.00 frames. ], tot_loss[loss=0.1934, simple_loss=0.2578, pruned_loss=0.06454, over 908025.55 frames. ], batch size: 43, lr: 3.80e-03, grad_scale: 32.0 2023-04-26 22:04:12,576 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=46430.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:04:14,309 INFO [optim.py:369] (6/7) 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 22:04:15,209 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=5.13 vs. limit=5.0 2023-04-26 22:04:20,970 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8903, 1.7726, 2.0954, 2.2322, 2.0071, 1.7707, 1.9005, 1.9328], device='cuda:6'), covar=tensor([0.6838, 0.8934, 0.9725, 0.9224, 0.7986, 1.2721, 1.2027, 1.0858], device='cuda:6'), in_proj_covar=tensor([0.0410, 0.0423, 0.0506, 0.0525, 0.0438, 0.0457, 0.0469, 0.0467], device='cuda:6'), out_proj_covar=tensor([9.9787e-05, 1.0488e-04, 1.1430e-04, 1.2469e-04, 1.0658e-04, 1.1063e-04, 1.1281e-04, 1.1309e-04], device='cuda:6') 2023-04-26 22:04:40,996 INFO [finetune.py:976] (6/7) Epoch 9, batch 650, loss[loss=0.1762, simple_loss=0.2587, pruned_loss=0.04686, over 4900.00 frames. ], tot_loss[loss=0.1954, simple_loss=0.2606, pruned_loss=0.06509, over 915739.21 frames. ], batch size: 43, lr: 3.80e-03, grad_scale: 32.0 2023-04-26 22:04:44,731 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=46478.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:05:10,664 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=46515.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:05:14,905 INFO [finetune.py:976] (6/7) Epoch 9, batch 700, loss[loss=0.1833, simple_loss=0.2609, pruned_loss=0.05288, over 4823.00 frames. ], tot_loss[loss=0.1986, simple_loss=0.2637, pruned_loss=0.06676, over 926252.48 frames. ], batch size: 49, lr: 3.80e-03, grad_scale: 32.0 2023-04-26 22:05:21,598 INFO [optim.py:369] (6/7) 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 22:05:43,205 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=46563.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:05:53,549 INFO [finetune.py:976] (6/7) Epoch 9, batch 750, loss[loss=0.1888, simple_loss=0.2655, pruned_loss=0.05605, over 4924.00 frames. ], tot_loss[loss=0.1979, simple_loss=0.2632, pruned_loss=0.06625, over 931882.59 frames. ], batch size: 41, lr: 3.80e-03, grad_scale: 32.0 2023-04-26 22:06:18,454 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=46595.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:06:28,896 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=46602.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:06:46,875 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6605, 2.1795, 1.8946, 2.0946, 1.7026, 1.8011, 1.7756, 1.4822], device='cuda:6'), covar=tensor([0.2161, 0.1155, 0.0859, 0.1232, 0.2960, 0.1202, 0.1934, 0.2531], device='cuda:6'), in_proj_covar=tensor([0.0300, 0.0321, 0.0229, 0.0292, 0.0318, 0.0274, 0.0261, 0.0284], device='cuda:6'), out_proj_covar=tensor([1.2169e-04, 1.2947e-04, 9.2450e-05, 1.1715e-04, 1.3053e-04, 1.1035e-04, 1.0677e-04, 1.1426e-04], device='cuda:6') 2023-04-26 22:06:58,617 INFO [finetune.py:976] (6/7) Epoch 9, batch 800, loss[loss=0.2296, simple_loss=0.2924, pruned_loss=0.08342, over 4744.00 frames. ], tot_loss[loss=0.1971, simple_loss=0.2625, pruned_loss=0.06587, over 936148.32 frames. ], batch size: 27, lr: 3.80e-03, grad_scale: 64.0 2023-04-26 22:07:09,963 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=46632.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:07:10,489 INFO [optim.py:369] (6/7) 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 22:07:14,905 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.1714, 2.5325, 1.0331, 1.3923, 1.9239, 1.2159, 3.2724, 1.7172], device='cuda:6'), covar=tensor([0.0620, 0.0630, 0.0779, 0.1235, 0.0486, 0.0974, 0.0236, 0.0624], device='cuda:6'), in_proj_covar=tensor([0.0052, 0.0068, 0.0050, 0.0048, 0.0052, 0.0053, 0.0079, 0.0051], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0011, 0.0007], device='cuda:6') 2023-04-26 22:07:26,685 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=46656.0, num_to_drop=1, layers_to_drop={3} 2023-04-26 22:07:27,355 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-04-26 22:07:31,868 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=46663.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:07:37,209 INFO [finetune.py:976] (6/7) Epoch 9, batch 850, loss[loss=0.2413, simple_loss=0.2971, pruned_loss=0.09274, over 4827.00 frames. ], tot_loss[loss=0.1978, simple_loss=0.2624, pruned_loss=0.06659, over 941326.85 frames. ], batch size: 38, lr: 3.80e-03, grad_scale: 64.0 2023-04-26 22:07:40,968 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5565, 1.2921, 0.6263, 1.2260, 1.4192, 1.4058, 1.3016, 1.3205], device='cuda:6'), covar=tensor([0.0515, 0.0393, 0.0424, 0.0567, 0.0305, 0.0515, 0.0500, 0.0594], device='cuda:6'), in_proj_covar=tensor([0.0030, 0.0026, 0.0024, 0.0030, 0.0021, 0.0029, 0.0029, 0.0030], device='cuda:6'), out_proj_covar=tensor([0.0049, 0.0044, 0.0038, 0.0050, 0.0037, 0.0048, 0.0048, 0.0050], device='cuda:6') 2023-04-26 22:07:44,598 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([0.7969, 1.8364, 1.6983, 1.5095, 2.0389, 1.5687, 2.5115, 1.5489], device='cuda:6'), covar=tensor([0.4313, 0.2016, 0.5285, 0.3417, 0.1843, 0.2879, 0.1560, 0.4897], device='cuda:6'), in_proj_covar=tensor([0.0348, 0.0352, 0.0434, 0.0366, 0.0391, 0.0386, 0.0385, 0.0423], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-26 22:08:10,360 INFO [finetune.py:976] (6/7) Epoch 9, batch 900, loss[loss=0.2166, simple_loss=0.2709, pruned_loss=0.08111, over 4733.00 frames. ], tot_loss[loss=0.1946, simple_loss=0.2588, pruned_loss=0.06518, over 944399.91 frames. ], batch size: 23, lr: 3.80e-03, grad_scale: 64.0 2023-04-26 22:08:17,021 INFO [optim.py:369] (6/7) 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 22:08:43,452 INFO [finetune.py:976] (6/7) Epoch 9, batch 950, loss[loss=0.1655, simple_loss=0.2263, pruned_loss=0.05231, over 4762.00 frames. ], tot_loss[loss=0.1935, simple_loss=0.2576, pruned_loss=0.0647, over 947630.94 frames. ], batch size: 27, lr: 3.80e-03, grad_scale: 64.0 2023-04-26 22:08:49,102 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-04-26 22:09:08,148 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 9, batch 1000, loss[loss=0.2252, simple_loss=0.2974, pruned_loss=0.0765, over 4819.00 frames. ], tot_loss[loss=0.1957, simple_loss=0.2603, pruned_loss=0.0655, over 950677.77 frames. ], batch size: 39, lr: 3.80e-03, grad_scale: 64.0 2023-04-26 22:09:22,997 INFO [optim.py:369] (6/7) 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 22:09:40,489 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5251, 1.2938, 1.6561, 1.6357, 1.3995, 1.2556, 1.3244, 0.7600], device='cuda:6'), covar=tensor([0.0569, 0.0868, 0.0591, 0.0664, 0.0688, 0.1265, 0.0612, 0.0935], device='cuda:6'), in_proj_covar=tensor([0.0066, 0.0073, 0.0072, 0.0067, 0.0076, 0.0096, 0.0079, 0.0074], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-26 22:09:49,147 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 9, batch 1050, loss[loss=0.2064, simple_loss=0.2722, pruned_loss=0.07026, over 4896.00 frames. ], tot_loss[loss=0.1971, simple_loss=0.2627, pruned_loss=0.06569, over 951610.52 frames. ], batch size: 43, lr: 3.80e-03, grad_scale: 64.0 2023-04-26 22:09:53,400 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.1828, 2.5741, 1.1447, 1.3690, 2.0676, 1.3096, 3.5584, 1.8379], device='cuda:6'), covar=tensor([0.0656, 0.0635, 0.0759, 0.1418, 0.0490, 0.0999, 0.0242, 0.0655], device='cuda:6'), in_proj_covar=tensor([0.0051, 0.0067, 0.0050, 0.0048, 0.0052, 0.0052, 0.0078, 0.0051], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0007], device='cuda:6') 2023-04-26 22:10:08,587 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.1706, 1.4660, 1.2657, 1.7502, 1.5266, 1.7422, 1.2837, 3.5806], device='cuda:6'), covar=tensor([0.0765, 0.1042, 0.1064, 0.1333, 0.0835, 0.0711, 0.0998, 0.0199], device='cuda:6'), in_proj_covar=tensor([0.0038, 0.0039, 0.0040, 0.0044, 0.0040, 0.0039, 0.0039, 0.0059], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0014, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:6') 2023-04-26 22:10:22,799 INFO [finetune.py:976] (6/7) Epoch 9, batch 1100, loss[loss=0.2225, simple_loss=0.2893, pruned_loss=0.07785, over 4885.00 frames. ], tot_loss[loss=0.199, simple_loss=0.2648, pruned_loss=0.06663, over 952182.17 frames. ], batch size: 35, lr: 3.80e-03, grad_scale: 32.0 2023-04-26 22:10:29,450 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=46932.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:10:30,586 INFO [optim.py:369] (6/7) 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 22:10:40,917 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=46951.0, num_to_drop=1, layers_to_drop={3} 2023-04-26 22:10:45,717 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=46958.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:10:56,083 INFO [finetune.py:976] (6/7) Epoch 9, batch 1150, loss[loss=0.1958, simple_loss=0.2427, pruned_loss=0.07446, over 3963.00 frames. ], tot_loss[loss=0.1997, simple_loss=0.2657, pruned_loss=0.0668, over 952349.04 frames. ], batch size: 17, lr: 3.80e-03, grad_scale: 32.0 2023-04-26 22:11:00,332 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.2466, 3.1837, 2.4988, 2.6741, 2.3471, 2.6916, 2.6195, 2.0047], device='cuda:6'), covar=tensor([0.2649, 0.1369, 0.1006, 0.1406, 0.3161, 0.1356, 0.2405, 0.3757], device='cuda:6'), in_proj_covar=tensor([0.0301, 0.0321, 0.0230, 0.0293, 0.0319, 0.0274, 0.0261, 0.0286], device='cuda:6'), out_proj_covar=tensor([1.2202e-04, 1.2918e-04, 9.2765e-05, 1.1734e-04, 1.3100e-04, 1.1041e-04, 1.0679e-04, 1.1489e-04], device='cuda:6') 2023-04-26 22:11:01,458 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=46980.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:11:21,913 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5682, 3.4243, 1.0563, 1.7651, 1.9835, 2.4815, 1.9591, 1.0488], device='cuda:6'), covar=tensor([0.1356, 0.1045, 0.1983, 0.1397, 0.1065, 0.1024, 0.1553, 0.1904], device='cuda:6'), in_proj_covar=tensor([0.0118, 0.0254, 0.0143, 0.0124, 0.0137, 0.0156, 0.0120, 0.0122], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-26 22:11:43,956 INFO [finetune.py:976] (6/7) Epoch 9, batch 1200, loss[loss=0.2033, simple_loss=0.2691, pruned_loss=0.06875, over 4780.00 frames. ], tot_loss[loss=0.1966, simple_loss=0.2627, pruned_loss=0.06522, over 953740.54 frames. ], batch size: 51, lr: 3.80e-03, grad_scale: 32.0 2023-04-26 22:12:03,719 INFO [optim.py:369] (6/7) 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 22:12:55,842 INFO [finetune.py:976] (6/7) Epoch 9, batch 1250, loss[loss=0.1552, simple_loss=0.2287, pruned_loss=0.04085, over 4862.00 frames. ], tot_loss[loss=0.193, simple_loss=0.2586, pruned_loss=0.06371, over 955029.06 frames. ], batch size: 31, lr: 3.80e-03, grad_scale: 32.0 2023-04-26 22:13:09,998 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.96 vs. limit=5.0 2023-04-26 22:13:48,239 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7143, 1.3828, 1.8376, 2.2256, 1.8314, 1.6509, 1.7304, 1.7774], device='cuda:6'), covar=tensor([0.6780, 0.9486, 0.9267, 0.9105, 0.8449, 1.0766, 1.1220, 1.0825], device='cuda:6'), in_proj_covar=tensor([0.0407, 0.0418, 0.0501, 0.0521, 0.0436, 0.0453, 0.0466, 0.0463], device='cuda:6'), out_proj_covar=tensor([9.8991e-05, 1.0376e-04, 1.1324e-04, 1.2381e-04, 1.0596e-04, 1.0972e-04, 1.1200e-04, 1.1191e-04], device='cuda:6') 2023-04-26 22:14:00,070 INFO [finetune.py:976] (6/7) Epoch 9, batch 1300, loss[loss=0.1847, simple_loss=0.2437, pruned_loss=0.0628, over 4824.00 frames. ], tot_loss[loss=0.1911, simple_loss=0.2561, pruned_loss=0.06307, over 955973.92 frames. ], batch size: 45, lr: 3.80e-03, grad_scale: 32.0 2023-04-26 22:14:15,264 INFO [optim.py:369] (6/7) 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 22:14:58,579 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 9, batch 1350, loss[loss=0.1605, simple_loss=0.2356, pruned_loss=0.04263, over 4865.00 frames. ], tot_loss[loss=0.1927, simple_loss=0.2578, pruned_loss=0.06383, over 957444.52 frames. ], batch size: 44, lr: 3.80e-03, grad_scale: 32.0 2023-04-26 22:15:58,008 INFO [finetune.py:976] (6/7) Epoch 9, batch 1400, loss[loss=0.2086, simple_loss=0.2821, pruned_loss=0.06754, over 4931.00 frames. ], tot_loss[loss=0.1961, simple_loss=0.2617, pruned_loss=0.06525, over 955646.88 frames. ], batch size: 36, lr: 3.80e-03, grad_scale: 32.0 2023-04-26 22:16:06,778 INFO [optim.py:369] (6/7) 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 22:16:11,711 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=47241.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:16:17,772 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=47251.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:16:22,023 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=47258.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:16:30,886 INFO [finetune.py:976] (6/7) Epoch 9, batch 1450, loss[loss=0.1935, simple_loss=0.258, pruned_loss=0.06454, over 4857.00 frames. ], tot_loss[loss=0.197, simple_loss=0.263, pruned_loss=0.06552, over 955178.99 frames. ], batch size: 31, lr: 3.80e-03, grad_scale: 32.0 2023-04-26 22:16:50,059 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=47299.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:16:51,914 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=47302.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:16:54,272 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=47306.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:17:03,914 INFO [finetune.py:976] (6/7) Epoch 9, batch 1500, loss[loss=0.1979, simple_loss=0.2647, pruned_loss=0.06552, over 4770.00 frames. ], tot_loss[loss=0.1974, simple_loss=0.2635, pruned_loss=0.06569, over 955084.86 frames. ], batch size: 28, lr: 3.80e-03, grad_scale: 32.0 2023-04-26 22:17:12,632 INFO [optim.py:369] (6/7) 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 22:17:22,221 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([4.7045, 4.6032, 3.2005, 5.4520, 4.8096, 4.6944, 2.0910, 4.6499], device='cuda:6'), covar=tensor([0.1650, 0.1106, 0.2980, 0.0790, 0.2577, 0.1697, 0.5544, 0.2205], device='cuda:6'), in_proj_covar=tensor([0.0243, 0.0216, 0.0251, 0.0306, 0.0300, 0.0251, 0.0270, 0.0271], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-26 22:18:03,817 INFO [finetune.py:976] (6/7) Epoch 9, batch 1550, loss[loss=0.1759, simple_loss=0.2363, pruned_loss=0.05772, over 4929.00 frames. ], tot_loss[loss=0.1961, simple_loss=0.2627, pruned_loss=0.0647, over 956045.22 frames. ], batch size: 33, lr: 3.80e-03, grad_scale: 32.0 2023-04-26 22:18:13,631 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.4103, 2.8979, 1.0464, 1.6080, 2.2479, 1.7679, 4.4154, 2.3780], device='cuda:6'), covar=tensor([0.0685, 0.0951, 0.0985, 0.1428, 0.0568, 0.0957, 0.0210, 0.0598], device='cuda:6'), in_proj_covar=tensor([0.0052, 0.0068, 0.0050, 0.0048, 0.0052, 0.0053, 0.0079, 0.0051], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0007], device='cuda:6') 2023-04-26 22:18:13,654 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=47379.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:18:36,066 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0314, 1.8991, 2.1180, 2.4502, 2.4700, 1.9883, 1.5514, 2.1244], device='cuda:6'), covar=tensor([0.0957, 0.1072, 0.0711, 0.0625, 0.0629, 0.0950, 0.0911, 0.0619], device='cuda:6'), in_proj_covar=tensor([0.0197, 0.0205, 0.0182, 0.0177, 0.0179, 0.0191, 0.0161, 0.0186], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-26 22:18:37,269 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6156, 1.5159, 1.7620, 1.9367, 2.0566, 1.5821, 1.2174, 1.7809], device='cuda:6'), covar=tensor([0.0807, 0.1117, 0.0649, 0.0564, 0.0539, 0.0850, 0.0834, 0.0603], device='cuda:6'), in_proj_covar=tensor([0.0197, 0.0205, 0.0182, 0.0177, 0.0178, 0.0191, 0.0161, 0.0186], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-26 22:19:10,272 INFO [finetune.py:976] (6/7) Epoch 9, batch 1600, loss[loss=0.2177, simple_loss=0.2734, pruned_loss=0.08101, over 4769.00 frames. ], tot_loss[loss=0.1939, simple_loss=0.2599, pruned_loss=0.06389, over 958502.95 frames. ], batch size: 26, lr: 3.80e-03, grad_scale: 32.0 2023-04-26 22:19:23,851 INFO [optim.py:369] (6/7) 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 22:19:34,993 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=47440.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:20:14,407 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=47466.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:20:17,939 INFO [finetune.py:976] (6/7) Epoch 9, batch 1650, loss[loss=0.1603, simple_loss=0.2399, pruned_loss=0.04035, over 4767.00 frames. ], tot_loss[loss=0.1916, simple_loss=0.2569, pruned_loss=0.06311, over 956359.38 frames. ], batch size: 26, lr: 3.80e-03, grad_scale: 32.0 2023-04-26 22:21:00,464 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=47514.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:21:05,271 INFO [finetune.py:976] (6/7) Epoch 9, batch 1700, loss[loss=0.2143, simple_loss=0.2731, pruned_loss=0.07779, over 4907.00 frames. ], tot_loss[loss=0.1915, simple_loss=0.2559, pruned_loss=0.06357, over 955486.70 frames. ], batch size: 36, lr: 3.80e-03, grad_scale: 32.0 2023-04-26 22:21:12,557 INFO [optim.py:369] (6/7) 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 22:21:38,484 INFO [finetune.py:976] (6/7) Epoch 9, batch 1750, loss[loss=0.2051, simple_loss=0.2484, pruned_loss=0.08092, over 4013.00 frames. ], tot_loss[loss=0.1945, simple_loss=0.2586, pruned_loss=0.06517, over 953848.72 frames. ], batch size: 17, lr: 3.80e-03, grad_scale: 32.0 2023-04-26 22:21:42,848 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.0691, 3.1656, 1.0129, 1.4752, 1.6212, 2.1658, 1.7358, 0.9379], device='cuda:6'), covar=tensor([0.2072, 0.1901, 0.2416, 0.2038, 0.1441, 0.1472, 0.1871, 0.2312], device='cuda:6'), in_proj_covar=tensor([0.0120, 0.0256, 0.0144, 0.0126, 0.0138, 0.0158, 0.0121, 0.0124], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-26 22:21:55,206 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=47597.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:22:08,354 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5987, 1.7490, 1.0043, 1.2833, 1.8536, 1.4708, 1.3866, 1.4773], device='cuda:6'), covar=tensor([0.0524, 0.0376, 0.0358, 0.0584, 0.0284, 0.0571, 0.0528, 0.0591], device='cuda:6'), in_proj_covar=tensor([0.0030, 0.0026, 0.0024, 0.0030, 0.0021, 0.0030, 0.0029, 0.0030], device='cuda:6'), out_proj_covar=tensor([0.0049, 0.0044, 0.0038, 0.0050, 0.0037, 0.0048, 0.0048, 0.0050], device='cuda:6') 2023-04-26 22:22:11,665 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-04-26 22:22:11,820 INFO [finetune.py:976] (6/7) Epoch 9, batch 1800, loss[loss=0.1765, simple_loss=0.2399, pruned_loss=0.05661, over 4760.00 frames. ], tot_loss[loss=0.196, simple_loss=0.2613, pruned_loss=0.06534, over 955578.88 frames. ], batch size: 28, lr: 3.79e-03, grad_scale: 32.0 2023-04-26 22:22:19,165 INFO [optim.py:369] (6/7) 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 22:22:45,313 INFO [finetune.py:976] (6/7) Epoch 9, batch 1850, loss[loss=0.2233, simple_loss=0.2865, pruned_loss=0.08006, over 4908.00 frames. ], tot_loss[loss=0.1972, simple_loss=0.2625, pruned_loss=0.066, over 955295.27 frames. ], batch size: 36, lr: 3.79e-03, grad_scale: 32.0 2023-04-26 22:23:08,250 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-04-26 22:23:44,012 INFO [finetune.py:976] (6/7) Epoch 9, batch 1900, loss[loss=0.1858, simple_loss=0.241, pruned_loss=0.06532, over 4828.00 frames. ], tot_loss[loss=0.1971, simple_loss=0.2627, pruned_loss=0.06576, over 954034.28 frames. ], batch size: 33, lr: 3.79e-03, grad_scale: 16.0 2023-04-26 22:24:02,108 INFO [optim.py:369] (6/7) 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 22:24:02,206 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=47735.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:24:04,104 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=47738.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:24:28,345 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0847, 2.6542, 1.9369, 2.0202, 1.5276, 1.4978, 2.1721, 1.5038], device='cuda:6'), covar=tensor([0.1533, 0.1563, 0.1542, 0.1851, 0.2298, 0.1830, 0.1026, 0.1942], device='cuda:6'), in_proj_covar=tensor([0.0193, 0.0215, 0.0170, 0.0204, 0.0204, 0.0182, 0.0160, 0.0186], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:6') 2023-04-26 22:24:30,130 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.1990, 2.6125, 0.9659, 1.4256, 1.9020, 1.2226, 3.5159, 1.8818], device='cuda:6'), covar=tensor([0.0626, 0.0808, 0.0876, 0.1159, 0.0521, 0.0965, 0.0191, 0.0559], device='cuda:6'), in_proj_covar=tensor([0.0052, 0.0068, 0.0050, 0.0047, 0.0052, 0.0053, 0.0079, 0.0051], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0007], device='cuda:6') 2023-04-26 22:24:33,493 INFO [finetune.py:976] (6/7) Epoch 9, batch 1950, loss[loss=0.1848, simple_loss=0.2508, pruned_loss=0.05937, over 4835.00 frames. ], tot_loss[loss=0.1964, simple_loss=0.2616, pruned_loss=0.0656, over 955011.96 frames. ], batch size: 47, lr: 3.79e-03, grad_scale: 16.0 2023-04-26 22:24:43,266 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=47788.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:24:50,427 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=47799.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:24:57,519 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=47809.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 22:25:11,664 INFO [finetune.py:976] (6/7) Epoch 9, batch 2000, loss[loss=0.2015, simple_loss=0.2617, pruned_loss=0.07069, over 4922.00 frames. ], tot_loss[loss=0.1953, simple_loss=0.2595, pruned_loss=0.06557, over 955084.95 frames. ], batch size: 36, lr: 3.79e-03, grad_scale: 16.0 2023-04-26 22:25:29,287 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2023-04-26 22:25:30,747 INFO [optim.py:369] (6/7) 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 22:25:50,216 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=47849.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:26:17,415 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=47870.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 22:26:24,628 INFO [finetune.py:976] (6/7) Epoch 9, batch 2050, loss[loss=0.2179, simple_loss=0.2576, pruned_loss=0.08907, over 3967.00 frames. ], tot_loss[loss=0.1925, simple_loss=0.2562, pruned_loss=0.06444, over 953871.37 frames. ], batch size: 17, lr: 3.79e-03, grad_scale: 16.0 2023-04-26 22:26:37,511 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.3354, 1.5898, 1.6340, 1.7997, 1.6896, 1.8797, 1.7831, 1.7646], device='cuda:6'), covar=tensor([0.4888, 0.6576, 0.6084, 0.5763, 0.6694, 0.9556, 0.7179, 0.6170], device='cuda:6'), in_proj_covar=tensor([0.0321, 0.0381, 0.0315, 0.0324, 0.0339, 0.0401, 0.0362, 0.0324], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-26 22:26:57,904 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=47897.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:27:19,792 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.3225, 1.1347, 1.6397, 1.4518, 1.2067, 1.0934, 1.2604, 0.8520], device='cuda:6'), covar=tensor([0.0680, 0.0909, 0.0503, 0.0772, 0.0971, 0.1467, 0.0735, 0.0849], device='cuda:6'), in_proj_covar=tensor([0.0066, 0.0073, 0.0071, 0.0067, 0.0076, 0.0096, 0.0078, 0.0074], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-26 22:27:22,671 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-04-26 22:27:30,785 INFO [finetune.py:976] (6/7) Epoch 9, batch 2100, loss[loss=0.1633, simple_loss=0.2325, pruned_loss=0.04706, over 4932.00 frames. ], tot_loss[loss=0.193, simple_loss=0.2564, pruned_loss=0.06476, over 955348.42 frames. ], batch size: 33, lr: 3.79e-03, grad_scale: 16.0 2023-04-26 22:27:39,268 INFO [optim.py:369] (6/7) 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 22:27:45,521 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=47945.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:28:20,277 INFO [finetune.py:976] (6/7) Epoch 9, batch 2150, loss[loss=0.2291, simple_loss=0.2954, pruned_loss=0.08135, over 4816.00 frames. ], tot_loss[loss=0.1967, simple_loss=0.2605, pruned_loss=0.06642, over 953692.92 frames. ], batch size: 40, lr: 3.79e-03, grad_scale: 16.0 2023-04-26 22:28:20,380 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.4764, 1.4397, 0.7307, 1.1726, 1.6561, 1.3238, 1.2696, 1.3658], device='cuda:6'), covar=tensor([0.0560, 0.0418, 0.0444, 0.0620, 0.0339, 0.0602, 0.0573, 0.0603], device='cuda:6'), in_proj_covar=tensor([0.0030, 0.0026, 0.0023, 0.0030, 0.0021, 0.0029, 0.0029, 0.0030], device='cuda:6'), out_proj_covar=tensor([0.0049, 0.0044, 0.0038, 0.0050, 0.0037, 0.0048, 0.0048, 0.0049], device='cuda:6') 2023-04-26 22:29:17,270 INFO [finetune.py:976] (6/7) Epoch 9, batch 2200, loss[loss=0.1866, simple_loss=0.2512, pruned_loss=0.06103, over 4794.00 frames. ], tot_loss[loss=0.1983, simple_loss=0.2633, pruned_loss=0.06665, over 955463.46 frames. ], batch size: 29, lr: 3.79e-03, grad_scale: 16.0 2023-04-26 22:29:20,756 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8673, 2.3718, 1.9432, 2.1460, 1.6610, 1.8734, 1.9165, 1.5625], device='cuda:6'), covar=tensor([0.2013, 0.1178, 0.0868, 0.1374, 0.3228, 0.1329, 0.1886, 0.2695], device='cuda:6'), in_proj_covar=tensor([0.0299, 0.0319, 0.0230, 0.0293, 0.0317, 0.0273, 0.0261, 0.0284], device='cuda:6'), out_proj_covar=tensor([1.2100e-04, 1.2868e-04, 9.2479e-05, 1.1715e-04, 1.3034e-04, 1.1018e-04, 1.0669e-04, 1.1403e-04], device='cuda:6') 2023-04-26 22:29:32,261 INFO [optim.py:369] (6/7) 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 22:29:32,373 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=48035.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:30:08,349 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-04-26 22:30:13,117 INFO [finetune.py:976] (6/7) Epoch 9, batch 2250, loss[loss=0.2368, simple_loss=0.2993, pruned_loss=0.08715, over 4832.00 frames. ], tot_loss[loss=0.2003, simple_loss=0.2654, pruned_loss=0.06755, over 954124.46 frames. ], batch size: 47, lr: 3.79e-03, grad_scale: 16.0 2023-04-26 22:30:18,388 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=48072.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:30:32,862 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=48083.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:30:45,748 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=48094.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:31:13,247 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.2188, 1.5476, 1.4417, 1.8373, 1.6214, 1.9085, 1.3608, 3.5274], device='cuda:6'), covar=tensor([0.0654, 0.0852, 0.0809, 0.1150, 0.0670, 0.0586, 0.0814, 0.0156], device='cuda:6'), in_proj_covar=tensor([0.0038, 0.0039, 0.0040, 0.0044, 0.0040, 0.0039, 0.0039, 0.0059], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0014, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:6') 2023-04-26 22:31:21,272 INFO [finetune.py:976] (6/7) Epoch 9, batch 2300, loss[loss=0.2113, simple_loss=0.2806, pruned_loss=0.07096, over 4897.00 frames. ], tot_loss[loss=0.1985, simple_loss=0.2649, pruned_loss=0.06609, over 956122.49 frames. ], batch size: 36, lr: 3.79e-03, grad_scale: 16.0 2023-04-26 22:31:35,707 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=48133.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:31:36,819 INFO [optim.py:369] (6/7) 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 22:31:50,261 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=48144.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:32:14,439 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=48165.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 22:32:25,261 INFO [finetune.py:976] (6/7) Epoch 9, batch 2350, loss[loss=0.1808, simple_loss=0.2466, pruned_loss=0.05754, over 4783.00 frames. ], tot_loss[loss=0.1973, simple_loss=0.2629, pruned_loss=0.06591, over 955948.90 frames. ], batch size: 29, lr: 3.79e-03, grad_scale: 16.0 2023-04-26 22:33:19,212 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=48212.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 22:33:30,440 INFO [finetune.py:976] (6/7) Epoch 9, batch 2400, loss[loss=0.1942, simple_loss=0.257, pruned_loss=0.06565, over 4731.00 frames. ], tot_loss[loss=0.1943, simple_loss=0.2591, pruned_loss=0.0648, over 956517.36 frames. ], batch size: 23, lr: 3.79e-03, grad_scale: 16.0 2023-04-26 22:33:45,095 INFO [optim.py:369] (6/7) 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 22:34:23,235 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.10 vs. limit=2.0 2023-04-26 22:34:34,226 INFO [finetune.py:976] (6/7) Epoch 9, batch 2450, loss[loss=0.1968, simple_loss=0.2502, pruned_loss=0.07171, over 4745.00 frames. ], tot_loss[loss=0.1915, simple_loss=0.2557, pruned_loss=0.06369, over 953482.12 frames. ], batch size: 23, lr: 3.79e-03, grad_scale: 16.0 2023-04-26 22:34:34,995 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=48273.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 22:34:37,853 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5291, 1.3610, 1.7861, 1.7323, 1.3411, 1.2132, 1.5409, 1.0091], device='cuda:6'), covar=tensor([0.0650, 0.0905, 0.0565, 0.0848, 0.1046, 0.1417, 0.0736, 0.0954], device='cuda:6'), in_proj_covar=tensor([0.0066, 0.0073, 0.0072, 0.0067, 0.0076, 0.0096, 0.0078, 0.0074], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-26 22:35:25,352 INFO [finetune.py:976] (6/7) Epoch 9, batch 2500, loss[loss=0.1763, simple_loss=0.2534, pruned_loss=0.04959, over 4756.00 frames. ], tot_loss[loss=0.1934, simple_loss=0.2583, pruned_loss=0.06428, over 954913.36 frames. ], batch size: 28, lr: 3.79e-03, grad_scale: 16.0 2023-04-26 22:35:26,636 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=48324.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:35:34,246 INFO [optim.py:369] (6/7) 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 22:36:09,269 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.4939, 1.2711, 0.4795, 1.2233, 1.4532, 1.3761, 1.2860, 1.3083], device='cuda:6'), covar=tensor([0.0551, 0.0434, 0.0447, 0.0605, 0.0305, 0.0571, 0.0554, 0.0651], device='cuda:6'), in_proj_covar=tensor([0.0030, 0.0026, 0.0023, 0.0030, 0.0021, 0.0030, 0.0029, 0.0030], device='cuda:6'), out_proj_covar=tensor([0.0049, 0.0044, 0.0038, 0.0050, 0.0038, 0.0048, 0.0048, 0.0050], device='cuda:6') 2023-04-26 22:36:14,628 INFO [finetune.py:976] (6/7) Epoch 9, batch 2550, loss[loss=0.1699, simple_loss=0.2437, pruned_loss=0.04804, over 4820.00 frames. ], tot_loss[loss=0.1958, simple_loss=0.2618, pruned_loss=0.06493, over 954227.62 frames. ], batch size: 40, lr: 3.79e-03, grad_scale: 16.0 2023-04-26 22:36:34,728 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=48385.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:36:53,744 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=48394.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:36:55,541 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 9, batch 2600, loss[loss=0.1929, simple_loss=0.2695, pruned_loss=0.05817, over 4813.00 frames. ], tot_loss[loss=0.1955, simple_loss=0.2618, pruned_loss=0.0646, over 953674.46 frames. ], batch size: 40, lr: 3.79e-03, grad_scale: 16.0 2023-04-26 22:37:28,734 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=48423.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:37:31,764 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=48428.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:37:41,884 INFO [optim.py:369] (6/7) 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 22:37:52,939 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=48442.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:37:54,693 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=48444.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:38:14,006 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=48458.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:38:24,209 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=48465.0, num_to_drop=1, layers_to_drop={2} 2023-04-26 22:38:34,662 INFO [finetune.py:976] (6/7) Epoch 9, batch 2650, loss[loss=0.1969, simple_loss=0.2631, pruned_loss=0.06539, over 4847.00 frames. ], tot_loss[loss=0.1964, simple_loss=0.2631, pruned_loss=0.06487, over 954597.60 frames. ], batch size: 44, lr: 3.79e-03, grad_scale: 16.0 2023-04-26 22:38:46,954 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=48484.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:38:57,603 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=48492.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:38:58,546 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.12 vs. limit=2.0 2023-04-26 22:39:26,506 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=48513.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 22:39:32,437 INFO [finetune.py:976] (6/7) Epoch 9, batch 2700, loss[loss=0.2072, simple_loss=0.2738, pruned_loss=0.07031, over 4901.00 frames. ], tot_loss[loss=0.1955, simple_loss=0.262, pruned_loss=0.06446, over 954628.97 frames. ], batch size: 36, lr: 3.79e-03, grad_scale: 16.0 2023-04-26 22:39:33,801 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-04-26 22:39:40,875 INFO [optim.py:369] (6/7) 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 22:40:07,257 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0928, 2.9833, 2.2898, 2.6813, 2.1494, 2.4463, 2.4914, 1.9853], device='cuda:6'), covar=tensor([0.2450, 0.1140, 0.0952, 0.1229, 0.3050, 0.1346, 0.2037, 0.2790], device='cuda:6'), in_proj_covar=tensor([0.0297, 0.0318, 0.0228, 0.0290, 0.0314, 0.0272, 0.0259, 0.0282], device='cuda:6'), out_proj_covar=tensor([1.2009e-04, 1.2804e-04, 9.1928e-05, 1.1619e-04, 1.2909e-04, 1.0955e-04, 1.0595e-04, 1.1299e-04], device='cuda:6') 2023-04-26 22:40:20,128 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=48568.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 22:40:28,390 INFO [finetune.py:976] (6/7) Epoch 9, batch 2750, loss[loss=0.1742, simple_loss=0.2361, pruned_loss=0.05617, over 4713.00 frames. ], tot_loss[loss=0.1926, simple_loss=0.2589, pruned_loss=0.06316, over 952179.95 frames. ], batch size: 23, lr: 3.79e-03, grad_scale: 16.0 2023-04-26 22:40:51,679 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.3960, 1.5355, 1.6376, 1.7926, 1.6592, 1.8508, 1.7996, 1.7290], device='cuda:6'), covar=tensor([0.4724, 0.7181, 0.5981, 0.5690, 0.7217, 0.9851, 0.6745, 0.6111], device='cuda:6'), in_proj_covar=tensor([0.0321, 0.0382, 0.0315, 0.0325, 0.0342, 0.0402, 0.0362, 0.0324], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-26 22:41:33,773 INFO [finetune.py:976] (6/7) Epoch 9, batch 2800, loss[loss=0.1789, simple_loss=0.2406, pruned_loss=0.05863, over 4830.00 frames. ], tot_loss[loss=0.1889, simple_loss=0.2551, pruned_loss=0.06139, over 953316.17 frames. ], batch size: 30, lr: 3.79e-03, grad_scale: 16.0 2023-04-26 22:41:46,304 INFO [optim.py:369] (6/7) 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 22:42:28,877 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.0498, 2.1039, 1.9014, 1.7611, 2.3715, 1.8181, 2.9343, 1.6804], device='cuda:6'), covar=tensor([0.4146, 0.1884, 0.4787, 0.3263, 0.1731, 0.2890, 0.1192, 0.4958], device='cuda:6'), in_proj_covar=tensor([0.0348, 0.0352, 0.0433, 0.0364, 0.0392, 0.0386, 0.0382, 0.0423], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-26 22:42:38,718 INFO [finetune.py:976] (6/7) Epoch 9, batch 2850, loss[loss=0.1821, simple_loss=0.246, pruned_loss=0.05912, over 4905.00 frames. ], tot_loss[loss=0.1886, simple_loss=0.2539, pruned_loss=0.06166, over 952517.76 frames. ], batch size: 32, lr: 3.79e-03, grad_scale: 16.0 2023-04-26 22:42:49,429 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=48680.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:43:00,136 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5970, 1.1462, 1.6587, 2.0819, 1.7531, 1.6070, 1.6111, 1.6506], device='cuda:6'), covar=tensor([0.6359, 0.8595, 0.8691, 0.8753, 0.7574, 1.0636, 1.0435, 0.9495], device='cuda:6'), in_proj_covar=tensor([0.0409, 0.0421, 0.0504, 0.0522, 0.0438, 0.0458, 0.0469, 0.0466], device='cuda:6'), out_proj_covar=tensor([9.9549e-05, 1.0445e-04, 1.1384e-04, 1.2413e-04, 1.0628e-04, 1.1068e-04, 1.1285e-04, 1.1269e-04], device='cuda:6') 2023-04-26 22:43:44,379 INFO [finetune.py:976] (6/7) Epoch 9, batch 2900, loss[loss=0.2092, simple_loss=0.2916, pruned_loss=0.06339, over 4901.00 frames. ], tot_loss[loss=0.1919, simple_loss=0.2576, pruned_loss=0.06311, over 953687.00 frames. ], batch size: 43, lr: 3.79e-03, grad_scale: 16.0 2023-04-26 22:43:48,148 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=48728.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:43:52,844 INFO [optim.py:369] (6/7) 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 22:44:04,272 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=48753.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:44:17,830 INFO [finetune.py:976] (6/7) Epoch 9, batch 2950, loss[loss=0.2967, simple_loss=0.3377, pruned_loss=0.1278, over 4147.00 frames. ], tot_loss[loss=0.196, simple_loss=0.2615, pruned_loss=0.06523, over 952171.37 frames. ], batch size: 67, lr: 3.79e-03, grad_scale: 16.0 2023-04-26 22:44:20,347 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=48776.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:44:22,240 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=48779.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:45:10,561 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=48821.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:45:11,039 INFO [finetune.py:976] (6/7) Epoch 9, batch 3000, loss[loss=0.167, simple_loss=0.2411, pruned_loss=0.04649, over 4722.00 frames. ], tot_loss[loss=0.197, simple_loss=0.263, pruned_loss=0.06549, over 953388.31 frames. ], batch size: 59, lr: 3.79e-03, grad_scale: 16.0 2023-04-26 22:45:11,039 INFO [finetune.py:1001] (6/7) Computing validation loss 2023-04-26 22:45:16,505 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.3144, 1.3265, 3.7778, 3.4572, 3.3793, 3.6611, 3.7076, 3.3686], device='cuda:6'), covar=tensor([0.7308, 0.5609, 0.1216, 0.1889, 0.1394, 0.1504, 0.0858, 0.1671], device='cuda:6'), in_proj_covar=tensor([0.0312, 0.0310, 0.0410, 0.0413, 0.0353, 0.0407, 0.0317, 0.0372], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-26 22:45:27,358 INFO [finetune.py:1010] (6/7) 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] (6/7) Maximum memory allocated so far is 6345MB 2023-04-26 22:45:46,032 INFO [optim.py:369] (6/7) 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 22:45:47,978 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=48838.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:45:56,964 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.1264, 2.1846, 1.7519, 1.8001, 2.3324, 1.8509, 2.8202, 1.5968], device='cuda:6'), covar=tensor([0.4209, 0.1825, 0.5260, 0.4045, 0.1845, 0.2940, 0.1459, 0.4989], device='cuda:6'), in_proj_covar=tensor([0.0350, 0.0354, 0.0436, 0.0365, 0.0394, 0.0388, 0.0385, 0.0425], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-26 22:46:30,388 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=48868.0, num_to_drop=1, layers_to_drop={2} 2023-04-26 22:46:32,698 INFO [finetune.py:976] (6/7) Epoch 9, batch 3050, loss[loss=0.1497, simple_loss=0.2248, pruned_loss=0.03729, over 4895.00 frames. ], tot_loss[loss=0.1961, simple_loss=0.2628, pruned_loss=0.06464, over 953399.97 frames. ], batch size: 32, lr: 3.79e-03, grad_scale: 16.0 2023-04-26 22:46:51,113 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=48882.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:47:12,130 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=48899.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:47:28,838 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=48916.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 22:47:32,442 INFO [finetune.py:976] (6/7) Epoch 9, batch 3100, loss[loss=0.2278, simple_loss=0.2779, pruned_loss=0.0888, over 4876.00 frames. ], tot_loss[loss=0.1961, simple_loss=0.2618, pruned_loss=0.06516, over 954822.52 frames. ], batch size: 31, lr: 3.79e-03, grad_scale: 16.0 2023-04-26 22:47:42,242 INFO [optim.py:369] (6/7) 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 22:48:00,739 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.48 vs. limit=2.0 2023-04-26 22:48:04,667 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.27 vs. limit=5.0 2023-04-26 22:48:05,675 INFO [finetune.py:976] (6/7) Epoch 9, batch 3150, loss[loss=0.1839, simple_loss=0.2523, pruned_loss=0.05775, over 4787.00 frames. ], tot_loss[loss=0.193, simple_loss=0.2587, pruned_loss=0.06365, over 955986.91 frames. ], batch size: 26, lr: 3.78e-03, grad_scale: 16.0 2023-04-26 22:48:11,548 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=48980.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:48:38,604 INFO [finetune.py:976] (6/7) Epoch 9, batch 3200, loss[loss=0.1554, simple_loss=0.2272, pruned_loss=0.04184, over 4768.00 frames. ], tot_loss[loss=0.1902, simple_loss=0.2558, pruned_loss=0.06229, over 958853.99 frames. ], batch size: 26, lr: 3.78e-03, grad_scale: 16.0 2023-04-26 22:48:42,810 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=49028.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:48:47,940 INFO [optim.py:369] (6/7) 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 22:48:59,931 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 9, batch 3250, loss[loss=0.1749, simple_loss=0.2553, pruned_loss=0.04731, over 4928.00 frames. ], tot_loss[loss=0.1904, simple_loss=0.2557, pruned_loss=0.06258, over 957248.63 frames. ], batch size: 38, lr: 3.78e-03, grad_scale: 16.0 2023-04-26 22:49:16,898 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=49079.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:49:25,074 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.63 vs. limit=5.0 2023-04-26 22:49:32,077 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=49101.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:49:45,282 INFO [finetune.py:976] (6/7) Epoch 9, batch 3300, loss[loss=0.2555, simple_loss=0.3123, pruned_loss=0.09932, over 4207.00 frames. ], tot_loss[loss=0.1943, simple_loss=0.2604, pruned_loss=0.06408, over 956873.51 frames. ], batch size: 65, lr: 3.78e-03, grad_scale: 16.0 2023-04-26 22:49:45,420 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([0.8243, 1.8809, 1.9274, 1.4724, 2.1774, 1.6415, 2.7089, 1.6647], device='cuda:6'), covar=tensor([0.3779, 0.1570, 0.4381, 0.2794, 0.1482, 0.2501, 0.1504, 0.4249], device='cuda:6'), in_proj_covar=tensor([0.0347, 0.0349, 0.0433, 0.0362, 0.0390, 0.0384, 0.0381, 0.0422], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-26 22:49:48,902 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=49127.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:49:54,142 INFO [optim.py:369] (6/7) 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 22:49:59,584 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.12 vs. limit=5.0 2023-04-26 22:50:04,879 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-04-26 22:50:18,719 INFO [finetune.py:976] (6/7) Epoch 9, batch 3350, loss[loss=0.1854, simple_loss=0.2515, pruned_loss=0.05961, over 4762.00 frames. ], tot_loss[loss=0.1954, simple_loss=0.2616, pruned_loss=0.06456, over 956405.30 frames. ], batch size: 27, lr: 3.78e-03, grad_scale: 16.0 2023-04-26 22:50:20,345 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-04-26 22:50:21,872 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=49177.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:50:38,277 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.60 vs. limit=2.0 2023-04-26 22:50:45,208 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6696, 1.0171, 1.5947, 2.1179, 1.8020, 1.6212, 1.6112, 1.7040], device='cuda:6'), covar=tensor([0.6076, 0.8061, 0.7954, 0.8360, 0.7376, 1.0205, 1.0305, 0.8393], device='cuda:6'), in_proj_covar=tensor([0.0409, 0.0421, 0.0502, 0.0523, 0.0438, 0.0457, 0.0468, 0.0464], device='cuda:6'), out_proj_covar=tensor([9.9418e-05, 1.0417e-04, 1.1343e-04, 1.2404e-04, 1.0618e-04, 1.1043e-04, 1.1241e-04, 1.1237e-04], device='cuda:6') 2023-04-26 22:50:46,869 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=49194.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:51:00,031 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([3.9745, 3.9874, 2.8055, 4.6232, 4.0668, 4.0332, 1.7915, 3.9997], device='cuda:6'), covar=tensor([0.1829, 0.1170, 0.2963, 0.1481, 0.3602, 0.1721, 0.5545, 0.2469], device='cuda:6'), in_proj_covar=tensor([0.0244, 0.0217, 0.0251, 0.0306, 0.0302, 0.0252, 0.0271, 0.0272], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-26 22:51:19,535 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7271, 1.3993, 0.6986, 1.3733, 1.6052, 1.5584, 1.4589, 1.4252], device='cuda:6'), covar=tensor([0.0510, 0.0439, 0.0442, 0.0592, 0.0307, 0.0560, 0.0576, 0.0589], device='cuda:6'), in_proj_covar=tensor([0.0030, 0.0026, 0.0024, 0.0030, 0.0021, 0.0029, 0.0029, 0.0030], device='cuda:6'), out_proj_covar=tensor([0.0049, 0.0045, 0.0038, 0.0050, 0.0038, 0.0048, 0.0048, 0.0050], device='cuda:6') 2023-04-26 22:51:20,618 INFO [finetune.py:976] (6/7) Epoch 9, batch 3400, loss[loss=0.1616, simple_loss=0.2353, pruned_loss=0.04397, over 4733.00 frames. ], tot_loss[loss=0.1954, simple_loss=0.262, pruned_loss=0.0644, over 956027.83 frames. ], batch size: 59, lr: 3.78e-03, grad_scale: 16.0 2023-04-26 22:51:38,493 INFO [optim.py:369] (6/7) 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 22:51:39,220 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=49236.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:51:53,740 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7526, 1.3561, 1.2955, 1.5195, 1.9531, 1.5366, 1.2706, 1.2776], device='cuda:6'), covar=tensor([0.1583, 0.1578, 0.2229, 0.1513, 0.0992, 0.1570, 0.2162, 0.2112], device='cuda:6'), in_proj_covar=tensor([0.0308, 0.0324, 0.0354, 0.0296, 0.0335, 0.0321, 0.0307, 0.0355], device='cuda:6'), out_proj_covar=tensor([6.5006e-05, 6.8722e-05, 7.6320e-05, 6.1246e-05, 7.0316e-05, 6.9002e-05, 6.6136e-05, 7.6297e-05], device='cuda:6') 2023-04-26 22:51:59,880 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([3.9514, 3.9540, 2.8251, 4.6152, 4.0326, 4.0391, 1.7582, 3.9507], device='cuda:6'), covar=tensor([0.1722, 0.1076, 0.2907, 0.1388, 0.3453, 0.1891, 0.5664, 0.2237], device='cuda:6'), in_proj_covar=tensor([0.0245, 0.0219, 0.0253, 0.0308, 0.0304, 0.0253, 0.0272, 0.0273], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-26 22:51:59,937 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0523, 2.5715, 2.0384, 2.4122, 1.9173, 2.0795, 2.0773, 1.6641], device='cuda:6'), covar=tensor([0.1994, 0.1385, 0.0973, 0.1242, 0.3249, 0.1638, 0.2077, 0.3057], device='cuda:6'), in_proj_covar=tensor([0.0295, 0.0315, 0.0227, 0.0287, 0.0311, 0.0270, 0.0256, 0.0280], device='cuda:6'), out_proj_covar=tensor([1.1941e-04, 1.2710e-04, 9.1349e-05, 1.1507e-04, 1.2795e-04, 1.0904e-04, 1.0502e-04, 1.1248e-04], device='cuda:6') 2023-04-26 22:52:05,151 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.50 vs. limit=2.0 2023-04-26 22:52:13,320 INFO [finetune.py:976] (6/7) Epoch 9, batch 3450, loss[loss=0.2111, simple_loss=0.2557, pruned_loss=0.08325, over 4701.00 frames. ], tot_loss[loss=0.195, simple_loss=0.2617, pruned_loss=0.06412, over 955450.46 frames. ], batch size: 23, lr: 3.78e-03, grad_scale: 16.0 2023-04-26 22:52:14,702 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8273, 2.2746, 1.8906, 2.1141, 1.7531, 1.7901, 1.8391, 1.4308], device='cuda:6'), covar=tensor([0.1798, 0.1276, 0.0955, 0.1078, 0.3178, 0.1408, 0.1902, 0.2600], device='cuda:6'), in_proj_covar=tensor([0.0296, 0.0316, 0.0227, 0.0288, 0.0312, 0.0271, 0.0257, 0.0281], device='cuda:6'), out_proj_covar=tensor([1.1978e-04, 1.2748e-04, 9.1471e-05, 1.1541e-04, 1.2826e-04, 1.0932e-04, 1.0520e-04, 1.1276e-04], device='cuda:6') 2023-04-26 22:52:30,781 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=49297.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:52:53,711 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([3.5984, 3.6025, 2.7687, 4.2172, 3.7144, 3.6428, 1.6495, 3.6210], device='cuda:6'), covar=tensor([0.1579, 0.1188, 0.2669, 0.1702, 0.2841, 0.1806, 0.5287, 0.2143], device='cuda:6'), in_proj_covar=tensor([0.0244, 0.0217, 0.0251, 0.0306, 0.0302, 0.0252, 0.0271, 0.0272], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-26 22:53:02,574 INFO [finetune.py:976] (6/7) Epoch 9, batch 3500, loss[loss=0.2114, simple_loss=0.2688, pruned_loss=0.07697, over 4847.00 frames. ], tot_loss[loss=0.1921, simple_loss=0.2583, pruned_loss=0.06292, over 954955.51 frames. ], batch size: 47, lr: 3.78e-03, grad_scale: 16.0 2023-04-26 22:53:12,844 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.3384, 1.9927, 2.2726, 2.5972, 2.5156, 2.0989, 1.6626, 2.2697], device='cuda:6'), covar=tensor([0.0814, 0.1063, 0.0621, 0.0587, 0.0605, 0.0907, 0.0805, 0.0647], device='cuda:6'), in_proj_covar=tensor([0.0197, 0.0205, 0.0182, 0.0177, 0.0179, 0.0190, 0.0160, 0.0186], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-26 22:53:15,826 INFO [optim.py:369] (6/7) 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 22:53:17,738 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5841, 1.5037, 1.8782, 1.8647, 1.4360, 1.2518, 1.5369, 1.0870], device='cuda:6'), covar=tensor([0.0761, 0.0944, 0.0556, 0.0867, 0.1097, 0.1429, 0.0920, 0.0957], device='cuda:6'), in_proj_covar=tensor([0.0067, 0.0074, 0.0072, 0.0067, 0.0077, 0.0097, 0.0078, 0.0074], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-26 22:53:38,333 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.2341, 1.7127, 1.4423, 1.8374, 1.6846, 1.9403, 1.4205, 3.5484], device='cuda:6'), covar=tensor([0.0702, 0.0775, 0.0818, 0.1185, 0.0646, 0.0529, 0.0752, 0.0138], device='cuda:6'), in_proj_covar=tensor([0.0038, 0.0039, 0.0040, 0.0044, 0.0040, 0.0039, 0.0039, 0.0059], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0014, 0.0013, 0.0012, 0.0013, 0.0016], device='cuda:6') 2023-04-26 22:54:00,998 INFO [finetune.py:976] (6/7) Epoch 9, batch 3550, loss[loss=0.1806, simple_loss=0.2345, pruned_loss=0.06342, over 4822.00 frames. ], tot_loss[loss=0.1906, simple_loss=0.2564, pruned_loss=0.06241, over 957682.41 frames. ], batch size: 39, lr: 3.78e-03, grad_scale: 16.0 2023-04-26 22:54:42,415 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5934, 0.6412, 1.3490, 1.9290, 1.6472, 1.5001, 1.4854, 1.5403], device='cuda:6'), covar=tensor([0.5645, 0.8120, 0.8400, 0.8603, 0.7336, 0.9123, 0.9250, 0.8679], device='cuda:6'), in_proj_covar=tensor([0.0409, 0.0420, 0.0504, 0.0522, 0.0438, 0.0457, 0.0467, 0.0466], device='cuda:6'), out_proj_covar=tensor([9.9408e-05, 1.0409e-04, 1.1372e-04, 1.2396e-04, 1.0630e-04, 1.1056e-04, 1.1234e-04, 1.1270e-04], device='cuda:6') 2023-04-26 22:55:07,558 INFO [finetune.py:976] (6/7) Epoch 9, batch 3600, loss[loss=0.2182, simple_loss=0.2791, pruned_loss=0.07867, over 4903.00 frames. ], tot_loss[loss=0.19, simple_loss=0.255, pruned_loss=0.06253, over 958116.54 frames. ], batch size: 37, lr: 3.78e-03, grad_scale: 16.0 2023-04-26 22:55:12,910 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=49422.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:55:24,161 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=49432.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 22:55:25,888 INFO [optim.py:369] (6/7) 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 22:56:18,900 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.81 vs. limit=5.0 2023-04-26 22:56:18,974 INFO [finetune.py:976] (6/7) Epoch 9, batch 3650, loss[loss=0.1773, simple_loss=0.2608, pruned_loss=0.04693, over 4791.00 frames. ], tot_loss[loss=0.1923, simple_loss=0.2573, pruned_loss=0.06367, over 956229.20 frames. ], batch size: 29, lr: 3.78e-03, grad_scale: 16.0 2023-04-26 22:56:19,073 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.3449, 1.3848, 4.0709, 3.7987, 3.5573, 3.8541, 3.8057, 3.5695], device='cuda:6'), covar=tensor([0.7490, 0.5895, 0.1159, 0.1943, 0.1205, 0.1872, 0.1769, 0.1752], device='cuda:6'), in_proj_covar=tensor([0.0312, 0.0309, 0.0408, 0.0413, 0.0353, 0.0405, 0.0317, 0.0372], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-26 22:56:22,163 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=49477.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:56:30,060 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=49483.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:56:36,681 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=49493.0, num_to_drop=1, layers_to_drop={3} 2023-04-26 22:56:37,252 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=49494.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:56:56,125 INFO [finetune.py:976] (6/7) Epoch 9, batch 3700, loss[loss=0.2178, simple_loss=0.2743, pruned_loss=0.08069, over 4814.00 frames. ], tot_loss[loss=0.1955, simple_loss=0.261, pruned_loss=0.06498, over 954394.01 frames. ], batch size: 30, lr: 3.78e-03, grad_scale: 16.0 2023-04-26 22:56:58,029 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=49525.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:57:01,755 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5915, 1.2610, 1.7673, 2.0502, 1.7078, 1.5852, 1.6958, 1.6477], device='cuda:6'), covar=tensor([0.6084, 0.8421, 0.8165, 0.8295, 0.7397, 1.0061, 1.0060, 0.9673], device='cuda:6'), in_proj_covar=tensor([0.0409, 0.0421, 0.0506, 0.0524, 0.0439, 0.0458, 0.0470, 0.0468], device='cuda:6'), out_proj_covar=tensor([9.9513e-05, 1.0424e-04, 1.1414e-04, 1.2456e-04, 1.0658e-04, 1.1085e-04, 1.1298e-04, 1.1313e-04], device='cuda:6') 2023-04-26 22:57:09,399 INFO [optim.py:369] (6/7) 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 22:57:13,693 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=49542.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:57:56,807 INFO [finetune.py:976] (6/7) Epoch 9, batch 3750, loss[loss=0.2077, simple_loss=0.2639, pruned_loss=0.07577, over 4854.00 frames. ], tot_loss[loss=0.1986, simple_loss=0.2641, pruned_loss=0.06655, over 955851.00 frames. ], batch size: 31, lr: 3.78e-03, grad_scale: 16.0 2023-04-26 22:58:23,857 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=49592.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 22:58:44,874 INFO [finetune.py:976] (6/7) Epoch 9, batch 3800, loss[loss=0.2001, simple_loss=0.2726, pruned_loss=0.06379, over 4820.00 frames. ], tot_loss[loss=0.1994, simple_loss=0.2649, pruned_loss=0.06689, over 954353.15 frames. ], batch size: 33, lr: 3.78e-03, grad_scale: 16.0 2023-04-26 22:58:50,082 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.99 vs. limit=2.0 2023-04-26 22:58:52,835 INFO [optim.py:369] (6/7) 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 22:59:17,904 INFO [finetune.py:976] (6/7) Epoch 9, batch 3850, loss[loss=0.2254, simple_loss=0.286, pruned_loss=0.08243, over 4204.00 frames. ], tot_loss[loss=0.1971, simple_loss=0.2628, pruned_loss=0.06573, over 954814.82 frames. ], batch size: 65, lr: 3.78e-03, grad_scale: 16.0 2023-04-26 22:59:22,264 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([0.8929, 2.0195, 1.7961, 1.5094, 2.1507, 1.8089, 2.6761, 1.6119], device='cuda:6'), covar=tensor([0.3900, 0.1598, 0.4604, 0.3556, 0.1714, 0.2480, 0.1174, 0.4681], device='cuda:6'), in_proj_covar=tensor([0.0347, 0.0350, 0.0434, 0.0365, 0.0391, 0.0385, 0.0383, 0.0422], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-26 22:59:42,998 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.43 vs. limit=5.0 2023-04-26 22:59:44,594 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.1847, 1.8505, 1.9989, 2.6049, 2.1366, 1.7055, 1.5955, 1.9864], device='cuda:6'), covar=tensor([0.2736, 0.3195, 0.1686, 0.2078, 0.2787, 0.2600, 0.4783, 0.2496], device='cuda:6'), in_proj_covar=tensor([0.0281, 0.0251, 0.0219, 0.0318, 0.0213, 0.0228, 0.0233, 0.0186], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:6') 2023-04-26 22:59:49,729 INFO [finetune.py:976] (6/7) Epoch 9, batch 3900, loss[loss=0.2023, simple_loss=0.26, pruned_loss=0.07232, over 4760.00 frames. ], tot_loss[loss=0.1947, simple_loss=0.2601, pruned_loss=0.0646, over 955609.04 frames. ], batch size: 27, lr: 3.78e-03, grad_scale: 32.0 2023-04-26 22:59:58,108 INFO [optim.py:369] (6/7) 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 23:00:04,893 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([4.3466, 4.1280, 3.0917, 5.0095, 4.2840, 4.4064, 1.7919, 4.2719], device='cuda:6'), covar=tensor([0.1538, 0.0974, 0.3781, 0.0889, 0.2820, 0.1639, 0.6220, 0.2226], device='cuda:6'), in_proj_covar=tensor([0.0245, 0.0218, 0.0252, 0.0306, 0.0303, 0.0253, 0.0273, 0.0273], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-26 23:00:16,195 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0079, 1.3680, 4.9733, 4.6428, 4.3444, 4.6976, 4.3127, 4.4242], device='cuda:6'), covar=tensor([0.6473, 0.6311, 0.1044, 0.1915, 0.1072, 0.1681, 0.1998, 0.1720], device='cuda:6'), in_proj_covar=tensor([0.0307, 0.0305, 0.0402, 0.0408, 0.0347, 0.0400, 0.0312, 0.0366], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-26 23:00:18,056 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=49766.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:00:21,573 INFO [finetune.py:976] (6/7) Epoch 9, batch 3950, loss[loss=0.1782, simple_loss=0.2389, pruned_loss=0.05876, over 4880.00 frames. ], tot_loss[loss=0.1919, simple_loss=0.2568, pruned_loss=0.06351, over 953803.48 frames. ], batch size: 31, lr: 3.78e-03, grad_scale: 32.0 2023-04-26 23:00:27,204 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=49778.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:00:33,307 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=49788.0, num_to_drop=1, layers_to_drop={3} 2023-04-26 23:00:48,271 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.0916, 1.5794, 1.3142, 1.7654, 1.6213, 1.7590, 1.3486, 3.2967], device='cuda:6'), covar=tensor([0.0681, 0.0737, 0.0828, 0.1161, 0.0610, 0.0579, 0.0737, 0.0157], device='cuda:6'), in_proj_covar=tensor([0.0038, 0.0039, 0.0040, 0.0044, 0.0040, 0.0039, 0.0039, 0.0059], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0014, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:6') 2023-04-26 23:00:49,670 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 2023-04-26 23:00:55,288 INFO [finetune.py:976] (6/7) Epoch 9, batch 4000, loss[loss=0.2461, simple_loss=0.306, pruned_loss=0.09312, over 4764.00 frames. ], tot_loss[loss=0.1923, simple_loss=0.2566, pruned_loss=0.06396, over 951658.21 frames. ], batch size: 54, lr: 3.78e-03, grad_scale: 32.0 2023-04-26 23:00:59,441 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=49827.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:01:03,451 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.1960, 2.8432, 2.2102, 2.2700, 1.5647, 1.5517, 2.3320, 1.5583], device='cuda:6'), covar=tensor([0.1790, 0.1720, 0.1520, 0.1833, 0.2593, 0.2158, 0.1205, 0.2207], device='cuda:6'), in_proj_covar=tensor([0.0193, 0.0216, 0.0170, 0.0205, 0.0205, 0.0183, 0.0160, 0.0188], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:6') 2023-04-26 23:01:05,132 INFO [optim.py:369] (6/7) 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 23:01:07,042 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([3.9614, 3.9069, 2.8076, 4.6496, 4.0435, 3.9858, 2.0580, 4.0042], device='cuda:6'), covar=tensor([0.1645, 0.1121, 0.3071, 0.1409, 0.2745, 0.1759, 0.5367, 0.2163], device='cuda:6'), in_proj_covar=tensor([0.0243, 0.0216, 0.0249, 0.0303, 0.0300, 0.0250, 0.0269, 0.0271], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-26 23:01:42,674 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-04-26 23:01:44,301 INFO [finetune.py:976] (6/7) Epoch 9, batch 4050, loss[loss=0.2073, simple_loss=0.2788, pruned_loss=0.06783, over 4900.00 frames. ], tot_loss[loss=0.1945, simple_loss=0.2596, pruned_loss=0.06476, over 951824.39 frames. ], batch size: 35, lr: 3.78e-03, grad_scale: 32.0 2023-04-26 23:02:14,668 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=49892.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:02:25,730 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9830, 1.6724, 1.9342, 2.1962, 2.3097, 1.8448, 1.6343, 2.0311], device='cuda:6'), covar=tensor([0.0841, 0.1189, 0.0691, 0.0647, 0.0560, 0.0891, 0.0841, 0.0572], device='cuda:6'), in_proj_covar=tensor([0.0195, 0.0205, 0.0181, 0.0175, 0.0177, 0.0189, 0.0159, 0.0183], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-26 23:02:49,743 INFO [finetune.py:976] (6/7) Epoch 9, batch 4100, loss[loss=0.1615, simple_loss=0.2176, pruned_loss=0.0527, over 4028.00 frames. ], tot_loss[loss=0.1953, simple_loss=0.2607, pruned_loss=0.06492, over 952387.67 frames. ], batch size: 17, lr: 3.78e-03, grad_scale: 32.0 2023-04-26 23:03:10,346 INFO [optim.py:369] (6/7) 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 23:03:19,227 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=49940.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:03:53,585 INFO [finetune.py:976] (6/7) Epoch 9, batch 4150, loss[loss=0.2292, simple_loss=0.301, pruned_loss=0.07868, over 4907.00 frames. ], tot_loss[loss=0.1971, simple_loss=0.2627, pruned_loss=0.06573, over 951725.32 frames. ], batch size: 37, lr: 3.78e-03, grad_scale: 32.0 2023-04-26 23:04:33,305 INFO [finetune.py:976] (6/7) Epoch 9, batch 4200, loss[loss=0.1973, simple_loss=0.261, pruned_loss=0.06683, over 4803.00 frames. ], tot_loss[loss=0.197, simple_loss=0.2629, pruned_loss=0.06562, over 951739.32 frames. ], batch size: 41, lr: 3.78e-03, grad_scale: 32.0 2023-04-26 23:04:41,638 INFO [optim.py:369] (6/7) 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 23:04:43,318 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.5181, 1.4319, 1.5371, 2.2331, 2.3205, 1.9101, 1.8635, 1.5703], device='cuda:6'), covar=tensor([0.1512, 0.2305, 0.2276, 0.1513, 0.1353, 0.2306, 0.2425, 0.2369], device='cuda:6'), in_proj_covar=tensor([0.0305, 0.0320, 0.0351, 0.0295, 0.0331, 0.0319, 0.0304, 0.0354], device='cuda:6'), out_proj_covar=tensor([6.4482e-05, 6.7865e-05, 7.5670e-05, 6.0831e-05, 6.9366e-05, 6.8483e-05, 6.5442e-05, 7.5915e-05], device='cuda:6') 2023-04-26 23:04:53,759 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-04-26 23:04:56,006 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6260, 1.4251, 4.5350, 4.2378, 3.9767, 4.2731, 4.0861, 3.9824], device='cuda:6'), covar=tensor([0.6957, 0.5918, 0.0909, 0.1582, 0.1059, 0.1538, 0.1246, 0.1448], device='cuda:6'), in_proj_covar=tensor([0.0311, 0.0309, 0.0405, 0.0411, 0.0351, 0.0405, 0.0316, 0.0370], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-26 23:05:01,570 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8833, 1.3614, 1.6595, 1.5444, 1.6158, 1.3386, 0.7677, 1.3485], device='cuda:6'), covar=tensor([0.3711, 0.4133, 0.2108, 0.2948, 0.3263, 0.3129, 0.4770, 0.2772], device='cuda:6'), in_proj_covar=tensor([0.0281, 0.0250, 0.0219, 0.0317, 0.0213, 0.0227, 0.0232, 0.0186], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:6') 2023-04-26 23:05:05,672 INFO [finetune.py:976] (6/7) Epoch 9, batch 4250, loss[loss=0.209, simple_loss=0.2701, pruned_loss=0.07392, over 4908.00 frames. ], tot_loss[loss=0.1955, simple_loss=0.2606, pruned_loss=0.06516, over 952543.51 frames. ], batch size: 36, lr: 3.78e-03, grad_scale: 32.0 2023-04-26 23:05:09,941 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=50078.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:05:16,026 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=50088.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 23:05:35,866 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.2674, 1.3366, 1.4588, 1.5681, 1.6737, 1.2381, 1.0067, 1.4618], device='cuda:6'), covar=tensor([0.0955, 0.1358, 0.0911, 0.0711, 0.0717, 0.0983, 0.0899, 0.0671], device='cuda:6'), in_proj_covar=tensor([0.0197, 0.0206, 0.0182, 0.0177, 0.0178, 0.0189, 0.0160, 0.0184], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-26 23:05:37,583 INFO [finetune.py:976] (6/7) Epoch 9, batch 4300, loss[loss=0.1519, simple_loss=0.2185, pruned_loss=0.04267, over 4822.00 frames. ], tot_loss[loss=0.1927, simple_loss=0.2577, pruned_loss=0.06383, over 952902.02 frames. ], batch size: 40, lr: 3.78e-03, grad_scale: 32.0 2023-04-26 23:05:37,655 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=50122.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:05:40,101 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=50126.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:05:41,358 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=50128.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:05:46,444 INFO [optim.py:369] (6/7) 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 23:05:47,119 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=50136.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 23:06:09,489 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-04-26 23:06:10,818 INFO [finetune.py:976] (6/7) Epoch 9, batch 4350, loss[loss=0.1591, simple_loss=0.2183, pruned_loss=0.04989, over 4754.00 frames. ], tot_loss[loss=0.1889, simple_loss=0.2534, pruned_loss=0.06219, over 954226.73 frames. ], batch size: 27, lr: 3.78e-03, grad_scale: 32.0 2023-04-26 23:06:22,169 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=50189.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:06:28,407 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=50197.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 23:06:37,584 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.96 vs. limit=5.0 2023-04-26 23:06:42,231 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6121, 1.6845, 0.8253, 1.3030, 1.9859, 1.4899, 1.3955, 1.4486], device='cuda:6'), covar=tensor([0.0505, 0.0367, 0.0370, 0.0556, 0.0261, 0.0531, 0.0504, 0.0567], device='cuda:6'), in_proj_covar=tensor([0.0030, 0.0026, 0.0023, 0.0030, 0.0021, 0.0029, 0.0029, 0.0030], device='cuda:6'), out_proj_covar=tensor([0.0049, 0.0044, 0.0038, 0.0050, 0.0037, 0.0048, 0.0048, 0.0049], device='cuda:6') 2023-04-26 23:06:44,561 INFO [finetune.py:976] (6/7) Epoch 9, batch 4400, loss[loss=0.1937, simple_loss=0.2596, pruned_loss=0.06388, over 4863.00 frames. ], tot_loss[loss=0.1908, simple_loss=0.2551, pruned_loss=0.0633, over 954656.73 frames. ], batch size: 31, lr: 3.78e-03, grad_scale: 32.0 2023-04-26 23:06:49,793 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-04-26 23:06:52,522 INFO [optim.py:369] (6/7) 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 23:06:57,874 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.9874, 2.2518, 2.1377, 2.3316, 2.0663, 2.3274, 2.2965, 2.1874], device='cuda:6'), covar=tensor([0.4575, 0.8269, 0.6740, 0.6059, 0.7226, 0.9244, 0.8105, 0.7234], device='cuda:6'), in_proj_covar=tensor([0.0322, 0.0380, 0.0313, 0.0324, 0.0338, 0.0400, 0.0359, 0.0322], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-26 23:07:13,997 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=50258.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 23:07:27,675 INFO [finetune.py:976] (6/7) Epoch 9, batch 4450, loss[loss=0.214, simple_loss=0.2775, pruned_loss=0.07525, over 4760.00 frames. ], tot_loss[loss=0.1932, simple_loss=0.2578, pruned_loss=0.06428, over 955241.59 frames. ], batch size: 26, lr: 3.77e-03, grad_scale: 32.0 2023-04-26 23:08:33,134 INFO [finetune.py:976] (6/7) Epoch 9, batch 4500, loss[loss=0.2132, simple_loss=0.2788, pruned_loss=0.07374, over 4870.00 frames. ], tot_loss[loss=0.1957, simple_loss=0.261, pruned_loss=0.06525, over 957024.96 frames. ], batch size: 31, lr: 3.77e-03, grad_scale: 32.0 2023-04-26 23:08:46,794 INFO [optim.py:369] (6/7) 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] (6/7) Epoch 9, batch 4550, loss[loss=0.1969, simple_loss=0.2663, pruned_loss=0.06376, over 4836.00 frames. ], tot_loss[loss=0.197, simple_loss=0.2625, pruned_loss=0.06573, over 956705.26 frames. ], batch size: 30, lr: 3.77e-03, grad_scale: 32.0 2023-04-26 23:09:56,968 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.3475, 2.5802, 1.2574, 1.5792, 2.2262, 1.4513, 3.3365, 1.8731], device='cuda:6'), covar=tensor([0.0520, 0.0619, 0.0673, 0.1137, 0.0408, 0.0855, 0.0237, 0.0574], device='cuda:6'), in_proj_covar=tensor([0.0051, 0.0066, 0.0049, 0.0047, 0.0051, 0.0052, 0.0078, 0.0051], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0007, 0.0008, 0.0008, 0.0010, 0.0007], device='cuda:6') 2023-04-26 23:10:20,888 INFO [finetune.py:976] (6/7) Epoch 9, batch 4600, loss[loss=0.1811, simple_loss=0.2529, pruned_loss=0.05469, over 4822.00 frames. ], tot_loss[loss=0.1977, simple_loss=0.2633, pruned_loss=0.06601, over 955728.41 frames. ], batch size: 33, lr: 3.77e-03, grad_scale: 32.0 2023-04-26 23:10:20,987 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=50422.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:10:39,536 INFO [optim.py:369] (6/7) 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 23:10:39,975 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.62 vs. limit=2.0 2023-04-26 23:11:09,606 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0876, 1.5063, 1.8891, 2.1885, 1.8521, 1.4859, 1.0487, 1.5639], device='cuda:6'), covar=tensor([0.3441, 0.3799, 0.1896, 0.2446, 0.3219, 0.3060, 0.4666, 0.2604], device='cuda:6'), in_proj_covar=tensor([0.0282, 0.0251, 0.0220, 0.0318, 0.0214, 0.0229, 0.0233, 0.0186], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:6') 2023-04-26 23:11:14,981 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=50470.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:11:16,116 INFO [finetune.py:976] (6/7) Epoch 9, batch 4650, loss[loss=0.1346, simple_loss=0.211, pruned_loss=0.02917, over 4782.00 frames. ], tot_loss[loss=0.195, simple_loss=0.2602, pruned_loss=0.06485, over 955061.36 frames. ], batch size: 28, lr: 3.77e-03, grad_scale: 32.0 2023-04-26 23:11:23,418 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=50484.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:11:46,341 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.1100, 1.3698, 1.2315, 1.6404, 1.4490, 1.5028, 1.2912, 2.4183], device='cuda:6'), covar=tensor([0.0612, 0.0805, 0.0790, 0.1208, 0.0671, 0.0527, 0.0754, 0.0230], device='cuda:6'), in_proj_covar=tensor([0.0038, 0.0039, 0.0040, 0.0044, 0.0040, 0.0039, 0.0039, 0.0058], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0014, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:6') 2023-04-26 23:11:49,303 INFO [finetune.py:976] (6/7) Epoch 9, batch 4700, loss[loss=0.1932, simple_loss=0.2479, pruned_loss=0.0693, over 4737.00 frames. ], tot_loss[loss=0.1921, simple_loss=0.2568, pruned_loss=0.06368, over 955265.90 frames. ], batch size: 23, lr: 3.77e-03, grad_scale: 32.0 2023-04-26 23:11:57,183 INFO [optim.py:369] (6/7) 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 23:12:04,106 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=50546.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:12:08,322 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=50553.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 23:12:22,117 INFO [finetune.py:976] (6/7) Epoch 9, batch 4750, loss[loss=0.1914, simple_loss=0.2612, pruned_loss=0.06078, over 4900.00 frames. ], tot_loss[loss=0.1901, simple_loss=0.2547, pruned_loss=0.0628, over 954873.88 frames. ], batch size: 35, lr: 3.77e-03, grad_scale: 32.0 2023-04-26 23:12:54,835 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=50607.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:13:16,246 INFO [finetune.py:976] (6/7) Epoch 9, batch 4800, loss[loss=0.1736, simple_loss=0.2362, pruned_loss=0.05549, over 4745.00 frames. ], tot_loss[loss=0.193, simple_loss=0.2583, pruned_loss=0.06385, over 954463.89 frames. ], batch size: 23, lr: 3.77e-03, grad_scale: 32.0 2023-04-26 23:13:30,156 INFO [optim.py:369] (6/7) 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 23:13:44,339 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.63 vs. limit=2.0 2023-04-26 23:13:55,578 INFO [finetune.py:976] (6/7) Epoch 9, batch 4850, loss[loss=0.1645, simple_loss=0.2373, pruned_loss=0.04587, over 4752.00 frames. ], tot_loss[loss=0.1962, simple_loss=0.2623, pruned_loss=0.06506, over 955473.28 frames. ], batch size: 27, lr: 3.77e-03, grad_scale: 32.0 2023-04-26 23:14:16,523 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-04-26 23:14:28,053 INFO [finetune.py:976] (6/7) Epoch 9, batch 4900, loss[loss=0.21, simple_loss=0.2733, pruned_loss=0.0733, over 4757.00 frames. ], tot_loss[loss=0.198, simple_loss=0.264, pruned_loss=0.06602, over 954074.73 frames. ], batch size: 54, lr: 3.77e-03, grad_scale: 32.0 2023-04-26 23:14:36,901 INFO [optim.py:369] (6/7) 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 23:15:14,090 INFO [finetune.py:976] (6/7) Epoch 9, batch 4950, loss[loss=0.1641, simple_loss=0.2427, pruned_loss=0.04278, over 4896.00 frames. ], tot_loss[loss=0.1966, simple_loss=0.2634, pruned_loss=0.06495, over 953975.51 frames. ], batch size: 36, lr: 3.77e-03, grad_scale: 32.0 2023-04-26 23:15:33,301 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=50782.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:15:34,472 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=50784.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:15:42,668 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7238, 2.0596, 1.8212, 1.9065, 1.5413, 1.6939, 1.8361, 1.4415], device='cuda:6'), covar=tensor([0.1489, 0.0889, 0.0665, 0.0974, 0.2539, 0.1006, 0.1374, 0.1950], device='cuda:6'), in_proj_covar=tensor([0.0296, 0.0316, 0.0227, 0.0289, 0.0316, 0.0272, 0.0258, 0.0280], device='cuda:6'), out_proj_covar=tensor([1.2012e-04, 1.2743e-04, 9.1247e-05, 1.1574e-04, 1.2932e-04, 1.0959e-04, 1.0535e-04, 1.1216e-04], device='cuda:6') 2023-04-26 23:15:44,777 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.14 vs. limit=2.0 2023-04-26 23:16:02,476 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=50808.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 23:16:11,331 INFO [finetune.py:976] (6/7) Epoch 9, batch 5000, loss[loss=0.2174, simple_loss=0.2654, pruned_loss=0.08472, over 4896.00 frames. ], tot_loss[loss=0.1952, simple_loss=0.2613, pruned_loss=0.06457, over 954031.40 frames. ], batch size: 35, lr: 3.77e-03, grad_scale: 32.0 2023-04-26 23:16:19,879 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=50832.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:16:21,617 INFO [optim.py:369] (6/7) 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 23:16:26,690 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=50843.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:16:32,731 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=50853.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 23:16:42,404 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=50869.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 23:16:44,514 INFO [finetune.py:976] (6/7) Epoch 9, batch 5050, loss[loss=0.1854, simple_loss=0.2623, pruned_loss=0.05423, over 4869.00 frames. ], tot_loss[loss=0.1927, simple_loss=0.2583, pruned_loss=0.06353, over 955353.77 frames. ], batch size: 31, lr: 3.77e-03, grad_scale: 32.0 2023-04-26 23:17:04,644 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=50901.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 23:17:05,267 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=50902.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:17:17,313 INFO [finetune.py:976] (6/7) Epoch 9, batch 5100, loss[loss=0.2287, simple_loss=0.2655, pruned_loss=0.09591, over 4073.00 frames. ], tot_loss[loss=0.1894, simple_loss=0.2546, pruned_loss=0.0621, over 956389.18 frames. ], batch size: 65, lr: 3.77e-03, grad_scale: 32.0 2023-04-26 23:17:26,163 INFO [optim.py:369] (6/7) 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 23:17:34,488 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=50946.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:17:50,266 INFO [finetune.py:976] (6/7) Epoch 9, batch 5150, loss[loss=0.242, simple_loss=0.2928, pruned_loss=0.09558, over 4905.00 frames. ], tot_loss[loss=0.1902, simple_loss=0.255, pruned_loss=0.06269, over 955324.04 frames. ], batch size: 43, lr: 3.77e-03, grad_scale: 32.0 2023-04-26 23:17:52,212 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.2574, 1.7039, 1.4935, 1.8208, 1.7845, 1.9908, 1.4013, 3.7852], device='cuda:6'), covar=tensor([0.0651, 0.0780, 0.0829, 0.1252, 0.0635, 0.0536, 0.0787, 0.0165], device='cuda:6'), in_proj_covar=tensor([0.0038, 0.0039, 0.0040, 0.0044, 0.0040, 0.0039, 0.0039, 0.0058], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0014, 0.0013, 0.0012, 0.0013, 0.0016], device='cuda:6') 2023-04-26 23:18:16,630 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8152, 1.3289, 1.6094, 1.6109, 1.6078, 1.2884, 0.7668, 1.2867], device='cuda:6'), covar=tensor([0.3393, 0.3518, 0.1765, 0.2352, 0.2624, 0.2769, 0.4451, 0.2390], device='cuda:6'), in_proj_covar=tensor([0.0280, 0.0248, 0.0218, 0.0315, 0.0212, 0.0227, 0.0231, 0.0184], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:6') 2023-04-26 23:18:25,497 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=51007.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:18:40,595 INFO [finetune.py:976] (6/7) Epoch 9, batch 5200, loss[loss=0.2125, simple_loss=0.2978, pruned_loss=0.06365, over 4935.00 frames. ], tot_loss[loss=0.1932, simple_loss=0.2587, pruned_loss=0.06387, over 953226.23 frames. ], batch size: 42, lr: 3.77e-03, grad_scale: 32.0 2023-04-26 23:18:49,051 INFO [optim.py:369] (6/7) 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 23:18:51,270 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.4979, 1.4631, 1.7851, 1.7910, 1.3944, 1.1990, 1.4141, 0.9574], device='cuda:6'), covar=tensor([0.0691, 0.0704, 0.0453, 0.0680, 0.0859, 0.1346, 0.0814, 0.0944], device='cuda:6'), in_proj_covar=tensor([0.0066, 0.0073, 0.0071, 0.0067, 0.0076, 0.0096, 0.0078, 0.0074], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-26 23:19:08,580 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=51062.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:19:14,486 INFO [finetune.py:976] (6/7) Epoch 9, batch 5250, loss[loss=0.2288, simple_loss=0.2878, pruned_loss=0.08487, over 4786.00 frames. ], tot_loss[loss=0.1951, simple_loss=0.2609, pruned_loss=0.06468, over 953502.56 frames. ], batch size: 29, lr: 3.77e-03, grad_scale: 32.0 2023-04-26 23:19:34,182 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-04-26 23:19:47,750 INFO [finetune.py:976] (6/7) Epoch 9, batch 5300, loss[loss=0.2215, simple_loss=0.2741, pruned_loss=0.08449, over 4116.00 frames. ], tot_loss[loss=0.1971, simple_loss=0.2628, pruned_loss=0.06572, over 954460.85 frames. ], batch size: 65, lr: 3.77e-03, grad_scale: 32.0 2023-04-26 23:19:48,488 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=51123.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:19:56,070 INFO [optim.py:369] (6/7) 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 23:19:57,981 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=51138.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:20:15,865 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=51164.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 23:20:20,664 INFO [finetune.py:976] (6/7) Epoch 9, batch 5350, loss[loss=0.1826, simple_loss=0.2566, pruned_loss=0.05428, over 4891.00 frames. ], tot_loss[loss=0.198, simple_loss=0.2634, pruned_loss=0.06632, over 954245.52 frames. ], batch size: 32, lr: 3.77e-03, grad_scale: 32.0 2023-04-26 23:20:20,782 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=51172.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 23:20:47,201 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=51202.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:21:10,255 INFO [finetune.py:976] (6/7) Epoch 9, batch 5400, loss[loss=0.1454, simple_loss=0.2108, pruned_loss=0.04004, over 4726.00 frames. ], tot_loss[loss=0.1942, simple_loss=0.2594, pruned_loss=0.06449, over 955436.26 frames. ], batch size: 59, lr: 3.77e-03, grad_scale: 32.0 2023-04-26 23:21:21,594 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=51233.0, num_to_drop=1, layers_to_drop={3} 2023-04-26 23:21:22,681 INFO [optim.py:369] (6/7) 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 23:21:44,188 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=51250.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:21:51,897 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([0.4219, 1.4107, 1.4682, 0.9671, 1.4240, 1.1388, 1.7769, 1.3470], device='cuda:6'), covar=tensor([0.4034, 0.1892, 0.5046, 0.3047, 0.1710, 0.2472, 0.2074, 0.5077], device='cuda:6'), in_proj_covar=tensor([0.0344, 0.0347, 0.0429, 0.0361, 0.0386, 0.0382, 0.0379, 0.0419], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-26 23:21:56,067 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.53 vs. limit=2.0 2023-04-26 23:22:14,588 INFO [finetune.py:976] (6/7) Epoch 9, batch 5450, loss[loss=0.1832, simple_loss=0.2512, pruned_loss=0.05763, over 4909.00 frames. ], tot_loss[loss=0.191, simple_loss=0.256, pruned_loss=0.063, over 956980.24 frames. ], batch size: 35, lr: 3.77e-03, grad_scale: 32.0 2023-04-26 23:22:47,581 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=51302.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:23:17,625 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-04-26 23:23:19,200 INFO [finetune.py:976] (6/7) Epoch 9, batch 5500, loss[loss=0.2113, simple_loss=0.2864, pruned_loss=0.06814, over 4905.00 frames. ], tot_loss[loss=0.1895, simple_loss=0.2541, pruned_loss=0.06241, over 957738.77 frames. ], batch size: 43, lr: 3.77e-03, grad_scale: 32.0 2023-04-26 23:23:32,912 INFO [optim.py:369] (6/7) 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 23:24:14,279 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5530, 1.2386, 1.7111, 2.0645, 1.7286, 1.5622, 1.5926, 1.6017], device='cuda:6'), covar=tensor([0.6203, 0.8576, 0.8546, 0.7924, 0.7034, 0.9965, 1.0298, 0.9923], device='cuda:6'), in_proj_covar=tensor([0.0409, 0.0418, 0.0504, 0.0521, 0.0437, 0.0456, 0.0468, 0.0465], device='cuda:6'), out_proj_covar=tensor([9.9642e-05, 1.0354e-04, 1.1359e-04, 1.2383e-04, 1.0618e-04, 1.1048e-04, 1.1257e-04, 1.1243e-04], device='cuda:6') 2023-04-26 23:24:14,746 INFO [finetune.py:976] (6/7) Epoch 9, batch 5550, loss[loss=0.1896, simple_loss=0.2675, pruned_loss=0.05584, over 4923.00 frames. ], tot_loss[loss=0.1932, simple_loss=0.2581, pruned_loss=0.06419, over 958123.36 frames. ], batch size: 42, lr: 3.77e-03, grad_scale: 32.0 2023-04-26 23:24:20,973 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=51382.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:24:26,995 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=51392.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:24:28,511 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-04-26 23:24:32,894 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.55 vs. limit=2.0 2023-04-26 23:24:42,935 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 9, batch 5600, loss[loss=0.2188, simple_loss=0.2745, pruned_loss=0.08152, over 4826.00 frames. ], tot_loss[loss=0.1955, simple_loss=0.2606, pruned_loss=0.06524, over 955115.93 frames. ], batch size: 30, lr: 3.77e-03, grad_scale: 32.0 2023-04-26 23:24:52,682 INFO [optim.py:369] (6/7) 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 23:24:54,551 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=51438.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:24:57,455 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=51443.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:25:03,232 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=51453.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:25:10,023 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=51464.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 23:25:15,158 INFO [finetune.py:976] (6/7) Epoch 9, batch 5650, loss[loss=0.1882, simple_loss=0.261, pruned_loss=0.05771, over 4813.00 frames. ], tot_loss[loss=0.1984, simple_loss=0.2642, pruned_loss=0.06636, over 954502.54 frames. ], batch size: 33, lr: 3.77e-03, grad_scale: 32.0 2023-04-26 23:25:23,729 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=51486.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:25:29,052 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=51495.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:25:39,098 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=51512.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 23:25:45,277 INFO [finetune.py:976] (6/7) Epoch 9, batch 5700, loss[loss=0.1808, simple_loss=0.2456, pruned_loss=0.05797, over 4153.00 frames. ], tot_loss[loss=0.1962, simple_loss=0.26, pruned_loss=0.06624, over 936415.70 frames. ], batch size: 18, lr: 3.77e-03, grad_scale: 32.0 2023-04-26 23:25:48,919 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=51528.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 23:25:53,110 INFO [optim.py:369] (6/7) 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 23:26:16,107 INFO [finetune.py:976] (6/7) Epoch 10, batch 0, loss[loss=0.1947, simple_loss=0.2639, pruned_loss=0.06277, over 4791.00 frames. ], tot_loss[loss=0.1947, simple_loss=0.2639, pruned_loss=0.06277, over 4791.00 frames. ], batch size: 25, lr: 3.76e-03, grad_scale: 32.0 2023-04-26 23:26:16,107 INFO [finetune.py:1001] (6/7) Computing validation loss 2023-04-26 23:26:31,822 INFO [finetune.py:1010] (6/7) Epoch 10, validation: loss=0.1558, simple_loss=0.2282, pruned_loss=0.04164, over 2265189.00 frames. 2023-04-26 23:26:31,823 INFO [finetune.py:1011] (6/7) Maximum memory allocated so far is 6345MB 2023-04-26 23:26:37,656 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=51556.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:27:02,355 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.46 vs. limit=5.0 2023-04-26 23:27:05,704 INFO [finetune.py:976] (6/7) Epoch 10, batch 50, loss[loss=0.1824, simple_loss=0.2504, pruned_loss=0.05724, over 4781.00 frames. ], tot_loss[loss=0.1953, simple_loss=0.2605, pruned_loss=0.06498, over 216196.45 frames. ], batch size: 29, lr: 3.76e-03, grad_scale: 32.0 2023-04-26 23:27:09,028 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=51602.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:27:31,933 INFO [optim.py:369] (6/7) 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 23:27:45,647 INFO [finetune.py:976] (6/7) Epoch 10, batch 100, loss[loss=0.1437, simple_loss=0.2163, pruned_loss=0.03555, over 4762.00 frames. ], tot_loss[loss=0.1901, simple_loss=0.2547, pruned_loss=0.06273, over 381763.77 frames. ], batch size: 28, lr: 3.76e-03, grad_scale: 16.0 2023-04-26 23:27:47,276 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=51650.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:28:38,560 INFO [finetune.py:976] (6/7) Epoch 10, batch 150, loss[loss=0.2096, simple_loss=0.2767, pruned_loss=0.07124, over 4818.00 frames. ], tot_loss[loss=0.1892, simple_loss=0.2526, pruned_loss=0.06294, over 510119.85 frames. ], batch size: 38, lr: 3.76e-03, grad_scale: 16.0 2023-04-26 23:29:10,037 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=51718.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:29:20,965 INFO [optim.py:369] (6/7) 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 23:29:22,218 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=51738.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:29:29,232 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=51747.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:29:29,800 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 10, batch 200, loss[loss=0.1905, simple_loss=0.2438, pruned_loss=0.0686, over 4813.00 frames. ], tot_loss[loss=0.1916, simple_loss=0.2541, pruned_loss=0.06451, over 607811.23 frames. ], batch size: 25, lr: 3.76e-03, grad_scale: 16.0 2023-04-26 23:29:42,681 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=51766.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:30:04,063 INFO [finetune.py:976] (6/7) Epoch 10, batch 250, loss[loss=0.1948, simple_loss=0.2697, pruned_loss=0.05993, over 4724.00 frames. ], tot_loss[loss=0.1921, simple_loss=0.256, pruned_loss=0.06413, over 685817.83 frames. ], batch size: 59, lr: 3.76e-03, grad_scale: 16.0 2023-04-26 23:30:11,186 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=51808.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:30:23,868 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=51828.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 23:30:28,670 INFO [optim.py:369] (6/7) 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 23:30:37,591 INFO [finetune.py:976] (6/7) Epoch 10, batch 300, loss[loss=0.1768, simple_loss=0.252, pruned_loss=0.05077, over 4813.00 frames. ], tot_loss[loss=0.1946, simple_loss=0.2599, pruned_loss=0.06461, over 746020.77 frames. ], batch size: 38, lr: 3.76e-03, grad_scale: 16.0 2023-04-26 23:30:39,366 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=51851.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:30:44,650 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6461, 1.3839, 1.8430, 2.0706, 1.8164, 1.6436, 1.7248, 1.7127], device='cuda:6'), covar=tensor([0.5613, 0.7689, 0.7447, 0.7845, 0.6733, 0.9547, 0.9565, 0.9162], device='cuda:6'), in_proj_covar=tensor([0.0409, 0.0416, 0.0502, 0.0520, 0.0437, 0.0456, 0.0467, 0.0465], device='cuda:6'), 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:6') 2023-04-26 23:30:56,482 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=51876.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 23:31:10,833 INFO [finetune.py:976] (6/7) Epoch 10, batch 350, loss[loss=0.237, simple_loss=0.3009, pruned_loss=0.08655, over 4885.00 frames. ], tot_loss[loss=0.1944, simple_loss=0.2604, pruned_loss=0.06424, over 793599.53 frames. ], batch size: 43, lr: 3.76e-03, grad_scale: 16.0 2023-04-26 23:31:26,194 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=3.49 vs. limit=5.0 2023-04-26 23:31:41,425 INFO [optim.py:369] (6/7) 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 23:32:00,718 INFO [finetune.py:976] (6/7) Epoch 10, batch 400, loss[loss=0.1854, simple_loss=0.2568, pruned_loss=0.05695, over 4820.00 frames. ], tot_loss[loss=0.1938, simple_loss=0.2604, pruned_loss=0.06359, over 828386.21 frames. ], batch size: 47, lr: 3.76e-03, grad_scale: 16.0 2023-04-26 23:32:15,697 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-04-26 23:32:17,894 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5712, 3.7660, 0.7625, 2.1613, 2.0834, 2.4629, 2.3002, 0.9733], device='cuda:6'), covar=tensor([0.1360, 0.0993, 0.2093, 0.1224, 0.1093, 0.1119, 0.1318, 0.2265], device='cuda:6'), in_proj_covar=tensor([0.0118, 0.0250, 0.0141, 0.0122, 0.0136, 0.0155, 0.0119, 0.0122], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-26 23:32:40,858 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=3.54 vs. limit=5.0 2023-04-26 23:32:51,605 INFO [finetune.py:976] (6/7) Epoch 10, batch 450, loss[loss=0.2142, simple_loss=0.2762, pruned_loss=0.07612, over 4870.00 frames. ], tot_loss[loss=0.1939, simple_loss=0.2605, pruned_loss=0.06365, over 857406.70 frames. ], batch size: 34, lr: 3.76e-03, grad_scale: 16.0 2023-04-26 23:33:31,101 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.82 vs. limit=5.0 2023-04-26 23:33:45,356 INFO [optim.py:369] (6/7) 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 23:33:46,684 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=52038.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:33:53,308 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=52040.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:34:03,621 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=52048.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:34:04,127 INFO [finetune.py:976] (6/7) Epoch 10, batch 500, loss[loss=0.2028, simple_loss=0.2628, pruned_loss=0.07134, over 4810.00 frames. ], tot_loss[loss=0.1915, simple_loss=0.2576, pruned_loss=0.06272, over 878841.65 frames. ], batch size: 39, lr: 3.76e-03, grad_scale: 16.0 2023-04-26 23:34:51,738 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=52086.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:35:09,316 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 10, batch 550, loss[loss=0.1946, simple_loss=0.2544, pruned_loss=0.06742, over 4911.00 frames. ], tot_loss[loss=0.1892, simple_loss=0.2548, pruned_loss=0.06177, over 898651.60 frames. ], batch size: 32, lr: 3.76e-03, grad_scale: 16.0 2023-04-26 23:35:12,504 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=52101.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:35:13,635 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=52103.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:35:22,470 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.8717, 2.1117, 2.0262, 2.2226, 1.9347, 2.0728, 2.0968, 1.9895], device='cuda:6'), covar=tensor([0.4863, 0.8214, 0.6274, 0.5164, 0.7103, 0.9480, 0.7852, 0.6999], device='cuda:6'), in_proj_covar=tensor([0.0327, 0.0385, 0.0318, 0.0328, 0.0342, 0.0405, 0.0362, 0.0326], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-26 23:36:05,085 INFO [optim.py:369] (6/7) 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 23:36:18,684 INFO [finetune.py:976] (6/7) Epoch 10, batch 600, loss[loss=0.2282, simple_loss=0.2915, pruned_loss=0.08245, over 4198.00 frames. ], tot_loss[loss=0.1889, simple_loss=0.2543, pruned_loss=0.06173, over 909125.17 frames. ], batch size: 65, lr: 3.76e-03, grad_scale: 16.0 2023-04-26 23:36:19,998 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=52151.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:37:10,897 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.3454, 1.7196, 2.2022, 2.8765, 2.1656, 1.6667, 1.7093, 2.0680], device='cuda:6'), covar=tensor([0.3896, 0.4148, 0.1890, 0.2908, 0.3544, 0.3170, 0.4644, 0.2730], device='cuda:6'), in_proj_covar=tensor([0.0284, 0.0251, 0.0220, 0.0319, 0.0214, 0.0227, 0.0234, 0.0186], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:6') 2023-04-26 23:37:12,638 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7138, 1.7579, 0.9541, 1.3619, 2.1199, 1.5986, 1.5621, 1.4518], device='cuda:6'), covar=tensor([0.0527, 0.0367, 0.0333, 0.0556, 0.0266, 0.0531, 0.0501, 0.0565], device='cuda:6'), in_proj_covar=tensor([0.0030, 0.0025, 0.0023, 0.0030, 0.0021, 0.0029, 0.0029, 0.0030], device='cuda:6'), out_proj_covar=tensor([0.0049, 0.0044, 0.0038, 0.0050, 0.0038, 0.0048, 0.0048, 0.0050], device='cuda:6') 2023-04-26 23:37:21,595 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 10, batch 650, loss[loss=0.1483, simple_loss=0.2309, pruned_loss=0.03281, over 4821.00 frames. ], tot_loss[loss=0.191, simple_loss=0.2568, pruned_loss=0.06264, over 918235.05 frames. ], batch size: 33, lr: 3.76e-03, grad_scale: 16.0 2023-04-26 23:37:25,286 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=52199.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:38:17,626 INFO [optim.py:369] (6/7) 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 23:38:18,947 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.0693, 1.4926, 1.3205, 1.6043, 1.5370, 1.8392, 1.3191, 3.4608], device='cuda:6'), covar=tensor([0.0677, 0.0813, 0.0794, 0.1250, 0.0687, 0.0570, 0.0805, 0.0133], device='cuda:6'), in_proj_covar=tensor([0.0038, 0.0039, 0.0040, 0.0044, 0.0040, 0.0039, 0.0039, 0.0059], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0014, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:6') 2023-04-26 23:38:30,843 INFO [finetune.py:976] (6/7) Epoch 10, batch 700, loss[loss=0.2014, simple_loss=0.2812, pruned_loss=0.0608, over 4912.00 frames. ], tot_loss[loss=0.1933, simple_loss=0.2597, pruned_loss=0.06343, over 926140.60 frames. ], batch size: 42, lr: 3.76e-03, grad_scale: 16.0 2023-04-26 23:38:38,227 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.14 vs. limit=2.0 2023-04-26 23:38:39,796 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=52254.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:39:43,996 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7022, 1.2718, 1.3673, 1.3667, 1.8914, 1.5333, 1.2110, 1.3107], device='cuda:6'), covar=tensor([0.1483, 0.1404, 0.2115, 0.1342, 0.0861, 0.1561, 0.2362, 0.1933], device='cuda:6'), in_proj_covar=tensor([0.0305, 0.0318, 0.0351, 0.0294, 0.0332, 0.0317, 0.0302, 0.0353], device='cuda:6'), out_proj_covar=tensor([6.4408e-05, 6.7312e-05, 7.5737e-05, 6.0495e-05, 6.9582e-05, 6.8011e-05, 6.4832e-05, 7.5754e-05], device='cuda:6') 2023-04-26 23:39:44,473 INFO [finetune.py:976] (6/7) Epoch 10, batch 750, loss[loss=0.2051, simple_loss=0.2721, pruned_loss=0.0691, over 4880.00 frames. ], tot_loss[loss=0.1952, simple_loss=0.262, pruned_loss=0.06423, over 932104.35 frames. ], batch size: 32, lr: 3.76e-03, grad_scale: 16.0 2023-04-26 23:39:56,806 INFO [zipformer.py:1188] (6/7) 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] (6/7) 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 23:40:22,485 INFO [finetune.py:976] (6/7) Epoch 10, batch 800, loss[loss=0.1422, simple_loss=0.2111, pruned_loss=0.03661, over 4830.00 frames. ], tot_loss[loss=0.1942, simple_loss=0.2614, pruned_loss=0.06355, over 936483.01 frames. ], batch size: 33, lr: 3.76e-03, grad_scale: 16.0 2023-04-26 23:40:25,684 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.1309, 1.5377, 1.3316, 1.6954, 1.5747, 1.8534, 1.3774, 3.4262], device='cuda:6'), covar=tensor([0.0653, 0.0780, 0.0812, 0.1185, 0.0644, 0.0660, 0.0761, 0.0168], device='cuda:6'), in_proj_covar=tensor([0.0038, 0.0039, 0.0040, 0.0044, 0.0040, 0.0039, 0.0039, 0.0059], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0014, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:6') 2023-04-26 23:40:36,560 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.56 vs. limit=5.0 2023-04-26 23:40:37,968 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=52373.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:40:54,341 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=52396.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:40:56,070 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 2023-04-26 23:40:56,137 INFO [finetune.py:976] (6/7) Epoch 10, batch 850, loss[loss=0.1696, simple_loss=0.2294, pruned_loss=0.05493, over 4665.00 frames. ], tot_loss[loss=0.1928, simple_loss=0.2595, pruned_loss=0.063, over 939575.62 frames. ], batch size: 23, lr: 3.76e-03, grad_scale: 16.0 2023-04-26 23:40:58,772 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=52403.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:41:03,875 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.66 vs. limit=2.0 2023-04-26 23:41:30,886 INFO [optim.py:369] (6/7) 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 23:41:44,735 INFO [finetune.py:976] (6/7) Epoch 10, batch 900, loss[loss=0.1901, simple_loss=0.2433, pruned_loss=0.06839, over 4131.00 frames. ], tot_loss[loss=0.1906, simple_loss=0.2565, pruned_loss=0.06231, over 940745.01 frames. ], batch size: 65, lr: 3.76e-03, grad_scale: 16.0 2023-04-26 23:41:51,264 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=52451.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:42:03,865 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.63 vs. limit=2.0 2023-04-26 23:42:16,411 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7337, 1.2009, 1.4728, 1.4409, 1.8846, 1.6109, 1.2685, 1.4448], device='cuda:6'), covar=tensor([0.1423, 0.1648, 0.1862, 0.1384, 0.0905, 0.1448, 0.2176, 0.1991], device='cuda:6'), in_proj_covar=tensor([0.0305, 0.0318, 0.0352, 0.0295, 0.0332, 0.0318, 0.0303, 0.0354], device='cuda:6'), out_proj_covar=tensor([6.4358e-05, 6.7424e-05, 7.6011e-05, 6.0737e-05, 6.9589e-05, 6.8042e-05, 6.5057e-05, 7.5950e-05], device='cuda:6') 2023-04-26 23:42:49,243 INFO [finetune.py:976] (6/7) Epoch 10, batch 950, loss[loss=0.203, simple_loss=0.2706, pruned_loss=0.06773, over 4899.00 frames. ], tot_loss[loss=0.19, simple_loss=0.2557, pruned_loss=0.06216, over 944760.75 frames. ], batch size: 35, lr: 3.76e-03, grad_scale: 16.0 2023-04-26 23:43:41,298 INFO [optim.py:369] (6/7) 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 23:43:55,595 INFO [finetune.py:976] (6/7) Epoch 10, batch 1000, loss[loss=0.2017, simple_loss=0.272, pruned_loss=0.06571, over 4871.00 frames. ], tot_loss[loss=0.1913, simple_loss=0.2577, pruned_loss=0.06245, over 947382.03 frames. ], batch size: 34, lr: 3.76e-03, grad_scale: 16.0 2023-04-26 23:44:01,406 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=52549.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:44:05,058 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=52555.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 23:45:07,088 INFO [finetune.py:976] (6/7) Epoch 10, batch 1050, loss[loss=0.1766, simple_loss=0.2488, pruned_loss=0.05223, over 4863.00 frames. ], tot_loss[loss=0.193, simple_loss=0.2599, pruned_loss=0.06306, over 949195.83 frames. ], batch size: 31, lr: 3.76e-03, grad_scale: 16.0 2023-04-26 23:45:11,802 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=3.90 vs. limit=5.0 2023-04-26 23:45:29,121 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=52616.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 23:45:52,967 INFO [optim.py:369] (6/7) 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 23:46:13,538 INFO [finetune.py:976] (6/7) Epoch 10, batch 1100, loss[loss=0.2008, simple_loss=0.2733, pruned_loss=0.06414, over 4771.00 frames. ], tot_loss[loss=0.1928, simple_loss=0.2602, pruned_loss=0.06268, over 950716.68 frames. ], batch size: 28, lr: 3.76e-03, grad_scale: 16.0 2023-04-26 23:46:36,537 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=52668.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:46:56,132 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=52696.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:46:57,844 INFO [finetune.py:976] (6/7) Epoch 10, batch 1150, loss[loss=0.2213, simple_loss=0.2952, pruned_loss=0.07366, over 4808.00 frames. ], tot_loss[loss=0.1926, simple_loss=0.2597, pruned_loss=0.06268, over 950714.53 frames. ], batch size: 40, lr: 3.75e-03, grad_scale: 16.0 2023-04-26 23:47:02,619 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9473, 2.4394, 1.9523, 2.3105, 1.7473, 1.9476, 2.0031, 1.6401], device='cuda:6'), covar=tensor([0.2029, 0.1497, 0.0910, 0.1350, 0.3467, 0.1395, 0.2251, 0.3075], device='cuda:6'), in_proj_covar=tensor([0.0292, 0.0314, 0.0225, 0.0286, 0.0314, 0.0270, 0.0255, 0.0277], device='cuda:6'), out_proj_covar=tensor([1.1825e-04, 1.2638e-04, 9.0611e-05, 1.1476e-04, 1.2870e-04, 1.0883e-04, 1.0421e-04, 1.1129e-04], device='cuda:6') 2023-04-26 23:47:03,254 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5945, 1.7674, 1.4974, 1.1141, 1.2052, 1.1692, 1.4646, 1.1243], device='cuda:6'), covar=tensor([0.1902, 0.1352, 0.1711, 0.1926, 0.2490, 0.2186, 0.1205, 0.2283], device='cuda:6'), in_proj_covar=tensor([0.0195, 0.0216, 0.0170, 0.0204, 0.0205, 0.0184, 0.0160, 0.0188], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:6') 2023-04-26 23:47:21,725 INFO [optim.py:369] (6/7) 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 23:47:33,674 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 10, batch 1200, loss[loss=0.1541, simple_loss=0.2181, pruned_loss=0.04502, over 4770.00 frames. ], tot_loss[loss=0.1914, simple_loss=0.2583, pruned_loss=0.06227, over 951976.43 frames. ], batch size: 29, lr: 3.75e-03, grad_scale: 16.0 2023-04-26 23:47:45,158 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=52752.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:48:25,366 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-04-26 23:48:48,804 INFO [finetune.py:976] (6/7) Epoch 10, batch 1250, loss[loss=0.1763, simple_loss=0.2408, pruned_loss=0.05589, over 4900.00 frames. ], tot_loss[loss=0.1886, simple_loss=0.2555, pruned_loss=0.06088, over 953502.60 frames. ], batch size: 32, lr: 3.75e-03, grad_scale: 16.0 2023-04-26 23:49:09,954 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=52813.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:49:22,321 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0793, 1.8020, 2.1481, 2.4888, 2.4437, 2.0067, 1.6366, 2.1123], device='cuda:6'), covar=tensor([0.0899, 0.1132, 0.0651, 0.0637, 0.0722, 0.0937, 0.0857, 0.0680], device='cuda:6'), in_proj_covar=tensor([0.0197, 0.0205, 0.0183, 0.0177, 0.0178, 0.0189, 0.0160, 0.0185], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-26 23:49:34,634 INFO [optim.py:369] (6/7) 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 23:49:48,759 INFO [finetune.py:976] (6/7) Epoch 10, batch 1300, loss[loss=0.1957, simple_loss=0.2548, pruned_loss=0.0683, over 4781.00 frames. ], tot_loss[loss=0.186, simple_loss=0.252, pruned_loss=0.06, over 952941.04 frames. ], batch size: 28, lr: 3.75e-03, grad_scale: 16.0 2023-04-26 23:49:48,870 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=52849.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:50:10,708 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.71 vs. limit=2.0 2023-04-26 23:50:18,747 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8120, 4.3076, 0.7925, 2.1593, 2.4493, 2.8321, 2.5186, 1.0354], device='cuda:6'), covar=tensor([0.1320, 0.1013, 0.2165, 0.1340, 0.0953, 0.1070, 0.1413, 0.2024], device='cuda:6'), in_proj_covar=tensor([0.0118, 0.0249, 0.0141, 0.0122, 0.0135, 0.0154, 0.0119, 0.0122], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-26 23:50:20,541 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=52897.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:50:21,746 INFO [finetune.py:976] (6/7) Epoch 10, batch 1350, loss[loss=0.2603, simple_loss=0.3239, pruned_loss=0.09834, over 4816.00 frames. ], tot_loss[loss=0.188, simple_loss=0.2539, pruned_loss=0.061, over 955098.67 frames. ], batch size: 39, lr: 3.75e-03, grad_scale: 16.0 2023-04-26 23:50:31,122 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=52911.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 23:50:46,658 INFO [optim.py:369] (6/7) 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 23:50:50,344 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-04-26 23:50:55,068 INFO [finetune.py:976] (6/7) Epoch 10, batch 1400, loss[loss=0.1976, simple_loss=0.2614, pruned_loss=0.06694, over 4896.00 frames. ], tot_loss[loss=0.1908, simple_loss=0.2572, pruned_loss=0.06224, over 952179.32 frames. ], batch size: 35, lr: 3.75e-03, grad_scale: 16.0 2023-04-26 23:51:09,500 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=52968.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:51:13,199 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.3253, 1.7715, 1.6160, 2.0682, 1.9596, 2.1682, 1.6852, 4.5601], device='cuda:6'), covar=tensor([0.0658, 0.0897, 0.0881, 0.1260, 0.0696, 0.0545, 0.0825, 0.0129], device='cuda:6'), in_proj_covar=tensor([0.0039, 0.0039, 0.0041, 0.0044, 0.0040, 0.0039, 0.0039, 0.0059], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:6') 2023-04-26 23:51:28,733 INFO [finetune.py:976] (6/7) Epoch 10, batch 1450, loss[loss=0.2014, simple_loss=0.2747, pruned_loss=0.06407, over 4891.00 frames. ], tot_loss[loss=0.1937, simple_loss=0.2604, pruned_loss=0.06346, over 953325.59 frames. ], batch size: 32, lr: 3.75e-03, grad_scale: 16.0 2023-04-26 23:51:41,188 INFO [zipformer.py:1188] (6/7) 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] (6/7) 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 23:52:07,465 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([3.7175, 3.6706, 2.6757, 4.2626, 3.8236, 3.6559, 1.6715, 3.6569], device='cuda:6'), covar=tensor([0.1987, 0.1346, 0.3392, 0.1680, 0.2061, 0.1977, 0.5713, 0.2397], device='cuda:6'), in_proj_covar=tensor([0.0241, 0.0215, 0.0246, 0.0300, 0.0297, 0.0247, 0.0266, 0.0268], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-26 23:52:07,526 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9684, 1.5272, 1.9369, 2.3025, 1.8908, 1.4485, 1.2304, 1.8024], device='cuda:6'), covar=tensor([0.4223, 0.4013, 0.1974, 0.3078, 0.3369, 0.3256, 0.5041, 0.2492], device='cuda:6'), in_proj_covar=tensor([0.0283, 0.0251, 0.0220, 0.0318, 0.0214, 0.0228, 0.0233, 0.0186], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:6') 2023-04-26 23:52:18,925 INFO [finetune.py:976] (6/7) Epoch 10, batch 1500, loss[loss=0.1685, simple_loss=0.2374, pruned_loss=0.04979, over 4821.00 frames. ], tot_loss[loss=0.1945, simple_loss=0.2614, pruned_loss=0.06381, over 953015.56 frames. ], batch size: 25, lr: 3.75e-03, grad_scale: 16.0 2023-04-26 23:52:28,792 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=53056.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:52:58,510 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.3715, 1.7107, 1.4264, 1.9330, 1.7447, 1.9593, 1.5743, 4.0739], device='cuda:6'), covar=tensor([0.0666, 0.0880, 0.0917, 0.1235, 0.0715, 0.0750, 0.0881, 0.0159], device='cuda:6'), in_proj_covar=tensor([0.0038, 0.0039, 0.0040, 0.0044, 0.0040, 0.0039, 0.0039, 0.0059], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0014, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:6') 2023-04-26 23:53:19,949 INFO [finetune.py:976] (6/7) Epoch 10, batch 1550, loss[loss=0.1804, simple_loss=0.248, pruned_loss=0.05638, over 4924.00 frames. ], tot_loss[loss=0.1929, simple_loss=0.26, pruned_loss=0.06288, over 952643.38 frames. ], batch size: 38, lr: 3.75e-03, grad_scale: 16.0 2023-04-26 23:53:25,471 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=53108.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:53:30,160 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6173, 1.3902, 1.8774, 1.8438, 1.4670, 1.2401, 1.5643, 1.0280], device='cuda:6'), covar=tensor([0.0587, 0.0788, 0.0430, 0.0751, 0.0861, 0.1225, 0.0617, 0.0775], device='cuda:6'), in_proj_covar=tensor([0.0066, 0.0073, 0.0070, 0.0067, 0.0075, 0.0095, 0.0077, 0.0073], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-26 23:53:32,429 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=53117.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:53:45,210 INFO [optim.py:369] (6/7) 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 23:53:53,129 INFO [finetune.py:976] (6/7) Epoch 10, batch 1600, loss[loss=0.1935, simple_loss=0.267, pruned_loss=0.06001, over 4920.00 frames. ], tot_loss[loss=0.1907, simple_loss=0.2575, pruned_loss=0.062, over 953502.32 frames. ], batch size: 43, lr: 3.75e-03, grad_scale: 16.0 2023-04-26 23:53:54,957 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5628, 1.7614, 1.3987, 1.0958, 1.1771, 1.1585, 1.3758, 1.0900], device='cuda:6'), covar=tensor([0.1782, 0.1422, 0.1676, 0.2061, 0.2557, 0.2083, 0.1205, 0.2282], device='cuda:6'), in_proj_covar=tensor([0.0196, 0.0217, 0.0171, 0.0205, 0.0206, 0.0186, 0.0161, 0.0190], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:6') 2023-04-26 23:54:44,152 INFO [finetune.py:976] (6/7) Epoch 10, batch 1650, loss[loss=0.1921, simple_loss=0.25, pruned_loss=0.06707, over 4380.00 frames. ], tot_loss[loss=0.1892, simple_loss=0.2552, pruned_loss=0.06157, over 954015.85 frames. ], batch size: 65, lr: 3.75e-03, grad_scale: 16.0 2023-04-26 23:54:52,153 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=53211.0, num_to_drop=1, layers_to_drop={2} 2023-04-26 23:55:08,519 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9238, 1.2992, 4.5967, 4.2972, 4.0327, 4.2655, 4.1284, 4.0885], device='cuda:6'), covar=tensor([0.6848, 0.6231, 0.1079, 0.1973, 0.1249, 0.1622, 0.2176, 0.1625], device='cuda:6'), in_proj_covar=tensor([0.0308, 0.0307, 0.0406, 0.0409, 0.0349, 0.0405, 0.0315, 0.0372], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-26 23:55:09,622 INFO [optim.py:369] (6/7) 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 23:55:12,841 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=53241.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:55:17,569 INFO [finetune.py:976] (6/7) Epoch 10, batch 1700, loss[loss=0.1838, simple_loss=0.2428, pruned_loss=0.06241, over 4817.00 frames. ], tot_loss[loss=0.1878, simple_loss=0.2535, pruned_loss=0.06108, over 955844.94 frames. ], batch size: 25, lr: 3.75e-03, grad_scale: 16.0 2023-04-26 23:55:24,105 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=53259.0, num_to_drop=1, layers_to_drop={0} 2023-04-26 23:55:51,480 INFO [finetune.py:976] (6/7) Epoch 10, batch 1750, loss[loss=0.2298, simple_loss=0.2974, pruned_loss=0.08114, over 4800.00 frames. ], tot_loss[loss=0.1899, simple_loss=0.256, pruned_loss=0.06184, over 956796.73 frames. ], batch size: 51, lr: 3.75e-03, grad_scale: 16.0 2023-04-26 23:55:53,417 INFO [zipformer.py:1188] (6/7) 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] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=53304.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:55:57,481 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.5308, 2.9561, 1.9739, 2.4359, 2.9981, 2.4165, 2.4405, 2.4914], device='cuda:6'), covar=tensor([0.0390, 0.0256, 0.0247, 0.0423, 0.0180, 0.0416, 0.0401, 0.0435], device='cuda:6'), in_proj_covar=tensor([0.0030, 0.0025, 0.0023, 0.0030, 0.0021, 0.0029, 0.0029, 0.0030], device='cuda:6'), out_proj_covar=tensor([0.0049, 0.0044, 0.0038, 0.0050, 0.0038, 0.0048, 0.0048, 0.0050], device='cuda:6') 2023-04-26 23:56:01,780 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=53315.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:56:02,469 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=3.61 vs. limit=5.0 2023-04-26 23:56:16,878 INFO [optim.py:369] (6/7) 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 23:56:25,438 INFO [finetune.py:976] (6/7) Epoch 10, batch 1800, loss[loss=0.2651, simple_loss=0.3148, pruned_loss=0.1076, over 4901.00 frames. ], tot_loss[loss=0.1915, simple_loss=0.2582, pruned_loss=0.06236, over 956528.55 frames. ], batch size: 43, lr: 3.75e-03, grad_scale: 16.0 2023-04-26 23:56:35,682 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=53365.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:56:36,893 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=53367.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:56:43,373 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 10, batch 1850, loss[loss=0.2018, simple_loss=0.2641, pruned_loss=0.06974, over 4862.00 frames. ], tot_loss[loss=0.1934, simple_loss=0.2602, pruned_loss=0.06332, over 956292.17 frames. ], batch size: 31, lr: 3.75e-03, grad_scale: 16.0 2023-04-26 23:57:10,977 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=53408.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:57:13,339 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7539, 1.1535, 1.3961, 1.4132, 1.8980, 1.5636, 1.2341, 1.3206], device='cuda:6'), covar=tensor([0.1548, 0.1737, 0.2111, 0.1519, 0.0913, 0.1367, 0.2186, 0.2245], device='cuda:6'), in_proj_covar=tensor([0.0305, 0.0320, 0.0354, 0.0296, 0.0334, 0.0320, 0.0305, 0.0357], device='cuda:6'), out_proj_covar=tensor([6.4267e-05, 6.7731e-05, 7.6398e-05, 6.1005e-05, 6.9799e-05, 6.8676e-05, 6.5490e-05, 7.6503e-05], device='cuda:6') 2023-04-26 23:57:13,889 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=53412.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:57:35,488 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=53428.0, num_to_drop=1, layers_to_drop={1} 2023-04-26 23:57:45,404 INFO [optim.py:369] (6/7) 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 23:58:05,921 INFO [finetune.py:976] (6/7) Epoch 10, batch 1900, loss[loss=0.1987, simple_loss=0.2847, pruned_loss=0.05633, over 4817.00 frames. ], tot_loss[loss=0.1935, simple_loss=0.261, pruned_loss=0.063, over 956675.99 frames. ], batch size: 39, lr: 3.75e-03, grad_scale: 16.0 2023-04-26 23:58:06,077 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9985, 1.5169, 1.8338, 2.0537, 1.7877, 1.3929, 1.0456, 1.5329], device='cuda:6'), covar=tensor([0.3792, 0.3846, 0.1997, 0.2908, 0.3086, 0.3096, 0.5047, 0.2622], device='cuda:6'), in_proj_covar=tensor([0.0284, 0.0251, 0.0221, 0.0318, 0.0214, 0.0229, 0.0234, 0.0186], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:6') 2023-04-26 23:58:15,453 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=53456.0, num_to_drop=0, layers_to_drop=set() 2023-04-26 23:58:40,362 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([3.6406, 3.5223, 2.5714, 4.2324, 3.6484, 3.5917, 1.5088, 3.7092], device='cuda:6'), covar=tensor([0.1670, 0.1298, 0.3011, 0.1785, 0.3146, 0.1898, 0.5942, 0.2380], device='cuda:6'), in_proj_covar=tensor([0.0243, 0.0218, 0.0250, 0.0303, 0.0301, 0.0251, 0.0270, 0.0271], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-26 23:58:48,693 INFO [finetune.py:976] (6/7) Epoch 10, batch 1950, loss[loss=0.1877, simple_loss=0.2506, pruned_loss=0.0624, over 4757.00 frames. ], tot_loss[loss=0.1918, simple_loss=0.2595, pruned_loss=0.06209, over 954386.43 frames. ], batch size: 27, lr: 3.75e-03, grad_scale: 16.0 2023-04-26 23:59:04,283 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2023-04-26 23:59:12,239 INFO [optim.py:369] (6/7) 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 23:59:22,127 INFO [finetune.py:976] (6/7) Epoch 10, batch 2000, loss[loss=0.191, simple_loss=0.2598, pruned_loss=0.06104, over 4818.00 frames. ], tot_loss[loss=0.1899, simple_loss=0.2568, pruned_loss=0.06148, over 954675.47 frames. ], batch size: 41, lr: 3.75e-03, grad_scale: 16.0 2023-04-27 00:00:04,619 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 10, batch 2050, loss[loss=0.1786, simple_loss=0.244, pruned_loss=0.0566, over 4826.00 frames. ], tot_loss[loss=0.1862, simple_loss=0.2528, pruned_loss=0.05977, over 955895.96 frames. ], batch size: 38, lr: 3.75e-03, grad_scale: 16.0 2023-04-27 00:00:34,503 INFO [optim.py:369] (6/7) 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 00:00:35,846 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.3652, 2.2925, 2.5553, 2.9741, 2.7696, 2.2110, 1.9067, 2.5338], device='cuda:6'), covar=tensor([0.0964, 0.1028, 0.0608, 0.0615, 0.0695, 0.1082, 0.0991, 0.0630], device='cuda:6'), in_proj_covar=tensor([0.0196, 0.0203, 0.0182, 0.0176, 0.0178, 0.0188, 0.0160, 0.0185], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 00:00:41,091 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([3.8968, 3.7819, 2.6471, 4.4717, 3.9542, 3.8610, 1.7319, 3.8664], device='cuda:6'), covar=tensor([0.1745, 0.1397, 0.3581, 0.1695, 0.4907, 0.2008, 0.6045, 0.2575], device='cuda:6'), in_proj_covar=tensor([0.0243, 0.0218, 0.0250, 0.0303, 0.0301, 0.0251, 0.0271, 0.0272], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 00:00:44,012 INFO [finetune.py:976] (6/7) Epoch 10, batch 2100, loss[loss=0.2653, simple_loss=0.3112, pruned_loss=0.1098, over 4067.00 frames. ], tot_loss[loss=0.1861, simple_loss=0.2525, pruned_loss=0.05986, over 956462.41 frames. ], batch size: 65, lr: 3.75e-03, grad_scale: 32.0 2023-04-27 00:00:51,833 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=53660.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 00:00:58,521 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=53671.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 00:01:03,989 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-04-27 00:01:16,444 INFO [finetune.py:976] (6/7) Epoch 10, batch 2150, loss[loss=0.2278, simple_loss=0.2959, pruned_loss=0.07985, over 4784.00 frames. ], tot_loss[loss=0.1923, simple_loss=0.2586, pruned_loss=0.06298, over 954504.92 frames. ], batch size: 29, lr: 3.75e-03, grad_scale: 32.0 2023-04-27 00:01:26,037 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=53712.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 00:01:31,968 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.74 vs. limit=2.0 2023-04-27 00:01:32,731 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=53723.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 00:01:38,784 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5823, 3.1988, 1.2941, 2.0185, 2.0217, 2.4413, 2.0526, 1.3575], device='cuda:6'), covar=tensor([0.1298, 0.0861, 0.1676, 0.1181, 0.0948, 0.0975, 0.1320, 0.1967], device='cuda:6'), in_proj_covar=tensor([0.0118, 0.0250, 0.0140, 0.0123, 0.0135, 0.0155, 0.0119, 0.0122], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-27 00:01:41,151 INFO [optim.py:369] (6/7) 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 00:01:49,682 INFO [finetune.py:976] (6/7) Epoch 10, batch 2200, loss[loss=0.214, simple_loss=0.2836, pruned_loss=0.0722, over 4823.00 frames. ], tot_loss[loss=0.1916, simple_loss=0.2585, pruned_loss=0.06237, over 953215.69 frames. ], batch size: 39, lr: 3.75e-03, grad_scale: 32.0 2023-04-27 00:01:55,697 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.60 vs. limit=5.0 2023-04-27 00:01:57,879 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=53760.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 00:02:22,061 INFO [finetune.py:976] (6/7) Epoch 10, batch 2250, loss[loss=0.1747, simple_loss=0.2516, pruned_loss=0.04886, over 4847.00 frames. ], tot_loss[loss=0.1929, simple_loss=0.2599, pruned_loss=0.06298, over 953733.66 frames. ], batch size: 31, lr: 3.75e-03, grad_scale: 32.0 2023-04-27 00:02:46,629 INFO [optim.py:369] (6/7) 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 00:03:06,366 INFO [finetune.py:976] (6/7) Epoch 10, batch 2300, loss[loss=0.1795, simple_loss=0.2538, pruned_loss=0.0526, over 4855.00 frames. ], tot_loss[loss=0.1929, simple_loss=0.2602, pruned_loss=0.06274, over 952966.54 frames. ], batch size: 31, lr: 3.75e-03, grad_scale: 32.0 2023-04-27 00:03:09,880 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9046, 1.7334, 2.2729, 2.3050, 1.6725, 1.4715, 1.8441, 1.1441], device='cuda:6'), covar=tensor([0.0735, 0.0928, 0.0571, 0.1024, 0.0885, 0.1289, 0.0769, 0.0962], device='cuda:6'), in_proj_covar=tensor([0.0066, 0.0073, 0.0070, 0.0066, 0.0076, 0.0095, 0.0077, 0.0073], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-27 00:04:01,733 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.60 vs. limit=5.0 2023-04-27 00:04:10,839 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=53897.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 00:04:11,998 INFO [finetune.py:976] (6/7) Epoch 10, batch 2350, loss[loss=0.152, simple_loss=0.218, pruned_loss=0.04295, over 4820.00 frames. ], tot_loss[loss=0.1907, simple_loss=0.258, pruned_loss=0.0617, over 954294.73 frames. ], batch size: 39, lr: 3.75e-03, grad_scale: 32.0 2023-04-27 00:04:57,265 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-04-27 00:04:57,666 INFO [optim.py:369] (6/7) 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 00:05:08,712 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=53945.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 00:05:16,065 INFO [finetune.py:976] (6/7) Epoch 10, batch 2400, loss[loss=0.2374, simple_loss=0.2859, pruned_loss=0.09443, over 4866.00 frames. ], tot_loss[loss=0.1902, simple_loss=0.2563, pruned_loss=0.06205, over 954835.65 frames. ], batch size: 31, lr: 3.74e-03, grad_scale: 32.0 2023-04-27 00:05:23,338 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=53960.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 00:05:30,759 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5959, 3.6669, 0.9070, 1.9067, 2.0689, 2.5819, 2.1295, 0.9886], device='cuda:6'), covar=tensor([0.1487, 0.1003, 0.2051, 0.1306, 0.1017, 0.1081, 0.1443, 0.2090], device='cuda:6'), in_proj_covar=tensor([0.0118, 0.0249, 0.0140, 0.0122, 0.0134, 0.0154, 0.0119, 0.0121], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-27 00:05:32,023 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=53971.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 00:05:37,704 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.96 vs. limit=2.0 2023-04-27 00:05:49,431 INFO [finetune.py:976] (6/7) Epoch 10, batch 2450, loss[loss=0.1592, simple_loss=0.2248, pruned_loss=0.04678, over 4810.00 frames. ], tot_loss[loss=0.1881, simple_loss=0.2535, pruned_loss=0.06136, over 954924.81 frames. ], batch size: 25, lr: 3.74e-03, grad_scale: 32.0 2023-04-27 00:05:56,751 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=54008.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 00:06:04,947 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=54019.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 00:06:07,916 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=54023.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 00:06:15,673 INFO [optim.py:369] (6/7) 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 00:06:24,054 INFO [finetune.py:976] (6/7) Epoch 10, batch 2500, loss[loss=0.1782, simple_loss=0.2511, pruned_loss=0.05267, over 4818.00 frames. ], tot_loss[loss=0.1889, simple_loss=0.2548, pruned_loss=0.06157, over 953827.71 frames. ], batch size: 33, lr: 3.74e-03, grad_scale: 32.0 2023-04-27 00:06:27,794 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.1863, 2.5285, 1.1416, 1.4166, 2.0109, 1.3114, 3.3064, 1.7091], device='cuda:6'), covar=tensor([0.0637, 0.0732, 0.0759, 0.1373, 0.0495, 0.1031, 0.0317, 0.0665], device='cuda:6'), in_proj_covar=tensor([0.0052, 0.0067, 0.0049, 0.0047, 0.0051, 0.0052, 0.0078, 0.0051], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0007], device='cuda:6') 2023-04-27 00:06:39,520 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=54071.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 00:06:57,697 INFO [finetune.py:976] (6/7) Epoch 10, batch 2550, loss[loss=0.1836, simple_loss=0.2497, pruned_loss=0.05877, over 4777.00 frames. ], tot_loss[loss=0.1916, simple_loss=0.2575, pruned_loss=0.06288, over 952000.60 frames. ], batch size: 28, lr: 3.74e-03, grad_scale: 32.0 2023-04-27 00:06:57,822 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9088, 2.4602, 2.1322, 2.2769, 1.8762, 2.1183, 2.0893, 1.5995], device='cuda:6'), covar=tensor([0.2371, 0.1687, 0.0919, 0.1491, 0.3486, 0.1363, 0.2147, 0.3022], device='cuda:6'), in_proj_covar=tensor([0.0293, 0.0317, 0.0228, 0.0288, 0.0314, 0.0270, 0.0259, 0.0277], device='cuda:6'), out_proj_covar=tensor([1.1872e-04, 1.2760e-04, 9.1509e-05, 1.1556e-04, 1.2848e-04, 1.0868e-04, 1.0575e-04, 1.1111e-04], device='cuda:6') 2023-04-27 00:06:59,054 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.0436, 2.0312, 1.6461, 1.6552, 2.1868, 1.7753, 2.5274, 1.5420], device='cuda:6'), covar=tensor([0.3747, 0.1614, 0.4625, 0.2938, 0.1494, 0.2217, 0.1399, 0.4440], device='cuda:6'), in_proj_covar=tensor([0.0344, 0.0348, 0.0432, 0.0362, 0.0387, 0.0384, 0.0380, 0.0420], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 00:06:59,680 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5808, 2.0953, 1.5213, 1.3694, 1.1569, 1.2015, 1.5713, 1.0978], device='cuda:6'), covar=tensor([0.1712, 0.1259, 0.1449, 0.1868, 0.2525, 0.2013, 0.1130, 0.2124], device='cuda:6'), in_proj_covar=tensor([0.0195, 0.0215, 0.0170, 0.0205, 0.0204, 0.0185, 0.0161, 0.0188], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:6') 2023-04-27 00:07:16,110 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=3.85 vs. limit=5.0 2023-04-27 00:07:22,641 INFO [optim.py:369] (6/7) 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] (6/7) Epoch 10, batch 2600, loss[loss=0.2005, simple_loss=0.264, pruned_loss=0.06847, over 4708.00 frames. ], tot_loss[loss=0.193, simple_loss=0.2592, pruned_loss=0.06342, over 951842.91 frames. ], batch size: 59, lr: 3.74e-03, grad_scale: 32.0 2023-04-27 00:07:31,601 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 2023-04-27 00:07:57,814 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6396, 1.2188, 1.3658, 1.4739, 1.7168, 1.5124, 1.2988, 1.2952], device='cuda:6'), covar=tensor([0.1338, 0.1072, 0.1350, 0.1008, 0.0770, 0.1243, 0.1997, 0.1577], device='cuda:6'), in_proj_covar=tensor([0.0305, 0.0321, 0.0353, 0.0297, 0.0334, 0.0319, 0.0307, 0.0358], device='cuda:6'), out_proj_covar=tensor([6.4296e-05, 6.7915e-05, 7.6172e-05, 6.1279e-05, 6.9727e-05, 6.8447e-05, 6.5796e-05, 7.6929e-05], device='cuda:6') 2023-04-27 00:08:02,059 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.2283, 3.0031, 0.7758, 1.6365, 1.6558, 2.1181, 1.7162, 0.9457], device='cuda:6'), covar=tensor([0.1572, 0.1074, 0.2224, 0.1396, 0.1197, 0.1108, 0.1646, 0.1982], device='cuda:6'), in_proj_covar=tensor([0.0118, 0.0250, 0.0140, 0.0123, 0.0135, 0.0155, 0.0120, 0.0122], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-27 00:08:04,394 INFO [finetune.py:976] (6/7) Epoch 10, batch 2650, loss[loss=0.1977, simple_loss=0.2822, pruned_loss=0.05663, over 4937.00 frames. ], tot_loss[loss=0.1937, simple_loss=0.2605, pruned_loss=0.06349, over 951648.37 frames. ], batch size: 42, lr: 3.74e-03, grad_scale: 32.0 2023-04-27 00:08:11,222 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5851, 1.9725, 1.7338, 1.8893, 1.4666, 1.5862, 1.7241, 1.2776], device='cuda:6'), covar=tensor([0.1856, 0.1101, 0.0796, 0.1092, 0.3245, 0.1192, 0.1587, 0.2269], device='cuda:6'), in_proj_covar=tensor([0.0293, 0.0316, 0.0228, 0.0288, 0.0314, 0.0270, 0.0258, 0.0278], device='cuda:6'), out_proj_covar=tensor([1.1866e-04, 1.2726e-04, 9.1590e-05, 1.1529e-04, 1.2850e-04, 1.0861e-04, 1.0553e-04, 1.1138e-04], device='cuda:6') 2023-04-27 00:08:39,511 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.14 vs. limit=2.0 2023-04-27 00:08:39,722 INFO [optim.py:369] (6/7) 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 00:08:49,391 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.2751, 1.6333, 1.5376, 2.2030, 1.8965, 1.9967, 1.7520, 4.3291], device='cuda:6'), covar=tensor([0.0793, 0.1126, 0.1103, 0.1332, 0.0835, 0.0738, 0.1033, 0.0196], device='cuda:6'), in_proj_covar=tensor([0.0039, 0.0039, 0.0040, 0.0044, 0.0040, 0.0039, 0.0039, 0.0059], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0017], device='cuda:6') 2023-04-27 00:08:52,969 INFO [finetune.py:976] (6/7) Epoch 10, batch 2700, loss[loss=0.1984, simple_loss=0.2565, pruned_loss=0.0702, over 4902.00 frames. ], tot_loss[loss=0.1929, simple_loss=0.2602, pruned_loss=0.06283, over 952135.99 frames. ], batch size: 35, lr: 3.74e-03, grad_scale: 32.0 2023-04-27 00:09:12,488 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6012, 1.1560, 1.7247, 2.1708, 1.7401, 1.5977, 1.6656, 1.6216], device='cuda:6'), covar=tensor([0.5638, 0.8125, 0.7694, 0.7047, 0.7182, 0.8977, 0.9762, 0.9544], device='cuda:6'), in_proj_covar=tensor([0.0408, 0.0413, 0.0497, 0.0516, 0.0437, 0.0455, 0.0465, 0.0463], device='cuda:6'), out_proj_covar=tensor([9.9243e-05, 1.0247e-04, 1.1213e-04, 1.2277e-04, 1.0601e-04, 1.1001e-04, 1.1154e-04, 1.1169e-04], device='cuda:6') 2023-04-27 00:09:34,249 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-04-27 00:10:03,552 INFO [finetune.py:976] (6/7) Epoch 10, batch 2750, loss[loss=0.2615, simple_loss=0.3015, pruned_loss=0.1107, over 4133.00 frames. ], tot_loss[loss=0.1903, simple_loss=0.2577, pruned_loss=0.06151, over 953055.79 frames. ], batch size: 65, lr: 3.74e-03, grad_scale: 32.0 2023-04-27 00:10:15,091 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8296, 1.1639, 1.6665, 1.9511, 1.6823, 1.3331, 0.8479, 1.3418], device='cuda:6'), covar=tensor([0.4338, 0.4883, 0.2374, 0.3222, 0.3010, 0.3319, 0.5635, 0.2849], device='cuda:6'), in_proj_covar=tensor([0.0281, 0.0249, 0.0219, 0.0314, 0.0212, 0.0227, 0.0232, 0.0184], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:6') 2023-04-27 00:10:49,707 INFO [optim.py:369] (6/7) 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 00:10:58,952 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.2961, 2.2184, 2.5060, 2.7719, 2.7981, 2.0394, 1.8765, 2.4423], device='cuda:6'), covar=tensor([0.0908, 0.1039, 0.0619, 0.0611, 0.0625, 0.0976, 0.0939, 0.0575], device='cuda:6'), in_proj_covar=tensor([0.0196, 0.0203, 0.0182, 0.0175, 0.0178, 0.0189, 0.0160, 0.0184], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 00:10:58,979 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.3336, 1.4972, 2.1145, 2.7733, 2.3182, 1.6509, 1.5876, 2.0519], device='cuda:6'), covar=tensor([0.3963, 0.4415, 0.2127, 0.3262, 0.3185, 0.3115, 0.5065, 0.2688], device='cuda:6'), in_proj_covar=tensor([0.0281, 0.0249, 0.0219, 0.0313, 0.0212, 0.0227, 0.0232, 0.0184], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:6') 2023-04-27 00:11:09,556 INFO [finetune.py:976] (6/7) Epoch 10, batch 2800, loss[loss=0.1768, simple_loss=0.2435, pruned_loss=0.05503, over 4808.00 frames. ], tot_loss[loss=0.1883, simple_loss=0.2549, pruned_loss=0.06085, over 953115.78 frames. ], batch size: 25, lr: 3.74e-03, grad_scale: 32.0 2023-04-27 00:11:48,128 INFO [finetune.py:976] (6/7) Epoch 10, batch 2850, loss[loss=0.1797, simple_loss=0.2413, pruned_loss=0.05902, over 4872.00 frames. ], tot_loss[loss=0.1872, simple_loss=0.2534, pruned_loss=0.06049, over 954525.95 frames. ], batch size: 31, lr: 3.74e-03, grad_scale: 32.0 2023-04-27 00:12:11,874 INFO [optim.py:369] (6/7) 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 00:12:21,807 INFO [finetune.py:976] (6/7) Epoch 10, batch 2900, loss[loss=0.1653, simple_loss=0.2288, pruned_loss=0.05092, over 4195.00 frames. ], tot_loss[loss=0.1916, simple_loss=0.2577, pruned_loss=0.06275, over 955371.15 frames. ], batch size: 18, lr: 3.74e-03, grad_scale: 32.0 2023-04-27 00:12:55,779 INFO [finetune.py:976] (6/7) Epoch 10, batch 2950, loss[loss=0.1935, simple_loss=0.2664, pruned_loss=0.0603, over 4901.00 frames. ], tot_loss[loss=0.1935, simple_loss=0.2606, pruned_loss=0.06317, over 956458.12 frames. ], batch size: 35, lr: 3.74e-03, grad_scale: 32.0 2023-04-27 00:13:19,230 INFO [optim.py:369] (6/7) 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 00:13:24,632 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.2526, 1.4732, 1.6584, 1.8057, 1.6689, 1.7851, 1.7931, 1.6815], device='cuda:6'), covar=tensor([0.5114, 0.6572, 0.5567, 0.5133, 0.6182, 0.8934, 0.5835, 0.6028], device='cuda:6'), in_proj_covar=tensor([0.0322, 0.0377, 0.0312, 0.0323, 0.0335, 0.0397, 0.0356, 0.0320], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 00:13:29,123 INFO [finetune.py:976] (6/7) Epoch 10, batch 3000, loss[loss=0.1985, simple_loss=0.2572, pruned_loss=0.06992, over 4918.00 frames. ], tot_loss[loss=0.1951, simple_loss=0.2624, pruned_loss=0.0639, over 955278.33 frames. ], batch size: 33, lr: 3.74e-03, grad_scale: 32.0 2023-04-27 00:13:29,123 INFO [finetune.py:1001] (6/7) Computing validation loss 2023-04-27 00:13:45,197 INFO [finetune.py:1010] (6/7) 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] (6/7) Maximum memory allocated so far is 6345MB 2023-04-27 00:14:32,348 INFO [finetune.py:976] (6/7) Epoch 10, batch 3050, loss[loss=0.2467, simple_loss=0.3131, pruned_loss=0.09014, over 4692.00 frames. ], tot_loss[loss=0.1944, simple_loss=0.2619, pruned_loss=0.0634, over 956315.44 frames. ], batch size: 59, lr: 3.74e-03, grad_scale: 32.0 2023-04-27 00:14:34,574 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=5.08 vs. limit=5.0 2023-04-27 00:14:45,726 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8157, 1.6184, 2.1030, 2.2123, 1.6297, 1.3279, 1.7097, 0.8792], device='cuda:6'), covar=tensor([0.0728, 0.1067, 0.0672, 0.1013, 0.1008, 0.1483, 0.1026, 0.1184], device='cuda:6'), in_proj_covar=tensor([0.0066, 0.0072, 0.0071, 0.0067, 0.0075, 0.0095, 0.0077, 0.0073], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-27 00:14:57,003 INFO [optim.py:369] (6/7) 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 00:15:05,020 INFO [finetune.py:976] (6/7) Epoch 10, batch 3100, loss[loss=0.1589, simple_loss=0.2282, pruned_loss=0.04481, over 4770.00 frames. ], tot_loss[loss=0.1931, simple_loss=0.2603, pruned_loss=0.06295, over 957960.82 frames. ], batch size: 26, lr: 3.74e-03, grad_scale: 32.0 2023-04-27 00:15:26,419 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-04-27 00:15:27,259 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=54664.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 00:16:11,753 INFO [finetune.py:976] (6/7) Epoch 10, batch 3150, loss[loss=0.1445, simple_loss=0.2164, pruned_loss=0.03629, over 4782.00 frames. ], tot_loss[loss=0.1901, simple_loss=0.2567, pruned_loss=0.06178, over 956497.76 frames. ], batch size: 26, lr: 3.74e-03, grad_scale: 32.0 2023-04-27 00:16:22,931 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-04-27 00:16:53,640 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=54725.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 00:17:04,897 INFO [optim.py:369] (6/7) 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 00:17:25,053 INFO [finetune.py:976] (6/7) Epoch 10, batch 3200, loss[loss=0.1839, simple_loss=0.2299, pruned_loss=0.06895, over 4816.00 frames. ], tot_loss[loss=0.1881, simple_loss=0.2541, pruned_loss=0.06112, over 958430.16 frames. ], batch size: 25, lr: 3.74e-03, grad_scale: 32.0 2023-04-27 00:17:46,488 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.3252, 1.7822, 2.2585, 2.6763, 2.1805, 1.7216, 1.4871, 1.9726], device='cuda:6'), covar=tensor([0.3939, 0.3772, 0.1820, 0.2914, 0.3224, 0.2989, 0.4671, 0.2557], device='cuda:6'), in_proj_covar=tensor([0.0283, 0.0250, 0.0219, 0.0316, 0.0213, 0.0227, 0.0233, 0.0185], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:6') 2023-04-27 00:18:08,048 INFO [finetune.py:976] (6/7) Epoch 10, batch 3250, loss[loss=0.2006, simple_loss=0.2703, pruned_loss=0.06541, over 4892.00 frames. ], tot_loss[loss=0.1886, simple_loss=0.2544, pruned_loss=0.06145, over 958214.70 frames. ], batch size: 35, lr: 3.74e-03, grad_scale: 32.0 2023-04-27 00:18:33,643 INFO [optim.py:369] (6/7) 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 00:18:38,506 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.1773, 2.0622, 2.2453, 2.5835, 2.6051, 1.9520, 1.8123, 2.1971], device='cuda:6'), covar=tensor([0.0884, 0.1097, 0.0648, 0.0594, 0.0680, 0.0894, 0.0900, 0.0651], device='cuda:6'), in_proj_covar=tensor([0.0196, 0.0204, 0.0183, 0.0177, 0.0179, 0.0189, 0.0160, 0.0185], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 00:18:41,640 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.22 vs. limit=5.0 2023-04-27 00:18:42,048 INFO [finetune.py:976] (6/7) Epoch 10, batch 3300, loss[loss=0.2129, simple_loss=0.2903, pruned_loss=0.06775, over 4817.00 frames. ], tot_loss[loss=0.1918, simple_loss=0.258, pruned_loss=0.06283, over 956888.66 frames. ], batch size: 38, lr: 3.74e-03, grad_scale: 32.0 2023-04-27 00:18:50,601 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=54862.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 00:19:08,381 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.4377, 1.7447, 1.7326, 1.9198, 1.7447, 1.8447, 1.8507, 1.7659], device='cuda:6'), covar=tensor([0.4582, 0.6902, 0.5852, 0.5360, 0.6409, 0.9389, 0.6773, 0.6510], device='cuda:6'), in_proj_covar=tensor([0.0325, 0.0379, 0.0315, 0.0324, 0.0338, 0.0399, 0.0358, 0.0323], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 00:19:15,374 INFO [finetune.py:976] (6/7) Epoch 10, batch 3350, loss[loss=0.209, simple_loss=0.2698, pruned_loss=0.07409, over 4787.00 frames. ], tot_loss[loss=0.1934, simple_loss=0.2602, pruned_loss=0.0633, over 956823.82 frames. ], batch size: 54, lr: 3.74e-03, grad_scale: 32.0 2023-04-27 00:19:31,045 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=54923.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 00:19:39,814 INFO [optim.py:369] (6/7) 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] (6/7) Epoch 10, batch 3400, loss[loss=0.1922, simple_loss=0.2647, pruned_loss=0.05984, over 4804.00 frames. ], tot_loss[loss=0.193, simple_loss=0.26, pruned_loss=0.063, over 954291.71 frames. ], batch size: 40, lr: 3.74e-03, grad_scale: 32.0 2023-04-27 00:20:20,498 INFO [finetune.py:976] (6/7) Epoch 10, batch 3450, loss[loss=0.187, simple_loss=0.251, pruned_loss=0.06154, over 4813.00 frames. ], tot_loss[loss=0.1916, simple_loss=0.2591, pruned_loss=0.06205, over 957022.64 frames. ], batch size: 25, lr: 3.74e-03, grad_scale: 32.0 2023-04-27 00:20:34,364 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=55020.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 00:20:45,488 INFO [optim.py:369] (6/7) 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 00:20:53,438 INFO [finetune.py:976] (6/7) Epoch 10, batch 3500, loss[loss=0.1824, simple_loss=0.2389, pruned_loss=0.06294, over 4833.00 frames. ], tot_loss[loss=0.1889, simple_loss=0.256, pruned_loss=0.0609, over 957710.13 frames. ], batch size: 33, lr: 3.74e-03, grad_scale: 32.0 2023-04-27 00:21:05,028 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.41 vs. limit=5.0 2023-04-27 00:21:37,261 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-04-27 00:21:37,474 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6756, 3.9856, 0.6385, 2.1342, 2.3567, 2.8402, 2.3689, 0.9404], device='cuda:6'), covar=tensor([0.1351, 0.0941, 0.2261, 0.1356, 0.0956, 0.0987, 0.1295, 0.2142], device='cuda:6'), in_proj_covar=tensor([0.0119, 0.0254, 0.0142, 0.0124, 0.0137, 0.0156, 0.0120, 0.0123], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-27 00:21:46,486 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-04-27 00:21:46,851 INFO [finetune.py:976] (6/7) Epoch 10, batch 3550, loss[loss=0.1739, simple_loss=0.2312, pruned_loss=0.05829, over 4931.00 frames. ], tot_loss[loss=0.1869, simple_loss=0.2531, pruned_loss=0.06034, over 956702.52 frames. ], batch size: 33, lr: 3.74e-03, grad_scale: 32.0 2023-04-27 00:22:27,076 INFO [optim.py:369] (6/7) 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 00:22:30,203 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-04-27 00:22:35,880 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 10, batch 3600, loss[loss=0.2278, simple_loss=0.284, pruned_loss=0.08577, over 4824.00 frames. ], tot_loss[loss=0.1865, simple_loss=0.2519, pruned_loss=0.06054, over 955845.42 frames. ], batch size: 40, lr: 3.74e-03, grad_scale: 32.0 2023-04-27 00:23:26,775 INFO [finetune.py:976] (6/7) Epoch 10, batch 3650, loss[loss=0.2274, simple_loss=0.2883, pruned_loss=0.0832, over 4733.00 frames. ], tot_loss[loss=0.1872, simple_loss=0.253, pruned_loss=0.06066, over 955801.12 frames. ], batch size: 54, lr: 3.74e-03, grad_scale: 32.0 2023-04-27 00:23:28,006 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-04-27 00:23:32,480 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([3.3318, 3.2408, 2.4974, 3.8722, 3.3777, 3.4089, 1.4031, 3.3214], device='cuda:6'), covar=tensor([0.1840, 0.1467, 0.3143, 0.2237, 0.4067, 0.1875, 0.5580, 0.2591], device='cuda:6'), in_proj_covar=tensor([0.0244, 0.0216, 0.0249, 0.0305, 0.0300, 0.0250, 0.0270, 0.0271], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 00:23:33,132 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=55209.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 00:23:34,355 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=55211.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 00:23:39,099 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=55218.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 00:23:41,366 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-04-27 00:23:45,731 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([3.3681, 3.3574, 2.5671, 4.0100, 3.4232, 3.4106, 1.4739, 3.3979], device='cuda:6'), covar=tensor([0.1678, 0.1396, 0.3687, 0.2002, 0.3308, 0.1818, 0.5482, 0.2772], device='cuda:6'), in_proj_covar=tensor([0.0243, 0.0216, 0.0249, 0.0306, 0.0300, 0.0250, 0.0270, 0.0271], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 00:23:50,724 INFO [optim.py:369] (6/7) 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 00:24:00,575 INFO [finetune.py:976] (6/7) Epoch 10, batch 3700, loss[loss=0.181, simple_loss=0.2503, pruned_loss=0.05591, over 4929.00 frames. ], tot_loss[loss=0.1914, simple_loss=0.2576, pruned_loss=0.06264, over 955004.37 frames. ], batch size: 33, lr: 3.73e-03, grad_scale: 32.0 2023-04-27 00:24:15,033 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=55272.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 00:24:33,931 INFO [finetune.py:976] (6/7) Epoch 10, batch 3750, loss[loss=0.2487, simple_loss=0.31, pruned_loss=0.09376, over 4891.00 frames. ], tot_loss[loss=0.1918, simple_loss=0.2583, pruned_loss=0.06271, over 953955.42 frames. ], batch size: 35, lr: 3.73e-03, grad_scale: 32.0 2023-04-27 00:24:47,359 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=55320.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 00:24:57,142 INFO [optim.py:369] (6/7) 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 00:25:07,103 INFO [finetune.py:976] (6/7) Epoch 10, batch 3800, loss[loss=0.179, simple_loss=0.2491, pruned_loss=0.05445, over 4789.00 frames. ], tot_loss[loss=0.1933, simple_loss=0.2603, pruned_loss=0.06319, over 953299.37 frames. ], batch size: 29, lr: 3.73e-03, grad_scale: 32.0 2023-04-27 00:25:16,098 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.2047, 2.9003, 0.8518, 1.5350, 1.8916, 1.3876, 3.8904, 1.7904], device='cuda:6'), covar=tensor([0.0704, 0.1005, 0.0935, 0.1210, 0.0576, 0.0988, 0.0235, 0.0652], device='cuda:6'), in_proj_covar=tensor([0.0052, 0.0067, 0.0050, 0.0047, 0.0051, 0.0052, 0.0078, 0.0051], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0007], device='cuda:6') 2023-04-27 00:25:19,741 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=55368.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 00:25:39,926 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.77 vs. limit=5.0 2023-04-27 00:25:40,049 INFO [finetune.py:976] (6/7) Epoch 10, batch 3850, loss[loss=0.1895, simple_loss=0.2456, pruned_loss=0.06669, over 4144.00 frames. ], tot_loss[loss=0.1932, simple_loss=0.2601, pruned_loss=0.06321, over 953231.15 frames. ], batch size: 18, lr: 3.73e-03, grad_scale: 32.0 2023-04-27 00:26:04,612 INFO [optim.py:369] (6/7) 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 00:26:08,262 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.1097, 1.8407, 2.2800, 2.5107, 2.1099, 1.9441, 2.1647, 2.0781], device='cuda:6'), covar=tensor([0.6115, 0.8492, 0.9004, 0.8121, 0.7845, 1.1320, 1.1734, 1.0620], device='cuda:6'), in_proj_covar=tensor([0.0408, 0.0414, 0.0497, 0.0517, 0.0437, 0.0456, 0.0467, 0.0464], device='cuda:6'), out_proj_covar=tensor([9.9328e-05, 1.0247e-04, 1.1221e-04, 1.2280e-04, 1.0602e-04, 1.1014e-04, 1.1184e-04, 1.1166e-04], device='cuda:6') 2023-04-27 00:26:12,962 INFO [finetune.py:976] (6/7) Epoch 10, batch 3900, loss[loss=0.1926, simple_loss=0.2557, pruned_loss=0.06471, over 4744.00 frames. ], tot_loss[loss=0.1896, simple_loss=0.2564, pruned_loss=0.06138, over 954813.01 frames. ], batch size: 54, lr: 3.73e-03, grad_scale: 32.0 2023-04-27 00:27:07,762 INFO [finetune.py:976] (6/7) Epoch 10, batch 3950, loss[loss=0.1752, simple_loss=0.2377, pruned_loss=0.05637, over 4782.00 frames. ], tot_loss[loss=0.1873, simple_loss=0.2533, pruned_loss=0.0606, over 957140.14 frames. ], batch size: 29, lr: 3.73e-03, grad_scale: 32.0 2023-04-27 00:27:15,659 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=55504.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 00:27:36,393 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=55518.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 00:27:47,672 INFO [optim.py:369] (6/7) 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] (6/7) Epoch 10, batch 4000, loss[loss=0.2193, simple_loss=0.2833, pruned_loss=0.0776, over 4840.00 frames. ], tot_loss[loss=0.188, simple_loss=0.2538, pruned_loss=0.06108, over 958396.95 frames. ], batch size: 39, lr: 3.73e-03, grad_scale: 32.0 2023-04-27 00:28:09,058 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=55566.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 00:28:12,128 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=55567.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 00:28:43,371 INFO [finetune.py:976] (6/7) Epoch 10, batch 4050, loss[loss=0.2251, simple_loss=0.2887, pruned_loss=0.08072, over 4748.00 frames. ], tot_loss[loss=0.1909, simple_loss=0.257, pruned_loss=0.06243, over 956160.72 frames. ], batch size: 54, lr: 3.73e-03, grad_scale: 32.0 2023-04-27 00:29:09,491 INFO [optim.py:369] (6/7) 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 00:29:15,159 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9403, 1.7679, 2.1251, 2.3839, 1.9826, 1.8851, 2.0369, 2.0060], device='cuda:6'), covar=tensor([0.6219, 0.8418, 0.8922, 0.7676, 0.7735, 1.0943, 1.1088, 1.0482], device='cuda:6'), in_proj_covar=tensor([0.0409, 0.0412, 0.0498, 0.0517, 0.0438, 0.0457, 0.0467, 0.0465], device='cuda:6'), out_proj_covar=tensor([9.9666e-05, 1.0221e-04, 1.1250e-04, 1.2282e-04, 1.0615e-04, 1.1043e-04, 1.1193e-04, 1.1191e-04], device='cuda:6') 2023-04-27 00:29:16,839 INFO [finetune.py:976] (6/7) Epoch 10, batch 4100, loss[loss=0.1875, simple_loss=0.2494, pruned_loss=0.06285, over 4837.00 frames. ], tot_loss[loss=0.1936, simple_loss=0.26, pruned_loss=0.06359, over 956937.55 frames. ], batch size: 30, lr: 3.73e-03, grad_scale: 32.0 2023-04-27 00:29:50,669 INFO [finetune.py:976] (6/7) Epoch 10, batch 4150, loss[loss=0.2216, simple_loss=0.2867, pruned_loss=0.07828, over 4790.00 frames. ], tot_loss[loss=0.1939, simple_loss=0.2611, pruned_loss=0.06338, over 957228.26 frames. ], batch size: 51, lr: 3.73e-03, grad_scale: 32.0 2023-04-27 00:30:16,063 INFO [optim.py:369] (6/7) 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 00:30:23,276 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=55748.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 00:30:23,764 INFO [finetune.py:976] (6/7) Epoch 10, batch 4200, loss[loss=0.1405, simple_loss=0.2127, pruned_loss=0.03415, over 4790.00 frames. ], tot_loss[loss=0.1928, simple_loss=0.2602, pruned_loss=0.0627, over 955186.24 frames. ], batch size: 29, lr: 3.73e-03, grad_scale: 32.0 2023-04-27 00:30:57,483 INFO [finetune.py:976] (6/7) Epoch 10, batch 4250, loss[loss=0.1816, simple_loss=0.2428, pruned_loss=0.06019, over 4684.00 frames. ], tot_loss[loss=0.1919, simple_loss=0.2587, pruned_loss=0.06257, over 956170.90 frames. ], batch size: 23, lr: 3.73e-03, grad_scale: 32.0 2023-04-27 00:30:57,795 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 2023-04-27 00:31:00,758 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=55804.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 00:31:04,336 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=55809.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 00:31:05,505 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-04-27 00:31:23,549 INFO [optim.py:369] (6/7) 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 00:31:30,789 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=55848.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 00:31:31,266 INFO [finetune.py:976] (6/7) Epoch 10, batch 4300, loss[loss=0.1933, simple_loss=0.2502, pruned_loss=0.06817, over 4796.00 frames. ], tot_loss[loss=0.1892, simple_loss=0.2556, pruned_loss=0.06136, over 956115.94 frames. ], batch size: 26, lr: 3.73e-03, grad_scale: 32.0 2023-04-27 00:31:33,044 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=55852.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 00:31:36,162 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6075, 3.3021, 0.8945, 1.7721, 1.8528, 2.2118, 1.9050, 0.9206], device='cuda:6'), covar=tensor([0.1264, 0.0881, 0.1947, 0.1343, 0.1078, 0.1080, 0.1420, 0.1933], device='cuda:6'), in_proj_covar=tensor([0.0118, 0.0251, 0.0142, 0.0123, 0.0136, 0.0154, 0.0118, 0.0122], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-27 00:31:39,198 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5613, 2.0313, 1.7024, 1.8168, 1.5195, 1.6374, 1.7309, 1.2844], device='cuda:6'), covar=tensor([0.1925, 0.1239, 0.0877, 0.1331, 0.2999, 0.1149, 0.1710, 0.2356], device='cuda:6'), in_proj_covar=tensor([0.0299, 0.0320, 0.0231, 0.0292, 0.0317, 0.0272, 0.0259, 0.0281], device='cuda:6'), out_proj_covar=tensor([1.2105e-04, 1.2879e-04, 9.2653e-05, 1.1695e-04, 1.2990e-04, 1.0964e-04, 1.0542e-04, 1.1287e-04], device='cuda:6') 2023-04-27 00:31:45,836 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.3070, 2.9569, 0.9135, 1.6902, 1.6636, 2.1098, 1.7722, 1.0343], device='cuda:6'), covar=tensor([0.1366, 0.1102, 0.1801, 0.1265, 0.1120, 0.1010, 0.1423, 0.1783], device='cuda:6'), in_proj_covar=tensor([0.0118, 0.0251, 0.0142, 0.0123, 0.0136, 0.0155, 0.0119, 0.0122], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-27 00:31:48,285 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=55867.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 00:31:59,066 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-04-27 00:32:31,698 INFO [finetune.py:976] (6/7) Epoch 10, batch 4350, loss[loss=0.1517, simple_loss=0.2096, pruned_loss=0.04692, over 4802.00 frames. ], tot_loss[loss=0.1865, simple_loss=0.2523, pruned_loss=0.06034, over 955033.86 frames. ], batch size: 23, lr: 3.73e-03, grad_scale: 32.0 2023-04-27 00:32:43,389 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=55909.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 00:32:46,928 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=55915.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 00:32:56,458 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6251, 1.6046, 1.8376, 2.0611, 2.0851, 1.5615, 1.2128, 1.8696], device='cuda:6'), covar=tensor([0.0930, 0.1187, 0.0758, 0.0598, 0.0617, 0.0927, 0.0986, 0.0603], device='cuda:6'), in_proj_covar=tensor([0.0198, 0.0205, 0.0183, 0.0177, 0.0179, 0.0190, 0.0161, 0.0185], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 00:33:02,712 INFO [optim.py:369] (6/7) 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] (6/7) Epoch 10, batch 4400, loss[loss=0.1976, simple_loss=0.2663, pruned_loss=0.06448, over 4829.00 frames. ], tot_loss[loss=0.1881, simple_loss=0.2539, pruned_loss=0.0611, over 953797.31 frames. ], batch size: 33, lr: 3.73e-03, grad_scale: 32.0 2023-04-27 00:33:20,591 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7629, 1.2560, 1.5795, 1.6135, 1.5332, 1.2544, 0.6813, 1.3057], device='cuda:6'), covar=tensor([0.3590, 0.3611, 0.1828, 0.2464, 0.2760, 0.2789, 0.4839, 0.2359], device='cuda:6'), in_proj_covar=tensor([0.0285, 0.0250, 0.0220, 0.0317, 0.0214, 0.0228, 0.0234, 0.0186], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:6') 2023-04-27 00:33:59,031 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.1718, 1.4662, 1.3184, 1.6299, 1.4991, 1.8262, 1.3598, 3.2226], device='cuda:6'), covar=tensor([0.0662, 0.0800, 0.0777, 0.1167, 0.0667, 0.0532, 0.0751, 0.0160], device='cuda:6'), in_proj_covar=tensor([0.0038, 0.0039, 0.0040, 0.0044, 0.0040, 0.0039, 0.0039, 0.0059], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0014, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:6') 2023-04-27 00:34:18,724 INFO [finetune.py:976] (6/7) Epoch 10, batch 4450, loss[loss=0.1546, simple_loss=0.2233, pruned_loss=0.04293, over 4727.00 frames. ], tot_loss[loss=0.1902, simple_loss=0.2568, pruned_loss=0.06177, over 951976.11 frames. ], batch size: 23, lr: 3.73e-03, grad_scale: 32.0 2023-04-27 00:34:46,901 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8740, 1.1820, 4.5612, 4.2470, 4.0187, 4.2175, 4.0883, 4.1031], device='cuda:6'), covar=tensor([0.6602, 0.6166, 0.1041, 0.1771, 0.0974, 0.1300, 0.2381, 0.1382], device='cuda:6'), in_proj_covar=tensor([0.0308, 0.0309, 0.0406, 0.0411, 0.0350, 0.0405, 0.0315, 0.0369], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-27 00:34:47,924 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-04-27 00:34:57,405 INFO [optim.py:369] (6/7) 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 00:35:05,272 INFO [finetune.py:976] (6/7) Epoch 10, batch 4500, loss[loss=0.1812, simple_loss=0.2713, pruned_loss=0.04558, over 4807.00 frames. ], tot_loss[loss=0.1929, simple_loss=0.2597, pruned_loss=0.063, over 954247.07 frames. ], batch size: 40, lr: 3.73e-03, grad_scale: 32.0 2023-04-27 00:35:10,710 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0081, 1.1252, 5.1368, 4.8165, 4.4991, 4.8497, 4.5553, 4.5866], device='cuda:6'), covar=tensor([0.6491, 0.7051, 0.0963, 0.1685, 0.0973, 0.1254, 0.1267, 0.1442], device='cuda:6'), in_proj_covar=tensor([0.0309, 0.0309, 0.0406, 0.0411, 0.0350, 0.0406, 0.0316, 0.0369], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-27 00:35:38,736 INFO [finetune.py:976] (6/7) Epoch 10, batch 4550, loss[loss=0.1955, simple_loss=0.2608, pruned_loss=0.06506, over 4858.00 frames. ], tot_loss[loss=0.1942, simple_loss=0.2615, pruned_loss=0.06348, over 954093.20 frames. ], batch size: 31, lr: 3.73e-03, grad_scale: 32.0 2023-04-27 00:35:41,837 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=56104.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 00:36:03,336 INFO [optim.py:369] (6/7) 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 00:36:09,761 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=56145.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 00:36:10,389 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8530, 2.3632, 1.9186, 2.1389, 1.5244, 1.9875, 1.9699, 1.5114], device='cuda:6'), covar=tensor([0.2113, 0.1151, 0.1027, 0.1440, 0.3589, 0.1194, 0.1942, 0.2529], device='cuda:6'), in_proj_covar=tensor([0.0300, 0.0322, 0.0232, 0.0293, 0.0320, 0.0274, 0.0260, 0.0284], device='cuda:6'), out_proj_covar=tensor([1.2117e-04, 1.2946e-04, 9.3210e-05, 1.1763e-04, 1.3099e-04, 1.1017e-04, 1.0590e-04, 1.1387e-04], device='cuda:6') 2023-04-27 00:36:12,140 INFO [finetune.py:976] (6/7) Epoch 10, batch 4600, loss[loss=0.161, simple_loss=0.2256, pruned_loss=0.04814, over 4757.00 frames. ], tot_loss[loss=0.1932, simple_loss=0.2601, pruned_loss=0.06312, over 953329.45 frames. ], batch size: 23, lr: 3.73e-03, grad_scale: 32.0 2023-04-27 00:36:45,850 INFO [finetune.py:976] (6/7) Epoch 10, batch 4650, loss[loss=0.1535, simple_loss=0.2184, pruned_loss=0.04431, over 4914.00 frames. ], tot_loss[loss=0.1908, simple_loss=0.2573, pruned_loss=0.06215, over 953277.80 frames. ], batch size: 37, lr: 3.73e-03, grad_scale: 32.0 2023-04-27 00:36:45,978 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8356, 1.3912, 1.4031, 1.6895, 2.0047, 1.6053, 1.3706, 1.2894], device='cuda:6'), covar=tensor([0.1481, 0.1627, 0.1957, 0.1180, 0.0850, 0.1787, 0.2344, 0.1976], device='cuda:6'), in_proj_covar=tensor([0.0307, 0.0322, 0.0355, 0.0300, 0.0339, 0.0321, 0.0307, 0.0361], device='cuda:6'), out_proj_covar=tensor([6.4541e-05, 6.8305e-05, 7.6510e-05, 6.1927e-05, 7.0988e-05, 6.8687e-05, 6.5663e-05, 7.7488e-05], device='cuda:6') 2023-04-27 00:36:49,601 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=56204.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 00:36:50,857 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=56206.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 00:37:16,886 INFO [optim.py:369] (6/7) 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 00:37:24,882 INFO [finetune.py:976] (6/7) Epoch 10, batch 4700, loss[loss=0.1566, simple_loss=0.2281, pruned_loss=0.04261, over 4823.00 frames. ], tot_loss[loss=0.1876, simple_loss=0.2537, pruned_loss=0.0607, over 952806.94 frames. ], batch size: 40, lr: 3.73e-03, grad_scale: 32.0 2023-04-27 00:37:28,574 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([0.8127, 1.8033, 1.7598, 1.4363, 1.9359, 1.6183, 2.4556, 1.5530], device='cuda:6'), covar=tensor([0.3853, 0.1933, 0.4533, 0.3086, 0.1866, 0.2405, 0.1543, 0.4494], device='cuda:6'), in_proj_covar=tensor([0.0345, 0.0349, 0.0431, 0.0363, 0.0387, 0.0385, 0.0383, 0.0421], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 00:37:57,894 INFO [finetune.py:976] (6/7) Epoch 10, batch 4750, loss[loss=0.1791, simple_loss=0.2342, pruned_loss=0.06205, over 4790.00 frames. ], tot_loss[loss=0.1875, simple_loss=0.2529, pruned_loss=0.06099, over 953259.26 frames. ], batch size: 29, lr: 3.73e-03, grad_scale: 32.0 2023-04-27 00:38:18,949 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7991, 2.4367, 1.7689, 1.7088, 1.2947, 1.2947, 1.8213, 1.2060], device='cuda:6'), covar=tensor([0.1697, 0.1381, 0.1428, 0.1869, 0.2433, 0.1984, 0.1122, 0.2164], device='cuda:6'), in_proj_covar=tensor([0.0195, 0.0213, 0.0169, 0.0203, 0.0204, 0.0184, 0.0159, 0.0186], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:6') 2023-04-27 00:38:23,669 INFO [optim.py:369] (6/7) 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 00:38:31,956 INFO [finetune.py:976] (6/7) Epoch 10, batch 4800, loss[loss=0.2026, simple_loss=0.271, pruned_loss=0.06709, over 4925.00 frames. ], tot_loss[loss=0.1901, simple_loss=0.2555, pruned_loss=0.06232, over 952137.84 frames. ], batch size: 38, lr: 3.73e-03, grad_scale: 32.0 2023-04-27 00:39:33,343 INFO [finetune.py:976] (6/7) Epoch 10, batch 4850, loss[loss=0.2223, simple_loss=0.2943, pruned_loss=0.07516, over 4925.00 frames. ], tot_loss[loss=0.1926, simple_loss=0.2592, pruned_loss=0.06304, over 954040.08 frames. ], batch size: 33, lr: 3.73e-03, grad_scale: 32.0 2023-04-27 00:39:43,310 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=56404.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 00:40:01,636 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.4285, 2.3888, 1.9226, 2.2112, 2.5126, 2.0919, 3.1963, 1.8140], device='cuda:6'), covar=tensor([0.3909, 0.2202, 0.5196, 0.3223, 0.2131, 0.2653, 0.1920, 0.4881], device='cuda:6'), in_proj_covar=tensor([0.0345, 0.0348, 0.0431, 0.0362, 0.0387, 0.0384, 0.0382, 0.0419], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 00:40:08,219 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6392, 1.2713, 1.7024, 2.0775, 1.7561, 1.5920, 1.6677, 1.7138], device='cuda:6'), covar=tensor([0.5630, 0.7894, 0.7937, 0.7948, 0.7203, 0.9595, 0.9830, 0.9457], device='cuda:6'), in_proj_covar=tensor([0.0410, 0.0413, 0.0497, 0.0517, 0.0437, 0.0458, 0.0468, 0.0467], device='cuda:6'), out_proj_covar=tensor([9.9839e-05, 1.0235e-04, 1.1229e-04, 1.2292e-04, 1.0589e-04, 1.1064e-04, 1.1218e-04, 1.1223e-04], device='cuda:6') 2023-04-27 00:40:14,074 INFO [optim.py:369] (6/7) 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 00:40:22,378 INFO [finetune.py:976] (6/7) Epoch 10, batch 4900, loss[loss=0.1958, simple_loss=0.2658, pruned_loss=0.06294, over 4735.00 frames. ], tot_loss[loss=0.1934, simple_loss=0.2607, pruned_loss=0.06307, over 956041.30 frames. ], batch size: 59, lr: 3.73e-03, grad_scale: 32.0 2023-04-27 00:40:24,261 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=56452.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 00:40:30,700 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0111, 2.7623, 1.9718, 1.9853, 1.4454, 1.4209, 2.1319, 1.4077], device='cuda:6'), covar=tensor([0.1730, 0.1485, 0.1517, 0.1862, 0.2449, 0.2027, 0.1068, 0.2094], device='cuda:6'), in_proj_covar=tensor([0.0194, 0.0213, 0.0168, 0.0202, 0.0203, 0.0184, 0.0159, 0.0185], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:6') 2023-04-27 00:40:36,723 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6836, 1.3082, 1.8004, 2.1528, 1.8596, 1.6924, 1.7629, 1.7450], device='cuda:6'), covar=tensor([0.5688, 0.8530, 0.8112, 0.7593, 0.7123, 0.9559, 0.9456, 0.9753], device='cuda:6'), in_proj_covar=tensor([0.0410, 0.0413, 0.0497, 0.0517, 0.0437, 0.0457, 0.0468, 0.0466], device='cuda:6'), out_proj_covar=tensor([9.9802e-05, 1.0233e-04, 1.1226e-04, 1.2295e-04, 1.0588e-04, 1.1053e-04, 1.1212e-04, 1.1211e-04], device='cuda:6') 2023-04-27 00:40:56,302 INFO [finetune.py:976] (6/7) Epoch 10, batch 4950, loss[loss=0.1925, simple_loss=0.2503, pruned_loss=0.06737, over 4721.00 frames. ], tot_loss[loss=0.194, simple_loss=0.2614, pruned_loss=0.0633, over 953609.48 frames. ], batch size: 23, lr: 3.72e-03, grad_scale: 32.0 2023-04-27 00:40:57,604 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=56501.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 00:40:59,475 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=56504.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 00:41:12,936 INFO [zipformer.py:1188] (6/7) 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] (6/7) 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 00:41:29,729 INFO [finetune.py:976] (6/7) Epoch 10, batch 5000, loss[loss=0.1757, simple_loss=0.2418, pruned_loss=0.05487, over 4833.00 frames. ], tot_loss[loss=0.1901, simple_loss=0.2573, pruned_loss=0.06148, over 953423.75 frames. ], batch size: 47, lr: 3.72e-03, grad_scale: 32.0 2023-04-27 00:41:32,069 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=56552.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 00:41:53,722 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=56584.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 00:42:03,262 INFO [finetune.py:976] (6/7) Epoch 10, batch 5050, loss[loss=0.1927, simple_loss=0.2552, pruned_loss=0.06515, over 4817.00 frames. ], tot_loss[loss=0.1879, simple_loss=0.2543, pruned_loss=0.06075, over 954375.44 frames. ], batch size: 38, lr: 3.72e-03, grad_scale: 32.0 2023-04-27 00:42:56,298 INFO [optim.py:369] (6/7) 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] (6/7) Epoch 10, batch 5100, loss[loss=0.1941, simple_loss=0.2519, pruned_loss=0.06813, over 4295.00 frames. ], tot_loss[loss=0.1837, simple_loss=0.2501, pruned_loss=0.05869, over 954729.02 frames. ], batch size: 18, lr: 3.72e-03, grad_scale: 32.0 2023-04-27 00:43:59,162 INFO [finetune.py:976] (6/7) Epoch 10, batch 5150, loss[loss=0.1875, simple_loss=0.2484, pruned_loss=0.06335, over 4152.00 frames. ], tot_loss[loss=0.1858, simple_loss=0.2514, pruned_loss=0.0601, over 953095.39 frames. ], batch size: 18, lr: 3.72e-03, grad_scale: 32.0 2023-04-27 00:44:00,546 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6844, 1.1535, 1.6849, 2.1270, 1.7744, 1.6434, 1.6694, 1.7342], device='cuda:6'), covar=tensor([0.6029, 0.8410, 0.8046, 0.8399, 0.7849, 1.0285, 0.9737, 0.8599], device='cuda:6'), in_proj_covar=tensor([0.0409, 0.0414, 0.0497, 0.0517, 0.0438, 0.0458, 0.0469, 0.0466], device='cuda:6'), out_proj_covar=tensor([9.9684e-05, 1.0259e-04, 1.1226e-04, 1.2294e-04, 1.0601e-04, 1.1071e-04, 1.1233e-04, 1.1224e-04], device='cuda:6') 2023-04-27 00:44:25,424 INFO [optim.py:369] (6/7) 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 00:44:26,780 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9229, 1.5214, 1.9664, 2.0546, 1.6979, 1.4166, 1.6603, 1.1121], device='cuda:6'), covar=tensor([0.0481, 0.1069, 0.0642, 0.0618, 0.0729, 0.1271, 0.0794, 0.0801], device='cuda:6'), in_proj_covar=tensor([0.0067, 0.0073, 0.0071, 0.0068, 0.0076, 0.0096, 0.0077, 0.0073], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-27 00:44:33,209 INFO [finetune.py:976] (6/7) Epoch 10, batch 5200, loss[loss=0.1754, simple_loss=0.2427, pruned_loss=0.05405, over 4745.00 frames. ], tot_loss[loss=0.1885, simple_loss=0.2546, pruned_loss=0.06122, over 950278.71 frames. ], batch size: 27, lr: 3.72e-03, grad_scale: 32.0 2023-04-27 00:45:10,355 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.3317, 1.7322, 2.1901, 2.6961, 2.2108, 1.6792, 1.5122, 2.0993], device='cuda:6'), covar=tensor([0.3693, 0.3673, 0.1731, 0.2683, 0.2876, 0.3001, 0.4348, 0.2209], device='cuda:6'), in_proj_covar=tensor([0.0283, 0.0248, 0.0218, 0.0314, 0.0213, 0.0226, 0.0231, 0.0184], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:6') 2023-04-27 00:45:18,108 INFO [finetune.py:976] (6/7) Epoch 10, batch 5250, loss[loss=0.2119, simple_loss=0.2792, pruned_loss=0.07232, over 4773.00 frames. ], tot_loss[loss=0.1906, simple_loss=0.2574, pruned_loss=0.06191, over 950409.81 frames. ], batch size: 54, lr: 3.72e-03, grad_scale: 32.0 2023-04-27 00:45:20,017 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=56801.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 00:45:21,232 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=56803.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 00:45:28,626 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.77 vs. limit=2.0 2023-04-27 00:45:33,856 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.2272, 1.1245, 1.2427, 1.5098, 1.6097, 1.2768, 0.9472, 1.4088], device='cuda:6'), covar=tensor([0.0940, 0.1735, 0.1089, 0.0663, 0.0657, 0.0910, 0.0991, 0.0709], device='cuda:6'), in_proj_covar=tensor([0.0194, 0.0202, 0.0181, 0.0174, 0.0177, 0.0187, 0.0159, 0.0183], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 00:45:42,565 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7170, 2.2325, 1.7400, 1.5752, 1.3376, 1.3384, 1.8010, 1.2526], device='cuda:6'), covar=tensor([0.1597, 0.1356, 0.1515, 0.1843, 0.2447, 0.1987, 0.1029, 0.2088], device='cuda:6'), in_proj_covar=tensor([0.0195, 0.0214, 0.0169, 0.0203, 0.0204, 0.0185, 0.0159, 0.0186], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:6') 2023-04-27 00:45:43,653 INFO [optim.py:369] (6/7) 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 00:45:50,981 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-04-27 00:45:51,393 INFO [finetune.py:976] (6/7) Epoch 10, batch 5300, loss[loss=0.2067, simple_loss=0.2825, pruned_loss=0.06546, over 4865.00 frames. ], tot_loss[loss=0.1932, simple_loss=0.26, pruned_loss=0.06322, over 951068.53 frames. ], batch size: 31, lr: 3.72e-03, grad_scale: 32.0 2023-04-27 00:45:51,456 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=56849.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 00:46:01,129 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=56864.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 00:46:11,636 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=56879.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 00:46:25,220 INFO [finetune.py:976] (6/7) Epoch 10, batch 5350, loss[loss=0.1803, simple_loss=0.2594, pruned_loss=0.05057, over 4778.00 frames. ], tot_loss[loss=0.193, simple_loss=0.26, pruned_loss=0.06301, over 953155.20 frames. ], batch size: 29, lr: 3.72e-03, grad_scale: 32.0 2023-04-27 00:46:31,345 INFO [zipformer.py:1188] (6/7) 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] (6/7) 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 00:46:58,346 INFO [finetune.py:976] (6/7) Epoch 10, batch 5400, loss[loss=0.1356, simple_loss=0.2068, pruned_loss=0.0322, over 4760.00 frames. ], tot_loss[loss=0.1909, simple_loss=0.2573, pruned_loss=0.06226, over 953479.72 frames. ], batch size: 27, lr: 3.72e-03, grad_scale: 32.0 2023-04-27 00:47:11,572 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=56969.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 00:47:28,578 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=56993.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 00:47:29,772 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=56995.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 00:47:32,094 INFO [finetune.py:976] (6/7) Epoch 10, batch 5450, loss[loss=0.2065, simple_loss=0.2674, pruned_loss=0.07282, over 4823.00 frames. ], tot_loss[loss=0.1887, simple_loss=0.2542, pruned_loss=0.06162, over 954329.44 frames. ], batch size: 38, lr: 3.72e-03, grad_scale: 32.0 2023-04-27 00:48:18,181 INFO [optim.py:369] (6/7) 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 00:48:36,193 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5204, 0.6337, 1.3833, 1.9150, 1.6131, 1.4258, 1.3779, 1.5220], device='cuda:6'), covar=tensor([0.5662, 0.7713, 0.7300, 0.7890, 0.6829, 0.9083, 0.9353, 0.8575], device='cuda:6'), in_proj_covar=tensor([0.0408, 0.0412, 0.0496, 0.0516, 0.0436, 0.0457, 0.0467, 0.0466], device='cuda:6'), out_proj_covar=tensor([9.9258e-05, 1.0203e-04, 1.1196e-04, 1.2270e-04, 1.0559e-04, 1.1032e-04, 1.1190e-04, 1.1202e-04], device='cuda:6') 2023-04-27 00:48:38,379 INFO [finetune.py:976] (6/7) Epoch 10, batch 5500, loss[loss=0.2019, simple_loss=0.2577, pruned_loss=0.07308, over 4764.00 frames. ], tot_loss[loss=0.1867, simple_loss=0.2514, pruned_loss=0.06095, over 953055.20 frames. ], batch size: 27, lr: 3.72e-03, grad_scale: 32.0 2023-04-27 00:48:47,146 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=57054.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 00:48:48,363 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=57056.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 00:49:00,509 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([0.8222, 1.9072, 1.7196, 1.4972, 2.0573, 1.6057, 2.5291, 1.5500], device='cuda:6'), covar=tensor([0.3822, 0.1824, 0.4870, 0.2994, 0.1503, 0.2417, 0.1439, 0.4562], device='cuda:6'), in_proj_covar=tensor([0.0339, 0.0343, 0.0426, 0.0358, 0.0381, 0.0380, 0.0377, 0.0413], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 00:49:12,695 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([4.3614, 4.1647, 3.0727, 4.9154, 4.3110, 4.2328, 1.6052, 4.1829], device='cuda:6'), covar=tensor([0.1495, 0.1031, 0.3410, 0.0986, 0.3916, 0.1640, 0.5976, 0.2140], device='cuda:6'), in_proj_covar=tensor([0.0241, 0.0214, 0.0248, 0.0303, 0.0299, 0.0248, 0.0267, 0.0270], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 00:49:35,990 INFO [finetune.py:976] (6/7) Epoch 10, batch 5550, loss[loss=0.1824, simple_loss=0.2437, pruned_loss=0.06059, over 4787.00 frames. ], tot_loss[loss=0.1886, simple_loss=0.2538, pruned_loss=0.06172, over 953517.23 frames. ], batch size: 26, lr: 3.72e-03, grad_scale: 32.0 2023-04-27 00:49:42,868 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8385, 1.4214, 1.8600, 2.2731, 1.9403, 1.8071, 1.8505, 1.8814], device='cuda:6'), covar=tensor([0.6106, 0.7935, 0.8803, 0.8304, 0.7753, 1.0207, 1.0487, 0.8574], device='cuda:6'), in_proj_covar=tensor([0.0408, 0.0413, 0.0497, 0.0517, 0.0438, 0.0457, 0.0468, 0.0466], device='cuda:6'), out_proj_covar=tensor([9.9377e-05, 1.0227e-04, 1.1212e-04, 1.2286e-04, 1.0591e-04, 1.1055e-04, 1.1214e-04, 1.1217e-04], device='cuda:6') 2023-04-27 00:49:59,958 INFO [optim.py:369] (6/7) 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] (6/7) Epoch 10, batch 5600, loss[loss=0.227, simple_loss=0.2732, pruned_loss=0.0904, over 4822.00 frames. ], tot_loss[loss=0.1912, simple_loss=0.2575, pruned_loss=0.06245, over 952951.78 frames. ], batch size: 30, lr: 3.72e-03, grad_scale: 32.0 2023-04-27 00:50:11,132 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-04-27 00:50:12,764 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=57159.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 00:50:23,771 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6277, 1.2110, 1.2957, 1.3812, 1.8893, 1.5468, 1.2706, 1.2717], device='cuda:6'), covar=tensor([0.1707, 0.1615, 0.1709, 0.1504, 0.0627, 0.1484, 0.1903, 0.2081], device='cuda:6'), in_proj_covar=tensor([0.0307, 0.0321, 0.0356, 0.0299, 0.0335, 0.0320, 0.0307, 0.0363], device='cuda:6'), out_proj_covar=tensor([6.4589e-05, 6.7941e-05, 7.6781e-05, 6.1708e-05, 6.9981e-05, 6.8409e-05, 6.5728e-05, 7.7947e-05], device='cuda:6') 2023-04-27 00:50:24,891 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=57179.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 00:50:29,401 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=57186.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 00:50:32,935 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-04-27 00:50:35,749 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 10, batch 5650, loss[loss=0.2193, simple_loss=0.289, pruned_loss=0.07477, over 4922.00 frames. ], tot_loss[loss=0.1924, simple_loss=0.2594, pruned_loss=0.06264, over 952385.24 frames. ], batch size: 38, lr: 3.72e-03, grad_scale: 32.0 2023-04-27 00:50:53,685 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=57227.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 00:50:59,541 INFO [optim.py:369] (6/7) 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 00:51:03,368 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.70 vs. limit=5.0 2023-04-27 00:51:03,826 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=57244.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 00:51:05,594 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=57247.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 00:51:06,701 INFO [finetune.py:976] (6/7) Epoch 10, batch 5700, loss[loss=0.1851, simple_loss=0.2447, pruned_loss=0.06273, over 4125.00 frames. ], tot_loss[loss=0.1894, simple_loss=0.2547, pruned_loss=0.06203, over 931168.64 frames. ], batch size: 17, lr: 3.72e-03, grad_scale: 32.0 2023-04-27 00:51:12,268 INFO [zipformer.py:1188] (6/7) 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] (6/7) 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] (6/7) Epoch 11, batch 0, loss[loss=0.1856, simple_loss=0.261, pruned_loss=0.05512, over 4899.00 frames. ], tot_loss[loss=0.1856, simple_loss=0.261, pruned_loss=0.05512, over 4899.00 frames. ], batch size: 46, lr: 3.72e-03, grad_scale: 32.0 2023-04-27 00:51:38,655 INFO [finetune.py:1001] (6/7) Computing validation loss 2023-04-27 00:51:55,309 INFO [finetune.py:1010] (6/7) Epoch 11, validation: loss=0.1558, simple_loss=0.2272, pruned_loss=0.04225, over 2265189.00 frames. 2023-04-27 00:51:55,309 INFO [finetune.py:1011] (6/7) Maximum memory allocated so far is 6345MB 2023-04-27 00:52:28,814 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=57305.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 00:52:42,959 INFO [finetune.py:976] (6/7) Epoch 11, batch 50, loss[loss=0.1724, simple_loss=0.2424, pruned_loss=0.05121, over 4815.00 frames. ], tot_loss[loss=0.1882, simple_loss=0.2559, pruned_loss=0.06028, over 215431.13 frames. ], batch size: 38, lr: 3.72e-03, grad_scale: 32.0 2023-04-27 00:52:49,892 INFO [optim.py:369] (6/7) 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 00:52:57,814 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=57349.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 00:52:59,005 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 11, batch 100, loss[loss=0.1873, simple_loss=0.2471, pruned_loss=0.06373, over 4721.00 frames. ], tot_loss[loss=0.1916, simple_loss=0.2567, pruned_loss=0.06326, over 382298.72 frames. ], batch size: 59, lr: 3.72e-03, grad_scale: 32.0 2023-04-27 00:53:38,536 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.3704, 1.7428, 1.6427, 1.7080, 1.3405, 1.5456, 1.5805, 1.2609], device='cuda:6'), covar=tensor([0.2388, 0.2112, 0.1253, 0.1805, 0.4057, 0.1827, 0.2044, 0.2822], device='cuda:6'), in_proj_covar=tensor([0.0297, 0.0317, 0.0229, 0.0287, 0.0316, 0.0269, 0.0255, 0.0278], device='cuda:6'), out_proj_covar=tensor([1.1976e-04, 1.2740e-04, 9.1591e-05, 1.1502e-04, 1.2924e-04, 1.0825e-04, 1.0390e-04, 1.1135e-04], device='cuda:6') 2023-04-27 00:53:40,240 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=57398.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 00:54:10,096 INFO [finetune.py:976] (6/7) Epoch 11, batch 150, loss[loss=0.223, simple_loss=0.2929, pruned_loss=0.07655, over 4830.00 frames. ], tot_loss[loss=0.1884, simple_loss=0.2532, pruned_loss=0.06178, over 509762.34 frames. ], batch size: 39, lr: 3.72e-03, grad_scale: 32.0 2023-04-27 00:54:27,173 INFO [optim.py:369] (6/7) 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 00:54:52,718 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=57459.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 00:54:52,739 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=57459.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 00:55:00,043 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=57462.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 00:55:10,526 INFO [finetune.py:976] (6/7) Epoch 11, batch 200, loss[loss=0.2046, simple_loss=0.2768, pruned_loss=0.06623, over 4873.00 frames. ], tot_loss[loss=0.1868, simple_loss=0.2517, pruned_loss=0.06097, over 609159.25 frames. ], batch size: 34, lr: 3.72e-03, grad_scale: 32.0 2023-04-27 00:55:22,560 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=57495.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 00:55:30,376 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=57507.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 00:55:34,065 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0207, 1.8870, 2.1417, 2.3473, 2.5445, 1.8962, 1.7022, 2.1506], device='cuda:6'), covar=tensor([0.0986, 0.1005, 0.0643, 0.0664, 0.0611, 0.0956, 0.0931, 0.0599], device='cuda:6'), in_proj_covar=tensor([0.0195, 0.0201, 0.0181, 0.0173, 0.0177, 0.0187, 0.0158, 0.0182], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 00:55:40,657 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=57523.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 00:55:42,942 INFO [finetune.py:976] (6/7) Epoch 11, batch 250, loss[loss=0.2237, simple_loss=0.2904, pruned_loss=0.07846, over 4831.00 frames. ], tot_loss[loss=0.1913, simple_loss=0.2572, pruned_loss=0.0627, over 687391.02 frames. ], batch size: 47, lr: 3.71e-03, grad_scale: 32.0 2023-04-27 00:55:50,073 INFO [optim.py:369] (6/7) 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 00:55:54,622 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=57542.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 00:56:01,293 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=57553.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 00:56:03,143 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=57556.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 00:56:08,523 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=57564.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 00:56:09,850 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-04-27 00:56:16,735 INFO [finetune.py:976] (6/7) Epoch 11, batch 300, loss[loss=0.1855, simple_loss=0.2443, pruned_loss=0.06335, over 4286.00 frames. ], tot_loss[loss=0.1919, simple_loss=0.2586, pruned_loss=0.0626, over 747077.25 frames. ], batch size: 19, lr: 3.71e-03, grad_scale: 32.0 2023-04-27 00:56:32,859 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=57600.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 00:56:40,097 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=57612.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 00:56:50,092 INFO [finetune.py:976] (6/7) Epoch 11, batch 350, loss[loss=0.1638, simple_loss=0.2439, pruned_loss=0.04182, over 4729.00 frames. ], tot_loss[loss=0.1923, simple_loss=0.2594, pruned_loss=0.06261, over 789950.09 frames. ], batch size: 54, lr: 3.71e-03, grad_scale: 32.0 2023-04-27 00:56:56,642 INFO [optim.py:369] (6/7) 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 00:57:11,901 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=57649.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 00:57:13,084 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 11, batch 400, loss[loss=0.2032, simple_loss=0.2743, pruned_loss=0.06606, over 4913.00 frames. ], tot_loss[loss=0.1937, simple_loss=0.2615, pruned_loss=0.06301, over 828248.64 frames. ], batch size: 38, lr: 3.71e-03, grad_scale: 64.0 2023-04-27 00:58:10,656 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=57697.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 00:58:12,861 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=57699.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 00:58:30,753 INFO [finetune.py:976] (6/7) Epoch 11, batch 450, loss[loss=0.1594, simple_loss=0.23, pruned_loss=0.04439, over 4867.00 frames. ], tot_loss[loss=0.1918, simple_loss=0.2593, pruned_loss=0.06211, over 855654.67 frames. ], batch size: 34, lr: 3.71e-03, grad_scale: 64.0 2023-04-27 00:58:35,002 INFO [zipformer.py:1188] (6/7) 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] (6/7) 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 00:59:00,417 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=57754.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 00:59:10,677 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.47 vs. limit=2.0 2023-04-27 00:59:10,877 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7062, 2.4078, 1.9468, 2.2281, 1.7606, 1.9792, 1.9743, 1.4343], device='cuda:6'), covar=tensor([0.2518, 0.1314, 0.0911, 0.1371, 0.3366, 0.1325, 0.2331, 0.2895], device='cuda:6'), in_proj_covar=tensor([0.0296, 0.0317, 0.0230, 0.0287, 0.0316, 0.0270, 0.0255, 0.0278], device='cuda:6'), out_proj_covar=tensor([1.1950e-04, 1.2772e-04, 9.2239e-05, 1.1485e-04, 1.2932e-04, 1.0861e-04, 1.0407e-04, 1.1161e-04], device='cuda:6') 2023-04-27 00:59:32,061 INFO [finetune.py:976] (6/7) Epoch 11, batch 500, loss[loss=0.169, simple_loss=0.2322, pruned_loss=0.05286, over 4760.00 frames. ], tot_loss[loss=0.1901, simple_loss=0.2568, pruned_loss=0.06167, over 878389.05 frames. ], batch size: 26, lr: 3.71e-03, grad_scale: 64.0 2023-04-27 00:59:43,846 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6774, 3.5806, 1.0496, 2.0075, 2.1208, 2.5325, 2.0647, 1.1750], device='cuda:6'), covar=tensor([0.1243, 0.0969, 0.1819, 0.1168, 0.0938, 0.0953, 0.1349, 0.1714], device='cuda:6'), in_proj_covar=tensor([0.0118, 0.0249, 0.0141, 0.0122, 0.0134, 0.0154, 0.0119, 0.0123], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-27 00:59:54,249 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=57794.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:00:25,886 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 11, batch 550, loss[loss=0.2147, simple_loss=0.2704, pruned_loss=0.07952, over 4851.00 frames. ], tot_loss[loss=0.1899, simple_loss=0.2552, pruned_loss=0.06234, over 896996.24 frames. ], batch size: 49, lr: 3.71e-03, grad_scale: 64.0 2023-04-27 01:00:38,243 INFO [optim.py:369] (6/7) 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 01:00:41,345 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=57842.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:00:44,268 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5209, 1.3070, 1.2051, 1.3389, 1.7432, 1.5106, 1.2677, 1.1583], device='cuda:6'), covar=tensor([0.1138, 0.1097, 0.1465, 0.1302, 0.0619, 0.1242, 0.1480, 0.1592], device='cuda:6'), in_proj_covar=tensor([0.0306, 0.0321, 0.0353, 0.0299, 0.0335, 0.0318, 0.0306, 0.0361], device='cuda:6'), out_proj_covar=tensor([6.4388e-05, 6.7825e-05, 7.5980e-05, 6.1830e-05, 7.0086e-05, 6.7849e-05, 6.5364e-05, 7.7424e-05], device='cuda:6') 2023-04-27 01:00:47,295 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=57851.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:00:49,056 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=57853.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 01:01:04,585 INFO [finetune.py:976] (6/7) Epoch 11, batch 600, loss[loss=0.2296, simple_loss=0.3029, pruned_loss=0.07815, over 4907.00 frames. ], tot_loss[loss=0.1904, simple_loss=0.2554, pruned_loss=0.0627, over 909913.05 frames. ], batch size: 37, lr: 3.71e-03, grad_scale: 64.0 2023-04-27 01:01:13,413 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=57890.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:01:20,164 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=57900.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:01:20,741 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=57901.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:01:38,049 INFO [finetune.py:976] (6/7) Epoch 11, batch 650, loss[loss=0.2389, simple_loss=0.313, pruned_loss=0.08239, over 4812.00 frames. ], tot_loss[loss=0.1918, simple_loss=0.258, pruned_loss=0.06282, over 919696.13 frames. ], batch size: 39, lr: 3.71e-03, grad_scale: 64.0 2023-04-27 01:01:45,068 INFO [optim.py:369] (6/7) 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 01:01:51,798 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=57948.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:01:53,674 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7660, 2.3605, 1.7122, 1.5520, 1.2792, 1.2867, 1.7023, 1.2360], device='cuda:6'), covar=tensor([0.1598, 0.1345, 0.1493, 0.1940, 0.2484, 0.2019, 0.1069, 0.2090], device='cuda:6'), in_proj_covar=tensor([0.0195, 0.0215, 0.0170, 0.0204, 0.0204, 0.0185, 0.0159, 0.0187], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:6') 2023-04-27 01:02:11,761 INFO [finetune.py:976] (6/7) Epoch 11, batch 700, loss[loss=0.1779, simple_loss=0.2564, pruned_loss=0.04966, over 4856.00 frames. ], tot_loss[loss=0.1945, simple_loss=0.2611, pruned_loss=0.06395, over 929815.89 frames. ], batch size: 31, lr: 3.71e-03, grad_scale: 64.0 2023-04-27 01:02:37,008 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.3956, 1.2931, 1.6658, 1.5794, 1.3113, 1.0845, 1.4057, 0.9737], device='cuda:6'), covar=tensor([0.0670, 0.0680, 0.0483, 0.0626, 0.0751, 0.0990, 0.0626, 0.0698], device='cuda:6'), in_proj_covar=tensor([0.0066, 0.0071, 0.0070, 0.0067, 0.0075, 0.0095, 0.0076, 0.0072], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-27 01:02:45,351 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8542, 2.5501, 1.9418, 1.8009, 1.3605, 1.3449, 1.8701, 1.3210], device='cuda:6'), covar=tensor([0.1669, 0.1431, 0.1444, 0.1821, 0.2425, 0.2050, 0.1061, 0.2089], device='cuda:6'), in_proj_covar=tensor([0.0193, 0.0213, 0.0168, 0.0202, 0.0203, 0.0184, 0.0158, 0.0185], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:6') 2023-04-27 01:02:53,357 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([3.8646, 3.7704, 2.6635, 4.4641, 3.8145, 3.8332, 1.7846, 3.7628], device='cuda:6'), covar=tensor([0.1667, 0.1291, 0.3025, 0.1660, 0.3852, 0.1778, 0.5564, 0.2405], device='cuda:6'), in_proj_covar=tensor([0.0240, 0.0214, 0.0248, 0.0300, 0.0295, 0.0248, 0.0267, 0.0268], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 01:02:56,395 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0941, 2.4002, 0.9826, 1.2416, 1.8053, 1.1404, 3.0439, 1.7018], device='cuda:6'), covar=tensor([0.0664, 0.0561, 0.0786, 0.1415, 0.0521, 0.1133, 0.0345, 0.0647], device='cuda:6'), in_proj_covar=tensor([0.0052, 0.0066, 0.0049, 0.0047, 0.0050, 0.0052, 0.0078, 0.0051], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0007, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:6') 2023-04-27 01:02:57,542 INFO [finetune.py:976] (6/7) Epoch 11, batch 750, loss[loss=0.2094, simple_loss=0.2942, pruned_loss=0.06232, over 4696.00 frames. ], tot_loss[loss=0.1978, simple_loss=0.2643, pruned_loss=0.06568, over 935773.65 frames. ], batch size: 54, lr: 3.71e-03, grad_scale: 32.0 2023-04-27 01:03:04,228 INFO [optim.py:369] (6/7) 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 01:03:14,991 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=58054.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:03:31,279 INFO [finetune.py:976] (6/7) Epoch 11, batch 800, loss[loss=0.2175, simple_loss=0.2757, pruned_loss=0.07964, over 4823.00 frames. ], tot_loss[loss=0.1962, simple_loss=0.2632, pruned_loss=0.06458, over 940578.76 frames. ], batch size: 33, lr: 3.71e-03, grad_scale: 32.0 2023-04-27 01:03:31,997 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.0518, 0.9585, 1.2636, 1.1699, 0.9610, 0.8618, 0.9701, 0.5072], device='cuda:6'), covar=tensor([0.0622, 0.0797, 0.0588, 0.0583, 0.0827, 0.1495, 0.0576, 0.0996], device='cuda:6'), in_proj_covar=tensor([0.0066, 0.0072, 0.0070, 0.0067, 0.0075, 0.0095, 0.0076, 0.0072], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-27 01:03:38,604 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=58089.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:03:47,437 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=58102.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:03:58,173 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=58118.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:04:04,031 INFO [finetune.py:976] (6/7) Epoch 11, batch 850, loss[loss=0.2265, simple_loss=0.2724, pruned_loss=0.09028, over 4800.00 frames. ], tot_loss[loss=0.1934, simple_loss=0.2599, pruned_loss=0.06349, over 943880.20 frames. ], batch size: 45, lr: 3.71e-03, grad_scale: 32.0 2023-04-27 01:04:10,688 INFO [optim.py:369] (6/7) 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 01:04:12,618 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=58141.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:04:14,898 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=58144.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:04:19,564 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=58151.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:04:29,489 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0576, 1.4390, 1.9099, 2.3782, 2.0264, 1.5564, 1.1879, 1.6493], device='cuda:6'), covar=tensor([0.4014, 0.4321, 0.2021, 0.2905, 0.3052, 0.3075, 0.5150, 0.2912], device='cuda:6'), in_proj_covar=tensor([0.0286, 0.0249, 0.0222, 0.0318, 0.0214, 0.0228, 0.0233, 0.0185], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:6') 2023-04-27 01:04:39,246 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=58166.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:04:52,831 INFO [finetune.py:976] (6/7) Epoch 11, batch 900, loss[loss=0.1587, simple_loss=0.2249, pruned_loss=0.04629, over 4899.00 frames. ], tot_loss[loss=0.1909, simple_loss=0.2564, pruned_loss=0.06266, over 947853.22 frames. ], batch size: 35, lr: 3.71e-03, grad_scale: 32.0 2023-04-27 01:05:20,283 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=58199.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:05:22,296 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=58202.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:05:28,581 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=58205.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:05:31,993 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.81 vs. limit=2.0 2023-04-27 01:05:54,766 INFO [finetune.py:976] (6/7) Epoch 11, batch 950, loss[loss=0.1878, simple_loss=0.244, pruned_loss=0.06586, over 4826.00 frames. ], tot_loss[loss=0.1882, simple_loss=0.2536, pruned_loss=0.06145, over 948658.01 frames. ], batch size: 25, lr: 3.71e-03, grad_scale: 32.0 2023-04-27 01:06:10,248 INFO [optim.py:369] (6/7) 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 01:06:13,048 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.14 vs. limit=5.0 2023-04-27 01:06:43,323 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0352, 1.5138, 1.9829, 2.2380, 1.9626, 1.5009, 1.2350, 1.7298], device='cuda:6'), covar=tensor([0.3326, 0.3426, 0.1705, 0.2563, 0.2725, 0.2808, 0.4443, 0.2209], device='cuda:6'), in_proj_covar=tensor([0.0284, 0.0249, 0.0221, 0.0317, 0.0213, 0.0227, 0.0232, 0.0185], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:6') 2023-04-27 01:06:46,851 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.5158, 3.0486, 0.7753, 1.5668, 2.1679, 1.5693, 4.3447, 2.1862], device='cuda:6'), covar=tensor([0.0615, 0.0737, 0.0925, 0.1285, 0.0563, 0.1066, 0.0240, 0.0609], device='cuda:6'), in_proj_covar=tensor([0.0052, 0.0066, 0.0049, 0.0047, 0.0051, 0.0053, 0.0078, 0.0052], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:6') 2023-04-27 01:06:47,387 INFO [finetune.py:976] (6/7) Epoch 11, batch 1000, loss[loss=0.1889, simple_loss=0.2421, pruned_loss=0.06785, over 4822.00 frames. ], tot_loss[loss=0.1894, simple_loss=0.2552, pruned_loss=0.06182, over 949781.98 frames. ], batch size: 30, lr: 3.71e-03, grad_scale: 32.0 2023-04-27 01:07:06,521 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8076, 1.1106, 1.4112, 1.5388, 1.5094, 1.5900, 1.4752, 1.4416], device='cuda:6'), covar=tensor([0.4823, 0.5977, 0.5365, 0.5228, 0.6053, 0.8411, 0.5753, 0.5550], device='cuda:6'), in_proj_covar=tensor([0.0329, 0.0380, 0.0314, 0.0327, 0.0339, 0.0402, 0.0360, 0.0324], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 01:07:20,925 INFO [finetune.py:976] (6/7) Epoch 11, batch 1050, loss[loss=0.2018, simple_loss=0.2771, pruned_loss=0.06323, over 4816.00 frames. ], tot_loss[loss=0.1906, simple_loss=0.2576, pruned_loss=0.06175, over 952477.88 frames. ], batch size: 38, lr: 3.71e-03, grad_scale: 32.0 2023-04-27 01:07:28,208 INFO [optim.py:369] (6/7) 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 01:07:52,946 INFO [finetune.py:976] (6/7) Epoch 11, batch 1100, loss[loss=0.208, simple_loss=0.2871, pruned_loss=0.06445, over 4758.00 frames. ], tot_loss[loss=0.1922, simple_loss=0.2595, pruned_loss=0.06245, over 953562.27 frames. ], batch size: 59, lr: 3.71e-03, grad_scale: 32.0 2023-04-27 01:08:01,691 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 11, batch 1150, loss[loss=0.2202, simple_loss=0.2877, pruned_loss=0.07637, over 4759.00 frames. ], tot_loss[loss=0.1927, simple_loss=0.2604, pruned_loss=0.06254, over 954024.17 frames. ], batch size: 27, lr: 3.71e-03, grad_scale: 32.0 2023-04-27 01:08:34,546 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=58437.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:08:35,082 INFO [optim.py:369] (6/7) 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 01:08:47,552 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.00 vs. limit=5.0 2023-04-27 01:08:59,898 INFO [finetune.py:976] (6/7) Epoch 11, batch 1200, loss[loss=0.1728, simple_loss=0.2313, pruned_loss=0.05715, over 4818.00 frames. ], tot_loss[loss=0.1914, simple_loss=0.2584, pruned_loss=0.06219, over 952898.06 frames. ], batch size: 30, lr: 3.71e-03, grad_scale: 32.0 2023-04-27 01:09:13,289 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.2786, 1.4218, 3.8121, 3.5615, 3.3930, 3.6029, 3.6117, 3.3916], device='cuda:6'), covar=tensor([0.6808, 0.5225, 0.1296, 0.1859, 0.1096, 0.1731, 0.1898, 0.1610], device='cuda:6'), in_proj_covar=tensor([0.0308, 0.0305, 0.0403, 0.0407, 0.0347, 0.0405, 0.0314, 0.0366], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-27 01:09:13,886 INFO [zipformer.py:1188] (6/7) 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] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=58500.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:09:32,954 INFO [finetune.py:976] (6/7) Epoch 11, batch 1250, loss[loss=0.1502, simple_loss=0.2258, pruned_loss=0.03725, over 4755.00 frames. ], tot_loss[loss=0.1908, simple_loss=0.2571, pruned_loss=0.0623, over 954166.74 frames. ], batch size: 26, lr: 3.71e-03, grad_scale: 32.0 2023-04-27 01:09:41,124 INFO [optim.py:369] (6/7) 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 01:10:22,920 INFO [finetune.py:976] (6/7) Epoch 11, batch 1300, loss[loss=0.1674, simple_loss=0.2409, pruned_loss=0.04693, over 4832.00 frames. ], tot_loss[loss=0.1884, simple_loss=0.2541, pruned_loss=0.06134, over 952498.63 frames. ], batch size: 30, lr: 3.71e-03, grad_scale: 32.0 2023-04-27 01:11:06,185 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-04-27 01:11:29,575 INFO [finetune.py:976] (6/7) Epoch 11, batch 1350, loss[loss=0.149, simple_loss=0.2094, pruned_loss=0.04429, over 4816.00 frames. ], tot_loss[loss=0.1893, simple_loss=0.2547, pruned_loss=0.06188, over 953669.89 frames. ], batch size: 25, lr: 3.71e-03, grad_scale: 32.0 2023-04-27 01:11:46,676 INFO [optim.py:369] (6/7) 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 01:11:48,846 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.25 vs. limit=5.0 2023-04-27 01:12:00,609 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6825, 2.1563, 1.7387, 1.9855, 1.6133, 1.7651, 1.7673, 1.4099], device='cuda:6'), covar=tensor([0.1914, 0.1164, 0.0954, 0.1305, 0.3330, 0.1373, 0.1930, 0.2635], device='cuda:6'), in_proj_covar=tensor([0.0294, 0.0315, 0.0228, 0.0285, 0.0314, 0.0270, 0.0254, 0.0276], device='cuda:6'), out_proj_covar=tensor([1.1861e-04, 1.2672e-04, 9.1438e-05, 1.1397e-04, 1.2851e-04, 1.0854e-04, 1.0345e-04, 1.1070e-04], device='cuda:6') 2023-04-27 01:12:22,086 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9162, 2.4326, 1.0374, 1.2177, 1.9192, 1.1459, 2.9566, 1.4654], device='cuda:6'), covar=tensor([0.0698, 0.0655, 0.0758, 0.1200, 0.0459, 0.0994, 0.0262, 0.0660], device='cuda:6'), in_proj_covar=tensor([0.0051, 0.0066, 0.0049, 0.0047, 0.0051, 0.0052, 0.0078, 0.0051], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:6') 2023-04-27 01:12:34,431 INFO [finetune.py:976] (6/7) Epoch 11, batch 1400, loss[loss=0.1826, simple_loss=0.2555, pruned_loss=0.05486, over 4941.00 frames. ], tot_loss[loss=0.1905, simple_loss=0.2572, pruned_loss=0.06194, over 952772.93 frames. ], batch size: 33, lr: 3.71e-03, grad_scale: 32.0 2023-04-27 01:12:42,442 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=58681.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:13:42,640 INFO [finetune.py:976] (6/7) Epoch 11, batch 1450, loss[loss=0.1875, simple_loss=0.2657, pruned_loss=0.05462, over 4813.00 frames. ], tot_loss[loss=0.191, simple_loss=0.2585, pruned_loss=0.06174, over 953066.42 frames. ], batch size: 39, lr: 3.71e-03, grad_scale: 32.0 2023-04-27 01:14:01,640 INFO [optim.py:369] (6/7) 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 01:14:03,411 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5841, 3.5346, 1.0097, 1.9684, 2.1717, 2.4262, 2.0860, 1.1675], device='cuda:6'), covar=tensor([0.1249, 0.0926, 0.1867, 0.1137, 0.0867, 0.0978, 0.1312, 0.1735], device='cuda:6'), in_proj_covar=tensor([0.0116, 0.0247, 0.0139, 0.0121, 0.0132, 0.0152, 0.0118, 0.0122], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-27 01:14:05,176 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=58742.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:14:22,214 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=3.32 vs. limit=5.0 2023-04-27 01:14:37,135 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 11, batch 1500, loss[loss=0.2159, simple_loss=0.2792, pruned_loss=0.07633, over 4859.00 frames. ], tot_loss[loss=0.1917, simple_loss=0.2593, pruned_loss=0.06203, over 954443.28 frames. ], batch size: 34, lr: 3.70e-03, grad_scale: 32.0 2023-04-27 01:14:44,484 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.34 vs. limit=5.0 2023-04-27 01:14:52,501 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=58797.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:14:54,830 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=58800.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:15:11,712 INFO [finetune.py:976] (6/7) Epoch 11, batch 1550, loss[loss=0.1473, simple_loss=0.2158, pruned_loss=0.03941, over 4829.00 frames. ], tot_loss[loss=0.1894, simple_loss=0.2576, pruned_loss=0.06063, over 955371.98 frames. ], batch size: 25, lr: 3.70e-03, grad_scale: 32.0 2023-04-27 01:15:18,197 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=58836.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:15:19,335 INFO [optim.py:369] (6/7) 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 01:15:24,655 INFO [zipformer.py:1188] (6/7) 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] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=58848.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:15:34,811 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.3730, 1.5569, 1.3152, 1.5529, 1.3641, 1.2715, 1.4085, 1.1312], device='cuda:6'), covar=tensor([0.1538, 0.1232, 0.0969, 0.1067, 0.3060, 0.1169, 0.1409, 0.1960], device='cuda:6'), in_proj_covar=tensor([0.0296, 0.0317, 0.0230, 0.0286, 0.0315, 0.0271, 0.0255, 0.0277], device='cuda:6'), out_proj_covar=tensor([1.1934e-04, 1.2762e-04, 9.2158e-05, 1.1442e-04, 1.2892e-04, 1.0885e-04, 1.0392e-04, 1.1092e-04], device='cuda:6') 2023-04-27 01:15:44,971 INFO [finetune.py:976] (6/7) Epoch 11, batch 1600, loss[loss=0.1984, simple_loss=0.2567, pruned_loss=0.07007, over 4827.00 frames. ], tot_loss[loss=0.1874, simple_loss=0.2552, pruned_loss=0.05979, over 957203.11 frames. ], batch size: 33, lr: 3.70e-03, grad_scale: 32.0 2023-04-27 01:16:00,300 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.4871, 1.0020, 0.4197, 1.1745, 1.0690, 1.3940, 1.2444, 1.2386], device='cuda:6'), covar=tensor([0.0546, 0.0437, 0.0428, 0.0593, 0.0310, 0.0554, 0.0530, 0.0589], device='cuda:6'), in_proj_covar=tensor([0.0030, 0.0025, 0.0023, 0.0029, 0.0020, 0.0029, 0.0029, 0.0029], device='cuda:6'), out_proj_covar=tensor([0.0049, 0.0044, 0.0037, 0.0050, 0.0037, 0.0048, 0.0048, 0.0049], device='cuda:6') 2023-04-27 01:16:05,034 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=58906.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:16:18,091 INFO [finetune.py:976] (6/7) Epoch 11, batch 1650, loss[loss=0.1709, simple_loss=0.237, pruned_loss=0.05237, over 4819.00 frames. ], tot_loss[loss=0.1856, simple_loss=0.2528, pruned_loss=0.05921, over 957149.27 frames. ], batch size: 39, lr: 3.70e-03, grad_scale: 32.0 2023-04-27 01:16:24,640 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.4195, 0.7244, 1.4944, 1.8675, 1.6026, 1.4321, 1.4464, 1.4764], device='cuda:6'), covar=tensor([0.4449, 0.6269, 0.5081, 0.5998, 0.5160, 0.6881, 0.6690, 0.6400], device='cuda:6'), in_proj_covar=tensor([0.0409, 0.0412, 0.0498, 0.0518, 0.0439, 0.0458, 0.0469, 0.0468], device='cuda:6'), out_proj_covar=tensor([9.9484e-05, 1.0225e-04, 1.1246e-04, 1.2294e-04, 1.0620e-04, 1.1067e-04, 1.1230e-04, 1.1231e-04], device='cuda:6') 2023-04-27 01:16:25,729 INFO [optim.py:369] (6/7) 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 01:17:07,497 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=58967.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:17:13,434 INFO [finetune.py:976] (6/7) Epoch 11, batch 1700, loss[loss=0.1935, simple_loss=0.2665, pruned_loss=0.06026, over 4835.00 frames. ], tot_loss[loss=0.1842, simple_loss=0.2506, pruned_loss=0.0589, over 957521.16 frames. ], batch size: 47, lr: 3.70e-03, grad_scale: 32.0 2023-04-27 01:17:21,647 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 2023-04-27 01:17:47,002 INFO [finetune.py:976] (6/7) Epoch 11, batch 1750, loss[loss=0.2587, simple_loss=0.3165, pruned_loss=0.1005, over 4202.00 frames. ], tot_loss[loss=0.1866, simple_loss=0.2527, pruned_loss=0.06021, over 954646.50 frames. ], batch size: 65, lr: 3.70e-03, grad_scale: 32.0 2023-04-27 01:17:53,221 INFO [zipformer.py:1188] (6/7) 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] (6/7) 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 01:18:15,569 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-04-27 01:18:19,033 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=3.62 vs. limit=5.0 2023-04-27 01:18:30,830 INFO [finetune.py:976] (6/7) Epoch 11, batch 1800, loss[loss=0.204, simple_loss=0.2724, pruned_loss=0.06777, over 4772.00 frames. ], tot_loss[loss=0.1887, simple_loss=0.256, pruned_loss=0.06068, over 954856.21 frames. ], batch size: 29, lr: 3.70e-03, grad_scale: 32.0 2023-04-27 01:18:48,970 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.2148, 1.4727, 1.5139, 1.6752, 1.5917, 1.7083, 1.7158, 1.5974], device='cuda:6'), covar=tensor([0.4464, 0.5797, 0.5005, 0.5022, 0.6221, 0.8418, 0.5510, 0.5499], device='cuda:6'), in_proj_covar=tensor([0.0327, 0.0380, 0.0315, 0.0326, 0.0339, 0.0399, 0.0359, 0.0323], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 01:19:09,376 INFO [finetune.py:976] (6/7) Epoch 11, batch 1850, loss[loss=0.1778, simple_loss=0.2406, pruned_loss=0.05751, over 4756.00 frames. ], tot_loss[loss=0.1898, simple_loss=0.2576, pruned_loss=0.06093, over 954219.78 frames. ], batch size: 26, lr: 3.70e-03, grad_scale: 32.0 2023-04-27 01:19:11,900 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=59131.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:19:21,280 INFO [optim.py:369] (6/7) 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 01:19:38,876 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-04-27 01:19:52,248 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=59164.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:20:10,993 INFO [finetune.py:976] (6/7) Epoch 11, batch 1900, loss[loss=0.2171, simple_loss=0.2752, pruned_loss=0.07947, over 4186.00 frames. ], tot_loss[loss=0.1923, simple_loss=0.2598, pruned_loss=0.06241, over 953375.98 frames. ], batch size: 65, lr: 3.70e-03, grad_scale: 32.0 2023-04-27 01:20:11,710 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([3.6946, 3.5696, 2.6678, 4.2237, 3.6579, 3.6977, 1.6588, 3.5898], device='cuda:6'), covar=tensor([0.1637, 0.1410, 0.3009, 0.1833, 0.2485, 0.1746, 0.5399, 0.2165], device='cuda:6'), in_proj_covar=tensor([0.0241, 0.0216, 0.0248, 0.0302, 0.0295, 0.0249, 0.0267, 0.0268], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 01:21:00,132 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=59225.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:21:01,214 INFO [finetune.py:976] (6/7) Epoch 11, batch 1950, loss[loss=0.1641, simple_loss=0.2458, pruned_loss=0.04124, over 4768.00 frames. ], tot_loss[loss=0.1906, simple_loss=0.258, pruned_loss=0.06163, over 952598.55 frames. ], batch size: 26, lr: 3.70e-03, grad_scale: 32.0 2023-04-27 01:21:08,361 INFO [optim.py:369] (6/7) 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 01:21:24,566 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=59262.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:21:35,027 INFO [finetune.py:976] (6/7) Epoch 11, batch 2000, loss[loss=0.1835, simple_loss=0.251, pruned_loss=0.05799, over 4744.00 frames. ], tot_loss[loss=0.1884, simple_loss=0.2555, pruned_loss=0.06072, over 954003.67 frames. ], batch size: 26, lr: 3.70e-03, grad_scale: 32.0 2023-04-27 01:21:42,690 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.65 vs. limit=5.0 2023-04-27 01:21:46,789 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-04-27 01:22:08,863 INFO [finetune.py:976] (6/7) Epoch 11, batch 2050, loss[loss=0.1608, simple_loss=0.2354, pruned_loss=0.04309, over 4749.00 frames. ], tot_loss[loss=0.1848, simple_loss=0.2515, pruned_loss=0.0591, over 953461.71 frames. ], batch size: 27, lr: 3.70e-03, grad_scale: 32.0 2023-04-27 01:22:15,380 INFO [zipformer.py:1188] (6/7) 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] (6/7) 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 01:22:48,348 INFO [finetune.py:976] (6/7) Epoch 11, batch 2100, loss[loss=0.2238, simple_loss=0.2891, pruned_loss=0.07927, over 4829.00 frames. ], tot_loss[loss=0.1861, simple_loss=0.2523, pruned_loss=0.05994, over 953779.96 frames. ], batch size: 33, lr: 3.70e-03, grad_scale: 32.0 2023-04-27 01:22:53,373 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=59385.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:23:06,273 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([3.3236, 3.2458, 2.5104, 3.8366, 3.3468, 3.3325, 1.4019, 3.2983], device='cuda:6'), covar=tensor([0.1933, 0.1482, 0.3342, 0.2421, 0.3775, 0.2070, 0.5829, 0.2503], device='cuda:6'), in_proj_covar=tensor([0.0241, 0.0215, 0.0248, 0.0302, 0.0295, 0.0248, 0.0268, 0.0268], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 01:23:12,730 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.6192, 2.4561, 2.7485, 2.9616, 2.9869, 2.3085, 2.0249, 2.7367], device='cuda:6'), covar=tensor([0.0896, 0.0912, 0.0547, 0.0629, 0.0540, 0.0868, 0.0901, 0.0525], device='cuda:6'), in_proj_covar=tensor([0.0194, 0.0202, 0.0182, 0.0174, 0.0177, 0.0187, 0.0159, 0.0183], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 01:23:32,963 INFO [finetune.py:976] (6/7) Epoch 11, batch 2150, loss[loss=0.2032, simple_loss=0.2691, pruned_loss=0.06862, over 4814.00 frames. ], tot_loss[loss=0.1891, simple_loss=0.256, pruned_loss=0.06111, over 954765.49 frames. ], batch size: 39, lr: 3.70e-03, grad_scale: 32.0 2023-04-27 01:23:35,513 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=59431.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:23:44,713 INFO [optim.py:369] (6/7) 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 01:24:02,514 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8673, 2.5257, 2.0037, 2.3489, 1.7562, 2.0761, 2.1836, 1.3625], device='cuda:6'), covar=tensor([0.2224, 0.1476, 0.0985, 0.1356, 0.3076, 0.1305, 0.1883, 0.3051], device='cuda:6'), in_proj_covar=tensor([0.0292, 0.0311, 0.0227, 0.0282, 0.0312, 0.0268, 0.0252, 0.0273], device='cuda:6'), out_proj_covar=tensor([1.1794e-04, 1.2508e-04, 9.0822e-05, 1.1309e-04, 1.2738e-04, 1.0761e-04, 1.0276e-04, 1.0922e-04], device='cuda:6') 2023-04-27 01:24:12,517 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6955, 1.3249, 1.8672, 2.0899, 1.7716, 1.7257, 1.7978, 1.8043], device='cuda:6'), covar=tensor([0.5913, 0.8281, 0.8310, 0.8811, 0.7606, 1.0420, 0.9968, 0.9865], device='cuda:6'), in_proj_covar=tensor([0.0408, 0.0409, 0.0497, 0.0515, 0.0438, 0.0456, 0.0464, 0.0466], device='cuda:6'), out_proj_covar=tensor([9.9137e-05, 1.0154e-04, 1.1216e-04, 1.2245e-04, 1.0598e-04, 1.1026e-04, 1.1137e-04, 1.1197e-04], device='cuda:6') 2023-04-27 01:24:15,855 INFO [finetune.py:976] (6/7) Epoch 11, batch 2200, loss[loss=0.1873, simple_loss=0.2578, pruned_loss=0.05839, over 4897.00 frames. ], tot_loss[loss=0.1901, simple_loss=0.2579, pruned_loss=0.06114, over 956044.58 frames. ], batch size: 36, lr: 3.70e-03, grad_scale: 32.0 2023-04-27 01:24:18,113 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=59479.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:24:37,676 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0920, 2.4595, 2.3218, 2.7675, 2.6682, 2.6563, 2.2566, 4.8925], device='cuda:6'), covar=tensor([0.0485, 0.0626, 0.0696, 0.0876, 0.0493, 0.0380, 0.0579, 0.0100], device='cuda:6'), in_proj_covar=tensor([0.0038, 0.0039, 0.0040, 0.0044, 0.0040, 0.0038, 0.0039, 0.0058], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0014, 0.0013, 0.0012, 0.0013, 0.0016], device='cuda:6') 2023-04-27 01:24:37,701 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=59502.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:24:45,232 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.2178, 1.2401, 1.3809, 1.5626, 1.5621, 1.2465, 0.8565, 1.3959], device='cuda:6'), covar=tensor([0.0838, 0.1250, 0.0836, 0.0601, 0.0640, 0.0837, 0.0910, 0.0593], device='cuda:6'), in_proj_covar=tensor([0.0196, 0.0204, 0.0184, 0.0176, 0.0179, 0.0189, 0.0160, 0.0185], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 01:24:54,722 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 11, batch 2250, loss[loss=0.1671, simple_loss=0.2257, pruned_loss=0.05419, over 4739.00 frames. ], tot_loss[loss=0.1924, simple_loss=0.2602, pruned_loss=0.06232, over 957739.49 frames. ], batch size: 23, lr: 3.70e-03, grad_scale: 32.0 2023-04-27 01:25:00,562 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([3.7335, 3.6338, 2.7000, 4.3053, 3.6701, 3.7113, 1.7335, 3.6763], device='cuda:6'), covar=tensor([0.1540, 0.1240, 0.3191, 0.1639, 0.2856, 0.1779, 0.5342, 0.2363], device='cuda:6'), in_proj_covar=tensor([0.0240, 0.0215, 0.0247, 0.0301, 0.0295, 0.0248, 0.0266, 0.0268], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 01:25:13,213 INFO [optim.py:369] (6/7) 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 01:25:45,852 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=59562.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:25:45,894 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.3573, 1.1308, 1.1648, 1.1332, 1.5138, 1.2752, 1.0600, 1.1320], device='cuda:6'), covar=tensor([0.1464, 0.1299, 0.1461, 0.1331, 0.0756, 0.1390, 0.1719, 0.1680], device='cuda:6'), in_proj_covar=tensor([0.0308, 0.0321, 0.0353, 0.0300, 0.0336, 0.0320, 0.0307, 0.0363], device='cuda:6'), out_proj_covar=tensor([6.4557e-05, 6.7986e-05, 7.5920e-05, 6.1870e-05, 7.0364e-05, 6.8401e-05, 6.5719e-05, 7.7906e-05], device='cuda:6') 2023-04-27 01:25:46,511 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=59563.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:25:55,722 INFO [finetune.py:976] (6/7) Epoch 11, batch 2300, loss[loss=0.1611, simple_loss=0.236, pruned_loss=0.0431, over 4801.00 frames. ], tot_loss[loss=0.1934, simple_loss=0.2616, pruned_loss=0.06261, over 956118.58 frames. ], batch size: 25, lr: 3.70e-03, grad_scale: 32.0 2023-04-27 01:26:03,402 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5726, 1.4721, 1.7506, 1.9123, 1.4390, 1.1390, 1.3710, 0.9232], device='cuda:6'), covar=tensor([0.0663, 0.0716, 0.0539, 0.0558, 0.0768, 0.1720, 0.0803, 0.0923], device='cuda:6'), in_proj_covar=tensor([0.0067, 0.0072, 0.0070, 0.0067, 0.0075, 0.0096, 0.0076, 0.0071], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-27 01:26:05,264 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-04-27 01:26:17,902 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=59610.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:26:29,151 INFO [finetune.py:976] (6/7) Epoch 11, batch 2350, loss[loss=0.1699, simple_loss=0.2387, pruned_loss=0.05057, over 4882.00 frames. ], tot_loss[loss=0.1903, simple_loss=0.2581, pruned_loss=0.06125, over 955688.16 frames. ], batch size: 32, lr: 3.70e-03, grad_scale: 32.0 2023-04-27 01:26:37,789 INFO [optim.py:369] (6/7) 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 01:27:02,576 INFO [finetune.py:976] (6/7) Epoch 11, batch 2400, loss[loss=0.1935, simple_loss=0.2512, pruned_loss=0.06791, over 4878.00 frames. ], tot_loss[loss=0.1875, simple_loss=0.2546, pruned_loss=0.0602, over 954404.83 frames. ], batch size: 34, lr: 3.70e-03, grad_scale: 32.0 2023-04-27 01:27:36,062 INFO [finetune.py:976] (6/7) Epoch 11, batch 2450, loss[loss=0.2141, simple_loss=0.2711, pruned_loss=0.07853, over 4904.00 frames. ], tot_loss[loss=0.185, simple_loss=0.2513, pruned_loss=0.05938, over 955438.07 frames. ], batch size: 32, lr: 3.70e-03, grad_scale: 32.0 2023-04-27 01:27:43,829 INFO [optim.py:369] (6/7) 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 01:27:48,420 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.4772, 3.3443, 0.9103, 1.7344, 1.8744, 2.2236, 1.8327, 1.0778], device='cuda:6'), covar=tensor([0.1460, 0.1027, 0.2133, 0.1387, 0.1148, 0.1115, 0.1701, 0.1961], device='cuda:6'), in_proj_covar=tensor([0.0117, 0.0249, 0.0140, 0.0122, 0.0134, 0.0154, 0.0118, 0.0122], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-27 01:28:09,958 INFO [finetune.py:976] (6/7) Epoch 11, batch 2500, loss[loss=0.2321, simple_loss=0.2965, pruned_loss=0.08381, over 4802.00 frames. ], tot_loss[loss=0.1881, simple_loss=0.2544, pruned_loss=0.06089, over 955498.86 frames. ], batch size: 51, lr: 3.70e-03, grad_scale: 32.0 2023-04-27 01:28:50,120 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=59820.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:28:54,835 INFO [finetune.py:976] (6/7) Epoch 11, batch 2550, loss[loss=0.2493, simple_loss=0.3037, pruned_loss=0.09743, over 4276.00 frames. ], tot_loss[loss=0.1902, simple_loss=0.2579, pruned_loss=0.06127, over 951885.19 frames. ], batch size: 65, lr: 3.70e-03, grad_scale: 32.0 2023-04-27 01:29:12,511 INFO [optim.py:369] (6/7) 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 01:29:16,932 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.50 vs. limit=2.0 2023-04-27 01:29:23,415 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=59848.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:29:35,321 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=59858.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:29:46,962 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=59868.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:29:57,641 INFO [finetune.py:976] (6/7) Epoch 11, batch 2600, loss[loss=0.1701, simple_loss=0.2325, pruned_loss=0.05386, over 4749.00 frames. ], tot_loss[loss=0.1907, simple_loss=0.2589, pruned_loss=0.06127, over 953691.09 frames. ], batch size: 27, lr: 3.70e-03, grad_scale: 32.0 2023-04-27 01:30:10,811 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=59889.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:30:29,461 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=59909.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:30:51,533 INFO [finetune.py:976] (6/7) Epoch 11, batch 2650, loss[loss=0.148, simple_loss=0.2107, pruned_loss=0.04267, over 4698.00 frames. ], tot_loss[loss=0.1919, simple_loss=0.26, pruned_loss=0.0619, over 953438.54 frames. ], batch size: 23, lr: 3.70e-03, grad_scale: 32.0 2023-04-27 01:30:54,979 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=3.84 vs. limit=5.0 2023-04-27 01:31:04,462 INFO [optim.py:369] (6/7) 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 01:31:21,124 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=59950.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:31:41,932 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.10 vs. limit=2.0 2023-04-27 01:31:43,485 INFO [finetune.py:976] (6/7) Epoch 11, batch 2700, loss[loss=0.1569, simple_loss=0.2193, pruned_loss=0.04728, over 4858.00 frames. ], tot_loss[loss=0.1893, simple_loss=0.2571, pruned_loss=0.06071, over 951522.24 frames. ], batch size: 31, lr: 3.70e-03, grad_scale: 32.0 2023-04-27 01:31:53,833 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.2892, 1.7208, 2.1644, 2.4042, 2.1208, 1.7205, 1.2432, 1.7950], device='cuda:6'), covar=tensor([0.3620, 0.3866, 0.1819, 0.2600, 0.2904, 0.2892, 0.5049, 0.2565], device='cuda:6'), in_proj_covar=tensor([0.0285, 0.0248, 0.0221, 0.0315, 0.0214, 0.0227, 0.0231, 0.0185], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:6') 2023-04-27 01:32:17,130 INFO [finetune.py:976] (6/7) Epoch 11, batch 2750, loss[loss=0.1448, simple_loss=0.2177, pruned_loss=0.03592, over 4765.00 frames. ], tot_loss[loss=0.1875, simple_loss=0.2547, pruned_loss=0.06015, over 952131.81 frames. ], batch size: 28, lr: 3.69e-03, grad_scale: 64.0 2023-04-27 01:32:24,355 INFO [optim.py:369] (6/7) 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] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=60040.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:32:33,976 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=60052.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:32:42,810 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=60065.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:32:50,418 INFO [finetune.py:976] (6/7) Epoch 11, batch 2800, loss[loss=0.1787, simple_loss=0.2356, pruned_loss=0.06089, over 4832.00 frames. ], tot_loss[loss=0.1843, simple_loss=0.2509, pruned_loss=0.05882, over 951281.53 frames. ], batch size: 33, lr: 3.69e-03, grad_scale: 64.0 2023-04-27 01:32:51,112 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=60078.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:33:01,573 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-04-27 01:33:05,639 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=60101.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 01:33:14,372 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=60113.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 01:33:22,800 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=60126.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:33:23,284 INFO [finetune.py:976] (6/7) Epoch 11, batch 2850, loss[loss=0.2465, simple_loss=0.3036, pruned_loss=0.09472, over 4131.00 frames. ], tot_loss[loss=0.185, simple_loss=0.2512, pruned_loss=0.05938, over 952114.60 frames. ], batch size: 65, lr: 3.69e-03, grad_scale: 64.0 2023-04-27 01:33:25,804 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([4.8457, 4.7646, 3.3175, 5.5454, 4.7723, 4.8241, 2.3925, 4.8716], device='cuda:6'), covar=tensor([0.1360, 0.0920, 0.2767, 0.0785, 0.3410, 0.1373, 0.5011, 0.1995], device='cuda:6'), in_proj_covar=tensor([0.0243, 0.0218, 0.0250, 0.0306, 0.0299, 0.0251, 0.0269, 0.0273], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 01:33:29,987 INFO [optim.py:369] (6/7) 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 01:33:30,706 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=60139.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:33:43,084 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=60158.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:33:46,493 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9563, 1.7032, 2.1765, 2.4320, 2.0662, 1.8968, 2.0641, 2.0327], device='cuda:6'), covar=tensor([0.6094, 0.8535, 0.8901, 0.7330, 0.7524, 1.1782, 1.0385, 0.9554], device='cuda:6'), in_proj_covar=tensor([0.0407, 0.0409, 0.0497, 0.0516, 0.0438, 0.0457, 0.0466, 0.0466], device='cuda:6'), out_proj_covar=tensor([9.8941e-05, 1.0151e-04, 1.1208e-04, 1.2256e-04, 1.0599e-04, 1.1045e-04, 1.1178e-04, 1.1190e-04], device='cuda:6') 2023-04-27 01:33:52,418 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.4431, 1.0769, 0.3916, 1.1554, 1.0991, 1.3502, 1.2447, 1.2393], device='cuda:6'), covar=tensor([0.0511, 0.0420, 0.0454, 0.0588, 0.0328, 0.0546, 0.0550, 0.0581], device='cuda:6'), in_proj_covar=tensor([0.0030, 0.0025, 0.0023, 0.0030, 0.0020, 0.0029, 0.0029, 0.0029], device='cuda:6'), out_proj_covar=tensor([0.0050, 0.0045, 0.0038, 0.0050, 0.0038, 0.0049, 0.0048, 0.0050], device='cuda:6') 2023-04-27 01:33:55,931 INFO [finetune.py:976] (6/7) Epoch 11, batch 2900, loss[loss=0.2564, simple_loss=0.3169, pruned_loss=0.09791, over 4896.00 frames. ], tot_loss[loss=0.1859, simple_loss=0.2525, pruned_loss=0.05962, over 949433.04 frames. ], batch size: 35, lr: 3.69e-03, grad_scale: 64.0 2023-04-27 01:34:15,764 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=60198.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:34:25,148 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=60204.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:34:26,338 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=60206.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:34:35,978 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-04-27 01:34:59,279 INFO [finetune.py:976] (6/7) Epoch 11, batch 2950, loss[loss=0.2098, simple_loss=0.2739, pruned_loss=0.07286, over 4828.00 frames. ], tot_loss[loss=0.1886, simple_loss=0.2558, pruned_loss=0.06071, over 949869.77 frames. ], batch size: 30, lr: 3.69e-03, grad_scale: 32.0 2023-04-27 01:35:03,030 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.0720, 2.2356, 1.8332, 1.6722, 2.1271, 1.6402, 2.6960, 1.4103], device='cuda:6'), covar=tensor([0.3938, 0.1739, 0.4169, 0.3272, 0.1855, 0.2729, 0.1521, 0.4862], device='cuda:6'), in_proj_covar=tensor([0.0340, 0.0347, 0.0423, 0.0356, 0.0381, 0.0379, 0.0377, 0.0418], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 01:35:12,689 INFO [optim.py:369] (6/7) 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 01:35:21,697 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=60245.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:35:23,415 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.80 vs. limit=2.0 2023-04-27 01:35:41,933 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=60259.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:36:05,412 INFO [finetune.py:976] (6/7) Epoch 11, batch 3000, loss[loss=0.1817, simple_loss=0.2513, pruned_loss=0.05606, over 4761.00 frames. ], tot_loss[loss=0.1885, simple_loss=0.2562, pruned_loss=0.06036, over 952068.62 frames. ], batch size: 28, lr: 3.69e-03, grad_scale: 32.0 2023-04-27 01:36:05,412 INFO [finetune.py:1001] (6/7) Computing validation loss 2023-04-27 01:36:11,651 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5525, 1.1761, 1.3082, 1.2279, 1.6997, 1.3810, 1.1072, 1.2566], device='cuda:6'), covar=tensor([0.1854, 0.1458, 0.2254, 0.1779, 0.1007, 0.1583, 0.2230, 0.2372], device='cuda:6'), in_proj_covar=tensor([0.0305, 0.0320, 0.0351, 0.0298, 0.0335, 0.0318, 0.0305, 0.0361], device='cuda:6'), out_proj_covar=tensor([6.3896e-05, 6.7707e-05, 7.5274e-05, 6.1380e-05, 7.0052e-05, 6.7741e-05, 6.5229e-05, 7.7438e-05], device='cuda:6') 2023-04-27 01:36:27,792 INFO [finetune.py:1010] (6/7) 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] (6/7) Maximum memory allocated so far is 6345MB 2023-04-27 01:36:36,573 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9103, 2.5097, 2.0891, 2.3340, 1.7928, 2.1004, 2.0583, 1.5876], device='cuda:6'), covar=tensor([0.2312, 0.1181, 0.0858, 0.1183, 0.3193, 0.1154, 0.2003, 0.2756], device='cuda:6'), in_proj_covar=tensor([0.0294, 0.0316, 0.0229, 0.0285, 0.0315, 0.0271, 0.0255, 0.0277], device='cuda:6'), out_proj_covar=tensor([1.1884e-04, 1.2700e-04, 9.1618e-05, 1.1423e-04, 1.2857e-04, 1.0874e-04, 1.0399e-04, 1.1092e-04], device='cuda:6') 2023-04-27 01:36:48,447 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 2023-04-27 01:37:01,935 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.22 vs. limit=5.0 2023-04-27 01:37:23,537 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.1826, 1.5536, 1.3820, 1.8219, 1.6709, 1.8754, 1.3930, 3.3363], device='cuda:6'), covar=tensor([0.0679, 0.0817, 0.0849, 0.1204, 0.0658, 0.0465, 0.0765, 0.0144], device='cuda:6'), in_proj_covar=tensor([0.0038, 0.0039, 0.0040, 0.0044, 0.0040, 0.0038, 0.0039, 0.0058], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0014, 0.0013, 0.0012, 0.0013, 0.0016], device='cuda:6') 2023-04-27 01:37:31,378 INFO [finetune.py:976] (6/7) Epoch 11, batch 3050, loss[loss=0.1992, simple_loss=0.2669, pruned_loss=0.06576, over 4823.00 frames. ], tot_loss[loss=0.1897, simple_loss=0.2575, pruned_loss=0.06093, over 952513.05 frames. ], batch size: 30, lr: 3.69e-03, grad_scale: 32.0 2023-04-27 01:37:45,692 INFO [optim.py:369] (6/7) 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 01:37:52,293 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.4190, 2.9468, 0.8706, 1.5851, 2.2210, 1.5989, 4.2781, 2.2402], device='cuda:6'), covar=tensor([0.0645, 0.0714, 0.0927, 0.1315, 0.0543, 0.0984, 0.0192, 0.0565], device='cuda:6'), in_proj_covar=tensor([0.0052, 0.0066, 0.0049, 0.0047, 0.0051, 0.0053, 0.0078, 0.0051], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:6') 2023-04-27 01:37:56,175 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 2023-04-27 01:38:26,536 INFO [finetune.py:976] (6/7) Epoch 11, batch 3100, loss[loss=0.2003, simple_loss=0.2589, pruned_loss=0.07083, over 4708.00 frames. ], tot_loss[loss=0.189, simple_loss=0.2562, pruned_loss=0.0609, over 952920.36 frames. ], batch size: 23, lr: 3.69e-03, grad_scale: 32.0 2023-04-27 01:38:40,716 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=60396.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 01:38:48,248 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=60408.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 01:38:56,655 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=60421.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:39:00,268 INFO [finetune.py:976] (6/7) Epoch 11, batch 3150, loss[loss=0.1933, simple_loss=0.262, pruned_loss=0.06227, over 4817.00 frames. ], tot_loss[loss=0.1879, simple_loss=0.2542, pruned_loss=0.06079, over 953878.22 frames. ], batch size: 40, lr: 3.69e-03, grad_scale: 32.0 2023-04-27 01:39:06,443 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=60434.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:39:09,792 INFO [optim.py:369] (6/7) 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 01:39:33,960 INFO [finetune.py:976] (6/7) Epoch 11, batch 3200, loss[loss=0.1942, simple_loss=0.2642, pruned_loss=0.06214, over 4938.00 frames. ], tot_loss[loss=0.1853, simple_loss=0.2514, pruned_loss=0.05959, over 954498.73 frames. ], batch size: 33, lr: 3.69e-03, grad_scale: 32.0 2023-04-27 01:39:53,415 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=60504.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:40:07,941 INFO [finetune.py:976] (6/7) Epoch 11, batch 3250, loss[loss=0.2001, simple_loss=0.2691, pruned_loss=0.06551, over 4926.00 frames. ], tot_loss[loss=0.1858, simple_loss=0.2519, pruned_loss=0.05987, over 955084.57 frames. ], batch size: 38, lr: 3.69e-03, grad_scale: 32.0 2023-04-27 01:40:15,718 INFO [optim.py:369] (6/7) 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 01:40:20,965 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=60545.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:40:25,628 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=60552.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:40:26,862 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=60554.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:40:52,022 INFO [finetune.py:976] (6/7) Epoch 11, batch 3300, loss[loss=0.1491, simple_loss=0.2238, pruned_loss=0.03723, over 4750.00 frames. ], tot_loss[loss=0.1884, simple_loss=0.2554, pruned_loss=0.06073, over 954005.98 frames. ], batch size: 26, lr: 3.69e-03, grad_scale: 32.0 2023-04-27 01:40:53,317 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.1202, 2.4628, 1.0966, 1.3365, 1.7788, 1.2812, 2.9809, 1.6321], device='cuda:6'), covar=tensor([0.0674, 0.0740, 0.0762, 0.1187, 0.0489, 0.0940, 0.0375, 0.0651], device='cuda:6'), in_proj_covar=tensor([0.0051, 0.0066, 0.0049, 0.0047, 0.0051, 0.0052, 0.0077, 0.0051], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:6') 2023-04-27 01:41:05,570 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.3949, 1.6361, 1.6808, 1.8480, 1.6293, 1.7856, 1.8168, 1.7093], device='cuda:6'), covar=tensor([0.5255, 0.7539, 0.5833, 0.5237, 0.6856, 0.9198, 0.7118, 0.6401], device='cuda:6'), in_proj_covar=tensor([0.0327, 0.0381, 0.0317, 0.0327, 0.0340, 0.0401, 0.0359, 0.0323], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 01:41:14,047 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=60593.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:41:34,509 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.0902, 0.7124, 0.9799, 0.7511, 1.2076, 0.9424, 0.8175, 1.0219], device='cuda:6'), covar=tensor([0.1420, 0.1559, 0.1985, 0.1555, 0.0954, 0.1366, 0.1789, 0.2022], device='cuda:6'), in_proj_covar=tensor([0.0304, 0.0318, 0.0350, 0.0295, 0.0332, 0.0317, 0.0303, 0.0359], device='cuda:6'), out_proj_covar=tensor([6.3731e-05, 6.7256e-05, 7.5150e-05, 6.0775e-05, 6.9480e-05, 6.7644e-05, 6.4741e-05, 7.7106e-05], device='cuda:6') 2023-04-27 01:41:58,104 INFO [finetune.py:976] (6/7) Epoch 11, batch 3350, loss[loss=0.1514, simple_loss=0.2219, pruned_loss=0.04045, over 4816.00 frames. ], tot_loss[loss=0.1907, simple_loss=0.2578, pruned_loss=0.06186, over 953770.33 frames. ], batch size: 25, lr: 3.69e-03, grad_scale: 32.0 2023-04-27 01:42:11,851 INFO [optim.py:369] (6/7) 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 01:42:48,688 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-04-27 01:42:52,677 INFO [finetune.py:976] (6/7) Epoch 11, batch 3400, loss[loss=0.2225, simple_loss=0.2878, pruned_loss=0.07861, over 4928.00 frames. ], tot_loss[loss=0.1904, simple_loss=0.2581, pruned_loss=0.06135, over 954820.08 frames. ], batch size: 42, lr: 3.69e-03, grad_scale: 32.0 2023-04-27 01:43:05,907 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=60696.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:43:14,618 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=60708.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:43:20,696 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=60718.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:43:22,444 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=60721.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:43:26,000 INFO [finetune.py:976] (6/7) Epoch 11, batch 3450, loss[loss=0.1533, simple_loss=0.2292, pruned_loss=0.03872, over 4805.00 frames. ], tot_loss[loss=0.1911, simple_loss=0.2589, pruned_loss=0.06164, over 955422.21 frames. ], batch size: 40, lr: 3.69e-03, grad_scale: 32.0 2023-04-27 01:43:30,798 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=60734.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:43:33,697 INFO [optim.py:369] (6/7) 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 01:43:35,283 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.68 vs. limit=5.0 2023-04-27 01:43:36,752 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=60744.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:43:45,936 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=60756.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:43:50,768 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5132, 1.4231, 1.7447, 1.8233, 1.3885, 1.0884, 1.3290, 0.9069], device='cuda:6'), covar=tensor([0.0695, 0.0676, 0.0513, 0.0675, 0.0823, 0.1587, 0.0929, 0.0943], device='cuda:6'), in_proj_covar=tensor([0.0068, 0.0073, 0.0072, 0.0068, 0.0076, 0.0098, 0.0078, 0.0072], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-27 01:43:54,320 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=60769.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:43:59,109 INFO [finetune.py:976] (6/7) Epoch 11, batch 3500, loss[loss=0.1545, simple_loss=0.2217, pruned_loss=0.04368, over 4929.00 frames. ], tot_loss[loss=0.1879, simple_loss=0.2552, pruned_loss=0.06029, over 957606.15 frames. ], batch size: 38, lr: 3.69e-03, grad_scale: 32.0 2023-04-27 01:44:00,483 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=60779.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:44:02,198 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=60782.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:44:29,923 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6774, 1.3257, 1.4105, 1.4462, 1.8433, 1.5093, 1.2844, 1.3568], device='cuda:6'), covar=tensor([0.1417, 0.1440, 0.1819, 0.1366, 0.0813, 0.1396, 0.1871, 0.2010], device='cuda:6'), in_proj_covar=tensor([0.0302, 0.0316, 0.0347, 0.0293, 0.0329, 0.0315, 0.0301, 0.0356], device='cuda:6'), out_proj_covar=tensor([6.3362e-05, 6.6803e-05, 7.4566e-05, 6.0185e-05, 6.8636e-05, 6.7127e-05, 6.4226e-05, 7.6509e-05], device='cuda:6') 2023-04-27 01:44:32,809 INFO [finetune.py:976] (6/7) Epoch 11, batch 3550, loss[loss=0.1787, simple_loss=0.2453, pruned_loss=0.05607, over 4896.00 frames. ], tot_loss[loss=0.1868, simple_loss=0.2536, pruned_loss=0.06005, over 957673.90 frames. ], batch size: 35, lr: 3.69e-03, grad_scale: 32.0 2023-04-27 01:44:40,092 INFO [optim.py:369] (6/7) 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 01:44:50,257 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=60854.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:44:50,288 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.2346, 1.9389, 2.1949, 2.5281, 2.6447, 1.9953, 1.6884, 2.3289], device='cuda:6'), covar=tensor([0.0820, 0.1113, 0.0695, 0.0624, 0.0500, 0.0897, 0.0977, 0.0562], device='cuda:6'), in_proj_covar=tensor([0.0198, 0.0204, 0.0184, 0.0177, 0.0181, 0.0190, 0.0160, 0.0186], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 01:45:06,041 INFO [finetune.py:976] (6/7) Epoch 11, batch 3600, loss[loss=0.1535, simple_loss=0.2322, pruned_loss=0.03738, over 4771.00 frames. ], tot_loss[loss=0.1845, simple_loss=0.251, pruned_loss=0.05905, over 958217.31 frames. ], batch size: 27, lr: 3.69e-03, grad_scale: 32.0 2023-04-27 01:45:21,747 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=60902.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:45:25,799 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-04-27 01:45:39,843 INFO [finetune.py:976] (6/7) Epoch 11, batch 3650, loss[loss=0.2652, simple_loss=0.3298, pruned_loss=0.1003, over 4746.00 frames. ], tot_loss[loss=0.189, simple_loss=0.2555, pruned_loss=0.06126, over 958186.44 frames. ], batch size: 54, lr: 3.69e-03, grad_scale: 32.0 2023-04-27 01:45:47,176 INFO [optim.py:369] (6/7) 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 01:45:53,884 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.63 vs. limit=5.0 2023-04-27 01:46:25,830 INFO [finetune.py:976] (6/7) Epoch 11, batch 3700, loss[loss=0.2005, simple_loss=0.265, pruned_loss=0.06801, over 4894.00 frames. ], tot_loss[loss=0.191, simple_loss=0.2584, pruned_loss=0.06178, over 957282.95 frames. ], batch size: 32, lr: 3.69e-03, grad_scale: 32.0 2023-04-27 01:47:05,592 INFO [finetune.py:976] (6/7) Epoch 11, batch 3750, loss[loss=0.2009, simple_loss=0.2627, pruned_loss=0.06954, over 4896.00 frames. ], tot_loss[loss=0.1902, simple_loss=0.2581, pruned_loss=0.06116, over 957209.42 frames. ], batch size: 43, lr: 3.69e-03, grad_scale: 32.0 2023-04-27 01:47:18,731 INFO [optim.py:369] (6/7) 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 01:47:40,517 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=61061.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:48:01,154 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=61074.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:48:05,290 INFO [finetune.py:976] (6/7) Epoch 11, batch 3800, loss[loss=0.1706, simple_loss=0.2444, pruned_loss=0.04838, over 4917.00 frames. ], tot_loss[loss=0.1905, simple_loss=0.2586, pruned_loss=0.06115, over 957730.18 frames. ], batch size: 33, lr: 3.69e-03, grad_scale: 32.0 2023-04-27 01:48:31,006 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8320, 2.4199, 1.8775, 1.7201, 1.3202, 1.3801, 1.9618, 1.2889], device='cuda:6'), covar=tensor([0.1684, 0.1495, 0.1440, 0.1787, 0.2433, 0.1945, 0.1012, 0.2048], device='cuda:6'), in_proj_covar=tensor([0.0194, 0.0213, 0.0168, 0.0204, 0.0202, 0.0184, 0.0158, 0.0185], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:6') 2023-04-27 01:48:42,574 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.77 vs. limit=2.0 2023-04-27 01:48:56,480 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=61122.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:48:59,871 INFO [finetune.py:976] (6/7) Epoch 11, batch 3850, loss[loss=0.1487, simple_loss=0.2236, pruned_loss=0.03692, over 4904.00 frames. ], tot_loss[loss=0.1892, simple_loss=0.257, pruned_loss=0.06068, over 956790.38 frames. ], batch size: 46, lr: 3.69e-03, grad_scale: 32.0 2023-04-27 01:49:08,085 INFO [optim.py:369] (6/7) 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 01:49:08,224 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([0.9722, 2.0111, 1.9105, 1.6710, 2.2371, 1.6686, 2.7547, 1.6878], device='cuda:6'), covar=tensor([0.3933, 0.1896, 0.4640, 0.3045, 0.1545, 0.2764, 0.1481, 0.4294], device='cuda:6'), in_proj_covar=tensor([0.0339, 0.0346, 0.0422, 0.0355, 0.0380, 0.0379, 0.0378, 0.0418], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 01:49:28,624 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6935, 2.0986, 1.6882, 2.0585, 1.5058, 1.8229, 1.9145, 1.4277], device='cuda:6'), covar=tensor([0.1967, 0.1860, 0.1362, 0.1398, 0.3138, 0.1515, 0.1776, 0.2500], device='cuda:6'), in_proj_covar=tensor([0.0295, 0.0316, 0.0228, 0.0286, 0.0315, 0.0272, 0.0256, 0.0277], device='cuda:6'), out_proj_covar=tensor([1.1921e-04, 1.2695e-04, 9.1185e-05, 1.1429e-04, 1.2872e-04, 1.0933e-04, 1.0443e-04, 1.1082e-04], device='cuda:6') 2023-04-27 01:49:31,455 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7644, 1.2651, 1.3771, 1.4752, 1.9147, 1.6183, 1.3454, 1.2565], device='cuda:6'), covar=tensor([0.1597, 0.1439, 0.1919, 0.1219, 0.0717, 0.1460, 0.2030, 0.2173], device='cuda:6'), in_proj_covar=tensor([0.0306, 0.0319, 0.0351, 0.0296, 0.0331, 0.0318, 0.0306, 0.0360], device='cuda:6'), out_proj_covar=tensor([6.4178e-05, 6.7427e-05, 7.5451e-05, 6.0912e-05, 6.9243e-05, 6.7814e-05, 6.5266e-05, 7.7390e-05], device='cuda:6') 2023-04-27 01:49:33,125 INFO [finetune.py:976] (6/7) Epoch 11, batch 3900, loss[loss=0.1957, simple_loss=0.2543, pruned_loss=0.06853, over 4806.00 frames. ], tot_loss[loss=0.1874, simple_loss=0.2546, pruned_loss=0.06014, over 957202.03 frames. ], batch size: 25, lr: 3.69e-03, grad_scale: 32.0 2023-04-27 01:50:04,201 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.4333, 1.7691, 1.8083, 1.9101, 1.8042, 1.9620, 1.9469, 1.8842], device='cuda:6'), covar=tensor([0.4354, 0.6391, 0.5305, 0.5438, 0.6440, 0.8601, 0.6229, 0.5814], device='cuda:6'), in_proj_covar=tensor([0.0325, 0.0376, 0.0314, 0.0325, 0.0338, 0.0399, 0.0357, 0.0321], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 01:50:05,884 INFO [finetune.py:976] (6/7) Epoch 11, batch 3950, loss[loss=0.23, simple_loss=0.2739, pruned_loss=0.09311, over 4868.00 frames. ], tot_loss[loss=0.1857, simple_loss=0.2522, pruned_loss=0.05965, over 957785.75 frames. ], batch size: 31, lr: 3.68e-03, grad_scale: 32.0 2023-04-27 01:50:15,546 INFO [optim.py:369] (6/7) 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 01:50:18,483 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 2023-04-27 01:50:22,435 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.18 vs. limit=5.0 2023-04-27 01:50:30,271 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.3205, 2.9167, 2.1628, 2.2546, 1.6917, 1.6411, 2.3520, 1.5983], device='cuda:6'), covar=tensor([0.1682, 0.1592, 0.1488, 0.1880, 0.2514, 0.2120, 0.1051, 0.2142], device='cuda:6'), in_proj_covar=tensor([0.0194, 0.0214, 0.0168, 0.0204, 0.0202, 0.0184, 0.0158, 0.0185], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:6') 2023-04-27 01:50:39,401 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-04-27 01:50:39,619 INFO [finetune.py:976] (6/7) Epoch 11, batch 4000, loss[loss=0.2181, simple_loss=0.2632, pruned_loss=0.08648, over 4062.00 frames. ], tot_loss[loss=0.185, simple_loss=0.2513, pruned_loss=0.05933, over 957107.44 frames. ], batch size: 18, lr: 3.68e-03, grad_scale: 32.0 2023-04-27 01:51:18,201 INFO [finetune.py:976] (6/7) Epoch 11, batch 4050, loss[loss=0.2033, simple_loss=0.28, pruned_loss=0.06332, over 4825.00 frames. ], tot_loss[loss=0.1891, simple_loss=0.2555, pruned_loss=0.06132, over 953538.31 frames. ], batch size: 39, lr: 3.68e-03, grad_scale: 32.0 2023-04-27 01:51:37,052 INFO [optim.py:369] (6/7) 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 01:51:38,030 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.75 vs. limit=2.0 2023-04-27 01:51:57,071 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([4.1224, 4.0281, 2.9117, 4.7377, 4.0132, 4.0739, 1.7288, 4.1842], device='cuda:6'), covar=tensor([0.1384, 0.1015, 0.3637, 0.0846, 0.2303, 0.1494, 0.5084, 0.1906], device='cuda:6'), in_proj_covar=tensor([0.0240, 0.0214, 0.0247, 0.0301, 0.0294, 0.0246, 0.0266, 0.0267], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 01:52:09,324 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.1903, 1.6140, 1.4113, 1.8686, 1.8078, 1.9682, 1.3903, 3.5743], device='cuda:6'), covar=tensor([0.0669, 0.0797, 0.0815, 0.1134, 0.0621, 0.0557, 0.0743, 0.0158], device='cuda:6'), in_proj_covar=tensor([0.0038, 0.0039, 0.0040, 0.0043, 0.0040, 0.0038, 0.0039, 0.0058], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0014, 0.0013, 0.0012, 0.0013, 0.0016], device='cuda:6') 2023-04-27 01:52:12,214 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.1804, 1.6674, 1.3950, 1.7439, 1.7897, 1.8746, 1.4090, 3.3748], device='cuda:6'), covar=tensor([0.0662, 0.0736, 0.0738, 0.1078, 0.0573, 0.0731, 0.0722, 0.0154], device='cuda:6'), in_proj_covar=tensor([0.0038, 0.0039, 0.0040, 0.0043, 0.0040, 0.0038, 0.0039, 0.0058], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0014, 0.0013, 0.0012, 0.0013, 0.0016], device='cuda:6') 2023-04-27 01:52:12,262 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.4128, 1.7033, 1.7277, 1.8420, 1.7334, 1.8442, 1.8427, 1.7541], device='cuda:6'), covar=tensor([0.4511, 0.6030, 0.5427, 0.4889, 0.5901, 0.8173, 0.6215, 0.5758], device='cuda:6'), in_proj_covar=tensor([0.0325, 0.0377, 0.0314, 0.0325, 0.0338, 0.0398, 0.0357, 0.0321], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 01:52:21,676 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 11, batch 4100, loss[loss=0.2575, simple_loss=0.3261, pruned_loss=0.09446, over 4815.00 frames. ], tot_loss[loss=0.193, simple_loss=0.2597, pruned_loss=0.06315, over 953482.15 frames. ], batch size: 51, lr: 3.68e-03, grad_scale: 32.0 2023-04-27 01:52:32,925 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7469, 1.6453, 2.0488, 2.0290, 1.7592, 1.6733, 1.7746, 1.8564], device='cuda:6'), covar=tensor([0.8179, 1.0888, 1.2225, 1.2118, 1.0123, 1.4848, 1.5558, 1.3757], device='cuda:6'), in_proj_covar=tensor([0.0412, 0.0412, 0.0500, 0.0518, 0.0442, 0.0461, 0.0472, 0.0471], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 01:53:16,502 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=61417.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 01:53:25,495 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 11, batch 4150, loss[loss=0.1878, simple_loss=0.2542, pruned_loss=0.06067, over 4747.00 frames. ], tot_loss[loss=0.192, simple_loss=0.2594, pruned_loss=0.06231, over 954097.02 frames. ], batch size: 27, lr: 3.68e-03, grad_scale: 32.0 2023-04-27 01:53:48,240 INFO [optim.py:369] (6/7) 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 01:54:37,125 INFO [finetune.py:976] (6/7) Epoch 11, batch 4200, loss[loss=0.1902, simple_loss=0.254, pruned_loss=0.06322, over 4740.00 frames. ], tot_loss[loss=0.1912, simple_loss=0.2591, pruned_loss=0.06168, over 954234.08 frames. ], batch size: 27, lr: 3.68e-03, grad_scale: 32.0 2023-04-27 01:54:52,765 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9567, 1.3862, 1.5873, 1.6260, 2.0994, 1.7998, 1.4083, 1.4406], device='cuda:6'), covar=tensor([0.1500, 0.1381, 0.1574, 0.1169, 0.0732, 0.1367, 0.2007, 0.1788], device='cuda:6'), in_proj_covar=tensor([0.0304, 0.0317, 0.0349, 0.0294, 0.0330, 0.0315, 0.0303, 0.0357], device='cuda:6'), out_proj_covar=tensor([6.3619e-05, 6.7010e-05, 7.4950e-05, 6.0449e-05, 6.8999e-05, 6.7212e-05, 6.4777e-05, 7.6657e-05], device='cuda:6') 2023-04-27 01:55:03,593 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-04-27 01:55:11,429 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.4915, 1.0093, 0.4375, 1.1950, 1.1131, 1.3919, 1.2186, 1.2594], device='cuda:6'), covar=tensor([0.0536, 0.0454, 0.0438, 0.0576, 0.0312, 0.0565, 0.0553, 0.0609], device='cuda:6'), in_proj_covar=tensor([0.0030, 0.0025, 0.0023, 0.0030, 0.0020, 0.0029, 0.0029, 0.0030], device='cuda:6'), out_proj_covar=tensor([0.0050, 0.0045, 0.0038, 0.0050, 0.0037, 0.0049, 0.0049, 0.0050], device='cuda:6') 2023-04-27 01:55:45,103 INFO [finetune.py:976] (6/7) Epoch 11, batch 4250, loss[loss=0.1356, simple_loss=0.206, pruned_loss=0.03256, over 4716.00 frames. ], tot_loss[loss=0.1886, simple_loss=0.2562, pruned_loss=0.06048, over 955685.61 frames. ], batch size: 23, lr: 3.68e-03, grad_scale: 32.0 2023-04-27 01:55:57,238 INFO [optim.py:369] (6/7) 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 01:56:30,631 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.4512, 1.3237, 1.6996, 1.6091, 1.3335, 1.2304, 1.3571, 0.8084], device='cuda:6'), covar=tensor([0.0679, 0.0889, 0.0508, 0.0737, 0.0871, 0.1410, 0.0614, 0.0768], device='cuda:6'), in_proj_covar=tensor([0.0068, 0.0072, 0.0071, 0.0067, 0.0075, 0.0097, 0.0076, 0.0072], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-27 01:56:41,099 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7558, 1.3257, 1.8682, 2.2647, 1.8624, 1.7306, 1.7982, 1.7748], device='cuda:6'), covar=tensor([0.5572, 0.7955, 0.8407, 0.7418, 0.7228, 0.9604, 0.9889, 0.9255], device='cuda:6'), in_proj_covar=tensor([0.0411, 0.0411, 0.0499, 0.0516, 0.0441, 0.0460, 0.0471, 0.0470], device='cuda:6'), out_proj_covar=tensor([9.9906e-05, 1.0181e-04, 1.1256e-04, 1.2262e-04, 1.0667e-04, 1.1130e-04, 1.1284e-04, 1.1282e-04], device='cuda:6') 2023-04-27 01:56:49,823 INFO [finetune.py:976] (6/7) Epoch 11, batch 4300, loss[loss=0.1809, simple_loss=0.2415, pruned_loss=0.06013, over 4811.00 frames. ], tot_loss[loss=0.1879, simple_loss=0.2544, pruned_loss=0.06066, over 954551.45 frames. ], batch size: 41, lr: 3.68e-03, grad_scale: 32.0 2023-04-27 01:57:24,122 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.8942, 2.8342, 2.2905, 3.3120, 2.8994, 2.8800, 1.2057, 2.7894], device='cuda:6'), covar=tensor([0.1954, 0.1603, 0.3269, 0.2767, 0.3310, 0.2176, 0.5901, 0.2911], device='cuda:6'), in_proj_covar=tensor([0.0239, 0.0215, 0.0248, 0.0302, 0.0294, 0.0248, 0.0266, 0.0267], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 01:57:57,465 INFO [finetune.py:976] (6/7) Epoch 11, batch 4350, loss[loss=0.2234, simple_loss=0.291, pruned_loss=0.07785, over 4810.00 frames. ], tot_loss[loss=0.1842, simple_loss=0.2504, pruned_loss=0.05895, over 954816.04 frames. ], batch size: 41, lr: 3.68e-03, grad_scale: 32.0 2023-04-27 01:58:10,460 INFO [optim.py:369] (6/7) 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 01:58:47,578 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0418, 1.5414, 1.8868, 2.2335, 2.4711, 1.8296, 1.6307, 2.1282], device='cuda:6'), covar=tensor([0.0795, 0.1263, 0.0710, 0.0592, 0.0498, 0.0884, 0.0833, 0.0524], device='cuda:6'), in_proj_covar=tensor([0.0193, 0.0201, 0.0181, 0.0173, 0.0178, 0.0186, 0.0157, 0.0182], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 01:59:02,122 INFO [finetune.py:976] (6/7) Epoch 11, batch 4400, loss[loss=0.2411, simple_loss=0.3036, pruned_loss=0.08925, over 4742.00 frames. ], tot_loss[loss=0.1861, simple_loss=0.2522, pruned_loss=0.05997, over 953386.83 frames. ], batch size: 59, lr: 3.68e-03, grad_scale: 32.0 2023-04-27 01:59:31,448 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7534, 2.7240, 2.3847, 2.6420, 3.0292, 2.5474, 3.8177, 2.2536], device='cuda:6'), covar=tensor([0.4134, 0.2236, 0.3946, 0.3081, 0.1680, 0.2762, 0.1619, 0.4067], device='cuda:6'), in_proj_covar=tensor([0.0346, 0.0353, 0.0430, 0.0363, 0.0388, 0.0386, 0.0385, 0.0424], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 01:59:45,408 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=5.04 vs. limit=5.0 2023-04-27 01:59:57,239 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=61717.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:00:08,368 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0153, 2.7380, 2.0831, 2.0003, 1.4145, 1.3545, 2.0671, 1.4129], device='cuda:6'), covar=tensor([0.1836, 0.1599, 0.1502, 0.1936, 0.2668, 0.2222, 0.1171, 0.2224], device='cuda:6'), in_proj_covar=tensor([0.0194, 0.0213, 0.0168, 0.0202, 0.0201, 0.0183, 0.0157, 0.0185], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:6') 2023-04-27 02:00:08,836 INFO [finetune.py:976] (6/7) Epoch 11, batch 4450, loss[loss=0.1897, simple_loss=0.2717, pruned_loss=0.05385, over 4817.00 frames. ], tot_loss[loss=0.1884, simple_loss=0.2554, pruned_loss=0.06073, over 952990.11 frames. ], batch size: 30, lr: 3.68e-03, grad_scale: 32.0 2023-04-27 02:00:17,353 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=61732.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:00:20,937 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.4305, 1.6897, 1.7418, 1.8807, 1.6643, 1.8025, 1.8074, 1.7585], device='cuda:6'), covar=tensor([0.4992, 0.6733, 0.5472, 0.5077, 0.6536, 0.8746, 0.7011, 0.6198], device='cuda:6'), in_proj_covar=tensor([0.0324, 0.0376, 0.0314, 0.0325, 0.0338, 0.0398, 0.0357, 0.0321], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 02:00:21,987 INFO [optim.py:369] (6/7) 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 02:00:24,561 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=61743.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:00:45,257 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-04-27 02:00:45,396 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=61765.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:01:03,254 INFO [finetune.py:976] (6/7) Epoch 11, batch 4500, loss[loss=0.2389, simple_loss=0.2971, pruned_loss=0.09033, over 4835.00 frames. ], tot_loss[loss=0.1902, simple_loss=0.2573, pruned_loss=0.0615, over 952787.71 frames. ], batch size: 49, lr: 3.68e-03, grad_scale: 32.0 2023-04-27 02:01:19,145 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=61793.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:01:26,333 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 11, batch 4550, loss[loss=0.2124, simple_loss=0.2814, pruned_loss=0.07169, over 4883.00 frames. ], tot_loss[loss=0.1901, simple_loss=0.2576, pruned_loss=0.06127, over 951896.21 frames. ], batch size: 32, lr: 3.68e-03, grad_scale: 32.0 2023-04-27 02:01:49,951 INFO [optim.py:369] (6/7) 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 02:02:15,276 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.2559, 1.3400, 1.6410, 1.7454, 1.6056, 1.7851, 1.7647, 1.6781], device='cuda:6'), covar=tensor([0.4725, 0.6128, 0.5411, 0.4897, 0.6422, 0.7820, 0.5934, 0.5855], device='cuda:6'), in_proj_covar=tensor([0.0324, 0.0375, 0.0312, 0.0324, 0.0336, 0.0397, 0.0356, 0.0320], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 02:02:16,335 INFO [finetune.py:976] (6/7) Epoch 11, batch 4600, loss[loss=0.213, simple_loss=0.2712, pruned_loss=0.07735, over 4738.00 frames. ], tot_loss[loss=0.1891, simple_loss=0.2569, pruned_loss=0.06069, over 952593.26 frames. ], batch size: 59, lr: 3.68e-03, grad_scale: 32.0 2023-04-27 02:02:40,686 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=61914.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:02:49,394 INFO [finetune.py:976] (6/7) Epoch 11, batch 4650, loss[loss=0.2063, simple_loss=0.2643, pruned_loss=0.07416, over 4910.00 frames. ], tot_loss[loss=0.1888, simple_loss=0.2555, pruned_loss=0.06103, over 950963.68 frames. ], batch size: 46, lr: 3.68e-03, grad_scale: 32.0 2023-04-27 02:02:56,648 INFO [optim.py:369] (6/7) 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 02:03:20,504 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=61975.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:03:22,014 INFO [finetune.py:976] (6/7) Epoch 11, batch 4700, loss[loss=0.155, simple_loss=0.2232, pruned_loss=0.04343, over 4828.00 frames. ], tot_loss[loss=0.1869, simple_loss=0.2529, pruned_loss=0.06047, over 952567.43 frames. ], batch size: 41, lr: 3.68e-03, grad_scale: 32.0 2023-04-27 02:03:39,667 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.4198, 1.6860, 1.8164, 1.9435, 1.8458, 1.9848, 1.8571, 1.8515], device='cuda:6'), covar=tensor([0.5125, 0.7117, 0.5982, 0.5748, 0.6608, 0.8745, 0.7090, 0.6249], device='cuda:6'), in_proj_covar=tensor([0.0326, 0.0376, 0.0313, 0.0325, 0.0337, 0.0398, 0.0357, 0.0321], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 02:03:58,802 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7251, 1.9411, 1.1844, 1.3985, 2.2530, 1.6962, 1.5744, 1.5809], device='cuda:6'), covar=tensor([0.0524, 0.0375, 0.0340, 0.0587, 0.0270, 0.0549, 0.0539, 0.0583], device='cuda:6'), in_proj_covar=tensor([0.0030, 0.0025, 0.0023, 0.0030, 0.0020, 0.0029, 0.0029, 0.0030], device='cuda:6'), out_proj_covar=tensor([0.0050, 0.0045, 0.0038, 0.0050, 0.0038, 0.0049, 0.0049, 0.0050], device='cuda:6') 2023-04-27 02:04:06,621 INFO [finetune.py:976] (6/7) Epoch 11, batch 4750, loss[loss=0.1992, simple_loss=0.2651, pruned_loss=0.06664, over 4908.00 frames. ], tot_loss[loss=0.1856, simple_loss=0.2511, pruned_loss=0.06007, over 952221.88 frames. ], batch size: 37, lr: 3.68e-03, grad_scale: 32.0 2023-04-27 02:04:20,384 INFO [optim.py:369] (6/7) 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 02:04:56,947 INFO [finetune.py:976] (6/7) Epoch 11, batch 4800, loss[loss=0.1991, simple_loss=0.2626, pruned_loss=0.06779, over 4816.00 frames. ], tot_loss[loss=0.1887, simple_loss=0.2548, pruned_loss=0.06128, over 952982.88 frames. ], batch size: 25, lr: 3.68e-03, grad_scale: 32.0 2023-04-27 02:05:05,180 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=62088.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:05:07,012 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.1476, 2.7886, 0.8541, 1.2812, 1.9237, 1.3550, 3.8863, 1.5849], device='cuda:6'), covar=tensor([0.0910, 0.0929, 0.1104, 0.1886, 0.0805, 0.1387, 0.0339, 0.1043], device='cuda:6'), in_proj_covar=tensor([0.0052, 0.0066, 0.0049, 0.0047, 0.0050, 0.0052, 0.0077, 0.0051], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0007], device='cuda:6') 2023-04-27 02:05:08,298 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5967, 0.9690, 1.5646, 2.0652, 1.6847, 1.5534, 1.5946, 1.6141], device='cuda:6'), covar=tensor([0.5660, 0.7845, 0.7703, 0.7202, 0.7222, 0.9217, 0.8895, 0.8916], device='cuda:6'), in_proj_covar=tensor([0.0411, 0.0410, 0.0498, 0.0515, 0.0442, 0.0459, 0.0469, 0.0469], device='cuda:6'), out_proj_covar=tensor([9.9815e-05, 1.0171e-04, 1.1250e-04, 1.2237e-04, 1.0703e-04, 1.1110e-04, 1.1244e-04, 1.1247e-04], device='cuda:6') 2023-04-27 02:05:10,751 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([0.9618, 0.9345, 1.0759, 1.0552, 0.9468, 0.8068, 0.8964, 0.4563], device='cuda:6'), covar=tensor([0.0638, 0.0582, 0.0594, 0.0562, 0.0760, 0.1251, 0.0547, 0.0942], device='cuda:6'), in_proj_covar=tensor([0.0067, 0.0072, 0.0070, 0.0067, 0.0075, 0.0097, 0.0076, 0.0072], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-27 02:05:11,913 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=62099.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:05:20,246 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.4464, 2.6800, 0.9938, 1.5196, 2.1724, 1.3836, 3.7684, 1.9868], device='cuda:6'), covar=tensor([0.0581, 0.0572, 0.0808, 0.1368, 0.0499, 0.1010, 0.0262, 0.0652], device='cuda:6'), in_proj_covar=tensor([0.0052, 0.0066, 0.0049, 0.0047, 0.0050, 0.0052, 0.0077, 0.0051], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0007], device='cuda:6') 2023-04-27 02:05:27,326 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=62122.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:05:30,802 INFO [finetune.py:976] (6/7) Epoch 11, batch 4850, loss[loss=0.2288, simple_loss=0.2934, pruned_loss=0.08208, over 4804.00 frames. ], tot_loss[loss=0.1915, simple_loss=0.2586, pruned_loss=0.06224, over 950231.98 frames. ], batch size: 51, lr: 3.68e-03, grad_scale: 32.0 2023-04-27 02:05:39,086 INFO [optim.py:369] (6/7) 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 02:06:08,696 INFO [finetune.py:976] (6/7) Epoch 11, batch 4900, loss[loss=0.1584, simple_loss=0.2323, pruned_loss=0.04232, over 4869.00 frames. ], tot_loss[loss=0.1921, simple_loss=0.2597, pruned_loss=0.06229, over 951743.05 frames. ], batch size: 34, lr: 3.68e-03, grad_scale: 32.0 2023-04-27 02:06:18,157 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=62183.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:06:52,813 INFO [finetune.py:976] (6/7) Epoch 11, batch 4950, loss[loss=0.1614, simple_loss=0.2299, pruned_loss=0.04647, over 4818.00 frames. ], tot_loss[loss=0.1932, simple_loss=0.2607, pruned_loss=0.06282, over 953522.56 frames. ], batch size: 33, lr: 3.68e-03, grad_scale: 64.0 2023-04-27 02:06:57,553 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0495, 2.4623, 1.0808, 1.3313, 1.8598, 1.2126, 3.0407, 1.6394], device='cuda:6'), covar=tensor([0.0653, 0.0540, 0.0681, 0.1231, 0.0458, 0.0953, 0.0326, 0.0611], device='cuda:6'), in_proj_covar=tensor([0.0051, 0.0066, 0.0048, 0.0047, 0.0050, 0.0052, 0.0077, 0.0051], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0007, 0.0008, 0.0008, 0.0010, 0.0007], device='cuda:6') 2023-04-27 02:07:01,574 INFO [optim.py:369] (6/7) 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 02:07:16,087 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.1498, 1.4224, 1.2800, 1.7634, 1.5605, 1.7733, 1.2951, 3.0228], device='cuda:6'), covar=tensor([0.0649, 0.0825, 0.0807, 0.1118, 0.0597, 0.0497, 0.0734, 0.0173], device='cuda:6'), in_proj_covar=tensor([0.0038, 0.0038, 0.0039, 0.0043, 0.0039, 0.0038, 0.0038, 0.0057], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0014, 0.0013, 0.0012, 0.0013, 0.0016], device='cuda:6') 2023-04-27 02:07:21,458 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=62270.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:07:23,326 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=62273.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:07:26,151 INFO [finetune.py:976] (6/7) Epoch 11, batch 5000, loss[loss=0.1957, simple_loss=0.2581, pruned_loss=0.06665, over 4925.00 frames. ], tot_loss[loss=0.1905, simple_loss=0.2578, pruned_loss=0.06157, over 954267.93 frames. ], batch size: 38, lr: 3.68e-03, grad_scale: 32.0 2023-04-27 02:07:46,937 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([4.0067, 3.8610, 2.8763, 4.5686, 4.0214, 3.9437, 1.7827, 3.8662], device='cuda:6'), covar=tensor([0.1442, 0.0976, 0.2978, 0.1372, 0.2173, 0.1632, 0.5118, 0.2176], device='cuda:6'), in_proj_covar=tensor([0.0244, 0.0219, 0.0253, 0.0307, 0.0300, 0.0250, 0.0271, 0.0271], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 02:07:58,884 INFO [finetune.py:976] (6/7) Epoch 11, batch 5050, loss[loss=0.1427, simple_loss=0.2128, pruned_loss=0.03631, over 4758.00 frames. ], tot_loss[loss=0.1885, simple_loss=0.2553, pruned_loss=0.06087, over 955186.69 frames. ], batch size: 28, lr: 3.68e-03, grad_scale: 32.0 2023-04-27 02:08:04,230 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=62334.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:08:08,229 INFO [optim.py:369] (6/7) 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 02:08:32,081 INFO [finetune.py:976] (6/7) Epoch 11, batch 5100, loss[loss=0.1558, simple_loss=0.2247, pruned_loss=0.04343, over 4935.00 frames. ], tot_loss[loss=0.1842, simple_loss=0.2512, pruned_loss=0.05857, over 956193.57 frames. ], batch size: 33, lr: 3.68e-03, grad_scale: 32.0 2023-04-27 02:08:37,461 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.3294, 1.4974, 1.3431, 1.8045, 1.6468, 2.0356, 1.3531, 3.3557], device='cuda:6'), covar=tensor([0.0606, 0.0781, 0.0803, 0.1100, 0.0628, 0.0486, 0.0738, 0.0157], device='cuda:6'), in_proj_covar=tensor([0.0038, 0.0038, 0.0040, 0.0043, 0.0039, 0.0038, 0.0038, 0.0057], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0014, 0.0013, 0.0012, 0.0013, 0.0016], device='cuda:6') 2023-04-27 02:08:40,389 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=62388.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:08:42,187 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=62391.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:08:47,995 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 11, batch 5150, loss[loss=0.2256, simple_loss=0.3134, pruned_loss=0.06886, over 4897.00 frames. ], tot_loss[loss=0.1851, simple_loss=0.2522, pruned_loss=0.05905, over 957729.85 frames. ], batch size: 43, lr: 3.67e-03, grad_scale: 32.0 2023-04-27 02:09:12,511 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=62436.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:09:14,848 INFO [optim.py:369] (6/7) 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 02:09:25,280 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=62447.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:09:33,245 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=62452.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:10:01,360 INFO [finetune.py:976] (6/7) Epoch 11, batch 5200, loss[loss=0.1382, simple_loss=0.2053, pruned_loss=0.03552, over 4491.00 frames. ], tot_loss[loss=0.189, simple_loss=0.2559, pruned_loss=0.06103, over 956333.14 frames. ], batch size: 19, lr: 3.67e-03, grad_scale: 32.0 2023-04-27 02:10:02,068 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=62478.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:10:58,658 INFO [finetune.py:976] (6/7) Epoch 11, batch 5250, loss[loss=0.1621, simple_loss=0.2203, pruned_loss=0.05193, over 4759.00 frames. ], tot_loss[loss=0.1906, simple_loss=0.2581, pruned_loss=0.06151, over 953526.71 frames. ], batch size: 26, lr: 3.67e-03, grad_scale: 32.0 2023-04-27 02:11:07,066 INFO [optim.py:369] (6/7) 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 02:11:10,694 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=62545.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:11:28,234 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=62570.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:11:32,446 INFO [finetune.py:976] (6/7) Epoch 11, batch 5300, loss[loss=0.2127, simple_loss=0.2811, pruned_loss=0.07213, over 4920.00 frames. ], tot_loss[loss=0.1917, simple_loss=0.2588, pruned_loss=0.06236, over 953392.67 frames. ], batch size: 33, lr: 3.67e-03, grad_scale: 32.0 2023-04-27 02:11:37,337 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=62585.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:11:52,277 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=62606.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:12:00,035 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=62618.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:12:05,913 INFO [finetune.py:976] (6/7) Epoch 11, batch 5350, loss[loss=0.1956, simple_loss=0.259, pruned_loss=0.06615, over 4701.00 frames. ], tot_loss[loss=0.1903, simple_loss=0.2579, pruned_loss=0.06137, over 951224.52 frames. ], batch size: 23, lr: 3.67e-03, grad_scale: 32.0 2023-04-27 02:12:07,195 INFO [zipformer.py:1188] (6/7) 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] (6/7) 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 02:12:18,129 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=62646.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:12:32,512 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=62666.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:12:37,832 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6879, 2.1386, 1.8414, 1.9744, 1.5873, 1.7215, 1.7038, 1.3995], device='cuda:6'), covar=tensor([0.1812, 0.1068, 0.0891, 0.1121, 0.3012, 0.1118, 0.1647, 0.2218], device='cuda:6'), in_proj_covar=tensor([0.0293, 0.0315, 0.0225, 0.0286, 0.0313, 0.0270, 0.0254, 0.0276], device='cuda:6'), out_proj_covar=tensor([1.1827e-04, 1.2646e-04, 9.0256e-05, 1.1428e-04, 1.2773e-04, 1.0847e-04, 1.0334e-04, 1.1042e-04], device='cuda:6') 2023-04-27 02:12:39,515 INFO [finetune.py:976] (6/7) Epoch 11, batch 5400, loss[loss=0.1441, simple_loss=0.2123, pruned_loss=0.03797, over 4803.00 frames. ], tot_loss[loss=0.1882, simple_loss=0.2554, pruned_loss=0.06053, over 951661.77 frames. ], batch size: 51, lr: 3.67e-03, grad_scale: 32.0 2023-04-27 02:13:02,223 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.51 vs. limit=5.0 2023-04-27 02:13:12,193 INFO [finetune.py:976] (6/7) Epoch 11, batch 5450, loss[loss=0.1667, simple_loss=0.2371, pruned_loss=0.04816, over 4828.00 frames. ], tot_loss[loss=0.1859, simple_loss=0.2527, pruned_loss=0.05955, over 951593.43 frames. ], batch size: 33, lr: 3.67e-03, grad_scale: 32.0 2023-04-27 02:13:12,309 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=62727.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:13:20,547 INFO [optim.py:369] (6/7) 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 02:13:25,401 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=62747.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:13:40,264 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-04-27 02:13:42,652 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 11, batch 5500, loss[loss=0.2046, simple_loss=0.2638, pruned_loss=0.07269, over 4907.00 frames. ], tot_loss[loss=0.1834, simple_loss=0.2498, pruned_loss=0.05857, over 952806.12 frames. ], batch size: 36, lr: 3.67e-03, grad_scale: 32.0 2023-04-27 02:13:46,159 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=62778.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:13:56,096 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-04-27 02:14:18,369 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 11, batch 5550, loss[loss=0.1936, simple_loss=0.2614, pruned_loss=0.06291, over 4798.00 frames. ], tot_loss[loss=0.1856, simple_loss=0.2523, pruned_loss=0.05945, over 953131.93 frames. ], batch size: 29, lr: 3.67e-03, grad_scale: 32.0 2023-04-27 02:14:23,732 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=62834.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:14:26,860 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.78 vs. limit=2.0 2023-04-27 02:14:27,255 INFO [optim.py:369] (6/7) 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 02:14:46,601 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9852, 1.4259, 1.8325, 2.0532, 1.8240, 1.4407, 1.0517, 1.5468], device='cuda:6'), covar=tensor([0.3615, 0.3666, 0.1831, 0.2697, 0.2889, 0.2968, 0.4400, 0.2246], device='cuda:6'), in_proj_covar=tensor([0.0284, 0.0248, 0.0221, 0.0314, 0.0213, 0.0227, 0.0229, 0.0185], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:6') 2023-04-27 02:15:05,419 INFO [finetune.py:976] (6/7) Epoch 11, batch 5600, loss[loss=0.1706, simple_loss=0.2556, pruned_loss=0.04276, over 4924.00 frames. ], tot_loss[loss=0.1863, simple_loss=0.2545, pruned_loss=0.05908, over 954241.83 frames. ], batch size: 38, lr: 3.67e-03, grad_scale: 32.0 2023-04-27 02:15:37,392 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 11, batch 5650, loss[loss=0.1706, simple_loss=0.2471, pruned_loss=0.0471, over 4868.00 frames. ], tot_loss[loss=0.1882, simple_loss=0.257, pruned_loss=0.05971, over 951598.64 frames. ], batch size: 44, lr: 3.67e-03, grad_scale: 32.0 2023-04-27 02:16:11,879 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=62929.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:16:23,344 INFO [optim.py:369] (6/7) 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 02:16:23,985 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=62941.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:17:05,514 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.3688, 1.8882, 2.2664, 2.5657, 2.1829, 1.8677, 1.3903, 1.9948], device='cuda:6'), covar=tensor([0.3183, 0.3176, 0.1591, 0.2363, 0.2995, 0.2537, 0.4126, 0.2184], device='cuda:6'), in_proj_covar=tensor([0.0284, 0.0248, 0.0221, 0.0314, 0.0213, 0.0228, 0.0229, 0.0185], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:6') 2023-04-27 02:17:06,581 INFO [finetune.py:976] (6/7) Epoch 11, batch 5700, loss[loss=0.1615, simple_loss=0.236, pruned_loss=0.04349, over 3536.00 frames. ], tot_loss[loss=0.1857, simple_loss=0.2529, pruned_loss=0.05925, over 932934.77 frames. ], batch size: 15, lr: 3.67e-03, grad_scale: 32.0 2023-04-27 02:17:06,613 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=62977.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:17:38,295 INFO [finetune.py:976] (6/7) Epoch 12, batch 0, loss[loss=0.1633, simple_loss=0.2428, pruned_loss=0.04192, over 4890.00 frames. ], tot_loss[loss=0.1633, simple_loss=0.2428, pruned_loss=0.04192, over 4890.00 frames. ], batch size: 37, lr: 3.67e-03, grad_scale: 32.0 2023-04-27 02:17:38,295 INFO [finetune.py:1001] (6/7) Computing validation loss 2023-04-27 02:17:51,977 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.3384, 1.5106, 1.8249, 1.9853, 1.9031, 2.1094, 1.8682, 1.9254], device='cuda:6'), covar=tensor([0.3964, 0.5481, 0.4731, 0.4775, 0.5941, 0.7773, 0.5616, 0.5076], device='cuda:6'), in_proj_covar=tensor([0.0327, 0.0379, 0.0315, 0.0327, 0.0339, 0.0400, 0.0359, 0.0324], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 02:17:53,690 INFO [finetune.py:1010] (6/7) Epoch 12, validation: loss=0.1544, simple_loss=0.2267, pruned_loss=0.04099, over 2265189.00 frames. 2023-04-27 02:17:53,690 INFO [finetune.py:1011] (6/7) Maximum memory allocated so far is 6345MB 2023-04-27 02:18:09,368 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8440, 1.5879, 1.9065, 2.2316, 2.2969, 1.8090, 1.5547, 2.0240], device='cuda:6'), covar=tensor([0.0846, 0.1158, 0.0704, 0.0550, 0.0546, 0.0886, 0.0895, 0.0550], device='cuda:6'), in_proj_covar=tensor([0.0197, 0.0204, 0.0183, 0.0175, 0.0179, 0.0189, 0.0159, 0.0184], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 02:18:18,024 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=63022.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:18:39,644 INFO [optim.py:369] (6/7) 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 02:18:48,797 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=63047.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:18:54,573 INFO [finetune.py:976] (6/7) Epoch 12, batch 50, loss[loss=0.2097, simple_loss=0.2838, pruned_loss=0.06778, over 4863.00 frames. ], tot_loss[loss=0.1912, simple_loss=0.2592, pruned_loss=0.06163, over 217157.43 frames. ], batch size: 34, lr: 3.67e-03, grad_scale: 32.0 2023-04-27 02:19:10,685 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-04-27 02:19:24,455 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.3252, 2.1899, 2.5601, 2.7952, 2.8434, 2.2221, 1.9160, 2.5060], device='cuda:6'), covar=tensor([0.0895, 0.1007, 0.0539, 0.0535, 0.0617, 0.0933, 0.0876, 0.0549], device='cuda:6'), in_proj_covar=tensor([0.0197, 0.0204, 0.0183, 0.0175, 0.0180, 0.0189, 0.0159, 0.0184], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 02:19:25,114 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-04-27 02:19:30,969 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=63095.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:19:37,864 INFO [finetune.py:976] (6/7) Epoch 12, batch 100, loss[loss=0.1786, simple_loss=0.2565, pruned_loss=0.05032, over 4819.00 frames. ], tot_loss[loss=0.1876, simple_loss=0.2536, pruned_loss=0.06079, over 381666.32 frames. ], batch size: 30, lr: 3.67e-03, grad_scale: 32.0 2023-04-27 02:19:54,525 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=63129.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:20:01,187 INFO [optim.py:369] (6/7) 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 02:20:03,608 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.4185, 1.6953, 1.6152, 2.0811, 2.0773, 2.2115, 1.5359, 4.1916], device='cuda:6'), covar=tensor([0.0562, 0.0785, 0.0786, 0.1122, 0.0569, 0.0513, 0.0761, 0.0120], device='cuda:6'), in_proj_covar=tensor([0.0038, 0.0038, 0.0040, 0.0043, 0.0039, 0.0038, 0.0038, 0.0057], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0014, 0.0013, 0.0012, 0.0013, 0.0016], device='cuda:6') 2023-04-27 02:20:11,747 INFO [finetune.py:976] (6/7) Epoch 12, batch 150, loss[loss=0.1499, simple_loss=0.2153, pruned_loss=0.04229, over 4706.00 frames. ], tot_loss[loss=0.1835, simple_loss=0.2495, pruned_loss=0.05871, over 509162.33 frames. ], batch size: 23, lr: 3.67e-03, grad_scale: 32.0 2023-04-27 02:20:22,172 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-04-27 02:20:30,382 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5717, 1.3045, 4.5028, 4.1427, 3.9749, 4.2539, 4.2493, 3.9756], device='cuda:6'), covar=tensor([0.6921, 0.6118, 0.1027, 0.1835, 0.1084, 0.1354, 0.1203, 0.1491], device='cuda:6'), in_proj_covar=tensor([0.0306, 0.0302, 0.0399, 0.0404, 0.0346, 0.0404, 0.0311, 0.0366], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-27 02:20:58,559 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=63201.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:21:01,605 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-04-27 02:21:01,803 INFO [finetune.py:976] (6/7) Epoch 12, batch 200, loss[loss=0.2154, simple_loss=0.2769, pruned_loss=0.07697, over 4188.00 frames. ], tot_loss[loss=0.1847, simple_loss=0.2497, pruned_loss=0.05989, over 608076.63 frames. ], batch size: 65, lr: 3.67e-03, grad_scale: 32.0 2023-04-27 02:21:13,196 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=63221.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:21:24,496 INFO [optim.py:369] (6/7) 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 02:21:25,236 INFO [zipformer.py:1188] (6/7) 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] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=63249.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:21:34,067 INFO [finetune.py:976] (6/7) Epoch 12, batch 250, loss[loss=0.1982, simple_loss=0.2728, pruned_loss=0.06177, over 4903.00 frames. ], tot_loss[loss=0.1848, simple_loss=0.25, pruned_loss=0.05983, over 684174.59 frames. ], batch size: 36, lr: 3.67e-03, grad_scale: 32.0 2023-04-27 02:21:52,374 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=63282.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:21:56,562 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=63289.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:21:57,602 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-04-27 02:22:00,265 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.0483, 1.3832, 1.2588, 1.6966, 1.5720, 1.5914, 1.3257, 2.4155], device='cuda:6'), covar=tensor([0.0697, 0.0858, 0.0901, 0.1252, 0.0665, 0.0489, 0.0793, 0.0245], device='cuda:6'), in_proj_covar=tensor([0.0038, 0.0039, 0.0040, 0.0043, 0.0040, 0.0038, 0.0038, 0.0057], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0014, 0.0013, 0.0012, 0.0013, 0.0016], device='cuda:6') 2023-04-27 02:22:11,660 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7161, 1.3009, 1.8256, 2.1508, 1.8221, 1.6793, 1.7337, 1.7610], device='cuda:6'), covar=tensor([0.5746, 0.7910, 0.7430, 0.7190, 0.6995, 0.9735, 0.8896, 0.8938], device='cuda:6'), in_proj_covar=tensor([0.0408, 0.0409, 0.0496, 0.0513, 0.0439, 0.0459, 0.0466, 0.0468], device='cuda:6'), 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:6') 2023-04-27 02:22:12,123 INFO [finetune.py:976] (6/7) Epoch 12, batch 300, loss[loss=0.1702, simple_loss=0.2443, pruned_loss=0.04809, over 4833.00 frames. ], tot_loss[loss=0.1884, simple_loss=0.2545, pruned_loss=0.06115, over 745814.71 frames. ], batch size: 33, lr: 3.67e-03, grad_scale: 32.0 2023-04-27 02:22:36,083 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=63322.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:22:43,408 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.1325, 1.4173, 1.2853, 1.6408, 1.5994, 1.7980, 1.3083, 3.0799], device='cuda:6'), covar=tensor([0.0658, 0.0797, 0.0840, 0.1215, 0.0630, 0.0484, 0.0741, 0.0204], device='cuda:6'), in_proj_covar=tensor([0.0038, 0.0039, 0.0040, 0.0043, 0.0040, 0.0038, 0.0038, 0.0057], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0014, 0.0013, 0.0012, 0.0013, 0.0016], device='cuda:6') 2023-04-27 02:22:46,896 INFO [optim.py:369] (6/7) 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] (6/7) Epoch 12, batch 350, loss[loss=0.1656, simple_loss=0.2404, pruned_loss=0.04545, over 4825.00 frames. ], tot_loss[loss=0.1889, simple_loss=0.2558, pruned_loss=0.06095, over 794052.19 frames. ], batch size: 47, lr: 3.67e-03, grad_scale: 32.0 2023-04-27 02:23:19,006 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=63370.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:23:48,728 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.4554, 1.0875, 0.4962, 1.1716, 1.1940, 1.3408, 1.2129, 1.1833], device='cuda:6'), covar=tensor([0.0525, 0.0429, 0.0451, 0.0585, 0.0329, 0.0553, 0.0551, 0.0598], device='cuda:6'), in_proj_covar=tensor([0.0030, 0.0025, 0.0023, 0.0030, 0.0020, 0.0029, 0.0029, 0.0030], device='cuda:6'), out_proj_covar=tensor([0.0050, 0.0045, 0.0038, 0.0050, 0.0038, 0.0049, 0.0049, 0.0050], device='cuda:6') 2023-04-27 02:23:57,901 INFO [finetune.py:976] (6/7) Epoch 12, batch 400, loss[loss=0.1443, simple_loss=0.2035, pruned_loss=0.04253, over 4822.00 frames. ], tot_loss[loss=0.1887, simple_loss=0.2561, pruned_loss=0.06064, over 829946.72 frames. ], batch size: 25, lr: 3.66e-03, grad_scale: 32.0 2023-04-27 02:24:19,302 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=63420.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:24:21,023 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=63422.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:24:31,390 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=63429.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:24:43,212 INFO [optim.py:369] (6/7) 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] (6/7) Epoch 12, batch 450, loss[loss=0.2025, simple_loss=0.2706, pruned_loss=0.06716, over 4714.00 frames. ], tot_loss[loss=0.1878, simple_loss=0.2556, pruned_loss=0.05997, over 857645.69 frames. ], batch size: 23, lr: 3.66e-03, grad_scale: 32.0 2023-04-27 02:25:38,225 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=63477.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:25:46,564 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=63481.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:25:47,772 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=63483.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:26:12,257 INFO [finetune.py:976] (6/7) Epoch 12, batch 500, loss[loss=0.2013, simple_loss=0.2664, pruned_loss=0.06817, over 4868.00 frames. ], tot_loss[loss=0.186, simple_loss=0.2533, pruned_loss=0.05935, over 878725.59 frames. ], batch size: 31, lr: 3.66e-03, grad_scale: 32.0 2023-04-27 02:26:41,243 INFO [optim.py:369] (6/7) 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] (6/7) Epoch 12, batch 550, loss[loss=0.2365, simple_loss=0.2773, pruned_loss=0.09782, over 4820.00 frames. ], tot_loss[loss=0.1838, simple_loss=0.2506, pruned_loss=0.05849, over 895200.78 frames. ], batch size: 41, lr: 3.66e-03, grad_scale: 32.0 2023-04-27 02:26:59,476 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=63569.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:27:05,806 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=63577.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:27:09,858 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([4.7038, 4.6170, 3.2710, 5.3592, 4.7283, 4.6766, 2.2744, 4.5598], device='cuda:6'), covar=tensor([0.1351, 0.0855, 0.3121, 0.0853, 0.3762, 0.1557, 0.5200, 0.1880], device='cuda:6'), in_proj_covar=tensor([0.0243, 0.0216, 0.0248, 0.0302, 0.0297, 0.0246, 0.0269, 0.0269], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 02:27:12,199 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=63586.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:27:23,755 INFO [finetune.py:976] (6/7) Epoch 12, batch 600, loss[loss=0.2392, simple_loss=0.3001, pruned_loss=0.08915, over 4920.00 frames. ], tot_loss[loss=0.1858, simple_loss=0.252, pruned_loss=0.05978, over 909686.51 frames. ], batch size: 38, lr: 3.66e-03, grad_scale: 32.0 2023-04-27 02:27:41,600 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=63630.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:27:48,535 INFO [optim.py:369] (6/7) 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 02:27:52,951 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=63647.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:27:57,757 INFO [finetune.py:976] (6/7) Epoch 12, batch 650, loss[loss=0.2148, simple_loss=0.284, pruned_loss=0.07279, over 4863.00 frames. ], tot_loss[loss=0.1894, simple_loss=0.2566, pruned_loss=0.06108, over 921351.44 frames. ], batch size: 34, lr: 3.66e-03, grad_scale: 32.0 2023-04-27 02:28:41,569 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7079, 1.3968, 1.6828, 2.0254, 2.0950, 1.5950, 1.2111, 1.9088], device='cuda:6'), covar=tensor([0.0759, 0.1328, 0.0744, 0.0520, 0.0511, 0.0845, 0.0901, 0.0539], device='cuda:6'), in_proj_covar=tensor([0.0197, 0.0203, 0.0184, 0.0175, 0.0179, 0.0189, 0.0160, 0.0184], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 02:28:42,667 INFO [finetune.py:976] (6/7) Epoch 12, batch 700, loss[loss=0.2023, simple_loss=0.2643, pruned_loss=0.07016, over 4919.00 frames. ], tot_loss[loss=0.1905, simple_loss=0.2578, pruned_loss=0.06164, over 928828.99 frames. ], batch size: 36, lr: 3.66e-03, grad_scale: 32.0 2023-04-27 02:29:23,284 INFO [optim.py:369] (6/7) 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 02:29:25,242 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.5737, 3.3045, 2.7167, 2.8270, 2.3196, 2.8077, 2.8530, 2.1301], device='cuda:6'), covar=tensor([0.2159, 0.1419, 0.0811, 0.1420, 0.3020, 0.1160, 0.1840, 0.2848], device='cuda:6'), in_proj_covar=tensor([0.0290, 0.0311, 0.0223, 0.0280, 0.0308, 0.0265, 0.0251, 0.0271], device='cuda:6'), 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:6') 2023-04-27 02:29:32,956 INFO [finetune.py:976] (6/7) Epoch 12, batch 750, loss[loss=0.1506, simple_loss=0.2339, pruned_loss=0.03363, over 4841.00 frames. ], tot_loss[loss=0.1902, simple_loss=0.2584, pruned_loss=0.06106, over 937005.04 frames. ], batch size: 44, lr: 3.66e-03, grad_scale: 32.0 2023-04-27 02:29:45,879 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=63776.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:29:47,608 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=63778.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:30:06,647 INFO [finetune.py:976] (6/7) Epoch 12, batch 800, loss[loss=0.194, simple_loss=0.2662, pruned_loss=0.06084, over 4828.00 frames. ], tot_loss[loss=0.1905, simple_loss=0.2588, pruned_loss=0.06108, over 940462.85 frames. ], batch size: 47, lr: 3.66e-03, grad_scale: 32.0 2023-04-27 02:30:09,838 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=63810.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:30:39,624 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([0.9268, 1.2097, 3.2353, 2.9940, 2.9185, 3.1027, 3.0966, 2.8623], device='cuda:6'), covar=tensor([0.6947, 0.5303, 0.1362, 0.1994, 0.1314, 0.1945, 0.2303, 0.1768], device='cuda:6'), in_proj_covar=tensor([0.0308, 0.0304, 0.0400, 0.0405, 0.0347, 0.0403, 0.0312, 0.0367], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-27 02:30:40,781 INFO [optim.py:369] (6/7) 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 02:30:49,737 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=63845.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:31:01,222 INFO [finetune.py:976] (6/7) Epoch 12, batch 850, loss[loss=0.1392, simple_loss=0.2194, pruned_loss=0.02948, over 4823.00 frames. ], tot_loss[loss=0.188, simple_loss=0.256, pruned_loss=0.06003, over 943349.41 frames. ], batch size: 38, lr: 3.66e-03, grad_scale: 32.0 2023-04-27 02:31:21,214 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=63871.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:31:30,635 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=63877.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:32:06,884 INFO [finetune.py:976] (6/7) Epoch 12, batch 900, loss[loss=0.1858, simple_loss=0.243, pruned_loss=0.06433, over 4835.00 frames. ], tot_loss[loss=0.187, simple_loss=0.254, pruned_loss=0.05999, over 946132.91 frames. ], batch size: 47, lr: 3.66e-03, grad_scale: 32.0 2023-04-27 02:32:07,626 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=63906.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:32:19,258 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=63916.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:32:24,726 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=63925.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:32:24,736 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=63925.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:32:34,616 INFO [optim.py:369] (6/7) 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 02:32:36,450 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=63942.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:32:45,689 INFO [finetune.py:976] (6/7) Epoch 12, batch 950, loss[loss=0.1766, simple_loss=0.2487, pruned_loss=0.05224, over 4890.00 frames. ], tot_loss[loss=0.1857, simple_loss=0.2528, pruned_loss=0.05933, over 948566.31 frames. ], batch size: 32, lr: 3.66e-03, grad_scale: 32.0 2023-04-27 02:32:59,650 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 12, batch 1000, loss[loss=0.206, simple_loss=0.2797, pruned_loss=0.06616, over 4856.00 frames. ], tot_loss[loss=0.1878, simple_loss=0.2548, pruned_loss=0.06035, over 950738.89 frames. ], batch size: 31, lr: 3.66e-03, grad_scale: 32.0 2023-04-27 02:33:28,304 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-04-27 02:33:32,466 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=64023.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:33:43,221 INFO [optim.py:369] (6/7) 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] (6/7) Epoch 12, batch 1050, loss[loss=0.1805, simple_loss=0.2603, pruned_loss=0.05036, over 4826.00 frames. ], tot_loss[loss=0.1885, simple_loss=0.2567, pruned_loss=0.06021, over 950824.00 frames. ], batch size: 33, lr: 3.66e-03, grad_scale: 32.0 2023-04-27 02:34:29,328 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=64076.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:34:30,595 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=64078.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:34:40,127 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=64084.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:34:51,350 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.3505, 2.2438, 2.5299, 2.8895, 2.7478, 2.3285, 1.8441, 2.5249], device='cuda:6'), covar=tensor([0.1001, 0.0992, 0.0684, 0.0616, 0.0695, 0.1102, 0.1037, 0.0646], device='cuda:6'), in_proj_covar=tensor([0.0194, 0.0200, 0.0182, 0.0172, 0.0177, 0.0186, 0.0158, 0.0182], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 02:35:05,820 INFO [finetune.py:976] (6/7) Epoch 12, batch 1100, loss[loss=0.1433, simple_loss=0.22, pruned_loss=0.03325, over 4757.00 frames. ], tot_loss[loss=0.1885, simple_loss=0.2567, pruned_loss=0.06018, over 951288.52 frames. ], batch size: 27, lr: 3.66e-03, grad_scale: 32.0 2023-04-27 02:35:11,674 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.3680, 2.3160, 2.0476, 2.0964, 2.5124, 2.0454, 3.3356, 1.8919], device='cuda:6'), covar=tensor([0.4517, 0.2438, 0.4687, 0.4031, 0.2066, 0.2992, 0.1785, 0.4737], device='cuda:6'), in_proj_covar=tensor([0.0343, 0.0349, 0.0431, 0.0361, 0.0385, 0.0382, 0.0377, 0.0423], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 02:35:28,512 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=64124.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:35:29,729 INFO [zipformer.py:1188] (6/7) 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] (6/7) 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] (6/7) Epoch 12, batch 1150, loss[loss=0.2198, simple_loss=0.2831, pruned_loss=0.07819, over 4870.00 frames. ], tot_loss[loss=0.1894, simple_loss=0.2577, pruned_loss=0.06055, over 953004.59 frames. ], batch size: 43, lr: 3.66e-03, grad_scale: 32.0 2023-04-27 02:35:57,681 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=64166.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:36:08,666 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.2974, 1.5852, 1.6667, 1.8015, 1.7245, 1.8000, 1.7754, 1.7388], device='cuda:6'), covar=tensor([0.4376, 0.6233, 0.5408, 0.5666, 0.6114, 0.8985, 0.6483, 0.5887], device='cuda:6'), in_proj_covar=tensor([0.0326, 0.0375, 0.0313, 0.0326, 0.0336, 0.0397, 0.0357, 0.0322], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 02:36:11,075 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0857, 1.6119, 1.9999, 2.2933, 1.9447, 1.5607, 1.2786, 1.7977], device='cuda:6'), covar=tensor([0.3450, 0.3211, 0.1689, 0.2310, 0.2588, 0.2724, 0.4249, 0.2116], device='cuda:6'), in_proj_covar=tensor([0.0283, 0.0246, 0.0219, 0.0312, 0.0211, 0.0226, 0.0227, 0.0183], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:6') 2023-04-27 02:36:19,851 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 12, batch 1200, loss[loss=0.1875, simple_loss=0.2488, pruned_loss=0.06303, over 4812.00 frames. ], tot_loss[loss=0.1882, simple_loss=0.2561, pruned_loss=0.06021, over 952365.98 frames. ], batch size: 40, lr: 3.66e-03, grad_scale: 32.0 2023-04-27 02:36:36,078 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=64225.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:36:40,398 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=64232.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:36:50,583 INFO [optim.py:369] (6/7) 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 02:36:51,914 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 12, batch 1250, loss[loss=0.1676, simple_loss=0.2355, pruned_loss=0.04988, over 4197.00 frames. ], tot_loss[loss=0.1867, simple_loss=0.254, pruned_loss=0.05975, over 951433.26 frames. ], batch size: 65, lr: 3.66e-03, grad_scale: 64.0 2023-04-27 02:37:33,224 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=64272.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:37:33,823 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=64273.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:37:55,440 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=64290.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:37:57,318 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=64293.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:38:16,203 INFO [finetune.py:976] (6/7) Epoch 12, batch 1300, loss[loss=0.1474, simple_loss=0.2136, pruned_loss=0.04058, over 4764.00 frames. ], tot_loss[loss=0.1841, simple_loss=0.2507, pruned_loss=0.05878, over 952853.84 frames. ], batch size: 26, lr: 3.66e-03, grad_scale: 32.0 2023-04-27 02:38:37,488 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-04-27 02:38:40,962 INFO [optim.py:369] (6/7) 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 02:38:43,502 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=64345.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:38:49,949 INFO [finetune.py:976] (6/7) Epoch 12, batch 1350, loss[loss=0.2892, simple_loss=0.3391, pruned_loss=0.1197, over 4141.00 frames. ], tot_loss[loss=0.1849, simple_loss=0.2511, pruned_loss=0.05931, over 952951.88 frames. ], batch size: 65, lr: 3.66e-03, grad_scale: 32.0 2023-04-27 02:39:01,821 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.1128, 2.6616, 1.9838, 2.0693, 1.4469, 1.4020, 2.2546, 1.3449], device='cuda:6'), covar=tensor([0.1698, 0.1577, 0.1553, 0.1850, 0.2425, 0.2054, 0.0975, 0.2184], device='cuda:6'), in_proj_covar=tensor([0.0193, 0.0213, 0.0168, 0.0203, 0.0200, 0.0183, 0.0157, 0.0186], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:6') 2023-04-27 02:39:02,784 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-04-27 02:39:07,146 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=64379.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:39:22,955 INFO [finetune.py:976] (6/7) Epoch 12, batch 1400, loss[loss=0.2298, simple_loss=0.2937, pruned_loss=0.08295, over 4816.00 frames. ], tot_loss[loss=0.1878, simple_loss=0.2548, pruned_loss=0.06041, over 954451.09 frames. ], batch size: 30, lr: 3.66e-03, grad_scale: 32.0 2023-04-27 02:39:24,216 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=64406.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:39:48,147 INFO [optim.py:369] (6/7) 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 02:40:07,658 INFO [finetune.py:976] (6/7) Epoch 12, batch 1450, loss[loss=0.2156, simple_loss=0.2801, pruned_loss=0.07555, over 4775.00 frames. ], tot_loss[loss=0.1887, simple_loss=0.2562, pruned_loss=0.06058, over 951382.10 frames. ], batch size: 51, lr: 3.66e-03, grad_scale: 32.0 2023-04-27 02:40:21,228 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=64465.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:40:21,840 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=64466.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:40:55,249 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=64501.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:40:57,611 INFO [finetune.py:976] (6/7) Epoch 12, batch 1500, loss[loss=0.1746, simple_loss=0.2554, pruned_loss=0.04694, over 4915.00 frames. ], tot_loss[loss=0.1898, simple_loss=0.2578, pruned_loss=0.06096, over 953154.86 frames. ], batch size: 38, lr: 3.66e-03, grad_scale: 32.0 2023-04-27 02:41:03,989 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=64514.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:41:06,600 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-04-27 02:41:12,852 INFO [zipformer.py:1188] (6/7) 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] (6/7) 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 02:41:27,205 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=64549.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:41:30,768 INFO [finetune.py:976] (6/7) Epoch 12, batch 1550, loss[loss=0.1581, simple_loss=0.2357, pruned_loss=0.04022, over 4854.00 frames. ], tot_loss[loss=0.1901, simple_loss=0.2586, pruned_loss=0.06076, over 954955.47 frames. ], batch size: 31, lr: 3.66e-03, grad_scale: 32.0 2023-04-27 02:41:34,586 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.86 vs. limit=5.0 2023-04-27 02:41:39,044 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9071, 1.6015, 1.8882, 2.1717, 2.1883, 1.8036, 1.4638, 2.0572], device='cuda:6'), covar=tensor([0.0694, 0.0959, 0.0512, 0.0444, 0.0490, 0.0690, 0.0778, 0.0413], device='cuda:6'), in_proj_covar=tensor([0.0195, 0.0201, 0.0182, 0.0174, 0.0177, 0.0188, 0.0159, 0.0182], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 02:41:42,600 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=64572.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:41:53,770 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=64588.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:41:54,914 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=64589.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:42:10,373 INFO [finetune.py:976] (6/7) Epoch 12, batch 1600, loss[loss=0.2226, simple_loss=0.2875, pruned_loss=0.07884, over 4909.00 frames. ], tot_loss[loss=0.1885, simple_loss=0.2561, pruned_loss=0.06042, over 953738.62 frames. ], batch size: 36, lr: 3.65e-03, grad_scale: 32.0 2023-04-27 02:42:32,225 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=64620.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:42:46,405 INFO [optim.py:369] (6/7) 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 02:42:56,552 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=64648.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:42:57,269 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2023-04-27 02:42:57,776 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=64650.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:43:06,463 INFO [finetune.py:976] (6/7) Epoch 12, batch 1650, loss[loss=0.2049, simple_loss=0.2516, pruned_loss=0.07914, over 4298.00 frames. ], tot_loss[loss=0.186, simple_loss=0.2533, pruned_loss=0.0593, over 955472.65 frames. ], batch size: 18, lr: 3.65e-03, grad_scale: 32.0 2023-04-27 02:43:29,824 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8136, 2.3691, 1.8038, 1.6318, 1.2997, 1.3465, 1.8658, 1.2445], device='cuda:6'), covar=tensor([0.1690, 0.1370, 0.1463, 0.1925, 0.2386, 0.2083, 0.1059, 0.2164], device='cuda:6'), in_proj_covar=tensor([0.0196, 0.0215, 0.0171, 0.0205, 0.0203, 0.0185, 0.0158, 0.0188], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:6') 2023-04-27 02:43:36,888 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=64679.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:44:09,944 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=64701.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:44:11,223 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=64703.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:44:12,324 INFO [finetune.py:976] (6/7) Epoch 12, batch 1700, loss[loss=0.151, simple_loss=0.2233, pruned_loss=0.0394, over 4832.00 frames. ], tot_loss[loss=0.1836, simple_loss=0.2508, pruned_loss=0.05822, over 956242.64 frames. ], batch size: 33, lr: 3.65e-03, grad_scale: 32.0 2023-04-27 02:44:14,922 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=64709.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:44:27,405 INFO [zipformer.py:1188] (6/7) 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] (6/7) 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 02:44:40,053 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=64745.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:44:44,754 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=64752.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:44:46,477 INFO [finetune.py:976] (6/7) Epoch 12, batch 1750, loss[loss=0.2208, simple_loss=0.2901, pruned_loss=0.07573, over 4818.00 frames. ], tot_loss[loss=0.1853, simple_loss=0.2526, pruned_loss=0.05901, over 956399.04 frames. ], batch size: 45, lr: 3.65e-03, grad_scale: 32.0 2023-04-27 02:44:51,987 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=64764.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:45:10,565 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0121, 2.4043, 0.9986, 1.3931, 1.7853, 1.2483, 3.1582, 1.5997], device='cuda:6'), covar=tensor([0.0641, 0.0712, 0.0761, 0.1169, 0.0480, 0.0959, 0.0260, 0.0617], device='cuda:6'), in_proj_covar=tensor([0.0051, 0.0066, 0.0049, 0.0047, 0.0051, 0.0053, 0.0076, 0.0051], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:6') 2023-04-27 02:45:20,204 INFO [finetune.py:976] (6/7) Epoch 12, batch 1800, loss[loss=0.1563, simple_loss=0.2307, pruned_loss=0.04101, over 4739.00 frames. ], tot_loss[loss=0.1881, simple_loss=0.2561, pruned_loss=0.06003, over 956990.26 frames. ], batch size: 27, lr: 3.65e-03, grad_scale: 32.0 2023-04-27 02:45:20,921 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=64806.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:45:25,129 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=64813.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:45:29,913 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=64821.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:45:39,293 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-04-27 02:46:00,006 INFO [optim.py:369] (6/7) 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] (6/7) Epoch 12, batch 1850, loss[loss=0.1952, simple_loss=0.2549, pruned_loss=0.06775, over 4733.00 frames. ], tot_loss[loss=0.1906, simple_loss=0.2585, pruned_loss=0.0614, over 955827.14 frames. ], batch size: 54, lr: 3.65e-03, grad_scale: 32.0 2023-04-27 02:46:58,523 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=64888.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:47:10,163 INFO [finetune.py:976] (6/7) Epoch 12, batch 1900, loss[loss=0.2199, simple_loss=0.2874, pruned_loss=0.0762, over 4858.00 frames. ], tot_loss[loss=0.1909, simple_loss=0.2593, pruned_loss=0.06128, over 954983.59 frames. ], batch size: 31, lr: 3.65e-03, grad_scale: 32.0 2023-04-27 02:47:35,552 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=64936.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:47:39,045 INFO [optim.py:369] (6/7) 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 02:47:48,349 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 12, batch 1950, loss[loss=0.1444, simple_loss=0.2274, pruned_loss=0.03067, over 4789.00 frames. ], tot_loss[loss=0.1893, simple_loss=0.2577, pruned_loss=0.06048, over 956438.06 frames. ], batch size: 29, lr: 3.65e-03, grad_scale: 32.0 2023-04-27 02:48:31,524 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=64976.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:48:42,270 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 2023-04-27 02:48:49,864 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=65001.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:48:51,659 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 12, batch 2000, loss[loss=0.192, simple_loss=0.2492, pruned_loss=0.06743, over 4232.00 frames. ], tot_loss[loss=0.1873, simple_loss=0.255, pruned_loss=0.05976, over 955272.07 frames. ], batch size: 65, lr: 3.65e-03, grad_scale: 32.0 2023-04-27 02:49:12,498 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=65037.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 02:49:15,251 INFO [optim.py:369] (6/7) 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 02:49:21,702 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=65049.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:49:26,346 INFO [finetune.py:976] (6/7) Epoch 12, batch 2050, loss[loss=0.1403, simple_loss=0.2064, pruned_loss=0.03708, over 4819.00 frames. ], tot_loss[loss=0.1852, simple_loss=0.252, pruned_loss=0.05923, over 955650.14 frames. ], batch size: 25, lr: 3.65e-03, grad_scale: 32.0 2023-04-27 02:49:26,468 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=65055.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:49:28,840 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=65059.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:49:56,652 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=65101.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:50:00,022 INFO [finetune.py:976] (6/7) Epoch 12, batch 2100, loss[loss=0.1724, simple_loss=0.2427, pruned_loss=0.05106, over 4755.00 frames. ], tot_loss[loss=0.1847, simple_loss=0.2511, pruned_loss=0.05908, over 955743.13 frames. ], batch size: 28, lr: 3.65e-03, grad_scale: 32.0 2023-04-27 02:50:00,241 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-04-27 02:50:01,930 INFO [zipformer.py:1188] (6/7) 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] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=65116.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:50:10,309 INFO [zipformer.py:1188] (6/7) 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] (6/7) 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 02:50:33,403 INFO [finetune.py:976] (6/7) Epoch 12, batch 2150, loss[loss=0.197, simple_loss=0.2831, pruned_loss=0.05547, over 4814.00 frames. ], tot_loss[loss=0.1873, simple_loss=0.2542, pruned_loss=0.06017, over 954464.17 frames. ], batch size: 40, lr: 3.65e-03, grad_scale: 32.0 2023-04-27 02:50:34,160 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9600, 1.3951, 1.7983, 2.1015, 1.7435, 1.3810, 1.0153, 1.5467], device='cuda:6'), covar=tensor([0.3231, 0.3325, 0.1777, 0.2461, 0.2714, 0.2796, 0.4621, 0.2295], device='cuda:6'), in_proj_covar=tensor([0.0288, 0.0250, 0.0223, 0.0317, 0.0214, 0.0230, 0.0232, 0.0187], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:6') 2023-04-27 02:50:42,399 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=65169.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:51:05,954 INFO [finetune.py:976] (6/7) Epoch 12, batch 2200, loss[loss=0.1523, simple_loss=0.2256, pruned_loss=0.03943, over 4744.00 frames. ], tot_loss[loss=0.1903, simple_loss=0.2575, pruned_loss=0.06158, over 952558.18 frames. ], batch size: 27, lr: 3.65e-03, grad_scale: 32.0 2023-04-27 02:51:35,435 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5056, 1.2938, 1.6973, 1.7410, 1.3301, 1.2156, 1.3985, 0.8668], device='cuda:6'), covar=tensor([0.0624, 0.0716, 0.0476, 0.0616, 0.0853, 0.1282, 0.0641, 0.0891], device='cuda:6'), in_proj_covar=tensor([0.0067, 0.0072, 0.0070, 0.0067, 0.0075, 0.0097, 0.0076, 0.0071], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-27 02:51:56,238 INFO [optim.py:369] (6/7) 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 02:51:59,300 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=65245.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:52:11,636 INFO [finetune.py:976] (6/7) Epoch 12, batch 2250, loss[loss=0.1483, simple_loss=0.2225, pruned_loss=0.03709, over 4834.00 frames. ], tot_loss[loss=0.1926, simple_loss=0.2601, pruned_loss=0.06256, over 954499.48 frames. ], batch size: 30, lr: 3.65e-03, grad_scale: 32.0 2023-04-27 02:52:53,430 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=65285.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:53:03,654 INFO [zipformer.py:1188] (6/7) 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] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=65304.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:53:21,922 INFO [finetune.py:976] (6/7) Epoch 12, batch 2300, loss[loss=0.1646, simple_loss=0.2389, pruned_loss=0.04512, over 4910.00 frames. ], tot_loss[loss=0.1921, simple_loss=0.2601, pruned_loss=0.06202, over 954119.72 frames. ], batch size: 36, lr: 3.65e-03, grad_scale: 32.0 2023-04-27 02:53:56,235 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=65332.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 02:54:07,779 INFO [optim.py:369] (6/7) 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 02:54:16,352 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=65346.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:54:20,876 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=65352.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:54:22,655 INFO [finetune.py:976] (6/7) Epoch 12, batch 2350, loss[loss=0.2109, simple_loss=0.2747, pruned_loss=0.07351, over 4872.00 frames. ], tot_loss[loss=0.1886, simple_loss=0.2562, pruned_loss=0.06051, over 954317.27 frames. ], batch size: 31, lr: 3.65e-03, grad_scale: 32.0 2023-04-27 02:54:25,651 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=65359.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:54:53,207 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=65401.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:54:55,973 INFO [finetune.py:976] (6/7) Epoch 12, batch 2400, loss[loss=0.2004, simple_loss=0.2626, pruned_loss=0.06904, over 4871.00 frames. ], tot_loss[loss=0.1862, simple_loss=0.2532, pruned_loss=0.05963, over 953271.78 frames. ], batch size: 34, lr: 3.65e-03, grad_scale: 32.0 2023-04-27 02:54:57,715 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=65407.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:54:58,361 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=65408.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:55:00,655 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=65411.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:55:20,645 INFO [optim.py:369] (6/7) 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 02:55:21,475 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-04-27 02:55:25,591 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=65449.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:55:26,231 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.3078, 1.5345, 1.3571, 1.4423, 1.2550, 1.2305, 1.3682, 1.0290], device='cuda:6'), covar=tensor([0.1274, 0.0946, 0.0795, 0.0995, 0.2411, 0.0888, 0.1257, 0.1702], device='cuda:6'), in_proj_covar=tensor([0.0287, 0.0308, 0.0220, 0.0278, 0.0308, 0.0261, 0.0250, 0.0270], device='cuda:6'), out_proj_covar=tensor([1.1580e-04, 1.2319e-04, 8.8054e-05, 1.1130e-04, 1.2573e-04, 1.0479e-04, 1.0152e-04, 1.0791e-04], device='cuda:6') 2023-04-27 02:55:29,227 INFO [finetune.py:976] (6/7) Epoch 12, batch 2450, loss[loss=0.202, simple_loss=0.2635, pruned_loss=0.07028, over 4827.00 frames. ], tot_loss[loss=0.1844, simple_loss=0.2509, pruned_loss=0.05896, over 952775.20 frames. ], batch size: 47, lr: 3.65e-03, grad_scale: 32.0 2023-04-27 02:55:30,349 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=65456.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:56:02,907 INFO [finetune.py:976] (6/7) Epoch 12, batch 2500, loss[loss=0.2093, simple_loss=0.265, pruned_loss=0.07678, over 4754.00 frames. ], tot_loss[loss=0.1846, simple_loss=0.2511, pruned_loss=0.05907, over 950389.51 frames. ], batch size: 26, lr: 3.65e-03, grad_scale: 32.0 2023-04-27 02:56:22,234 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=65532.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 02:56:28,028 INFO [optim.py:369] (6/7) 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 02:56:33,710 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7585, 2.2283, 1.6873, 1.4948, 1.3662, 1.3443, 1.8580, 1.2356], device='cuda:6'), covar=tensor([0.1904, 0.1482, 0.1564, 0.1955, 0.2540, 0.2105, 0.1091, 0.2179], device='cuda:6'), in_proj_covar=tensor([0.0196, 0.0214, 0.0169, 0.0203, 0.0203, 0.0184, 0.0157, 0.0186], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:6') 2023-04-27 02:56:36,599 INFO [finetune.py:976] (6/7) Epoch 12, batch 2550, loss[loss=0.1452, simple_loss=0.2172, pruned_loss=0.03658, over 4771.00 frames. ], tot_loss[loss=0.188, simple_loss=0.2555, pruned_loss=0.06023, over 951700.24 frames. ], batch size: 26, lr: 3.65e-03, grad_scale: 16.0 2023-04-27 02:56:49,183 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8485, 2.5217, 2.0373, 1.9072, 1.3883, 1.4328, 2.3595, 1.3514], device='cuda:6'), covar=tensor([0.1710, 0.1455, 0.1386, 0.1679, 0.2316, 0.1886, 0.0875, 0.1996], device='cuda:6'), in_proj_covar=tensor([0.0195, 0.0212, 0.0168, 0.0202, 0.0201, 0.0183, 0.0156, 0.0184], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:6') 2023-04-27 02:57:02,284 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=65593.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 02:57:05,155 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.3793, 3.2078, 2.6100, 2.8484, 2.3310, 2.6537, 2.8361, 2.0919], device='cuda:6'), covar=tensor([0.2058, 0.1283, 0.0869, 0.1231, 0.3043, 0.1161, 0.2010, 0.2701], device='cuda:6'), in_proj_covar=tensor([0.0288, 0.0309, 0.0221, 0.0279, 0.0310, 0.0262, 0.0250, 0.0271], device='cuda:6'), out_proj_covar=tensor([1.1617e-04, 1.2347e-04, 8.8475e-05, 1.1162e-04, 1.2658e-04, 1.0522e-04, 1.0173e-04, 1.0820e-04], device='cuda:6') 2023-04-27 02:57:10,048 INFO [finetune.py:976] (6/7) Epoch 12, batch 2600, loss[loss=0.1852, simple_loss=0.2534, pruned_loss=0.05854, over 4697.00 frames. ], tot_loss[loss=0.1887, simple_loss=0.256, pruned_loss=0.06067, over 952120.66 frames. ], batch size: 23, lr: 3.65e-03, grad_scale: 16.0 2023-04-27 02:57:17,905 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6075, 1.1549, 4.3448, 4.0468, 3.8082, 4.1069, 4.0279, 3.7590], device='cuda:6'), covar=tensor([0.7179, 0.6393, 0.1085, 0.1784, 0.1216, 0.1549, 0.1525, 0.1550], device='cuda:6'), in_proj_covar=tensor([0.0307, 0.0303, 0.0400, 0.0403, 0.0346, 0.0404, 0.0314, 0.0366], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-27 02:57:29,334 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=65632.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:57:31,113 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.1420, 1.8806, 2.1624, 2.5037, 2.5949, 1.9727, 1.6610, 2.1902], device='cuda:6'), covar=tensor([0.0874, 0.1055, 0.0683, 0.0530, 0.0510, 0.0932, 0.0836, 0.0559], device='cuda:6'), in_proj_covar=tensor([0.0193, 0.0200, 0.0181, 0.0173, 0.0176, 0.0187, 0.0157, 0.0181], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 02:57:32,925 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0542, 1.3975, 1.6335, 1.6606, 2.1781, 1.7734, 1.4525, 1.4431], device='cuda:6'), covar=tensor([0.1480, 0.1823, 0.1981, 0.1499, 0.0960, 0.1780, 0.2702, 0.2356], device='cuda:6'), in_proj_covar=tensor([0.0309, 0.0321, 0.0356, 0.0299, 0.0335, 0.0320, 0.0309, 0.0364], device='cuda:6'), out_proj_covar=tensor([6.4590e-05, 6.7411e-05, 7.6549e-05, 6.1438e-05, 6.9954e-05, 6.8153e-05, 6.5882e-05, 7.7972e-05], device='cuda:6') 2023-04-27 02:57:34,723 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=65641.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:57:35,231 INFO [optim.py:369] (6/7) 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] (6/7) Epoch 12, batch 2650, loss[loss=0.1967, simple_loss=0.2658, pruned_loss=0.06379, over 4914.00 frames. ], tot_loss[loss=0.188, simple_loss=0.2554, pruned_loss=0.0603, over 949313.18 frames. ], batch size: 38, lr: 3.65e-03, grad_scale: 16.0 2023-04-27 02:58:26,146 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=65680.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:58:37,483 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.0754, 2.5195, 0.8225, 1.4843, 1.5255, 1.9213, 1.6333, 0.8098], device='cuda:6'), covar=tensor([0.1500, 0.1151, 0.1882, 0.1376, 0.1159, 0.0897, 0.1437, 0.1692], device='cuda:6'), in_proj_covar=tensor([0.0118, 0.0247, 0.0141, 0.0122, 0.0134, 0.0154, 0.0118, 0.0120], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-27 02:58:46,716 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-04-27 02:58:54,676 INFO [finetune.py:976] (6/7) Epoch 12, batch 2700, loss[loss=0.2056, simple_loss=0.2672, pruned_loss=0.07207, over 4822.00 frames. ], tot_loss[loss=0.1866, simple_loss=0.2541, pruned_loss=0.05953, over 948893.81 frames. ], batch size: 33, lr: 3.65e-03, grad_scale: 16.0 2023-04-27 02:59:03,970 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=65711.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 02:59:42,189 INFO [optim.py:369] (6/7) 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 03:00:01,387 INFO [finetune.py:976] (6/7) Epoch 12, batch 2750, loss[loss=0.2183, simple_loss=0.2674, pruned_loss=0.08456, over 4832.00 frames. ], tot_loss[loss=0.1861, simple_loss=0.2526, pruned_loss=0.05979, over 949614.49 frames. ], batch size: 41, lr: 3.65e-03, grad_scale: 16.0 2023-04-27 03:00:02,925 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.41 vs. limit=5.0 2023-04-27 03:00:09,046 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=65759.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 03:00:14,038 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-04-27 03:01:06,633 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.2890, 3.1138, 2.4916, 2.6915, 2.1061, 2.4932, 2.6390, 1.9532], device='cuda:6'), covar=tensor([0.2171, 0.1247, 0.0882, 0.1200, 0.3074, 0.1157, 0.1940, 0.2780], device='cuda:6'), in_proj_covar=tensor([0.0291, 0.0311, 0.0223, 0.0282, 0.0312, 0.0264, 0.0252, 0.0273], device='cuda:6'), out_proj_covar=tensor([1.1709e-04, 1.2466e-04, 8.9197e-05, 1.1263e-04, 1.2730e-04, 1.0609e-04, 1.0230e-04, 1.0920e-04], device='cuda:6') 2023-04-27 03:01:07,124 INFO [finetune.py:976] (6/7) Epoch 12, batch 2800, loss[loss=0.1642, simple_loss=0.2303, pruned_loss=0.04904, over 4922.00 frames. ], tot_loss[loss=0.1831, simple_loss=0.2493, pruned_loss=0.05845, over 950874.67 frames. ], batch size: 37, lr: 3.64e-03, grad_scale: 16.0 2023-04-27 03:01:51,938 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=65838.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 03:01:59,671 INFO [optim.py:369] (6/7) 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 03:02:08,133 INFO [finetune.py:976] (6/7) Epoch 12, batch 2850, loss[loss=0.1567, simple_loss=0.2272, pruned_loss=0.04305, over 4717.00 frames. ], tot_loss[loss=0.1818, simple_loss=0.2474, pruned_loss=0.05806, over 949713.76 frames. ], batch size: 23, lr: 3.64e-03, grad_scale: 16.0 2023-04-27 03:02:29,546 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=65888.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 03:02:38,178 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 12, batch 2900, loss[loss=0.1994, simple_loss=0.2858, pruned_loss=0.05651, over 4824.00 frames. ], tot_loss[loss=0.1848, simple_loss=0.2513, pruned_loss=0.05911, over 950240.77 frames. ], batch size: 51, lr: 3.64e-03, grad_scale: 16.0 2023-04-27 03:03:02,661 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([0.6511, 1.6954, 1.6234, 1.3359, 1.8548, 1.4902, 2.1982, 1.4116], device='cuda:6'), covar=tensor([0.3588, 0.1528, 0.4490, 0.2897, 0.1319, 0.2083, 0.1395, 0.4391], device='cuda:6'), in_proj_covar=tensor([0.0337, 0.0343, 0.0424, 0.0355, 0.0380, 0.0379, 0.0373, 0.0418], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 03:03:05,047 INFO [zipformer.py:1188] (6/7) 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] (6/7) 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 03:03:15,316 INFO [finetune.py:976] (6/7) Epoch 12, batch 2950, loss[loss=0.2094, simple_loss=0.272, pruned_loss=0.07336, over 4898.00 frames. ], tot_loss[loss=0.1869, simple_loss=0.2542, pruned_loss=0.05987, over 950319.21 frames. ], batch size: 36, lr: 3.64e-03, grad_scale: 16.0 2023-04-27 03:03:24,183 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.23 vs. limit=5.0 2023-04-27 03:03:36,944 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 12, batch 3000, loss[loss=0.243, simple_loss=0.2903, pruned_loss=0.09783, over 4714.00 frames. ], tot_loss[loss=0.189, simple_loss=0.2566, pruned_loss=0.06072, over 952637.80 frames. ], batch size: 23, lr: 3.64e-03, grad_scale: 16.0 2023-04-27 03:03:49,756 INFO [finetune.py:1001] (6/7) Computing validation loss 2023-04-27 03:03:55,168 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([0.7237, 1.8370, 1.7663, 1.4258, 1.9684, 1.6107, 2.3438, 1.5646], device='cuda:6'), covar=tensor([0.3171, 0.1686, 0.4596, 0.2757, 0.1367, 0.2146, 0.1406, 0.4566], device='cuda:6'), in_proj_covar=tensor([0.0338, 0.0344, 0.0426, 0.0356, 0.0381, 0.0380, 0.0375, 0.0419], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 03:04:00,410 INFO [finetune.py:1010] (6/7) 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] (6/7) Maximum memory allocated so far is 6345MB 2023-04-27 03:04:22,349 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-04-27 03:04:29,444 INFO [optim.py:369] (6/7) 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 03:04:38,279 INFO [finetune.py:976] (6/7) Epoch 12, batch 3050, loss[loss=0.1816, simple_loss=0.2585, pruned_loss=0.05236, over 4804.00 frames. ], tot_loss[loss=0.1882, simple_loss=0.2563, pruned_loss=0.06004, over 952347.54 frames. ], batch size: 39, lr: 3.64e-03, grad_scale: 16.0 2023-04-27 03:04:43,339 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.11 vs. limit=5.0 2023-04-27 03:05:10,513 INFO [finetune.py:976] (6/7) Epoch 12, batch 3100, loss[loss=0.1724, simple_loss=0.2298, pruned_loss=0.0575, over 4776.00 frames. ], tot_loss[loss=0.1868, simple_loss=0.2548, pruned_loss=0.0594, over 950231.45 frames. ], batch size: 26, lr: 3.64e-03, grad_scale: 16.0 2023-04-27 03:05:40,811 INFO [optim.py:369] (6/7) 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 03:05:48,050 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=66144.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 03:05:54,642 INFO [finetune.py:976] (6/7) Epoch 12, batch 3150, loss[loss=0.2044, simple_loss=0.2641, pruned_loss=0.07232, over 4820.00 frames. ], tot_loss[loss=0.1861, simple_loss=0.2531, pruned_loss=0.05952, over 950803.44 frames. ], batch size: 33, lr: 3.64e-03, grad_scale: 16.0 2023-04-27 03:05:58,713 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=66160.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 03:06:27,521 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=66188.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 03:06:36,569 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=66194.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 03:06:38,979 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([3.8735, 3.7268, 2.8023, 4.4593, 3.9463, 3.8352, 1.9029, 3.8274], device='cuda:6'), covar=tensor([0.1646, 0.1081, 0.2796, 0.1846, 0.2458, 0.1812, 0.5649, 0.2216], device='cuda:6'), in_proj_covar=tensor([0.0242, 0.0216, 0.0248, 0.0302, 0.0297, 0.0246, 0.0269, 0.0268], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 03:06:49,145 INFO [finetune.py:976] (6/7) Epoch 12, batch 3200, loss[loss=0.1684, simple_loss=0.2325, pruned_loss=0.05218, over 4825.00 frames. ], tot_loss[loss=0.183, simple_loss=0.2496, pruned_loss=0.05822, over 953330.36 frames. ], batch size: 39, lr: 3.64e-03, grad_scale: 16.0 2023-04-27 03:06:49,261 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=66205.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 03:07:12,317 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=66221.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 03:07:34,235 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=66236.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 03:07:36,206 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.64 vs. limit=2.0 2023-04-27 03:07:43,555 INFO [optim.py:369] (6/7) 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 03:07:55,611 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.52 vs. limit=2.0 2023-04-27 03:07:57,817 INFO [finetune.py:976] (6/7) Epoch 12, batch 3250, loss[loss=0.2212, simple_loss=0.2892, pruned_loss=0.07656, over 4910.00 frames. ], tot_loss[loss=0.1862, simple_loss=0.2526, pruned_loss=0.05997, over 953323.89 frames. ], batch size: 35, lr: 3.64e-03, grad_scale: 16.0 2023-04-27 03:07:58,528 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6475, 1.2678, 4.4415, 4.1758, 3.8422, 4.2163, 4.0862, 3.8453], device='cuda:6'), covar=tensor([0.6851, 0.5797, 0.1168, 0.1775, 0.1185, 0.1436, 0.1606, 0.1520], device='cuda:6'), in_proj_covar=tensor([0.0311, 0.0306, 0.0404, 0.0408, 0.0352, 0.0409, 0.0316, 0.0370], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-27 03:08:05,459 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=66258.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 03:08:29,831 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0359, 2.6597, 0.9205, 1.3930, 2.0630, 1.3286, 3.3790, 1.7095], device='cuda:6'), covar=tensor([0.0736, 0.0820, 0.0979, 0.1392, 0.0553, 0.1004, 0.0309, 0.0647], device='cuda:6'), in_proj_covar=tensor([0.0051, 0.0066, 0.0049, 0.0047, 0.0051, 0.0052, 0.0077, 0.0051], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0007], device='cuda:6') 2023-04-27 03:08:43,233 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([3.8349, 3.8623, 2.8764, 4.4368, 3.9041, 3.8290, 1.8370, 3.7834], device='cuda:6'), covar=tensor([0.1665, 0.1247, 0.2809, 0.1697, 0.3783, 0.1808, 0.5522, 0.2367], device='cuda:6'), in_proj_covar=tensor([0.0242, 0.0215, 0.0247, 0.0302, 0.0296, 0.0247, 0.0269, 0.0268], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 03:08:47,969 INFO [finetune.py:976] (6/7) Epoch 12, batch 3300, loss[loss=0.2103, simple_loss=0.2927, pruned_loss=0.06389, over 4844.00 frames. ], tot_loss[loss=0.1894, simple_loss=0.2562, pruned_loss=0.06126, over 952114.72 frames. ], batch size: 49, lr: 3.64e-03, grad_scale: 16.0 2023-04-27 03:08:57,536 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=66319.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 03:09:13,362 INFO [optim.py:369] (6/7) 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 03:09:21,690 INFO [finetune.py:976] (6/7) Epoch 12, batch 3350, loss[loss=0.1646, simple_loss=0.2366, pruned_loss=0.04627, over 4799.00 frames. ], tot_loss[loss=0.1917, simple_loss=0.259, pruned_loss=0.06218, over 954275.65 frames. ], batch size: 26, lr: 3.64e-03, grad_scale: 16.0 2023-04-27 03:09:33,601 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.1864, 2.5750, 1.0407, 1.4131, 2.0438, 1.3234, 3.3984, 1.7380], device='cuda:6'), covar=tensor([0.0726, 0.0651, 0.0788, 0.1343, 0.0520, 0.1020, 0.0282, 0.0675], device='cuda:6'), in_proj_covar=tensor([0.0052, 0.0067, 0.0049, 0.0047, 0.0051, 0.0052, 0.0078, 0.0051], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:6') 2023-04-27 03:09:54,777 INFO [finetune.py:976] (6/7) Epoch 12, batch 3400, loss[loss=0.1972, simple_loss=0.2762, pruned_loss=0.05913, over 4076.00 frames. ], tot_loss[loss=0.1924, simple_loss=0.2602, pruned_loss=0.06228, over 954023.99 frames. ], batch size: 65, lr: 3.64e-03, grad_scale: 16.0 2023-04-27 03:10:02,222 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6488, 1.7908, 0.8350, 1.3973, 1.9535, 1.5477, 1.4436, 1.5169], device='cuda:6'), covar=tensor([0.0518, 0.0370, 0.0364, 0.0579, 0.0262, 0.0550, 0.0515, 0.0600], device='cuda:6'), in_proj_covar=tensor([0.0030, 0.0025, 0.0023, 0.0030, 0.0020, 0.0029, 0.0029, 0.0029], device='cuda:6'), out_proj_covar=tensor([0.0050, 0.0045, 0.0038, 0.0050, 0.0038, 0.0049, 0.0048, 0.0050], device='cuda:6') 2023-04-27 03:10:04,017 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6389, 1.7818, 0.8548, 1.4052, 1.7762, 1.5679, 1.4805, 1.4902], device='cuda:6'), covar=tensor([0.0522, 0.0380, 0.0402, 0.0585, 0.0285, 0.0558, 0.0530, 0.0614], device='cuda:6'), in_proj_covar=tensor([0.0030, 0.0025, 0.0023, 0.0030, 0.0020, 0.0029, 0.0029, 0.0029], device='cuda:6'), out_proj_covar=tensor([0.0050, 0.0045, 0.0038, 0.0050, 0.0038, 0.0049, 0.0049, 0.0050], device='cuda:6') 2023-04-27 03:10:06,332 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.3249, 1.6650, 1.7525, 1.8812, 1.7116, 1.8547, 1.8342, 1.7839], device='cuda:6'), covar=tensor([0.4824, 0.6680, 0.5510, 0.5525, 0.6673, 0.9075, 0.6620, 0.6026], device='cuda:6'), in_proj_covar=tensor([0.0328, 0.0378, 0.0317, 0.0327, 0.0339, 0.0399, 0.0358, 0.0326], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 03:10:20,217 INFO [optim.py:369] (6/7) 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 03:10:21,563 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5655, 2.0453, 1.7538, 1.9810, 1.5624, 1.7297, 1.7144, 1.3650], device='cuda:6'), covar=tensor([0.1933, 0.1176, 0.0721, 0.1068, 0.2996, 0.0994, 0.1628, 0.2167], device='cuda:6'), in_proj_covar=tensor([0.0289, 0.0310, 0.0221, 0.0281, 0.0310, 0.0261, 0.0250, 0.0270], device='cuda:6'), out_proj_covar=tensor([1.1658e-04, 1.2392e-04, 8.8545e-05, 1.1229e-04, 1.2649e-04, 1.0459e-04, 1.0170e-04, 1.0797e-04], device='cuda:6') 2023-04-27 03:10:28,165 INFO [finetune.py:976] (6/7) Epoch 12, batch 3450, loss[loss=0.1789, simple_loss=0.2434, pruned_loss=0.05716, over 4680.00 frames. ], tot_loss[loss=0.1902, simple_loss=0.2583, pruned_loss=0.06102, over 953844.04 frames. ], batch size: 23, lr: 3.64e-03, grad_scale: 16.0 2023-04-27 03:10:55,252 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=66494.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 03:10:58,895 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=66500.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 03:11:01,855 INFO [finetune.py:976] (6/7) Epoch 12, batch 3500, loss[loss=0.1675, simple_loss=0.2274, pruned_loss=0.05376, over 4894.00 frames. ], tot_loss[loss=0.188, simple_loss=0.2552, pruned_loss=0.06034, over 953705.22 frames. ], batch size: 35, lr: 3.64e-03, grad_scale: 16.0 2023-04-27 03:11:09,121 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=66516.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 03:11:12,202 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0983, 1.6831, 1.9525, 2.3649, 2.4697, 1.8600, 1.6740, 2.1112], device='cuda:6'), covar=tensor([0.0748, 0.1084, 0.0652, 0.0445, 0.0481, 0.0823, 0.0727, 0.0532], device='cuda:6'), in_proj_covar=tensor([0.0194, 0.0202, 0.0184, 0.0174, 0.0179, 0.0187, 0.0157, 0.0182], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 03:11:12,214 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=66521.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 03:11:26,633 INFO [optim.py:369] (6/7) 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 03:11:27,202 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=66542.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 03:11:40,349 INFO [finetune.py:976] (6/7) Epoch 12, batch 3550, loss[loss=0.1752, simple_loss=0.2441, pruned_loss=0.05313, over 4841.00 frames. ], tot_loss[loss=0.1867, simple_loss=0.2527, pruned_loss=0.06036, over 952683.72 frames. ], batch size: 47, lr: 3.64e-03, grad_scale: 16.0 2023-04-27 03:12:13,289 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=66582.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 03:12:22,657 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.0873, 1.4730, 1.3053, 1.6531, 1.5212, 1.6849, 1.3425, 2.9897], device='cuda:6'), covar=tensor([0.0710, 0.0735, 0.0788, 0.1171, 0.0624, 0.0580, 0.0725, 0.0196], device='cuda:6'), in_proj_covar=tensor([0.0038, 0.0039, 0.0040, 0.0044, 0.0040, 0.0038, 0.0039, 0.0058], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:6') 2023-04-27 03:12:45,654 INFO [finetune.py:976] (6/7) Epoch 12, batch 3600, loss[loss=0.2883, simple_loss=0.3233, pruned_loss=0.1266, over 4156.00 frames. ], tot_loss[loss=0.1849, simple_loss=0.2506, pruned_loss=0.05961, over 953037.85 frames. ], batch size: 65, lr: 3.64e-03, grad_scale: 16.0 2023-04-27 03:12:57,587 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=66614.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 03:13:16,320 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-04-27 03:13:31,411 INFO [optim.py:369] (6/7) 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 03:13:45,503 INFO [finetune.py:976] (6/7) Epoch 12, batch 3650, loss[loss=0.1722, simple_loss=0.2438, pruned_loss=0.05029, over 4868.00 frames. ], tot_loss[loss=0.1865, simple_loss=0.2525, pruned_loss=0.06025, over 953540.47 frames. ], batch size: 31, lr: 3.64e-03, grad_scale: 16.0 2023-04-27 03:13:57,389 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.75 vs. limit=5.0 2023-04-27 03:13:59,537 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9901, 1.7676, 2.0065, 2.3919, 2.4273, 1.8132, 1.6474, 2.1743], device='cuda:6'), covar=tensor([0.0838, 0.1050, 0.0712, 0.0540, 0.0544, 0.0916, 0.0777, 0.0532], device='cuda:6'), in_proj_covar=tensor([0.0193, 0.0202, 0.0184, 0.0173, 0.0179, 0.0187, 0.0157, 0.0181], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 03:14:15,826 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.1022, 2.5647, 2.1789, 2.4109, 1.7605, 2.1788, 2.1494, 1.7043], device='cuda:6'), covar=tensor([0.2123, 0.1081, 0.0752, 0.1204, 0.3169, 0.0966, 0.1946, 0.2595], device='cuda:6'), in_proj_covar=tensor([0.0291, 0.0311, 0.0223, 0.0282, 0.0310, 0.0261, 0.0251, 0.0270], device='cuda:6'), out_proj_covar=tensor([1.1722e-04, 1.2417e-04, 8.9115e-05, 1.1257e-04, 1.2670e-04, 1.0465e-04, 1.0222e-04, 1.0788e-04], device='cuda:6') 2023-04-27 03:14:19,349 INFO [finetune.py:976] (6/7) Epoch 12, batch 3700, loss[loss=0.1749, simple_loss=0.2455, pruned_loss=0.05213, over 4769.00 frames. ], tot_loss[loss=0.1887, simple_loss=0.2558, pruned_loss=0.06084, over 953990.44 frames. ], batch size: 28, lr: 3.64e-03, grad_scale: 16.0 2023-04-27 03:14:35,822 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=66731.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 03:14:43,337 INFO [optim.py:369] (6/7) 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 03:14:46,878 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6038, 1.7580, 0.8576, 1.2947, 1.9184, 1.4782, 1.3661, 1.4325], device='cuda:6'), covar=tensor([0.0522, 0.0372, 0.0357, 0.0565, 0.0261, 0.0514, 0.0516, 0.0583], device='cuda:6'), in_proj_covar=tensor([0.0029, 0.0025, 0.0023, 0.0030, 0.0020, 0.0029, 0.0029, 0.0029], device='cuda:6'), out_proj_covar=tensor([0.0050, 0.0045, 0.0038, 0.0050, 0.0038, 0.0049, 0.0049, 0.0050], device='cuda:6') 2023-04-27 03:14:52,769 INFO [finetune.py:976] (6/7) Epoch 12, batch 3750, loss[loss=0.2127, simple_loss=0.2787, pruned_loss=0.0733, over 4906.00 frames. ], tot_loss[loss=0.1903, simple_loss=0.2575, pruned_loss=0.06155, over 951235.72 frames. ], batch size: 46, lr: 3.64e-03, grad_scale: 16.0 2023-04-27 03:15:08,550 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.3015, 2.9798, 2.5344, 2.6158, 2.1942, 2.5645, 2.6338, 2.0403], device='cuda:6'), covar=tensor([0.2348, 0.1219, 0.0738, 0.1306, 0.2884, 0.1029, 0.2034, 0.2545], device='cuda:6'), in_proj_covar=tensor([0.0290, 0.0310, 0.0222, 0.0281, 0.0310, 0.0260, 0.0251, 0.0269], device='cuda:6'), out_proj_covar=tensor([1.1672e-04, 1.2388e-04, 8.8685e-05, 1.1234e-04, 1.2643e-04, 1.0434e-04, 1.0194e-04, 1.0732e-04], device='cuda:6') 2023-04-27 03:15:16,256 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=66792.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 03:15:22,029 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=66800.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 03:15:25,458 INFO [finetune.py:976] (6/7) Epoch 12, batch 3800, loss[loss=0.1836, simple_loss=0.2551, pruned_loss=0.05598, over 4857.00 frames. ], tot_loss[loss=0.1919, simple_loss=0.2591, pruned_loss=0.06236, over 951589.56 frames. ], batch size: 44, lr: 3.64e-03, grad_scale: 16.0 2023-04-27 03:15:31,991 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.64 vs. limit=5.0 2023-04-27 03:15:32,299 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=66816.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 03:15:38,363 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5717, 1.3876, 1.7872, 1.7401, 1.3379, 1.2548, 1.4061, 0.8653], device='cuda:6'), covar=tensor([0.0539, 0.0738, 0.0414, 0.0589, 0.0793, 0.1285, 0.0668, 0.0774], device='cuda:6'), in_proj_covar=tensor([0.0067, 0.0071, 0.0071, 0.0067, 0.0075, 0.0097, 0.0076, 0.0071], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-27 03:15:48,321 INFO [optim.py:369] (6/7) 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 03:15:52,969 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 12, batch 3850, loss[loss=0.2311, simple_loss=0.2892, pruned_loss=0.08653, over 4830.00 frames. ], tot_loss[loss=0.19, simple_loss=0.2574, pruned_loss=0.06126, over 952672.39 frames. ], batch size: 30, lr: 3.64e-03, grad_scale: 16.0 2023-04-27 03:16:04,614 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=66864.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 03:16:12,476 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=66877.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 03:16:23,346 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8269, 2.4508, 1.8291, 1.7708, 1.3445, 1.3542, 1.9178, 1.2216], device='cuda:6'), covar=tensor([0.1841, 0.1446, 0.1675, 0.1858, 0.2641, 0.2202, 0.1151, 0.2333], device='cuda:6'), in_proj_covar=tensor([0.0196, 0.0215, 0.0171, 0.0204, 0.0204, 0.0185, 0.0158, 0.0186], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:6') 2023-04-27 03:16:31,352 INFO [finetune.py:976] (6/7) Epoch 12, batch 3900, loss[loss=0.1461, simple_loss=0.2278, pruned_loss=0.03216, over 4763.00 frames. ], tot_loss[loss=0.1871, simple_loss=0.2543, pruned_loss=0.05997, over 953728.86 frames. ], batch size: 28, lr: 3.64e-03, grad_scale: 16.0 2023-04-27 03:16:38,404 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=66914.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 03:17:12,164 INFO [optim.py:369] (6/7) 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 03:17:22,161 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 2023-04-27 03:17:32,058 INFO [finetune.py:976] (6/7) Epoch 12, batch 3950, loss[loss=0.1349, simple_loss=0.2042, pruned_loss=0.03274, over 4832.00 frames. ], tot_loss[loss=0.1825, simple_loss=0.2492, pruned_loss=0.05785, over 954353.20 frames. ], batch size: 25, lr: 3.64e-03, grad_scale: 16.0 2023-04-27 03:17:33,391 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=66957.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 03:17:42,930 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=66962.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 03:18:04,835 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.1961, 2.0214, 2.4822, 2.6456, 1.8900, 1.8401, 2.1413, 1.2166], device='cuda:6'), covar=tensor([0.0500, 0.0908, 0.0471, 0.0747, 0.0886, 0.1153, 0.0829, 0.0874], device='cuda:6'), in_proj_covar=tensor([0.0067, 0.0071, 0.0070, 0.0067, 0.0075, 0.0096, 0.0076, 0.0071], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-27 03:18:39,569 INFO [finetune.py:976] (6/7) Epoch 12, batch 4000, loss[loss=0.1768, simple_loss=0.244, pruned_loss=0.05484, over 4924.00 frames. ], tot_loss[loss=0.1819, simple_loss=0.2487, pruned_loss=0.05756, over 954884.98 frames. ], batch size: 37, lr: 3.63e-03, grad_scale: 16.0 2023-04-27 03:18:58,933 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=67018.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 03:19:14,013 INFO [optim.py:369] (6/7) 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 03:19:22,923 INFO [finetune.py:976] (6/7) Epoch 12, batch 4050, loss[loss=0.2037, simple_loss=0.2781, pruned_loss=0.06466, over 4800.00 frames. ], tot_loss[loss=0.1852, simple_loss=0.2529, pruned_loss=0.05872, over 954278.72 frames. ], batch size: 45, lr: 3.63e-03, grad_scale: 16.0 2023-04-27 03:19:42,059 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=67083.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 03:19:44,509 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=67087.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 03:19:46,353 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.2224, 1.4958, 1.3696, 1.6701, 1.5929, 1.7801, 1.4377, 3.4213], device='cuda:6'), covar=tensor([0.0626, 0.0809, 0.0803, 0.1221, 0.0655, 0.0591, 0.0739, 0.0151], device='cuda:6'), in_proj_covar=tensor([0.0038, 0.0039, 0.0040, 0.0043, 0.0040, 0.0038, 0.0038, 0.0057], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0014, 0.0013, 0.0012, 0.0013, 0.0016], device='cuda:6') 2023-04-27 03:19:56,228 INFO [finetune.py:976] (6/7) Epoch 12, batch 4100, loss[loss=0.2438, simple_loss=0.3091, pruned_loss=0.0892, over 4831.00 frames. ], tot_loss[loss=0.1896, simple_loss=0.2573, pruned_loss=0.06097, over 956343.02 frames. ], batch size: 49, lr: 3.63e-03, grad_scale: 16.0 2023-04-27 03:19:58,996 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-04-27 03:20:05,103 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.3027, 1.4021, 1.2577, 1.6239, 1.5060, 1.8416, 1.3340, 3.3626], device='cuda:6'), covar=tensor([0.0678, 0.0957, 0.0942, 0.1370, 0.0751, 0.0662, 0.0902, 0.0201], device='cuda:6'), in_proj_covar=tensor([0.0038, 0.0039, 0.0040, 0.0043, 0.0040, 0.0038, 0.0038, 0.0057], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0014, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:6') 2023-04-27 03:20:17,038 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.4931, 1.3294, 1.5967, 1.7943, 1.3441, 1.0567, 1.3944, 0.8694], device='cuda:6'), covar=tensor([0.0722, 0.0655, 0.0591, 0.0622, 0.0804, 0.1743, 0.0822, 0.0972], device='cuda:6'), in_proj_covar=tensor([0.0068, 0.0071, 0.0071, 0.0067, 0.0076, 0.0097, 0.0076, 0.0071], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-27 03:20:21,139 INFO [optim.py:369] (6/7) 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 03:20:21,265 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.1025, 1.6993, 2.0414, 2.4556, 2.5514, 1.9715, 1.7875, 2.1248], device='cuda:6'), covar=tensor([0.0794, 0.1084, 0.0649, 0.0474, 0.0494, 0.0836, 0.0749, 0.0493], device='cuda:6'), in_proj_covar=tensor([0.0192, 0.0200, 0.0182, 0.0173, 0.0177, 0.0185, 0.0155, 0.0179], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 03:20:22,511 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=67144.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 03:20:29,534 INFO [finetune.py:976] (6/7) Epoch 12, batch 4150, loss[loss=0.1811, simple_loss=0.2474, pruned_loss=0.05743, over 4225.00 frames. ], tot_loss[loss=0.1899, simple_loss=0.2578, pruned_loss=0.06099, over 955242.76 frames. ], batch size: 65, lr: 3.63e-03, grad_scale: 16.0 2023-04-27 03:20:37,308 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5420, 1.4093, 0.5782, 1.2108, 1.4581, 1.3881, 1.2715, 1.2883], device='cuda:6'), covar=tensor([0.0532, 0.0381, 0.0419, 0.0581, 0.0328, 0.0564, 0.0521, 0.0622], device='cuda:6'), in_proj_covar=tensor([0.0029, 0.0025, 0.0023, 0.0029, 0.0020, 0.0029, 0.0029, 0.0029], device='cuda:6'), out_proj_covar=tensor([0.0050, 0.0045, 0.0038, 0.0050, 0.0038, 0.0049, 0.0048, 0.0050], device='cuda:6') 2023-04-27 03:20:45,902 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=67177.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 03:21:03,552 INFO [finetune.py:976] (6/7) Epoch 12, batch 4200, loss[loss=0.2027, simple_loss=0.2763, pruned_loss=0.06449, over 4866.00 frames. ], tot_loss[loss=0.1903, simple_loss=0.2589, pruned_loss=0.06084, over 956861.49 frames. ], batch size: 34, lr: 3.63e-03, grad_scale: 16.0 2023-04-27 03:21:18,108 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=67225.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 03:21:28,836 INFO [optim.py:369] (6/7) 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] (6/7) Epoch 12, batch 4250, loss[loss=0.1512, simple_loss=0.2239, pruned_loss=0.03927, over 4798.00 frames. ], tot_loss[loss=0.1874, simple_loss=0.2558, pruned_loss=0.05951, over 957470.75 frames. ], batch size: 51, lr: 3.63e-03, grad_scale: 16.0 2023-04-27 03:22:10,189 INFO [finetune.py:976] (6/7) Epoch 12, batch 4300, loss[loss=0.1577, simple_loss=0.2347, pruned_loss=0.04033, over 4902.00 frames. ], tot_loss[loss=0.1851, simple_loss=0.2527, pruned_loss=0.05876, over 956722.35 frames. ], batch size: 35, lr: 3.63e-03, grad_scale: 16.0 2023-04-27 03:22:15,636 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=67313.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 03:22:36,120 INFO [optim.py:369] (6/7) 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 03:22:43,958 INFO [finetune.py:976] (6/7) Epoch 12, batch 4350, loss[loss=0.1639, simple_loss=0.2303, pruned_loss=0.04873, over 4793.00 frames. ], tot_loss[loss=0.1825, simple_loss=0.2494, pruned_loss=0.05778, over 955744.13 frames. ], batch size: 29, lr: 3.63e-03, grad_scale: 16.0 2023-04-27 03:23:29,029 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=67387.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 03:23:51,844 INFO [finetune.py:976] (6/7) Epoch 12, batch 4400, loss[loss=0.2732, simple_loss=0.3168, pruned_loss=0.1148, over 4915.00 frames. ], tot_loss[loss=0.1833, simple_loss=0.2501, pruned_loss=0.0582, over 956873.91 frames. ], batch size: 36, lr: 3.63e-03, grad_scale: 16.0 2023-04-27 03:24:15,511 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6572, 1.2808, 1.2797, 1.5314, 1.8510, 1.5700, 1.3711, 1.2090], device='cuda:6'), covar=tensor([0.1670, 0.1528, 0.1575, 0.1440, 0.0904, 0.1645, 0.1914, 0.1966], device='cuda:6'), in_proj_covar=tensor([0.0307, 0.0318, 0.0352, 0.0295, 0.0332, 0.0315, 0.0305, 0.0361], device='cuda:6'), out_proj_covar=tensor([6.4040e-05, 6.6979e-05, 7.5505e-05, 6.0679e-05, 6.9404e-05, 6.6934e-05, 6.4883e-05, 7.7344e-05], device='cuda:6') 2023-04-27 03:24:26,733 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=67435.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 03:24:34,558 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=67439.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 03:24:36,307 INFO [optim.py:369] (6/7) 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 03:24:56,594 INFO [finetune.py:976] (6/7) Epoch 12, batch 4450, loss[loss=0.215, simple_loss=0.2803, pruned_loss=0.07484, over 4841.00 frames. ], tot_loss[loss=0.1878, simple_loss=0.255, pruned_loss=0.06034, over 956664.28 frames. ], batch size: 49, lr: 3.63e-03, grad_scale: 16.0 2023-04-27 03:25:33,504 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=3.64 vs. limit=5.0 2023-04-27 03:25:48,054 INFO [finetune.py:976] (6/7) Epoch 12, batch 4500, loss[loss=0.1659, simple_loss=0.2411, pruned_loss=0.04532, over 4760.00 frames. ], tot_loss[loss=0.1887, simple_loss=0.2561, pruned_loss=0.06072, over 955867.61 frames. ], batch size: 27, lr: 3.63e-03, grad_scale: 16.0 2023-04-27 03:25:53,084 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=67513.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 03:26:12,842 INFO [optim.py:369] (6/7) 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 03:26:21,715 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([0.9638, 1.1585, 3.1152, 2.8657, 2.7973, 3.0349, 3.0530, 2.7742], device='cuda:6'), covar=tensor([0.6729, 0.5156, 0.1468, 0.2102, 0.1488, 0.1991, 0.1119, 0.1508], device='cuda:6'), in_proj_covar=tensor([0.0311, 0.0307, 0.0403, 0.0412, 0.0351, 0.0410, 0.0314, 0.0370], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-27 03:26:22,245 INFO [finetune.py:976] (6/7) Epoch 12, batch 4550, loss[loss=0.1834, simple_loss=0.2361, pruned_loss=0.06532, over 4406.00 frames. ], tot_loss[loss=0.1897, simple_loss=0.2577, pruned_loss=0.06086, over 954434.16 frames. ], batch size: 19, lr: 3.63e-03, grad_scale: 32.0 2023-04-27 03:26:28,002 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.27 vs. limit=5.0 2023-04-27 03:26:34,573 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=67574.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 03:26:51,833 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-04-27 03:26:56,263 INFO [finetune.py:976] (6/7) Epoch 12, batch 4600, loss[loss=0.1713, simple_loss=0.2431, pruned_loss=0.04973, over 4809.00 frames. ], tot_loss[loss=0.189, simple_loss=0.2572, pruned_loss=0.06035, over 953626.49 frames. ], batch size: 41, lr: 3.63e-03, grad_scale: 32.0 2023-04-27 03:27:01,312 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=67613.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 03:27:15,115 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8749, 1.4060, 1.4460, 1.6229, 2.0310, 1.6500, 1.3709, 1.3713], device='cuda:6'), covar=tensor([0.1500, 0.1399, 0.2027, 0.1213, 0.0770, 0.1512, 0.2032, 0.2032], device='cuda:6'), in_proj_covar=tensor([0.0306, 0.0317, 0.0352, 0.0295, 0.0333, 0.0315, 0.0304, 0.0361], device='cuda:6'), out_proj_covar=tensor([6.3849e-05, 6.6771e-05, 7.5619e-05, 6.0570e-05, 6.9552e-05, 6.6873e-05, 6.4821e-05, 7.7292e-05], device='cuda:6') 2023-04-27 03:27:20,973 INFO [optim.py:369] (6/7) 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 03:27:29,856 INFO [finetune.py:976] (6/7) Epoch 12, batch 4650, loss[loss=0.1547, simple_loss=0.2331, pruned_loss=0.0382, over 4765.00 frames. ], tot_loss[loss=0.1866, simple_loss=0.2544, pruned_loss=0.05942, over 954732.88 frames. ], batch size: 28, lr: 3.63e-03, grad_scale: 16.0 2023-04-27 03:27:34,155 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=67661.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 03:27:37,156 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-04-27 03:27:46,419 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=67680.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 03:27:54,878 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 12, batch 4700, loss[loss=0.1722, simple_loss=0.2406, pruned_loss=0.05183, over 4771.00 frames. ], tot_loss[loss=0.1844, simple_loss=0.2517, pruned_loss=0.05853, over 956260.63 frames. ], batch size: 26, lr: 3.63e-03, grad_scale: 16.0 2023-04-27 03:28:26,054 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.4326, 1.5421, 5.4138, 5.1292, 4.8091, 5.1753, 4.7034, 4.9268], device='cuda:6'), covar=tensor([0.6225, 0.6297, 0.0977, 0.1742, 0.1035, 0.1525, 0.1418, 0.1394], device='cuda:6'), in_proj_covar=tensor([0.0312, 0.0307, 0.0403, 0.0412, 0.0350, 0.0409, 0.0315, 0.0369], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-27 03:28:37,301 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=67739.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 03:28:38,543 INFO [zipformer.py:1188] (6/7) 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] (6/7) 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 03:28:58,096 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=67753.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 03:28:58,928 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.48 vs. limit=2.0 2023-04-27 03:28:59,231 INFO [finetune.py:976] (6/7) Epoch 12, batch 4750, loss[loss=0.179, simple_loss=0.2322, pruned_loss=0.0629, over 4255.00 frames. ], tot_loss[loss=0.1825, simple_loss=0.2494, pruned_loss=0.05784, over 957295.59 frames. ], batch size: 18, lr: 3.63e-03, grad_scale: 16.0 2023-04-27 03:29:11,203 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6309, 1.1334, 1.7166, 2.1564, 1.7569, 1.6339, 1.6478, 1.6801], device='cuda:6'), covar=tensor([0.5012, 0.7174, 0.6540, 0.6231, 0.6107, 0.8282, 0.8437, 0.8200], device='cuda:6'), in_proj_covar=tensor([0.0411, 0.0406, 0.0496, 0.0514, 0.0441, 0.0462, 0.0469, 0.0471], device='cuda:6'), out_proj_covar=tensor([9.9664e-05, 1.0074e-04, 1.1173e-04, 1.2194e-04, 1.0679e-04, 1.1151e-04, 1.1234e-04, 1.1262e-04], device='cuda:6') 2023-04-27 03:29:23,488 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-04-27 03:29:42,911 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=67787.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 03:30:04,061 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.1703, 1.3744, 1.6329, 1.7579, 1.5665, 1.6286, 1.7214, 1.6663], device='cuda:6'), covar=tensor([0.4747, 0.6667, 0.5094, 0.4933, 0.6161, 0.8908, 0.5690, 0.5361], device='cuda:6'), in_proj_covar=tensor([0.0328, 0.0376, 0.0314, 0.0326, 0.0339, 0.0398, 0.0357, 0.0324], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 03:30:06,234 INFO [finetune.py:976] (6/7) Epoch 12, batch 4800, loss[loss=0.2017, simple_loss=0.2728, pruned_loss=0.06531, over 4904.00 frames. ], tot_loss[loss=0.1867, simple_loss=0.2535, pruned_loss=0.0599, over 957033.93 frames. ], batch size: 35, lr: 3.63e-03, grad_scale: 16.0 2023-04-27 03:30:23,025 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.3030, 2.7685, 1.0950, 1.5814, 2.2314, 1.4344, 3.8251, 1.9292], device='cuda:6'), covar=tensor([0.0639, 0.0642, 0.0743, 0.1235, 0.0455, 0.0907, 0.0385, 0.0606], device='cuda:6'), in_proj_covar=tensor([0.0051, 0.0066, 0.0048, 0.0047, 0.0050, 0.0052, 0.0076, 0.0051], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0007], device='cuda:6') 2023-04-27 03:30:35,556 INFO [optim.py:369] (6/7) 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] (6/7) Epoch 12, batch 4850, loss[loss=0.1862, simple_loss=0.2641, pruned_loss=0.05415, over 4776.00 frames. ], tot_loss[loss=0.1884, simple_loss=0.2561, pruned_loss=0.06035, over 955882.42 frames. ], batch size: 28, lr: 3.63e-03, grad_scale: 16.0 2023-04-27 03:30:54,465 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=67869.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 03:31:05,502 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8630, 2.5975, 1.8827, 1.7516, 1.3179, 1.3630, 2.0118, 1.2618], device='cuda:6'), covar=tensor([0.1862, 0.1310, 0.1599, 0.2008, 0.2596, 0.2216, 0.1110, 0.2259], device='cuda:6'), in_proj_covar=tensor([0.0196, 0.0214, 0.0170, 0.0204, 0.0203, 0.0185, 0.0157, 0.0185], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:6') 2023-04-27 03:31:17,822 INFO [finetune.py:976] (6/7) Epoch 12, batch 4900, loss[loss=0.2074, simple_loss=0.26, pruned_loss=0.07735, over 4838.00 frames. ], tot_loss[loss=0.1899, simple_loss=0.2581, pruned_loss=0.0609, over 955692.47 frames. ], batch size: 31, lr: 3.63e-03, grad_scale: 16.0 2023-04-27 03:31:43,017 INFO [optim.py:369] (6/7) 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 03:31:51,277 INFO [finetune.py:976] (6/7) Epoch 12, batch 4950, loss[loss=0.1693, simple_loss=0.224, pruned_loss=0.05733, over 3907.00 frames. ], tot_loss[loss=0.1898, simple_loss=0.258, pruned_loss=0.06078, over 953929.66 frames. ], batch size: 17, lr: 3.63e-03, grad_scale: 16.0 2023-04-27 03:32:01,243 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.2338, 1.8930, 2.2839, 2.6009, 1.8894, 1.7464, 1.9782, 1.0149], device='cuda:6'), covar=tensor([0.0565, 0.0895, 0.0579, 0.0692, 0.0912, 0.1288, 0.1026, 0.1035], device='cuda:6'), in_proj_covar=tensor([0.0067, 0.0071, 0.0070, 0.0067, 0.0075, 0.0096, 0.0075, 0.0070], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-27 03:32:09,668 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=67981.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 03:32:26,420 INFO [finetune.py:976] (6/7) Epoch 12, batch 5000, loss[loss=0.1895, simple_loss=0.2513, pruned_loss=0.06387, over 4823.00 frames. ], tot_loss[loss=0.1879, simple_loss=0.2558, pruned_loss=0.06, over 955594.94 frames. ], batch size: 33, lr: 3.63e-03, grad_scale: 16.0 2023-04-27 03:32:36,349 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.3425, 1.2351, 1.6191, 1.4839, 1.2142, 1.1155, 1.2798, 0.9524], device='cuda:6'), covar=tensor([0.0650, 0.0792, 0.0454, 0.0734, 0.0837, 0.1181, 0.0667, 0.0650], device='cuda:6'), in_proj_covar=tensor([0.0067, 0.0070, 0.0070, 0.0067, 0.0075, 0.0096, 0.0075, 0.0070], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-27 03:32:47,898 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=68036.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 03:32:51,648 INFO [zipformer.py:1188] (6/7) 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] (6/7) 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 03:32:55,705 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=68048.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 03:32:59,806 INFO [finetune.py:976] (6/7) Epoch 12, batch 5050, loss[loss=0.1869, simple_loss=0.25, pruned_loss=0.06193, over 4905.00 frames. ], tot_loss[loss=0.1861, simple_loss=0.2531, pruned_loss=0.05954, over 955179.13 frames. ], batch size: 32, lr: 3.63e-03, grad_scale: 16.0 2023-04-27 03:33:22,293 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=68087.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 03:33:23,466 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5154, 1.3722, 4.3342, 4.0425, 3.8251, 4.1477, 4.0499, 3.8004], device='cuda:6'), covar=tensor([0.6638, 0.5965, 0.1057, 0.1839, 0.1162, 0.1395, 0.1064, 0.1590], device='cuda:6'), in_proj_covar=tensor([0.0313, 0.0308, 0.0404, 0.0412, 0.0352, 0.0410, 0.0316, 0.0373], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-27 03:33:26,549 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.3487, 2.9255, 2.2366, 2.1297, 1.6111, 1.5995, 2.3922, 1.5966], device='cuda:6'), covar=tensor([0.1629, 0.1479, 0.1429, 0.1889, 0.2462, 0.2026, 0.1058, 0.2063], device='cuda:6'), in_proj_covar=tensor([0.0195, 0.0212, 0.0170, 0.0203, 0.0202, 0.0184, 0.0157, 0.0184], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:6') 2023-04-27 03:33:27,720 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([3.2800, 3.2588, 2.5321, 3.7927, 3.2960, 3.2348, 1.3200, 3.0925], device='cuda:6'), covar=tensor([0.2124, 0.1502, 0.3692, 0.2451, 0.3259, 0.2201, 0.6276, 0.3015], device='cuda:6'), in_proj_covar=tensor([0.0243, 0.0215, 0.0248, 0.0303, 0.0298, 0.0247, 0.0271, 0.0267], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 03:33:33,616 INFO [finetune.py:976] (6/7) Epoch 12, batch 5100, loss[loss=0.1468, simple_loss=0.221, pruned_loss=0.03631, over 4869.00 frames. ], tot_loss[loss=0.183, simple_loss=0.2497, pruned_loss=0.05813, over 954024.51 frames. ], batch size: 31, lr: 3.63e-03, grad_scale: 16.0 2023-04-27 03:33:43,241 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([0.9560, 1.9641, 1.7623, 1.5947, 2.1649, 1.7447, 2.6161, 1.4989], device='cuda:6'), covar=tensor([0.4144, 0.2078, 0.4958, 0.3292, 0.1954, 0.2641, 0.1455, 0.4858], device='cuda:6'), in_proj_covar=tensor([0.0341, 0.0345, 0.0427, 0.0355, 0.0378, 0.0379, 0.0373, 0.0417], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 03:34:11,406 INFO [optim.py:369] (6/7) 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 03:34:14,592 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=68148.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 03:34:18,644 INFO [finetune.py:976] (6/7) Epoch 12, batch 5150, loss[loss=0.1917, simple_loss=0.2606, pruned_loss=0.0614, over 4843.00 frames. ], tot_loss[loss=0.1851, simple_loss=0.2513, pruned_loss=0.05948, over 953477.45 frames. ], batch size: 33, lr: 3.63e-03, grad_scale: 16.0 2023-04-27 03:34:34,059 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=68169.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 03:34:44,974 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.2671, 2.9354, 0.9608, 1.5702, 1.6092, 2.2171, 1.7197, 0.8791], device='cuda:6'), covar=tensor([0.1399, 0.1043, 0.1858, 0.1326, 0.1103, 0.0924, 0.1515, 0.1954], device='cuda:6'), in_proj_covar=tensor([0.0117, 0.0248, 0.0139, 0.0121, 0.0133, 0.0153, 0.0117, 0.0120], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-27 03:35:01,107 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.74 vs. limit=5.0 2023-04-27 03:35:04,839 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-04-27 03:35:16,279 INFO [finetune.py:976] (6/7) Epoch 12, batch 5200, loss[loss=0.24, simple_loss=0.3192, pruned_loss=0.08042, over 4841.00 frames. ], tot_loss[loss=0.1888, simple_loss=0.2554, pruned_loss=0.06113, over 952004.82 frames. ], batch size: 51, lr: 3.62e-03, grad_scale: 16.0 2023-04-27 03:35:22,601 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7270, 1.5422, 1.8818, 2.1347, 2.2711, 1.7048, 1.4526, 1.8826], device='cuda:6'), covar=tensor([0.0976, 0.1237, 0.0724, 0.0637, 0.0606, 0.0977, 0.0929, 0.0664], device='cuda:6'), in_proj_covar=tensor([0.0195, 0.0204, 0.0184, 0.0176, 0.0180, 0.0187, 0.0157, 0.0182], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 03:35:34,919 INFO [zipformer.py:1188] (6/7) 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] (6/7) 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 03:36:22,259 INFO [finetune.py:976] (6/7) Epoch 12, batch 5250, loss[loss=0.2228, simple_loss=0.2968, pruned_loss=0.07433, over 4720.00 frames. ], tot_loss[loss=0.1897, simple_loss=0.2573, pruned_loss=0.06099, over 953579.48 frames. ], batch size: 59, lr: 3.62e-03, grad_scale: 16.0 2023-04-27 03:37:07,994 INFO [finetune.py:976] (6/7) Epoch 12, batch 5300, loss[loss=0.2013, simple_loss=0.2657, pruned_loss=0.06846, over 4240.00 frames. ], tot_loss[loss=0.1897, simple_loss=0.258, pruned_loss=0.06071, over 954938.11 frames. ], batch size: 66, lr: 3.62e-03, grad_scale: 16.0 2023-04-27 03:37:29,896 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=68336.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 03:37:30,469 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=68337.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 03:37:34,483 INFO [optim.py:369] (6/7) 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 03:37:37,607 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 12, batch 5350, loss[loss=0.1754, simple_loss=0.2472, pruned_loss=0.0518, over 4700.00 frames. ], tot_loss[loss=0.1881, simple_loss=0.2568, pruned_loss=0.0597, over 954719.90 frames. ], batch size: 59, lr: 3.62e-03, grad_scale: 16.0 2023-04-27 03:38:00,880 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=68384.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 03:38:10,027 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=68396.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 03:38:11,939 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=68399.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 03:38:15,655 INFO [finetune.py:976] (6/7) Epoch 12, batch 5400, loss[loss=0.1717, simple_loss=0.237, pruned_loss=0.05324, over 4805.00 frames. ], tot_loss[loss=0.1875, simple_loss=0.2555, pruned_loss=0.05976, over 953330.04 frames. ], batch size: 45, lr: 3.62e-03, grad_scale: 16.0 2023-04-27 03:38:22,482 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=3.39 vs. limit=5.0 2023-04-27 03:38:41,372 INFO [optim.py:369] (6/7) 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 03:38:41,454 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=68443.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 03:38:49,104 INFO [finetune.py:976] (6/7) Epoch 12, batch 5450, loss[loss=0.1519, simple_loss=0.2228, pruned_loss=0.04056, over 4824.00 frames. ], tot_loss[loss=0.1847, simple_loss=0.2519, pruned_loss=0.0587, over 954341.22 frames. ], batch size: 38, lr: 3.62e-03, grad_scale: 16.0 2023-04-27 03:38:52,242 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=68460.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 03:39:26,613 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.2439, 1.6458, 1.4909, 1.7793, 1.7135, 1.8310, 1.4473, 3.3789], device='cuda:6'), covar=tensor([0.0622, 0.0753, 0.0756, 0.1152, 0.0577, 0.0525, 0.0736, 0.0154], device='cuda:6'), in_proj_covar=tensor([0.0038, 0.0039, 0.0040, 0.0044, 0.0040, 0.0038, 0.0038, 0.0057], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:6') 2023-04-27 03:39:33,461 INFO [finetune.py:976] (6/7) Epoch 12, batch 5500, loss[loss=0.1716, simple_loss=0.2467, pruned_loss=0.04828, over 4822.00 frames. ], tot_loss[loss=0.1817, simple_loss=0.2489, pruned_loss=0.05727, over 956268.75 frames. ], batch size: 51, lr: 3.62e-03, grad_scale: 16.0 2023-04-27 03:39:36,769 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 2023-04-27 03:39:58,193 INFO [optim.py:369] (6/7) 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] (6/7) Epoch 12, batch 5550, loss[loss=0.1982, simple_loss=0.2718, pruned_loss=0.0623, over 4834.00 frames. ], tot_loss[loss=0.1826, simple_loss=0.2496, pruned_loss=0.05782, over 953989.42 frames. ], batch size: 49, lr: 3.62e-03, grad_scale: 16.0 2023-04-27 03:40:21,097 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=68578.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 03:40:54,928 INFO [finetune.py:976] (6/7) Epoch 12, batch 5600, loss[loss=0.1469, simple_loss=0.2214, pruned_loss=0.03615, over 4813.00 frames. ], tot_loss[loss=0.1853, simple_loss=0.2533, pruned_loss=0.05869, over 953759.42 frames. ], batch size: 25, lr: 3.62e-03, grad_scale: 16.0 2023-04-27 03:41:02,580 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2023-04-27 03:41:24,531 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.3049, 1.1633, 1.5852, 1.4981, 1.2079, 1.1383, 1.2456, 0.8658], device='cuda:6'), covar=tensor([0.0529, 0.0641, 0.0394, 0.0567, 0.0769, 0.1119, 0.0484, 0.0642], device='cuda:6'), in_proj_covar=tensor([0.0068, 0.0071, 0.0071, 0.0067, 0.0075, 0.0096, 0.0075, 0.0070], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-27 03:41:36,479 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=68637.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 03:41:37,659 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=68639.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 03:41:46,137 INFO [optim.py:369] (6/7) 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 03:41:58,627 INFO [finetune.py:976] (6/7) Epoch 12, batch 5650, loss[loss=0.2073, simple_loss=0.2871, pruned_loss=0.06378, over 4909.00 frames. ], tot_loss[loss=0.1882, simple_loss=0.2569, pruned_loss=0.05971, over 954456.59 frames. ], batch size: 43, lr: 3.62e-03, grad_scale: 16.0 2023-04-27 03:42:33,437 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=68685.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 03:42:33,498 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=68685.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 03:42:45,294 INFO [finetune.py:976] (6/7) Epoch 12, batch 5700, loss[loss=0.1861, simple_loss=0.2485, pruned_loss=0.0619, over 4306.00 frames. ], tot_loss[loss=0.1859, simple_loss=0.2537, pruned_loss=0.0591, over 938607.27 frames. ], batch size: 18, lr: 3.62e-03, grad_scale: 16.0 2023-04-27 03:42:54,862 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([3.0774, 2.1680, 2.1322, 2.9157, 2.9748, 2.6087, 2.5977, 2.3092], device='cuda:6'), covar=tensor([0.1566, 0.1524, 0.1788, 0.1516, 0.1051, 0.1348, 0.1861, 0.1742], device='cuda:6'), in_proj_covar=tensor([0.0309, 0.0319, 0.0353, 0.0296, 0.0333, 0.0317, 0.0309, 0.0362], device='cuda:6'), out_proj_covar=tensor([6.4438e-05, 6.7198e-05, 7.5791e-05, 6.0680e-05, 6.9643e-05, 6.7296e-05, 6.5737e-05, 7.7424e-05], device='cuda:6') 2023-04-27 03:42:58,337 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.9853, 2.7808, 3.3097, 3.4954, 3.3074, 3.0500, 2.4232, 3.1471], device='cuda:6'), covar=tensor([0.0855, 0.0881, 0.0489, 0.0556, 0.0562, 0.0787, 0.0823, 0.0533], device='cuda:6'), in_proj_covar=tensor([0.0192, 0.0201, 0.0182, 0.0173, 0.0178, 0.0184, 0.0155, 0.0179], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 03:43:16,427 INFO [finetune.py:976] (6/7) Epoch 13, batch 0, loss[loss=0.1752, simple_loss=0.2448, pruned_loss=0.05277, over 4835.00 frames. ], tot_loss[loss=0.1752, simple_loss=0.2448, pruned_loss=0.05277, over 4835.00 frames. ], batch size: 49, lr: 3.62e-03, grad_scale: 16.0 2023-04-27 03:43:16,427 INFO [finetune.py:1001] (6/7) Computing validation loss 2023-04-27 03:43:19,227 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([3.4029, 3.3567, 2.5801, 3.8745, 3.5097, 3.4302, 1.6417, 3.3880], device='cuda:6'), covar=tensor([0.1405, 0.1519, 0.2611, 0.2076, 0.2747, 0.1897, 0.4990, 0.2372], device='cuda:6'), in_proj_covar=tensor([0.0244, 0.0216, 0.0248, 0.0304, 0.0298, 0.0248, 0.0271, 0.0270], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 03:43:32,175 INFO [finetune.py:1010] (6/7) 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] (6/7) Maximum memory allocated so far is 6345MB 2023-04-27 03:43:49,890 INFO [optim.py:369] (6/7) 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 03:43:49,989 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=68743.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 03:43:51,894 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=68746.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 03:43:57,268 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=68755.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 03:43:59,154 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6139, 1.4026, 0.7935, 1.3005, 1.7873, 1.4794, 1.3786, 1.3499], device='cuda:6'), covar=tensor([0.0508, 0.0393, 0.0344, 0.0587, 0.0280, 0.0532, 0.0517, 0.0601], device='cuda:6'), in_proj_covar=tensor([0.0029, 0.0025, 0.0023, 0.0029, 0.0020, 0.0029, 0.0028, 0.0029], device='cuda:6'), out_proj_covar=tensor([0.0049, 0.0044, 0.0037, 0.0050, 0.0037, 0.0048, 0.0048, 0.0050], device='cuda:6') 2023-04-27 03:44:15,627 INFO [finetune.py:976] (6/7) Epoch 13, batch 50, loss[loss=0.1327, simple_loss=0.2152, pruned_loss=0.02507, over 4782.00 frames. ], tot_loss[loss=0.1941, simple_loss=0.2608, pruned_loss=0.06364, over 215400.49 frames. ], batch size: 29, lr: 3.62e-03, grad_scale: 16.0 2023-04-27 03:44:21,434 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=68791.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 03:44:24,212 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.63 vs. limit=5.0 2023-04-27 03:44:34,707 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-04-27 03:44:47,502 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([3.4249, 3.2349, 2.5010, 3.8831, 3.3396, 3.4071, 1.4275, 3.3202], device='cuda:6'), covar=tensor([0.1736, 0.1401, 0.3023, 0.2192, 0.2840, 0.1701, 0.5523, 0.2488], device='cuda:6'), in_proj_covar=tensor([0.0244, 0.0216, 0.0248, 0.0303, 0.0298, 0.0248, 0.0271, 0.0270], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 03:44:48,040 INFO [finetune.py:976] (6/7) Epoch 13, batch 100, loss[loss=0.2068, simple_loss=0.2688, pruned_loss=0.07236, over 4820.00 frames. ], tot_loss[loss=0.1877, simple_loss=0.2543, pruned_loss=0.06059, over 379050.39 frames. ], batch size: 38, lr: 3.62e-03, grad_scale: 16.0 2023-04-27 03:44:55,551 INFO [optim.py:369] (6/7) 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 03:44:57,543 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.2454, 1.6070, 1.3520, 1.4952, 1.3242, 1.2013, 1.4349, 1.0398], device='cuda:6'), covar=tensor([0.1441, 0.1122, 0.0852, 0.1015, 0.3047, 0.0995, 0.1395, 0.1886], device='cuda:6'), in_proj_covar=tensor([0.0291, 0.0309, 0.0225, 0.0282, 0.0311, 0.0262, 0.0253, 0.0271], device='cuda:6'), out_proj_covar=tensor([1.1726e-04, 1.2344e-04, 8.9668e-05, 1.1268e-04, 1.2716e-04, 1.0482e-04, 1.0273e-04, 1.0830e-04], device='cuda:6') 2023-04-27 03:45:11,652 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5035, 1.4268, 4.0806, 3.8187, 3.5246, 3.8342, 3.8116, 3.5582], device='cuda:6'), covar=tensor([0.6637, 0.5509, 0.1075, 0.1715, 0.1218, 0.1676, 0.1449, 0.1416], device='cuda:6'), in_proj_covar=tensor([0.0309, 0.0301, 0.0398, 0.0405, 0.0346, 0.0404, 0.0310, 0.0366], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-27 03:45:21,011 INFO [finetune.py:976] (6/7) Epoch 13, batch 150, loss[loss=0.1384, simple_loss=0.2067, pruned_loss=0.03499, over 4775.00 frames. ], tot_loss[loss=0.1834, simple_loss=0.2493, pruned_loss=0.05869, over 507468.97 frames. ], batch size: 26, lr: 3.62e-03, grad_scale: 16.0 2023-04-27 03:45:53,958 INFO [finetune.py:976] (6/7) Epoch 13, batch 200, loss[loss=0.2001, simple_loss=0.275, pruned_loss=0.0626, over 4850.00 frames. ], tot_loss[loss=0.1814, simple_loss=0.2472, pruned_loss=0.0578, over 605491.43 frames. ], batch size: 44, lr: 3.62e-03, grad_scale: 16.0 2023-04-27 03:45:54,543 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.1963, 1.5292, 1.3421, 1.9233, 1.6458, 1.8062, 1.4024, 3.4017], device='cuda:6'), covar=tensor([0.0732, 0.0846, 0.0915, 0.1194, 0.0704, 0.0555, 0.0807, 0.0176], device='cuda:6'), in_proj_covar=tensor([0.0039, 0.0039, 0.0041, 0.0044, 0.0040, 0.0038, 0.0039, 0.0058], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:6') 2023-04-27 03:45:55,105 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=68934.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 03:46:01,423 INFO [optim.py:369] (6/7) 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] (6/7) Epoch 13, batch 250, loss[loss=0.2131, simple_loss=0.2929, pruned_loss=0.06666, over 4849.00 frames. ], tot_loss[loss=0.1828, simple_loss=0.2491, pruned_loss=0.05828, over 683875.91 frames. ], batch size: 44, lr: 3.62e-03, grad_scale: 16.0 2023-04-27 03:47:12,558 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.3833, 4.6503, 1.2428, 2.6920, 2.9880, 3.1895, 2.8814, 1.1870], device='cuda:6'), covar=tensor([0.1188, 0.1104, 0.2004, 0.1165, 0.0900, 0.0992, 0.1317, 0.1948], device='cuda:6'), in_proj_covar=tensor([0.0117, 0.0247, 0.0138, 0.0120, 0.0132, 0.0152, 0.0117, 0.0119], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-27 03:47:36,762 INFO [finetune.py:976] (6/7) Epoch 13, batch 300, loss[loss=0.2406, simple_loss=0.3034, pruned_loss=0.08889, over 4870.00 frames. ], tot_loss[loss=0.1882, simple_loss=0.2548, pruned_loss=0.06086, over 744012.31 frames. ], batch size: 34, lr: 3.62e-03, grad_scale: 16.0 2023-04-27 03:47:47,290 INFO [zipformer.py:1188] (6/7) 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] (6/7) 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 03:47:48,559 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=69043.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 03:47:59,635 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6747, 1.6509, 1.9495, 2.0466, 1.5976, 1.3628, 1.7275, 1.0713], device='cuda:6'), covar=tensor([0.0822, 0.0700, 0.0628, 0.0835, 0.0931, 0.1233, 0.0857, 0.0875], device='cuda:6'), in_proj_covar=tensor([0.0068, 0.0071, 0.0071, 0.0067, 0.0075, 0.0096, 0.0075, 0.0070], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-27 03:48:05,319 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=69055.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 03:48:39,072 INFO [finetune.py:976] (6/7) Epoch 13, batch 350, loss[loss=0.1937, simple_loss=0.2559, pruned_loss=0.06579, over 4113.00 frames. ], tot_loss[loss=0.1895, simple_loss=0.2568, pruned_loss=0.06115, over 791709.49 frames. ], batch size: 17, lr: 3.62e-03, grad_scale: 16.0 2023-04-27 03:48:59,549 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=69103.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 03:49:00,173 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=69104.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 03:49:17,727 INFO [finetune.py:976] (6/7) Epoch 13, batch 400, loss[loss=0.23, simple_loss=0.2857, pruned_loss=0.08715, over 4727.00 frames. ], tot_loss[loss=0.189, simple_loss=0.2572, pruned_loss=0.06037, over 828823.05 frames. ], batch size: 54, lr: 3.61e-03, grad_scale: 16.0 2023-04-27 03:49:24,722 INFO [optim.py:369] (6/7) 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 03:49:38,356 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.13 vs. limit=2.0 2023-04-27 03:49:38,761 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=69162.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 03:49:42,425 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=69168.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 03:49:51,422 INFO [finetune.py:976] (6/7) Epoch 13, batch 450, loss[loss=0.229, simple_loss=0.2717, pruned_loss=0.09322, over 4930.00 frames. ], tot_loss[loss=0.1885, simple_loss=0.2566, pruned_loss=0.0602, over 855629.08 frames. ], batch size: 38, lr: 3.61e-03, grad_scale: 16.0 2023-04-27 03:49:57,550 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8900, 2.5379, 1.8210, 1.7225, 1.3299, 1.3727, 1.8316, 1.2146], device='cuda:6'), covar=tensor([0.1692, 0.1212, 0.1471, 0.1810, 0.2312, 0.1894, 0.1032, 0.2072], device='cuda:6'), in_proj_covar=tensor([0.0197, 0.0214, 0.0170, 0.0204, 0.0204, 0.0185, 0.0157, 0.0185], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:6') 2023-04-27 03:49:58,692 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-04-27 03:50:19,120 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=69223.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 03:50:20,344 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=69225.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 03:50:22,768 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 13, batch 500, loss[loss=0.193, simple_loss=0.2579, pruned_loss=0.06405, over 4721.00 frames. ], tot_loss[loss=0.1852, simple_loss=0.2531, pruned_loss=0.05859, over 878671.00 frames. ], batch size: 59, lr: 3.61e-03, grad_scale: 16.0 2023-04-27 03:50:25,826 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=69234.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 03:50:31,195 INFO [optim.py:369] (6/7) 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 03:50:32,791 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-04-27 03:50:32,982 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=69245.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 03:50:38,163 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.6446, 1.9081, 1.9619, 2.1160, 1.9810, 2.0919, 2.0790, 2.0576], device='cuda:6'), covar=tensor([0.3949, 0.5896, 0.4766, 0.4882, 0.5498, 0.7441, 0.6142, 0.5665], device='cuda:6'), in_proj_covar=tensor([0.0327, 0.0377, 0.0316, 0.0328, 0.0341, 0.0398, 0.0358, 0.0325], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 03:50:52,427 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=69273.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 03:50:57,810 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=69282.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 03:50:58,365 INFO [finetune.py:976] (6/7) Epoch 13, batch 550, loss[loss=0.1587, simple_loss=0.2278, pruned_loss=0.04477, over 4808.00 frames. ], tot_loss[loss=0.1829, simple_loss=0.2505, pruned_loss=0.05768, over 895711.65 frames. ], batch size: 25, lr: 3.61e-03, grad_scale: 16.0 2023-04-27 03:51:00,308 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=69286.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 03:51:14,903 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=69306.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 03:51:24,247 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=69320.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 03:51:30,958 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.1299, 2.7335, 2.3046, 2.5280, 1.8961, 2.2356, 2.3283, 1.7499], device='cuda:6'), covar=tensor([0.2196, 0.1161, 0.0800, 0.1257, 0.3573, 0.1222, 0.2162, 0.2849], device='cuda:6'), in_proj_covar=tensor([0.0292, 0.0312, 0.0225, 0.0284, 0.0313, 0.0263, 0.0255, 0.0273], device='cuda:6'), out_proj_covar=tensor([1.1779e-04, 1.2482e-04, 8.9838e-05, 1.1334e-04, 1.2786e-04, 1.0548e-04, 1.0345e-04, 1.0926e-04], device='cuda:6') 2023-04-27 03:51:32,073 INFO [finetune.py:976] (6/7) Epoch 13, batch 600, loss[loss=0.2187, simple_loss=0.2814, pruned_loss=0.07806, over 4749.00 frames. ], tot_loss[loss=0.1837, simple_loss=0.251, pruned_loss=0.05817, over 908750.30 frames. ], batch size: 59, lr: 3.61e-03, grad_scale: 16.0 2023-04-27 03:51:32,824 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=69334.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 03:51:37,049 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=69341.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 03:51:38,135 INFO [optim.py:369] (6/7) 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 03:51:50,629 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-04-27 03:51:56,894 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=69368.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 03:52:04,771 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=69381.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 03:52:05,915 INFO [finetune.py:976] (6/7) Epoch 13, batch 650, loss[loss=0.1667, simple_loss=0.2493, pruned_loss=0.04207, over 4898.00 frames. ], tot_loss[loss=0.1869, simple_loss=0.2549, pruned_loss=0.05944, over 919958.88 frames. ], batch size: 37, lr: 3.61e-03, grad_scale: 16.0 2023-04-27 03:52:08,436 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0432, 2.6437, 1.0443, 1.3968, 1.9969, 1.2040, 3.5396, 1.6694], device='cuda:6'), covar=tensor([0.0766, 0.0741, 0.0887, 0.1413, 0.0543, 0.1090, 0.0253, 0.0689], device='cuda:6'), in_proj_covar=tensor([0.0051, 0.0066, 0.0049, 0.0047, 0.0051, 0.0052, 0.0076, 0.0051], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:6') 2023-04-27 03:52:09,607 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=69389.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 03:52:12,663 INFO [zipformer.py:1188] (6/7) 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] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=69399.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 03:52:36,501 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=69429.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 03:52:44,420 INFO [finetune.py:976] (6/7) Epoch 13, batch 700, loss[loss=0.1789, simple_loss=0.2455, pruned_loss=0.05616, over 4916.00 frames. ], tot_loss[loss=0.1881, simple_loss=0.2563, pruned_loss=0.05993, over 929405.91 frames. ], batch size: 36, lr: 3.61e-03, grad_scale: 16.0 2023-04-27 03:52:56,109 INFO [optim.py:369] (6/7) 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 03:53:09,893 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 13, batch 750, loss[loss=0.1724, simple_loss=0.231, pruned_loss=0.05691, over 4000.00 frames. ], tot_loss[loss=0.1888, simple_loss=0.2576, pruned_loss=0.06, over 936506.18 frames. ], batch size: 17, lr: 3.61e-03, grad_scale: 16.0 2023-04-27 03:53:59,734 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-04-27 03:54:42,282 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=69518.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 03:54:46,364 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 13, batch 800, loss[loss=0.1755, simple_loss=0.2413, pruned_loss=0.05487, over 4905.00 frames. ], tot_loss[loss=0.1884, simple_loss=0.2573, pruned_loss=0.05974, over 940959.98 frames. ], batch size: 37, lr: 3.61e-03, grad_scale: 16.0 2023-04-27 03:55:09,400 INFO [optim.py:369] (6/7) 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 03:55:15,873 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.76 vs. limit=2.0 2023-04-27 03:55:18,271 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 2023-04-27 03:55:35,210 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=69581.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 03:55:36,384 INFO [finetune.py:976] (6/7) Epoch 13, batch 850, loss[loss=0.1489, simple_loss=0.2279, pruned_loss=0.03491, over 4852.00 frames. ], tot_loss[loss=0.1858, simple_loss=0.2544, pruned_loss=0.05854, over 944027.51 frames. ], batch size: 31, lr: 3.61e-03, grad_scale: 16.0 2023-04-27 03:55:47,527 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=69601.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 03:56:07,729 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=69629.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 03:56:10,108 INFO [finetune.py:976] (6/7) Epoch 13, batch 900, loss[loss=0.1923, simple_loss=0.2413, pruned_loss=0.07169, over 4695.00 frames. ], tot_loss[loss=0.1849, simple_loss=0.253, pruned_loss=0.05839, over 948189.70 frames. ], batch size: 23, lr: 3.61e-03, grad_scale: 32.0 2023-04-27 03:56:16,242 INFO [optim.py:369] (6/7) 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 03:56:37,658 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=69674.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 03:56:39,833 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=69676.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 03:56:44,054 INFO [finetune.py:976] (6/7) Epoch 13, batch 950, loss[loss=0.1331, simple_loss=0.1894, pruned_loss=0.03845, over 4032.00 frames. ], tot_loss[loss=0.1846, simple_loss=0.2521, pruned_loss=0.05853, over 950585.55 frames. ], batch size: 17, lr: 3.61e-03, grad_scale: 32.0 2023-04-27 03:56:49,100 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7287, 2.0083, 1.0768, 1.3619, 2.1659, 1.5896, 1.5260, 1.5513], device='cuda:6'), covar=tensor([0.0485, 0.0347, 0.0326, 0.0571, 0.0258, 0.0515, 0.0491, 0.0571], device='cuda:6'), in_proj_covar=tensor([0.0029, 0.0025, 0.0023, 0.0029, 0.0020, 0.0029, 0.0028, 0.0029], device='cuda:6'), out_proj_covar=tensor([0.0049, 0.0044, 0.0037, 0.0050, 0.0037, 0.0048, 0.0048, 0.0050], device='cuda:6') 2023-04-27 03:56:53,915 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=69699.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 03:57:11,099 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=69724.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 03:57:12,881 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5453, 1.1075, 1.2968, 1.2231, 1.6711, 1.3509, 1.0545, 1.2682], device='cuda:6'), covar=tensor([0.1566, 0.1432, 0.2100, 0.1510, 0.0863, 0.1508, 0.2081, 0.2176], device='cuda:6'), in_proj_covar=tensor([0.0305, 0.0316, 0.0348, 0.0291, 0.0330, 0.0313, 0.0304, 0.0358], device='cuda:6'), 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:6') 2023-04-27 03:57:18,039 INFO [finetune.py:976] (6/7) Epoch 13, batch 1000, loss[loss=0.2416, simple_loss=0.3201, pruned_loss=0.08156, over 4799.00 frames. ], tot_loss[loss=0.1858, simple_loss=0.2535, pruned_loss=0.05911, over 951933.71 frames. ], batch size: 51, lr: 3.61e-03, grad_scale: 32.0 2023-04-27 03:57:19,398 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=69735.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 03:57:24,158 INFO [optim.py:369] (6/7) 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 03:57:26,751 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=69747.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 03:57:28,612 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=69750.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 03:57:30,658 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.24 vs. limit=5.0 2023-04-27 03:57:50,563 INFO [finetune.py:976] (6/7) Epoch 13, batch 1050, loss[loss=0.2068, simple_loss=0.2887, pruned_loss=0.06241, over 4903.00 frames. ], tot_loss[loss=0.188, simple_loss=0.2566, pruned_loss=0.05972, over 953138.64 frames. ], batch size: 36, lr: 3.61e-03, grad_scale: 32.0 2023-04-27 03:58:13,053 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=69818.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 03:58:16,688 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=69824.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 03:58:23,480 INFO [finetune.py:976] (6/7) Epoch 13, batch 1100, loss[loss=0.1563, simple_loss=0.2261, pruned_loss=0.04318, over 4749.00 frames. ], tot_loss[loss=0.1882, simple_loss=0.2571, pruned_loss=0.05962, over 953641.09 frames. ], batch size: 54, lr: 3.61e-03, grad_scale: 32.0 2023-04-27 03:58:30,702 INFO [optim.py:369] (6/7) 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 03:58:50,917 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=69866.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 03:59:00,300 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=69872.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 03:59:00,621 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.13 vs. limit=2.0 2023-04-27 03:59:12,155 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 13, batch 1150, loss[loss=0.2427, simple_loss=0.3048, pruned_loss=0.09027, over 4895.00 frames. ], tot_loss[loss=0.1877, simple_loss=0.2571, pruned_loss=0.05917, over 953338.62 frames. ], batch size: 43, lr: 3.61e-03, grad_scale: 32.0 2023-04-27 03:59:41,635 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=69901.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 04:00:04,402 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5285, 1.4419, 0.5738, 1.2367, 1.5190, 1.3530, 1.3197, 1.3189], device='cuda:6'), covar=tensor([0.0594, 0.0348, 0.0415, 0.0610, 0.0298, 0.0670, 0.0603, 0.0638], device='cuda:6'), in_proj_covar=tensor([0.0029, 0.0025, 0.0023, 0.0029, 0.0020, 0.0029, 0.0028, 0.0029], device='cuda:6'), out_proj_covar=tensor([0.0050, 0.0045, 0.0038, 0.0050, 0.0037, 0.0049, 0.0048, 0.0050], device='cuda:6') 2023-04-27 04:00:16,221 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=69929.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 04:00:16,250 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=69929.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 04:00:18,639 INFO [finetune.py:976] (6/7) Epoch 13, batch 1200, loss[loss=0.1713, simple_loss=0.2386, pruned_loss=0.05198, over 4841.00 frames. ], tot_loss[loss=0.1878, simple_loss=0.2564, pruned_loss=0.05955, over 953633.06 frames. ], batch size: 49, lr: 3.61e-03, grad_scale: 32.0 2023-04-27 04:00:32,298 INFO [optim.py:369] (6/7) 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 04:00:36,039 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=69949.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 04:00:47,626 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.3845, 2.6190, 1.4973, 2.2084, 2.9061, 2.3310, 2.2951, 2.3582], device='cuda:6'), covar=tensor([0.0431, 0.0296, 0.0257, 0.0455, 0.0191, 0.0438, 0.0408, 0.0459], device='cuda:6'), in_proj_covar=tensor([0.0029, 0.0025, 0.0023, 0.0030, 0.0020, 0.0029, 0.0029, 0.0029], device='cuda:6'), out_proj_covar=tensor([0.0050, 0.0045, 0.0038, 0.0050, 0.0038, 0.0049, 0.0049, 0.0050], device='cuda:6') 2023-04-27 04:00:48,210 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([0.8923, 1.2865, 3.2719, 3.0317, 2.9702, 3.1786, 3.1755, 2.8909], device='cuda:6'), covar=tensor([0.6876, 0.5098, 0.1467, 0.2153, 0.1208, 0.1656, 0.1641, 0.1691], device='cuda:6'), in_proj_covar=tensor([0.0308, 0.0302, 0.0398, 0.0406, 0.0345, 0.0404, 0.0310, 0.0364], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-27 04:00:52,495 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=69976.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 04:00:53,027 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=69977.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 04:00:54,946 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.0641, 1.0219, 1.2569, 1.1468, 1.0406, 0.8824, 0.9216, 0.4125], device='cuda:6'), covar=tensor([0.0559, 0.0633, 0.0495, 0.0569, 0.0690, 0.1338, 0.0506, 0.0790], device='cuda:6'), in_proj_covar=tensor([0.0067, 0.0070, 0.0071, 0.0066, 0.0075, 0.0096, 0.0075, 0.0069], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-27 04:00:57,154 INFO [finetune.py:976] (6/7) Epoch 13, batch 1250, loss[loss=0.1786, simple_loss=0.2424, pruned_loss=0.05737, over 4867.00 frames. ], tot_loss[loss=0.1864, simple_loss=0.2544, pruned_loss=0.05922, over 954207.63 frames. ], batch size: 31, lr: 3.61e-03, grad_scale: 32.0 2023-04-27 04:01:26,387 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=70024.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 04:01:26,414 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=70024.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 04:01:30,022 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 13, batch 1300, loss[loss=0.1568, simple_loss=0.2214, pruned_loss=0.04606, over 4152.00 frames. ], tot_loss[loss=0.1842, simple_loss=0.2516, pruned_loss=0.05839, over 955193.05 frames. ], batch size: 65, lr: 3.61e-03, grad_scale: 32.0 2023-04-27 04:01:39,419 INFO [optim.py:369] (6/7) 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 04:01:40,633 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.3001, 2.9834, 0.9147, 1.7270, 1.5618, 2.1607, 1.7413, 0.9161], device='cuda:6'), covar=tensor([0.1484, 0.1306, 0.1949, 0.1259, 0.1273, 0.1048, 0.1489, 0.1964], device='cuda:6'), in_proj_covar=tensor([0.0117, 0.0245, 0.0138, 0.0121, 0.0133, 0.0152, 0.0116, 0.0119], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-27 04:01:44,809 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-04-27 04:01:45,194 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=70050.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 04:01:58,567 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=70072.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 04:01:59,206 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9029, 4.3342, 0.7961, 2.3078, 2.2701, 2.8290, 2.3476, 1.0328], device='cuda:6'), covar=tensor([0.1406, 0.0958, 0.2255, 0.1191, 0.1149, 0.1057, 0.1432, 0.2049], device='cuda:6'), in_proj_covar=tensor([0.0117, 0.0245, 0.0138, 0.0121, 0.0133, 0.0152, 0.0117, 0.0119], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-27 04:02:05,230 INFO [finetune.py:976] (6/7) Epoch 13, batch 1350, loss[loss=0.1729, simple_loss=0.2424, pruned_loss=0.05171, over 4843.00 frames. ], tot_loss[loss=0.1832, simple_loss=0.2507, pruned_loss=0.05782, over 955472.49 frames. ], batch size: 33, lr: 3.61e-03, grad_scale: 32.0 2023-04-27 04:02:16,792 INFO [zipformer.py:1188] (6/7) 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] (6/7) 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] (6/7) Epoch 13, batch 1400, loss[loss=0.1932, simple_loss=0.2781, pruned_loss=0.05415, over 4848.00 frames. ], tot_loss[loss=0.1851, simple_loss=0.2533, pruned_loss=0.05841, over 954492.46 frames. ], batch size: 47, lr: 3.61e-03, grad_scale: 32.0 2023-04-27 04:02:43,374 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.2854, 1.7089, 2.1746, 2.6921, 2.1745, 1.6734, 1.4191, 1.9812], device='cuda:6'), covar=tensor([0.3749, 0.3808, 0.1884, 0.2715, 0.3071, 0.3061, 0.4666, 0.2390], device='cuda:6'), in_proj_covar=tensor([0.0285, 0.0247, 0.0222, 0.0314, 0.0212, 0.0227, 0.0230, 0.0184], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:6') 2023-04-27 04:02:45,036 INFO [optim.py:369] (6/7) 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 04:03:04,709 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=70171.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 04:03:11,828 INFO [finetune.py:976] (6/7) Epoch 13, batch 1450, loss[loss=0.1705, simple_loss=0.2308, pruned_loss=0.05508, over 4730.00 frames. ], tot_loss[loss=0.1864, simple_loss=0.2552, pruned_loss=0.0588, over 954460.06 frames. ], batch size: 23, lr: 3.61e-03, grad_scale: 32.0 2023-04-27 04:03:45,320 INFO [finetune.py:976] (6/7) Epoch 13, batch 1500, loss[loss=0.1811, simple_loss=0.2416, pruned_loss=0.06036, over 4054.00 frames. ], tot_loss[loss=0.1878, simple_loss=0.2567, pruned_loss=0.05941, over 952814.37 frames. ], batch size: 17, lr: 3.61e-03, grad_scale: 32.0 2023-04-27 04:03:51,425 INFO [optim.py:369] (6/7) 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 04:04:10,457 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=70260.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 04:04:40,472 INFO [finetune.py:976] (6/7) Epoch 13, batch 1550, loss[loss=0.141, simple_loss=0.2046, pruned_loss=0.03873, over 4243.00 frames. ], tot_loss[loss=0.1884, simple_loss=0.2573, pruned_loss=0.05979, over 952961.16 frames. ], batch size: 18, lr: 3.61e-03, grad_scale: 32.0 2023-04-27 04:04:57,585 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.78 vs. limit=5.0 2023-04-27 04:05:16,073 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.74 vs. limit=2.0 2023-04-27 04:05:17,658 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=70321.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 04:05:28,447 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=70330.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 04:05:30,177 INFO [finetune.py:976] (6/7) Epoch 13, batch 1600, loss[loss=0.1462, simple_loss=0.226, pruned_loss=0.0332, over 4750.00 frames. ], tot_loss[loss=0.1865, simple_loss=0.255, pruned_loss=0.05903, over 953502.09 frames. ], batch size: 26, lr: 3.60e-03, grad_scale: 32.0 2023-04-27 04:05:41,008 INFO [optim.py:369] (6/7) 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 04:05:50,525 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([3.7732, 3.7484, 2.7416, 4.3503, 3.8432, 3.7702, 1.5262, 3.7711], device='cuda:6'), covar=tensor([0.1860, 0.1289, 0.3153, 0.1710, 0.2846, 0.2048, 0.6081, 0.2377], device='cuda:6'), in_proj_covar=tensor([0.0243, 0.0215, 0.0249, 0.0302, 0.0298, 0.0247, 0.0270, 0.0269], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 04:06:16,258 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=70378.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 04:06:19,132 INFO [finetune.py:976] (6/7) Epoch 13, batch 1650, loss[loss=0.1778, simple_loss=0.2425, pruned_loss=0.05656, over 4767.00 frames. ], tot_loss[loss=0.1844, simple_loss=0.2522, pruned_loss=0.05832, over 953873.84 frames. ], batch size: 26, lr: 3.60e-03, grad_scale: 32.0 2023-04-27 04:06:52,942 INFO [finetune.py:976] (6/7) Epoch 13, batch 1700, loss[loss=0.1505, simple_loss=0.2249, pruned_loss=0.03802, over 4777.00 frames. ], tot_loss[loss=0.1826, simple_loss=0.2501, pruned_loss=0.05751, over 955097.42 frames. ], batch size: 26, lr: 3.60e-03, grad_scale: 32.0 2023-04-27 04:06:59,064 INFO [optim.py:369] (6/7) 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 04:07:16,035 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=70466.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 04:07:17,987 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7455, 2.3820, 1.8015, 1.6386, 1.2917, 1.2921, 1.9220, 1.2632], device='cuda:6'), covar=tensor([0.1543, 0.1339, 0.1325, 0.1761, 0.2229, 0.1857, 0.0916, 0.1888], device='cuda:6'), in_proj_covar=tensor([0.0196, 0.0214, 0.0169, 0.0204, 0.0202, 0.0184, 0.0157, 0.0186], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:6') 2023-04-27 04:07:18,559 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6994, 1.4888, 4.5437, 4.2467, 3.9394, 4.2340, 4.1367, 3.9900], device='cuda:6'), covar=tensor([0.6810, 0.5538, 0.0866, 0.1754, 0.1188, 0.1741, 0.1452, 0.1521], device='cuda:6'), in_proj_covar=tensor([0.0306, 0.0303, 0.0398, 0.0405, 0.0345, 0.0403, 0.0309, 0.0363], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-27 04:07:22,152 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.64 vs. limit=2.0 2023-04-27 04:07:26,839 INFO [finetune.py:976] (6/7) Epoch 13, batch 1750, loss[loss=0.2353, simple_loss=0.3042, pruned_loss=0.08317, over 4813.00 frames. ], tot_loss[loss=0.184, simple_loss=0.2515, pruned_loss=0.05827, over 954621.36 frames. ], batch size: 45, lr: 3.60e-03, grad_scale: 32.0 2023-04-27 04:08:00,058 INFO [finetune.py:976] (6/7) Epoch 13, batch 1800, loss[loss=0.1905, simple_loss=0.256, pruned_loss=0.06251, over 4834.00 frames. ], tot_loss[loss=0.1858, simple_loss=0.2543, pruned_loss=0.05865, over 955180.59 frames. ], batch size: 47, lr: 3.60e-03, grad_scale: 32.0 2023-04-27 04:08:06,008 INFO [optim.py:369] (6/7) 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 04:08:14,716 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.2977, 1.5297, 1.3746, 1.7489, 1.5421, 1.7846, 1.3604, 3.3458], device='cuda:6'), covar=tensor([0.0672, 0.0844, 0.0904, 0.1280, 0.0714, 0.0573, 0.0807, 0.0188], device='cuda:6'), in_proj_covar=tensor([0.0038, 0.0039, 0.0040, 0.0044, 0.0040, 0.0038, 0.0039, 0.0058], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:6') 2023-04-27 04:08:33,362 INFO [finetune.py:976] (6/7) Epoch 13, batch 1850, loss[loss=0.198, simple_loss=0.2714, pruned_loss=0.06235, over 4817.00 frames. ], tot_loss[loss=0.1887, simple_loss=0.2576, pruned_loss=0.05988, over 954760.09 frames. ], batch size: 38, lr: 3.60e-03, grad_scale: 32.0 2023-04-27 04:08:40,200 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8374, 2.4727, 2.0601, 2.2083, 1.7152, 1.9946, 2.0820, 1.6114], device='cuda:6'), covar=tensor([0.2105, 0.1377, 0.0921, 0.1441, 0.3133, 0.1272, 0.1914, 0.2751], device='cuda:6'), in_proj_covar=tensor([0.0291, 0.0311, 0.0225, 0.0283, 0.0313, 0.0263, 0.0255, 0.0272], device='cuda:6'), out_proj_covar=tensor([1.1727e-04, 1.2409e-04, 8.9652e-05, 1.1316e-04, 1.2761e-04, 1.0531e-04, 1.0347e-04, 1.0874e-04], device='cuda:6') 2023-04-27 04:08:40,533 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.35 vs. limit=5.0 2023-04-27 04:08:42,063 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9920, 1.5461, 1.8693, 2.1459, 1.8783, 1.4956, 1.2393, 1.6162], device='cuda:6'), covar=tensor([0.3026, 0.3208, 0.1545, 0.2179, 0.2434, 0.2406, 0.4266, 0.2112], device='cuda:6'), in_proj_covar=tensor([0.0284, 0.0246, 0.0220, 0.0313, 0.0211, 0.0227, 0.0229, 0.0183], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:6') 2023-04-27 04:08:54,139 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 13, batch 1900, loss[loss=0.1955, simple_loss=0.2716, pruned_loss=0.05968, over 4704.00 frames. ], tot_loss[loss=0.1892, simple_loss=0.2585, pruned_loss=0.05996, over 954013.99 frames. ], batch size: 59, lr: 3.60e-03, grad_scale: 32.0 2023-04-27 04:09:12,477 INFO [optim.py:369] (6/7) 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 04:09:16,861 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0828, 2.9026, 2.2864, 2.5200, 1.9368, 2.5046, 2.5104, 1.8320], device='cuda:6'), covar=tensor([0.2457, 0.1472, 0.0894, 0.1599, 0.3345, 0.1215, 0.2191, 0.2993], device='cuda:6'), in_proj_covar=tensor([0.0292, 0.0311, 0.0225, 0.0284, 0.0313, 0.0264, 0.0255, 0.0273], device='cuda:6'), out_proj_covar=tensor([1.1762e-04, 1.2436e-04, 8.9849e-05, 1.1335e-04, 1.2760e-04, 1.0560e-04, 1.0357e-04, 1.0893e-04], device='cuda:6') 2023-04-27 04:10:00,433 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9317, 1.5182, 2.0904, 2.2897, 1.7378, 1.4724, 1.6939, 1.2746], device='cuda:6'), covar=tensor([0.0517, 0.0963, 0.0534, 0.0666, 0.0710, 0.1249, 0.0704, 0.0723], device='cuda:6'), in_proj_covar=tensor([0.0067, 0.0069, 0.0070, 0.0066, 0.0074, 0.0096, 0.0074, 0.0069], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-27 04:10:02,034 INFO [finetune.py:976] (6/7) Epoch 13, batch 1950, loss[loss=0.2184, simple_loss=0.2736, pruned_loss=0.0816, over 4719.00 frames. ], tot_loss[loss=0.1875, simple_loss=0.2566, pruned_loss=0.05919, over 953042.91 frames. ], batch size: 59, lr: 3.60e-03, grad_scale: 32.0 2023-04-27 04:10:07,010 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([0.9286, 2.1209, 1.8630, 1.6984, 2.2254, 1.7306, 2.7382, 1.6669], device='cuda:6'), covar=tensor([0.3814, 0.1608, 0.4706, 0.3194, 0.1811, 0.2550, 0.1276, 0.4415], device='cuda:6'), in_proj_covar=tensor([0.0339, 0.0340, 0.0422, 0.0352, 0.0376, 0.0375, 0.0368, 0.0414], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 04:10:39,638 INFO [finetune.py:976] (6/7) Epoch 13, batch 2000, loss[loss=0.1918, simple_loss=0.2599, pruned_loss=0.06187, over 4936.00 frames. ], tot_loss[loss=0.1845, simple_loss=0.2531, pruned_loss=0.05793, over 951085.65 frames. ], batch size: 38, lr: 3.60e-03, grad_scale: 32.0 2023-04-27 04:10:52,179 INFO [optim.py:369] (6/7) 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 04:11:10,621 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 2023-04-27 04:11:15,906 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=70766.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 04:11:28,147 INFO [finetune.py:976] (6/7) Epoch 13, batch 2050, loss[loss=0.1561, simple_loss=0.2226, pruned_loss=0.04482, over 4664.00 frames. ], tot_loss[loss=0.182, simple_loss=0.2502, pruned_loss=0.05689, over 951783.98 frames. ], batch size: 23, lr: 3.60e-03, grad_scale: 32.0 2023-04-27 04:11:48,345 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=70814.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 04:12:01,814 INFO [finetune.py:976] (6/7) Epoch 13, batch 2100, loss[loss=0.2019, simple_loss=0.2716, pruned_loss=0.06606, over 4765.00 frames. ], tot_loss[loss=0.1833, simple_loss=0.2507, pruned_loss=0.05792, over 951223.87 frames. ], batch size: 59, lr: 3.60e-03, grad_scale: 32.0 2023-04-27 04:12:08,818 INFO [optim.py:369] (6/7) 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 04:12:08,979 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8167, 1.3732, 1.6372, 1.6614, 1.5806, 1.2875, 0.7487, 1.3655], device='cuda:6'), covar=tensor([0.3257, 0.3295, 0.1751, 0.2325, 0.2425, 0.2679, 0.4369, 0.2064], device='cuda:6'), in_proj_covar=tensor([0.0284, 0.0245, 0.0220, 0.0313, 0.0212, 0.0227, 0.0229, 0.0183], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:6') 2023-04-27 04:12:27,887 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9112, 2.3770, 1.0062, 1.3188, 1.8015, 1.1583, 2.8569, 1.4979], device='cuda:6'), covar=tensor([0.0664, 0.0677, 0.0761, 0.1229, 0.0478, 0.0987, 0.0296, 0.0651], device='cuda:6'), in_proj_covar=tensor([0.0051, 0.0066, 0.0049, 0.0047, 0.0050, 0.0052, 0.0076, 0.0051], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0007, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:6') 2023-04-27 04:12:35,713 INFO [finetune.py:976] (6/7) Epoch 13, batch 2150, loss[loss=0.2506, simple_loss=0.3136, pruned_loss=0.09379, over 4915.00 frames. ], tot_loss[loss=0.1863, simple_loss=0.254, pruned_loss=0.05932, over 950599.69 frames. ], batch size: 36, lr: 3.60e-03, grad_scale: 32.0 2023-04-27 04:12:56,787 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=70916.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 04:13:03,909 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.90 vs. limit=5.0 2023-04-27 04:13:05,874 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([0.9055, 1.8373, 1.6093, 1.5071, 1.9425, 1.5686, 2.4704, 1.4122], device='cuda:6'), covar=tensor([0.3877, 0.1780, 0.4900, 0.3201, 0.1684, 0.2460, 0.1357, 0.4878], device='cuda:6'), in_proj_covar=tensor([0.0338, 0.0338, 0.0421, 0.0351, 0.0375, 0.0374, 0.0368, 0.0413], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 04:13:08,518 INFO [finetune.py:976] (6/7) Epoch 13, batch 2200, loss[loss=0.1722, simple_loss=0.2332, pruned_loss=0.05561, over 4777.00 frames. ], tot_loss[loss=0.1859, simple_loss=0.2542, pruned_loss=0.05879, over 951915.56 frames. ], batch size: 29, lr: 3.60e-03, grad_scale: 32.0 2023-04-27 04:13:16,545 INFO [optim.py:369] (6/7) 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 04:13:29,558 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=70964.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 04:13:33,306 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8874, 1.1269, 5.0496, 4.7278, 4.4279, 4.8262, 4.4301, 4.4874], device='cuda:6'), covar=tensor([0.6941, 0.6596, 0.0900, 0.1674, 0.1020, 0.1131, 0.1262, 0.1548], device='cuda:6'), in_proj_covar=tensor([0.0310, 0.0304, 0.0400, 0.0406, 0.0347, 0.0405, 0.0311, 0.0366], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-27 04:13:41,742 INFO [finetune.py:976] (6/7) Epoch 13, batch 2250, loss[loss=0.1886, simple_loss=0.2509, pruned_loss=0.06318, over 4776.00 frames. ], tot_loss[loss=0.1858, simple_loss=0.2544, pruned_loss=0.05862, over 951090.95 frames. ], batch size: 28, lr: 3.60e-03, grad_scale: 32.0 2023-04-27 04:13:49,257 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=70992.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 04:13:54,649 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.47 vs. limit=5.0 2023-04-27 04:14:13,348 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=71030.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 04:14:14,997 INFO [finetune.py:976] (6/7) Epoch 13, batch 2300, loss[loss=0.1995, simple_loss=0.259, pruned_loss=0.06999, over 4920.00 frames. ], tot_loss[loss=0.1864, simple_loss=0.2552, pruned_loss=0.05877, over 951400.12 frames. ], batch size: 38, lr: 3.60e-03, grad_scale: 32.0 2023-04-27 04:14:23,496 INFO [optim.py:369] (6/7) 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 04:14:30,598 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=71053.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 04:15:10,018 INFO [finetune.py:976] (6/7) Epoch 13, batch 2350, loss[loss=0.1848, simple_loss=0.2558, pruned_loss=0.05688, over 4902.00 frames. ], tot_loss[loss=0.1847, simple_loss=0.2535, pruned_loss=0.05799, over 952349.93 frames. ], batch size: 36, lr: 3.60e-03, grad_scale: 32.0 2023-04-27 04:15:21,364 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=71091.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 04:15:57,363 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.35 vs. limit=5.0 2023-04-27 04:16:15,378 INFO [finetune.py:976] (6/7) Epoch 13, batch 2400, loss[loss=0.1417, simple_loss=0.2122, pruned_loss=0.03563, over 4855.00 frames. ], tot_loss[loss=0.182, simple_loss=0.2503, pruned_loss=0.05682, over 953456.24 frames. ], batch size: 49, lr: 3.60e-03, grad_scale: 32.0 2023-04-27 04:16:26,844 INFO [optim.py:369] (6/7) 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 04:16:54,088 INFO [finetune.py:976] (6/7) Epoch 13, batch 2450, loss[loss=0.194, simple_loss=0.2609, pruned_loss=0.06361, over 4904.00 frames. ], tot_loss[loss=0.1801, simple_loss=0.248, pruned_loss=0.05606, over 954673.27 frames. ], batch size: 43, lr: 3.60e-03, grad_scale: 32.0 2023-04-27 04:17:17,912 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.77 vs. limit=5.0 2023-04-27 04:17:28,065 INFO [finetune.py:976] (6/7) Epoch 13, batch 2500, loss[loss=0.1676, simple_loss=0.238, pruned_loss=0.04863, over 4819.00 frames. ], tot_loss[loss=0.1811, simple_loss=0.2488, pruned_loss=0.05673, over 954125.66 frames. ], batch size: 30, lr: 3.60e-03, grad_scale: 32.0 2023-04-27 04:17:30,010 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5042, 0.9118, 1.5602, 1.9712, 1.6024, 1.4981, 1.5105, 1.5041], device='cuda:6'), covar=tensor([0.5040, 0.6938, 0.6863, 0.6696, 0.6096, 0.8049, 0.8165, 0.8829], device='cuda:6'), in_proj_covar=tensor([0.0413, 0.0410, 0.0498, 0.0518, 0.0444, 0.0465, 0.0472, 0.0476], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 04:17:34,107 INFO [optim.py:369] (6/7) 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 04:17:53,036 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.1251, 2.4837, 0.9436, 1.5307, 1.4178, 1.9692, 1.5956, 0.8894], device='cuda:6'), covar=tensor([0.1387, 0.1087, 0.1561, 0.1248, 0.1155, 0.0827, 0.1375, 0.1798], device='cuda:6'), in_proj_covar=tensor([0.0118, 0.0247, 0.0140, 0.0122, 0.0133, 0.0153, 0.0118, 0.0120], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-27 04:17:57,040 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-04-27 04:18:01,397 INFO [finetune.py:976] (6/7) Epoch 13, batch 2550, loss[loss=0.1529, simple_loss=0.2323, pruned_loss=0.03674, over 4825.00 frames. ], tot_loss[loss=0.1852, simple_loss=0.2537, pruned_loss=0.0584, over 952959.41 frames. ], batch size: 33, lr: 3.60e-03, grad_scale: 32.0 2023-04-27 04:18:03,640 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-04-27 04:18:34,377 INFO [finetune.py:976] (6/7) Epoch 13, batch 2600, loss[loss=0.2389, simple_loss=0.3009, pruned_loss=0.08844, over 4762.00 frames. ], tot_loss[loss=0.1859, simple_loss=0.2549, pruned_loss=0.05843, over 953699.52 frames. ], batch size: 59, lr: 3.60e-03, grad_scale: 32.0 2023-04-27 04:18:40,525 INFO [optim.py:369] (6/7) 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 04:18:43,645 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=71348.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 04:18:54,995 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7405, 1.7974, 1.0487, 1.4267, 2.0760, 1.6383, 1.5378, 1.5791], device='cuda:6'), covar=tensor([0.0479, 0.0338, 0.0307, 0.0535, 0.0259, 0.0496, 0.0505, 0.0527], device='cuda:6'), in_proj_covar=tensor([0.0029, 0.0025, 0.0023, 0.0029, 0.0020, 0.0029, 0.0028, 0.0029], device='cuda:6'), out_proj_covar=tensor([0.0050, 0.0044, 0.0038, 0.0050, 0.0037, 0.0048, 0.0048, 0.0050], device='cuda:6') 2023-04-27 04:18:55,039 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.6995, 1.9382, 1.9117, 2.0840, 1.8693, 1.9697, 1.9488, 1.9082], device='cuda:6'), covar=tensor([0.4279, 0.7478, 0.5947, 0.5272, 0.6370, 0.8263, 0.7425, 0.6619], device='cuda:6'), in_proj_covar=tensor([0.0329, 0.0378, 0.0317, 0.0331, 0.0341, 0.0399, 0.0359, 0.0325], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 04:19:05,546 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.74 vs. limit=5.0 2023-04-27 04:19:08,106 INFO [finetune.py:976] (6/7) Epoch 13, batch 2650, loss[loss=0.1843, simple_loss=0.2665, pruned_loss=0.05102, over 4822.00 frames. ], tot_loss[loss=0.187, simple_loss=0.2562, pruned_loss=0.05887, over 953483.93 frames. ], batch size: 49, lr: 3.60e-03, grad_scale: 32.0 2023-04-27 04:19:10,020 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=71386.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 04:19:17,410 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.2355, 3.0226, 1.0072, 1.6288, 1.6977, 2.1744, 1.8400, 0.9496], device='cuda:6'), covar=tensor([0.1527, 0.0893, 0.1860, 0.1363, 0.1122, 0.0999, 0.1503, 0.1988], device='cuda:6'), in_proj_covar=tensor([0.0118, 0.0246, 0.0139, 0.0122, 0.0133, 0.0153, 0.0117, 0.0120], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-27 04:19:41,958 INFO [finetune.py:976] (6/7) Epoch 13, batch 2700, loss[loss=0.1669, simple_loss=0.2324, pruned_loss=0.05067, over 4770.00 frames. ], tot_loss[loss=0.1858, simple_loss=0.255, pruned_loss=0.05827, over 953207.82 frames. ], batch size: 26, lr: 3.60e-03, grad_scale: 32.0 2023-04-27 04:19:48,072 INFO [optim.py:369] (6/7) 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 04:19:57,116 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5111, 1.3266, 4.3706, 4.0627, 3.7898, 4.1677, 3.9695, 3.8755], device='cuda:6'), covar=tensor([0.7370, 0.6206, 0.1110, 0.1852, 0.1124, 0.1594, 0.1656, 0.1629], device='cuda:6'), in_proj_covar=tensor([0.0311, 0.0307, 0.0403, 0.0409, 0.0348, 0.0406, 0.0313, 0.0368], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-27 04:19:58,765 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.3534, 1.5421, 1.5143, 1.8559, 1.6986, 1.9854, 1.4382, 3.5673], device='cuda:6'), covar=tensor([0.0673, 0.0904, 0.0956, 0.1285, 0.0704, 0.0502, 0.0860, 0.0222], device='cuda:6'), in_proj_covar=tensor([0.0038, 0.0039, 0.0040, 0.0043, 0.0040, 0.0038, 0.0038, 0.0057], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0012, 0.0013, 0.0016], device='cuda:6') 2023-04-27 04:19:58,789 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([0.9951, 2.0380, 1.9595, 1.7247, 2.3019, 1.6830, 2.8224, 1.6918], device='cuda:6'), covar=tensor([0.3954, 0.1928, 0.4945, 0.3301, 0.1599, 0.2728, 0.1235, 0.4507], device='cuda:6'), in_proj_covar=tensor([0.0342, 0.0342, 0.0426, 0.0356, 0.0378, 0.0379, 0.0372, 0.0415], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 04:20:30,300 INFO [finetune.py:976] (6/7) Epoch 13, batch 2750, loss[loss=0.1999, simple_loss=0.2609, pruned_loss=0.06949, over 4830.00 frames. ], tot_loss[loss=0.1843, simple_loss=0.2528, pruned_loss=0.05787, over 953720.50 frames. ], batch size: 33, lr: 3.59e-03, grad_scale: 32.0 2023-04-27 04:21:21,453 INFO [finetune.py:976] (6/7) Epoch 13, batch 2800, loss[loss=0.1554, simple_loss=0.224, pruned_loss=0.04338, over 4829.00 frames. ], tot_loss[loss=0.1828, simple_loss=0.2507, pruned_loss=0.05741, over 953458.10 frames. ], batch size: 39, lr: 3.59e-03, grad_scale: 32.0 2023-04-27 04:21:33,191 INFO [optim.py:369] (6/7) 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 04:21:46,944 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([4.3984, 4.2793, 2.9306, 5.0785, 4.4303, 4.3889, 1.9880, 4.3277], device='cuda:6'), covar=tensor([0.1630, 0.0917, 0.3541, 0.0996, 0.4286, 0.1741, 0.5781, 0.2107], device='cuda:6'), in_proj_covar=tensor([0.0242, 0.0214, 0.0248, 0.0300, 0.0295, 0.0245, 0.0269, 0.0268], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 04:22:06,642 INFO [finetune.py:976] (6/7) Epoch 13, batch 2850, loss[loss=0.2001, simple_loss=0.2675, pruned_loss=0.06632, over 4766.00 frames. ], tot_loss[loss=0.1818, simple_loss=0.2496, pruned_loss=0.05703, over 955224.60 frames. ], batch size: 28, lr: 3.59e-03, grad_scale: 32.0 2023-04-27 04:22:08,635 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.1619, 1.6214, 2.0750, 2.4935, 2.0910, 1.6232, 1.3201, 1.8828], device='cuda:6'), covar=tensor([0.3649, 0.3674, 0.1902, 0.2639, 0.2904, 0.2896, 0.4624, 0.2434], device='cuda:6'), in_proj_covar=tensor([0.0284, 0.0245, 0.0221, 0.0313, 0.0213, 0.0228, 0.0228, 0.0183], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:6') 2023-04-27 04:22:13,474 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5376, 1.3499, 1.7024, 1.7285, 1.3701, 1.2292, 1.3090, 0.8739], device='cuda:6'), covar=tensor([0.0562, 0.0898, 0.0479, 0.0795, 0.0779, 0.1292, 0.0733, 0.0756], device='cuda:6'), in_proj_covar=tensor([0.0067, 0.0070, 0.0070, 0.0067, 0.0075, 0.0096, 0.0075, 0.0069], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-27 04:22:21,565 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.4526, 1.7553, 1.3069, 1.0474, 1.0901, 1.0616, 1.3376, 1.0417], device='cuda:6'), covar=tensor([0.1842, 0.1296, 0.1610, 0.1886, 0.2628, 0.2092, 0.1143, 0.2199], device='cuda:6'), in_proj_covar=tensor([0.0195, 0.0212, 0.0168, 0.0202, 0.0201, 0.0183, 0.0156, 0.0184], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:6') 2023-04-27 04:22:28,018 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=71616.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 04:22:30,853 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.60 vs. limit=2.0 2023-04-27 04:22:40,694 INFO [finetune.py:976] (6/7) Epoch 13, batch 2900, loss[loss=0.1699, simple_loss=0.2474, pruned_loss=0.04619, over 4825.00 frames. ], tot_loss[loss=0.1848, simple_loss=0.2529, pruned_loss=0.05832, over 955521.21 frames. ], batch size: 33, lr: 3.59e-03, grad_scale: 64.0 2023-04-27 04:22:41,461 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.3258, 1.6283, 1.6907, 1.8694, 1.6829, 1.7400, 1.8104, 1.7514], device='cuda:6'), covar=tensor([0.4810, 0.6430, 0.5314, 0.4853, 0.6124, 0.8019, 0.5897, 0.5711], device='cuda:6'), in_proj_covar=tensor([0.0330, 0.0377, 0.0318, 0.0331, 0.0341, 0.0399, 0.0359, 0.0325], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 04:22:46,792 INFO [optim.py:369] (6/7) 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 04:22:49,912 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=71648.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 04:23:09,473 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=71677.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 04:23:13,482 INFO [finetune.py:976] (6/7) Epoch 13, batch 2950, loss[loss=0.2243, simple_loss=0.2918, pruned_loss=0.07844, over 4823.00 frames. ], tot_loss[loss=0.1874, simple_loss=0.2563, pruned_loss=0.05923, over 956545.75 frames. ], batch size: 51, lr: 3.59e-03, grad_scale: 64.0 2023-04-27 04:23:15,369 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=71686.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 04:23:19,635 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.1683, 2.5501, 0.7153, 1.4764, 1.5352, 1.8432, 1.6450, 0.8564], device='cuda:6'), covar=tensor([0.1392, 0.1080, 0.1799, 0.1288, 0.1097, 0.0946, 0.1336, 0.1596], device='cuda:6'), in_proj_covar=tensor([0.0117, 0.0246, 0.0139, 0.0121, 0.0133, 0.0153, 0.0117, 0.0120], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-27 04:23:21,437 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 13, batch 3000, loss[loss=0.222, simple_loss=0.3033, pruned_loss=0.07034, over 4812.00 frames. ], tot_loss[loss=0.1887, simple_loss=0.2579, pruned_loss=0.05978, over 958245.19 frames. ], batch size: 40, lr: 3.59e-03, grad_scale: 64.0 2023-04-27 04:23:45,590 INFO [finetune.py:1001] (6/7) Computing validation loss 2023-04-27 04:23:50,963 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7859, 1.6258, 1.8947, 2.1629, 2.1460, 1.6669, 1.2985, 1.9703], device='cuda:6'), covar=tensor([0.0789, 0.1177, 0.0687, 0.0576, 0.0558, 0.0923, 0.0898, 0.0509], device='cuda:6'), in_proj_covar=tensor([0.0193, 0.0202, 0.0184, 0.0175, 0.0180, 0.0185, 0.0157, 0.0181], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 04:23:56,065 INFO [finetune.py:1010] (6/7) 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] (6/7) Maximum memory allocated so far is 6345MB 2023-04-27 04:23:56,746 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=71734.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 04:24:03,111 INFO [optim.py:369] (6/7) 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 04:24:19,775 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0689, 1.9526, 2.4354, 2.5625, 1.8333, 1.6673, 1.9390, 1.0713], device='cuda:6'), covar=tensor([0.0622, 0.0849, 0.0493, 0.0840, 0.0917, 0.1371, 0.0790, 0.1001], device='cuda:6'), in_proj_covar=tensor([0.0068, 0.0070, 0.0070, 0.0068, 0.0075, 0.0097, 0.0075, 0.0069], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-27 04:24:27,545 INFO [finetune.py:976] (6/7) Epoch 13, batch 3050, loss[loss=0.1893, simple_loss=0.2583, pruned_loss=0.06016, over 4825.00 frames. ], tot_loss[loss=0.1887, simple_loss=0.2585, pruned_loss=0.05946, over 958713.66 frames. ], batch size: 49, lr: 3.59e-03, grad_scale: 64.0 2023-04-27 04:25:00,525 INFO [finetune.py:976] (6/7) Epoch 13, batch 3100, loss[loss=0.1545, simple_loss=0.2231, pruned_loss=0.04298, over 4821.00 frames. ], tot_loss[loss=0.187, simple_loss=0.2562, pruned_loss=0.05894, over 957891.78 frames. ], batch size: 38, lr: 3.59e-03, grad_scale: 64.0 2023-04-27 04:25:08,972 INFO [optim.py:369] (6/7) 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 04:25:54,816 INFO [finetune.py:976] (6/7) Epoch 13, batch 3150, loss[loss=0.1967, simple_loss=0.2601, pruned_loss=0.06671, over 4930.00 frames. ], tot_loss[loss=0.1855, simple_loss=0.2535, pruned_loss=0.05877, over 957082.37 frames. ], batch size: 43, lr: 3.59e-03, grad_scale: 64.0 2023-04-27 04:26:39,778 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5660, 2.6065, 2.0002, 2.3691, 2.6665, 2.1179, 3.2082, 1.8650], device='cuda:6'), covar=tensor([0.3577, 0.1889, 0.4045, 0.3350, 0.1613, 0.2556, 0.1826, 0.4200], device='cuda:6'), in_proj_covar=tensor([0.0343, 0.0345, 0.0427, 0.0356, 0.0379, 0.0381, 0.0373, 0.0417], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 04:26:48,055 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=71919.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 04:27:02,052 INFO [finetune.py:976] (6/7) Epoch 13, batch 3200, loss[loss=0.1614, simple_loss=0.2469, pruned_loss=0.03796, over 4899.00 frames. ], tot_loss[loss=0.1821, simple_loss=0.25, pruned_loss=0.05706, over 956259.43 frames. ], batch size: 43, lr: 3.59e-03, grad_scale: 64.0 2023-04-27 04:27:08,183 INFO [optim.py:369] (6/7) 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 04:27:24,651 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.2065, 1.6297, 2.1478, 2.6829, 2.1719, 1.6591, 1.4483, 1.9435], device='cuda:6'), covar=tensor([0.3436, 0.3394, 0.1574, 0.2286, 0.2797, 0.2770, 0.4419, 0.2298], device='cuda:6'), in_proj_covar=tensor([0.0284, 0.0246, 0.0221, 0.0314, 0.0213, 0.0228, 0.0229, 0.0184], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:6') 2023-04-27 04:27:25,187 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([4.6965, 4.6294, 3.3063, 5.3575, 4.6672, 4.6401, 1.8576, 4.5077], device='cuda:6'), covar=tensor([0.1594, 0.0966, 0.3152, 0.0927, 0.4352, 0.1686, 0.6083, 0.2473], device='cuda:6'), in_proj_covar=tensor([0.0242, 0.0213, 0.0247, 0.0300, 0.0296, 0.0246, 0.0269, 0.0268], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 04:27:28,804 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=71972.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 04:27:31,972 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([3.1955, 2.2695, 2.3100, 2.3885, 2.0905, 2.3083, 2.4136, 2.3952], device='cuda:6'), covar=tensor([0.4475, 0.7355, 0.5806, 0.5789, 0.6848, 0.8227, 0.6325, 0.6024], device='cuda:6'), in_proj_covar=tensor([0.0331, 0.0379, 0.0319, 0.0332, 0.0343, 0.0401, 0.0359, 0.0326], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 04:27:33,776 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 13, batch 3250, loss[loss=0.1891, simple_loss=0.2598, pruned_loss=0.05922, over 4846.00 frames. ], tot_loss[loss=0.1831, simple_loss=0.251, pruned_loss=0.05762, over 957031.96 frames. ], batch size: 44, lr: 3.59e-03, grad_scale: 64.0 2023-04-27 04:27:36,187 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.4052, 2.1435, 1.7592, 1.9111, 2.2139, 1.8400, 2.4347, 1.6080], device='cuda:6'), covar=tensor([0.2780, 0.1374, 0.3864, 0.2241, 0.1452, 0.1900, 0.1393, 0.3461], device='cuda:6'), in_proj_covar=tensor([0.0344, 0.0346, 0.0428, 0.0358, 0.0380, 0.0383, 0.0373, 0.0419], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 04:28:10,359 INFO [finetune.py:976] (6/7) Epoch 13, batch 3300, loss[loss=0.1933, simple_loss=0.2682, pruned_loss=0.05921, over 4819.00 frames. ], tot_loss[loss=0.1854, simple_loss=0.2537, pruned_loss=0.05851, over 953250.50 frames. ], batch size: 33, lr: 3.59e-03, grad_scale: 32.0 2023-04-27 04:28:15,338 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5500, 0.9477, 1.6409, 1.9882, 1.6679, 1.5181, 1.5429, 1.5721], device='cuda:6'), covar=tensor([0.5093, 0.7085, 0.6697, 0.6908, 0.6366, 0.8441, 0.8409, 0.8723], device='cuda:6'), in_proj_covar=tensor([0.0409, 0.0406, 0.0493, 0.0511, 0.0439, 0.0460, 0.0467, 0.0470], device='cuda:6'), out_proj_covar=tensor([9.9172e-05, 1.0055e-04, 1.1083e-04, 1.2138e-04, 1.0606e-04, 1.1089e-04, 1.1161e-04, 1.1254e-04], device='cuda:6') 2023-04-27 04:28:16,993 INFO [optim.py:369] (6/7) 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] (6/7) Epoch 13, batch 3350, loss[loss=0.1957, simple_loss=0.2702, pruned_loss=0.06058, over 4779.00 frames. ], tot_loss[loss=0.1867, simple_loss=0.2561, pruned_loss=0.05861, over 954496.91 frames. ], batch size: 54, lr: 3.59e-03, grad_scale: 32.0 2023-04-27 04:28:44,120 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8765, 1.5574, 1.4352, 1.7635, 2.0909, 1.7261, 1.4513, 1.3731], device='cuda:6'), covar=tensor([0.1778, 0.1537, 0.1821, 0.1415, 0.0894, 0.1555, 0.2233, 0.2163], device='cuda:6'), in_proj_covar=tensor([0.0307, 0.0318, 0.0353, 0.0294, 0.0331, 0.0314, 0.0308, 0.0362], device='cuda:6'), out_proj_covar=tensor([6.3995e-05, 6.6667e-05, 7.5836e-05, 6.0224e-05, 6.9170e-05, 6.6727e-05, 6.5314e-05, 7.7567e-05], device='cuda:6') 2023-04-27 04:28:51,448 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.1547, 2.1730, 1.7431, 1.7861, 2.1615, 1.6721, 2.6120, 1.4901], device='cuda:6'), covar=tensor([0.3575, 0.1776, 0.4473, 0.2968, 0.1901, 0.2519, 0.1625, 0.4409], device='cuda:6'), in_proj_covar=tensor([0.0345, 0.0347, 0.0429, 0.0358, 0.0381, 0.0383, 0.0373, 0.0420], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 04:29:15,398 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=72129.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 04:29:17,730 INFO [finetune.py:976] (6/7) Epoch 13, batch 3400, loss[loss=0.147, simple_loss=0.2127, pruned_loss=0.04061, over 4725.00 frames. ], tot_loss[loss=0.1889, simple_loss=0.2577, pruned_loss=0.06009, over 952912.67 frames. ], batch size: 23, lr: 3.59e-03, grad_scale: 32.0 2023-04-27 04:29:24,412 INFO [optim.py:369] (6/7) 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 04:29:25,752 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8516, 1.3328, 1.4523, 1.5570, 1.9806, 1.6251, 1.3103, 1.3636], device='cuda:6'), covar=tensor([0.1466, 0.1433, 0.1902, 0.1264, 0.0797, 0.1299, 0.1983, 0.1881], device='cuda:6'), in_proj_covar=tensor([0.0306, 0.0316, 0.0351, 0.0293, 0.0329, 0.0313, 0.0306, 0.0362], device='cuda:6'), out_proj_covar=tensor([6.3770e-05, 6.6433e-05, 7.5445e-05, 6.0048e-05, 6.8753e-05, 6.6468e-05, 6.4993e-05, 7.7360e-05], device='cuda:6') 2023-04-27 04:29:51,377 INFO [finetune.py:976] (6/7) Epoch 13, batch 3450, loss[loss=0.1894, simple_loss=0.2596, pruned_loss=0.0596, over 4800.00 frames. ], tot_loss[loss=0.1884, simple_loss=0.2577, pruned_loss=0.05958, over 953830.23 frames. ], batch size: 25, lr: 3.59e-03, grad_scale: 32.0 2023-04-27 04:29:55,737 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=72190.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 04:30:01,989 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=3.68 vs. limit=5.0 2023-04-27 04:30:02,524 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.3522, 1.2684, 1.6972, 1.6026, 1.2755, 1.1239, 1.3538, 0.9634], device='cuda:6'), covar=tensor([0.0605, 0.0622, 0.0430, 0.0553, 0.0733, 0.1157, 0.0602, 0.0618], device='cuda:6'), in_proj_covar=tensor([0.0068, 0.0070, 0.0070, 0.0068, 0.0075, 0.0096, 0.0075, 0.0069], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-27 04:30:24,560 INFO [finetune.py:976] (6/7) Epoch 13, batch 3500, loss[loss=0.1446, simple_loss=0.2204, pruned_loss=0.03446, over 4902.00 frames. ], tot_loss[loss=0.1863, simple_loss=0.2547, pruned_loss=0.05895, over 953050.03 frames. ], batch size: 36, lr: 3.59e-03, grad_scale: 32.0 2023-04-27 04:30:31,098 INFO [optim.py:369] (6/7) 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 04:31:07,545 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.63 vs. limit=2.0 2023-04-27 04:31:08,049 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=72272.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 04:31:09,865 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=72275.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 04:31:12,380 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.1448, 1.5437, 1.4774, 1.7534, 1.5788, 1.8890, 1.3856, 3.2745], device='cuda:6'), covar=tensor([0.0643, 0.0751, 0.0760, 0.1120, 0.0662, 0.0530, 0.0729, 0.0160], device='cuda:6'), in_proj_covar=tensor([0.0038, 0.0039, 0.0040, 0.0043, 0.0040, 0.0038, 0.0038, 0.0057], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0014, 0.0013, 0.0012, 0.0013, 0.0016], device='cuda:6') 2023-04-27 04:31:20,405 INFO [finetune.py:976] (6/7) Epoch 13, batch 3550, loss[loss=0.1569, simple_loss=0.2319, pruned_loss=0.04097, over 4772.00 frames. ], tot_loss[loss=0.1822, simple_loss=0.2502, pruned_loss=0.0571, over 953434.14 frames. ], batch size: 28, lr: 3.59e-03, grad_scale: 32.0 2023-04-27 04:32:06,132 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=72320.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 04:32:16,768 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 13, batch 3600, loss[loss=0.156, simple_loss=0.2287, pruned_loss=0.04162, over 4762.00 frames. ], tot_loss[loss=0.1815, simple_loss=0.2492, pruned_loss=0.05689, over 954816.43 frames. ], batch size: 28, lr: 3.59e-03, grad_scale: 32.0 2023-04-27 04:32:28,561 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-04-27 04:32:33,138 INFO [optim.py:369] (6/7) 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 04:32:59,900 INFO [finetune.py:976] (6/7) Epoch 13, batch 3650, loss[loss=0.2024, simple_loss=0.2785, pruned_loss=0.06318, over 4829.00 frames. ], tot_loss[loss=0.1838, simple_loss=0.2517, pruned_loss=0.05798, over 953061.41 frames. ], batch size: 40, lr: 3.59e-03, grad_scale: 32.0 2023-04-27 04:33:02,536 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=72387.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 04:33:03,190 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.1858, 1.6610, 2.0286, 2.2848, 2.0362, 1.6117, 1.1248, 1.7214], device='cuda:6'), covar=tensor([0.3695, 0.3598, 0.1797, 0.2371, 0.2746, 0.2873, 0.4406, 0.2293], device='cuda:6'), in_proj_covar=tensor([0.0284, 0.0246, 0.0221, 0.0314, 0.0213, 0.0228, 0.0229, 0.0184], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:6') 2023-04-27 04:33:13,810 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.1861, 1.5990, 2.0938, 2.5101, 2.0100, 1.5749, 1.3152, 1.8787], device='cuda:6'), covar=tensor([0.3394, 0.3608, 0.1652, 0.2319, 0.3028, 0.2891, 0.4340, 0.2126], device='cuda:6'), in_proj_covar=tensor([0.0285, 0.0247, 0.0222, 0.0316, 0.0214, 0.0229, 0.0230, 0.0185], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:6') 2023-04-27 04:33:33,725 INFO [finetune.py:976] (6/7) Epoch 13, batch 3700, loss[loss=0.1805, simple_loss=0.2602, pruned_loss=0.05034, over 4819.00 frames. ], tot_loss[loss=0.184, simple_loss=0.2527, pruned_loss=0.05772, over 952921.47 frames. ], batch size: 39, lr: 3.59e-03, grad_scale: 32.0 2023-04-27 04:33:40,457 INFO [optim.py:369] (6/7) 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 04:34:07,019 INFO [finetune.py:976] (6/7) Epoch 13, batch 3750, loss[loss=0.2167, simple_loss=0.2959, pruned_loss=0.06873, over 4850.00 frames. ], tot_loss[loss=0.1871, simple_loss=0.2555, pruned_loss=0.05932, over 952596.47 frames. ], batch size: 49, lr: 3.59e-03, grad_scale: 32.0 2023-04-27 04:34:08,311 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=72485.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 04:34:39,242 INFO [finetune.py:976] (6/7) Epoch 13, batch 3800, loss[loss=0.1544, simple_loss=0.2367, pruned_loss=0.03606, over 4921.00 frames. ], tot_loss[loss=0.1872, simple_loss=0.2561, pruned_loss=0.05913, over 952890.69 frames. ], batch size: 42, lr: 3.59e-03, grad_scale: 32.0 2023-04-27 04:34:46,448 INFO [optim.py:369] (6/7) 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 04:34:59,143 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.57 vs. limit=2.0 2023-04-27 04:35:06,307 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=72575.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 04:35:09,159 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-04-27 04:35:11,559 INFO [finetune.py:976] (6/7) Epoch 13, batch 3850, loss[loss=0.1404, simple_loss=0.2206, pruned_loss=0.03014, over 4776.00 frames. ], tot_loss[loss=0.1853, simple_loss=0.2545, pruned_loss=0.05807, over 953010.79 frames. ], batch size: 28, lr: 3.59e-03, grad_scale: 32.0 2023-04-27 04:35:37,011 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 13, batch 3900, loss[loss=0.1644, simple_loss=0.2385, pruned_loss=0.04511, over 4911.00 frames. ], tot_loss[loss=0.1824, simple_loss=0.2514, pruned_loss=0.05671, over 954470.05 frames. ], batch size: 43, lr: 3.59e-03, grad_scale: 32.0 2023-04-27 04:35:48,122 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8679, 2.0781, 1.7352, 1.4207, 1.3371, 1.3719, 1.7204, 1.2413], device='cuda:6'), covar=tensor([0.1688, 0.1388, 0.1573, 0.1984, 0.2471, 0.2149, 0.1175, 0.2232], device='cuda:6'), in_proj_covar=tensor([0.0195, 0.0214, 0.0169, 0.0204, 0.0202, 0.0184, 0.0157, 0.0186], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:6') 2023-04-27 04:35:51,448 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([5.0344, 4.9653, 3.5605, 5.7535, 4.9796, 5.0483, 2.8889, 5.0004], device='cuda:6'), covar=tensor([0.1673, 0.0783, 0.2484, 0.0836, 0.2978, 0.1656, 0.4479, 0.1877], device='cuda:6'), in_proj_covar=tensor([0.0242, 0.0214, 0.0249, 0.0302, 0.0295, 0.0248, 0.0270, 0.0271], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 04:35:51,958 INFO [optim.py:369] (6/7) 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 04:36:32,498 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=72682.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 04:36:33,066 INFO [finetune.py:976] (6/7) Epoch 13, batch 3950, loss[loss=0.1576, simple_loss=0.2253, pruned_loss=0.04496, over 4760.00 frames. ], tot_loss[loss=0.1795, simple_loss=0.2478, pruned_loss=0.05557, over 953030.40 frames. ], batch size: 28, lr: 3.58e-03, grad_scale: 32.0 2023-04-27 04:37:38,716 INFO [finetune.py:976] (6/7) Epoch 13, batch 4000, loss[loss=0.1807, simple_loss=0.2556, pruned_loss=0.05295, over 4859.00 frames. ], tot_loss[loss=0.1781, simple_loss=0.2459, pruned_loss=0.05516, over 953123.90 frames. ], batch size: 49, lr: 3.58e-03, grad_scale: 32.0 2023-04-27 04:37:56,032 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([3.7603, 3.7576, 2.6502, 4.4392, 3.8812, 3.7640, 1.6946, 3.8641], device='cuda:6'), covar=tensor([0.1650, 0.1128, 0.3015, 0.1642, 0.3868, 0.1814, 0.5963, 0.2468], device='cuda:6'), in_proj_covar=tensor([0.0241, 0.0213, 0.0249, 0.0301, 0.0295, 0.0247, 0.0269, 0.0269], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 04:37:57,171 INFO [optim.py:369] (6/7) 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 04:38:21,827 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0424, 1.7971, 2.0599, 2.4548, 2.4732, 2.0492, 1.6415, 2.1249], device='cuda:6'), covar=tensor([0.0918, 0.1148, 0.0692, 0.0556, 0.0571, 0.0777, 0.0789, 0.0577], device='cuda:6'), in_proj_covar=tensor([0.0190, 0.0201, 0.0180, 0.0172, 0.0176, 0.0181, 0.0154, 0.0178], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 04:38:27,849 INFO [finetune.py:976] (6/7) Epoch 13, batch 4050, loss[loss=0.1974, simple_loss=0.2717, pruned_loss=0.06154, over 4912.00 frames. ], tot_loss[loss=0.1823, simple_loss=0.2506, pruned_loss=0.05701, over 952799.13 frames. ], batch size: 43, lr: 3.58e-03, grad_scale: 32.0 2023-04-27 04:38:29,646 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=72785.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 04:38:30,329 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.3432, 1.5459, 1.7052, 1.8594, 1.6244, 1.7003, 1.7902, 1.7391], device='cuda:6'), covar=tensor([0.4168, 0.6348, 0.5258, 0.5001, 0.6249, 0.8508, 0.6372, 0.5558], device='cuda:6'), in_proj_covar=tensor([0.0330, 0.0377, 0.0319, 0.0330, 0.0343, 0.0399, 0.0358, 0.0326], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 04:38:58,992 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.2524, 1.6605, 1.5535, 1.8378, 1.7937, 2.0696, 1.5516, 3.9823], device='cuda:6'), covar=tensor([0.0559, 0.0759, 0.0744, 0.1180, 0.0598, 0.0548, 0.0690, 0.0128], device='cuda:6'), in_proj_covar=tensor([0.0038, 0.0039, 0.0039, 0.0043, 0.0040, 0.0038, 0.0038, 0.0057], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0014, 0.0013, 0.0012, 0.0013, 0.0016], device='cuda:6') 2023-04-27 04:39:00,859 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.4292, 1.4738, 1.8890, 1.8288, 1.3743, 1.2099, 1.6208, 1.0684], device='cuda:6'), covar=tensor([0.0689, 0.0771, 0.0475, 0.0838, 0.1026, 0.1255, 0.0760, 0.0750], device='cuda:6'), in_proj_covar=tensor([0.0068, 0.0070, 0.0070, 0.0068, 0.0075, 0.0096, 0.0075, 0.0069], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-27 04:39:01,351 INFO [finetune.py:976] (6/7) Epoch 13, batch 4100, loss[loss=0.179, simple_loss=0.2494, pruned_loss=0.0543, over 4818.00 frames. ], tot_loss[loss=0.1836, simple_loss=0.2531, pruned_loss=0.05704, over 955602.44 frames. ], batch size: 40, lr: 3.58e-03, grad_scale: 32.0 2023-04-27 04:39:01,417 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=72833.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 04:39:09,031 INFO [optim.py:369] (6/7) 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] (6/7) Epoch 13, batch 4150, loss[loss=0.1977, simple_loss=0.2628, pruned_loss=0.06629, over 4725.00 frames. ], tot_loss[loss=0.1851, simple_loss=0.2544, pruned_loss=0.05796, over 955616.93 frames. ], batch size: 54, lr: 3.58e-03, grad_scale: 32.0 2023-04-27 04:39:42,048 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=72894.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 04:40:08,514 INFO [finetune.py:976] (6/7) Epoch 13, batch 4200, loss[loss=0.1774, simple_loss=0.253, pruned_loss=0.05092, over 4765.00 frames. ], tot_loss[loss=0.185, simple_loss=0.2547, pruned_loss=0.05764, over 956378.53 frames. ], batch size: 28, lr: 3.58e-03, grad_scale: 32.0 2023-04-27 04:40:15,133 INFO [optim.py:369] (6/7) 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 04:40:20,267 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8568, 1.7146, 2.0026, 2.2253, 2.2703, 1.8509, 1.4196, 1.9422], device='cuda:6'), covar=tensor([0.0851, 0.1187, 0.0642, 0.0611, 0.0621, 0.0832, 0.0834, 0.0579], device='cuda:6'), in_proj_covar=tensor([0.0190, 0.0201, 0.0181, 0.0172, 0.0176, 0.0182, 0.0155, 0.0179], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 04:40:24,271 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=72955.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 04:40:30,242 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=72964.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 04:40:30,439 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.21 vs. limit=5.0 2023-04-27 04:40:41,148 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=72982.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 04:40:41,682 INFO [finetune.py:976] (6/7) Epoch 13, batch 4250, loss[loss=0.1469, simple_loss=0.2099, pruned_loss=0.04191, over 4161.00 frames. ], tot_loss[loss=0.1832, simple_loss=0.2523, pruned_loss=0.05706, over 955336.31 frames. ], batch size: 18, lr: 3.58e-03, grad_scale: 32.0 2023-04-27 04:41:10,639 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=73025.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 04:41:13,600 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=73030.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 04:41:15,360 INFO [finetune.py:976] (6/7) Epoch 13, batch 4300, loss[loss=0.2066, simple_loss=0.2607, pruned_loss=0.07623, over 4910.00 frames. ], tot_loss[loss=0.1816, simple_loss=0.2502, pruned_loss=0.05652, over 955408.80 frames. ], batch size: 43, lr: 3.58e-03, grad_scale: 32.0 2023-04-27 04:41:22,016 INFO [optim.py:369] (6/7) 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 04:41:59,414 INFO [finetune.py:976] (6/7) Epoch 13, batch 4350, loss[loss=0.165, simple_loss=0.2291, pruned_loss=0.0504, over 4912.00 frames. ], tot_loss[loss=0.18, simple_loss=0.248, pruned_loss=0.05596, over 956884.49 frames. ], batch size: 36, lr: 3.58e-03, grad_scale: 32.0 2023-04-27 04:42:11,859 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8273, 2.0132, 1.1331, 1.6094, 1.8845, 1.6991, 1.6341, 1.7200], device='cuda:6'), covar=tensor([0.0493, 0.0316, 0.0332, 0.0523, 0.0246, 0.0494, 0.0468, 0.0516], device='cuda:6'), in_proj_covar=tensor([0.0029, 0.0025, 0.0023, 0.0029, 0.0020, 0.0029, 0.0028, 0.0029], device='cuda:6'), out_proj_covar=tensor([0.0050, 0.0044, 0.0037, 0.0050, 0.0037, 0.0048, 0.0048, 0.0050], device='cuda:6') 2023-04-27 04:42:38,502 INFO [finetune.py:976] (6/7) Epoch 13, batch 4400, loss[loss=0.2074, simple_loss=0.2705, pruned_loss=0.07219, over 4814.00 frames. ], tot_loss[loss=0.1833, simple_loss=0.2504, pruned_loss=0.05813, over 954486.82 frames. ], batch size: 33, lr: 3.58e-03, grad_scale: 32.0 2023-04-27 04:42:47,734 INFO [optim.py:369] (6/7) 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 04:43:39,919 INFO [finetune.py:976] (6/7) Epoch 13, batch 4450, loss[loss=0.178, simple_loss=0.2488, pruned_loss=0.05356, over 4768.00 frames. ], tot_loss[loss=0.1855, simple_loss=0.2531, pruned_loss=0.05896, over 952969.14 frames. ], batch size: 26, lr: 3.58e-03, grad_scale: 32.0 2023-04-27 04:44:03,210 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.1800, 2.6256, 1.0705, 1.4324, 2.0397, 1.3315, 3.4922, 1.8690], device='cuda:6'), covar=tensor([0.0613, 0.0617, 0.0791, 0.1282, 0.0530, 0.0987, 0.0319, 0.0657], device='cuda:6'), in_proj_covar=tensor([0.0051, 0.0066, 0.0049, 0.0047, 0.0051, 0.0052, 0.0076, 0.0051], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:6') 2023-04-27 04:44:27,565 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.22 vs. limit=5.0 2023-04-27 04:44:30,377 INFO [finetune.py:976] (6/7) Epoch 13, batch 4500, loss[loss=0.2094, simple_loss=0.2827, pruned_loss=0.06803, over 4793.00 frames. ], tot_loss[loss=0.1864, simple_loss=0.2545, pruned_loss=0.0592, over 950941.93 frames. ], batch size: 45, lr: 3.58e-03, grad_scale: 32.0 2023-04-27 04:44:37,126 INFO [optim.py:369] (6/7) 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] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=73250.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 04:45:04,241 INFO [finetune.py:976] (6/7) Epoch 13, batch 4550, loss[loss=0.2282, simple_loss=0.2916, pruned_loss=0.0824, over 4790.00 frames. ], tot_loss[loss=0.1888, simple_loss=0.257, pruned_loss=0.06027, over 951048.27 frames. ], batch size: 45, lr: 3.58e-03, grad_scale: 32.0 2023-04-27 04:45:10,384 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=73293.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 04:45:22,429 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.0294, 1.0266, 1.2520, 1.1509, 0.9989, 0.9105, 0.9716, 0.6071], device='cuda:6'), covar=tensor([0.0626, 0.0595, 0.0544, 0.0593, 0.0739, 0.1127, 0.0514, 0.0730], device='cuda:6'), in_proj_covar=tensor([0.0068, 0.0070, 0.0070, 0.0068, 0.0075, 0.0096, 0.0075, 0.0069], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-27 04:45:28,146 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=73320.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 04:45:29,274 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.4280, 3.0647, 2.2848, 2.3314, 1.8101, 1.6011, 2.4784, 1.7068], device='cuda:6'), covar=tensor([0.1556, 0.1316, 0.1489, 0.1682, 0.2176, 0.1927, 0.0992, 0.2043], device='cuda:6'), in_proj_covar=tensor([0.0194, 0.0212, 0.0168, 0.0202, 0.0200, 0.0182, 0.0156, 0.0185], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:6') 2023-04-27 04:45:37,748 INFO [finetune.py:976] (6/7) Epoch 13, batch 4600, loss[loss=0.1896, simple_loss=0.2447, pruned_loss=0.06731, over 4863.00 frames. ], tot_loss[loss=0.187, simple_loss=0.2553, pruned_loss=0.05932, over 952361.34 frames. ], batch size: 31, lr: 3.58e-03, grad_scale: 32.0 2023-04-27 04:45:44,462 INFO [optim.py:369] (6/7) 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 04:45:50,683 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=73354.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 04:45:54,997 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.5377, 1.8397, 2.3004, 2.8797, 2.3360, 1.8897, 1.8784, 2.2378], device='cuda:6'), covar=tensor([0.3035, 0.3354, 0.1700, 0.2378, 0.2735, 0.2688, 0.3801, 0.2378], device='cuda:6'), in_proj_covar=tensor([0.0289, 0.0250, 0.0224, 0.0318, 0.0216, 0.0231, 0.0233, 0.0187], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:6') 2023-04-27 04:46:04,243 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.4835, 1.0894, 1.2951, 1.1740, 1.6481, 1.3013, 1.0741, 1.2152], device='cuda:6'), covar=tensor([0.1477, 0.1432, 0.1786, 0.1416, 0.0788, 0.1402, 0.1817, 0.2069], device='cuda:6'), in_proj_covar=tensor([0.0305, 0.0316, 0.0350, 0.0292, 0.0327, 0.0314, 0.0304, 0.0359], device='cuda:6'), out_proj_covar=tensor([6.3568e-05, 6.6461e-05, 7.5114e-05, 5.9676e-05, 6.8215e-05, 6.6706e-05, 6.4409e-05, 7.6727e-05], device='cuda:6') 2023-04-27 04:46:10,741 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2023-04-27 04:46:11,138 INFO [finetune.py:976] (6/7) Epoch 13, batch 4650, loss[loss=0.1499, simple_loss=0.2198, pruned_loss=0.03998, over 4767.00 frames. ], tot_loss[loss=0.1843, simple_loss=0.2525, pruned_loss=0.05804, over 953938.20 frames. ], batch size: 28, lr: 3.58e-03, grad_scale: 32.0 2023-04-27 04:46:26,467 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6433, 1.2820, 0.5541, 1.3109, 1.3842, 1.4957, 1.3838, 1.3923], device='cuda:6'), covar=tensor([0.0485, 0.0373, 0.0388, 0.0543, 0.0280, 0.0492, 0.0483, 0.0568], device='cuda:6'), in_proj_covar=tensor([0.0029, 0.0025, 0.0023, 0.0029, 0.0020, 0.0029, 0.0028, 0.0029], device='cuda:6'), out_proj_covar=tensor([0.0050, 0.0044, 0.0037, 0.0050, 0.0037, 0.0049, 0.0049, 0.0050], device='cuda:6') 2023-04-27 04:46:30,134 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9991, 2.8171, 1.8581, 2.0033, 1.4904, 1.4086, 1.9364, 1.2943], device='cuda:6'), covar=tensor([0.1999, 0.1554, 0.1721, 0.1939, 0.2648, 0.2321, 0.1198, 0.2423], device='cuda:6'), in_proj_covar=tensor([0.0195, 0.0212, 0.0168, 0.0203, 0.0201, 0.0183, 0.0156, 0.0186], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:6') 2023-04-27 04:46:32,570 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.68 vs. limit=2.0 2023-04-27 04:46:44,505 INFO [finetune.py:976] (6/7) Epoch 13, batch 4700, loss[loss=0.1976, simple_loss=0.2683, pruned_loss=0.06348, over 4005.00 frames. ], tot_loss[loss=0.1815, simple_loss=0.2494, pruned_loss=0.05677, over 953623.00 frames. ], batch size: 17, lr: 3.58e-03, grad_scale: 32.0 2023-04-27 04:46:44,882 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 2023-04-27 04:46:51,531 INFO [optim.py:369] (6/7) 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 04:47:33,204 INFO [finetune.py:976] (6/7) Epoch 13, batch 4750, loss[loss=0.2034, simple_loss=0.2675, pruned_loss=0.06965, over 4874.00 frames. ], tot_loss[loss=0.1805, simple_loss=0.248, pruned_loss=0.05653, over 955612.39 frames. ], batch size: 34, lr: 3.58e-03, grad_scale: 32.0 2023-04-27 04:48:13,162 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([0.9989, 1.0323, 1.3180, 1.1513, 0.9496, 0.8866, 0.9870, 0.6528], device='cuda:6'), covar=tensor([0.0592, 0.0675, 0.0555, 0.0626, 0.0853, 0.1337, 0.0561, 0.0786], device='cuda:6'), in_proj_covar=tensor([0.0068, 0.0070, 0.0070, 0.0067, 0.0075, 0.0096, 0.0075, 0.0069], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-27 04:48:28,608 INFO [finetune.py:976] (6/7) Epoch 13, batch 4800, loss[loss=0.1708, simple_loss=0.2498, pruned_loss=0.04588, over 4752.00 frames. ], tot_loss[loss=0.1829, simple_loss=0.2509, pruned_loss=0.05746, over 956601.77 frames. ], batch size: 54, lr: 3.58e-03, grad_scale: 32.0 2023-04-27 04:48:36,791 INFO [optim.py:369] (6/7) 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 04:48:41,060 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=73550.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 04:48:49,664 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.4109, 1.7412, 1.7476, 1.8873, 1.6972, 1.7692, 1.8429, 1.7486], device='cuda:6'), covar=tensor([0.4750, 0.6406, 0.5437, 0.4981, 0.6355, 0.7774, 0.6377, 0.6069], device='cuda:6'), in_proj_covar=tensor([0.0326, 0.0371, 0.0313, 0.0326, 0.0337, 0.0393, 0.0353, 0.0320], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 04:48:58,838 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 2023-04-27 04:49:02,206 INFO [finetune.py:976] (6/7) Epoch 13, batch 4850, loss[loss=0.2184, simple_loss=0.2931, pruned_loss=0.07188, over 4843.00 frames. ], tot_loss[loss=0.1858, simple_loss=0.2542, pruned_loss=0.05871, over 953748.69 frames. ], batch size: 49, lr: 3.58e-03, grad_scale: 32.0 2023-04-27 04:49:17,648 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=73590.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 04:49:18,790 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.5076, 1.7894, 1.8586, 1.9443, 1.8017, 1.8962, 1.9611, 1.8815], device='cuda:6'), covar=tensor([0.4213, 0.5763, 0.5220, 0.5143, 0.5797, 0.7784, 0.6044, 0.5467], device='cuda:6'), in_proj_covar=tensor([0.0327, 0.0372, 0.0314, 0.0327, 0.0337, 0.0394, 0.0354, 0.0321], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 04:49:28,460 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=73598.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 04:49:40,996 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=2.13 vs. limit=2.0 2023-04-27 04:49:51,070 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.0826, 2.0411, 1.6439, 1.7958, 2.1712, 1.6869, 2.5886, 1.4798], device='cuda:6'), covar=tensor([0.3889, 0.1753, 0.4778, 0.2836, 0.1613, 0.2586, 0.1377, 0.4597], device='cuda:6'), in_proj_covar=tensor([0.0344, 0.0350, 0.0429, 0.0357, 0.0385, 0.0385, 0.0375, 0.0420], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 04:49:53,456 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=73620.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 04:50:13,624 INFO [finetune.py:976] (6/7) Epoch 13, batch 4900, loss[loss=0.1668, simple_loss=0.2373, pruned_loss=0.0482, over 4911.00 frames. ], tot_loss[loss=0.1859, simple_loss=0.2546, pruned_loss=0.05856, over 952834.91 frames. ], batch size: 29, lr: 3.58e-03, grad_scale: 32.0 2023-04-27 04:50:28,449 INFO [optim.py:369] (6/7) 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 04:50:38,434 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=73649.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 04:50:39,753 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=73651.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 04:50:50,304 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.35 vs. limit=5.0 2023-04-27 04:50:52,068 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.3780, 1.7181, 2.2480, 2.8148, 2.2065, 1.7401, 1.8245, 2.0832], device='cuda:6'), covar=tensor([0.3559, 0.4002, 0.1810, 0.2666, 0.3122, 0.3056, 0.4222, 0.2708], device='cuda:6'), in_proj_covar=tensor([0.0288, 0.0247, 0.0222, 0.0316, 0.0215, 0.0229, 0.0231, 0.0186], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:6') 2023-04-27 04:50:55,037 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=73668.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 04:51:04,151 INFO [finetune.py:976] (6/7) Epoch 13, batch 4950, loss[loss=0.1785, simple_loss=0.2496, pruned_loss=0.05369, over 4882.00 frames. ], tot_loss[loss=0.1866, simple_loss=0.2557, pruned_loss=0.05873, over 953768.20 frames. ], batch size: 32, lr: 3.58e-03, grad_scale: 32.0 2023-04-27 04:51:20,513 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.11 vs. limit=2.0 2023-04-27 04:51:23,732 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.2778, 1.8674, 2.2066, 2.6384, 2.6423, 2.1135, 1.6807, 2.3653], device='cuda:6'), covar=tensor([0.0785, 0.1176, 0.0718, 0.0517, 0.0537, 0.0840, 0.0800, 0.0534], device='cuda:6'), in_proj_covar=tensor([0.0190, 0.0202, 0.0182, 0.0173, 0.0178, 0.0183, 0.0156, 0.0180], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 04:51:36,605 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=73732.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 04:51:37,096 INFO [finetune.py:976] (6/7) Epoch 13, batch 5000, loss[loss=0.1714, simple_loss=0.244, pruned_loss=0.04946, over 4812.00 frames. ], tot_loss[loss=0.1851, simple_loss=0.2541, pruned_loss=0.05806, over 954250.05 frames. ], batch size: 41, lr: 3.58e-03, grad_scale: 32.0 2023-04-27 04:51:45,204 INFO [optim.py:369] (6/7) 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 04:52:11,152 INFO [finetune.py:976] (6/7) Epoch 13, batch 5050, loss[loss=0.1734, simple_loss=0.2398, pruned_loss=0.05354, over 4921.00 frames. ], tot_loss[loss=0.1829, simple_loss=0.2509, pruned_loss=0.05748, over 955072.75 frames. ], batch size: 43, lr: 3.58e-03, grad_scale: 32.0 2023-04-27 04:52:17,790 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 13, batch 5100, loss[loss=0.1616, simple_loss=0.244, pruned_loss=0.03963, over 4833.00 frames. ], tot_loss[loss=0.1797, simple_loss=0.2478, pruned_loss=0.05585, over 957198.98 frames. ], batch size: 47, lr: 3.57e-03, grad_scale: 32.0 2023-04-27 04:52:58,492 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.1340, 2.6927, 2.0668, 1.9894, 1.6773, 1.5135, 2.1816, 1.4960], device='cuda:6'), covar=tensor([0.1598, 0.1519, 0.1465, 0.1890, 0.2255, 0.1923, 0.1022, 0.2084], device='cuda:6'), in_proj_covar=tensor([0.0196, 0.0213, 0.0169, 0.0204, 0.0202, 0.0184, 0.0157, 0.0187], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:6') 2023-04-27 04:53:03,636 INFO [optim.py:369] (6/7) 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 04:53:14,202 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.13 vs. limit=2.0 2023-04-27 04:53:44,072 INFO [finetune.py:976] (6/7) Epoch 13, batch 5150, loss[loss=0.2051, simple_loss=0.2648, pruned_loss=0.07271, over 4842.00 frames. ], tot_loss[loss=0.18, simple_loss=0.2482, pruned_loss=0.05591, over 957494.09 frames. ], batch size: 49, lr: 3.57e-03, grad_scale: 32.0 2023-04-27 04:54:18,052 INFO [finetune.py:976] (6/7) Epoch 13, batch 5200, loss[loss=0.1937, simple_loss=0.2732, pruned_loss=0.05709, over 4903.00 frames. ], tot_loss[loss=0.1837, simple_loss=0.2524, pruned_loss=0.05748, over 955543.86 frames. ], batch size: 36, lr: 3.57e-03, grad_scale: 32.0 2023-04-27 04:54:24,746 INFO [optim.py:369] (6/7) 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 04:54:26,022 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=73946.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 04:54:28,373 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=73949.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 04:55:07,560 INFO [finetune.py:976] (6/7) Epoch 13, batch 5250, loss[loss=0.1744, simple_loss=0.2447, pruned_loss=0.05199, over 4928.00 frames. ], tot_loss[loss=0.1856, simple_loss=0.2547, pruned_loss=0.05831, over 954658.81 frames. ], batch size: 33, lr: 3.57e-03, grad_scale: 32.0 2023-04-27 04:55:10,088 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7145, 1.3945, 1.3404, 1.4444, 1.8794, 1.4912, 1.2658, 1.2634], device='cuda:6'), covar=tensor([0.1677, 0.1422, 0.1961, 0.1235, 0.0877, 0.1810, 0.2487, 0.2020], device='cuda:6'), in_proj_covar=tensor([0.0309, 0.0319, 0.0355, 0.0294, 0.0333, 0.0317, 0.0308, 0.0364], device='cuda:6'), out_proj_covar=tensor([6.4266e-05, 6.7115e-05, 7.6166e-05, 6.0176e-05, 6.9456e-05, 6.7352e-05, 6.5384e-05, 7.7913e-05], device='cuda:6') 2023-04-27 04:55:15,588 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([0.9044, 1.3178, 3.3243, 3.1364, 3.0313, 3.2312, 3.2211, 2.9358], device='cuda:6'), covar=tensor([0.7279, 0.5310, 0.1597, 0.2126, 0.1331, 0.2063, 0.1407, 0.1731], device='cuda:6'), in_proj_covar=tensor([0.0309, 0.0304, 0.0401, 0.0405, 0.0343, 0.0402, 0.0312, 0.0364], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-27 04:55:16,184 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=73997.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 04:55:34,603 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6492, 1.2987, 4.2439, 3.9780, 3.8031, 3.9756, 3.8975, 3.7498], device='cuda:6'), covar=tensor([0.7036, 0.5916, 0.0953, 0.1609, 0.0958, 0.1325, 0.1630, 0.1357], device='cuda:6'), in_proj_covar=tensor([0.0309, 0.0305, 0.0401, 0.0405, 0.0343, 0.0402, 0.0312, 0.0364], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-27 04:55:42,306 INFO [finetune.py:976] (6/7) Epoch 13, batch 5300, loss[loss=0.1292, simple_loss=0.2052, pruned_loss=0.02654, over 4758.00 frames. ], tot_loss[loss=0.1863, simple_loss=0.2556, pruned_loss=0.05849, over 954232.36 frames. ], batch size: 28, lr: 3.57e-03, grad_scale: 64.0 2023-04-27 04:55:54,435 INFO [optim.py:369] (6/7) 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 04:56:19,409 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=74064.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 04:56:36,837 INFO [finetune.py:976] (6/7) Epoch 13, batch 5350, loss[loss=0.1966, simple_loss=0.2628, pruned_loss=0.0652, over 4888.00 frames. ], tot_loss[loss=0.1867, simple_loss=0.256, pruned_loss=0.05865, over 955661.33 frames. ], batch size: 32, lr: 3.57e-03, grad_scale: 64.0 2023-04-27 04:56:39,880 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=74088.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 04:56:46,557 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.1638, 1.4580, 1.3304, 1.7316, 1.5318, 1.8363, 1.4066, 3.3005], device='cuda:6'), covar=tensor([0.0645, 0.0868, 0.0848, 0.1237, 0.0704, 0.0508, 0.0767, 0.0156], device='cuda:6'), in_proj_covar=tensor([0.0038, 0.0039, 0.0040, 0.0044, 0.0040, 0.0038, 0.0038, 0.0057], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:6') 2023-04-27 04:56:48,007 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=3.55 vs. limit=5.0 2023-04-27 04:57:05,257 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=74125.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 04:57:10,418 INFO [finetune.py:976] (6/7) Epoch 13, batch 5400, loss[loss=0.1602, simple_loss=0.225, pruned_loss=0.04771, over 4928.00 frames. ], tot_loss[loss=0.1854, simple_loss=0.2541, pruned_loss=0.05835, over 954823.93 frames. ], batch size: 38, lr: 3.57e-03, grad_scale: 64.0 2023-04-27 04:57:17,119 INFO [optim.py:369] (6/7) 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] (6/7) Epoch 13, batch 5450, loss[loss=0.1322, simple_loss=0.2043, pruned_loss=0.03004, over 4729.00 frames. ], tot_loss[loss=0.1829, simple_loss=0.2513, pruned_loss=0.05725, over 954510.94 frames. ], batch size: 59, lr: 3.57e-03, grad_scale: 64.0 2023-04-27 04:58:35,036 INFO [finetune.py:976] (6/7) Epoch 13, batch 5500, loss[loss=0.1207, simple_loss=0.1904, pruned_loss=0.02549, over 4743.00 frames. ], tot_loss[loss=0.1787, simple_loss=0.2469, pruned_loss=0.05527, over 956074.30 frames. ], batch size: 23, lr: 3.57e-03, grad_scale: 64.0 2023-04-27 04:58:39,445 INFO [zipformer.py:1188] (6/7) 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] (6/7) 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 04:58:43,609 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 13, batch 5550, loss[loss=0.177, simple_loss=0.2523, pruned_loss=0.05078, over 4762.00 frames. ], tot_loss[loss=0.1801, simple_loss=0.2482, pruned_loss=0.05602, over 957513.00 frames. ], batch size: 26, lr: 3.57e-03, grad_scale: 64.0 2023-04-27 04:59:15,401 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=74294.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 04:59:20,229 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=74301.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 04:59:39,644 INFO [finetune.py:976] (6/7) Epoch 13, batch 5600, loss[loss=0.2052, simple_loss=0.264, pruned_loss=0.07317, over 4801.00 frames. ], tot_loss[loss=0.1828, simple_loss=0.2515, pruned_loss=0.05705, over 956491.75 frames. ], batch size: 51, lr: 3.57e-03, grad_scale: 64.0 2023-04-27 04:59:46,068 INFO [optim.py:369] (6/7) 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 05:00:16,160 INFO [finetune.py:976] (6/7) Epoch 13, batch 5650, loss[loss=0.2158, simple_loss=0.2867, pruned_loss=0.07247, over 4709.00 frames. ], tot_loss[loss=0.1834, simple_loss=0.253, pruned_loss=0.05687, over 957170.86 frames. ], batch size: 59, lr: 3.57e-03, grad_scale: 64.0 2023-04-27 05:00:25,012 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=74388.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 05:00:44,667 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=74420.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 05:00:47,738 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-04-27 05:00:51,210 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([4.0822, 4.0593, 3.2343, 4.7630, 4.1214, 4.0630, 2.2919, 4.1110], device='cuda:6'), covar=tensor([0.1861, 0.1092, 0.2782, 0.1420, 0.4024, 0.1930, 0.5781, 0.2331], device='cuda:6'), in_proj_covar=tensor([0.0243, 0.0214, 0.0248, 0.0302, 0.0296, 0.0245, 0.0270, 0.0272], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 05:00:52,321 INFO [finetune.py:976] (6/7) Epoch 13, batch 5700, loss[loss=0.1355, simple_loss=0.2027, pruned_loss=0.03414, over 4316.00 frames. ], tot_loss[loss=0.1832, simple_loss=0.2509, pruned_loss=0.05773, over 940587.20 frames. ], batch size: 18, lr: 3.57e-03, grad_scale: 64.0 2023-04-27 05:00:54,137 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=74436.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 05:00:58,756 INFO [optim.py:369] (6/7) 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 05:01:23,647 INFO [finetune.py:976] (6/7) Epoch 14, batch 0, loss[loss=0.1552, simple_loss=0.2202, pruned_loss=0.04507, over 4810.00 frames. ], tot_loss[loss=0.1552, simple_loss=0.2202, pruned_loss=0.04507, over 4810.00 frames. ], batch size: 25, lr: 3.57e-03, grad_scale: 64.0 2023-04-27 05:01:23,648 INFO [finetune.py:1001] (6/7) Computing validation loss 2023-04-27 05:01:45,227 INFO [finetune.py:1010] (6/7) Epoch 14, validation: loss=0.1535, simple_loss=0.226, pruned_loss=0.04054, over 2265189.00 frames. 2023-04-27 05:01:45,227 INFO [finetune.py:1011] (6/7) Maximum memory allocated so far is 6345MB 2023-04-27 05:01:57,202 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6020, 1.6877, 0.7173, 1.3476, 1.5947, 1.4587, 1.3674, 1.4396], device='cuda:6'), covar=tensor([0.0509, 0.0354, 0.0387, 0.0558, 0.0281, 0.0540, 0.0512, 0.0560], device='cuda:6'), in_proj_covar=tensor([0.0029, 0.0025, 0.0023, 0.0029, 0.0020, 0.0029, 0.0028, 0.0029], device='cuda:6'), out_proj_covar=tensor([0.0050, 0.0044, 0.0038, 0.0050, 0.0037, 0.0049, 0.0049, 0.0050], device='cuda:6') 2023-04-27 05:02:18,139 INFO [finetune.py:976] (6/7) Epoch 14, batch 50, loss[loss=0.1874, simple_loss=0.2629, pruned_loss=0.05595, over 4903.00 frames. ], tot_loss[loss=0.1932, simple_loss=0.2605, pruned_loss=0.06293, over 216491.55 frames. ], batch size: 46, lr: 3.57e-03, grad_scale: 32.0 2023-04-27 05:02:41,160 INFO [optim.py:369] (6/7) 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 05:02:51,311 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([0.6122, 1.7311, 1.7381, 1.2896, 1.7839, 1.5300, 2.3240, 1.4947], device='cuda:6'), covar=tensor([0.3984, 0.1516, 0.4664, 0.2808, 0.1574, 0.2383, 0.1469, 0.4600], device='cuda:6'), in_proj_covar=tensor([0.0339, 0.0342, 0.0422, 0.0351, 0.0376, 0.0378, 0.0368, 0.0413], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 05:02:51,797 INFO [finetune.py:976] (6/7) Epoch 14, batch 100, loss[loss=0.1844, simple_loss=0.2452, pruned_loss=0.06175, over 4826.00 frames. ], tot_loss[loss=0.1833, simple_loss=0.251, pruned_loss=0.05778, over 378875.66 frames. ], batch size: 33, lr: 3.57e-03, grad_scale: 32.0 2023-04-27 05:02:53,147 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 2023-04-27 05:03:08,549 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.5091, 1.7487, 1.8523, 1.9781, 1.8103, 1.9555, 1.9829, 1.9093], device='cuda:6'), covar=tensor([0.3845, 0.5834, 0.4995, 0.4821, 0.5907, 0.7924, 0.5900, 0.5249], device='cuda:6'), in_proj_covar=tensor([0.0329, 0.0374, 0.0316, 0.0328, 0.0340, 0.0397, 0.0355, 0.0323], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 05:03:29,465 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=74595.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 05:03:30,508 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=74596.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 05:03:51,944 INFO [finetune.py:976] (6/7) Epoch 14, batch 150, loss[loss=0.1924, simple_loss=0.2534, pruned_loss=0.06565, over 4936.00 frames. ], tot_loss[loss=0.1787, simple_loss=0.2454, pruned_loss=0.05598, over 507406.19 frames. ], batch size: 33, lr: 3.57e-03, grad_scale: 32.0 2023-04-27 05:04:20,051 INFO [optim.py:369] (6/7) 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 05:04:27,986 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 14, batch 200, loss[loss=0.1687, simple_loss=0.2413, pruned_loss=0.04802, over 4170.00 frames. ], tot_loss[loss=0.1796, simple_loss=0.2461, pruned_loss=0.05659, over 605888.12 frames. ], batch size: 65, lr: 3.57e-03, grad_scale: 32.0 2023-04-27 05:04:48,835 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-04-27 05:05:05,315 INFO [finetune.py:976] (6/7) Epoch 14, batch 250, loss[loss=0.2216, simple_loss=0.2902, pruned_loss=0.07653, over 4816.00 frames. ], tot_loss[loss=0.1809, simple_loss=0.2483, pruned_loss=0.05678, over 684213.36 frames. ], batch size: 39, lr: 3.57e-03, grad_scale: 32.0 2023-04-27 05:05:11,831 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=74720.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 05:05:39,269 INFO [optim.py:369] (6/7) 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 05:05:49,927 INFO [finetune.py:976] (6/7) Epoch 14, batch 300, loss[loss=0.1807, simple_loss=0.2263, pruned_loss=0.06756, over 4049.00 frames. ], tot_loss[loss=0.1824, simple_loss=0.2508, pruned_loss=0.057, over 743670.92 frames. ], batch size: 17, lr: 3.56e-03, grad_scale: 32.0 2023-04-27 05:05:50,077 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.5194, 1.8062, 1.7250, 2.3905, 2.6287, 2.0397, 1.9726, 1.7773], device='cuda:6'), covar=tensor([0.1422, 0.1576, 0.1795, 0.1377, 0.0910, 0.1702, 0.2156, 0.2067], device='cuda:6'), in_proj_covar=tensor([0.0307, 0.0316, 0.0351, 0.0292, 0.0329, 0.0314, 0.0305, 0.0362], device='cuda:6'), 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:6') 2023-04-27 05:05:54,767 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 14, batch 350, loss[loss=0.2112, simple_loss=0.2713, pruned_loss=0.0756, over 4856.00 frames. ], tot_loss[loss=0.1837, simple_loss=0.2522, pruned_loss=0.05761, over 790922.94 frames. ], batch size: 44, lr: 3.56e-03, grad_scale: 32.0 2023-04-27 05:07:09,533 INFO [optim.py:369] (6/7) 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 05:07:25,453 INFO [finetune.py:976] (6/7) Epoch 14, batch 400, loss[loss=0.2305, simple_loss=0.2841, pruned_loss=0.08844, over 4850.00 frames. ], tot_loss[loss=0.1854, simple_loss=0.2539, pruned_loss=0.05843, over 827682.48 frames. ], batch size: 31, lr: 3.56e-03, grad_scale: 32.0 2023-04-27 05:07:47,703 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.48 vs. limit=5.0 2023-04-27 05:07:49,827 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=74896.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 05:07:57,757 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=74909.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 05:07:59,305 INFO [finetune.py:976] (6/7) Epoch 14, batch 450, loss[loss=0.2315, simple_loss=0.2981, pruned_loss=0.08252, over 4722.00 frames. ], tot_loss[loss=0.1852, simple_loss=0.2534, pruned_loss=0.05848, over 856498.82 frames. ], batch size: 54, lr: 3.56e-03, grad_scale: 32.0 2023-04-27 05:08:37,891 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=74944.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 05:08:38,410 INFO [optim.py:369] (6/7) 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 05:08:47,686 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=74951.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 05:08:53,661 INFO [finetune.py:976] (6/7) Epoch 14, batch 500, loss[loss=0.1815, simple_loss=0.2447, pruned_loss=0.05912, over 4932.00 frames. ], tot_loss[loss=0.1838, simple_loss=0.2515, pruned_loss=0.05799, over 879015.08 frames. ], batch size: 38, lr: 3.56e-03, grad_scale: 32.0 2023-04-27 05:08:56,095 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=74964.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 05:08:57,947 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.8470, 2.1063, 2.0368, 2.2231, 1.9808, 2.0858, 2.1643, 2.0399], device='cuda:6'), covar=tensor([0.4292, 0.7447, 0.5527, 0.5391, 0.6145, 0.7891, 0.6666, 0.6450], device='cuda:6'), in_proj_covar=tensor([0.0329, 0.0374, 0.0317, 0.0329, 0.0341, 0.0399, 0.0355, 0.0324], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 05:08:59,785 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=74970.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 05:09:16,549 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.50 vs. limit=2.0 2023-04-27 05:09:18,274 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.81 vs. limit=2.0 2023-04-27 05:09:27,831 INFO [finetune.py:976] (6/7) Epoch 14, batch 550, loss[loss=0.1535, simple_loss=0.2356, pruned_loss=0.0357, over 4824.00 frames. ], tot_loss[loss=0.1811, simple_loss=0.2485, pruned_loss=0.05688, over 895745.96 frames. ], batch size: 33, lr: 3.56e-03, grad_scale: 32.0 2023-04-27 05:09:37,443 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=75025.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 05:09:51,446 INFO [optim.py:369] (6/7) 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 05:10:01,743 INFO [finetune.py:976] (6/7) Epoch 14, batch 600, loss[loss=0.1582, simple_loss=0.2425, pruned_loss=0.037, over 4931.00 frames. ], tot_loss[loss=0.1809, simple_loss=0.2486, pruned_loss=0.05663, over 907596.40 frames. ], batch size: 38, lr: 3.56e-03, grad_scale: 32.0 2023-04-27 05:10:07,890 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6209, 1.6346, 0.8337, 1.3145, 1.7194, 1.5035, 1.4163, 1.4201], device='cuda:6'), covar=tensor([0.0531, 0.0384, 0.0363, 0.0584, 0.0282, 0.0508, 0.0494, 0.0599], device='cuda:6'), in_proj_covar=tensor([0.0029, 0.0025, 0.0023, 0.0029, 0.0020, 0.0029, 0.0028, 0.0029], device='cuda:6'), out_proj_covar=tensor([0.0050, 0.0045, 0.0038, 0.0051, 0.0038, 0.0049, 0.0049, 0.0050], device='cuda:6') 2023-04-27 05:10:35,744 INFO [finetune.py:976] (6/7) Epoch 14, batch 650, loss[loss=0.221, simple_loss=0.2956, pruned_loss=0.07322, over 4902.00 frames. ], tot_loss[loss=0.1837, simple_loss=0.2523, pruned_loss=0.05756, over 917724.78 frames. ], batch size: 37, lr: 3.56e-03, grad_scale: 32.0 2023-04-27 05:10:35,877 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.6412, 3.4312, 2.8966, 3.0469, 2.4616, 2.9513, 2.9788, 2.4016], device='cuda:6'), covar=tensor([0.2229, 0.1159, 0.0766, 0.1298, 0.3012, 0.1069, 0.1937, 0.2818], device='cuda:6'), in_proj_covar=tensor([0.0293, 0.0311, 0.0227, 0.0285, 0.0315, 0.0269, 0.0258, 0.0273], device='cuda:6'), out_proj_covar=tensor([1.1790e-04, 1.2396e-04, 9.0807e-05, 1.1378e-04, 1.2831e-04, 1.0753e-04, 1.0428e-04, 1.0902e-04], device='cuda:6') 2023-04-27 05:10:38,910 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5350, 1.4345, 4.2185, 3.9647, 3.7494, 3.9912, 3.9704, 3.7187], device='cuda:6'), covar=tensor([0.6631, 0.5712, 0.1134, 0.1811, 0.1076, 0.1564, 0.1154, 0.1549], device='cuda:6'), in_proj_covar=tensor([0.0308, 0.0307, 0.0403, 0.0403, 0.0346, 0.0403, 0.0312, 0.0366], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-27 05:10:42,800 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-04-27 05:10:59,289 INFO [optim.py:369] (6/7) 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 05:11:09,465 INFO [finetune.py:976] (6/7) Epoch 14, batch 700, loss[loss=0.1518, simple_loss=0.2326, pruned_loss=0.03547, over 4811.00 frames. ], tot_loss[loss=0.1835, simple_loss=0.2522, pruned_loss=0.05743, over 926068.27 frames. ], batch size: 45, lr: 3.56e-03, grad_scale: 32.0 2023-04-27 05:11:27,284 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-04-27 05:11:30,432 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7236, 1.3982, 1.2971, 1.5834, 1.8978, 1.6168, 1.4656, 1.2159], device='cuda:6'), covar=tensor([0.1205, 0.1254, 0.1393, 0.1139, 0.0714, 0.1255, 0.1767, 0.1599], device='cuda:6'), in_proj_covar=tensor([0.0307, 0.0317, 0.0352, 0.0293, 0.0331, 0.0314, 0.0306, 0.0362], device='cuda:6'), out_proj_covar=tensor([6.3782e-05, 6.6524e-05, 7.5506e-05, 5.9830e-05, 6.8999e-05, 6.6681e-05, 6.4819e-05, 7.7300e-05], device='cuda:6') 2023-04-27 05:11:46,819 INFO [finetune.py:976] (6/7) Epoch 14, batch 750, loss[loss=0.2014, simple_loss=0.2629, pruned_loss=0.06999, over 4756.00 frames. ], tot_loss[loss=0.1838, simple_loss=0.2532, pruned_loss=0.05723, over 931668.12 frames. ], batch size: 27, lr: 3.56e-03, grad_scale: 32.0 2023-04-27 05:12:09,268 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=75228.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 05:12:31,821 INFO [optim.py:369] (6/7) 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 05:12:35,595 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=75251.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 05:12:41,571 INFO [finetune.py:976] (6/7) Epoch 14, batch 800, loss[loss=0.2119, simple_loss=0.2757, pruned_loss=0.07404, over 4837.00 frames. ], tot_loss[loss=0.1834, simple_loss=0.253, pruned_loss=0.05695, over 934250.72 frames. ], batch size: 49, lr: 3.56e-03, grad_scale: 32.0 2023-04-27 05:12:42,835 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.9025, 2.1464, 2.1097, 2.1983, 2.0168, 2.2027, 2.1617, 2.0474], device='cuda:6'), covar=tensor([0.4122, 0.7036, 0.5666, 0.5153, 0.6515, 0.7925, 0.6840, 0.6767], device='cuda:6'), in_proj_covar=tensor([0.0326, 0.0369, 0.0313, 0.0324, 0.0337, 0.0393, 0.0351, 0.0320], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 05:12:44,559 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=75265.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 05:12:47,636 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=75270.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 05:13:00,639 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=75289.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 05:13:01,210 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.8399, 2.8457, 2.2382, 3.3039, 2.8821, 2.8870, 1.2326, 2.7379], device='cuda:6'), covar=tensor([0.2171, 0.1571, 0.3080, 0.2585, 0.3384, 0.2097, 0.5678, 0.3029], device='cuda:6'), in_proj_covar=tensor([0.0243, 0.0213, 0.0246, 0.0301, 0.0296, 0.0244, 0.0269, 0.0270], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 05:13:07,657 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=75299.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 05:13:11,944 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.2110, 3.0714, 0.8953, 1.7745, 1.6557, 2.4283, 1.8100, 1.0689], device='cuda:6'), covar=tensor([0.1579, 0.1286, 0.2056, 0.1370, 0.1270, 0.0998, 0.1524, 0.2025], device='cuda:6'), in_proj_covar=tensor([0.0118, 0.0245, 0.0137, 0.0121, 0.0133, 0.0152, 0.0118, 0.0120], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-27 05:13:14,865 INFO [finetune.py:976] (6/7) Epoch 14, batch 850, loss[loss=0.2294, simple_loss=0.2858, pruned_loss=0.08649, over 4777.00 frames. ], tot_loss[loss=0.1827, simple_loss=0.252, pruned_loss=0.05672, over 938292.05 frames. ], batch size: 26, lr: 3.56e-03, grad_scale: 32.0 2023-04-27 05:13:26,892 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=75320.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 05:13:40,194 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=75331.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 05:13:48,386 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7763, 1.7115, 0.8118, 1.3777, 1.8653, 1.6577, 1.5188, 1.5462], device='cuda:6'), covar=tensor([0.0501, 0.0373, 0.0365, 0.0574, 0.0266, 0.0512, 0.0505, 0.0568], device='cuda:6'), in_proj_covar=tensor([0.0029, 0.0025, 0.0023, 0.0029, 0.0020, 0.0029, 0.0028, 0.0029], device='cuda:6'), out_proj_covar=tensor([0.0050, 0.0045, 0.0038, 0.0050, 0.0037, 0.0049, 0.0048, 0.0050], device='cuda:6') 2023-04-27 05:13:56,951 INFO [optim.py:369] (6/7) 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 05:14:00,691 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7342, 1.3642, 1.4242, 1.4307, 1.9100, 1.5607, 1.2508, 1.3697], device='cuda:6'), covar=tensor([0.1554, 0.1303, 0.2167, 0.1443, 0.0890, 0.1477, 0.1761, 0.2318], device='cuda:6'), in_proj_covar=tensor([0.0308, 0.0318, 0.0354, 0.0294, 0.0332, 0.0315, 0.0307, 0.0363], device='cuda:6'), out_proj_covar=tensor([6.4046e-05, 6.6824e-05, 7.5892e-05, 6.0025e-05, 6.9237e-05, 6.6805e-05, 6.4959e-05, 7.7662e-05], device='cuda:6') 2023-04-27 05:14:12,837 INFO [finetune.py:976] (6/7) Epoch 14, batch 900, loss[loss=0.1851, simple_loss=0.2481, pruned_loss=0.061, over 4942.00 frames. ], tot_loss[loss=0.1805, simple_loss=0.2491, pruned_loss=0.05599, over 942273.94 frames. ], batch size: 33, lr: 3.56e-03, grad_scale: 32.0 2023-04-27 05:14:57,517 INFO [finetune.py:976] (6/7) Epoch 14, batch 950, loss[loss=0.1788, simple_loss=0.252, pruned_loss=0.05284, over 4818.00 frames. ], tot_loss[loss=0.1812, simple_loss=0.249, pruned_loss=0.05666, over 945278.06 frames. ], batch size: 38, lr: 3.56e-03, grad_scale: 32.0 2023-04-27 05:15:02,532 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7320, 2.1596, 0.8503, 1.1030, 1.4892, 1.0324, 2.4373, 1.2426], device='cuda:6'), covar=tensor([0.0761, 0.0620, 0.0679, 0.1310, 0.0505, 0.1074, 0.0288, 0.0737], device='cuda:6'), in_proj_covar=tensor([0.0051, 0.0066, 0.0049, 0.0047, 0.0050, 0.0052, 0.0075, 0.0051], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0007], device='cuda:6') 2023-04-27 05:15:20,049 INFO [optim.py:369] (6/7) 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 05:15:30,322 INFO [finetune.py:976] (6/7) Epoch 14, batch 1000, loss[loss=0.2187, simple_loss=0.2937, pruned_loss=0.07179, over 4721.00 frames. ], tot_loss[loss=0.183, simple_loss=0.2511, pruned_loss=0.05743, over 946535.22 frames. ], batch size: 59, lr: 3.56e-03, grad_scale: 32.0 2023-04-27 05:15:56,223 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.3684, 1.8396, 2.1615, 2.9317, 2.1944, 1.7399, 1.7676, 2.1315], device='cuda:6'), covar=tensor([0.3458, 0.3521, 0.1777, 0.2491, 0.3127, 0.2781, 0.3904, 0.2260], device='cuda:6'), in_proj_covar=tensor([0.0286, 0.0247, 0.0222, 0.0315, 0.0214, 0.0229, 0.0229, 0.0185], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:6') 2023-04-27 05:16:03,291 INFO [finetune.py:976] (6/7) Epoch 14, batch 1050, loss[loss=0.1527, simple_loss=0.2304, pruned_loss=0.03752, over 4746.00 frames. ], tot_loss[loss=0.1837, simple_loss=0.2525, pruned_loss=0.05746, over 948356.67 frames. ], batch size: 27, lr: 3.56e-03, grad_scale: 32.0 2023-04-27 05:16:05,264 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.1918, 1.6769, 1.9649, 2.5492, 2.0554, 1.5465, 1.2516, 1.9234], device='cuda:6'), covar=tensor([0.3410, 0.3436, 0.1840, 0.2454, 0.2662, 0.2941, 0.4461, 0.2231], device='cuda:6'), in_proj_covar=tensor([0.0287, 0.0247, 0.0222, 0.0315, 0.0215, 0.0229, 0.0229, 0.0185], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:6') 2023-04-27 05:16:25,347 INFO [optim.py:369] (6/7) 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 05:16:36,807 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.77 vs. limit=5.0 2023-04-27 05:16:36,966 INFO [finetune.py:976] (6/7) Epoch 14, batch 1100, loss[loss=0.226, simple_loss=0.2974, pruned_loss=0.07726, over 4901.00 frames. ], tot_loss[loss=0.1856, simple_loss=0.2549, pruned_loss=0.05813, over 950787.17 frames. ], batch size: 43, lr: 3.56e-03, grad_scale: 32.0 2023-04-27 05:16:39,548 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=75565.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 05:16:41,422 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.0520, 1.0477, 1.2383, 1.1571, 1.0010, 0.9461, 0.9956, 0.4703], device='cuda:6'), covar=tensor([0.0619, 0.0517, 0.0541, 0.0565, 0.0791, 0.1251, 0.0468, 0.0789], device='cuda:6'), in_proj_covar=tensor([0.0069, 0.0070, 0.0070, 0.0068, 0.0076, 0.0097, 0.0076, 0.0070], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-27 05:16:57,393 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 14, batch 1150, loss[loss=0.1818, simple_loss=0.2517, pruned_loss=0.05602, over 4917.00 frames. ], tot_loss[loss=0.1872, simple_loss=0.2563, pruned_loss=0.05906, over 953038.62 frames. ], batch size: 33, lr: 3.56e-03, grad_scale: 32.0 2023-04-27 05:17:32,674 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=75613.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 05:17:36,938 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=75620.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 05:17:40,643 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=75626.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 05:17:42,671 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2023-04-27 05:17:53,273 INFO [optim.py:369] (6/7) 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] (6/7) Epoch 14, batch 1200, loss[loss=0.1728, simple_loss=0.2351, pruned_loss=0.05523, over 4798.00 frames. ], tot_loss[loss=0.1853, simple_loss=0.254, pruned_loss=0.05828, over 952661.46 frames. ], batch size: 25, lr: 3.56e-03, grad_scale: 32.0 2023-04-27 05:18:09,671 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=75668.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 05:18:43,841 INFO [finetune.py:976] (6/7) Epoch 14, batch 1250, loss[loss=0.1614, simple_loss=0.2306, pruned_loss=0.04609, over 4935.00 frames. ], tot_loss[loss=0.1833, simple_loss=0.2515, pruned_loss=0.05749, over 953704.17 frames. ], batch size: 38, lr: 3.56e-03, grad_scale: 32.0 2023-04-27 05:19:24,355 INFO [optim.py:369] (6/7) 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 05:19:31,213 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-04-27 05:19:45,791 INFO [finetune.py:976] (6/7) Epoch 14, batch 1300, loss[loss=0.1513, simple_loss=0.2229, pruned_loss=0.03984, over 4893.00 frames. ], tot_loss[loss=0.1803, simple_loss=0.2484, pruned_loss=0.05609, over 955760.22 frames. ], batch size: 32, lr: 3.56e-03, grad_scale: 32.0 2023-04-27 05:20:50,516 INFO [finetune.py:976] (6/7) Epoch 14, batch 1350, loss[loss=0.1367, simple_loss=0.209, pruned_loss=0.03226, over 4767.00 frames. ], tot_loss[loss=0.1826, simple_loss=0.25, pruned_loss=0.05757, over 952687.71 frames. ], batch size: 26, lr: 3.56e-03, grad_scale: 32.0 2023-04-27 05:21:39,792 INFO [optim.py:369] (6/7) 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 05:21:55,187 INFO [finetune.py:976] (6/7) Epoch 14, batch 1400, loss[loss=0.1543, simple_loss=0.2309, pruned_loss=0.03881, over 4809.00 frames. ], tot_loss[loss=0.1828, simple_loss=0.2514, pruned_loss=0.05706, over 954703.19 frames. ], batch size: 25, lr: 3.56e-03, grad_scale: 32.0 2023-04-27 05:22:34,307 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=75884.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 05:22:49,500 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6649, 1.3178, 4.3549, 4.0488, 3.7987, 4.1106, 4.0123, 3.8813], device='cuda:6'), covar=tensor([0.7091, 0.6044, 0.1118, 0.1955, 0.1124, 0.1894, 0.1923, 0.1445], device='cuda:6'), in_proj_covar=tensor([0.0307, 0.0305, 0.0400, 0.0400, 0.0344, 0.0402, 0.0310, 0.0362], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-27 05:22:51,294 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6414, 1.3373, 4.1195, 3.8750, 3.5761, 3.7662, 3.7405, 3.6514], device='cuda:6'), covar=tensor([0.6916, 0.5764, 0.0975, 0.1608, 0.1087, 0.1728, 0.1902, 0.1371], device='cuda:6'), in_proj_covar=tensor([0.0307, 0.0305, 0.0400, 0.0400, 0.0344, 0.0402, 0.0310, 0.0362], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-27 05:22:51,836 INFO [finetune.py:976] (6/7) Epoch 14, batch 1450, loss[loss=0.1689, simple_loss=0.2495, pruned_loss=0.04413, over 4727.00 frames. ], tot_loss[loss=0.184, simple_loss=0.253, pruned_loss=0.05751, over 952740.40 frames. ], batch size: 54, lr: 3.55e-03, grad_scale: 32.0 2023-04-27 05:22:52,641 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.30 vs. limit=5.0 2023-04-27 05:22:52,995 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.50 vs. limit=2.0 2023-04-27 05:23:03,095 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=75926.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 05:23:06,748 INFO [zipformer.py:1188] (6/7) 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] (6/7) 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] (6/7) Epoch 14, batch 1500, loss[loss=0.1857, simple_loss=0.2592, pruned_loss=0.05613, over 3982.00 frames. ], tot_loss[loss=0.185, simple_loss=0.2538, pruned_loss=0.05805, over 951503.30 frames. ], batch size: 17, lr: 3.55e-03, grad_scale: 32.0 2023-04-27 05:23:29,221 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-04-27 05:23:34,211 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=75974.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 05:24:11,482 INFO [finetune.py:976] (6/7) Epoch 14, batch 1550, loss[loss=0.1938, simple_loss=0.2421, pruned_loss=0.07272, over 4313.00 frames. ], tot_loss[loss=0.1848, simple_loss=0.254, pruned_loss=0.0578, over 952748.81 frames. ], batch size: 19, lr: 3.55e-03, grad_scale: 32.0 2023-04-27 05:24:40,106 INFO [optim.py:369] (6/7) 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 05:24:42,704 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.3740, 1.2287, 1.6331, 1.5217, 1.2752, 1.1366, 1.3791, 1.0095], device='cuda:6'), covar=tensor([0.0588, 0.0655, 0.0379, 0.0627, 0.0711, 0.1147, 0.0476, 0.0553], device='cuda:6'), in_proj_covar=tensor([0.0068, 0.0071, 0.0070, 0.0068, 0.0076, 0.0097, 0.0076, 0.0070], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-27 05:25:01,381 INFO [finetune.py:976] (6/7) Epoch 14, batch 1600, loss[loss=0.163, simple_loss=0.2373, pruned_loss=0.04433, over 4818.00 frames. ], tot_loss[loss=0.1837, simple_loss=0.2523, pruned_loss=0.05758, over 953389.50 frames. ], batch size: 40, lr: 3.55e-03, grad_scale: 32.0 2023-04-27 05:25:01,497 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.4087, 1.9523, 1.9003, 2.3109, 2.1710, 2.1879, 1.7550, 4.6225], device='cuda:6'), covar=tensor([0.0552, 0.0726, 0.0731, 0.1049, 0.0581, 0.0466, 0.0677, 0.0123], device='cuda:6'), in_proj_covar=tensor([0.0038, 0.0039, 0.0040, 0.0043, 0.0040, 0.0038, 0.0038, 0.0057], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0014, 0.0013, 0.0012, 0.0013, 0.0016], device='cuda:6') 2023-04-27 05:25:34,938 INFO [finetune.py:976] (6/7) Epoch 14, batch 1650, loss[loss=0.1882, simple_loss=0.2683, pruned_loss=0.05403, over 4901.00 frames. ], tot_loss[loss=0.1808, simple_loss=0.249, pruned_loss=0.05629, over 955346.67 frames. ], batch size: 35, lr: 3.55e-03, grad_scale: 32.0 2023-04-27 05:25:58,459 INFO [optim.py:369] (6/7) 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 05:26:08,245 INFO [finetune.py:976] (6/7) Epoch 14, batch 1700, loss[loss=0.1963, simple_loss=0.2553, pruned_loss=0.06862, over 4799.00 frames. ], tot_loss[loss=0.1794, simple_loss=0.2471, pruned_loss=0.05581, over 955738.01 frames. ], batch size: 41, lr: 3.55e-03, grad_scale: 32.0 2023-04-27 05:26:42,125 INFO [finetune.py:976] (6/7) Epoch 14, batch 1750, loss[loss=0.2, simple_loss=0.2501, pruned_loss=0.07496, over 4697.00 frames. ], tot_loss[loss=0.1803, simple_loss=0.2485, pruned_loss=0.05607, over 955167.99 frames. ], batch size: 23, lr: 3.55e-03, grad_scale: 32.0 2023-04-27 05:26:44,571 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.1128, 1.6580, 1.5975, 2.0116, 1.8083, 2.0490, 1.5640, 4.2086], device='cuda:6'), covar=tensor([0.0608, 0.0791, 0.0800, 0.1270, 0.0650, 0.0529, 0.0751, 0.0115], device='cuda:6'), in_proj_covar=tensor([0.0038, 0.0039, 0.0040, 0.0044, 0.0040, 0.0038, 0.0038, 0.0057], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:6') 2023-04-27 05:27:06,454 INFO [optim.py:369] (6/7) 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 05:27:16,247 INFO [finetune.py:976] (6/7) Epoch 14, batch 1800, loss[loss=0.1964, simple_loss=0.2663, pruned_loss=0.06323, over 4862.00 frames. ], tot_loss[loss=0.1834, simple_loss=0.2526, pruned_loss=0.05704, over 956160.06 frames. ], batch size: 44, lr: 3.55e-03, grad_scale: 32.0 2023-04-27 05:27:27,569 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=76269.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 05:28:02,598 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=76294.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 05:28:12,930 INFO [finetune.py:976] (6/7) Epoch 14, batch 1850, loss[loss=0.1888, simple_loss=0.2576, pruned_loss=0.06002, over 4896.00 frames. ], tot_loss[loss=0.1841, simple_loss=0.2536, pruned_loss=0.0573, over 957843.55 frames. ], batch size: 43, lr: 3.55e-03, grad_scale: 32.0 2023-04-27 05:28:20,921 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=76323.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 05:28:25,685 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=76330.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 05:28:36,248 INFO [optim.py:369] (6/7) 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 05:28:43,356 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 14, batch 1900, loss[loss=0.1607, simple_loss=0.2455, pruned_loss=0.03798, over 4774.00 frames. ], tot_loss[loss=0.1848, simple_loss=0.2551, pruned_loss=0.05719, over 955851.22 frames. ], batch size: 27, lr: 3.55e-03, grad_scale: 32.0 2023-04-27 05:29:01,986 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=76384.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 05:29:17,126 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0080, 2.3974, 2.0382, 2.2809, 1.7526, 2.0149, 1.9535, 1.5987], device='cuda:6'), covar=tensor([0.1834, 0.1356, 0.0870, 0.1195, 0.3165, 0.1179, 0.2030, 0.2610], device='cuda:6'), in_proj_covar=tensor([0.0291, 0.0312, 0.0226, 0.0284, 0.0313, 0.0267, 0.0256, 0.0272], device='cuda:6'), out_proj_covar=tensor([1.1725e-04, 1.2424e-04, 9.0302e-05, 1.1342e-04, 1.2758e-04, 1.0689e-04, 1.0364e-04, 1.0866e-04], device='cuda:6') 2023-04-27 05:29:20,620 INFO [finetune.py:976] (6/7) Epoch 14, batch 1950, loss[loss=0.1648, simple_loss=0.2459, pruned_loss=0.04186, over 4792.00 frames. ], tot_loss[loss=0.1829, simple_loss=0.2532, pruned_loss=0.05632, over 954811.74 frames. ], batch size: 26, lr: 3.55e-03, grad_scale: 32.0 2023-04-27 05:29:42,667 INFO [optim.py:369] (6/7) 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 05:29:55,872 INFO [finetune.py:976] (6/7) Epoch 14, batch 2000, loss[loss=0.1728, simple_loss=0.2425, pruned_loss=0.05151, over 4908.00 frames. ], tot_loss[loss=0.1819, simple_loss=0.2513, pruned_loss=0.05625, over 956618.84 frames. ], batch size: 43, lr: 3.55e-03, grad_scale: 32.0 2023-04-27 05:30:16,722 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.1039, 2.4343, 0.9748, 1.2247, 1.9169, 1.2382, 3.0763, 1.5652], device='cuda:6'), covar=tensor([0.0628, 0.0723, 0.0812, 0.1275, 0.0492, 0.0959, 0.0261, 0.0687], device='cuda:6'), in_proj_covar=tensor([0.0051, 0.0066, 0.0049, 0.0046, 0.0050, 0.0051, 0.0075, 0.0051], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0007, 0.0008, 0.0008, 0.0010, 0.0007], device='cuda:6') 2023-04-27 05:30:58,158 INFO [finetune.py:976] (6/7) Epoch 14, batch 2050, loss[loss=0.1773, simple_loss=0.2446, pruned_loss=0.05505, over 4914.00 frames. ], tot_loss[loss=0.1788, simple_loss=0.2474, pruned_loss=0.05507, over 956508.30 frames. ], batch size: 36, lr: 3.55e-03, grad_scale: 64.0 2023-04-27 05:31:09,490 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-04-27 05:31:18,747 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.2847, 1.8845, 1.8713, 2.2233, 2.1193, 2.0648, 1.7626, 4.5813], device='cuda:6'), covar=tensor([0.0571, 0.0741, 0.0761, 0.1125, 0.0575, 0.0499, 0.0677, 0.0103], device='cuda:6'), in_proj_covar=tensor([0.0038, 0.0038, 0.0039, 0.0043, 0.0040, 0.0038, 0.0038, 0.0056], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0014, 0.0013, 0.0012, 0.0013, 0.0016], device='cuda:6') 2023-04-27 05:31:19,798 INFO [optim.py:369] (6/7) 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 05:31:32,054 INFO [finetune.py:976] (6/7) Epoch 14, batch 2100, loss[loss=0.2422, simple_loss=0.299, pruned_loss=0.09269, over 4854.00 frames. ], tot_loss[loss=0.1789, simple_loss=0.2471, pruned_loss=0.05536, over 956970.09 frames. ], batch size: 44, lr: 3.55e-03, grad_scale: 64.0 2023-04-27 05:32:06,102 INFO [finetune.py:976] (6/7) Epoch 14, batch 2150, loss[loss=0.1764, simple_loss=0.2591, pruned_loss=0.04681, over 4726.00 frames. ], tot_loss[loss=0.181, simple_loss=0.2501, pruned_loss=0.05589, over 956660.21 frames. ], batch size: 59, lr: 3.55e-03, grad_scale: 64.0 2023-04-27 05:32:14,746 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=76625.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 05:32:27,240 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.0373, 1.5683, 1.4749, 1.7187, 1.6480, 1.8978, 1.3268, 3.5663], device='cuda:6'), covar=tensor([0.0687, 0.0803, 0.0768, 0.1179, 0.0663, 0.0552, 0.0767, 0.0139], device='cuda:6'), in_proj_covar=tensor([0.0038, 0.0039, 0.0039, 0.0043, 0.0040, 0.0038, 0.0038, 0.0056], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0014, 0.0013, 0.0012, 0.0013, 0.0016], device='cuda:6') 2023-04-27 05:32:27,726 INFO [optim.py:369] (6/7) 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 05:32:30,859 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=76650.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 05:32:44,024 INFO [finetune.py:976] (6/7) Epoch 14, batch 2200, loss[loss=0.2473, simple_loss=0.3059, pruned_loss=0.09432, over 4856.00 frames. ], tot_loss[loss=0.1831, simple_loss=0.2524, pruned_loss=0.05685, over 955783.68 frames. ], batch size: 44, lr: 3.55e-03, grad_scale: 64.0 2023-04-27 05:33:06,766 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=76679.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 05:33:07,212 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-04-27 05:33:18,430 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.1608, 2.7633, 2.2308, 2.6457, 1.9234, 2.3591, 2.4078, 1.7762], device='cuda:6'), covar=tensor([0.2077, 0.1292, 0.0919, 0.1177, 0.3171, 0.1184, 0.2092, 0.2634], device='cuda:6'), in_proj_covar=tensor([0.0294, 0.0313, 0.0227, 0.0287, 0.0314, 0.0268, 0.0257, 0.0273], device='cuda:6'), out_proj_covar=tensor([1.1854e-04, 1.2488e-04, 9.0773e-05, 1.1441e-04, 1.2822e-04, 1.0712e-04, 1.0438e-04, 1.0908e-04], device='cuda:6') 2023-04-27 05:33:46,580 INFO [finetune.py:976] (6/7) Epoch 14, batch 2250, loss[loss=0.1889, simple_loss=0.2603, pruned_loss=0.05874, over 4844.00 frames. ], tot_loss[loss=0.1839, simple_loss=0.2538, pruned_loss=0.05696, over 957580.31 frames. ], batch size: 49, lr: 3.55e-03, grad_scale: 64.0 2023-04-27 05:33:59,872 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-04-27 05:34:10,668 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6374, 3.2856, 0.9392, 1.8327, 1.9883, 2.5067, 1.9391, 1.0853], device='cuda:6'), covar=tensor([0.1401, 0.1186, 0.2142, 0.1344, 0.1131, 0.1054, 0.1577, 0.1930], device='cuda:6'), in_proj_covar=tensor([0.0118, 0.0247, 0.0139, 0.0121, 0.0133, 0.0153, 0.0119, 0.0120], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-27 05:34:11,936 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7763, 2.4893, 2.1442, 1.8295, 1.3080, 1.3645, 2.2492, 1.3755], device='cuda:6'), covar=tensor([0.1749, 0.1433, 0.1332, 0.1733, 0.2356, 0.1972, 0.0881, 0.2019], device='cuda:6'), in_proj_covar=tensor([0.0196, 0.0214, 0.0169, 0.0205, 0.0202, 0.0184, 0.0157, 0.0187], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:6') 2023-04-27 05:34:30,283 INFO [optim.py:369] (6/7) 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] (6/7) Epoch 14, batch 2300, loss[loss=0.1762, simple_loss=0.2412, pruned_loss=0.05566, over 4918.00 frames. ], tot_loss[loss=0.184, simple_loss=0.2538, pruned_loss=0.05703, over 956746.26 frames. ], batch size: 33, lr: 3.55e-03, grad_scale: 64.0 2023-04-27 05:34:47,785 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5612, 1.5871, 0.7795, 1.2664, 1.4258, 1.4264, 1.3469, 1.3653], device='cuda:6'), covar=tensor([0.0538, 0.0353, 0.0393, 0.0588, 0.0301, 0.0532, 0.0514, 0.0585], device='cuda:6'), in_proj_covar=tensor([0.0029, 0.0024, 0.0023, 0.0029, 0.0020, 0.0029, 0.0028, 0.0029], device='cuda:6'), out_proj_covar=tensor([0.0050, 0.0044, 0.0038, 0.0051, 0.0038, 0.0049, 0.0049, 0.0050], device='cuda:6') 2023-04-27 05:34:52,944 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-04-27 05:34:55,893 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-04-27 05:35:26,140 INFO [finetune.py:976] (6/7) Epoch 14, batch 2350, loss[loss=0.1641, simple_loss=0.2322, pruned_loss=0.04804, over 4928.00 frames. ], tot_loss[loss=0.1838, simple_loss=0.253, pruned_loss=0.05728, over 957560.09 frames. ], batch size: 33, lr: 3.55e-03, grad_scale: 64.0 2023-04-27 05:36:06,015 INFO [zipformer.py:1188] (6/7) 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] (6/7) 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 05:36:20,266 INFO [finetune.py:976] (6/7) Epoch 14, batch 2400, loss[loss=0.1651, simple_loss=0.2392, pruned_loss=0.04546, over 4908.00 frames. ], tot_loss[loss=0.1815, simple_loss=0.2501, pruned_loss=0.05643, over 956746.62 frames. ], batch size: 43, lr: 3.55e-03, grad_scale: 64.0 2023-04-27 05:36:25,012 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-04-27 05:36:32,934 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-04-27 05:36:46,483 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=76900.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 05:36:52,972 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0669, 2.5554, 1.2515, 1.7478, 2.4775, 1.9095, 1.8015, 1.8797], device='cuda:6'), covar=tensor([0.0472, 0.0320, 0.0305, 0.0534, 0.0223, 0.0483, 0.0530, 0.0537], device='cuda:6'), in_proj_covar=tensor([0.0029, 0.0025, 0.0023, 0.0029, 0.0020, 0.0029, 0.0028, 0.0029], device='cuda:6'), out_proj_covar=tensor([0.0050, 0.0045, 0.0038, 0.0051, 0.0038, 0.0049, 0.0049, 0.0050], device='cuda:6') 2023-04-27 05:36:54,080 INFO [finetune.py:976] (6/7) Epoch 14, batch 2450, loss[loss=0.1673, simple_loss=0.2317, pruned_loss=0.05143, over 4798.00 frames. ], tot_loss[loss=0.1781, simple_loss=0.2462, pruned_loss=0.05502, over 957489.99 frames. ], batch size: 51, lr: 3.55e-03, grad_scale: 64.0 2023-04-27 05:37:04,093 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=76925.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 05:37:06,486 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5659, 0.9221, 1.6577, 2.0532, 1.6745, 1.5645, 1.6118, 1.6236], device='cuda:6'), covar=tensor([0.5057, 0.7070, 0.6364, 0.6616, 0.6146, 0.8253, 0.7474, 0.8549], device='cuda:6'), in_proj_covar=tensor([0.0415, 0.0408, 0.0495, 0.0511, 0.0444, 0.0465, 0.0471, 0.0475], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 05:37:15,277 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([4.1203, 4.0633, 2.9332, 4.6673, 4.1456, 4.0422, 1.7790, 4.0243], device='cuda:6'), covar=tensor([0.1494, 0.1005, 0.3008, 0.1249, 0.4352, 0.1665, 0.5971, 0.2403], device='cuda:6'), in_proj_covar=tensor([0.0242, 0.0212, 0.0247, 0.0300, 0.0295, 0.0244, 0.0269, 0.0270], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 05:37:17,026 INFO [optim.py:369] (6/7) 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 05:37:20,126 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=76950.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 05:37:27,667 INFO [finetune.py:976] (6/7) Epoch 14, batch 2500, loss[loss=0.1672, simple_loss=0.2487, pruned_loss=0.04286, over 4826.00 frames. ], tot_loss[loss=0.1824, simple_loss=0.2496, pruned_loss=0.05762, over 956339.60 frames. ], batch size: 49, lr: 3.55e-03, grad_scale: 64.0 2023-04-27 05:37:36,501 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=76973.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 05:37:40,692 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=76979.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 05:37:52,651 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=76998.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 05:38:07,211 INFO [finetune.py:976] (6/7) Epoch 14, batch 2550, loss[loss=0.1978, simple_loss=0.259, pruned_loss=0.0683, over 4928.00 frames. ], tot_loss[loss=0.1843, simple_loss=0.252, pruned_loss=0.05831, over 953515.08 frames. ], batch size: 33, lr: 3.55e-03, grad_scale: 64.0 2023-04-27 05:38:18,429 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=77027.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 05:38:30,331 INFO [optim.py:369] (6/7) 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 05:38:34,691 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([0.8671, 1.6971, 1.6949, 1.4067, 1.7543, 1.5220, 2.1660, 1.3865], device='cuda:6'), covar=tensor([0.2528, 0.1158, 0.3082, 0.1900, 0.1089, 0.1663, 0.1133, 0.2904], device='cuda:6'), in_proj_covar=tensor([0.0344, 0.0349, 0.0428, 0.0360, 0.0384, 0.0385, 0.0374, 0.0422], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 05:38:40,005 INFO [finetune.py:976] (6/7) Epoch 14, batch 2600, loss[loss=0.1388, simple_loss=0.2112, pruned_loss=0.03318, over 4772.00 frames. ], tot_loss[loss=0.1848, simple_loss=0.2529, pruned_loss=0.05832, over 952093.27 frames. ], batch size: 26, lr: 3.55e-03, grad_scale: 64.0 2023-04-27 05:39:20,550 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=77095.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 05:39:41,103 INFO [finetune.py:976] (6/7) Epoch 14, batch 2650, loss[loss=0.1478, simple_loss=0.2198, pruned_loss=0.03797, over 4763.00 frames. ], tot_loss[loss=0.1852, simple_loss=0.2538, pruned_loss=0.05827, over 952185.77 frames. ], batch size: 27, lr: 3.54e-03, grad_scale: 64.0 2023-04-27 05:40:19,437 INFO [optim.py:369] (6/7) 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 05:40:21,598 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=3.92 vs. limit=5.0 2023-04-27 05:40:26,291 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=77156.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 05:40:29,227 INFO [finetune.py:976] (6/7) Epoch 14, batch 2700, loss[loss=0.1428, simple_loss=0.2213, pruned_loss=0.03216, over 4932.00 frames. ], tot_loss[loss=0.1849, simple_loss=0.2536, pruned_loss=0.05805, over 951563.06 frames. ], batch size: 33, lr: 3.54e-03, grad_scale: 64.0 2023-04-27 05:40:53,253 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 14, batch 2750, loss[loss=0.2113, simple_loss=0.2649, pruned_loss=0.07885, over 4858.00 frames. ], tot_loss[loss=0.1823, simple_loss=0.2505, pruned_loss=0.05706, over 950998.14 frames. ], batch size: 49, lr: 3.54e-03, grad_scale: 64.0 2023-04-27 05:41:03,202 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=77211.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 05:41:05,633 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.1722, 2.5147, 1.0013, 1.4718, 2.0070, 1.2461, 3.5376, 1.8159], device='cuda:6'), covar=tensor([0.0664, 0.0702, 0.0905, 0.1276, 0.0529, 0.1015, 0.0244, 0.0638], device='cuda:6'), in_proj_covar=tensor([0.0052, 0.0067, 0.0049, 0.0047, 0.0051, 0.0052, 0.0076, 0.0052], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:6') 2023-04-27 05:41:37,790 INFO [optim.py:369] (6/7) 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 05:41:39,725 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=3.97 vs. limit=5.0 2023-04-27 05:41:48,209 INFO [finetune.py:976] (6/7) Epoch 14, batch 2800, loss[loss=0.1527, simple_loss=0.2265, pruned_loss=0.03943, over 4701.00 frames. ], tot_loss[loss=0.1806, simple_loss=0.248, pruned_loss=0.05662, over 950554.32 frames. ], batch size: 23, lr: 3.54e-03, grad_scale: 64.0 2023-04-27 05:41:55,155 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=77272.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 05:41:55,750 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=77273.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 05:41:58,595 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([3.4023, 3.3721, 2.5680, 3.8885, 3.3760, 3.3739, 1.4379, 3.2571], device='cuda:6'), covar=tensor([0.1495, 0.1279, 0.2879, 0.1990, 0.3307, 0.1775, 0.5576, 0.2559], device='cuda:6'), in_proj_covar=tensor([0.0244, 0.0212, 0.0248, 0.0302, 0.0296, 0.0245, 0.0270, 0.0271], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 05:41:58,647 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=77277.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 05:42:22,266 INFO [finetune.py:976] (6/7) Epoch 14, batch 2850, loss[loss=0.2651, simple_loss=0.3073, pruned_loss=0.1115, over 4811.00 frames. ], tot_loss[loss=0.18, simple_loss=0.2475, pruned_loss=0.05624, over 951787.63 frames. ], batch size: 51, lr: 3.54e-03, grad_scale: 64.0 2023-04-27 05:42:37,427 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=77334.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 05:42:39,861 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=77338.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 05:42:44,406 INFO [optim.py:369] (6/7) 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] (6/7) Epoch 14, batch 2900, loss[loss=0.2555, simple_loss=0.3099, pruned_loss=0.1006, over 4184.00 frames. ], tot_loss[loss=0.183, simple_loss=0.2514, pruned_loss=0.05732, over 951350.64 frames. ], batch size: 65, lr: 3.54e-03, grad_scale: 64.0 2023-04-27 05:43:03,957 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-04-27 05:43:07,211 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.77 vs. limit=2.0 2023-04-27 05:43:29,374 INFO [finetune.py:976] (6/7) Epoch 14, batch 2950, loss[loss=0.1968, simple_loss=0.2617, pruned_loss=0.06596, over 4822.00 frames. ], tot_loss[loss=0.185, simple_loss=0.2539, pruned_loss=0.05802, over 953589.06 frames. ], batch size: 33, lr: 3.54e-03, grad_scale: 64.0 2023-04-27 05:43:50,925 INFO [optim.py:369] (6/7) 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 05:43:56,098 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=77451.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 05:44:03,066 INFO [finetune.py:976] (6/7) Epoch 14, batch 3000, loss[loss=0.2197, simple_loss=0.2865, pruned_loss=0.07647, over 4896.00 frames. ], tot_loss[loss=0.1886, simple_loss=0.2573, pruned_loss=0.05994, over 953946.44 frames. ], batch size: 43, lr: 3.54e-03, grad_scale: 64.0 2023-04-27 05:44:03,066 INFO [finetune.py:1001] (6/7) Computing validation loss 2023-04-27 05:44:11,434 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.3123, 1.3810, 3.8817, 3.5997, 3.4871, 3.7194, 3.7779, 3.4276], device='cuda:6'), covar=tensor([0.7055, 0.5136, 0.1140, 0.2002, 0.1398, 0.1450, 0.0845, 0.1646], device='cuda:6'), in_proj_covar=tensor([0.0311, 0.0308, 0.0405, 0.0409, 0.0348, 0.0407, 0.0315, 0.0367], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-27 05:44:19,534 INFO [finetune.py:1010] (6/7) Epoch 14, validation: loss=0.1527, simple_loss=0.224, pruned_loss=0.04073, over 2265189.00 frames. 2023-04-27 05:44:19,534 INFO [finetune.py:1011] (6/7) Maximum memory allocated so far is 6345MB 2023-04-27 05:44:30,641 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-04-27 05:45:02,932 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=77495.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 05:45:24,240 INFO [finetune.py:976] (6/7) Epoch 14, batch 3050, loss[loss=0.1488, simple_loss=0.2268, pruned_loss=0.03539, over 4728.00 frames. ], tot_loss[loss=0.1886, simple_loss=0.2573, pruned_loss=0.05993, over 954459.25 frames. ], batch size: 59, lr: 3.54e-03, grad_scale: 64.0 2023-04-27 05:45:26,739 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.2204, 2.1534, 1.8164, 1.8827, 2.3244, 1.7765, 2.7097, 1.6393], device='cuda:6'), covar=tensor([0.3918, 0.1992, 0.4468, 0.3178, 0.1648, 0.2693, 0.1572, 0.4413], device='cuda:6'), in_proj_covar=tensor([0.0344, 0.0348, 0.0430, 0.0356, 0.0384, 0.0385, 0.0374, 0.0423], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 05:45:45,785 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=77543.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 05:45:46,942 INFO [optim.py:369] (6/7) 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] (6/7) Epoch 14, batch 3100, loss[loss=0.146, simple_loss=0.2187, pruned_loss=0.0367, over 4816.00 frames. ], tot_loss[loss=0.1846, simple_loss=0.2536, pruned_loss=0.05779, over 956108.15 frames. ], batch size: 41, lr: 3.54e-03, grad_scale: 64.0 2023-04-27 05:46:02,399 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=77567.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 05:46:36,287 INFO [finetune.py:976] (6/7) Epoch 14, batch 3150, loss[loss=0.1711, simple_loss=0.2338, pruned_loss=0.05424, over 4940.00 frames. ], tot_loss[loss=0.1835, simple_loss=0.2516, pruned_loss=0.05774, over 955974.46 frames. ], batch size: 33, lr: 3.54e-03, grad_scale: 32.0 2023-04-27 05:46:45,250 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=77616.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 05:47:05,390 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=77629.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 05:47:07,284 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7874, 1.8429, 1.0501, 1.4801, 1.9716, 1.6850, 1.5206, 1.5975], device='cuda:6'), covar=tensor([0.0507, 0.0374, 0.0336, 0.0541, 0.0257, 0.0523, 0.0517, 0.0567], device='cuda:6'), in_proj_covar=tensor([0.0029, 0.0025, 0.0023, 0.0029, 0.0020, 0.0029, 0.0028, 0.0029], device='cuda:6'), out_proj_covar=tensor([0.0050, 0.0045, 0.0038, 0.0051, 0.0038, 0.0049, 0.0049, 0.0050], device='cuda:6') 2023-04-27 05:47:07,854 INFO [zipformer.py:1188] (6/7) 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] (6/7) 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 05:47:31,071 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8890, 2.3742, 1.9109, 1.7474, 1.3807, 1.4028, 1.9740, 1.3033], device='cuda:6'), covar=tensor([0.1773, 0.1437, 0.1566, 0.1937, 0.2494, 0.2233, 0.1069, 0.2295], device='cuda:6'), in_proj_covar=tensor([0.0195, 0.0213, 0.0168, 0.0205, 0.0200, 0.0184, 0.0156, 0.0186], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:6') 2023-04-27 05:47:38,389 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.1851, 2.5970, 2.1291, 2.4885, 1.7984, 2.0651, 2.2424, 1.6673], device='cuda:6'), covar=tensor([0.1869, 0.1083, 0.0896, 0.1086, 0.3023, 0.1236, 0.1769, 0.2291], device='cuda:6'), in_proj_covar=tensor([0.0290, 0.0309, 0.0224, 0.0282, 0.0311, 0.0264, 0.0254, 0.0270], device='cuda:6'), out_proj_covar=tensor([1.1703e-04, 1.2317e-04, 8.9263e-05, 1.1249e-04, 1.2690e-04, 1.0575e-04, 1.0276e-04, 1.0786e-04], device='cuda:6') 2023-04-27 05:47:41,184 INFO [finetune.py:976] (6/7) Epoch 14, batch 3200, loss[loss=0.1518, simple_loss=0.2224, pruned_loss=0.04062, over 4787.00 frames. ], tot_loss[loss=0.1807, simple_loss=0.2485, pruned_loss=0.05648, over 955889.05 frames. ], batch size: 26, lr: 3.54e-03, grad_scale: 32.0 2023-04-27 05:48:10,696 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=77677.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 05:48:10,705 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=77677.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 05:48:22,202 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-04-27 05:48:37,446 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.5059, 1.8319, 1.8503, 1.9675, 1.8282, 1.9466, 1.9758, 1.9173], device='cuda:6'), covar=tensor([0.4460, 0.5578, 0.5249, 0.5084, 0.5741, 0.7639, 0.5611, 0.5378], device='cuda:6'), in_proj_covar=tensor([0.0330, 0.0375, 0.0316, 0.0329, 0.0340, 0.0399, 0.0355, 0.0325], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 05:48:48,229 INFO [finetune.py:976] (6/7) Epoch 14, batch 3250, loss[loss=0.1929, simple_loss=0.2659, pruned_loss=0.05998, over 4828.00 frames. ], tot_loss[loss=0.1806, simple_loss=0.2483, pruned_loss=0.05644, over 954451.07 frames. ], batch size: 30, lr: 3.54e-03, grad_scale: 32.0 2023-04-27 05:49:28,791 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=77738.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 05:49:33,563 INFO [optim.py:369] (6/7) 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 05:49:36,703 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 14, batch 3300, loss[loss=0.1787, simple_loss=0.2393, pruned_loss=0.05912, over 4732.00 frames. ], tot_loss[loss=0.1839, simple_loss=0.2521, pruned_loss=0.0578, over 955241.20 frames. ], batch size: 23, lr: 3.54e-03, grad_scale: 32.0 2023-04-27 05:50:08,404 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=77799.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 05:50:09,193 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9220, 1.6634, 4.9129, 4.5699, 4.3303, 4.6695, 4.4388, 4.4032], device='cuda:6'), covar=tensor([0.7224, 0.5608, 0.1003, 0.2019, 0.1144, 0.1317, 0.1493, 0.1606], device='cuda:6'), in_proj_covar=tensor([0.0315, 0.0310, 0.0408, 0.0411, 0.0350, 0.0410, 0.0317, 0.0369], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-27 05:50:09,571 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-04-27 05:50:15,838 INFO [finetune.py:976] (6/7) Epoch 14, batch 3350, loss[loss=0.2186, simple_loss=0.2807, pruned_loss=0.07829, over 4888.00 frames. ], tot_loss[loss=0.1845, simple_loss=0.2535, pruned_loss=0.05775, over 955139.78 frames. ], batch size: 35, lr: 3.54e-03, grad_scale: 32.0 2023-04-27 05:50:38,295 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.21 vs. limit=5.0 2023-04-27 05:50:39,909 INFO [optim.py:369] (6/7) 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] (6/7) Epoch 14, batch 3400, loss[loss=0.1821, simple_loss=0.2671, pruned_loss=0.04858, over 4821.00 frames. ], tot_loss[loss=0.1847, simple_loss=0.2541, pruned_loss=0.05762, over 956247.43 frames. ], batch size: 47, lr: 3.54e-03, grad_scale: 32.0 2023-04-27 05:50:52,864 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=77867.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 05:51:00,589 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([3.8054, 3.7128, 2.8547, 4.4304, 3.8572, 3.7086, 1.8388, 3.7459], device='cuda:6'), covar=tensor([0.1715, 0.1245, 0.3085, 0.1367, 0.2642, 0.1746, 0.5963, 0.2393], device='cuda:6'), in_proj_covar=tensor([0.0246, 0.0216, 0.0252, 0.0307, 0.0300, 0.0249, 0.0274, 0.0275], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 05:51:00,656 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.1343, 1.5919, 1.9514, 2.4694, 1.9715, 1.5025, 1.2983, 1.7464], device='cuda:6'), covar=tensor([0.3812, 0.3879, 0.2035, 0.2726, 0.3107, 0.3327, 0.4730, 0.2426], device='cuda:6'), in_proj_covar=tensor([0.0288, 0.0247, 0.0223, 0.0317, 0.0215, 0.0230, 0.0229, 0.0185], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:6') 2023-04-27 05:51:22,436 INFO [finetune.py:976] (6/7) Epoch 14, batch 3450, loss[loss=0.1728, simple_loss=0.2343, pruned_loss=0.05567, over 4824.00 frames. ], tot_loss[loss=0.1851, simple_loss=0.2546, pruned_loss=0.05779, over 956119.44 frames. ], batch size: 47, lr: 3.54e-03, grad_scale: 32.0 2023-04-27 05:51:24,874 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=77915.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 05:51:33,833 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=77929.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 05:51:36,784 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=77933.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 05:51:45,494 INFO [optim.py:369] (6/7) 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 05:51:54,680 INFO [finetune.py:976] (6/7) Epoch 14, batch 3500, loss[loss=0.1883, simple_loss=0.2576, pruned_loss=0.05946, over 4228.00 frames. ], tot_loss[loss=0.1831, simple_loss=0.2518, pruned_loss=0.05719, over 955972.78 frames. ], batch size: 65, lr: 3.54e-03, grad_scale: 32.0 2023-04-27 05:52:01,406 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=77972.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 05:52:04,956 INFO [zipformer.py:1188] (6/7) 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] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=77981.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 05:52:28,916 INFO [finetune.py:976] (6/7) Epoch 14, batch 3550, loss[loss=0.1938, simple_loss=0.2565, pruned_loss=0.06558, over 4820.00 frames. ], tot_loss[loss=0.1822, simple_loss=0.2504, pruned_loss=0.05702, over 956572.54 frames. ], batch size: 38, lr: 3.54e-03, grad_scale: 32.0 2023-04-27 05:52:42,973 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=78033.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 05:52:56,890 INFO [optim.py:369] (6/7) 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] (6/7) Epoch 14, batch 3600, loss[loss=0.1858, simple_loss=0.248, pruned_loss=0.0618, over 4856.00 frames. ], tot_loss[loss=0.18, simple_loss=0.2476, pruned_loss=0.05624, over 956420.53 frames. ], batch size: 44, lr: 3.54e-03, grad_scale: 32.0 2023-04-27 05:53:19,448 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5582, 3.3116, 0.9214, 1.9287, 1.8464, 2.4391, 1.9191, 0.9970], device='cuda:6'), covar=tensor([0.1250, 0.0805, 0.1978, 0.1108, 0.1047, 0.0978, 0.1378, 0.2125], device='cuda:6'), in_proj_covar=tensor([0.0117, 0.0245, 0.0138, 0.0121, 0.0131, 0.0153, 0.0118, 0.0120], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-27 05:53:39,817 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5902, 1.5483, 1.9745, 1.9273, 1.4958, 1.3038, 1.5907, 1.1314], device='cuda:6'), covar=tensor([0.0613, 0.0614, 0.0390, 0.0576, 0.0750, 0.1170, 0.0535, 0.0666], device='cuda:6'), in_proj_covar=tensor([0.0068, 0.0070, 0.0069, 0.0068, 0.0075, 0.0097, 0.0075, 0.0069], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-27 05:54:18,584 INFO [finetune.py:976] (6/7) Epoch 14, batch 3650, loss[loss=0.2227, simple_loss=0.3019, pruned_loss=0.07177, over 4829.00 frames. ], tot_loss[loss=0.1822, simple_loss=0.2501, pruned_loss=0.05714, over 954262.37 frames. ], batch size: 40, lr: 3.54e-03, grad_scale: 32.0 2023-04-27 05:54:31,746 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5615, 1.4528, 1.8172, 1.7371, 1.3662, 1.2491, 1.4296, 0.9357], device='cuda:6'), covar=tensor([0.0559, 0.0759, 0.0424, 0.0623, 0.0822, 0.1137, 0.0597, 0.0684], device='cuda:6'), in_proj_covar=tensor([0.0068, 0.0070, 0.0070, 0.0068, 0.0076, 0.0097, 0.0076, 0.0070], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-27 05:54:56,796 INFO [optim.py:369] (6/7) 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 05:54:57,710 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.74 vs. limit=2.0 2023-04-27 05:55:18,203 INFO [finetune.py:976] (6/7) Epoch 14, batch 3700, loss[loss=0.2212, simple_loss=0.2962, pruned_loss=0.07306, over 4918.00 frames. ], tot_loss[loss=0.1847, simple_loss=0.2536, pruned_loss=0.05791, over 955017.99 frames. ], batch size: 42, lr: 3.54e-03, grad_scale: 32.0 2023-04-27 05:55:56,841 INFO [finetune.py:976] (6/7) Epoch 14, batch 3750, loss[loss=0.1934, simple_loss=0.262, pruned_loss=0.0624, over 4893.00 frames. ], tot_loss[loss=0.1851, simple_loss=0.2546, pruned_loss=0.05783, over 953963.29 frames. ], batch size: 37, lr: 3.54e-03, grad_scale: 32.0 2023-04-27 05:56:09,781 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=78232.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 05:56:18,616 INFO [optim.py:369] (6/7) 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] (6/7) Epoch 14, batch 3800, loss[loss=0.2083, simple_loss=0.2778, pruned_loss=0.06936, over 4823.00 frames. ], tot_loss[loss=0.1865, simple_loss=0.256, pruned_loss=0.05853, over 952014.35 frames. ], batch size: 33, lr: 3.53e-03, grad_scale: 32.0 2023-04-27 05:56:37,435 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=78272.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 05:56:38,678 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.4420, 3.2139, 1.1761, 1.9085, 1.7928, 2.3857, 1.8941, 1.0326], device='cuda:6'), covar=tensor([0.1344, 0.0946, 0.1713, 0.1124, 0.1043, 0.0921, 0.1403, 0.2056], device='cuda:6'), in_proj_covar=tensor([0.0117, 0.0245, 0.0138, 0.0121, 0.0132, 0.0153, 0.0118, 0.0121], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-27 05:56:50,291 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=78293.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 05:57:03,399 INFO [finetune.py:976] (6/7) Epoch 14, batch 3850, loss[loss=0.2075, simple_loss=0.2659, pruned_loss=0.07457, over 4823.00 frames. ], tot_loss[loss=0.1862, simple_loss=0.2555, pruned_loss=0.05845, over 952674.25 frames. ], batch size: 30, lr: 3.53e-03, grad_scale: 32.0 2023-04-27 05:57:09,830 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=78320.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 05:57:16,682 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8098, 1.1219, 1.5711, 1.7012, 1.6601, 1.7349, 1.5861, 1.5880], device='cuda:6'), covar=tensor([0.4078, 0.5335, 0.4773, 0.4785, 0.5494, 0.7730, 0.5070, 0.4705], device='cuda:6'), in_proj_covar=tensor([0.0332, 0.0376, 0.0319, 0.0331, 0.0343, 0.0399, 0.0355, 0.0326], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 05:57:17,857 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=78333.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 05:57:21,543 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.6387, 2.3420, 2.7772, 3.0699, 3.0107, 2.4956, 2.0285, 2.5324], device='cuda:6'), covar=tensor([0.0783, 0.0970, 0.0473, 0.0517, 0.0551, 0.0806, 0.0705, 0.0584], device='cuda:6'), in_proj_covar=tensor([0.0191, 0.0202, 0.0182, 0.0172, 0.0177, 0.0182, 0.0153, 0.0180], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 05:57:25,785 INFO [optim.py:369] (6/7) 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 05:57:32,962 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-04-27 05:57:35,702 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7872, 1.7550, 0.8550, 1.4090, 1.8369, 1.6328, 1.5105, 1.5813], device='cuda:6'), covar=tensor([0.0490, 0.0366, 0.0354, 0.0555, 0.0259, 0.0511, 0.0489, 0.0564], device='cuda:6'), in_proj_covar=tensor([0.0029, 0.0025, 0.0023, 0.0029, 0.0020, 0.0029, 0.0028, 0.0029], device='cuda:6'), out_proj_covar=tensor([0.0050, 0.0045, 0.0038, 0.0051, 0.0038, 0.0049, 0.0049, 0.0050], device='cuda:6') 2023-04-27 05:57:36,829 INFO [finetune.py:976] (6/7) Epoch 14, batch 3900, loss[loss=0.178, simple_loss=0.2379, pruned_loss=0.05907, over 4827.00 frames. ], tot_loss[loss=0.1842, simple_loss=0.2533, pruned_loss=0.05759, over 953604.64 frames. ], batch size: 25, lr: 3.53e-03, grad_scale: 32.0 2023-04-27 05:57:44,791 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-04-27 05:57:50,411 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=78381.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 05:57:50,443 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5367, 3.3614, 0.9932, 1.8666, 1.8596, 2.4034, 1.9334, 0.9704], device='cuda:6'), covar=tensor([0.1300, 0.0959, 0.1908, 0.1219, 0.1030, 0.1047, 0.1398, 0.2043], device='cuda:6'), in_proj_covar=tensor([0.0117, 0.0245, 0.0138, 0.0121, 0.0132, 0.0153, 0.0118, 0.0120], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-27 05:57:51,338 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.79 vs. limit=5.0 2023-04-27 05:58:04,664 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.70 vs. limit=2.0 2023-04-27 05:58:09,806 INFO [finetune.py:976] (6/7) Epoch 14, batch 3950, loss[loss=0.211, simple_loss=0.2665, pruned_loss=0.07776, over 4766.00 frames. ], tot_loss[loss=0.1809, simple_loss=0.2493, pruned_loss=0.05622, over 954016.30 frames. ], batch size: 28, lr: 3.53e-03, grad_scale: 32.0 2023-04-27 05:58:17,804 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8480, 2.0552, 1.0919, 1.5189, 2.1292, 1.7159, 1.6159, 1.6890], device='cuda:6'), covar=tensor([0.0497, 0.0335, 0.0307, 0.0528, 0.0242, 0.0510, 0.0513, 0.0522], device='cuda:6'), in_proj_covar=tensor([0.0029, 0.0025, 0.0023, 0.0029, 0.0020, 0.0029, 0.0028, 0.0029], device='cuda:6'), out_proj_covar=tensor([0.0050, 0.0045, 0.0038, 0.0051, 0.0038, 0.0049, 0.0049, 0.0050], device='cuda:6') 2023-04-27 05:58:29,222 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.1004, 1.6022, 1.6454, 2.0157, 1.9338, 1.9354, 1.5198, 4.2260], device='cuda:6'), covar=tensor([0.0647, 0.0885, 0.0865, 0.1225, 0.0646, 0.0635, 0.0825, 0.0152], device='cuda:6'), in_proj_covar=tensor([0.0038, 0.0039, 0.0040, 0.0043, 0.0040, 0.0038, 0.0038, 0.0057], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:6') 2023-04-27 05:58:50,266 INFO [optim.py:369] (6/7) 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 05:59:04,633 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=78460.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 05:59:05,120 INFO [finetune.py:976] (6/7) Epoch 14, batch 4000, loss[loss=0.151, simple_loss=0.228, pruned_loss=0.03699, over 4808.00 frames. ], tot_loss[loss=0.1801, simple_loss=0.2483, pruned_loss=0.05596, over 956407.65 frames. ], batch size: 38, lr: 3.53e-03, grad_scale: 32.0 2023-04-27 05:59:58,041 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.65 vs. limit=2.0 2023-04-27 06:00:00,330 INFO [finetune.py:976] (6/7) Epoch 14, batch 4050, loss[loss=0.1487, simple_loss=0.2303, pruned_loss=0.03351, over 4859.00 frames. ], tot_loss[loss=0.1805, simple_loss=0.249, pruned_loss=0.05598, over 956730.84 frames. ], batch size: 31, lr: 3.53e-03, grad_scale: 32.0 2023-04-27 06:00:19,673 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=78521.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:00:51,060 INFO [optim.py:369] (6/7) 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 06:01:03,062 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 14, batch 4100, loss[loss=0.2245, simple_loss=0.2815, pruned_loss=0.08374, over 4850.00 frames. ], tot_loss[loss=0.1849, simple_loss=0.2536, pruned_loss=0.05812, over 955203.19 frames. ], batch size: 31, lr: 3.53e-03, grad_scale: 32.0 2023-04-27 06:01:32,157 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.14 vs. limit=2.0 2023-04-27 06:01:33,669 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9724, 2.6623, 0.9866, 1.4195, 2.1477, 1.1949, 3.5855, 1.6911], device='cuda:6'), covar=tensor([0.0711, 0.0666, 0.0835, 0.1241, 0.0491, 0.1028, 0.0230, 0.0670], device='cuda:6'), in_proj_covar=tensor([0.0051, 0.0066, 0.0049, 0.0047, 0.0050, 0.0052, 0.0076, 0.0051], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:6') 2023-04-27 06:01:45,287 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=78588.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 06:02:05,270 INFO [finetune.py:976] (6/7) Epoch 14, batch 4150, loss[loss=0.1827, simple_loss=0.2615, pruned_loss=0.05199, over 4885.00 frames. ], tot_loss[loss=0.1855, simple_loss=0.2543, pruned_loss=0.05839, over 954434.03 frames. ], batch size: 43, lr: 3.53e-03, grad_scale: 32.0 2023-04-27 06:02:09,572 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=78617.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:02:19,991 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-04-27 06:02:29,495 INFO [optim.py:369] (6/7) 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 06:02:32,948 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.68 vs. limit=5.0 2023-04-27 06:02:35,188 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-04-27 06:02:38,717 INFO [finetune.py:976] (6/7) Epoch 14, batch 4200, loss[loss=0.1882, simple_loss=0.2578, pruned_loss=0.05926, over 4798.00 frames. ], tot_loss[loss=0.1859, simple_loss=0.2556, pruned_loss=0.05805, over 954062.53 frames. ], batch size: 40, lr: 3.53e-03, grad_scale: 32.0 2023-04-27 06:03:08,462 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0174, 2.3716, 0.9622, 1.2324, 1.8200, 1.2213, 2.9679, 1.5020], device='cuda:6'), covar=tensor([0.0654, 0.0613, 0.0799, 0.1312, 0.0493, 0.1060, 0.0277, 0.0704], device='cuda:6'), in_proj_covar=tensor([0.0051, 0.0066, 0.0049, 0.0047, 0.0050, 0.0052, 0.0076, 0.0051], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:6') 2023-04-27 06:03:12,003 INFO [finetune.py:976] (6/7) Epoch 14, batch 4250, loss[loss=0.1976, simple_loss=0.2674, pruned_loss=0.06397, over 4878.00 frames. ], tot_loss[loss=0.1835, simple_loss=0.2532, pruned_loss=0.05688, over 954053.82 frames. ], batch size: 43, lr: 3.53e-03, grad_scale: 32.0 2023-04-27 06:03:29,645 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6327, 0.7268, 1.4410, 1.9594, 1.7202, 1.5481, 1.5810, 1.5589], device='cuda:6'), covar=tensor([0.4375, 0.6255, 0.5984, 0.6295, 0.5980, 0.7240, 0.6942, 0.7541], device='cuda:6'), in_proj_covar=tensor([0.0416, 0.0408, 0.0498, 0.0512, 0.0447, 0.0466, 0.0474, 0.0477], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 06:03:36,181 INFO [optim.py:369] (6/7) 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 06:03:45,439 INFO [finetune.py:976] (6/7) Epoch 14, batch 4300, loss[loss=0.1418, simple_loss=0.2205, pruned_loss=0.03155, over 4831.00 frames. ], tot_loss[loss=0.1809, simple_loss=0.2502, pruned_loss=0.05579, over 954901.99 frames. ], batch size: 30, lr: 3.53e-03, grad_scale: 32.0 2023-04-27 06:04:19,292 INFO [finetune.py:976] (6/7) Epoch 14, batch 4350, loss[loss=0.2113, simple_loss=0.2576, pruned_loss=0.08249, over 4741.00 frames. ], tot_loss[loss=0.1782, simple_loss=0.2471, pruned_loss=0.05464, over 955112.96 frames. ], batch size: 54, lr: 3.53e-03, grad_scale: 32.0 2023-04-27 06:04:22,431 INFO [zipformer.py:1188] (6/7) 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] (6/7) 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] (6/7) Epoch 14, batch 4400, loss[loss=0.2395, simple_loss=0.3051, pruned_loss=0.08699, over 4905.00 frames. ], tot_loss[loss=0.1802, simple_loss=0.2486, pruned_loss=0.05591, over 953834.37 frames. ], batch size: 43, lr: 3.53e-03, grad_scale: 32.0 2023-04-27 06:05:38,864 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=78888.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:06:11,508 INFO [finetune.py:976] (6/7) Epoch 14, batch 4450, loss[loss=0.1999, simple_loss=0.2763, pruned_loss=0.06171, over 4860.00 frames. ], tot_loss[loss=0.1829, simple_loss=0.252, pruned_loss=0.05693, over 952708.39 frames. ], batch size: 49, lr: 3.53e-03, grad_scale: 32.0 2023-04-27 06:06:12,194 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=78912.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:06:44,185 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=78936.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:07:03,296 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.61 vs. limit=2.0 2023-04-27 06:07:03,710 INFO [optim.py:369] (6/7) 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 06:07:18,902 INFO [finetune.py:976] (6/7) Epoch 14, batch 4500, loss[loss=0.1869, simple_loss=0.26, pruned_loss=0.05689, over 4242.00 frames. ], tot_loss[loss=0.1831, simple_loss=0.2526, pruned_loss=0.05675, over 952556.79 frames. ], batch size: 65, lr: 3.53e-03, grad_scale: 32.0 2023-04-27 06:07:21,402 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=78965.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:07:52,199 INFO [finetune.py:976] (6/7) Epoch 14, batch 4550, loss[loss=0.1796, simple_loss=0.254, pruned_loss=0.05259, over 4732.00 frames. ], tot_loss[loss=0.1846, simple_loss=0.2545, pruned_loss=0.05735, over 954276.47 frames. ], batch size: 27, lr: 3.53e-03, grad_scale: 32.0 2023-04-27 06:08:01,472 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=79026.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:08:14,993 INFO [optim.py:369] (6/7) 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 06:08:26,106 INFO [finetune.py:976] (6/7) Epoch 14, batch 4600, loss[loss=0.1577, simple_loss=0.2316, pruned_loss=0.0419, over 4745.00 frames. ], tot_loss[loss=0.1846, simple_loss=0.2545, pruned_loss=0.05734, over 954942.55 frames. ], batch size: 27, lr: 3.53e-03, grad_scale: 32.0 2023-04-27 06:08:27,959 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9567, 1.4336, 5.1751, 4.7859, 4.5245, 4.9179, 4.5246, 4.6404], device='cuda:6'), covar=tensor([0.6723, 0.6015, 0.0848, 0.1709, 0.0968, 0.1332, 0.1403, 0.1458], device='cuda:6'), in_proj_covar=tensor([0.0311, 0.0306, 0.0402, 0.0407, 0.0347, 0.0407, 0.0312, 0.0365], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-27 06:08:59,227 INFO [finetune.py:976] (6/7) Epoch 14, batch 4650, loss[loss=0.1928, simple_loss=0.252, pruned_loss=0.06677, over 4894.00 frames. ], tot_loss[loss=0.1838, simple_loss=0.2527, pruned_loss=0.05748, over 955259.86 frames. ], batch size: 32, lr: 3.53e-03, grad_scale: 32.0 2023-04-27 06:09:00,347 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-04-27 06:09:02,384 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=79116.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:09:21,597 INFO [optim.py:369] (6/7) 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 06:09:32,657 INFO [finetune.py:976] (6/7) Epoch 14, batch 4700, loss[loss=0.1629, simple_loss=0.2242, pruned_loss=0.05077, over 4921.00 frames. ], tot_loss[loss=0.1796, simple_loss=0.2478, pruned_loss=0.05567, over 955196.53 frames. ], batch size: 46, lr: 3.53e-03, grad_scale: 32.0 2023-04-27 06:09:34,554 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=79164.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:10:05,833 INFO [finetune.py:976] (6/7) Epoch 14, batch 4750, loss[loss=0.1782, simple_loss=0.2445, pruned_loss=0.05596, over 4829.00 frames. ], tot_loss[loss=0.1783, simple_loss=0.2459, pruned_loss=0.05532, over 954939.91 frames. ], batch size: 49, lr: 3.53e-03, grad_scale: 32.0 2023-04-27 06:10:06,535 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=79212.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:10:13,167 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5239, 0.6865, 1.4396, 1.9261, 1.6233, 1.4620, 1.4571, 1.4813], device='cuda:6'), covar=tensor([0.4686, 0.6535, 0.6118, 0.5988, 0.5690, 0.7131, 0.7135, 0.7483], device='cuda:6'), in_proj_covar=tensor([0.0415, 0.0407, 0.0497, 0.0511, 0.0445, 0.0466, 0.0473, 0.0477], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 06:10:37,980 INFO [optim.py:369] (6/7) 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 06:10:58,676 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=79260.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:10:59,216 INFO [finetune.py:976] (6/7) Epoch 14, batch 4800, loss[loss=0.2125, simple_loss=0.2781, pruned_loss=0.07343, over 4821.00 frames. ], tot_loss[loss=0.1806, simple_loss=0.2488, pruned_loss=0.05614, over 957294.29 frames. ], batch size: 47, lr: 3.53e-03, grad_scale: 32.0 2023-04-27 06:11:30,778 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=79285.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 06:11:58,652 INFO [finetune.py:976] (6/7) Epoch 14, batch 4850, loss[loss=0.1641, simple_loss=0.2427, pruned_loss=0.04279, over 4789.00 frames. ], tot_loss[loss=0.1829, simple_loss=0.2516, pruned_loss=0.05711, over 956376.61 frames. ], batch size: 29, lr: 3.53e-03, grad_scale: 32.0 2023-04-27 06:12:10,416 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=79321.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:12:23,689 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=79333.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:12:34,624 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([3.3916, 2.6539, 2.5524, 2.6648, 2.3838, 2.6737, 2.7109, 2.6274], device='cuda:6'), covar=tensor([0.3813, 0.6015, 0.4990, 0.4922, 0.5828, 0.6844, 0.5726, 0.5633], device='cuda:6'), in_proj_covar=tensor([0.0335, 0.0379, 0.0319, 0.0334, 0.0343, 0.0402, 0.0359, 0.0329], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 06:12:36,867 INFO [optim.py:369] (6/7) 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 06:12:42,260 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=79346.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 06:12:55,208 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0871, 2.3785, 0.9951, 1.2718, 1.7656, 1.1673, 3.0714, 1.5347], device='cuda:6'), covar=tensor([0.0701, 0.0708, 0.0794, 0.1488, 0.0516, 0.1147, 0.0315, 0.0745], device='cuda:6'), in_proj_covar=tensor([0.0051, 0.0066, 0.0049, 0.0047, 0.0050, 0.0052, 0.0076, 0.0051], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:6') 2023-04-27 06:12:56,921 INFO [finetune.py:976] (6/7) Epoch 14, batch 4900, loss[loss=0.215, simple_loss=0.2903, pruned_loss=0.06986, over 4809.00 frames. ], tot_loss[loss=0.1847, simple_loss=0.2533, pruned_loss=0.058, over 953828.36 frames. ], batch size: 41, lr: 3.53e-03, grad_scale: 32.0 2023-04-27 06:13:00,224 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=79364.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:13:07,486 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=79375.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 06:13:19,225 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 14, batch 4950, loss[loss=0.2166, simple_loss=0.2682, pruned_loss=0.08246, over 4160.00 frames. ], tot_loss[loss=0.1861, simple_loss=0.2548, pruned_loss=0.05873, over 953110.53 frames. ], batch size: 65, lr: 3.52e-03, grad_scale: 32.0 2023-04-27 06:13:39,240 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6303, 1.2994, 4.6520, 4.3403, 4.0126, 4.4536, 4.2483, 4.1109], device='cuda:6'), covar=tensor([0.6987, 0.6378, 0.0948, 0.1586, 0.1077, 0.1768, 0.1103, 0.1495], device='cuda:6'), in_proj_covar=tensor([0.0312, 0.0307, 0.0403, 0.0407, 0.0349, 0.0407, 0.0313, 0.0366], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-27 06:13:40,487 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=79425.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:13:47,626 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=79436.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 06:13:53,550 INFO [optim.py:369] (6/7) 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 06:14:03,223 INFO [finetune.py:976] (6/7) Epoch 14, batch 5000, loss[loss=0.1716, simple_loss=0.2421, pruned_loss=0.05054, over 4683.00 frames. ], tot_loss[loss=0.1845, simple_loss=0.2532, pruned_loss=0.05796, over 953631.31 frames. ], batch size: 23, lr: 3.52e-03, grad_scale: 32.0 2023-04-27 06:14:13,811 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.3038, 1.7566, 1.7364, 1.8483, 1.8651, 2.1395, 1.5867, 3.2950], device='cuda:6'), covar=tensor([0.0577, 0.0618, 0.0617, 0.0996, 0.0507, 0.0515, 0.0629, 0.0197], device='cuda:6'), in_proj_covar=tensor([0.0038, 0.0039, 0.0040, 0.0044, 0.0040, 0.0038, 0.0038, 0.0057], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:6') 2023-04-27 06:14:35,550 INFO [finetune.py:976] (6/7) Epoch 14, batch 5050, loss[loss=0.1912, simple_loss=0.2565, pruned_loss=0.06295, over 4817.00 frames. ], tot_loss[loss=0.1816, simple_loss=0.2501, pruned_loss=0.0566, over 955404.94 frames. ], batch size: 41, lr: 3.52e-03, grad_scale: 16.0 2023-04-27 06:14:59,517 INFO [optim.py:369] (6/7) 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 06:15:02,724 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.1620, 2.5446, 0.9132, 1.3667, 1.7609, 1.1907, 3.3345, 1.7552], device='cuda:6'), covar=tensor([0.0661, 0.0670, 0.0830, 0.1271, 0.0554, 0.1058, 0.0469, 0.0670], device='cuda:6'), in_proj_covar=tensor([0.0051, 0.0066, 0.0048, 0.0047, 0.0050, 0.0052, 0.0075, 0.0051], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:6') 2023-04-27 06:15:03,533 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.23 vs. limit=5.0 2023-04-27 06:15:08,091 INFO [finetune.py:976] (6/7) Epoch 14, batch 5100, loss[loss=0.1777, simple_loss=0.2333, pruned_loss=0.06109, over 4839.00 frames. ], tot_loss[loss=0.1789, simple_loss=0.2469, pruned_loss=0.0554, over 956580.89 frames. ], batch size: 47, lr: 3.52e-03, grad_scale: 16.0 2023-04-27 06:15:14,412 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.47 vs. limit=2.0 2023-04-27 06:15:17,662 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([3.0674, 2.4041, 2.2180, 2.4956, 2.2506, 2.4305, 2.4328, 2.3031], device='cuda:6'), covar=tensor([0.4239, 0.6719, 0.5868, 0.5087, 0.6233, 0.7725, 0.6438, 0.5920], device='cuda:6'), in_proj_covar=tensor([0.0334, 0.0378, 0.0319, 0.0333, 0.0343, 0.0400, 0.0357, 0.0328], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 06:15:25,183 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=79585.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:15:41,438 INFO [finetune.py:976] (6/7) Epoch 14, batch 5150, loss[loss=0.1972, simple_loss=0.2677, pruned_loss=0.06329, over 4710.00 frames. ], tot_loss[loss=0.1802, simple_loss=0.2481, pruned_loss=0.05621, over 954605.17 frames. ], batch size: 59, lr: 3.52e-03, grad_scale: 16.0 2023-04-27 06:15:48,505 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.2977, 4.4925, 1.2258, 2.2310, 2.6589, 3.1066, 2.5794, 1.0610], device='cuda:6'), covar=tensor([0.1254, 0.0858, 0.2081, 0.1358, 0.0969, 0.1087, 0.1464, 0.2129], device='cuda:6'), in_proj_covar=tensor([0.0118, 0.0247, 0.0139, 0.0122, 0.0132, 0.0154, 0.0119, 0.0121], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-27 06:15:48,521 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=79621.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:15:50,940 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.2006, 2.5324, 0.8325, 1.4422, 1.5927, 1.9022, 1.6814, 0.8062], device='cuda:6'), covar=tensor([0.1544, 0.1402, 0.1818, 0.1452, 0.1144, 0.1047, 0.1640, 0.1695], device='cuda:6'), in_proj_covar=tensor([0.0118, 0.0246, 0.0139, 0.0122, 0.0132, 0.0154, 0.0119, 0.0121], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-27 06:16:02,441 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=79641.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 06:16:06,014 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=79646.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:16:06,492 INFO [optim.py:369] (6/7) 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 06:16:12,611 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=79657.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:16:14,972 INFO [finetune.py:976] (6/7) Epoch 14, batch 5200, loss[loss=0.2036, simple_loss=0.2755, pruned_loss=0.0659, over 4901.00 frames. ], tot_loss[loss=0.1826, simple_loss=0.251, pruned_loss=0.05713, over 954359.67 frames. ], batch size: 35, lr: 3.52e-03, grad_scale: 16.0 2023-04-27 06:16:19,855 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=79669.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:16:23,282 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5574, 1.4919, 1.7250, 1.9958, 2.1042, 1.5652, 1.2812, 1.7905], device='cuda:6'), covar=tensor([0.0940, 0.1240, 0.0821, 0.0619, 0.0624, 0.0905, 0.0874, 0.0605], device='cuda:6'), in_proj_covar=tensor([0.0190, 0.0201, 0.0182, 0.0172, 0.0177, 0.0182, 0.0154, 0.0180], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 06:16:34,496 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 14, batch 5250, loss[loss=0.2122, simple_loss=0.2842, pruned_loss=0.07012, over 4818.00 frames. ], tot_loss[loss=0.1852, simple_loss=0.254, pruned_loss=0.05827, over 955472.54 frames. ], batch size: 40, lr: 3.52e-03, grad_scale: 16.0 2023-04-27 06:17:03,885 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=79718.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:17:05,052 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=79720.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:17:19,838 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=79731.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 06:17:31,767 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-04-27 06:17:36,848 INFO [optim.py:369] (6/7) 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 06:17:44,807 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6514, 2.1424, 1.8371, 2.0314, 1.5686, 1.7910, 1.7775, 1.4242], device='cuda:6'), covar=tensor([0.1960, 0.1336, 0.0891, 0.1194, 0.3400, 0.1094, 0.1743, 0.2537], device='cuda:6'), in_proj_covar=tensor([0.0291, 0.0312, 0.0225, 0.0283, 0.0313, 0.0265, 0.0255, 0.0270], device='cuda:6'), out_proj_covar=tensor([1.1730e-04, 1.2424e-04, 8.9584e-05, 1.1290e-04, 1.2757e-04, 1.0591e-04, 1.0308e-04, 1.0792e-04], device='cuda:6') 2023-04-27 06:17:45,877 INFO [finetune.py:976] (6/7) Epoch 14, batch 5300, loss[loss=0.1816, simple_loss=0.2514, pruned_loss=0.0559, over 4272.00 frames. ], tot_loss[loss=0.1861, simple_loss=0.2557, pruned_loss=0.05827, over 954794.79 frames. ], batch size: 65, lr: 3.52e-03, grad_scale: 16.0 2023-04-27 06:17:46,601 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=79762.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:17:52,734 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.67 vs. limit=2.0 2023-04-27 06:18:38,536 INFO [finetune.py:976] (6/7) Epoch 14, batch 5350, loss[loss=0.2759, simple_loss=0.3352, pruned_loss=0.1082, over 4150.00 frames. ], tot_loss[loss=0.1859, simple_loss=0.256, pruned_loss=0.05796, over 955603.88 frames. ], batch size: 65, lr: 3.52e-03, grad_scale: 16.0 2023-04-27 06:18:45,995 INFO [zipformer.py:1188] (6/7) 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] (6/7) 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 06:19:07,902 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.2398, 2.2758, 1.8539, 1.8456, 2.2861, 1.7275, 2.7875, 1.4506], device='cuda:6'), covar=tensor([0.3427, 0.1661, 0.3988, 0.2953, 0.1705, 0.2593, 0.1324, 0.4746], device='cuda:6'), in_proj_covar=tensor([0.0342, 0.0347, 0.0427, 0.0355, 0.0382, 0.0380, 0.0372, 0.0419], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 06:19:11,410 INFO [finetune.py:976] (6/7) Epoch 14, batch 5400, loss[loss=0.1919, simple_loss=0.261, pruned_loss=0.06135, over 4898.00 frames. ], tot_loss[loss=0.1829, simple_loss=0.2526, pruned_loss=0.0566, over 955629.93 frames. ], batch size: 35, lr: 3.52e-03, grad_scale: 16.0 2023-04-27 06:19:45,249 INFO [finetune.py:976] (6/7) Epoch 14, batch 5450, loss[loss=0.1627, simple_loss=0.2261, pruned_loss=0.04967, over 4736.00 frames. ], tot_loss[loss=0.179, simple_loss=0.248, pruned_loss=0.05503, over 955366.72 frames. ], batch size: 59, lr: 3.52e-03, grad_scale: 16.0 2023-04-27 06:20:04,594 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=79941.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:20:04,611 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=79941.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 06:20:05,914 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-04-27 06:20:09,136 INFO [optim.py:369] (6/7) 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] (6/7) Epoch 14, batch 5500, loss[loss=0.2067, simple_loss=0.2677, pruned_loss=0.07283, over 4771.00 frames. ], tot_loss[loss=0.1756, simple_loss=0.2446, pruned_loss=0.05333, over 954824.72 frames. ], batch size: 54, lr: 3.52e-03, grad_scale: 16.0 2023-04-27 06:20:36,321 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=79989.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 06:20:36,339 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=79989.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:20:53,195 INFO [finetune.py:976] (6/7) Epoch 14, batch 5550, loss[loss=0.1789, simple_loss=0.2545, pruned_loss=0.05166, over 4921.00 frames. ], tot_loss[loss=0.1778, simple_loss=0.2465, pruned_loss=0.05458, over 953562.55 frames. ], batch size: 38, lr: 3.52e-03, grad_scale: 16.0 2023-04-27 06:20:54,498 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=80013.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:20:59,230 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=80020.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:21:04,825 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.4163, 1.6920, 1.8340, 1.9508, 1.7711, 1.8021, 1.8045, 1.8123], device='cuda:6'), covar=tensor([0.4573, 0.6957, 0.5686, 0.4878, 0.6316, 0.8236, 0.6649, 0.6091], device='cuda:6'), in_proj_covar=tensor([0.0334, 0.0378, 0.0322, 0.0334, 0.0344, 0.0401, 0.0358, 0.0330], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 06:21:06,009 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=80031.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 06:21:09,686 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=80037.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:21:16,567 INFO [optim.py:369] (6/7) 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 06:21:24,825 INFO [finetune.py:976] (6/7) Epoch 14, batch 5600, loss[loss=0.2066, simple_loss=0.2776, pruned_loss=0.06779, over 4884.00 frames. ], tot_loss[loss=0.1828, simple_loss=0.2524, pruned_loss=0.0566, over 954946.80 frames. ], batch size: 32, lr: 3.52e-03, grad_scale: 16.0 2023-04-27 06:21:28,975 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=80068.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:21:35,804 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=80079.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 06:21:45,043 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7211, 2.1433, 1.0523, 1.4663, 2.1072, 1.5113, 1.5280, 1.5846], device='cuda:6'), covar=tensor([0.0519, 0.0338, 0.0321, 0.0553, 0.0263, 0.0538, 0.0508, 0.0574], device='cuda:6'), in_proj_covar=tensor([0.0029, 0.0024, 0.0023, 0.0029, 0.0020, 0.0029, 0.0028, 0.0029], device='cuda:6'), out_proj_covar=tensor([0.0050, 0.0044, 0.0038, 0.0050, 0.0037, 0.0049, 0.0048, 0.0050], device='cuda:6') 2023-04-27 06:21:45,070 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.3902, 1.6125, 1.5794, 2.1692, 2.4150, 1.9687, 1.9094, 1.6301], device='cuda:6'), covar=tensor([0.1905, 0.1872, 0.1919, 0.1601, 0.1186, 0.1789, 0.2411, 0.2330], device='cuda:6'), in_proj_covar=tensor([0.0311, 0.0318, 0.0355, 0.0292, 0.0331, 0.0316, 0.0306, 0.0365], device='cuda:6'), out_proj_covar=tensor([6.4659e-05, 6.6679e-05, 7.6181e-05, 5.9648e-05, 6.8981e-05, 6.6888e-05, 6.4641e-05, 7.7914e-05], device='cuda:6') 2023-04-27 06:22:01,084 INFO [finetune.py:976] (6/7) Epoch 14, batch 5650, loss[loss=0.2133, simple_loss=0.2714, pruned_loss=0.07764, over 4857.00 frames. ], tot_loss[loss=0.184, simple_loss=0.2539, pruned_loss=0.05708, over 955272.32 frames. ], batch size: 31, lr: 3.52e-03, grad_scale: 16.0 2023-04-27 06:22:11,110 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=80118.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:22:28,366 INFO [optim.py:369] (6/7) 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] (6/7) Epoch 14, batch 5700, loss[loss=0.216, simple_loss=0.2575, pruned_loss=0.08728, over 4278.00 frames. ], tot_loss[loss=0.1823, simple_loss=0.2509, pruned_loss=0.05681, over 936782.57 frames. ], batch size: 18, lr: 3.52e-03, grad_scale: 16.0 2023-04-27 06:22:55,001 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.6546, 1.7837, 2.0748, 2.1771, 1.9883, 2.1823, 2.1770, 2.1632], device='cuda:6'), covar=tensor([0.3630, 0.5477, 0.4938, 0.4686, 0.6172, 0.7085, 0.5216, 0.4656], device='cuda:6'), in_proj_covar=tensor([0.0334, 0.0378, 0.0321, 0.0334, 0.0344, 0.0403, 0.0357, 0.0329], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 06:22:58,324 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.65 vs. limit=5.0 2023-04-27 06:23:36,604 INFO [finetune.py:976] (6/7) Epoch 15, batch 0, loss[loss=0.1989, simple_loss=0.2684, pruned_loss=0.06469, over 4857.00 frames. ], tot_loss[loss=0.1989, simple_loss=0.2684, pruned_loss=0.06469, over 4857.00 frames. ], batch size: 44, lr: 3.52e-03, grad_scale: 16.0 2023-04-27 06:23:36,604 INFO [finetune.py:1001] (6/7) Computing validation loss 2023-04-27 06:23:43,548 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([3.3824, 3.4364, 2.5513, 3.8496, 3.5168, 3.4159, 1.4157, 3.4482], device='cuda:6'), covar=tensor([0.1487, 0.1339, 0.2629, 0.1949, 0.2452, 0.1677, 0.5121, 0.1997], device='cuda:6'), in_proj_covar=tensor([0.0244, 0.0214, 0.0248, 0.0300, 0.0296, 0.0246, 0.0269, 0.0269], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 06:23:46,656 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7601, 1.4389, 0.6613, 1.3864, 1.5708, 1.5864, 1.4892, 1.4422], device='cuda:6'), covar=tensor([0.0478, 0.0415, 0.0373, 0.0546, 0.0269, 0.0533, 0.0512, 0.0554], device='cuda:6'), in_proj_covar=tensor([0.0029, 0.0024, 0.0023, 0.0029, 0.0020, 0.0029, 0.0028, 0.0029], device='cuda:6'), out_proj_covar=tensor([0.0050, 0.0044, 0.0038, 0.0050, 0.0037, 0.0049, 0.0048, 0.0050], device='cuda:6') 2023-04-27 06:23:47,231 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.2079, 1.7440, 2.0638, 2.3538, 2.0325, 1.6635, 1.1638, 1.7798], device='cuda:6'), covar=tensor([0.3700, 0.3743, 0.1987, 0.2551, 0.3242, 0.3100, 0.4579, 0.2257], device='cuda:6'), in_proj_covar=tensor([0.0290, 0.0249, 0.0225, 0.0318, 0.0216, 0.0231, 0.0231, 0.0187], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:6') 2023-04-27 06:23:53,230 INFO [finetune.py:1010] (6/7) 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] (6/7) Maximum memory allocated so far is 6365MB 2023-04-27 06:23:53,377 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.1814, 1.6829, 2.0118, 2.4221, 2.0269, 1.6229, 1.1653, 1.7610], device='cuda:6'), covar=tensor([0.3235, 0.3511, 0.1716, 0.2448, 0.2805, 0.2718, 0.4749, 0.2222], device='cuda:6'), in_proj_covar=tensor([0.0290, 0.0249, 0.0225, 0.0318, 0.0216, 0.0231, 0.0231, 0.0187], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:6') 2023-04-27 06:24:44,603 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9958, 1.7530, 1.9571, 2.3118, 2.4523, 1.8342, 1.4999, 2.1048], device='cuda:6'), covar=tensor([0.0847, 0.1191, 0.0794, 0.0640, 0.0557, 0.0914, 0.0807, 0.0582], device='cuda:6'), in_proj_covar=tensor([0.0191, 0.0202, 0.0183, 0.0172, 0.0178, 0.0183, 0.0155, 0.0180], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 06:24:56,492 INFO [finetune.py:976] (6/7) Epoch 15, batch 50, loss[loss=0.1458, simple_loss=0.2128, pruned_loss=0.03942, over 4829.00 frames. ], tot_loss[loss=0.1857, simple_loss=0.2544, pruned_loss=0.05847, over 216446.88 frames. ], batch size: 47, lr: 3.52e-03, grad_scale: 16.0 2023-04-27 06:25:05,169 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=80241.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:25:08,720 INFO [optim.py:369] (6/7) 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 06:25:36,894 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=80283.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:25:40,292 INFO [finetune.py:976] (6/7) Epoch 15, batch 100, loss[loss=0.2211, simple_loss=0.275, pruned_loss=0.08357, over 4900.00 frames. ], tot_loss[loss=0.1796, simple_loss=0.2482, pruned_loss=0.05552, over 381410.11 frames. ], batch size: 35, lr: 3.52e-03, grad_scale: 16.0 2023-04-27 06:25:41,452 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=80289.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:25:57,015 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=80313.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:25:57,622 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0982, 2.4447, 0.9482, 1.3021, 1.7660, 1.2628, 2.9994, 1.5875], device='cuda:6'), covar=tensor([0.0655, 0.0616, 0.0748, 0.1336, 0.0540, 0.1067, 0.0312, 0.0713], device='cuda:6'), in_proj_covar=tensor([0.0051, 0.0066, 0.0048, 0.0047, 0.0050, 0.0052, 0.0075, 0.0051], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0007, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:6') 2023-04-27 06:26:12,986 INFO [finetune.py:976] (6/7) Epoch 15, batch 150, loss[loss=0.1708, simple_loss=0.2448, pruned_loss=0.04843, over 4711.00 frames. ], tot_loss[loss=0.1751, simple_loss=0.2425, pruned_loss=0.05389, over 509295.34 frames. ], batch size: 23, lr: 3.51e-03, grad_scale: 16.0 2023-04-27 06:26:18,198 INFO [zipformer.py:1188] (6/7) 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] (6/7) 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 06:26:28,923 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=80361.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:26:46,350 INFO [finetune.py:976] (6/7) Epoch 15, batch 200, loss[loss=0.1716, simple_loss=0.2294, pruned_loss=0.05687, over 4790.00 frames. ], tot_loss[loss=0.1748, simple_loss=0.2421, pruned_loss=0.0538, over 606773.72 frames. ], batch size: 26, lr: 3.51e-03, grad_scale: 16.0 2023-04-27 06:26:49,312 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5294, 1.1368, 4.3828, 4.0660, 3.8762, 4.1902, 4.0748, 3.8891], device='cuda:6'), covar=tensor([0.7564, 0.6766, 0.1096, 0.1957, 0.1252, 0.1521, 0.1548, 0.1654], device='cuda:6'), in_proj_covar=tensor([0.0313, 0.0308, 0.0403, 0.0407, 0.0350, 0.0406, 0.0314, 0.0366], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-27 06:26:50,411 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=80393.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:27:06,638 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=80418.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:27:07,243 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([0.8489, 1.1155, 3.2677, 3.0279, 2.9725, 3.2020, 3.1713, 2.9114], device='cuda:6'), covar=tensor([0.7636, 0.5583, 0.1529, 0.2472, 0.1420, 0.1842, 0.1777, 0.1760], device='cuda:6'), in_proj_covar=tensor([0.0312, 0.0308, 0.0402, 0.0406, 0.0350, 0.0406, 0.0314, 0.0365], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-27 06:27:12,125 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7687, 2.4958, 1.8351, 1.9061, 1.2966, 1.2879, 1.8953, 1.2916], device='cuda:6'), covar=tensor([0.1745, 0.1626, 0.1490, 0.1952, 0.2442, 0.2087, 0.1130, 0.2138], device='cuda:6'), in_proj_covar=tensor([0.0196, 0.0213, 0.0169, 0.0204, 0.0200, 0.0185, 0.0156, 0.0185], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:6') 2023-04-27 06:27:12,189 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-04-27 06:27:18,651 INFO [finetune.py:976] (6/7) Epoch 15, batch 250, loss[loss=0.1173, simple_loss=0.181, pruned_loss=0.02681, over 4809.00 frames. ], tot_loss[loss=0.1787, simple_loss=0.2462, pruned_loss=0.05558, over 685060.67 frames. ], batch size: 25, lr: 3.51e-03, grad_scale: 16.0 2023-04-27 06:27:25,585 INFO [optim.py:369] (6/7) 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 06:27:31,415 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=80454.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:27:36,897 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=80463.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:27:38,599 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=80466.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:27:51,746 INFO [finetune.py:976] (6/7) Epoch 15, batch 300, loss[loss=0.1633, simple_loss=0.234, pruned_loss=0.04629, over 4729.00 frames. ], tot_loss[loss=0.1817, simple_loss=0.2501, pruned_loss=0.05667, over 742521.09 frames. ], batch size: 54, lr: 3.51e-03, grad_scale: 16.0 2023-04-27 06:27:56,471 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=80494.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:28:27,727 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=80524.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:28:36,208 INFO [finetune.py:976] (6/7) Epoch 15, batch 350, loss[loss=0.2059, simple_loss=0.2836, pruned_loss=0.06412, over 4898.00 frames. ], tot_loss[loss=0.1846, simple_loss=0.2529, pruned_loss=0.05811, over 790086.40 frames. ], batch size: 43, lr: 3.51e-03, grad_scale: 16.0 2023-04-27 06:28:38,181 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.2316, 1.7149, 1.7521, 1.9842, 2.3483, 2.0089, 1.7567, 1.6154], device='cuda:6'), covar=tensor([0.1459, 0.1449, 0.1757, 0.1289, 0.1004, 0.1265, 0.1833, 0.1970], device='cuda:6'), in_proj_covar=tensor([0.0307, 0.0315, 0.0350, 0.0289, 0.0329, 0.0311, 0.0302, 0.0360], device='cuda:6'), out_proj_covar=tensor([6.3662e-05, 6.5969e-05, 7.5116e-05, 5.8942e-05, 6.8463e-05, 6.5943e-05, 6.3828e-05, 7.6901e-05], device='cuda:6') 2023-04-27 06:28:42,200 INFO [optim.py:369] (6/7) 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 06:28:44,053 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8927, 2.6024, 1.8815, 2.0340, 1.4725, 1.3993, 2.0850, 1.4116], device='cuda:6'), covar=tensor([0.1802, 0.1632, 0.1570, 0.1800, 0.2352, 0.2087, 0.1058, 0.2162], device='cuda:6'), in_proj_covar=tensor([0.0195, 0.0212, 0.0169, 0.0203, 0.0199, 0.0184, 0.0156, 0.0185], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:6') 2023-04-27 06:28:48,614 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=80555.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:29:15,645 INFO [finetune.py:976] (6/7) Epoch 15, batch 400, loss[loss=0.1971, simple_loss=0.2756, pruned_loss=0.0593, over 4744.00 frames. ], tot_loss[loss=0.1849, simple_loss=0.2541, pruned_loss=0.0579, over 828002.72 frames. ], batch size: 54, lr: 3.51e-03, grad_scale: 16.0 2023-04-27 06:30:07,682 INFO [finetune.py:976] (6/7) Epoch 15, batch 450, loss[loss=0.1767, simple_loss=0.2422, pruned_loss=0.05566, over 4895.00 frames. ], tot_loss[loss=0.1838, simple_loss=0.2529, pruned_loss=0.05733, over 856820.48 frames. ], batch size: 35, lr: 3.51e-03, grad_scale: 16.0 2023-04-27 06:30:13,737 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=80639.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:30:18,557 INFO [optim.py:369] (6/7) 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] (6/7) Epoch 15, batch 500, loss[loss=0.1609, simple_loss=0.2338, pruned_loss=0.04396, over 4784.00 frames. ], tot_loss[loss=0.181, simple_loss=0.2499, pruned_loss=0.05608, over 876763.47 frames. ], batch size: 51, lr: 3.51e-03, grad_scale: 16.0 2023-04-27 06:32:06,655 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=80736.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:32:07,803 INFO [finetune.py:976] (6/7) Epoch 15, batch 550, loss[loss=0.1661, simple_loss=0.2416, pruned_loss=0.04528, over 4840.00 frames. ], tot_loss[loss=0.1785, simple_loss=0.2469, pruned_loss=0.05506, over 894150.72 frames. ], batch size: 49, lr: 3.51e-03, grad_scale: 16.0 2023-04-27 06:32:13,248 INFO [optim.py:369] (6/7) 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 06:32:14,563 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=80749.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:32:27,231 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.3176, 1.4982, 1.3625, 1.6907, 1.6094, 1.7960, 1.4133, 3.4606], device='cuda:6'), covar=tensor([0.0619, 0.0791, 0.0802, 0.1167, 0.0617, 0.0591, 0.0718, 0.0135], device='cuda:6'), in_proj_covar=tensor([0.0038, 0.0039, 0.0040, 0.0044, 0.0040, 0.0038, 0.0038, 0.0057], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:6') 2023-04-27 06:32:32,574 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 15, batch 600, loss[loss=0.1921, simple_loss=0.2696, pruned_loss=0.05734, over 4822.00 frames. ], tot_loss[loss=0.1798, simple_loss=0.2479, pruned_loss=0.0559, over 908439.39 frames. ], batch size: 40, lr: 3.51e-03, grad_scale: 16.0 2023-04-27 06:32:47,153 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=80797.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:33:03,036 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=80819.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:33:06,990 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.49 vs. limit=2.0 2023-04-27 06:33:12,675 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=80834.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:33:13,383 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-04-27 06:33:15,002 INFO [finetune.py:976] (6/7) Epoch 15, batch 650, loss[loss=0.1578, simple_loss=0.2453, pruned_loss=0.03518, over 4797.00 frames. ], tot_loss[loss=0.1822, simple_loss=0.2507, pruned_loss=0.05684, over 918698.89 frames. ], batch size: 29, lr: 3.51e-03, grad_scale: 16.0 2023-04-27 06:33:20,404 INFO [optim.py:369] (6/7) 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 06:33:22,323 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=80850.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:33:31,199 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8818, 2.0502, 1.0632, 1.5985, 1.9977, 1.7237, 1.5851, 1.6746], device='cuda:6'), covar=tensor([0.0479, 0.0343, 0.0348, 0.0528, 0.0270, 0.0522, 0.0541, 0.0552], device='cuda:6'), in_proj_covar=tensor([0.0029, 0.0024, 0.0023, 0.0029, 0.0020, 0.0029, 0.0028, 0.0029], device='cuda:6'), out_proj_covar=tensor([0.0050, 0.0045, 0.0038, 0.0051, 0.0038, 0.0049, 0.0049, 0.0050], device='cuda:6') 2023-04-27 06:33:33,608 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-04-27 06:33:35,799 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0310, 2.6887, 2.0450, 2.3149, 1.8149, 2.2450, 2.3222, 1.7445], device='cuda:6'), covar=tensor([0.2602, 0.1888, 0.1319, 0.1968, 0.3578, 0.1623, 0.2180, 0.2981], device='cuda:6'), in_proj_covar=tensor([0.0291, 0.0313, 0.0225, 0.0282, 0.0314, 0.0264, 0.0255, 0.0271], device='cuda:6'), out_proj_covar=tensor([1.1736e-04, 1.2447e-04, 8.9664e-05, 1.1226e-04, 1.2776e-04, 1.0548e-04, 1.0330e-04, 1.0792e-04], device='cuda:6') 2023-04-27 06:33:48,176 INFO [finetune.py:976] (6/7) Epoch 15, batch 700, loss[loss=0.3065, simple_loss=0.353, pruned_loss=0.13, over 4158.00 frames. ], tot_loss[loss=0.1834, simple_loss=0.2522, pruned_loss=0.05728, over 925705.47 frames. ], batch size: 65, lr: 3.51e-03, grad_scale: 16.0 2023-04-27 06:34:21,750 INFO [finetune.py:976] (6/7) Epoch 15, batch 750, loss[loss=0.1696, simple_loss=0.2393, pruned_loss=0.04998, over 4749.00 frames. ], tot_loss[loss=0.1843, simple_loss=0.2532, pruned_loss=0.05772, over 930754.45 frames. ], batch size: 54, lr: 3.51e-03, grad_scale: 16.0 2023-04-27 06:34:22,441 INFO [zipformer.py:1188] (6/7) 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] (6/7) 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 06:34:35,861 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.4411, 2.2467, 2.5725, 2.8222, 2.9024, 2.2662, 2.0086, 2.4492], device='cuda:6'), covar=tensor([0.0857, 0.0975, 0.0561, 0.0554, 0.0571, 0.0842, 0.0750, 0.0529], device='cuda:6'), in_proj_covar=tensor([0.0192, 0.0203, 0.0184, 0.0173, 0.0179, 0.0184, 0.0155, 0.0181], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 06:34:54,976 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=80987.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:34:55,528 INFO [finetune.py:976] (6/7) Epoch 15, batch 800, loss[loss=0.1632, simple_loss=0.2414, pruned_loss=0.04248, over 4922.00 frames. ], tot_loss[loss=0.1834, simple_loss=0.253, pruned_loss=0.05687, over 937998.93 frames. ], batch size: 38, lr: 3.51e-03, grad_scale: 16.0 2023-04-27 06:35:33,309 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.4538, 1.6210, 1.5721, 1.8949, 1.8097, 1.9937, 1.5306, 3.8817], device='cuda:6'), covar=tensor([0.0562, 0.0764, 0.0764, 0.1095, 0.0607, 0.0475, 0.0700, 0.0115], device='cuda:6'), in_proj_covar=tensor([0.0038, 0.0039, 0.0040, 0.0044, 0.0040, 0.0038, 0.0038, 0.0057], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:6') 2023-04-27 06:35:34,231 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=5.08 vs. limit=5.0 2023-04-27 06:35:35,654 INFO [finetune.py:976] (6/7) Epoch 15, batch 850, loss[loss=0.1656, simple_loss=0.2297, pruned_loss=0.05073, over 4770.00 frames. ], tot_loss[loss=0.1813, simple_loss=0.2506, pruned_loss=0.05602, over 943309.78 frames. ], batch size: 28, lr: 3.51e-03, grad_scale: 16.0 2023-04-27 06:35:43,985 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-04-27 06:35:46,466 INFO [optim.py:369] (6/7) 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 06:35:47,797 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=81049.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:36:30,277 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-04-27 06:36:41,754 INFO [finetune.py:976] (6/7) Epoch 15, batch 900, loss[loss=0.1624, simple_loss=0.2348, pruned_loss=0.04498, over 4835.00 frames. ], tot_loss[loss=0.1787, simple_loss=0.2477, pruned_loss=0.05482, over 944864.47 frames. ], batch size: 30, lr: 3.51e-03, grad_scale: 16.0 2023-04-27 06:36:49,902 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=81092.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:36:52,938 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=81097.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:37:15,247 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=81116.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:37:22,516 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=81119.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:37:35,538 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=81129.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:37:48,251 INFO [finetune.py:976] (6/7) Epoch 15, batch 950, loss[loss=0.1549, simple_loss=0.2289, pruned_loss=0.04045, over 4731.00 frames. ], tot_loss[loss=0.1784, simple_loss=0.247, pruned_loss=0.05494, over 949925.68 frames. ], batch size: 54, lr: 3.51e-03, grad_scale: 16.0 2023-04-27 06:38:04,672 INFO [optim.py:369] (6/7) 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 06:38:06,668 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=81150.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:38:17,045 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=81167.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:38:23,608 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=81177.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:38:32,220 INFO [finetune.py:976] (6/7) Epoch 15, batch 1000, loss[loss=0.1602, simple_loss=0.2522, pruned_loss=0.03409, over 4780.00 frames. ], tot_loss[loss=0.1821, simple_loss=0.2503, pruned_loss=0.0569, over 952570.91 frames. ], batch size: 29, lr: 3.51e-03, grad_scale: 16.0 2023-04-27 06:38:38,872 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=81198.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:38:45,753 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([0.8134, 1.8556, 1.8581, 1.5687, 2.1263, 1.6443, 2.6021, 1.5994], device='cuda:6'), covar=tensor([0.3813, 0.1922, 0.4345, 0.2849, 0.1530, 0.2572, 0.1334, 0.4472], device='cuda:6'), in_proj_covar=tensor([0.0345, 0.0350, 0.0430, 0.0356, 0.0385, 0.0384, 0.0374, 0.0422], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 06:39:05,803 INFO [finetune.py:976] (6/7) Epoch 15, batch 1050, loss[loss=0.1563, simple_loss=0.2405, pruned_loss=0.03603, over 4845.00 frames. ], tot_loss[loss=0.1843, simple_loss=0.2537, pruned_loss=0.0575, over 953687.57 frames. ], batch size: 49, lr: 3.51e-03, grad_scale: 16.0 2023-04-27 06:39:11,771 INFO [optim.py:369] (6/7) 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 06:39:13,062 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0743, 1.7687, 5.3326, 5.0000, 4.6117, 5.1269, 4.7425, 4.7370], device='cuda:6'), covar=tensor([0.7013, 0.5534, 0.0981, 0.1813, 0.1053, 0.1673, 0.0991, 0.1458], device='cuda:6'), in_proj_covar=tensor([0.0310, 0.0307, 0.0400, 0.0405, 0.0349, 0.0404, 0.0314, 0.0361], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-27 06:39:25,964 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5131, 0.6363, 1.4180, 1.9027, 1.5775, 1.4306, 1.4275, 1.4702], device='cuda:6'), covar=tensor([0.4668, 0.6585, 0.6388, 0.6667, 0.5925, 0.7726, 0.7440, 0.7981], device='cuda:6'), in_proj_covar=tensor([0.0413, 0.0406, 0.0492, 0.0507, 0.0442, 0.0465, 0.0471, 0.0474], device='cuda:6'), out_proj_covar=tensor([9.9862e-05, 1.0040e-04, 1.1060e-04, 1.2049e-04, 1.0633e-04, 1.1200e-04, 1.1237e-04, 1.1280e-04], device='cuda:6') 2023-04-27 06:39:38,634 INFO [finetune.py:976] (6/7) Epoch 15, batch 1100, loss[loss=0.2204, simple_loss=0.2966, pruned_loss=0.07212, over 4808.00 frames. ], tot_loss[loss=0.1849, simple_loss=0.255, pruned_loss=0.05735, over 954653.71 frames. ], batch size: 45, lr: 3.51e-03, grad_scale: 16.0 2023-04-27 06:39:46,740 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=81299.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:39:47,950 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=81301.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:40:11,236 INFO [finetune.py:976] (6/7) Epoch 15, batch 1150, loss[loss=0.1961, simple_loss=0.2706, pruned_loss=0.06083, over 4894.00 frames. ], tot_loss[loss=0.1855, simple_loss=0.2558, pruned_loss=0.05756, over 955129.91 frames. ], batch size: 36, lr: 3.51e-03, grad_scale: 16.0 2023-04-27 06:40:18,549 INFO [optim.py:369] (6/7) 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 06:40:25,272 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.9672, 2.2104, 2.1279, 2.2735, 2.1012, 2.1581, 2.1529, 2.1178], device='cuda:6'), covar=tensor([0.4343, 0.6942, 0.5472, 0.5106, 0.6021, 0.8041, 0.6596, 0.6390], device='cuda:6'), in_proj_covar=tensor([0.0327, 0.0373, 0.0316, 0.0330, 0.0340, 0.0398, 0.0353, 0.0325], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 06:40:27,090 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=81360.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:40:28,326 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=81362.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:40:36,027 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.41 vs. limit=5.0 2023-04-27 06:40:44,965 INFO [finetune.py:976] (6/7) Epoch 15, batch 1200, loss[loss=0.1919, simple_loss=0.2683, pruned_loss=0.05774, over 4823.00 frames. ], tot_loss[loss=0.1839, simple_loss=0.2541, pruned_loss=0.05683, over 954939.86 frames. ], batch size: 33, lr: 3.51e-03, grad_scale: 16.0 2023-04-27 06:40:48,489 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=81392.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:41:12,566 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 15, batch 1250, loss[loss=0.1705, simple_loss=0.2399, pruned_loss=0.0505, over 4709.00 frames. ], tot_loss[loss=0.1817, simple_loss=0.251, pruned_loss=0.05617, over 956364.97 frames. ], batch size: 23, lr: 3.51e-03, grad_scale: 16.0 2023-04-27 06:41:20,062 INFO [zipformer.py:1188] (6/7) 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] (6/7) 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 06:41:55,795 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.1493, 1.3787, 1.3309, 1.6779, 1.5137, 1.6266, 1.3492, 2.4359], device='cuda:6'), covar=tensor([0.0604, 0.0829, 0.0795, 0.1192, 0.0654, 0.0437, 0.0730, 0.0231], device='cuda:6'), in_proj_covar=tensor([0.0038, 0.0039, 0.0040, 0.0043, 0.0040, 0.0038, 0.0038, 0.0056], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0014, 0.0013, 0.0012, 0.0013, 0.0016], device='cuda:6') 2023-04-27 06:42:05,042 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=81472.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:42:08,091 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=81477.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:42:20,542 INFO [finetune.py:976] (6/7) Epoch 15, batch 1300, loss[loss=0.181, simple_loss=0.2505, pruned_loss=0.05569, over 4751.00 frames. ], tot_loss[loss=0.1787, simple_loss=0.2474, pruned_loss=0.05499, over 955685.40 frames. ], batch size: 59, lr: 3.50e-03, grad_scale: 32.0 2023-04-27 06:43:12,244 INFO [finetune.py:976] (6/7) Epoch 15, batch 1350, loss[loss=0.1832, simple_loss=0.246, pruned_loss=0.06021, over 4097.00 frames. ], tot_loss[loss=0.1804, simple_loss=0.2485, pruned_loss=0.05615, over 955634.79 frames. ], batch size: 65, lr: 3.50e-03, grad_scale: 32.0 2023-04-27 06:43:24,088 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([3.6023, 3.5122, 2.5665, 4.1375, 3.5016, 3.5798, 1.5091, 3.5127], device='cuda:6'), covar=tensor([0.1549, 0.1224, 0.3345, 0.1751, 0.3298, 0.1737, 0.5616, 0.2360], device='cuda:6'), in_proj_covar=tensor([0.0242, 0.0213, 0.0248, 0.0300, 0.0295, 0.0246, 0.0269, 0.0270], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 06:43:24,613 INFO [optim.py:369] (6/7) 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 06:43:44,117 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.2197, 1.4183, 1.3992, 1.7220, 1.6125, 1.7762, 1.3318, 3.0693], device='cuda:6'), covar=tensor([0.0674, 0.0843, 0.0809, 0.1203, 0.0651, 0.0463, 0.0742, 0.0181], device='cuda:6'), in_proj_covar=tensor([0.0038, 0.0039, 0.0040, 0.0043, 0.0040, 0.0038, 0.0038, 0.0057], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:6') 2023-04-27 06:43:47,126 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=3.93 vs. limit=5.0 2023-04-27 06:44:07,693 INFO [finetune.py:976] (6/7) Epoch 15, batch 1400, loss[loss=0.2098, simple_loss=0.3024, pruned_loss=0.05858, over 4806.00 frames. ], tot_loss[loss=0.1843, simple_loss=0.2532, pruned_loss=0.05773, over 956969.01 frames. ], batch size: 39, lr: 3.50e-03, grad_scale: 32.0 2023-04-27 06:44:18,461 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9598, 1.6139, 1.9387, 2.2846, 2.3394, 1.8935, 1.6229, 2.0198], device='cuda:6'), covar=tensor([0.0816, 0.1177, 0.0688, 0.0524, 0.0588, 0.0803, 0.0790, 0.0564], device='cuda:6'), in_proj_covar=tensor([0.0192, 0.0203, 0.0184, 0.0174, 0.0180, 0.0184, 0.0155, 0.0182], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 06:44:41,267 INFO [finetune.py:976] (6/7) Epoch 15, batch 1450, loss[loss=0.1725, simple_loss=0.2354, pruned_loss=0.05476, over 4799.00 frames. ], tot_loss[loss=0.1846, simple_loss=0.2541, pruned_loss=0.05753, over 957911.88 frames. ], batch size: 25, lr: 3.50e-03, grad_scale: 32.0 2023-04-27 06:44:47,199 INFO [optim.py:369] (6/7) 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 06:44:53,609 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=81655.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:44:55,286 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=81657.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:45:04,224 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=81670.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:45:15,083 INFO [finetune.py:976] (6/7) Epoch 15, batch 1500, loss[loss=0.2036, simple_loss=0.2663, pruned_loss=0.07044, over 4044.00 frames. ], tot_loss[loss=0.1855, simple_loss=0.2556, pruned_loss=0.05772, over 958062.95 frames. ], batch size: 65, lr: 3.50e-03, grad_scale: 32.0 2023-04-27 06:45:44,092 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=81731.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 06:45:48,175 INFO [finetune.py:976] (6/7) Epoch 15, batch 1550, loss[loss=0.1721, simple_loss=0.2457, pruned_loss=0.04928, over 4843.00 frames. ], tot_loss[loss=0.1855, simple_loss=0.2554, pruned_loss=0.05784, over 957820.36 frames. ], batch size: 44, lr: 3.50e-03, grad_scale: 32.0 2023-04-27 06:45:53,673 INFO [optim.py:369] (6/7) 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 06:46:11,800 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=81772.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:46:21,441 INFO [finetune.py:976] (6/7) Epoch 15, batch 1600, loss[loss=0.1751, simple_loss=0.2352, pruned_loss=0.0575, over 4915.00 frames. ], tot_loss[loss=0.1835, simple_loss=0.2525, pruned_loss=0.05722, over 954895.43 frames. ], batch size: 43, lr: 3.50e-03, grad_scale: 32.0 2023-04-27 06:46:34,752 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6775, 1.3539, 0.6890, 1.3164, 1.4444, 1.5278, 1.3845, 1.4070], device='cuda:6'), covar=tensor([0.0472, 0.0407, 0.0377, 0.0566, 0.0295, 0.0520, 0.0530, 0.0573], device='cuda:6'), in_proj_covar=tensor([0.0029, 0.0024, 0.0023, 0.0029, 0.0020, 0.0029, 0.0028, 0.0029], device='cuda:6'), out_proj_covar=tensor([0.0050, 0.0045, 0.0038, 0.0051, 0.0038, 0.0049, 0.0049, 0.0050], device='cuda:6') 2023-04-27 06:46:42,457 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 15, batch 1650, loss[loss=0.1382, simple_loss=0.2098, pruned_loss=0.03329, over 4821.00 frames. ], tot_loss[loss=0.1803, simple_loss=0.249, pruned_loss=0.05577, over 956133.50 frames. ], batch size: 40, lr: 3.50e-03, grad_scale: 32.0 2023-04-27 06:47:04,625 INFO [optim.py:369] (6/7) 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 06:47:27,299 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.0407, 0.9964, 1.2379, 1.1804, 1.0258, 0.8930, 0.9168, 0.5257], device='cuda:6'), covar=tensor([0.0595, 0.0780, 0.0561, 0.0548, 0.0766, 0.1384, 0.0559, 0.0809], device='cuda:6'), in_proj_covar=tensor([0.0067, 0.0070, 0.0069, 0.0067, 0.0075, 0.0096, 0.0074, 0.0069], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-27 06:47:59,312 INFO [finetune.py:976] (6/7) Epoch 15, batch 1700, loss[loss=0.2299, simple_loss=0.285, pruned_loss=0.08745, over 4128.00 frames. ], tot_loss[loss=0.1784, simple_loss=0.2471, pruned_loss=0.05487, over 953965.73 frames. ], batch size: 65, lr: 3.50e-03, grad_scale: 32.0 2023-04-27 06:49:06,083 INFO [finetune.py:976] (6/7) Epoch 15, batch 1750, loss[loss=0.1895, simple_loss=0.2562, pruned_loss=0.06136, over 4869.00 frames. ], tot_loss[loss=0.1801, simple_loss=0.2487, pruned_loss=0.05573, over 952988.74 frames. ], batch size: 31, lr: 3.50e-03, grad_scale: 32.0 2023-04-27 06:49:16,877 INFO [optim.py:369] (6/7) 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 06:49:25,071 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.6169, 2.9956, 0.8827, 1.6837, 2.4811, 1.6061, 4.3530, 2.0989], device='cuda:6'), covar=tensor([0.0609, 0.0717, 0.0901, 0.1351, 0.0524, 0.1016, 0.0199, 0.0664], device='cuda:6'), in_proj_covar=tensor([0.0052, 0.0067, 0.0049, 0.0047, 0.0050, 0.0053, 0.0077, 0.0052], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:6') 2023-04-27 06:49:26,288 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=81955.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:49:27,520 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=81957.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:49:33,447 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7601, 1.2307, 4.9849, 4.6700, 4.3691, 4.8062, 4.4380, 4.4190], device='cuda:6'), covar=tensor([0.7622, 0.6677, 0.1072, 0.1959, 0.1116, 0.1313, 0.1526, 0.1631], device='cuda:6'), in_proj_covar=tensor([0.0314, 0.0309, 0.0404, 0.0407, 0.0351, 0.0407, 0.0316, 0.0365], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-27 06:49:48,416 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=81987.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:49:48,923 INFO [finetune.py:976] (6/7) Epoch 15, batch 1800, loss[loss=0.1833, simple_loss=0.2695, pruned_loss=0.04854, over 4819.00 frames. ], tot_loss[loss=0.1817, simple_loss=0.2512, pruned_loss=0.05613, over 953108.40 frames. ], batch size: 40, lr: 3.50e-03, grad_scale: 32.0 2023-04-27 06:49:59,487 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=82003.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:50:00,738 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=82005.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:50:01,392 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.2185, 2.5770, 0.9839, 1.4851, 2.1833, 1.3321, 3.6359, 1.8229], device='cuda:6'), covar=tensor([0.0641, 0.0715, 0.0824, 0.1303, 0.0490, 0.1002, 0.0237, 0.0624], device='cuda:6'), in_proj_covar=tensor([0.0051, 0.0066, 0.0049, 0.0047, 0.0050, 0.0052, 0.0076, 0.0052], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:6') 2023-04-27 06:50:16,420 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=82026.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 06:50:24,046 INFO [finetune.py:976] (6/7) Epoch 15, batch 1850, loss[loss=0.1932, simple_loss=0.2552, pruned_loss=0.06561, over 4915.00 frames. ], tot_loss[loss=0.1822, simple_loss=0.2522, pruned_loss=0.0561, over 953128.22 frames. ], batch size: 38, lr: 3.50e-03, grad_scale: 32.0 2023-04-27 06:50:29,488 INFO [optim.py:369] (6/7) 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 06:50:30,246 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=82048.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:50:34,524 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=82055.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:50:56,620 INFO [finetune.py:976] (6/7) Epoch 15, batch 1900, loss[loss=0.188, simple_loss=0.2574, pruned_loss=0.05934, over 4882.00 frames. ], tot_loss[loss=0.1824, simple_loss=0.2528, pruned_loss=0.05599, over 953200.06 frames. ], batch size: 32, lr: 3.50e-03, grad_scale: 32.0 2023-04-27 06:51:14,327 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=82116.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:51:30,521 INFO [finetune.py:976] (6/7) Epoch 15, batch 1950, loss[loss=0.1518, simple_loss=0.2307, pruned_loss=0.03641, over 4783.00 frames. ], tot_loss[loss=0.1825, simple_loss=0.2523, pruned_loss=0.05633, over 954128.57 frames. ], batch size: 26, lr: 3.50e-03, grad_scale: 32.0 2023-04-27 06:51:36,461 INFO [optim.py:369] (6/7) 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 06:52:04,204 INFO [finetune.py:976] (6/7) Epoch 15, batch 2000, loss[loss=0.1815, simple_loss=0.2438, pruned_loss=0.05962, over 4761.00 frames. ], tot_loss[loss=0.1815, simple_loss=0.2501, pruned_loss=0.05642, over 952591.29 frames. ], batch size: 54, lr: 3.50e-03, grad_scale: 32.0 2023-04-27 06:52:12,922 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([3.6339, 3.5335, 2.8246, 4.1859, 3.6187, 3.6551, 1.6026, 3.5945], device='cuda:6'), covar=tensor([0.1686, 0.1237, 0.2771, 0.1784, 0.2649, 0.1781, 0.5778, 0.2448], device='cuda:6'), in_proj_covar=tensor([0.0243, 0.0214, 0.0249, 0.0300, 0.0295, 0.0246, 0.0270, 0.0271], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 06:52:15,371 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.1258, 1.1943, 3.8740, 3.6398, 3.4500, 3.6940, 3.6840, 3.4421], device='cuda:6'), covar=tensor([0.7241, 0.5700, 0.1144, 0.1725, 0.1138, 0.1483, 0.1455, 0.1490], device='cuda:6'), in_proj_covar=tensor([0.0310, 0.0307, 0.0399, 0.0403, 0.0347, 0.0404, 0.0312, 0.0361], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-27 06:52:37,841 INFO [finetune.py:976] (6/7) Epoch 15, batch 2050, loss[loss=0.1477, simple_loss=0.2138, pruned_loss=0.04085, over 4850.00 frames. ], tot_loss[loss=0.1787, simple_loss=0.2469, pruned_loss=0.05522, over 956114.85 frames. ], batch size: 44, lr: 3.50e-03, grad_scale: 32.0 2023-04-27 06:52:43,309 INFO [optim.py:369] (6/7) 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 06:53:07,005 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.3690, 1.6709, 1.7519, 1.9021, 1.6987, 1.8702, 1.8678, 1.8074], device='cuda:6'), covar=tensor([0.4399, 0.6403, 0.5497, 0.5057, 0.6264, 0.8002, 0.6135, 0.5909], device='cuda:6'), in_proj_covar=tensor([0.0327, 0.0372, 0.0316, 0.0333, 0.0341, 0.0398, 0.0352, 0.0325], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 06:53:21,147 INFO [finetune.py:976] (6/7) Epoch 15, batch 2100, loss[loss=0.2058, simple_loss=0.2728, pruned_loss=0.06938, over 4854.00 frames. ], tot_loss[loss=0.1791, simple_loss=0.2474, pruned_loss=0.05538, over 955838.74 frames. ], batch size: 49, lr: 3.50e-03, grad_scale: 32.0 2023-04-27 06:53:28,950 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6133, 1.5443, 1.9558, 1.8858, 1.4310, 1.3033, 1.4678, 0.9216], device='cuda:6'), covar=tensor([0.0492, 0.0828, 0.0449, 0.0740, 0.0877, 0.1207, 0.0759, 0.0761], device='cuda:6'), in_proj_covar=tensor([0.0067, 0.0069, 0.0069, 0.0067, 0.0075, 0.0096, 0.0074, 0.0068], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-27 06:54:04,956 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=82326.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:54:24,279 INFO [finetune.py:976] (6/7) Epoch 15, batch 2150, loss[loss=0.1885, simple_loss=0.2694, pruned_loss=0.05386, over 4908.00 frames. ], tot_loss[loss=0.1828, simple_loss=0.2517, pruned_loss=0.0569, over 955036.99 frames. ], batch size: 43, lr: 3.50e-03, grad_scale: 32.0 2023-04-27 06:54:33,147 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=82343.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:54:35,063 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=82346.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:54:35,534 INFO [optim.py:369] (6/7) 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 06:54:57,182 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.3156, 2.1401, 1.7749, 1.7705, 2.1371, 1.7922, 2.5301, 1.5748], device='cuda:6'), covar=tensor([0.3190, 0.1494, 0.3866, 0.2705, 0.1633, 0.2201, 0.1480, 0.3994], device='cuda:6'), in_proj_covar=tensor([0.0344, 0.0348, 0.0429, 0.0356, 0.0383, 0.0385, 0.0376, 0.0420], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 06:54:58,342 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=82374.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:55:13,106 INFO [finetune.py:976] (6/7) Epoch 15, batch 2200, loss[loss=0.1712, simple_loss=0.2445, pruned_loss=0.04894, over 4872.00 frames. ], tot_loss[loss=0.1816, simple_loss=0.2515, pruned_loss=0.05587, over 955173.97 frames. ], batch size: 31, lr: 3.50e-03, grad_scale: 32.0 2023-04-27 06:55:42,647 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=82407.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:55:45,045 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 15, batch 2250, loss[loss=0.2168, simple_loss=0.284, pruned_loss=0.07477, over 4820.00 frames. ], tot_loss[loss=0.1832, simple_loss=0.2532, pruned_loss=0.05659, over 957434.95 frames. ], batch size: 47, lr: 3.50e-03, grad_scale: 32.0 2023-04-27 06:56:19,194 INFO [optim.py:369] (6/7) 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 06:56:44,270 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 2023-04-27 06:56:45,466 INFO [finetune.py:976] (6/7) Epoch 15, batch 2300, loss[loss=0.2094, simple_loss=0.2729, pruned_loss=0.07295, over 4838.00 frames. ], tot_loss[loss=0.1837, simple_loss=0.2536, pruned_loss=0.05688, over 957958.99 frames. ], batch size: 49, lr: 3.50e-03, grad_scale: 32.0 2023-04-27 06:56:49,015 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=82493.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 06:57:18,144 INFO [finetune.py:976] (6/7) Epoch 15, batch 2350, loss[loss=0.1763, simple_loss=0.2386, pruned_loss=0.05701, over 4922.00 frames. ], tot_loss[loss=0.1818, simple_loss=0.2513, pruned_loss=0.05618, over 958494.20 frames. ], batch size: 38, lr: 3.50e-03, grad_scale: 32.0 2023-04-27 06:57:20,781 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.73 vs. limit=2.0 2023-04-27 06:57:25,042 INFO [optim.py:369] (6/7) 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 06:57:30,359 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=82554.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 06:57:40,469 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=82569.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:57:51,902 INFO [finetune.py:976] (6/7) Epoch 15, batch 2400, loss[loss=0.1762, simple_loss=0.2328, pruned_loss=0.05982, over 4938.00 frames. ], tot_loss[loss=0.1797, simple_loss=0.2486, pruned_loss=0.05542, over 958360.83 frames. ], batch size: 33, lr: 3.50e-03, grad_scale: 32.0 2023-04-27 06:57:56,907 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-04-27 06:58:21,020 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=82630.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:58:25,773 INFO [finetune.py:976] (6/7) Epoch 15, batch 2450, loss[loss=0.1882, simple_loss=0.2386, pruned_loss=0.06891, over 4831.00 frames. ], tot_loss[loss=0.1764, simple_loss=0.245, pruned_loss=0.05391, over 957674.91 frames. ], batch size: 33, lr: 3.49e-03, grad_scale: 32.0 2023-04-27 06:58:28,862 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=82643.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:58:31,162 INFO [optim.py:369] (6/7) 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 06:59:04,416 INFO [finetune.py:976] (6/7) Epoch 15, batch 2500, loss[loss=0.168, simple_loss=0.248, pruned_loss=0.04402, over 4866.00 frames. ], tot_loss[loss=0.1773, simple_loss=0.2459, pruned_loss=0.05437, over 955897.06 frames. ], batch size: 44, lr: 3.49e-03, grad_scale: 32.0 2023-04-27 06:59:11,877 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=82691.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:59:25,496 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=82702.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 06:59:42,700 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=82711.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:00:09,104 INFO [finetune.py:976] (6/7) Epoch 15, batch 2550, loss[loss=0.213, simple_loss=0.2923, pruned_loss=0.06683, over 4933.00 frames. ], tot_loss[loss=0.1799, simple_loss=0.2499, pruned_loss=0.055, over 956869.19 frames. ], batch size: 42, lr: 3.49e-03, grad_scale: 32.0 2023-04-27 07:00:14,538 INFO [optim.py:369] (6/7) 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 07:00:24,403 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=82759.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:00:47,935 INFO [finetune.py:976] (6/7) Epoch 15, batch 2600, loss[loss=0.2041, simple_loss=0.2781, pruned_loss=0.06503, over 4889.00 frames. ], tot_loss[loss=0.1819, simple_loss=0.2519, pruned_loss=0.05589, over 957040.70 frames. ], batch size: 35, lr: 3.49e-03, grad_scale: 32.0 2023-04-27 07:00:57,698 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([0.8434, 1.7961, 1.5655, 1.4300, 1.7064, 1.5013, 2.1363, 1.1932], device='cuda:6'), covar=tensor([0.3156, 0.1311, 0.4034, 0.2232, 0.1468, 0.1949, 0.1370, 0.4727], device='cuda:6'), in_proj_covar=tensor([0.0348, 0.0351, 0.0433, 0.0359, 0.0387, 0.0386, 0.0376, 0.0424], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 07:01:43,239 INFO [finetune.py:976] (6/7) Epoch 15, batch 2650, loss[loss=0.1682, simple_loss=0.2339, pruned_loss=0.05128, over 4851.00 frames. ], tot_loss[loss=0.1848, simple_loss=0.2551, pruned_loss=0.05726, over 956711.58 frames. ], batch size: 47, lr: 3.49e-03, grad_scale: 32.0 2023-04-27 07:01:48,667 INFO [optim.py:369] (6/7) 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 07:01:49,965 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=82849.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 07:02:33,354 INFO [finetune.py:976] (6/7) Epoch 15, batch 2700, loss[loss=0.1842, simple_loss=0.2579, pruned_loss=0.05523, over 4816.00 frames. ], tot_loss[loss=0.1843, simple_loss=0.254, pruned_loss=0.05723, over 955488.42 frames. ], batch size: 40, lr: 3.49e-03, grad_scale: 32.0 2023-04-27 07:03:04,632 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=82925.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:03:12,572 INFO [finetune.py:976] (6/7) Epoch 15, batch 2750, loss[loss=0.1547, simple_loss=0.2032, pruned_loss=0.05314, over 4058.00 frames. ], tot_loss[loss=0.181, simple_loss=0.2502, pruned_loss=0.05587, over 955699.29 frames. ], batch size: 17, lr: 3.49e-03, grad_scale: 32.0 2023-04-27 07:03:18,528 INFO [optim.py:369] (6/7) 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 07:03:28,536 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=82963.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:03:33,218 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.4211, 2.9447, 1.9944, 2.1906, 2.7188, 2.3120, 2.1833, 2.3653], device='cuda:6'), covar=tensor([0.0440, 0.0296, 0.0261, 0.0516, 0.0217, 0.0469, 0.0499, 0.0518], device='cuda:6'), in_proj_covar=tensor([0.0029, 0.0024, 0.0023, 0.0029, 0.0020, 0.0029, 0.0028, 0.0029], device='cuda:6'), out_proj_covar=tensor([0.0050, 0.0045, 0.0038, 0.0051, 0.0038, 0.0049, 0.0049, 0.0050], device='cuda:6') 2023-04-27 07:03:43,729 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6900, 3.6259, 0.8073, 1.9549, 2.1939, 2.6373, 1.9631, 0.9796], device='cuda:6'), covar=tensor([0.1410, 0.0851, 0.2232, 0.1309, 0.1021, 0.1041, 0.1724, 0.2125], device='cuda:6'), in_proj_covar=tensor([0.0118, 0.0246, 0.0137, 0.0121, 0.0131, 0.0153, 0.0118, 0.0120], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-27 07:03:46,114 INFO [finetune.py:976] (6/7) Epoch 15, batch 2800, loss[loss=0.1524, simple_loss=0.227, pruned_loss=0.03888, over 4910.00 frames. ], tot_loss[loss=0.1775, simple_loss=0.2465, pruned_loss=0.05427, over 955181.78 frames. ], batch size: 32, lr: 3.49e-03, grad_scale: 32.0 2023-04-27 07:03:55,378 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=83002.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:04:09,803 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=83024.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:04:19,800 INFO [finetune.py:976] (6/7) Epoch 15, batch 2850, loss[loss=0.1902, simple_loss=0.2735, pruned_loss=0.05348, over 4913.00 frames. ], tot_loss[loss=0.1763, simple_loss=0.2452, pruned_loss=0.05366, over 955883.02 frames. ], batch size: 43, lr: 3.49e-03, grad_scale: 32.0 2023-04-27 07:04:25,725 INFO [optim.py:369] (6/7) 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 07:04:27,628 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=83050.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:04:28,925 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9250, 1.7592, 2.1788, 2.2659, 1.7129, 1.5854, 1.7908, 1.0211], device='cuda:6'), covar=tensor([0.0728, 0.0955, 0.0535, 0.0793, 0.0952, 0.1216, 0.0929, 0.0976], device='cuda:6'), in_proj_covar=tensor([0.0067, 0.0070, 0.0069, 0.0067, 0.0075, 0.0096, 0.0075, 0.0069], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-27 07:04:33,157 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([4.5974, 4.6043, 3.2184, 5.2860, 4.7222, 4.5726, 1.9330, 4.5765], device='cuda:6'), covar=tensor([0.1527, 0.0971, 0.3217, 0.0989, 0.4206, 0.1688, 0.6042, 0.2183], device='cuda:6'), in_proj_covar=tensor([0.0244, 0.0214, 0.0250, 0.0301, 0.0298, 0.0246, 0.0270, 0.0271], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 07:05:03,760 INFO [finetune.py:976] (6/7) Epoch 15, batch 2900, loss[loss=0.2074, simple_loss=0.2801, pruned_loss=0.06737, over 4738.00 frames. ], tot_loss[loss=0.1794, simple_loss=0.2485, pruned_loss=0.05514, over 954583.81 frames. ], batch size: 54, lr: 3.49e-03, grad_scale: 32.0 2023-04-27 07:05:45,875 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8802, 1.8112, 2.2839, 2.3603, 1.6902, 1.5387, 1.8166, 1.1214], device='cuda:6'), covar=tensor([0.0652, 0.0750, 0.0471, 0.0813, 0.0902, 0.1338, 0.0738, 0.0898], device='cuda:6'), in_proj_covar=tensor([0.0067, 0.0070, 0.0069, 0.0067, 0.0075, 0.0096, 0.0075, 0.0069], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-27 07:06:07,041 INFO [finetune.py:976] (6/7) Epoch 15, batch 2950, loss[loss=0.1483, simple_loss=0.2263, pruned_loss=0.0352, over 4776.00 frames. ], tot_loss[loss=0.1813, simple_loss=0.2513, pruned_loss=0.05561, over 955419.04 frames. ], batch size: 26, lr: 3.49e-03, grad_scale: 32.0 2023-04-27 07:06:18,121 INFO [optim.py:369] (6/7) 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 07:06:19,456 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=83149.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 07:06:50,580 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7622, 2.4407, 1.8159, 1.7850, 1.2868, 1.3372, 1.8783, 1.2214], device='cuda:6'), covar=tensor([0.1674, 0.1424, 0.1494, 0.1759, 0.2473, 0.2127, 0.1014, 0.2186], device='cuda:6'), in_proj_covar=tensor([0.0196, 0.0214, 0.0169, 0.0205, 0.0201, 0.0184, 0.0156, 0.0185], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:6') 2023-04-27 07:06:56,295 INFO [finetune.py:976] (6/7) Epoch 15, batch 3000, loss[loss=0.1934, simple_loss=0.2626, pruned_loss=0.06211, over 4748.00 frames. ], tot_loss[loss=0.1834, simple_loss=0.2533, pruned_loss=0.05669, over 956490.84 frames. ], batch size: 27, lr: 3.49e-03, grad_scale: 32.0 2023-04-27 07:06:56,295 INFO [finetune.py:1001] (6/7) Computing validation loss 2023-04-27 07:07:06,875 INFO [finetune.py:1010] (6/7) 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] (6/7) Maximum memory allocated so far is 6365MB 2023-04-27 07:07:12,341 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=83196.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:07:12,902 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=83197.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 07:07:22,751 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9428, 4.5130, 1.0648, 2.4010, 2.6490, 2.9901, 2.5883, 1.0488], device='cuda:6'), covar=tensor([0.1417, 0.0749, 0.1968, 0.1167, 0.0956, 0.1075, 0.1440, 0.2094], device='cuda:6'), in_proj_covar=tensor([0.0118, 0.0246, 0.0137, 0.0121, 0.0131, 0.0152, 0.0118, 0.0119], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-27 07:07:30,742 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=83225.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:07:36,617 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=5.02 vs. limit=5.0 2023-04-27 07:07:38,611 INFO [finetune.py:976] (6/7) Epoch 15, batch 3050, loss[loss=0.1501, simple_loss=0.2281, pruned_loss=0.03605, over 4915.00 frames. ], tot_loss[loss=0.1825, simple_loss=0.2532, pruned_loss=0.05593, over 956597.06 frames. ], batch size: 38, lr: 3.49e-03, grad_scale: 32.0 2023-04-27 07:07:50,803 INFO [optim.py:369] (6/7) 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 07:08:02,499 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=83257.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:08:04,633 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-04-27 07:08:24,058 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=83273.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:08:36,041 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=83284.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:08:37,219 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.3737, 1.5885, 1.5359, 1.8270, 1.7751, 2.0000, 1.4809, 3.6421], device='cuda:6'), covar=tensor([0.0605, 0.0756, 0.0830, 0.1191, 0.0605, 0.0467, 0.0744, 0.0168], device='cuda:6'), in_proj_covar=tensor([0.0038, 0.0039, 0.0040, 0.0044, 0.0040, 0.0038, 0.0038, 0.0057], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:6') 2023-04-27 07:08:42,489 INFO [finetune.py:976] (6/7) Epoch 15, batch 3100, loss[loss=0.2056, simple_loss=0.268, pruned_loss=0.07162, over 4845.00 frames. ], tot_loss[loss=0.1816, simple_loss=0.2521, pruned_loss=0.05553, over 958665.61 frames. ], batch size: 49, lr: 3.49e-03, grad_scale: 32.0 2023-04-27 07:08:42,815 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-04-27 07:09:12,185 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5825, 1.8946, 0.7722, 1.3047, 1.8674, 1.4378, 1.3809, 1.4248], device='cuda:6'), covar=tensor([0.0664, 0.0350, 0.0383, 0.0639, 0.0295, 0.0668, 0.0654, 0.0696], device='cuda:6'), in_proj_covar=tensor([0.0029, 0.0024, 0.0023, 0.0029, 0.0020, 0.0029, 0.0028, 0.0029], device='cuda:6'), out_proj_covar=tensor([0.0050, 0.0045, 0.0038, 0.0050, 0.0038, 0.0049, 0.0049, 0.0050], device='cuda:6') 2023-04-27 07:09:14,461 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([0.8567, 1.2751, 3.2868, 3.0632, 2.9714, 3.1862, 3.1804, 2.8925], device='cuda:6'), covar=tensor([0.7377, 0.5221, 0.1465, 0.2167, 0.1474, 0.1684, 0.1555, 0.1703], device='cuda:6'), in_proj_covar=tensor([0.0312, 0.0308, 0.0402, 0.0407, 0.0349, 0.0405, 0.0314, 0.0361], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-27 07:09:15,039 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=83319.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:09:26,591 INFO [finetune.py:976] (6/7) Epoch 15, batch 3150, loss[loss=0.1995, simple_loss=0.263, pruned_loss=0.06798, over 4867.00 frames. ], tot_loss[loss=0.1793, simple_loss=0.2493, pruned_loss=0.05463, over 959362.36 frames. ], batch size: 34, lr: 3.49e-03, grad_scale: 32.0 2023-04-27 07:09:31,532 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=83345.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:09:32,641 INFO [optim.py:369] (6/7) 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 07:09:59,877 INFO [finetune.py:976] (6/7) Epoch 15, batch 3200, loss[loss=0.2014, simple_loss=0.2616, pruned_loss=0.07066, over 4771.00 frames. ], tot_loss[loss=0.1779, simple_loss=0.2471, pruned_loss=0.05436, over 960149.86 frames. ], batch size: 28, lr: 3.49e-03, grad_scale: 32.0 2023-04-27 07:10:45,248 INFO [finetune.py:976] (6/7) Epoch 15, batch 3250, loss[loss=0.1876, simple_loss=0.2582, pruned_loss=0.0585, over 4824.00 frames. ], tot_loss[loss=0.1778, simple_loss=0.2473, pruned_loss=0.05418, over 961001.65 frames. ], batch size: 30, lr: 3.49e-03, grad_scale: 32.0 2023-04-27 07:10:56,129 INFO [optim.py:369] (6/7) 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 07:11:38,462 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5370, 3.3050, 1.0908, 1.8360, 1.9208, 2.2484, 1.8341, 0.9647], device='cuda:6'), covar=tensor([0.1407, 0.1179, 0.1930, 0.1302, 0.1089, 0.1109, 0.1637, 0.2029], device='cuda:6'), in_proj_covar=tensor([0.0117, 0.0245, 0.0137, 0.0121, 0.0131, 0.0152, 0.0118, 0.0119], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-27 07:11:50,145 INFO [finetune.py:976] (6/7) Epoch 15, batch 3300, loss[loss=0.1934, simple_loss=0.2725, pruned_loss=0.05718, over 4910.00 frames. ], tot_loss[loss=0.1817, simple_loss=0.2517, pruned_loss=0.05585, over 958015.34 frames. ], batch size: 36, lr: 3.49e-03, grad_scale: 64.0 2023-04-27 07:12:00,630 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.1860, 1.6281, 1.5251, 1.9646, 1.8603, 1.8224, 1.5918, 4.3107], device='cuda:6'), covar=tensor([0.0575, 0.0773, 0.0848, 0.1199, 0.0640, 0.0566, 0.0713, 0.0098], device='cuda:6'), in_proj_covar=tensor([0.0038, 0.0039, 0.0040, 0.0044, 0.0040, 0.0039, 0.0039, 0.0057], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:6') 2023-04-27 07:12:14,151 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5807, 1.3377, 4.0673, 3.8462, 3.5920, 3.7957, 3.7101, 3.6175], device='cuda:6'), covar=tensor([0.6956, 0.5473, 0.1017, 0.1448, 0.1032, 0.1583, 0.2619, 0.1541], device='cuda:6'), in_proj_covar=tensor([0.0309, 0.0304, 0.0398, 0.0403, 0.0345, 0.0402, 0.0312, 0.0358], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-27 07:12:23,031 INFO [finetune.py:976] (6/7) Epoch 15, batch 3350, loss[loss=0.1916, simple_loss=0.2672, pruned_loss=0.05804, over 4777.00 frames. ], tot_loss[loss=0.1826, simple_loss=0.2532, pruned_loss=0.05603, over 957881.12 frames. ], batch size: 26, lr: 3.49e-03, grad_scale: 64.0 2023-04-27 07:12:28,446 INFO [optim.py:369] (6/7) 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 07:12:31,581 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=83552.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:12:55,063 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7019, 1.2912, 1.7871, 2.2281, 1.7900, 1.6804, 1.7564, 1.7041], device='cuda:6'), covar=tensor([0.5215, 0.7461, 0.7107, 0.6292, 0.6656, 0.8760, 0.8947, 0.9522], device='cuda:6'), in_proj_covar=tensor([0.0414, 0.0405, 0.0492, 0.0506, 0.0443, 0.0465, 0.0472, 0.0476], device='cuda:6'), out_proj_covar=tensor([9.9905e-05, 1.0014e-04, 1.1073e-04, 1.2037e-04, 1.0663e-04, 1.1197e-04, 1.1226e-04, 1.1314e-04], device='cuda:6') 2023-04-27 07:12:56,721 INFO [finetune.py:976] (6/7) Epoch 15, batch 3400, loss[loss=0.1895, simple_loss=0.2586, pruned_loss=0.06017, over 4792.00 frames. ], tot_loss[loss=0.1835, simple_loss=0.2541, pruned_loss=0.05642, over 956289.60 frames. ], batch size: 29, lr: 3.49e-03, grad_scale: 64.0 2023-04-27 07:13:18,187 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=83619.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:13:21,273 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8042, 1.2762, 1.8143, 2.3470, 1.8834, 1.7221, 1.8042, 1.7614], device='cuda:6'), covar=tensor([0.4976, 0.7084, 0.6863, 0.6225, 0.6218, 0.8084, 0.8657, 0.9519], device='cuda:6'), in_proj_covar=tensor([0.0416, 0.0407, 0.0494, 0.0507, 0.0446, 0.0467, 0.0474, 0.0478], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 07:13:30,155 INFO [finetune.py:976] (6/7) Epoch 15, batch 3450, loss[loss=0.1945, simple_loss=0.2637, pruned_loss=0.06263, over 4813.00 frames. ], tot_loss[loss=0.1837, simple_loss=0.2545, pruned_loss=0.05645, over 957045.17 frames. ], batch size: 40, lr: 3.49e-03, grad_scale: 64.0 2023-04-27 07:13:31,421 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=83640.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:13:35,583 INFO [optim.py:369] (6/7) 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 07:13:49,724 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=83667.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:14:13,519 INFO [finetune.py:976] (6/7) Epoch 15, batch 3500, loss[loss=0.1388, simple_loss=0.2161, pruned_loss=0.03075, over 4839.00 frames. ], tot_loss[loss=0.1813, simple_loss=0.2515, pruned_loss=0.05554, over 956420.63 frames. ], batch size: 47, lr: 3.49e-03, grad_scale: 64.0 2023-04-27 07:15:08,597 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.69 vs. limit=2.0 2023-04-27 07:15:15,484 INFO [finetune.py:976] (6/7) Epoch 15, batch 3550, loss[loss=0.1667, simple_loss=0.2357, pruned_loss=0.0488, over 4829.00 frames. ], tot_loss[loss=0.179, simple_loss=0.2485, pruned_loss=0.05481, over 956168.61 frames. ], batch size: 30, lr: 3.49e-03, grad_scale: 64.0 2023-04-27 07:15:17,725 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=3.82 vs. limit=5.0 2023-04-27 07:15:21,402 INFO [optim.py:369] (6/7) 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 07:15:22,253 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-04-27 07:15:29,365 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=83760.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:15:29,375 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.5207, 1.7493, 1.7290, 2.3064, 2.5279, 2.0036, 1.9992, 1.7457], device='cuda:6'), covar=tensor([0.1603, 0.1785, 0.1757, 0.1525, 0.1150, 0.1986, 0.2448, 0.2128], device='cuda:6'), in_proj_covar=tensor([0.0312, 0.0319, 0.0356, 0.0296, 0.0333, 0.0317, 0.0308, 0.0367], device='cuda:6'), out_proj_covar=tensor([6.4644e-05, 6.6848e-05, 7.6078e-05, 6.0468e-05, 6.9416e-05, 6.6904e-05, 6.5201e-05, 7.8251e-05], device='cuda:6') 2023-04-27 07:15:31,813 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.3424, 1.9983, 2.3788, 2.7599, 2.7109, 2.2767, 1.8676, 2.3776], device='cuda:6'), covar=tensor([0.0763, 0.1075, 0.0552, 0.0513, 0.0654, 0.0779, 0.0800, 0.0530], device='cuda:6'), in_proj_covar=tensor([0.0188, 0.0200, 0.0179, 0.0170, 0.0176, 0.0181, 0.0152, 0.0177], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 07:15:49,271 INFO [finetune.py:976] (6/7) Epoch 15, batch 3600, loss[loss=0.1494, simple_loss=0.2118, pruned_loss=0.04344, over 4762.00 frames. ], tot_loss[loss=0.1771, simple_loss=0.2457, pruned_loss=0.05423, over 956844.26 frames. ], batch size: 54, lr: 3.49e-03, grad_scale: 64.0 2023-04-27 07:16:32,021 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=83821.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:16:55,916 INFO [finetune.py:976] (6/7) Epoch 15, batch 3650, loss[loss=0.1995, simple_loss=0.2734, pruned_loss=0.06282, over 4771.00 frames. ], tot_loss[loss=0.1805, simple_loss=0.2487, pruned_loss=0.05615, over 954290.48 frames. ], batch size: 54, lr: 3.48e-03, grad_scale: 64.0 2023-04-27 07:17:01,431 INFO [optim.py:369] (6/7) 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 07:17:05,011 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=83852.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:17:29,864 INFO [finetune.py:976] (6/7) Epoch 15, batch 3700, loss[loss=0.1377, simple_loss=0.2073, pruned_loss=0.03401, over 4788.00 frames. ], tot_loss[loss=0.1825, simple_loss=0.2515, pruned_loss=0.05675, over 954536.70 frames. ], batch size: 25, lr: 3.48e-03, grad_scale: 32.0 2023-04-27 07:17:37,201 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=83900.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:17:43,135 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5473, 1.6522, 0.8483, 1.3178, 1.8098, 1.4324, 1.3436, 1.4414], device='cuda:6'), covar=tensor([0.0491, 0.0366, 0.0358, 0.0534, 0.0298, 0.0498, 0.0473, 0.0548], device='cuda:6'), in_proj_covar=tensor([0.0029, 0.0024, 0.0023, 0.0029, 0.0020, 0.0028, 0.0028, 0.0029], device='cuda:6'), out_proj_covar=tensor([0.0050, 0.0045, 0.0038, 0.0050, 0.0038, 0.0049, 0.0049, 0.0050], device='cuda:6') 2023-04-27 07:17:56,249 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([0.3791, 1.3435, 1.4602, 0.9175, 1.3538, 1.2151, 1.7979, 1.4272], device='cuda:6'), covar=tensor([0.3523, 0.1741, 0.4570, 0.2528, 0.1463, 0.2200, 0.1718, 0.4361], device='cuda:6'), in_proj_covar=tensor([0.0342, 0.0345, 0.0424, 0.0355, 0.0382, 0.0380, 0.0372, 0.0418], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 07:18:03,518 INFO [finetune.py:976] (6/7) Epoch 15, batch 3750, loss[loss=0.2242, simple_loss=0.2824, pruned_loss=0.08302, over 4887.00 frames. ], tot_loss[loss=0.1851, simple_loss=0.254, pruned_loss=0.05806, over 954408.87 frames. ], batch size: 43, lr: 3.48e-03, grad_scale: 32.0 2023-04-27 07:18:04,868 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=83940.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:18:09,571 INFO [optim.py:369] (6/7) 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 07:18:10,280 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=83949.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:18:14,401 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5206, 1.3478, 0.6314, 1.2404, 1.4775, 1.4011, 1.2948, 1.3296], device='cuda:6'), covar=tensor([0.0476, 0.0366, 0.0397, 0.0529, 0.0297, 0.0495, 0.0457, 0.0550], device='cuda:6'), in_proj_covar=tensor([0.0029, 0.0024, 0.0023, 0.0029, 0.0020, 0.0029, 0.0028, 0.0029], device='cuda:6'), out_proj_covar=tensor([0.0050, 0.0045, 0.0038, 0.0051, 0.0038, 0.0049, 0.0049, 0.0050], device='cuda:6') 2023-04-27 07:18:26,660 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7099, 1.3038, 1.8693, 2.1178, 1.7355, 1.6779, 1.8214, 1.8159], device='cuda:6'), covar=tensor([0.5328, 0.7571, 0.7280, 0.6566, 0.6777, 0.8772, 0.8919, 0.9397], device='cuda:6'), in_proj_covar=tensor([0.0415, 0.0407, 0.0494, 0.0508, 0.0446, 0.0467, 0.0475, 0.0479], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 07:18:34,078 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=83985.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:18:36,295 INFO [finetune.py:976] (6/7) Epoch 15, batch 3800, loss[loss=0.1799, simple_loss=0.2518, pruned_loss=0.05398, over 4882.00 frames. ], tot_loss[loss=0.1865, simple_loss=0.2554, pruned_loss=0.05879, over 951410.84 frames. ], batch size: 32, lr: 3.48e-03, grad_scale: 32.0 2023-04-27 07:18:36,355 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=83988.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:18:51,932 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=84010.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 07:19:00,362 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5523, 1.6009, 0.6115, 1.2809, 1.5614, 1.4293, 1.3472, 1.3677], device='cuda:6'), covar=tensor([0.0499, 0.0376, 0.0400, 0.0561, 0.0309, 0.0516, 0.0513, 0.0581], device='cuda:6'), in_proj_covar=tensor([0.0029, 0.0024, 0.0023, 0.0029, 0.0020, 0.0028, 0.0028, 0.0029], device='cuda:6'), out_proj_covar=tensor([0.0050, 0.0045, 0.0038, 0.0050, 0.0038, 0.0049, 0.0049, 0.0050], device='cuda:6') 2023-04-27 07:19:10,673 INFO [finetune.py:976] (6/7) Epoch 15, batch 3850, loss[loss=0.1569, simple_loss=0.2215, pruned_loss=0.04616, over 4223.00 frames. ], tot_loss[loss=0.1848, simple_loss=0.2541, pruned_loss=0.05775, over 952384.75 frames. ], batch size: 65, lr: 3.48e-03, grad_scale: 32.0 2023-04-27 07:19:21,658 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=84046.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:19:22,270 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7503, 1.5459, 1.7960, 2.1236, 2.0949, 1.8372, 1.4264, 1.8991], device='cuda:6'), covar=tensor([0.0855, 0.1178, 0.0710, 0.0545, 0.0676, 0.0750, 0.0790, 0.0580], device='cuda:6'), in_proj_covar=tensor([0.0187, 0.0199, 0.0178, 0.0169, 0.0175, 0.0180, 0.0151, 0.0177], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 07:19:22,760 INFO [optim.py:369] (6/7) 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 07:19:31,188 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-04-27 07:20:14,211 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0081, 2.6019, 0.9503, 1.4318, 1.9303, 1.1677, 3.3296, 1.6002], device='cuda:6'), covar=tensor([0.0676, 0.0709, 0.0852, 0.1145, 0.0507, 0.0974, 0.0209, 0.0633], device='cuda:6'), in_proj_covar=tensor([0.0051, 0.0066, 0.0048, 0.0046, 0.0050, 0.0052, 0.0075, 0.0051], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0007, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:6') 2023-04-27 07:20:15,363 INFO [finetune.py:976] (6/7) Epoch 15, batch 3900, loss[loss=0.1559, simple_loss=0.2294, pruned_loss=0.04126, over 4792.00 frames. ], tot_loss[loss=0.1832, simple_loss=0.2518, pruned_loss=0.05731, over 953430.22 frames. ], batch size: 26, lr: 3.48e-03, grad_scale: 32.0 2023-04-27 07:20:27,923 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=84097.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:20:28,552 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.4675, 1.7284, 1.8306, 1.9526, 1.7973, 1.9313, 1.8877, 1.8344], device='cuda:6'), covar=tensor([0.4585, 0.6453, 0.4899, 0.4857, 0.6133, 0.7703, 0.5569, 0.5584], device='cuda:6'), in_proj_covar=tensor([0.0327, 0.0371, 0.0317, 0.0332, 0.0341, 0.0396, 0.0352, 0.0325], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 07:20:35,449 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.66 vs. limit=2.0 2023-04-27 07:20:49,186 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2023-04-27 07:20:50,574 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=84116.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:21:20,859 INFO [finetune.py:976] (6/7) Epoch 15, batch 3950, loss[loss=0.1605, simple_loss=0.2206, pruned_loss=0.05024, over 4752.00 frames. ], tot_loss[loss=0.1796, simple_loss=0.2479, pruned_loss=0.05571, over 954627.16 frames. ], batch size: 59, lr: 3.48e-03, grad_scale: 32.0 2023-04-27 07:21:31,293 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.1549, 2.8218, 2.0721, 2.0413, 1.4961, 1.5106, 2.2730, 1.6243], device='cuda:6'), covar=tensor([0.1646, 0.1353, 0.1403, 0.1767, 0.2288, 0.1936, 0.0977, 0.1960], device='cuda:6'), in_proj_covar=tensor([0.0197, 0.0215, 0.0169, 0.0205, 0.0202, 0.0185, 0.0157, 0.0187], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:6') 2023-04-27 07:21:34,732 INFO [optim.py:369] (6/7) 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 07:21:41,478 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=84158.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:22:12,111 INFO [finetune.py:976] (6/7) Epoch 15, batch 4000, loss[loss=0.2646, simple_loss=0.3082, pruned_loss=0.1105, over 4260.00 frames. ], tot_loss[loss=0.1788, simple_loss=0.2465, pruned_loss=0.05557, over 952489.86 frames. ], batch size: 65, lr: 3.48e-03, grad_scale: 32.0 2023-04-27 07:22:23,814 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7222, 1.3251, 1.8169, 2.2395, 1.8794, 1.6871, 1.7766, 1.7275], device='cuda:6'), covar=tensor([0.5130, 0.7068, 0.6700, 0.6445, 0.6028, 0.8891, 0.9070, 0.8839], device='cuda:6'), in_proj_covar=tensor([0.0416, 0.0406, 0.0494, 0.0506, 0.0445, 0.0468, 0.0475, 0.0477], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 07:22:31,738 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=84199.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:23:00,597 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 15, batch 4050, loss[loss=0.2297, simple_loss=0.2997, pruned_loss=0.07984, over 4800.00 frames. ], tot_loss[loss=0.1806, simple_loss=0.2487, pruned_loss=0.05622, over 953408.28 frames. ], batch size: 45, lr: 3.48e-03, grad_scale: 32.0 2023-04-27 07:23:09,758 INFO [optim.py:369] (6/7) 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 07:23:18,189 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=84260.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:23:34,049 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-04-27 07:23:35,684 INFO [finetune.py:976] (6/7) Epoch 15, batch 4100, loss[loss=0.2131, simple_loss=0.284, pruned_loss=0.07109, over 4811.00 frames. ], tot_loss[loss=0.1824, simple_loss=0.2514, pruned_loss=0.05669, over 953923.80 frames. ], batch size: 38, lr: 3.48e-03, grad_scale: 32.0 2023-04-27 07:23:42,203 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=84297.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:23:48,048 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=84305.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 07:24:08,400 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=84336.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:24:09,504 INFO [finetune.py:976] (6/7) Epoch 15, batch 4150, loss[loss=0.2189, simple_loss=0.2906, pruned_loss=0.07359, over 4888.00 frames. ], tot_loss[loss=0.1848, simple_loss=0.2539, pruned_loss=0.05784, over 951107.61 frames. ], batch size: 35, lr: 3.48e-03, grad_scale: 32.0 2023-04-27 07:24:11,378 INFO [zipformer.py:1188] (6/7) 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] (6/7) 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 07:24:43,282 INFO [finetune.py:976] (6/7) Epoch 15, batch 4200, loss[loss=0.1565, simple_loss=0.2136, pruned_loss=0.04969, over 4730.00 frames. ], tot_loss[loss=0.1857, simple_loss=0.255, pruned_loss=0.05821, over 952391.73 frames. ], batch size: 59, lr: 3.48e-03, grad_scale: 32.0 2023-04-27 07:24:49,272 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=84397.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:24:59,711 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-04-27 07:25:03,281 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=84416.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:25:22,955 INFO [finetune.py:976] (6/7) Epoch 15, batch 4250, loss[loss=0.1524, simple_loss=0.2236, pruned_loss=0.04058, over 4844.00 frames. ], tot_loss[loss=0.1839, simple_loss=0.2528, pruned_loss=0.05749, over 953686.38 frames. ], batch size: 25, lr: 3.48e-03, grad_scale: 32.0 2023-04-27 07:25:33,108 INFO [optim.py:369] (6/7) 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 07:25:42,010 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=84453.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:25:56,125 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=84464.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:25:57,514 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.67 vs. limit=2.0 2023-04-27 07:26:28,014 INFO [finetune.py:976] (6/7) Epoch 15, batch 4300, loss[loss=0.176, simple_loss=0.2408, pruned_loss=0.05566, over 4912.00 frames. ], tot_loss[loss=0.1798, simple_loss=0.2482, pruned_loss=0.05573, over 953278.57 frames. ], batch size: 37, lr: 3.48e-03, grad_scale: 32.0 2023-04-27 07:27:19,022 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=84525.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:27:32,536 INFO [finetune.py:976] (6/7) Epoch 15, batch 4350, loss[loss=0.1802, simple_loss=0.236, pruned_loss=0.06223, over 4070.00 frames. ], tot_loss[loss=0.1777, simple_loss=0.2459, pruned_loss=0.05479, over 954832.65 frames. ], batch size: 17, lr: 3.48e-03, grad_scale: 32.0 2023-04-27 07:27:44,743 INFO [optim.py:369] (6/7) 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 07:27:54,633 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=84555.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:28:26,836 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=84586.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:28:27,952 INFO [finetune.py:976] (6/7) Epoch 15, batch 4400, loss[loss=0.1744, simple_loss=0.2547, pruned_loss=0.04704, over 4825.00 frames. ], tot_loss[loss=0.1783, simple_loss=0.247, pruned_loss=0.05477, over 955621.47 frames. ], batch size: 40, lr: 3.48e-03, grad_scale: 32.0 2023-04-27 07:28:30,487 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=84592.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:28:39,056 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=84605.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:28:59,174 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=84633.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:29:02,150 INFO [finetune.py:976] (6/7) Epoch 15, batch 4450, loss[loss=0.2149, simple_loss=0.2837, pruned_loss=0.07307, over 4737.00 frames. ], tot_loss[loss=0.1805, simple_loss=0.2497, pruned_loss=0.05566, over 953574.81 frames. ], batch size: 54, lr: 3.48e-03, grad_scale: 32.0 2023-04-27 07:29:04,104 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=84641.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:29:08,293 INFO [optim.py:369] (6/7) 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] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=84653.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:29:36,304 INFO [finetune.py:976] (6/7) Epoch 15, batch 4500, loss[loss=0.1922, simple_loss=0.2598, pruned_loss=0.06229, over 4790.00 frames. ], tot_loss[loss=0.1817, simple_loss=0.2516, pruned_loss=0.05593, over 955940.51 frames. ], batch size: 29, lr: 3.48e-03, grad_scale: 32.0 2023-04-27 07:29:37,002 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=84689.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:29:38,816 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=84692.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:29:40,113 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=84694.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:29:58,537 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8555, 2.5959, 1.9886, 2.2981, 1.7328, 2.0013, 2.1511, 1.5900], device='cuda:6'), covar=tensor([0.2157, 0.0991, 0.0838, 0.1218, 0.3117, 0.1286, 0.1891, 0.2427], device='cuda:6'), in_proj_covar=tensor([0.0287, 0.0308, 0.0219, 0.0281, 0.0311, 0.0264, 0.0250, 0.0266], device='cuda:6'), out_proj_covar=tensor([1.1547e-04, 1.2243e-04, 8.7448e-05, 1.1185e-04, 1.2658e-04, 1.0516e-04, 1.0123e-04, 1.0598e-04], device='cuda:6') 2023-04-27 07:30:09,977 INFO [finetune.py:976] (6/7) Epoch 15, batch 4550, loss[loss=0.2167, simple_loss=0.287, pruned_loss=0.07316, over 4814.00 frames. ], tot_loss[loss=0.1841, simple_loss=0.2541, pruned_loss=0.057, over 955723.01 frames. ], batch size: 39, lr: 3.48e-03, grad_scale: 32.0 2023-04-27 07:30:16,094 INFO [optim.py:369] (6/7) 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 07:30:19,285 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=84753.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:30:42,052 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.1474, 1.2935, 1.6718, 1.7858, 1.6576, 1.7615, 1.7022, 1.6953], device='cuda:6'), covar=tensor([0.3694, 0.5428, 0.4275, 0.4224, 0.5409, 0.7163, 0.5063, 0.4765], device='cuda:6'), in_proj_covar=tensor([0.0332, 0.0375, 0.0319, 0.0334, 0.0346, 0.0400, 0.0356, 0.0326], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 07:30:43,750 INFO [finetune.py:976] (6/7) Epoch 15, batch 4600, loss[loss=0.1493, simple_loss=0.2349, pruned_loss=0.03192, over 4811.00 frames. ], tot_loss[loss=0.1823, simple_loss=0.2529, pruned_loss=0.05579, over 957315.13 frames. ], batch size: 41, lr: 3.48e-03, grad_scale: 32.0 2023-04-27 07:30:51,295 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=84800.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:30:51,853 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=84801.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:31:22,560 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8990, 2.5139, 1.9547, 1.8363, 1.4069, 1.4405, 2.0870, 1.3947], device='cuda:6'), covar=tensor([0.1658, 0.1363, 0.1432, 0.1746, 0.2355, 0.2003, 0.1029, 0.2022], device='cuda:6'), in_proj_covar=tensor([0.0196, 0.0213, 0.0168, 0.0204, 0.0201, 0.0184, 0.0155, 0.0185], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:6') 2023-04-27 07:31:33,985 INFO [finetune.py:976] (6/7) Epoch 15, batch 4650, loss[loss=0.188, simple_loss=0.2675, pruned_loss=0.05423, over 4864.00 frames. ], tot_loss[loss=0.1794, simple_loss=0.2493, pruned_loss=0.05475, over 956802.28 frames. ], batch size: 34, lr: 3.48e-03, grad_scale: 32.0 2023-04-27 07:31:45,944 INFO [optim.py:369] (6/7) 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 07:31:48,008 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.61 vs. limit=2.0 2023-04-27 07:31:55,712 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=84855.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:32:05,683 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=84861.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:32:30,053 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=84881.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:32:39,701 INFO [finetune.py:976] (6/7) Epoch 15, batch 4700, loss[loss=0.1529, simple_loss=0.2184, pruned_loss=0.04368, over 4769.00 frames. ], tot_loss[loss=0.1779, simple_loss=0.2469, pruned_loss=0.05443, over 958307.36 frames. ], batch size: 26, lr: 3.48e-03, grad_scale: 32.0 2023-04-27 07:32:48,394 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=84892.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:33:00,660 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=84903.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:33:46,043 INFO [finetune.py:976] (6/7) Epoch 15, batch 4750, loss[loss=0.1704, simple_loss=0.2312, pruned_loss=0.05481, over 4902.00 frames. ], tot_loss[loss=0.1764, simple_loss=0.245, pruned_loss=0.05386, over 958230.27 frames. ], batch size: 32, lr: 3.48e-03, grad_scale: 32.0 2023-04-27 07:33:47,830 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=84940.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:33:53,645 INFO [optim.py:369] (6/7) 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 07:33:55,672 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=84951.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:34:00,311 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-04-27 07:34:19,537 INFO [finetune.py:976] (6/7) Epoch 15, batch 4800, loss[loss=0.1586, simple_loss=0.2461, pruned_loss=0.03552, over 4788.00 frames. ], tot_loss[loss=0.1785, simple_loss=0.2481, pruned_loss=0.05445, over 957985.46 frames. ], batch size: 29, lr: 3.47e-03, grad_scale: 32.0 2023-04-27 07:34:20,717 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=84989.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:34:23,034 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=84992.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:34:28,075 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-04-27 07:34:32,837 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-04-27 07:34:36,376 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=85012.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:34:47,615 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-04-27 07:34:53,702 INFO [finetune.py:976] (6/7) Epoch 15, batch 4850, loss[loss=0.1629, simple_loss=0.2461, pruned_loss=0.03989, over 4852.00 frames. ], tot_loss[loss=0.1814, simple_loss=0.2519, pruned_loss=0.05545, over 957783.14 frames. ], batch size: 44, lr: 3.47e-03, grad_scale: 32.0 2023-04-27 07:34:55,433 INFO [zipformer.py:1188] (6/7) 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] (6/7) 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 07:35:03,620 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.2963, 1.6204, 1.4462, 1.4986, 1.3899, 1.2327, 1.3505, 1.1072], device='cuda:6'), covar=tensor([0.1621, 0.1254, 0.0914, 0.1187, 0.3302, 0.1361, 0.1785, 0.2114], device='cuda:6'), in_proj_covar=tensor([0.0287, 0.0308, 0.0220, 0.0281, 0.0313, 0.0264, 0.0251, 0.0266], device='cuda:6'), out_proj_covar=tensor([1.1560e-04, 1.2271e-04, 8.7787e-05, 1.1179e-04, 1.2746e-04, 1.0530e-04, 1.0171e-04, 1.0601e-04], device='cuda:6') 2023-04-27 07:35:27,218 INFO [finetune.py:976] (6/7) Epoch 15, batch 4900, loss[loss=0.1816, simple_loss=0.2589, pruned_loss=0.05212, over 4843.00 frames. ], tot_loss[loss=0.1839, simple_loss=0.2541, pruned_loss=0.05686, over 957729.05 frames. ], batch size: 49, lr: 3.47e-03, grad_scale: 32.0 2023-04-27 07:35:29,716 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9057, 2.5392, 0.9277, 1.2780, 1.7440, 1.2291, 3.2828, 1.5489], device='cuda:6'), covar=tensor([0.0877, 0.0776, 0.0989, 0.1694, 0.0732, 0.1410, 0.0358, 0.0936], device='cuda:6'), in_proj_covar=tensor([0.0051, 0.0066, 0.0048, 0.0047, 0.0050, 0.0052, 0.0075, 0.0052], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:6') 2023-04-27 07:35:40,344 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=85107.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:35:45,210 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=85115.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:36:00,015 INFO [finetune.py:976] (6/7) Epoch 15, batch 4950, loss[loss=0.2103, simple_loss=0.2825, pruned_loss=0.06907, over 4821.00 frames. ], tot_loss[loss=0.1847, simple_loss=0.2553, pruned_loss=0.05703, over 957850.09 frames. ], batch size: 39, lr: 3.47e-03, grad_scale: 32.0 2023-04-27 07:36:07,087 INFO [optim.py:369] (6/7) 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 07:36:12,984 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=85156.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:36:20,329 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=85168.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 07:36:25,197 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=85176.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:36:28,193 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=85181.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:36:33,336 INFO [finetune.py:976] (6/7) Epoch 15, batch 5000, loss[loss=0.1942, simple_loss=0.2518, pruned_loss=0.06836, over 3986.00 frames. ], tot_loss[loss=0.1832, simple_loss=0.2535, pruned_loss=0.05649, over 955972.89 frames. ], batch size: 17, lr: 3.47e-03, grad_scale: 32.0 2023-04-27 07:37:00,634 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=85229.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:37:12,536 INFO [finetune.py:976] (6/7) Epoch 15, batch 5050, loss[loss=0.1352, simple_loss=0.212, pruned_loss=0.02923, over 4784.00 frames. ], tot_loss[loss=0.1812, simple_loss=0.2506, pruned_loss=0.05588, over 956426.34 frames. ], batch size: 26, lr: 3.47e-03, grad_scale: 32.0 2023-04-27 07:37:25,144 INFO [optim.py:369] (6/7) 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 07:37:47,117 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.4444, 2.9973, 0.8787, 1.6350, 1.8127, 2.1311, 1.7922, 0.9638], device='cuda:6'), covar=tensor([0.1440, 0.1076, 0.1921, 0.1351, 0.1114, 0.1090, 0.1629, 0.1803], device='cuda:6'), in_proj_covar=tensor([0.0117, 0.0245, 0.0137, 0.0121, 0.0132, 0.0152, 0.0119, 0.0120], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-27 07:38:08,699 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.56 vs. limit=2.0 2023-04-27 07:38:17,993 INFO [finetune.py:976] (6/7) Epoch 15, batch 5100, loss[loss=0.1439, simple_loss=0.2172, pruned_loss=0.03535, over 4801.00 frames. ], tot_loss[loss=0.1775, simple_loss=0.2467, pruned_loss=0.05416, over 955458.09 frames. ], batch size: 29, lr: 3.47e-03, grad_scale: 32.0 2023-04-27 07:38:18,684 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=85289.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:38:20,386 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-04-27 07:38:41,597 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=85307.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:38:49,295 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5910, 3.4011, 0.9494, 1.8846, 1.9831, 2.3628, 1.8707, 1.0203], device='cuda:6'), covar=tensor([0.1281, 0.0838, 0.1940, 0.1243, 0.0961, 0.0983, 0.1455, 0.1832], device='cuda:6'), in_proj_covar=tensor([0.0116, 0.0242, 0.0136, 0.0120, 0.0130, 0.0151, 0.0118, 0.0119], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-27 07:38:55,363 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.3314, 1.1372, 1.2399, 1.5916, 1.6733, 1.3630, 0.8822, 1.5060], device='cuda:6'), covar=tensor([0.0912, 0.1757, 0.1194, 0.0738, 0.0750, 0.0925, 0.1116, 0.0720], device='cuda:6'), in_proj_covar=tensor([0.0193, 0.0207, 0.0185, 0.0177, 0.0182, 0.0186, 0.0157, 0.0184], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 07:39:06,788 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=85337.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:39:07,343 INFO [finetune.py:976] (6/7) Epoch 15, batch 5150, loss[loss=0.2163, simple_loss=0.2742, pruned_loss=0.07925, over 4825.00 frames. ], tot_loss[loss=0.1776, simple_loss=0.2463, pruned_loss=0.05442, over 952675.31 frames. ], batch size: 40, lr: 3.47e-03, grad_scale: 32.0 2023-04-27 07:39:20,283 INFO [optim.py:369] (6/7) 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 07:39:38,585 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5518, 0.9230, 1.5645, 2.0765, 1.6353, 1.5187, 1.5761, 1.5427], device='cuda:6'), covar=tensor([0.4321, 0.6303, 0.5646, 0.5482, 0.5361, 0.7395, 0.7026, 0.7311], device='cuda:6'), in_proj_covar=tensor([0.0416, 0.0407, 0.0494, 0.0506, 0.0445, 0.0469, 0.0475, 0.0478], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 07:39:47,578 INFO [finetune.py:976] (6/7) Epoch 15, batch 5200, loss[loss=0.2019, simple_loss=0.2609, pruned_loss=0.07144, over 4839.00 frames. ], tot_loss[loss=0.1814, simple_loss=0.2506, pruned_loss=0.05608, over 952744.34 frames. ], batch size: 25, lr: 3.47e-03, grad_scale: 32.0 2023-04-27 07:39:50,123 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.1977, 2.9362, 0.8996, 1.5164, 2.3364, 1.3598, 4.0529, 1.9331], device='cuda:6'), covar=tensor([0.0727, 0.0803, 0.0948, 0.1322, 0.0503, 0.1078, 0.0257, 0.0641], device='cuda:6'), in_proj_covar=tensor([0.0051, 0.0066, 0.0049, 0.0047, 0.0050, 0.0052, 0.0076, 0.0052], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:6') 2023-04-27 07:40:05,472 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.60 vs. limit=5.0 2023-04-27 07:40:14,861 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.4380, 1.6642, 1.2227, 1.0867, 1.0939, 1.0787, 1.2344, 1.0136], device='cuda:6'), covar=tensor([0.1989, 0.1313, 0.1912, 0.1974, 0.2640, 0.2375, 0.1174, 0.2288], device='cuda:6'), in_proj_covar=tensor([0.0195, 0.0212, 0.0168, 0.0203, 0.0201, 0.0184, 0.0155, 0.0185], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:6') 2023-04-27 07:40:15,418 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 15, batch 5250, loss[loss=0.1778, simple_loss=0.2474, pruned_loss=0.05414, over 4801.00 frames. ], tot_loss[loss=0.1824, simple_loss=0.2519, pruned_loss=0.05642, over 952544.56 frames. ], batch size: 51, lr: 3.47e-03, grad_scale: 32.0 2023-04-27 07:40:27,422 INFO [optim.py:369] (6/7) 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 07:40:33,765 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=85456.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:40:38,414 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=85463.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 07:40:44,665 INFO [zipformer.py:1188] (6/7) 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:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=85483.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:40:54,958 INFO [finetune.py:976] (6/7) Epoch 15, batch 5300, loss[loss=0.1493, simple_loss=0.2224, pruned_loss=0.03813, over 4766.00 frames. ], tot_loss[loss=0.1846, simple_loss=0.2544, pruned_loss=0.0574, over 955062.12 frames. ], batch size: 27, lr: 3.47e-03, grad_scale: 32.0 2023-04-27 07:40:55,707 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=85489.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:41:04,897 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=85504.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:41:23,647 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=85530.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:41:28,361 INFO [finetune.py:976] (6/7) Epoch 15, batch 5350, loss[loss=0.1685, simple_loss=0.2446, pruned_loss=0.04618, over 4840.00 frames. ], tot_loss[loss=0.1842, simple_loss=0.2538, pruned_loss=0.05726, over 954642.22 frames. ], batch size: 44, lr: 3.47e-03, grad_scale: 32.0 2023-04-27 07:41:32,143 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=85544.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:41:34,430 INFO [optim.py:369] (6/7) 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 07:41:38,544 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-04-27 07:42:02,178 INFO [finetune.py:976] (6/7) Epoch 15, batch 5400, loss[loss=0.2077, simple_loss=0.268, pruned_loss=0.07372, over 4922.00 frames. ], tot_loss[loss=0.1819, simple_loss=0.2514, pruned_loss=0.05622, over 952782.35 frames. ], batch size: 33, lr: 3.47e-03, grad_scale: 32.0 2023-04-27 07:42:04,160 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=85591.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:42:13,927 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=85607.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:42:27,862 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=85625.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:42:41,672 INFO [finetune.py:976] (6/7) Epoch 15, batch 5450, loss[loss=0.1586, simple_loss=0.2316, pruned_loss=0.04277, over 4913.00 frames. ], tot_loss[loss=0.1794, simple_loss=0.2482, pruned_loss=0.05529, over 952272.25 frames. ], batch size: 36, lr: 3.47e-03, grad_scale: 32.0 2023-04-27 07:42:53,055 INFO [optim.py:369] (6/7) 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 07:43:02,874 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=85655.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:43:45,383 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=85686.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 07:43:46,491 INFO [finetune.py:976] (6/7) Epoch 15, batch 5500, loss[loss=0.1583, simple_loss=0.2255, pruned_loss=0.04553, over 4932.00 frames. ], tot_loss[loss=0.1769, simple_loss=0.245, pruned_loss=0.05447, over 952759.11 frames. ], batch size: 38, lr: 3.47e-03, grad_scale: 32.0 2023-04-27 07:44:51,826 INFO [finetune.py:976] (6/7) Epoch 15, batch 5550, loss[loss=0.1961, simple_loss=0.2593, pruned_loss=0.0664, over 4926.00 frames. ], tot_loss[loss=0.1782, simple_loss=0.2465, pruned_loss=0.05493, over 954233.10 frames. ], batch size: 38, lr: 3.47e-03, grad_scale: 32.0 2023-04-27 07:45:02,125 INFO [optim.py:369] (6/7) 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 07:45:11,411 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([4.0382, 4.0230, 2.8617, 4.5604, 4.0216, 3.9708, 1.8129, 3.9479], device='cuda:6'), covar=tensor([0.1620, 0.1100, 0.3217, 0.1393, 0.4485, 0.1802, 0.5657, 0.2289], device='cuda:6'), in_proj_covar=tensor([0.0243, 0.0212, 0.0249, 0.0301, 0.0295, 0.0246, 0.0269, 0.0269], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 07:45:11,454 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=85763.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:45:16,814 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=85771.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:45:24,807 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 15, batch 5600, loss[loss=0.1972, simple_loss=0.271, pruned_loss=0.0617, over 4902.00 frames. ], tot_loss[loss=0.181, simple_loss=0.2501, pruned_loss=0.05593, over 954084.91 frames. ], batch size: 35, lr: 3.47e-03, grad_scale: 32.0 2023-04-27 07:45:38,909 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([3.4792, 3.4771, 2.9149, 4.0106, 3.3258, 3.4463, 1.8595, 3.4996], device='cuda:6'), covar=tensor([0.1717, 0.1136, 0.3774, 0.1391, 0.3313, 0.1708, 0.4864, 0.2099], device='cuda:6'), in_proj_covar=tensor([0.0243, 0.0212, 0.0249, 0.0301, 0.0295, 0.0246, 0.0269, 0.0269], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 07:45:40,675 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=85811.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:45:44,887 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9242, 1.4634, 1.7466, 1.7440, 1.7435, 1.4321, 0.7660, 1.3995], device='cuda:6'), covar=tensor([0.3406, 0.3449, 0.1818, 0.2296, 0.2582, 0.2783, 0.4321, 0.2178], device='cuda:6'), in_proj_covar=tensor([0.0290, 0.0246, 0.0223, 0.0315, 0.0216, 0.0230, 0.0229, 0.0186], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:6') 2023-04-27 07:45:45,402 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=85819.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:45:57,427 INFO [finetune.py:976] (6/7) Epoch 15, batch 5650, loss[loss=0.2125, simple_loss=0.2764, pruned_loss=0.07432, over 4874.00 frames. ], tot_loss[loss=0.1832, simple_loss=0.2528, pruned_loss=0.05679, over 952958.76 frames. ], batch size: 34, lr: 3.47e-03, grad_scale: 32.0 2023-04-27 07:45:58,082 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=85839.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:46:03,465 INFO [optim.py:369] (6/7) 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 07:46:07,659 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6821, 2.0581, 1.6183, 1.4923, 1.2547, 1.2720, 1.6816, 1.2207], device='cuda:6'), covar=tensor([0.1623, 0.1305, 0.1385, 0.1689, 0.2287, 0.1908, 0.1008, 0.1987], device='cuda:6'), in_proj_covar=tensor([0.0196, 0.0213, 0.0169, 0.0204, 0.0201, 0.0184, 0.0156, 0.0186], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:6') 2023-04-27 07:46:10,915 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7372, 1.7231, 0.8939, 1.4163, 1.8204, 1.5966, 1.4935, 1.5382], device='cuda:6'), covar=tensor([0.0508, 0.0377, 0.0355, 0.0570, 0.0264, 0.0510, 0.0510, 0.0559], device='cuda:6'), in_proj_covar=tensor([0.0029, 0.0024, 0.0022, 0.0029, 0.0020, 0.0028, 0.0028, 0.0029], device='cuda:6'), out_proj_covar=tensor([0.0050, 0.0044, 0.0038, 0.0051, 0.0038, 0.0049, 0.0049, 0.0050], device='cuda:6') 2023-04-27 07:46:19,871 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=85875.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:46:21,128 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=3.63 vs. limit=5.0 2023-04-27 07:46:21,768 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-04-27 07:46:24,098 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.2269, 2.4878, 2.6021, 2.9210, 2.7503, 2.8222, 2.5004, 4.9221], device='cuda:6'), covar=tensor([0.0485, 0.0679, 0.0638, 0.0939, 0.0541, 0.0382, 0.0606, 0.0112], device='cuda:6'), in_proj_covar=tensor([0.0038, 0.0039, 0.0040, 0.0044, 0.0040, 0.0038, 0.0038, 0.0056], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:6') 2023-04-27 07:46:26,480 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=85886.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:46:27,650 INFO [finetune.py:976] (6/7) Epoch 15, batch 5700, loss[loss=0.1263, simple_loss=0.1903, pruned_loss=0.03117, over 4361.00 frames. ], tot_loss[loss=0.18, simple_loss=0.249, pruned_loss=0.05553, over 938783.65 frames. ], batch size: 19, lr: 3.47e-03, grad_scale: 64.0 2023-04-27 07:46:59,232 INFO [finetune.py:976] (6/7) Epoch 16, batch 0, loss[loss=0.1803, simple_loss=0.2344, pruned_loss=0.06312, over 4516.00 frames. ], tot_loss[loss=0.1803, simple_loss=0.2344, pruned_loss=0.06312, over 4516.00 frames. ], batch size: 19, lr: 3.46e-03, grad_scale: 64.0 2023-04-27 07:46:59,233 INFO [finetune.py:1001] (6/7) Computing validation loss 2023-04-27 07:47:13,072 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.1893, 2.5331, 0.9703, 1.4333, 1.9735, 1.3577, 3.0463, 1.7243], device='cuda:6'), covar=tensor([0.0617, 0.0563, 0.0751, 0.1247, 0.0427, 0.0918, 0.0275, 0.0586], device='cuda:6'), in_proj_covar=tensor([0.0051, 0.0066, 0.0048, 0.0046, 0.0050, 0.0052, 0.0075, 0.0051], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:6') 2023-04-27 07:47:15,739 INFO [finetune.py:1010] (6/7) Epoch 16, validation: loss=0.1534, simple_loss=0.2252, pruned_loss=0.04076, over 2265189.00 frames. 2023-04-27 07:47:15,740 INFO [finetune.py:1011] (6/7) Maximum memory allocated so far is 6365MB 2023-04-27 07:47:26,822 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5302, 1.4446, 1.7889, 1.8105, 1.3789, 1.2738, 1.5795, 0.9428], device='cuda:6'), covar=tensor([0.0578, 0.0782, 0.0424, 0.0664, 0.0770, 0.1145, 0.0663, 0.0757], device='cuda:6'), in_proj_covar=tensor([0.0068, 0.0070, 0.0070, 0.0068, 0.0076, 0.0097, 0.0076, 0.0069], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-27 07:47:30,146 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.10 vs. limit=2.0 2023-04-27 07:47:30,517 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=85930.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 07:47:32,917 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6527, 1.6990, 0.8584, 1.3569, 1.8052, 1.5274, 1.4513, 1.4470], device='cuda:6'), covar=tensor([0.0450, 0.0323, 0.0334, 0.0519, 0.0271, 0.0460, 0.0433, 0.0508], device='cuda:6'), in_proj_covar=tensor([0.0029, 0.0024, 0.0022, 0.0029, 0.0020, 0.0028, 0.0028, 0.0028], device='cuda:6'), out_proj_covar=tensor([0.0050, 0.0044, 0.0037, 0.0050, 0.0037, 0.0049, 0.0049, 0.0050], device='cuda:6') 2023-04-27 07:47:34,156 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=85936.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:47:41,351 INFO [optim.py:369] (6/7) 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 07:47:43,373 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.5670, 1.8690, 2.2943, 3.1080, 2.3427, 1.7673, 1.8733, 2.3531], device='cuda:6'), covar=tensor([0.3283, 0.3445, 0.1643, 0.2257, 0.2911, 0.2804, 0.3823, 0.2096], device='cuda:6'), in_proj_covar=tensor([0.0291, 0.0247, 0.0224, 0.0316, 0.0216, 0.0230, 0.0229, 0.0186], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:6') 2023-04-27 07:47:47,002 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=85957.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:47:53,861 INFO [finetune.py:976] (6/7) Epoch 16, batch 50, loss[loss=0.1741, simple_loss=0.244, pruned_loss=0.05206, over 4885.00 frames. ], tot_loss[loss=0.1846, simple_loss=0.2546, pruned_loss=0.0573, over 216239.19 frames. ], batch size: 32, lr: 3.46e-03, grad_scale: 64.0 2023-04-27 07:48:04,130 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=85981.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 07:48:10,284 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=85991.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 07:48:23,355 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=86002.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:48:31,119 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 2023-04-27 07:48:44,015 INFO [finetune.py:976] (6/7) Epoch 16, batch 100, loss[loss=0.1723, simple_loss=0.2367, pruned_loss=0.05395, over 4734.00 frames. ], tot_loss[loss=0.1788, simple_loss=0.2476, pruned_loss=0.05499, over 379628.76 frames. ], batch size: 59, lr: 3.46e-03, grad_scale: 64.0 2023-04-27 07:48:50,971 INFO [zipformer.py:1188] (6/7) 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] (6/7) 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 07:49:34,861 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=86052.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:49:47,415 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 16, batch 150, loss[loss=0.1894, simple_loss=0.2452, pruned_loss=0.06683, over 4913.00 frames. ], tot_loss[loss=0.1765, simple_loss=0.2436, pruned_loss=0.05466, over 508808.71 frames. ], batch size: 43, lr: 3.46e-03, grad_scale: 64.0 2023-04-27 07:49:52,566 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=86069.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:49:57,299 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-04-27 07:50:02,665 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=86084.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:50:13,589 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.3057, 1.5617, 1.4423, 1.7643, 1.6964, 2.0124, 1.3976, 3.7520], device='cuda:6'), covar=tensor([0.0598, 0.0805, 0.0811, 0.1199, 0.0657, 0.0481, 0.0750, 0.0119], device='cuda:6'), in_proj_covar=tensor([0.0037, 0.0039, 0.0040, 0.0043, 0.0040, 0.0038, 0.0038, 0.0056], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:6') 2023-04-27 07:50:37,996 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=86113.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:50:45,308 INFO [finetune.py:976] (6/7) Epoch 16, batch 200, loss[loss=0.132, simple_loss=0.1969, pruned_loss=0.03352, over 4766.00 frames. ], tot_loss[loss=0.1742, simple_loss=0.2418, pruned_loss=0.05332, over 609962.77 frames. ], batch size: 27, lr: 3.46e-03, grad_scale: 64.0 2023-04-27 07:51:06,765 INFO [zipformer.py:1188] (6/7) 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] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=86132.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:51:16,788 INFO [zipformer.py:1188] (6/7) 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] (6/7) 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 07:51:33,769 INFO [finetune.py:976] (6/7) Epoch 16, batch 250, loss[loss=0.1751, simple_loss=0.2477, pruned_loss=0.05126, over 4903.00 frames. ], tot_loss[loss=0.1771, simple_loss=0.2459, pruned_loss=0.05415, over 688051.68 frames. ], batch size: 43, lr: 3.46e-03, grad_scale: 64.0 2023-04-27 07:51:41,370 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.2746, 1.2142, 1.3219, 1.6085, 1.5316, 1.2720, 0.9446, 1.4399], device='cuda:6'), covar=tensor([0.0902, 0.1329, 0.0855, 0.0636, 0.0698, 0.0854, 0.0950, 0.0657], device='cuda:6'), in_proj_covar=tensor([0.0191, 0.0205, 0.0183, 0.0175, 0.0179, 0.0185, 0.0155, 0.0181], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 07:51:48,353 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=86186.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:51:48,949 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=86187.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:51:52,584 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.2693, 1.3792, 3.8872, 3.6033, 3.4544, 3.7355, 3.7160, 3.4202], device='cuda:6'), covar=tensor([0.7611, 0.5809, 0.1178, 0.1844, 0.1319, 0.1822, 0.1455, 0.1504], device='cuda:6'), in_proj_covar=tensor([0.0312, 0.0306, 0.0401, 0.0408, 0.0347, 0.0406, 0.0311, 0.0366], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-27 07:52:06,461 INFO [finetune.py:976] (6/7) Epoch 16, batch 300, loss[loss=0.1882, simple_loss=0.263, pruned_loss=0.05673, over 4820.00 frames. ], tot_loss[loss=0.1814, simple_loss=0.2512, pruned_loss=0.05584, over 747121.56 frames. ], batch size: 40, lr: 3.46e-03, grad_scale: 64.0 2023-04-27 07:52:17,998 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=86231.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:52:19,823 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=86234.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:52:20,594 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.85 vs. limit=5.0 2023-04-27 07:52:27,213 INFO [zipformer.py:1188] (6/7) 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] (6/7) 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 07:52:30,221 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([3.9415, 3.9145, 2.9618, 4.4811, 3.9307, 3.9665, 1.9458, 3.8478], device='cuda:6'), covar=tensor([0.1781, 0.1028, 0.2954, 0.1637, 0.3759, 0.1795, 0.5329, 0.2416], device='cuda:6'), in_proj_covar=tensor([0.0244, 0.0214, 0.0251, 0.0304, 0.0297, 0.0247, 0.0271, 0.0270], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 07:52:39,296 INFO [finetune.py:976] (6/7) Epoch 16, batch 350, loss[loss=0.2265, simple_loss=0.2938, pruned_loss=0.07958, over 4792.00 frames. ], tot_loss[loss=0.1838, simple_loss=0.2533, pruned_loss=0.0571, over 792450.21 frames. ], batch size: 51, lr: 3.46e-03, grad_scale: 32.0 2023-04-27 07:52:50,032 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=86281.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:52:54,333 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=86286.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 07:52:56,804 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([0.7737, 1.7539, 1.5854, 1.3558, 1.8130, 1.5435, 2.2075, 1.3929], device='cuda:6'), covar=tensor([0.3448, 0.1635, 0.4622, 0.2793, 0.1515, 0.2063, 0.1508, 0.4586], device='cuda:6'), in_proj_covar=tensor([0.0344, 0.0351, 0.0430, 0.0359, 0.0387, 0.0382, 0.0374, 0.0420], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 07:53:07,241 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=86307.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:53:10,808 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=86313.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:53:11,517 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.2899, 1.5917, 1.6791, 1.8238, 1.6867, 1.8265, 1.7502, 1.7729], device='cuda:6'), covar=tensor([0.3652, 0.5291, 0.4643, 0.4256, 0.5463, 0.7166, 0.5093, 0.4681], device='cuda:6'), in_proj_covar=tensor([0.0331, 0.0374, 0.0319, 0.0334, 0.0345, 0.0400, 0.0354, 0.0326], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 07:53:12,610 INFO [finetune.py:976] (6/7) Epoch 16, batch 400, loss[loss=0.1721, simple_loss=0.2529, pruned_loss=0.04568, over 4898.00 frames. ], tot_loss[loss=0.184, simple_loss=0.2546, pruned_loss=0.05666, over 830936.06 frames. ], batch size: 46, lr: 3.46e-03, grad_scale: 32.0 2023-04-27 07:53:13,317 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.1082, 1.5597, 5.3701, 5.0936, 4.6833, 5.1509, 4.6213, 4.7820], device='cuda:6'), covar=tensor([0.6808, 0.5945, 0.0809, 0.1476, 0.1042, 0.1343, 0.1133, 0.1350], device='cuda:6'), in_proj_covar=tensor([0.0312, 0.0305, 0.0402, 0.0408, 0.0348, 0.0407, 0.0311, 0.0367], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-27 07:53:21,590 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=86329.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:53:35,568 INFO [optim.py:369] (6/7) 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 07:53:41,144 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=86358.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:53:44,230 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9393, 1.3182, 1.3012, 1.6427, 2.0301, 1.5400, 1.4230, 1.2635], device='cuda:6'), covar=tensor([0.1594, 0.2101, 0.2518, 0.1599, 0.1107, 0.1969, 0.2108, 0.2493], device='cuda:6'), in_proj_covar=tensor([0.0305, 0.0313, 0.0350, 0.0289, 0.0329, 0.0311, 0.0299, 0.0360], device='cuda:6'), out_proj_covar=tensor([6.3130e-05, 6.5404e-05, 7.4863e-05, 5.9019e-05, 6.8602e-05, 6.5603e-05, 6.3106e-05, 7.6865e-05], device='cuda:6') 2023-04-27 07:53:45,903 INFO [finetune.py:976] (6/7) Epoch 16, batch 450, loss[loss=0.193, simple_loss=0.2589, pruned_loss=0.06357, over 4859.00 frames. ], tot_loss[loss=0.1826, simple_loss=0.2533, pruned_loss=0.05592, over 859813.82 frames. ], batch size: 34, lr: 3.46e-03, grad_scale: 32.0 2023-04-27 07:54:14,885 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=86408.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:54:19,721 INFO [finetune.py:976] (6/7) Epoch 16, batch 500, loss[loss=0.1644, simple_loss=0.2312, pruned_loss=0.04883, over 4430.00 frames. ], tot_loss[loss=0.1812, simple_loss=0.2511, pruned_loss=0.05566, over 881503.88 frames. ], batch size: 19, lr: 3.46e-03, grad_scale: 32.0 2023-04-27 07:54:25,191 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=86425.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:54:42,238 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=86440.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:54:45,737 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=86445.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:54:52,709 INFO [optim.py:369] (6/7) 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 07:55:04,269 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.32 vs. limit=5.0 2023-04-27 07:55:08,755 INFO [finetune.py:976] (6/7) Epoch 16, batch 550, loss[loss=0.1679, simple_loss=0.2373, pruned_loss=0.04926, over 4928.00 frames. ], tot_loss[loss=0.1783, simple_loss=0.2478, pruned_loss=0.05442, over 899300.51 frames. ], batch size: 33, lr: 3.46e-03, grad_scale: 32.0 2023-04-27 07:55:27,659 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-04-27 07:55:29,843 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.10 vs. limit=2.0 2023-04-27 07:56:00,653 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=86501.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:56:09,264 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-04-27 07:56:09,552 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=86506.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 07:56:20,905 INFO [finetune.py:976] (6/7) Epoch 16, batch 600, loss[loss=0.1953, simple_loss=0.2634, pruned_loss=0.06355, over 4742.00 frames. ], tot_loss[loss=0.1785, simple_loss=0.2477, pruned_loss=0.05466, over 911795.56 frames. ], batch size: 54, lr: 3.46e-03, grad_scale: 32.0 2023-04-27 07:56:32,806 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7586, 1.4433, 1.2919, 1.5701, 1.9780, 1.5508, 1.3746, 1.2665], device='cuda:6'), covar=tensor([0.1485, 0.1485, 0.1730, 0.1272, 0.0837, 0.1715, 0.1976, 0.2208], device='cuda:6'), in_proj_covar=tensor([0.0306, 0.0312, 0.0350, 0.0289, 0.0329, 0.0311, 0.0299, 0.0360], device='cuda:6'), out_proj_covar=tensor([6.3267e-05, 6.5197e-05, 7.4795e-05, 5.9126e-05, 6.8614e-05, 6.5673e-05, 6.3203e-05, 7.6881e-05], device='cuda:6') 2023-04-27 07:56:35,706 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=86531.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:56:45,566 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.3120, 1.9188, 2.1848, 2.6262, 2.4862, 2.1525, 1.7619, 2.4478], device='cuda:6'), covar=tensor([0.0690, 0.1086, 0.0563, 0.0477, 0.0545, 0.0706, 0.0721, 0.0429], device='cuda:6'), in_proj_covar=tensor([0.0190, 0.0203, 0.0182, 0.0173, 0.0177, 0.0184, 0.0154, 0.0180], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 07:56:54,679 INFO [optim.py:369] (6/7) 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 07:57:05,049 INFO [finetune.py:976] (6/7) Epoch 16, batch 650, loss[loss=0.1836, simple_loss=0.262, pruned_loss=0.05254, over 4902.00 frames. ], tot_loss[loss=0.1818, simple_loss=0.2513, pruned_loss=0.05613, over 920840.60 frames. ], batch size: 43, lr: 3.46e-03, grad_scale: 32.0 2023-04-27 07:57:12,976 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=86579.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:57:13,630 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=86580.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:57:18,209 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=86586.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 07:57:20,575 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.3126, 1.5454, 1.3865, 1.4835, 1.2544, 1.4207, 1.3597, 1.0530], device='cuda:6'), covar=tensor([0.1809, 0.1291, 0.0973, 0.1106, 0.3421, 0.1016, 0.1797, 0.2266], device='cuda:6'), in_proj_covar=tensor([0.0289, 0.0310, 0.0222, 0.0282, 0.0315, 0.0265, 0.0252, 0.0268], device='cuda:6'), out_proj_covar=tensor([1.1635e-04, 1.2343e-04, 8.8465e-05, 1.1218e-04, 1.2824e-04, 1.0553e-04, 1.0205e-04, 1.0663e-04], device='cuda:6') 2023-04-27 07:57:29,904 INFO [zipformer.py:1188] (6/7) 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] (6/7) 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] (6/7) Epoch 16, batch 700, loss[loss=0.1703, simple_loss=0.2477, pruned_loss=0.04645, over 4249.00 frames. ], tot_loss[loss=0.1825, simple_loss=0.2528, pruned_loss=0.05606, over 928933.95 frames. ], batch size: 65, lr: 3.46e-03, grad_scale: 32.0 2023-04-27 07:57:49,364 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=86634.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 07:57:54,610 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=86641.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:57:55,232 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.3799, 1.5294, 1.7542, 1.8712, 1.7265, 1.8244, 1.8660, 1.8161], device='cuda:6'), covar=tensor([0.3756, 0.5233, 0.4446, 0.4448, 0.5892, 0.7213, 0.4860, 0.4771], device='cuda:6'), in_proj_covar=tensor([0.0330, 0.0371, 0.0318, 0.0332, 0.0344, 0.0397, 0.0352, 0.0325], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 07:58:00,292 INFO [optim.py:369] (6/7) 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 07:58:06,398 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=86658.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:58:08,177 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=86661.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:58:11,210 INFO [finetune.py:976] (6/7) Epoch 16, batch 750, loss[loss=0.182, simple_loss=0.2785, pruned_loss=0.04278, over 4891.00 frames. ], tot_loss[loss=0.1821, simple_loss=0.2527, pruned_loss=0.05574, over 933224.03 frames. ], batch size: 43, lr: 3.46e-03, grad_scale: 32.0 2023-04-27 07:58:38,829 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=86706.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:58:40,076 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=86708.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:58:43,925 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 2023-04-27 07:58:44,306 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.0722, 2.0841, 1.7763, 1.7019, 2.1857, 1.7194, 2.5840, 1.5763], device='cuda:6'), covar=tensor([0.3878, 0.1817, 0.4207, 0.2996, 0.1608, 0.2427, 0.1232, 0.4282], device='cuda:6'), in_proj_covar=tensor([0.0342, 0.0349, 0.0428, 0.0359, 0.0385, 0.0383, 0.0373, 0.0421], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 07:58:44,816 INFO [finetune.py:976] (6/7) Epoch 16, batch 800, loss[loss=0.1538, simple_loss=0.2206, pruned_loss=0.04355, over 4900.00 frames. ], tot_loss[loss=0.1821, simple_loss=0.2526, pruned_loss=0.05575, over 938142.91 frames. ], batch size: 46, lr: 3.46e-03, grad_scale: 32.0 2023-04-27 07:58:50,341 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=86725.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:59:05,807 INFO [optim.py:369] (6/7) 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 07:59:11,626 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=86756.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:59:17,589 INFO [finetune.py:976] (6/7) Epoch 16, batch 850, loss[loss=0.1716, simple_loss=0.2381, pruned_loss=0.05254, over 4864.00 frames. ], tot_loss[loss=0.1798, simple_loss=0.2499, pruned_loss=0.05481, over 943397.60 frames. ], batch size: 31, lr: 3.46e-03, grad_scale: 32.0 2023-04-27 07:59:21,930 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=86773.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:59:36,380 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=86796.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:59:39,994 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=86801.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 07:59:49,823 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=86814.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 07:59:50,501 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8460, 1.3550, 1.8754, 2.3112, 1.9458, 1.7961, 1.8786, 1.8525], device='cuda:6'), covar=tensor([0.4797, 0.6692, 0.6387, 0.5763, 0.6160, 0.7631, 0.7769, 0.7722], device='cuda:6'), in_proj_covar=tensor([0.0419, 0.0409, 0.0496, 0.0507, 0.0448, 0.0471, 0.0478, 0.0481], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 07:59:50,970 INFO [finetune.py:976] (6/7) Epoch 16, batch 900, loss[loss=0.2191, simple_loss=0.2713, pruned_loss=0.08344, over 4738.00 frames. ], tot_loss[loss=0.1779, simple_loss=0.2473, pruned_loss=0.0542, over 947899.03 frames. ], batch size: 23, lr: 3.46e-03, grad_scale: 32.0 2023-04-27 07:59:58,452 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0248, 1.4899, 1.9211, 2.1534, 1.8539, 1.4773, 1.1224, 1.6471], device='cuda:6'), covar=tensor([0.3121, 0.3243, 0.1552, 0.2116, 0.2372, 0.2621, 0.4266, 0.2060], device='cuda:6'), in_proj_covar=tensor([0.0293, 0.0249, 0.0225, 0.0318, 0.0219, 0.0231, 0.0230, 0.0188], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:6') 2023-04-27 08:00:22,184 INFO [optim.py:369] (6/7) 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 08:00:52,314 INFO [finetune.py:976] (6/7) Epoch 16, batch 950, loss[loss=0.1117, simple_loss=0.1703, pruned_loss=0.02656, over 3258.00 frames. ], tot_loss[loss=0.1768, simple_loss=0.2461, pruned_loss=0.05372, over 949497.44 frames. ], batch size: 14, lr: 3.46e-03, grad_scale: 32.0 2023-04-27 08:01:03,393 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=86875.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:01:05,152 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=86878.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:01:05,186 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7975, 2.4033, 1.7976, 1.8193, 1.3368, 1.3501, 1.9543, 1.3085], device='cuda:6'), covar=tensor([0.1688, 0.1421, 0.1496, 0.1739, 0.2337, 0.1940, 0.0945, 0.2016], device='cuda:6'), in_proj_covar=tensor([0.0195, 0.0212, 0.0168, 0.0203, 0.0199, 0.0183, 0.0155, 0.0184], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:6') 2023-04-27 08:01:37,255 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=86902.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:01:58,432 INFO [finetune.py:976] (6/7) Epoch 16, batch 1000, loss[loss=0.1697, simple_loss=0.2552, pruned_loss=0.04211, over 4854.00 frames. ], tot_loss[loss=0.18, simple_loss=0.2492, pruned_loss=0.05541, over 950402.56 frames. ], batch size: 44, lr: 3.46e-03, grad_scale: 32.0 2023-04-27 08:02:18,019 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=86936.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:02:19,922 INFO [zipformer.py:1188] (6/7) 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] (6/7) 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 08:02:37,036 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=86950.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:02:59,847 INFO [finetune.py:976] (6/7) Epoch 16, batch 1050, loss[loss=0.2077, simple_loss=0.2776, pruned_loss=0.06886, over 4913.00 frames. ], tot_loss[loss=0.1809, simple_loss=0.2511, pruned_loss=0.05533, over 950837.81 frames. ], batch size: 36, lr: 3.46e-03, grad_scale: 32.0 2023-04-27 08:03:05,920 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5547, 2.0287, 1.8032, 1.9159, 1.5306, 1.7942, 1.7566, 1.3848], device='cuda:6'), covar=tensor([0.1990, 0.1282, 0.0839, 0.1174, 0.3198, 0.1150, 0.1768, 0.2371], device='cuda:6'), in_proj_covar=tensor([0.0291, 0.0315, 0.0224, 0.0285, 0.0317, 0.0268, 0.0255, 0.0271], device='cuda:6'), out_proj_covar=tensor([1.1730e-04, 1.2519e-04, 8.9292e-05, 1.1343e-04, 1.2901e-04, 1.0679e-04, 1.0303e-04, 1.0780e-04], device='cuda:6') 2023-04-27 08:03:28,657 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=87002.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 08:03:43,717 INFO [finetune.py:976] (6/7) Epoch 16, batch 1100, loss[loss=0.1584, simple_loss=0.2485, pruned_loss=0.03413, over 4922.00 frames. ], tot_loss[loss=0.1823, simple_loss=0.2524, pruned_loss=0.05607, over 951528.09 frames. ], batch size: 38, lr: 3.45e-03, grad_scale: 32.0 2023-04-27 08:04:06,134 INFO [optim.py:369] (6/7) 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 08:04:15,689 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=87063.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 08:04:17,378 INFO [finetune.py:976] (6/7) Epoch 16, batch 1150, loss[loss=0.1226, simple_loss=0.1958, pruned_loss=0.02472, over 4741.00 frames. ], tot_loss[loss=0.1835, simple_loss=0.2537, pruned_loss=0.05668, over 954253.78 frames. ], batch size: 26, lr: 3.45e-03, grad_scale: 32.0 2023-04-27 08:04:37,608 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=87096.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:04:40,686 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=87101.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 08:04:50,576 INFO [finetune.py:976] (6/7) Epoch 16, batch 1200, loss[loss=0.1691, simple_loss=0.2302, pruned_loss=0.05399, over 4763.00 frames. ], tot_loss[loss=0.1808, simple_loss=0.2508, pruned_loss=0.0554, over 954996.34 frames. ], batch size: 28, lr: 3.45e-03, grad_scale: 32.0 2023-04-27 08:05:10,107 INFO [zipformer.py:1188] (6/7) 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] (6/7) 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] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=87149.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:05:24,522 INFO [finetune.py:976] (6/7) Epoch 16, batch 1250, loss[loss=0.1744, simple_loss=0.2404, pruned_loss=0.05417, over 4758.00 frames. ], tot_loss[loss=0.1786, simple_loss=0.2485, pruned_loss=0.05433, over 956953.42 frames. ], batch size: 54, lr: 3.45e-03, grad_scale: 32.0 2023-04-27 08:05:27,049 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=87170.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:05:58,113 INFO [finetune.py:976] (6/7) Epoch 16, batch 1300, loss[loss=0.1868, simple_loss=0.247, pruned_loss=0.06325, over 4829.00 frames. ], tot_loss[loss=0.1765, simple_loss=0.2456, pruned_loss=0.05368, over 955808.84 frames. ], batch size: 39, lr: 3.45e-03, grad_scale: 32.0 2023-04-27 08:06:06,460 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8251, 1.2356, 1.9022, 2.3387, 1.9519, 1.7571, 1.8655, 1.8171], device='cuda:6'), covar=tensor([0.5142, 0.7181, 0.6882, 0.6049, 0.6266, 0.8568, 0.8562, 0.8935], device='cuda:6'), in_proj_covar=tensor([0.0417, 0.0407, 0.0494, 0.0505, 0.0446, 0.0469, 0.0475, 0.0478], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 08:06:09,220 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=87230.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:06:12,071 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=87234.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:06:13,311 INFO [zipformer.py:1188] (6/7) 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] (6/7) 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 08:06:31,977 INFO [finetune.py:976] (6/7) Epoch 16, batch 1350, loss[loss=0.1362, simple_loss=0.2004, pruned_loss=0.03595, over 4237.00 frames. ], tot_loss[loss=0.1763, simple_loss=0.2456, pruned_loss=0.05351, over 956363.99 frames. ], batch size: 18, lr: 3.45e-03, grad_scale: 32.0 2023-04-27 08:06:45,593 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=87284.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:06:50,330 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=87291.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:07:16,736 INFO [finetune.py:976] (6/7) Epoch 16, batch 1400, loss[loss=0.2045, simple_loss=0.2647, pruned_loss=0.0721, over 4757.00 frames. ], tot_loss[loss=0.1792, simple_loss=0.2488, pruned_loss=0.05478, over 955079.71 frames. ], batch size: 27, lr: 3.45e-03, grad_scale: 32.0 2023-04-27 08:07:25,329 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6678, 1.1786, 1.7476, 2.1458, 1.7467, 1.6664, 1.7473, 1.6728], device='cuda:6'), covar=tensor([0.4772, 0.6659, 0.6249, 0.5725, 0.5974, 0.7985, 0.7466, 0.8354], device='cuda:6'), in_proj_covar=tensor([0.0417, 0.0408, 0.0495, 0.0506, 0.0447, 0.0471, 0.0476, 0.0479], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 08:07:58,929 INFO [optim.py:369] (6/7) 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 08:08:15,347 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=87358.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 08:08:17,786 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=87362.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:08:20,152 INFO [finetune.py:976] (6/7) Epoch 16, batch 1450, loss[loss=0.1797, simple_loss=0.2437, pruned_loss=0.05779, over 4900.00 frames. ], tot_loss[loss=0.1803, simple_loss=0.2504, pruned_loss=0.05512, over 954794.98 frames. ], batch size: 35, lr: 3.45e-03, grad_scale: 32.0 2023-04-27 08:08:25,467 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.14 vs. limit=2.0 2023-04-27 08:09:12,956 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.2065, 1.4410, 1.3415, 1.6189, 1.5103, 1.8442, 1.3232, 3.1862], device='cuda:6'), covar=tensor([0.0649, 0.0788, 0.0775, 0.1189, 0.0649, 0.0546, 0.0778, 0.0161], device='cuda:6'), in_proj_covar=tensor([0.0038, 0.0039, 0.0040, 0.0044, 0.0040, 0.0038, 0.0038, 0.0056], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:6') 2023-04-27 08:09:14,072 INFO [finetune.py:976] (6/7) Epoch 16, batch 1500, loss[loss=0.1939, simple_loss=0.2625, pruned_loss=0.06259, over 4888.00 frames. ], tot_loss[loss=0.1824, simple_loss=0.2524, pruned_loss=0.05622, over 956136.67 frames. ], batch size: 32, lr: 3.45e-03, grad_scale: 32.0 2023-04-27 08:09:17,423 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.08 vs. limit=5.0 2023-04-27 08:09:18,472 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=87423.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:09:36,463 INFO [optim.py:369] (6/7) 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 08:09:43,846 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 16, batch 1550, loss[loss=0.2084, simple_loss=0.2826, pruned_loss=0.0671, over 4901.00 frames. ], tot_loss[loss=0.182, simple_loss=0.2521, pruned_loss=0.05594, over 952697.63 frames. ], batch size: 46, lr: 3.45e-03, grad_scale: 32.0 2023-04-27 08:09:49,332 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=87470.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:09:57,980 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 2023-04-27 08:10:09,485 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2023-04-27 08:10:30,165 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=87505.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:10:42,326 INFO [finetune.py:976] (6/7) Epoch 16, batch 1600, loss[loss=0.1536, simple_loss=0.2264, pruned_loss=0.0404, over 4767.00 frames. ], tot_loss[loss=0.179, simple_loss=0.2492, pruned_loss=0.05438, over 952251.23 frames. ], batch size: 26, lr: 3.45e-03, grad_scale: 32.0 2023-04-27 08:10:43,626 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=87518.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:10:46,082 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=87522.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:10:54,399 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=87534.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:11:05,883 INFO [optim.py:369] (6/7) 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 08:11:16,207 INFO [finetune.py:976] (6/7) Epoch 16, batch 1650, loss[loss=0.1233, simple_loss=0.2019, pruned_loss=0.02236, over 4935.00 frames. ], tot_loss[loss=0.1765, simple_loss=0.2464, pruned_loss=0.05333, over 954061.29 frames. ], batch size: 33, lr: 3.45e-03, grad_scale: 32.0 2023-04-27 08:11:16,331 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=87566.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 08:11:26,487 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=87582.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:11:28,261 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=87584.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:11:29,410 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=87586.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:11:29,475 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.5603, 1.6686, 1.6780, 2.3596, 2.5538, 2.1087, 1.9783, 1.8223], device='cuda:6'), covar=tensor([0.1807, 0.1920, 0.2638, 0.1587, 0.1438, 0.1949, 0.2598, 0.2600], device='cuda:6'), in_proj_covar=tensor([0.0304, 0.0310, 0.0349, 0.0288, 0.0327, 0.0310, 0.0299, 0.0359], device='cuda:6'), out_proj_covar=tensor([6.2941e-05, 6.4801e-05, 7.4613e-05, 5.8759e-05, 6.8196e-05, 6.5451e-05, 6.3297e-05, 7.6657e-05], device='cuda:6') 2023-04-27 08:11:31,290 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=87589.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:11:41,648 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2023-04-27 08:11:43,112 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=87605.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:11:49,704 INFO [finetune.py:976] (6/7) Epoch 16, batch 1700, loss[loss=0.1633, simple_loss=0.2349, pruned_loss=0.04581, over 4756.00 frames. ], tot_loss[loss=0.1742, simple_loss=0.2436, pruned_loss=0.05242, over 952405.46 frames. ], batch size: 27, lr: 3.45e-03, grad_scale: 32.0 2023-04-27 08:12:02,940 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.3383, 1.9961, 2.1864, 2.6361, 2.5364, 2.1371, 1.8112, 2.3473], device='cuda:6'), covar=tensor([0.0752, 0.1027, 0.0633, 0.0518, 0.0563, 0.0801, 0.0774, 0.0494], device='cuda:6'), in_proj_covar=tensor([0.0190, 0.0203, 0.0184, 0.0173, 0.0177, 0.0183, 0.0154, 0.0180], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 08:12:08,949 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=87645.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 08:12:12,239 INFO [optim.py:369] (6/7) 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 08:12:13,466 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=87650.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:12:18,812 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=87658.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 08:12:19,775 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-04-27 08:12:23,602 INFO [finetune.py:976] (6/7) Epoch 16, batch 1750, loss[loss=0.1471, simple_loss=0.2272, pruned_loss=0.03346, over 4798.00 frames. ], tot_loss[loss=0.1771, simple_loss=0.2463, pruned_loss=0.0539, over 952256.15 frames. ], batch size: 29, lr: 3.45e-03, grad_scale: 32.0 2023-04-27 08:12:23,731 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=87666.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 08:12:26,746 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8558, 3.6996, 0.8910, 1.9305, 2.1489, 2.6048, 2.1659, 1.1484], device='cuda:6'), covar=tensor([0.1215, 0.0896, 0.2174, 0.1246, 0.0973, 0.1033, 0.1479, 0.1844], device='cuda:6'), in_proj_covar=tensor([0.0119, 0.0246, 0.0138, 0.0122, 0.0131, 0.0154, 0.0118, 0.0121], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-27 08:12:55,812 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 16, batch 1800, loss[loss=0.1781, simple_loss=0.2536, pruned_loss=0.05124, over 4232.00 frames. ], tot_loss[loss=0.1801, simple_loss=0.2501, pruned_loss=0.05503, over 951356.07 frames. ], batch size: 65, lr: 3.45e-03, grad_scale: 32.0 2023-04-27 08:13:07,730 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8901, 2.2774, 1.8737, 2.1213, 1.5649, 1.9209, 1.9246, 1.3577], device='cuda:6'), covar=tensor([0.1800, 0.1102, 0.0850, 0.1146, 0.3366, 0.1084, 0.1758, 0.2741], device='cuda:6'), in_proj_covar=tensor([0.0285, 0.0308, 0.0220, 0.0281, 0.0312, 0.0261, 0.0251, 0.0266], device='cuda:6'), out_proj_covar=tensor([1.1465e-04, 1.2270e-04, 8.7674e-05, 1.1159e-04, 1.2707e-04, 1.0413e-04, 1.0141e-04, 1.0568e-04], device='cuda:6') 2023-04-27 08:13:14,249 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=87718.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:13:49,466 INFO [optim.py:369] (6/7) 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 08:14:12,970 INFO [finetune.py:976] (6/7) Epoch 16, batch 1850, loss[loss=0.1646, simple_loss=0.2484, pruned_loss=0.04035, over 4794.00 frames. ], tot_loss[loss=0.1797, simple_loss=0.2504, pruned_loss=0.05449, over 954384.59 frames. ], batch size: 51, lr: 3.45e-03, grad_scale: 32.0 2023-04-27 08:14:20,542 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.1155, 1.2325, 5.0257, 4.6475, 4.3960, 4.7986, 4.4903, 4.4627], device='cuda:6'), covar=tensor([0.6392, 0.6594, 0.1104, 0.1915, 0.1216, 0.1116, 0.1326, 0.1547], device='cuda:6'), in_proj_covar=tensor([0.0311, 0.0303, 0.0401, 0.0405, 0.0347, 0.0405, 0.0308, 0.0363], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-27 08:14:40,487 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.9117, 2.2354, 2.0119, 2.1858, 1.9392, 2.1280, 2.0547, 2.0566], device='cuda:6'), covar=tensor([0.4171, 0.6703, 0.6043, 0.5038, 0.6359, 0.8084, 0.7367, 0.6564], device='cuda:6'), in_proj_covar=tensor([0.0331, 0.0372, 0.0318, 0.0332, 0.0344, 0.0397, 0.0353, 0.0326], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 08:15:07,316 INFO [finetune.py:976] (6/7) Epoch 16, batch 1900, loss[loss=0.1725, simple_loss=0.2586, pruned_loss=0.04323, over 4811.00 frames. ], tot_loss[loss=0.1809, simple_loss=0.2519, pruned_loss=0.05498, over 954276.03 frames. ], batch size: 38, lr: 3.45e-03, grad_scale: 32.0 2023-04-27 08:15:08,021 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=87817.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:15:28,307 INFO [optim.py:369] (6/7) 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 08:15:37,111 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=87861.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 08:15:40,073 INFO [finetune.py:976] (6/7) Epoch 16, batch 1950, loss[loss=0.1701, simple_loss=0.2449, pruned_loss=0.04771, over 4710.00 frames. ], tot_loss[loss=0.18, simple_loss=0.2511, pruned_loss=0.05447, over 956275.14 frames. ], batch size: 54, lr: 3.45e-03, grad_scale: 32.0 2023-04-27 08:15:47,612 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7793, 1.5959, 1.7088, 2.0308, 2.0622, 1.6420, 1.3752, 1.8984], device='cuda:6'), covar=tensor([0.0831, 0.1191, 0.0798, 0.0609, 0.0639, 0.0860, 0.0799, 0.0536], device='cuda:6'), in_proj_covar=tensor([0.0191, 0.0204, 0.0184, 0.0174, 0.0178, 0.0183, 0.0155, 0.0181], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 08:16:10,452 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=87886.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:16:47,482 INFO [finetune.py:976] (6/7) Epoch 16, batch 2000, loss[loss=0.1876, simple_loss=0.2524, pruned_loss=0.06144, over 4901.00 frames. ], tot_loss[loss=0.1793, simple_loss=0.2494, pruned_loss=0.05463, over 956956.89 frames. ], batch size: 32, lr: 3.45e-03, grad_scale: 32.0 2023-04-27 08:16:48,885 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8511, 1.4283, 1.9477, 2.4143, 1.9758, 1.8919, 1.9743, 1.8813], device='cuda:6'), covar=tensor([0.4415, 0.6288, 0.6387, 0.5335, 0.5588, 0.7403, 0.7195, 0.7858], device='cuda:6'), in_proj_covar=tensor([0.0416, 0.0405, 0.0494, 0.0505, 0.0444, 0.0469, 0.0474, 0.0479], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 08:16:55,341 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6023, 3.3869, 1.0675, 1.8716, 1.9491, 2.4458, 2.0117, 1.0836], device='cuda:6'), covar=tensor([0.1368, 0.0916, 0.1916, 0.1325, 0.1147, 0.1051, 0.1480, 0.1741], device='cuda:6'), in_proj_covar=tensor([0.0118, 0.0245, 0.0137, 0.0121, 0.0130, 0.0153, 0.0118, 0.0120], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-27 08:16:56,642 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.5785, 1.8157, 1.9119, 2.0257, 1.8135, 1.9586, 1.9681, 1.9423], device='cuda:6'), covar=tensor([0.4019, 0.5913, 0.5060, 0.4561, 0.5704, 0.7357, 0.5418, 0.5233], device='cuda:6'), in_proj_covar=tensor([0.0330, 0.0373, 0.0319, 0.0333, 0.0344, 0.0396, 0.0354, 0.0326], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 08:16:58,998 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=87934.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:17:02,679 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=87940.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 08:17:06,233 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=87945.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:17:08,485 INFO [optim.py:369] (6/7) 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 08:17:16,741 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7761, 1.7425, 0.7467, 1.4138, 1.7714, 1.6610, 1.4727, 1.5397], device='cuda:6'), covar=tensor([0.0489, 0.0376, 0.0371, 0.0549, 0.0277, 0.0518, 0.0523, 0.0605], device='cuda:6'), in_proj_covar=tensor([0.0029, 0.0024, 0.0022, 0.0029, 0.0020, 0.0028, 0.0028, 0.0029], device='cuda:6'), out_proj_covar=tensor([0.0050, 0.0044, 0.0038, 0.0051, 0.0038, 0.0049, 0.0049, 0.0050], device='cuda:6') 2023-04-27 08:17:17,324 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=87961.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 08:17:21,295 INFO [finetune.py:976] (6/7) Epoch 16, batch 2050, loss[loss=0.1768, simple_loss=0.2429, pruned_loss=0.0553, over 4810.00 frames. ], tot_loss[loss=0.1763, simple_loss=0.2461, pruned_loss=0.05323, over 956614.39 frames. ], batch size: 38, lr: 3.45e-03, grad_scale: 32.0 2023-04-27 08:17:21,998 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.2486, 1.5299, 1.3599, 1.7628, 1.5310, 1.7933, 1.4236, 3.5244], device='cuda:6'), covar=tensor([0.0642, 0.0875, 0.0895, 0.1285, 0.0742, 0.0555, 0.0832, 0.0171], device='cuda:6'), in_proj_covar=tensor([0.0038, 0.0039, 0.0040, 0.0044, 0.0040, 0.0038, 0.0038, 0.0056], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:6') 2023-04-27 08:17:30,006 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.54 vs. limit=2.0 2023-04-27 08:17:41,592 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-04-27 08:17:55,289 INFO [finetune.py:976] (6/7) Epoch 16, batch 2100, loss[loss=0.1788, simple_loss=0.2477, pruned_loss=0.05492, over 4861.00 frames. ], tot_loss[loss=0.1761, simple_loss=0.2455, pruned_loss=0.0533, over 956106.91 frames. ], batch size: 31, lr: 3.45e-03, grad_scale: 32.0 2023-04-27 08:17:57,083 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=88018.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:18:16,279 INFO [optim.py:369] (6/7) 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 08:18:28,510 INFO [finetune.py:976] (6/7) Epoch 16, batch 2150, loss[loss=0.1568, simple_loss=0.2171, pruned_loss=0.04824, over 4223.00 frames. ], tot_loss[loss=0.1788, simple_loss=0.2487, pruned_loss=0.05441, over 954520.00 frames. ], batch size: 17, lr: 3.45e-03, grad_scale: 32.0 2023-04-27 08:18:29,056 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=88066.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:18:50,564 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=88100.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:19:01,084 INFO [finetune.py:976] (6/7) Epoch 16, batch 2200, loss[loss=0.1967, simple_loss=0.2825, pruned_loss=0.05546, over 4842.00 frames. ], tot_loss[loss=0.1802, simple_loss=0.2504, pruned_loss=0.05494, over 953238.95 frames. ], batch size: 49, lr: 3.45e-03, grad_scale: 32.0 2023-04-27 08:19:02,774 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=88117.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:19:34,550 INFO [optim.py:369] (6/7) 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 08:19:47,596 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=88161.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 08:19:47,619 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=88161.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:19:55,584 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 16, batch 2250, loss[loss=0.1586, simple_loss=0.2245, pruned_loss=0.04635, over 4759.00 frames. ], tot_loss[loss=0.1795, simple_loss=0.2503, pruned_loss=0.05434, over 954856.23 frames. ], batch size: 26, lr: 3.45e-03, grad_scale: 32.0 2023-04-27 08:20:10,637 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-04-27 08:20:29,816 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.12 vs. limit=2.0 2023-04-27 08:20:51,261 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=88209.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:21:01,647 INFO [finetune.py:976] (6/7) Epoch 16, batch 2300, loss[loss=0.2015, simple_loss=0.2607, pruned_loss=0.07112, over 4258.00 frames. ], tot_loss[loss=0.1803, simple_loss=0.2511, pruned_loss=0.05472, over 954339.69 frames. ], batch size: 66, lr: 3.44e-03, grad_scale: 32.0 2023-04-27 08:21:07,837 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2023-04-27 08:21:18,879 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=88240.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:21:21,950 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=88245.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:21:24,249 INFO [optim.py:369] (6/7) 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 08:21:32,159 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=88261.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 08:21:41,585 INFO [finetune.py:976] (6/7) Epoch 16, batch 2350, loss[loss=0.1336, simple_loss=0.2125, pruned_loss=0.02736, over 4754.00 frames. ], tot_loss[loss=0.1785, simple_loss=0.2488, pruned_loss=0.05411, over 955253.97 frames. ], batch size: 27, lr: 3.44e-03, grad_scale: 64.0 2023-04-27 08:21:43,259 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.22 vs. limit=5.0 2023-04-27 08:22:14,473 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=88288.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:22:16,073 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.70 vs. limit=2.0 2023-04-27 08:22:16,382 INFO [zipformer.py:1188] (6/7) 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] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=88293.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:22:38,202 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=88309.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:22:48,485 INFO [finetune.py:976] (6/7) Epoch 16, batch 2400, loss[loss=0.1766, simple_loss=0.2338, pruned_loss=0.05968, over 4829.00 frames. ], tot_loss[loss=0.1767, simple_loss=0.2466, pruned_loss=0.05342, over 956222.23 frames. ], batch size: 33, lr: 3.44e-03, grad_scale: 64.0 2023-04-27 08:23:23,049 INFO [optim.py:369] (6/7) 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 08:23:24,125 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-04-27 08:23:25,031 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=88352.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 08:23:33,497 INFO [finetune.py:976] (6/7) Epoch 16, batch 2450, loss[loss=0.176, simple_loss=0.2421, pruned_loss=0.05499, over 4754.00 frames. ], tot_loss[loss=0.1764, simple_loss=0.245, pruned_loss=0.0539, over 955241.41 frames. ], batch size: 27, lr: 3.44e-03, grad_scale: 64.0 2023-04-27 08:23:59,621 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.68 vs. limit=5.0 2023-04-27 08:24:30,565 INFO [finetune.py:976] (6/7) Epoch 16, batch 2500, loss[loss=0.1367, simple_loss=0.2198, pruned_loss=0.02684, over 4788.00 frames. ], tot_loss[loss=0.1788, simple_loss=0.2477, pruned_loss=0.05495, over 956269.84 frames. ], batch size: 29, lr: 3.44e-03, grad_scale: 32.0 2023-04-27 08:24:54,799 INFO [optim.py:369] (6/7) 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 08:24:58,649 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=88456.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:25:04,201 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([3.5327, 3.5597, 2.6907, 4.1454, 3.5558, 3.5288, 1.6046, 3.5178], device='cuda:6'), covar=tensor([0.1901, 0.1274, 0.4187, 0.1561, 0.3470, 0.1969, 0.5708, 0.2538], device='cuda:6'), in_proj_covar=tensor([0.0242, 0.0212, 0.0251, 0.0302, 0.0297, 0.0247, 0.0269, 0.0271], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 08:25:04,747 INFO [finetune.py:976] (6/7) Epoch 16, batch 2550, loss[loss=0.2012, simple_loss=0.2679, pruned_loss=0.06725, over 4790.00 frames. ], tot_loss[loss=0.182, simple_loss=0.2516, pruned_loss=0.05618, over 956433.96 frames. ], batch size: 54, lr: 3.44e-03, grad_scale: 32.0 2023-04-27 08:25:38,576 INFO [finetune.py:976] (6/7) Epoch 16, batch 2600, loss[loss=0.1847, simple_loss=0.2459, pruned_loss=0.06176, over 4862.00 frames. ], tot_loss[loss=0.1835, simple_loss=0.2533, pruned_loss=0.05681, over 953714.40 frames. ], batch size: 31, lr: 3.44e-03, grad_scale: 32.0 2023-04-27 08:25:39,877 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.3004, 2.8896, 0.9019, 1.5713, 2.0748, 1.3260, 3.6970, 1.6003], device='cuda:6'), covar=tensor([0.0687, 0.0838, 0.0909, 0.1151, 0.0497, 0.0958, 0.0221, 0.0641], device='cuda:6'), in_proj_covar=tensor([0.0051, 0.0066, 0.0049, 0.0047, 0.0050, 0.0052, 0.0075, 0.0052], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:6') 2023-04-27 08:26:18,045 INFO [optim.py:369] (6/7) 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] (6/7) Epoch 16, batch 2650, loss[loss=0.1643, simple_loss=0.2329, pruned_loss=0.04786, over 4924.00 frames. ], tot_loss[loss=0.1837, simple_loss=0.2539, pruned_loss=0.05675, over 954859.95 frames. ], batch size: 33, lr: 3.44e-03, grad_scale: 32.0 2023-04-27 08:26:59,838 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.62 vs. limit=2.0 2023-04-27 08:27:29,040 INFO [finetune.py:976] (6/7) Epoch 16, batch 2700, loss[loss=0.1828, simple_loss=0.248, pruned_loss=0.05881, over 4734.00 frames. ], tot_loss[loss=0.1798, simple_loss=0.2501, pruned_loss=0.05481, over 954364.14 frames. ], batch size: 54, lr: 3.44e-03, grad_scale: 32.0 2023-04-27 08:28:09,276 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([0.8626, 1.1841, 3.2863, 2.9887, 2.9631, 3.1919, 3.1531, 2.8766], device='cuda:6'), covar=tensor([0.7367, 0.5448, 0.1573, 0.2429, 0.1511, 0.1892, 0.1855, 0.1848], device='cuda:6'), in_proj_covar=tensor([0.0305, 0.0297, 0.0395, 0.0400, 0.0341, 0.0400, 0.0304, 0.0357], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-27 08:28:11,603 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=88647.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 08:28:13,346 INFO [optim.py:369] (6/7) 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 08:28:36,534 INFO [finetune.py:976] (6/7) Epoch 16, batch 2750, loss[loss=0.1788, simple_loss=0.2445, pruned_loss=0.05656, over 4909.00 frames. ], tot_loss[loss=0.1778, simple_loss=0.2477, pruned_loss=0.05395, over 955960.52 frames. ], batch size: 37, lr: 3.44e-03, grad_scale: 32.0 2023-04-27 08:28:37,258 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.2065, 1.5479, 1.3647, 1.7148, 1.6114, 1.7405, 1.4603, 3.3582], device='cuda:6'), covar=tensor([0.0623, 0.0791, 0.0816, 0.1203, 0.0636, 0.0563, 0.0728, 0.0170], device='cuda:6'), in_proj_covar=tensor([0.0038, 0.0039, 0.0040, 0.0044, 0.0040, 0.0038, 0.0038, 0.0056], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:6') 2023-04-27 08:29:05,434 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.21 vs. limit=5.0 2023-04-27 08:29:06,444 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8730, 1.3097, 1.4361, 1.5713, 1.9748, 1.6488, 1.3312, 1.3387], device='cuda:6'), covar=tensor([0.1248, 0.1365, 0.1808, 0.1210, 0.0770, 0.1393, 0.1928, 0.1996], device='cuda:6'), in_proj_covar=tensor([0.0309, 0.0314, 0.0353, 0.0292, 0.0332, 0.0314, 0.0304, 0.0364], device='cuda:6'), out_proj_covar=tensor([6.4015e-05, 6.5562e-05, 7.5416e-05, 5.9628e-05, 6.9159e-05, 6.6278e-05, 6.4198e-05, 7.7770e-05], device='cuda:6') 2023-04-27 08:29:21,112 INFO [finetune.py:976] (6/7) Epoch 16, batch 2800, loss[loss=0.1646, simple_loss=0.2474, pruned_loss=0.04092, over 4771.00 frames. ], tot_loss[loss=0.176, simple_loss=0.2449, pruned_loss=0.05353, over 954518.40 frames. ], batch size: 28, lr: 3.44e-03, grad_scale: 32.0 2023-04-27 08:29:42,735 INFO [optim.py:369] (6/7) 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 08:29:47,975 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 16, batch 2850, loss[loss=0.216, simple_loss=0.2872, pruned_loss=0.07238, over 4901.00 frames. ], tot_loss[loss=0.1745, simple_loss=0.2436, pruned_loss=0.05272, over 955573.43 frames. ], batch size: 35, lr: 3.44e-03, grad_scale: 32.0 2023-04-27 08:29:59,317 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.0766, 1.0240, 1.1653, 1.2147, 1.0842, 0.9390, 0.8921, 0.4357], device='cuda:6'), covar=tensor([0.0518, 0.0554, 0.0520, 0.0534, 0.0667, 0.1416, 0.0529, 0.0876], device='cuda:6'), in_proj_covar=tensor([0.0067, 0.0070, 0.0069, 0.0068, 0.0075, 0.0096, 0.0075, 0.0068], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-27 08:29:59,933 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.2890, 1.6306, 1.5014, 1.7745, 1.6727, 1.8192, 1.4686, 3.3983], device='cuda:6'), covar=tensor([0.0573, 0.0744, 0.0754, 0.1120, 0.0600, 0.0478, 0.0679, 0.0143], device='cuda:6'), in_proj_covar=tensor([0.0038, 0.0039, 0.0040, 0.0043, 0.0040, 0.0038, 0.0038, 0.0056], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:6') 2023-04-27 08:30:15,460 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.2229, 2.9452, 1.0182, 1.5744, 1.6506, 2.2792, 1.7002, 0.9052], device='cuda:6'), covar=tensor([0.1603, 0.0968, 0.1899, 0.1367, 0.1252, 0.0912, 0.1622, 0.1926], device='cuda:6'), in_proj_covar=tensor([0.0118, 0.0246, 0.0139, 0.0122, 0.0132, 0.0154, 0.0119, 0.0120], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-27 08:30:20,170 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 16, batch 2900, loss[loss=0.1975, simple_loss=0.2716, pruned_loss=0.06168, over 4912.00 frames. ], tot_loss[loss=0.1775, simple_loss=0.2472, pruned_loss=0.05394, over 955075.41 frames. ], batch size: 36, lr: 3.44e-03, grad_scale: 32.0 2023-04-27 08:30:29,767 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0146, 2.6002, 1.1854, 1.4545, 2.2855, 1.2776, 3.3748, 1.6643], device='cuda:6'), covar=tensor([0.0707, 0.0723, 0.0812, 0.1173, 0.0425, 0.0971, 0.0208, 0.0619], device='cuda:6'), in_proj_covar=tensor([0.0052, 0.0066, 0.0049, 0.0046, 0.0050, 0.0052, 0.0076, 0.0052], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:6') 2023-04-27 08:30:49,620 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8990, 2.1844, 1.9720, 2.1292, 1.6026, 1.9552, 2.0344, 1.5132], device='cuda:6'), covar=tensor([0.1558, 0.1067, 0.0728, 0.1076, 0.2827, 0.1081, 0.1530, 0.2331], device='cuda:6'), in_proj_covar=tensor([0.0287, 0.0311, 0.0221, 0.0281, 0.0314, 0.0263, 0.0252, 0.0267], device='cuda:6'), out_proj_covar=tensor([1.1546e-04, 1.2392e-04, 8.8027e-05, 1.1185e-04, 1.2792e-04, 1.0474e-04, 1.0195e-04, 1.0643e-04], device='cuda:6') 2023-04-27 08:30:50,707 INFO [optim.py:369] (6/7) 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 08:30:50,796 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([4.4463, 4.3930, 3.0930, 5.0814, 4.3525, 4.3840, 1.7950, 4.3591], device='cuda:6'), covar=tensor([0.1477, 0.0983, 0.3584, 0.0939, 0.2748, 0.1558, 0.5434, 0.2118], device='cuda:6'), in_proj_covar=tensor([0.0242, 0.0212, 0.0251, 0.0302, 0.0296, 0.0246, 0.0269, 0.0270], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 08:31:02,488 INFO [finetune.py:976] (6/7) Epoch 16, batch 2950, loss[loss=0.1757, simple_loss=0.256, pruned_loss=0.04775, over 4754.00 frames. ], tot_loss[loss=0.1806, simple_loss=0.2504, pruned_loss=0.0554, over 955503.83 frames. ], batch size: 27, lr: 3.44e-03, grad_scale: 32.0 2023-04-27 08:31:03,178 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([3.6708, 3.7317, 2.8005, 4.3542, 3.6831, 3.6845, 1.6657, 3.7765], device='cuda:6'), covar=tensor([0.1811, 0.1137, 0.3476, 0.1435, 0.3354, 0.1798, 0.5663, 0.2290], device='cuda:6'), in_proj_covar=tensor([0.0242, 0.0212, 0.0251, 0.0302, 0.0296, 0.0246, 0.0269, 0.0270], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 08:31:58,219 INFO [finetune.py:976] (6/7) Epoch 16, batch 3000, loss[loss=0.1594, simple_loss=0.2299, pruned_loss=0.04446, over 4778.00 frames. ], tot_loss[loss=0.1818, simple_loss=0.2518, pruned_loss=0.05596, over 956221.39 frames. ], batch size: 28, lr: 3.44e-03, grad_scale: 32.0 2023-04-27 08:31:58,219 INFO [finetune.py:1001] (6/7) Computing validation loss 2023-04-27 08:32:14,558 INFO [finetune.py:1010] (6/7) Epoch 16, validation: loss=0.1523, simple_loss=0.2234, pruned_loss=0.04062, over 2265189.00 frames. 2023-04-27 08:32:14,559 INFO [finetune.py:1011] (6/7) Maximum memory allocated so far is 6365MB 2023-04-27 08:32:30,247 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8305, 2.2694, 0.8986, 1.2137, 1.4858, 1.1371, 2.4745, 1.3641], device='cuda:6'), covar=tensor([0.0727, 0.0553, 0.0702, 0.1284, 0.0493, 0.1042, 0.0382, 0.0700], device='cuda:6'), in_proj_covar=tensor([0.0052, 0.0066, 0.0049, 0.0046, 0.0050, 0.0053, 0.0076, 0.0052], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:6') 2023-04-27 08:32:47,309 INFO [zipformer.py:1188] (6/7) 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] (6/7) 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] (6/7) Epoch 16, batch 3050, loss[loss=0.2157, simple_loss=0.27, pruned_loss=0.08069, over 4169.00 frames. ], tot_loss[loss=0.1831, simple_loss=0.2534, pruned_loss=0.05634, over 956610.36 frames. ], batch size: 65, lr: 3.44e-03, grad_scale: 32.0 2023-04-27 08:33:30,464 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 16, batch 3100, loss[loss=0.1577, simple_loss=0.2302, pruned_loss=0.04264, over 4849.00 frames. ], tot_loss[loss=0.1806, simple_loss=0.2504, pruned_loss=0.05536, over 954839.96 frames. ], batch size: 49, lr: 3.44e-03, grad_scale: 32.0 2023-04-27 08:34:05,003 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6566, 1.4704, 0.7436, 1.2990, 1.6205, 1.5384, 1.4023, 1.4499], device='cuda:6'), covar=tensor([0.0501, 0.0393, 0.0364, 0.0587, 0.0276, 0.0547, 0.0508, 0.0580], device='cuda:6'), in_proj_covar=tensor([0.0028, 0.0024, 0.0022, 0.0029, 0.0020, 0.0028, 0.0028, 0.0028], device='cuda:6'), out_proj_covar=tensor([0.0049, 0.0044, 0.0037, 0.0050, 0.0038, 0.0049, 0.0049, 0.0050], device='cuda:6') 2023-04-27 08:34:13,218 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=89027.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:34:38,479 INFO [optim.py:369] (6/7) 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 08:34:39,244 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7878, 1.4069, 1.8644, 2.3558, 1.9442, 1.7723, 1.9126, 1.8411], device='cuda:6'), covar=tensor([0.4994, 0.7000, 0.6875, 0.5835, 0.6534, 0.8634, 0.8714, 0.8837], device='cuda:6'), in_proj_covar=tensor([0.0417, 0.0407, 0.0493, 0.0507, 0.0447, 0.0472, 0.0477, 0.0480], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 08:34:59,019 INFO [finetune.py:976] (6/7) Epoch 16, batch 3150, loss[loss=0.1921, simple_loss=0.2668, pruned_loss=0.05867, over 4826.00 frames. ], tot_loss[loss=0.1784, simple_loss=0.2476, pruned_loss=0.05463, over 955818.63 frames. ], batch size: 40, lr: 3.44e-03, grad_scale: 32.0 2023-04-27 08:35:02,293 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.63 vs. limit=2.0 2023-04-27 08:35:14,726 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=89088.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:35:21,955 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5130, 1.3637, 1.6847, 1.7489, 1.4081, 1.2292, 1.4511, 0.8930], device='cuda:6'), covar=tensor([0.0576, 0.0706, 0.0395, 0.0667, 0.0736, 0.1124, 0.0587, 0.0650], device='cuda:6'), in_proj_covar=tensor([0.0069, 0.0071, 0.0070, 0.0069, 0.0077, 0.0098, 0.0076, 0.0069], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-27 08:35:37,805 INFO [finetune.py:976] (6/7) Epoch 16, batch 3200, loss[loss=0.1592, simple_loss=0.2313, pruned_loss=0.04358, over 4785.00 frames. ], tot_loss[loss=0.1756, simple_loss=0.2444, pruned_loss=0.05336, over 956400.01 frames. ], batch size: 29, lr: 3.44e-03, grad_scale: 32.0 2023-04-27 08:36:23,957 INFO [optim.py:369] (6/7) 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 08:36:44,941 INFO [finetune.py:976] (6/7) Epoch 16, batch 3250, loss[loss=0.2024, simple_loss=0.2725, pruned_loss=0.06616, over 4824.00 frames. ], tot_loss[loss=0.1755, simple_loss=0.2446, pruned_loss=0.05321, over 955726.68 frames. ], batch size: 33, lr: 3.44e-03, grad_scale: 32.0 2023-04-27 08:36:46,285 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([0.6433, 1.7590, 1.7846, 1.3213, 1.9174, 1.5461, 2.3119, 1.5448], device='cuda:6'), covar=tensor([0.3932, 0.1816, 0.4361, 0.3044, 0.1449, 0.2432, 0.1417, 0.4476], device='cuda:6'), in_proj_covar=tensor([0.0344, 0.0348, 0.0430, 0.0357, 0.0385, 0.0383, 0.0372, 0.0421], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 08:36:46,915 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=89169.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:37:35,633 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0623, 1.5436, 1.9340, 2.4245, 1.8890, 1.5137, 1.2339, 1.6516], device='cuda:6'), covar=tensor([0.3018, 0.3136, 0.1604, 0.1986, 0.2495, 0.2671, 0.4337, 0.2131], device='cuda:6'), in_proj_covar=tensor([0.0287, 0.0244, 0.0222, 0.0311, 0.0215, 0.0229, 0.0226, 0.0184], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:6') 2023-04-27 08:37:46,006 INFO [finetune.py:976] (6/7) Epoch 16, batch 3300, loss[loss=0.1888, simple_loss=0.2722, pruned_loss=0.05273, over 4804.00 frames. ], tot_loss[loss=0.178, simple_loss=0.2479, pruned_loss=0.05403, over 956240.18 frames. ], batch size: 45, lr: 3.44e-03, grad_scale: 32.0 2023-04-27 08:38:07,676 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=89230.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:38:21,745 INFO [optim.py:369] (6/7) 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 08:38:43,926 INFO [finetune.py:976] (6/7) Epoch 16, batch 3350, loss[loss=0.1481, simple_loss=0.227, pruned_loss=0.03458, over 4775.00 frames. ], tot_loss[loss=0.181, simple_loss=0.2515, pruned_loss=0.05529, over 956633.99 frames. ], batch size: 29, lr: 3.44e-03, grad_scale: 32.0 2023-04-27 08:38:54,319 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=3.95 vs. limit=5.0 2023-04-27 08:39:17,175 INFO [finetune.py:976] (6/7) Epoch 16, batch 3400, loss[loss=0.1998, simple_loss=0.2786, pruned_loss=0.06048, over 4845.00 frames. ], tot_loss[loss=0.1826, simple_loss=0.2528, pruned_loss=0.05623, over 955089.10 frames. ], batch size: 44, lr: 3.44e-03, grad_scale: 32.0 2023-04-27 08:39:31,341 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.76 vs. limit=2.0 2023-04-27 08:39:40,163 INFO [optim.py:369] (6/7) 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 08:39:53,109 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0082, 1.2598, 5.0579, 4.7041, 4.4534, 4.8644, 4.4849, 4.5385], device='cuda:6'), covar=tensor([0.6652, 0.6117, 0.0915, 0.1751, 0.1001, 0.1095, 0.1336, 0.1297], device='cuda:6'), in_proj_covar=tensor([0.0306, 0.0300, 0.0397, 0.0399, 0.0342, 0.0402, 0.0305, 0.0357], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-27 08:39:53,663 INFO [finetune.py:976] (6/7) Epoch 16, batch 3450, loss[loss=0.1464, simple_loss=0.2253, pruned_loss=0.03379, over 4898.00 frames. ], tot_loss[loss=0.1822, simple_loss=0.2529, pruned_loss=0.05574, over 956478.13 frames. ], batch size: 36, lr: 3.43e-03, grad_scale: 32.0 2023-04-27 08:40:04,279 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=89374.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:40:16,199 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=89383.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:40:54,853 INFO [finetune.py:976] (6/7) Epoch 16, batch 3500, loss[loss=0.1698, simple_loss=0.2327, pruned_loss=0.05341, over 4320.00 frames. ], tot_loss[loss=0.1797, simple_loss=0.2501, pruned_loss=0.0546, over 956774.23 frames. ], batch size: 19, lr: 3.43e-03, grad_scale: 32.0 2023-04-27 08:41:00,466 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=89425.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:41:07,516 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=89435.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:41:09,970 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=89439.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 08:41:18,075 INFO [optim.py:369] (6/7) 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 08:41:18,398 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.58 vs. limit=5.0 2023-04-27 08:41:22,508 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5726, 1.4761, 1.7401, 1.8787, 1.4644, 1.2683, 1.5545, 0.9740], device='cuda:6'), covar=tensor([0.0591, 0.0737, 0.0447, 0.0727, 0.0779, 0.1194, 0.0592, 0.0660], device='cuda:6'), in_proj_covar=tensor([0.0069, 0.0071, 0.0070, 0.0069, 0.0077, 0.0098, 0.0076, 0.0069], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-27 08:41:34,093 INFO [finetune.py:976] (6/7) Epoch 16, batch 3550, loss[loss=0.1669, simple_loss=0.2406, pruned_loss=0.04656, over 4752.00 frames. ], tot_loss[loss=0.1776, simple_loss=0.2474, pruned_loss=0.05393, over 956094.96 frames. ], batch size: 23, lr: 3.43e-03, grad_scale: 32.0 2023-04-27 08:41:58,277 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=89486.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:42:06,477 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.1830, 1.6221, 2.0530, 2.3652, 1.9367, 1.5855, 1.1933, 1.7663], device='cuda:6'), covar=tensor([0.3208, 0.3258, 0.1588, 0.2236, 0.2734, 0.2726, 0.4346, 0.2137], device='cuda:6'), in_proj_covar=tensor([0.0288, 0.0245, 0.0223, 0.0312, 0.0216, 0.0229, 0.0227, 0.0185], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:6') 2023-04-27 08:42:08,153 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=89500.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 08:42:20,193 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=89509.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:42:29,611 INFO [finetune.py:976] (6/7) Epoch 16, batch 3600, loss[loss=0.1531, simple_loss=0.2275, pruned_loss=0.03937, over 4847.00 frames. ], tot_loss[loss=0.1738, simple_loss=0.2435, pruned_loss=0.05207, over 955987.07 frames. ], batch size: 49, lr: 3.43e-03, grad_scale: 32.0 2023-04-27 08:42:41,051 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=89525.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:43:13,816 INFO [optim.py:369] (6/7) 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 08:43:36,162 INFO [finetune.py:976] (6/7) Epoch 16, batch 3650, loss[loss=0.1915, simple_loss=0.2759, pruned_loss=0.05358, over 4826.00 frames. ], tot_loss[loss=0.1767, simple_loss=0.2467, pruned_loss=0.05338, over 957104.98 frames. ], batch size: 47, lr: 3.43e-03, grad_scale: 32.0 2023-04-27 08:43:38,733 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=89570.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:43:48,406 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.9677, 2.9242, 2.3864, 3.3156, 2.9436, 2.9168, 1.1681, 2.9155], device='cuda:6'), covar=tensor([0.2052, 0.1614, 0.2755, 0.2564, 0.3314, 0.2268, 0.5851, 0.2643], device='cuda:6'), in_proj_covar=tensor([0.0247, 0.0214, 0.0254, 0.0306, 0.0300, 0.0250, 0.0274, 0.0274], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 08:43:48,464 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.0676, 1.0585, 1.2056, 1.2173, 1.0645, 0.8865, 1.0809, 0.7269], device='cuda:6'), covar=tensor([0.0529, 0.0541, 0.0507, 0.0451, 0.0611, 0.1258, 0.0446, 0.0749], device='cuda:6'), in_proj_covar=tensor([0.0069, 0.0071, 0.0070, 0.0068, 0.0077, 0.0098, 0.0076, 0.0069], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-27 08:43:57,871 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6971, 3.2148, 1.1303, 1.9131, 1.9342, 2.5497, 1.9523, 1.2557], device='cuda:6'), covar=tensor([0.1127, 0.0843, 0.1724, 0.1111, 0.0941, 0.0810, 0.1302, 0.1983], device='cuda:6'), in_proj_covar=tensor([0.0118, 0.0247, 0.0139, 0.0122, 0.0132, 0.0154, 0.0119, 0.0121], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-27 08:44:42,833 INFO [finetune.py:976] (6/7) Epoch 16, batch 3700, loss[loss=0.1646, simple_loss=0.2462, pruned_loss=0.04148, over 4908.00 frames. ], tot_loss[loss=0.179, simple_loss=0.2495, pruned_loss=0.05425, over 954711.60 frames. ], batch size: 36, lr: 3.43e-03, grad_scale: 32.0 2023-04-27 08:45:26,133 INFO [optim.py:369] (6/7) 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 08:45:48,269 INFO [finetune.py:976] (6/7) Epoch 16, batch 3750, loss[loss=0.1654, simple_loss=0.2141, pruned_loss=0.05839, over 4022.00 frames. ], tot_loss[loss=0.1802, simple_loss=0.251, pruned_loss=0.05468, over 954690.39 frames. ], batch size: 17, lr: 3.43e-03, grad_scale: 32.0 2023-04-27 08:46:02,535 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=89680.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:46:10,288 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=89683.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:46:39,696 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.75 vs. limit=2.0 2023-04-27 08:46:49,774 INFO [finetune.py:976] (6/7) Epoch 16, batch 3800, loss[loss=0.2193, simple_loss=0.2931, pruned_loss=0.07277, over 4100.00 frames. ], tot_loss[loss=0.1801, simple_loss=0.2513, pruned_loss=0.05447, over 954806.83 frames. ], batch size: 65, lr: 3.43e-03, grad_scale: 32.0 2023-04-27 08:47:03,122 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.72 vs. limit=2.0 2023-04-27 08:47:09,444 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=89730.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:47:10,029 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=89731.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:47:13,144 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5702, 1.4055, 0.5610, 1.2493, 1.4516, 1.4682, 1.2963, 1.3577], device='cuda:6'), covar=tensor([0.0514, 0.0414, 0.0403, 0.0583, 0.0292, 0.0547, 0.0535, 0.0591], device='cuda:6'), in_proj_covar=tensor([0.0029, 0.0024, 0.0022, 0.0029, 0.0020, 0.0028, 0.0028, 0.0028], device='cuda:6'), out_proj_covar=tensor([0.0050, 0.0044, 0.0038, 0.0051, 0.0038, 0.0049, 0.0049, 0.0050], device='cuda:6') 2023-04-27 08:47:14,494 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.98 vs. limit=2.0 2023-04-27 08:47:21,770 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=89741.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:47:31,965 INFO [optim.py:369] (6/7) 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 08:47:43,651 INFO [finetune.py:976] (6/7) Epoch 16, batch 3850, loss[loss=0.2042, simple_loss=0.2642, pruned_loss=0.07211, over 4711.00 frames. ], tot_loss[loss=0.1789, simple_loss=0.2496, pruned_loss=0.05413, over 953088.71 frames. ], batch size: 23, lr: 3.43e-03, grad_scale: 32.0 2023-04-27 08:47:58,306 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=89781.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:48:12,384 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=89795.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 08:48:26,721 INFO [finetune.py:976] (6/7) Epoch 16, batch 3900, loss[loss=0.1738, simple_loss=0.2547, pruned_loss=0.04639, over 4828.00 frames. ], tot_loss[loss=0.1795, simple_loss=0.2493, pruned_loss=0.05491, over 955915.48 frames. ], batch size: 33, lr: 3.43e-03, grad_scale: 32.0 2023-04-27 08:48:33,315 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=89825.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:48:46,574 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.3729, 1.8448, 2.2606, 2.8178, 2.2295, 1.7850, 1.6013, 1.9411], device='cuda:6'), covar=tensor([0.3628, 0.3874, 0.1904, 0.2540, 0.3101, 0.3040, 0.4367, 0.2438], device='cuda:6'), in_proj_covar=tensor([0.0290, 0.0248, 0.0225, 0.0315, 0.0217, 0.0230, 0.0229, 0.0187], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:6') 2023-04-27 08:48:48,866 INFO [optim.py:369] (6/7) 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 08:48:59,087 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=89865.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:48:59,120 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=89865.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:48:59,622 INFO [finetune.py:976] (6/7) Epoch 16, batch 3950, loss[loss=0.1326, simple_loss=0.2125, pruned_loss=0.0264, over 4851.00 frames. ], tot_loss[loss=0.1766, simple_loss=0.2462, pruned_loss=0.05353, over 956159.53 frames. ], batch size: 47, lr: 3.43e-03, grad_scale: 32.0 2023-04-27 08:49:05,867 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=89873.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:49:27,732 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.3560, 3.0995, 2.6036, 2.8476, 2.2146, 2.6397, 2.8600, 2.0455], device='cuda:6'), covar=tensor([0.2259, 0.1194, 0.0870, 0.1278, 0.3067, 0.1415, 0.1723, 0.2896], device='cuda:6'), in_proj_covar=tensor([0.0288, 0.0311, 0.0223, 0.0282, 0.0314, 0.0264, 0.0252, 0.0267], device='cuda:6'), out_proj_covar=tensor([1.1587e-04, 1.2387e-04, 8.8797e-05, 1.1202e-04, 1.2781e-04, 1.0524e-04, 1.0206e-04, 1.0645e-04], device='cuda:6') 2023-04-27 08:49:33,475 INFO [finetune.py:976] (6/7) Epoch 16, batch 4000, loss[loss=0.2082, simple_loss=0.2746, pruned_loss=0.07093, over 4897.00 frames. ], tot_loss[loss=0.1762, simple_loss=0.2457, pruned_loss=0.05333, over 956835.70 frames. ], batch size: 35, lr: 3.43e-03, grad_scale: 32.0 2023-04-27 08:49:47,304 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=89926.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 08:50:11,838 INFO [optim.py:369] (6/7) 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 08:50:33,228 INFO [finetune.py:976] (6/7) Epoch 16, batch 4050, loss[loss=0.1882, simple_loss=0.2527, pruned_loss=0.06189, over 4082.00 frames. ], tot_loss[loss=0.1796, simple_loss=0.2489, pruned_loss=0.05511, over 954803.97 frames. ], batch size: 17, lr: 3.43e-03, grad_scale: 32.0 2023-04-27 08:51:40,298 INFO [finetune.py:976] (6/7) Epoch 16, batch 4100, loss[loss=0.1922, simple_loss=0.2665, pruned_loss=0.05896, over 4905.00 frames. ], tot_loss[loss=0.18, simple_loss=0.2499, pruned_loss=0.05505, over 952494.95 frames. ], batch size: 37, lr: 3.43e-03, grad_scale: 32.0 2023-04-27 08:52:03,174 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=90030.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:52:12,525 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=90036.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:52:14,364 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8260, 2.2308, 0.8804, 1.2217, 1.4422, 1.2405, 2.4863, 1.3639], device='cuda:6'), covar=tensor([0.0668, 0.0536, 0.0657, 0.1350, 0.0496, 0.1014, 0.0340, 0.0706], device='cuda:6'), in_proj_covar=tensor([0.0051, 0.0065, 0.0048, 0.0046, 0.0050, 0.0052, 0.0074, 0.0051], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0007, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:6') 2023-04-27 08:52:27,087 INFO [optim.py:369] (6/7) 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 08:52:48,658 INFO [finetune.py:976] (6/7) Epoch 16, batch 4150, loss[loss=0.1925, simple_loss=0.2522, pruned_loss=0.06641, over 4803.00 frames. ], tot_loss[loss=0.1814, simple_loss=0.2516, pruned_loss=0.05563, over 954046.43 frames. ], batch size: 45, lr: 3.43e-03, grad_scale: 32.0 2023-04-27 08:53:06,918 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=90078.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:53:09,310 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=90081.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:53:11,931 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.4378, 3.3865, 1.0281, 1.8654, 1.8405, 2.5322, 1.8556, 1.0550], device='cuda:6'), covar=tensor([0.1370, 0.0810, 0.1875, 0.1199, 0.1114, 0.0897, 0.1518, 0.2070], device='cuda:6'), in_proj_covar=tensor([0.0119, 0.0247, 0.0139, 0.0122, 0.0133, 0.0154, 0.0119, 0.0122], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-27 08:53:26,050 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=90095.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 08:53:50,595 INFO [finetune.py:976] (6/7) Epoch 16, batch 4200, loss[loss=0.1428, simple_loss=0.2327, pruned_loss=0.02648, over 4767.00 frames. ], tot_loss[loss=0.1811, simple_loss=0.252, pruned_loss=0.05515, over 955743.84 frames. ], batch size: 28, lr: 3.43e-03, grad_scale: 32.0 2023-04-27 08:53:59,901 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=90129.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:54:01,065 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=90130.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:54:14,880 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=90143.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 08:54:24,334 INFO [optim.py:369] (6/7) 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 08:54:27,612 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=3.89 vs. limit=5.0 2023-04-27 08:54:34,103 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=90165.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:54:34,612 INFO [finetune.py:976] (6/7) Epoch 16, batch 4250, loss[loss=0.2407, simple_loss=0.2913, pruned_loss=0.09501, over 4753.00 frames. ], tot_loss[loss=0.1789, simple_loss=0.2495, pruned_loss=0.05413, over 956261.97 frames. ], batch size: 27, lr: 3.43e-03, grad_scale: 32.0 2023-04-27 08:54:52,890 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=90191.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:55:06,210 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 16, batch 4300, loss[loss=0.1342, simple_loss=0.2121, pruned_loss=0.02816, over 4853.00 frames. ], tot_loss[loss=0.1777, simple_loss=0.2477, pruned_loss=0.05379, over 955637.81 frames. ], batch size: 44, lr: 3.43e-03, grad_scale: 32.0 2023-04-27 08:55:11,530 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=90221.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 08:55:19,263 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=90232.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:55:32,139 INFO [optim.py:369] (6/7) 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 08:55:41,853 INFO [finetune.py:976] (6/7) Epoch 16, batch 4350, loss[loss=0.2152, simple_loss=0.2751, pruned_loss=0.07768, over 4731.00 frames. ], tot_loss[loss=0.1771, simple_loss=0.246, pruned_loss=0.05414, over 956249.28 frames. ], batch size: 59, lr: 3.43e-03, grad_scale: 32.0 2023-04-27 08:55:41,928 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.8820, 2.8204, 2.2550, 3.3189, 2.8986, 2.8892, 1.1501, 2.7840], device='cuda:6'), covar=tensor([0.2208, 0.1772, 0.3191, 0.2764, 0.3800, 0.2341, 0.5769, 0.2979], device='cuda:6'), in_proj_covar=tensor([0.0245, 0.0213, 0.0252, 0.0303, 0.0299, 0.0249, 0.0272, 0.0273], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 08:55:53,154 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9356, 1.6734, 3.9715, 3.7521, 3.5213, 3.5819, 3.5337, 3.5575], device='cuda:6'), covar=tensor([0.6117, 0.4686, 0.0979, 0.1451, 0.0981, 0.2006, 0.3201, 0.1289], device='cuda:6'), in_proj_covar=tensor([0.0306, 0.0300, 0.0397, 0.0397, 0.0342, 0.0402, 0.0304, 0.0358], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-27 08:56:00,614 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=90293.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:56:15,531 INFO [finetune.py:976] (6/7) Epoch 16, batch 4400, loss[loss=0.1606, simple_loss=0.2246, pruned_loss=0.04833, over 4712.00 frames. ], tot_loss[loss=0.1782, simple_loss=0.247, pruned_loss=0.0547, over 954751.37 frames. ], batch size: 23, lr: 3.43e-03, grad_scale: 32.0 2023-04-27 08:56:34,957 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=90336.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:56:55,344 INFO [optim.py:369] (6/7) 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 08:56:56,111 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=90351.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:57:06,782 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8976, 1.7259, 2.0973, 2.3961, 1.9782, 1.8197, 1.9635, 1.9438], device='cuda:6'), covar=tensor([0.5390, 0.7740, 0.7942, 0.6395, 0.6449, 0.9378, 1.0184, 1.0707], device='cuda:6'), in_proj_covar=tensor([0.0419, 0.0405, 0.0493, 0.0506, 0.0446, 0.0471, 0.0476, 0.0480], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 08:57:10,219 INFO [finetune.py:976] (6/7) Epoch 16, batch 4450, loss[loss=0.2238, simple_loss=0.2978, pruned_loss=0.0749, over 4839.00 frames. ], tot_loss[loss=0.181, simple_loss=0.2508, pruned_loss=0.05559, over 954626.31 frames. ], batch size: 47, lr: 3.43e-03, grad_scale: 32.0 2023-04-27 08:57:33,655 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=90384.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:58:04,790 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 16, batch 4500, loss[loss=0.2036, simple_loss=0.2806, pruned_loss=0.06327, over 4850.00 frames. ], tot_loss[loss=0.1814, simple_loss=0.2516, pruned_loss=0.05558, over 955132.07 frames. ], batch size: 44, lr: 3.43e-03, grad_scale: 64.0 2023-04-27 08:58:18,451 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.1092, 1.5663, 1.3851, 1.6858, 1.6548, 1.9551, 1.4169, 3.6031], device='cuda:6'), covar=tensor([0.0622, 0.0794, 0.0803, 0.1246, 0.0633, 0.0502, 0.0764, 0.0139], device='cuda:6'), in_proj_covar=tensor([0.0038, 0.0039, 0.0040, 0.0044, 0.0040, 0.0038, 0.0039, 0.0057], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0013, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:6') 2023-04-27 08:58:40,121 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.15 vs. limit=5.0 2023-04-27 08:58:52,082 INFO [optim.py:369] (6/7) 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 08:59:00,289 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8360, 1.3835, 1.3306, 1.6403, 2.0169, 1.6164, 1.3556, 1.2189], device='cuda:6'), covar=tensor([0.1603, 0.1725, 0.2018, 0.1419, 0.0828, 0.1593, 0.2312, 0.2403], device='cuda:6'), in_proj_covar=tensor([0.0308, 0.0312, 0.0353, 0.0292, 0.0329, 0.0313, 0.0303, 0.0366], device='cuda:6'), out_proj_covar=tensor([6.3812e-05, 6.5138e-05, 7.5329e-05, 5.9442e-05, 6.8455e-05, 6.5891e-05, 6.4023e-05, 7.8362e-05], device='cuda:6') 2023-04-27 08:59:04,432 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=90459.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:59:14,178 INFO [finetune.py:976] (6/7) Epoch 16, batch 4550, loss[loss=0.1699, simple_loss=0.2382, pruned_loss=0.05083, over 4848.00 frames. ], tot_loss[loss=0.1816, simple_loss=0.2525, pruned_loss=0.05538, over 957192.21 frames. ], batch size: 25, lr: 3.43e-03, grad_scale: 64.0 2023-04-27 08:59:35,899 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5177, 1.6153, 1.3576, 1.0797, 1.1183, 1.1362, 1.3991, 1.0712], device='cuda:6'), covar=tensor([0.1850, 0.1480, 0.1677, 0.1859, 0.2427, 0.2135, 0.1092, 0.2089], device='cuda:6'), in_proj_covar=tensor([0.0196, 0.0213, 0.0169, 0.0204, 0.0201, 0.0184, 0.0156, 0.0185], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:6') 2023-04-27 08:59:43,241 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=90486.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 08:59:55,957 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=90496.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:00:15,724 INFO [finetune.py:976] (6/7) Epoch 16, batch 4600, loss[loss=0.1839, simple_loss=0.2493, pruned_loss=0.05922, over 4746.00 frames. ], tot_loss[loss=0.1812, simple_loss=0.2521, pruned_loss=0.05514, over 957783.45 frames. ], batch size: 59, lr: 3.43e-03, grad_scale: 64.0 2023-04-27 09:00:18,329 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=90520.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:00:18,920 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=90521.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 09:00:37,736 INFO [optim.py:369] (6/7) 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 09:00:43,609 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=90557.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:00:49,349 INFO [finetune.py:976] (6/7) Epoch 16, batch 4650, loss[loss=0.147, simple_loss=0.2165, pruned_loss=0.03878, over 4816.00 frames. ], tot_loss[loss=0.179, simple_loss=0.2488, pruned_loss=0.05459, over 957788.63 frames. ], batch size: 40, lr: 3.42e-03, grad_scale: 64.0 2023-04-27 09:00:51,266 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=90569.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:01:03,251 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=90588.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:01:23,261 INFO [finetune.py:976] (6/7) Epoch 16, batch 4700, loss[loss=0.1706, simple_loss=0.241, pruned_loss=0.05011, over 4702.00 frames. ], tot_loss[loss=0.177, simple_loss=0.2463, pruned_loss=0.05384, over 955634.38 frames. ], batch size: 23, lr: 3.42e-03, grad_scale: 64.0 2023-04-27 09:01:36,021 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.3131, 2.9901, 0.9761, 1.5974, 1.7365, 2.2135, 1.7708, 0.9679], device='cuda:6'), covar=tensor([0.1473, 0.0912, 0.1818, 0.1306, 0.1118, 0.0920, 0.1512, 0.1909], device='cuda:6'), in_proj_covar=tensor([0.0117, 0.0244, 0.0137, 0.0121, 0.0131, 0.0153, 0.0118, 0.0120], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-27 09:01:45,414 INFO [optim.py:369] (6/7) 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 09:02:01,940 INFO [finetune.py:976] (6/7) Epoch 16, batch 4750, loss[loss=0.2114, simple_loss=0.2717, pruned_loss=0.07561, over 4927.00 frames. ], tot_loss[loss=0.1748, simple_loss=0.2435, pruned_loss=0.05302, over 954653.06 frames. ], batch size: 38, lr: 3.42e-03, grad_scale: 32.0 2023-04-27 09:02:57,065 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 16, batch 4800, loss[loss=0.1765, simple_loss=0.2339, pruned_loss=0.05958, over 4709.00 frames. ], tot_loss[loss=0.1768, simple_loss=0.2458, pruned_loss=0.05392, over 954492.80 frames. ], batch size: 23, lr: 3.42e-03, grad_scale: 32.0 2023-04-27 09:03:36,199 INFO [optim.py:369] (6/7) 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 09:03:47,210 INFO [finetune.py:976] (6/7) Epoch 16, batch 4850, loss[loss=0.2394, simple_loss=0.2979, pruned_loss=0.09041, over 4061.00 frames. ], tot_loss[loss=0.179, simple_loss=0.2487, pruned_loss=0.05466, over 952996.58 frames. ], batch size: 66, lr: 3.42e-03, grad_scale: 32.0 2023-04-27 09:04:00,343 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=90786.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:04:36,235 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 16, batch 4900, loss[loss=0.2094, simple_loss=0.2828, pruned_loss=0.06801, over 4155.00 frames. ], tot_loss[loss=0.1802, simple_loss=0.2502, pruned_loss=0.05513, over 952326.85 frames. ], batch size: 65, lr: 3.42e-03, grad_scale: 32.0 2023-04-27 09:05:01,145 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=90834.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:05:21,463 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=90849.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:05:22,606 INFO [optim.py:369] (6/7) 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 09:05:23,822 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=90852.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:05:38,748 INFO [finetune.py:976] (6/7) Epoch 16, batch 4950, loss[loss=0.1765, simple_loss=0.2521, pruned_loss=0.05044, over 4900.00 frames. ], tot_loss[loss=0.1816, simple_loss=0.2519, pruned_loss=0.05565, over 952158.52 frames. ], batch size: 37, lr: 3.42e-03, grad_scale: 32.0 2023-04-27 09:05:39,487 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8309, 1.4998, 1.4448, 1.5563, 2.0423, 1.6134, 1.2815, 1.3876], device='cuda:6'), covar=tensor([0.1589, 0.1305, 0.1971, 0.1529, 0.0783, 0.1497, 0.2005, 0.2171], device='cuda:6'), in_proj_covar=tensor([0.0302, 0.0307, 0.0347, 0.0285, 0.0323, 0.0307, 0.0298, 0.0361], device='cuda:6'), out_proj_covar=tensor([6.2487e-05, 6.3897e-05, 7.4004e-05, 5.8032e-05, 6.7340e-05, 6.4773e-05, 6.2898e-05, 7.7140e-05], device='cuda:6') 2023-04-27 09:05:54,240 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=90888.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:06:08,377 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=90910.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:06:12,377 INFO [finetune.py:976] (6/7) Epoch 16, batch 5000, loss[loss=0.1675, simple_loss=0.2418, pruned_loss=0.04661, over 4814.00 frames. ], tot_loss[loss=0.1822, simple_loss=0.2519, pruned_loss=0.05625, over 953709.81 frames. ], batch size: 30, lr: 3.42e-03, grad_scale: 32.0 2023-04-27 09:06:27,004 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=90936.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:06:36,100 INFO [optim.py:369] (6/7) 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] (6/7) Epoch 16, batch 5050, loss[loss=0.1384, simple_loss=0.2111, pruned_loss=0.03287, over 4866.00 frames. ], tot_loss[loss=0.1803, simple_loss=0.2495, pruned_loss=0.0556, over 955911.48 frames. ], batch size: 34, lr: 3.42e-03, grad_scale: 32.0 2023-04-27 09:07:13,891 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=91007.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:07:19,798 INFO [finetune.py:976] (6/7) Epoch 16, batch 5100, loss[loss=0.1426, simple_loss=0.2136, pruned_loss=0.03579, over 4712.00 frames. ], tot_loss[loss=0.1761, simple_loss=0.245, pruned_loss=0.05365, over 954823.27 frames. ], batch size: 23, lr: 3.42e-03, grad_scale: 32.0 2023-04-27 09:07:25,201 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.8839, 2.9040, 2.2729, 3.3309, 2.8886, 2.9314, 1.0363, 2.8111], device='cuda:6'), covar=tensor([0.2205, 0.1585, 0.3226, 0.3023, 0.3796, 0.2164, 0.6399, 0.3000], device='cuda:6'), in_proj_covar=tensor([0.0244, 0.0211, 0.0250, 0.0304, 0.0295, 0.0247, 0.0271, 0.0270], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 09:07:43,901 INFO [optim.py:369] (6/7) 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 09:07:46,435 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=91055.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:07:53,572 INFO [finetune.py:976] (6/7) Epoch 16, batch 5150, loss[loss=0.2298, simple_loss=0.3001, pruned_loss=0.07975, over 4907.00 frames. ], tot_loss[loss=0.1746, simple_loss=0.2436, pruned_loss=0.05285, over 953521.15 frames. ], batch size: 43, lr: 3.42e-03, grad_scale: 32.0 2023-04-27 09:07:59,046 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6912, 2.4286, 1.5923, 1.7006, 1.2334, 1.2693, 1.6520, 1.1760], device='cuda:6'), covar=tensor([0.1957, 0.1391, 0.1773, 0.1940, 0.2660, 0.2347, 0.1156, 0.2274], device='cuda:6'), in_proj_covar=tensor([0.0197, 0.0214, 0.0169, 0.0205, 0.0202, 0.0185, 0.0157, 0.0186], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:6') 2023-04-27 09:08:48,457 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 16, batch 5200, loss[loss=0.1703, simple_loss=0.2452, pruned_loss=0.04767, over 4787.00 frames. ], tot_loss[loss=0.1782, simple_loss=0.2476, pruned_loss=0.05436, over 952875.01 frames. ], batch size: 29, lr: 3.42e-03, grad_scale: 32.0 2023-04-27 09:09:38,603 INFO [optim.py:369] (6/7) 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 09:09:39,318 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=91152.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:09:51,505 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=91163.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:09:53,285 INFO [finetune.py:976] (6/7) Epoch 16, batch 5250, loss[loss=0.2303, simple_loss=0.2947, pruned_loss=0.08298, over 4903.00 frames. ], tot_loss[loss=0.1803, simple_loss=0.2505, pruned_loss=0.05503, over 953895.26 frames. ], batch size: 37, lr: 3.42e-03, grad_scale: 32.0 2023-04-27 09:09:53,465 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9378, 1.6445, 2.1475, 2.4028, 2.0089, 1.8289, 1.9941, 1.9268], device='cuda:6'), covar=tensor([0.5141, 0.7780, 0.7370, 0.6361, 0.6614, 0.9287, 0.9301, 1.0529], device='cuda:6'), in_proj_covar=tensor([0.0420, 0.0406, 0.0494, 0.0506, 0.0448, 0.0472, 0.0478, 0.0483], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 09:10:33,361 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6880, 1.4551, 1.9680, 1.9522, 1.4311, 1.2995, 1.5784, 1.1040], device='cuda:6'), covar=tensor([0.0528, 0.0865, 0.0374, 0.0620, 0.0810, 0.1182, 0.0665, 0.0700], device='cuda:6'), in_proj_covar=tensor([0.0067, 0.0070, 0.0069, 0.0068, 0.0075, 0.0097, 0.0075, 0.0067], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-27 09:10:36,886 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.0597, 1.0412, 1.1813, 1.1998, 1.0126, 0.9374, 0.9672, 0.6046], device='cuda:6'), covar=tensor([0.0537, 0.0574, 0.0472, 0.0527, 0.0807, 0.1228, 0.0521, 0.0744], device='cuda:6'), in_proj_covar=tensor([0.0067, 0.0070, 0.0069, 0.0068, 0.0075, 0.0097, 0.0075, 0.0067], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-27 09:10:44,541 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=91200.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:10:47,648 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=91205.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:11:00,425 INFO [finetune.py:976] (6/7) Epoch 16, batch 5300, loss[loss=0.1662, simple_loss=0.2395, pruned_loss=0.04647, over 4712.00 frames. ], tot_loss[loss=0.1815, simple_loss=0.252, pruned_loss=0.05547, over 955041.99 frames. ], batch size: 54, lr: 3.42e-03, grad_scale: 32.0 2023-04-27 09:11:25,281 INFO [optim.py:369] (6/7) 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 09:11:34,434 INFO [finetune.py:976] (6/7) Epoch 16, batch 5350, loss[loss=0.1378, simple_loss=0.2139, pruned_loss=0.0308, over 4762.00 frames. ], tot_loss[loss=0.1815, simple_loss=0.2519, pruned_loss=0.05558, over 952347.40 frames. ], batch size: 28, lr: 3.42e-03, grad_scale: 32.0 2023-04-27 09:11:47,130 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 16, batch 5400, loss[loss=0.1384, simple_loss=0.2267, pruned_loss=0.02509, over 4814.00 frames. ], tot_loss[loss=0.1802, simple_loss=0.2502, pruned_loss=0.05512, over 952929.21 frames. ], batch size: 38, lr: 3.42e-03, grad_scale: 32.0 2023-04-27 09:12:28,681 INFO [zipformer.py:1188] (6/7) 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] (6/7) 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 09:12:42,145 INFO [finetune.py:976] (6/7) Epoch 16, batch 5450, loss[loss=0.1874, simple_loss=0.2526, pruned_loss=0.06107, over 4022.00 frames. ], tot_loss[loss=0.178, simple_loss=0.2472, pruned_loss=0.05435, over 951666.05 frames. ], batch size: 17, lr: 3.42e-03, grad_scale: 32.0 2023-04-27 09:12:50,731 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.29 vs. limit=5.0 2023-04-27 09:12:51,864 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.1950, 1.8768, 2.1506, 2.5680, 2.4808, 2.0493, 1.7287, 2.2695], device='cuda:6'), covar=tensor([0.0783, 0.0974, 0.0605, 0.0504, 0.0579, 0.0919, 0.0786, 0.0556], device='cuda:6'), in_proj_covar=tensor([0.0191, 0.0203, 0.0183, 0.0174, 0.0179, 0.0184, 0.0154, 0.0182], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 09:13:13,759 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.98 vs. limit=2.0 2023-04-27 09:13:15,973 INFO [finetune.py:976] (6/7) Epoch 16, batch 5500, loss[loss=0.14, simple_loss=0.2181, pruned_loss=0.031, over 4768.00 frames. ], tot_loss[loss=0.1747, simple_loss=0.2438, pruned_loss=0.05276, over 954181.44 frames. ], batch size: 26, lr: 3.42e-03, grad_scale: 32.0 2023-04-27 09:13:38,283 INFO [optim.py:369] (6/7) 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 09:13:49,968 INFO [finetune.py:976] (6/7) Epoch 16, batch 5550, loss[loss=0.2303, simple_loss=0.3042, pruned_loss=0.07822, over 4843.00 frames. ], tot_loss[loss=0.1761, simple_loss=0.2457, pruned_loss=0.0533, over 954930.08 frames. ], batch size: 49, lr: 3.42e-03, grad_scale: 32.0 2023-04-27 09:13:59,333 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8290, 1.2920, 1.8769, 2.3120, 1.9131, 1.7746, 1.8091, 1.8151], device='cuda:6'), covar=tensor([0.5176, 0.7054, 0.6632, 0.6510, 0.6705, 0.8373, 0.8477, 0.8396], device='cuda:6'), in_proj_covar=tensor([0.0419, 0.0407, 0.0493, 0.0504, 0.0447, 0.0470, 0.0476, 0.0483], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 09:14:20,555 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=91505.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:14:32,820 INFO [finetune.py:976] (6/7) Epoch 16, batch 5600, loss[loss=0.1964, simple_loss=0.2673, pruned_loss=0.06277, over 4855.00 frames. ], tot_loss[loss=0.1783, simple_loss=0.2487, pruned_loss=0.054, over 953629.92 frames. ], batch size: 31, lr: 3.42e-03, grad_scale: 32.0 2023-04-27 09:14:32,881 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([3.9235, 3.8899, 2.7839, 4.5540, 3.9754, 3.8378, 1.8246, 3.9032], device='cuda:6'), covar=tensor([0.1515, 0.1019, 0.3186, 0.1190, 0.3213, 0.1688, 0.5558, 0.2090], device='cuda:6'), in_proj_covar=tensor([0.0245, 0.0212, 0.0252, 0.0304, 0.0297, 0.0248, 0.0272, 0.0271], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 09:15:04,819 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=91542.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 09:15:15,394 INFO [optim.py:369] (6/7) 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 09:15:16,632 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=91553.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:15:18,466 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6999, 2.2955, 1.6989, 1.5679, 1.2674, 1.2677, 1.8348, 1.2149], device='cuda:6'), covar=tensor([0.1599, 0.1286, 0.1386, 0.1761, 0.2301, 0.1887, 0.0912, 0.2038], device='cuda:6'), in_proj_covar=tensor([0.0197, 0.0214, 0.0170, 0.0205, 0.0202, 0.0186, 0.0156, 0.0186], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:6') 2023-04-27 09:15:29,905 INFO [finetune.py:976] (6/7) Epoch 16, batch 5650, loss[loss=0.1656, simple_loss=0.2559, pruned_loss=0.03761, over 4771.00 frames. ], tot_loss[loss=0.1811, simple_loss=0.2523, pruned_loss=0.05495, over 952713.97 frames. ], batch size: 54, lr: 3.42e-03, grad_scale: 32.0 2023-04-27 09:15:51,520 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-04-27 09:16:20,696 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=91603.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 09:16:33,753 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6084, 1.0187, 1.4526, 1.5566, 1.4774, 1.5813, 1.4553, 1.4787], device='cuda:6'), covar=tensor([0.3139, 0.4445, 0.3380, 0.3827, 0.4637, 0.5743, 0.3916, 0.3686], device='cuda:6'), in_proj_covar=tensor([0.0331, 0.0371, 0.0317, 0.0332, 0.0344, 0.0395, 0.0352, 0.0323], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 09:16:34,226 INFO [finetune.py:976] (6/7) Epoch 16, batch 5700, loss[loss=0.1421, simple_loss=0.1946, pruned_loss=0.04484, over 3874.00 frames. ], tot_loss[loss=0.1784, simple_loss=0.2483, pruned_loss=0.0543, over 934092.26 frames. ], batch size: 16, lr: 3.42e-03, grad_scale: 32.0 2023-04-27 09:17:06,022 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=91641.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:17:19,302 INFO [finetune.py:976] (6/7) Epoch 17, batch 0, loss[loss=0.2065, simple_loss=0.276, pruned_loss=0.06848, over 4890.00 frames. ], tot_loss[loss=0.2065, simple_loss=0.276, pruned_loss=0.06848, over 4890.00 frames. ], batch size: 37, lr: 3.41e-03, grad_scale: 32.0 2023-04-27 09:17:19,303 INFO [finetune.py:1001] (6/7) Computing validation loss 2023-04-27 09:17:22,552 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8452, 1.0739, 1.7197, 2.3026, 1.9409, 1.7628, 1.7481, 1.7394], device='cuda:6'), covar=tensor([0.5023, 0.7332, 0.7136, 0.6499, 0.7012, 0.8520, 0.8625, 0.9000], device='cuda:6'), in_proj_covar=tensor([0.0420, 0.0408, 0.0496, 0.0506, 0.0449, 0.0471, 0.0479, 0.0485], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 09:17:22,616 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.3871, 1.3752, 3.9100, 3.6017, 3.4959, 3.7296, 3.8074, 3.4602], device='cuda:6'), covar=tensor([0.7161, 0.5209, 0.1258, 0.2119, 0.1188, 0.1494, 0.0793, 0.1576], device='cuda:6'), in_proj_covar=tensor([0.0308, 0.0303, 0.0400, 0.0400, 0.0345, 0.0408, 0.0307, 0.0361], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-27 09:17:23,135 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5020, 1.3605, 1.7383, 1.7177, 1.3863, 1.2884, 1.3875, 0.9328], device='cuda:6'), covar=tensor([0.0555, 0.0674, 0.0423, 0.0473, 0.0748, 0.1114, 0.0559, 0.0600], device='cuda:6'), in_proj_covar=tensor([0.0068, 0.0070, 0.0069, 0.0068, 0.0075, 0.0097, 0.0075, 0.0067], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-27 09:17:40,814 INFO [finetune.py:1010] (6/7) 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] (6/7) Maximum memory allocated so far is 6365MB 2023-04-27 09:17:45,698 INFO [optim.py:369] (6/7) 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 09:18:20,578 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.63 vs. limit=2.0 2023-04-27 09:18:24,712 INFO [finetune.py:976] (6/7) Epoch 17, batch 50, loss[loss=0.1748, simple_loss=0.2512, pruned_loss=0.04917, over 4897.00 frames. ], tot_loss[loss=0.1839, simple_loss=0.2543, pruned_loss=0.05679, over 216490.98 frames. ], batch size: 36, lr: 3.41e-03, grad_scale: 32.0 2023-04-27 09:18:42,092 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.46 vs. limit=5.0 2023-04-27 09:18:50,718 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.70 vs. limit=5.0 2023-04-27 09:18:57,085 INFO [finetune.py:976] (6/7) Epoch 17, batch 100, loss[loss=0.1377, simple_loss=0.2141, pruned_loss=0.03064, over 4766.00 frames. ], tot_loss[loss=0.1778, simple_loss=0.2463, pruned_loss=0.05463, over 381869.40 frames. ], batch size: 27, lr: 3.41e-03, grad_scale: 32.0 2023-04-27 09:18:59,525 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.1379, 2.1144, 1.8085, 1.8268, 2.2059, 1.6027, 2.6125, 1.5636], device='cuda:6'), covar=tensor([0.3707, 0.1945, 0.4851, 0.3075, 0.1645, 0.2852, 0.1510, 0.4517], device='cuda:6'), in_proj_covar=tensor([0.0342, 0.0348, 0.0431, 0.0358, 0.0384, 0.0382, 0.0374, 0.0422], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 09:19:02,005 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.4392, 1.7369, 1.8178, 1.9620, 1.8149, 1.9399, 1.8495, 1.8413], device='cuda:6'), covar=tensor([0.3474, 0.5161, 0.4399, 0.4081, 0.5329, 0.6787, 0.5395, 0.4784], device='cuda:6'), in_proj_covar=tensor([0.0332, 0.0372, 0.0317, 0.0333, 0.0344, 0.0395, 0.0353, 0.0324], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 09:19:02,438 INFO [optim.py:369] (6/7) 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 09:19:30,055 INFO [finetune.py:976] (6/7) Epoch 17, batch 150, loss[loss=0.1596, simple_loss=0.2269, pruned_loss=0.04613, over 4675.00 frames. ], tot_loss[loss=0.1742, simple_loss=0.2422, pruned_loss=0.05306, over 508689.58 frames. ], batch size: 23, lr: 3.41e-03, grad_scale: 32.0 2023-04-27 09:19:42,235 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.2867, 2.8122, 2.1710, 2.1645, 1.6805, 1.6276, 2.3117, 1.5045], device='cuda:6'), covar=tensor([0.1650, 0.1406, 0.1397, 0.1626, 0.2295, 0.1900, 0.0952, 0.1976], device='cuda:6'), in_proj_covar=tensor([0.0197, 0.0214, 0.0170, 0.0205, 0.0202, 0.0186, 0.0157, 0.0186], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:6') 2023-04-27 09:20:03,533 INFO [finetune.py:976] (6/7) Epoch 17, batch 200, loss[loss=0.2087, simple_loss=0.2601, pruned_loss=0.07862, over 4193.00 frames. ], tot_loss[loss=0.174, simple_loss=0.242, pruned_loss=0.053, over 608301.21 frames. ], batch size: 65, lr: 3.41e-03, grad_scale: 32.0 2023-04-27 09:20:08,958 INFO [optim.py:369] (6/7) 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 09:20:35,899 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=91891.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:20:36,998 INFO [finetune.py:976] (6/7) Epoch 17, batch 250, loss[loss=0.2133, simple_loss=0.2641, pruned_loss=0.08129, over 4896.00 frames. ], tot_loss[loss=0.1762, simple_loss=0.2446, pruned_loss=0.05384, over 685307.16 frames. ], batch size: 32, lr: 3.41e-03, grad_scale: 32.0 2023-04-27 09:20:40,162 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=91898.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 09:20:56,904 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-04-27 09:21:08,453 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=91941.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:21:09,223 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 2023-04-27 09:21:10,055 INFO [finetune.py:976] (6/7) Epoch 17, batch 300, loss[loss=0.1956, simple_loss=0.2652, pruned_loss=0.06294, over 4737.00 frames. ], tot_loss[loss=0.1772, simple_loss=0.2468, pruned_loss=0.05382, over 744950.96 frames. ], batch size: 59, lr: 3.41e-03, grad_scale: 32.0 2023-04-27 09:21:15,846 INFO [optim.py:369] (6/7) 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 09:21:16,638 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=91952.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:21:17,243 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([3.0755, 2.1830, 2.1771, 2.8697, 2.9988, 2.6158, 2.5502, 2.2379], device='cuda:6'), covar=tensor([0.1535, 0.1548, 0.1411, 0.1468, 0.0854, 0.1274, 0.1670, 0.1557], device='cuda:6'), in_proj_covar=tensor([0.0303, 0.0309, 0.0347, 0.0286, 0.0323, 0.0308, 0.0298, 0.0360], device='cuda:6'), out_proj_covar=tensor([6.2627e-05, 6.4407e-05, 7.4042e-05, 5.8197e-05, 6.7055e-05, 6.4836e-05, 6.2912e-05, 7.7004e-05], device='cuda:6') 2023-04-27 09:21:18,433 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=91955.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:21:44,421 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8482, 1.1864, 1.4075, 1.5686, 1.5278, 1.6027, 1.4870, 1.4680], device='cuda:6'), covar=tensor([0.3089, 0.3810, 0.3610, 0.3271, 0.4175, 0.5831, 0.3646, 0.3728], device='cuda:6'), in_proj_covar=tensor([0.0331, 0.0372, 0.0318, 0.0332, 0.0345, 0.0395, 0.0353, 0.0323], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 09:21:51,277 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=91981.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 09:22:01,153 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=91989.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:22:04,087 INFO [finetune.py:976] (6/7) Epoch 17, batch 350, loss[loss=0.2255, simple_loss=0.2882, pruned_loss=0.08139, over 4174.00 frames. ], tot_loss[loss=0.1798, simple_loss=0.2495, pruned_loss=0.05504, over 789893.97 frames. ], batch size: 65, lr: 3.41e-03, grad_scale: 32.0 2023-04-27 09:22:06,542 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5283, 1.0055, 0.5010, 1.2001, 1.1393, 1.4200, 1.3036, 1.2519], device='cuda:6'), covar=tensor([0.0501, 0.0429, 0.0402, 0.0575, 0.0298, 0.0530, 0.0514, 0.0617], device='cuda:6'), in_proj_covar=tensor([0.0028, 0.0024, 0.0022, 0.0029, 0.0020, 0.0028, 0.0028, 0.0028], device='cuda:6'), out_proj_covar=tensor([0.0050, 0.0044, 0.0038, 0.0051, 0.0038, 0.0049, 0.0049, 0.0050], device='cuda:6') 2023-04-27 09:22:36,907 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=92016.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:22:38,218 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-04-27 09:23:09,164 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=92042.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 09:23:09,647 INFO [finetune.py:976] (6/7) Epoch 17, batch 400, loss[loss=0.1751, simple_loss=0.2322, pruned_loss=0.05902, over 4692.00 frames. ], tot_loss[loss=0.1792, simple_loss=0.2499, pruned_loss=0.05429, over 826415.98 frames. ], batch size: 23, lr: 3.41e-03, grad_scale: 32.0 2023-04-27 09:23:21,384 INFO [optim.py:369] (6/7) 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 09:23:51,453 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-04-27 09:24:03,171 INFO [finetune.py:976] (6/7) Epoch 17, batch 450, loss[loss=0.1842, simple_loss=0.2567, pruned_loss=0.05587, over 4728.00 frames. ], tot_loss[loss=0.179, simple_loss=0.2492, pruned_loss=0.05443, over 853964.39 frames. ], batch size: 59, lr: 3.41e-03, grad_scale: 32.0 2023-04-27 09:24:11,148 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0225, 2.4634, 0.9658, 1.2434, 1.6004, 1.1912, 2.9599, 1.5346], device='cuda:6'), covar=tensor([0.0839, 0.0683, 0.0937, 0.1740, 0.0680, 0.1355, 0.0382, 0.0957], device='cuda:6'), in_proj_covar=tensor([0.0050, 0.0065, 0.0048, 0.0046, 0.0049, 0.0051, 0.0074, 0.0051], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0007, 0.0008, 0.0008, 0.0010, 0.0007], device='cuda:6') 2023-04-27 09:24:14,777 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.1326, 2.7046, 1.9927, 2.2154, 1.5457, 1.5002, 2.2462, 1.3951], device='cuda:6'), covar=tensor([0.1674, 0.1544, 0.1457, 0.1641, 0.2296, 0.1923, 0.0923, 0.2049], device='cuda:6'), in_proj_covar=tensor([0.0197, 0.0215, 0.0170, 0.0206, 0.0203, 0.0186, 0.0157, 0.0187], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:6') 2023-04-27 09:24:32,044 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7222, 1.4929, 4.2402, 3.9814, 3.7098, 3.9750, 3.8679, 3.7017], device='cuda:6'), covar=tensor([0.7249, 0.5798, 0.1150, 0.1832, 0.1178, 0.1516, 0.2120, 0.1679], device='cuda:6'), in_proj_covar=tensor([0.0309, 0.0305, 0.0400, 0.0403, 0.0347, 0.0408, 0.0308, 0.0364], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-27 09:24:36,858 INFO [finetune.py:976] (6/7) Epoch 17, batch 500, loss[loss=0.173, simple_loss=0.2454, pruned_loss=0.05032, over 4898.00 frames. ], tot_loss[loss=0.1776, simple_loss=0.2474, pruned_loss=0.05387, over 877360.44 frames. ], batch size: 36, lr: 3.41e-03, grad_scale: 32.0 2023-04-27 09:24:42,153 INFO [optim.py:369] (6/7) 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 09:24:52,363 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0933, 1.6209, 1.9932, 2.2056, 1.9396, 1.5433, 1.1709, 1.6769], device='cuda:6'), covar=tensor([0.3411, 0.3184, 0.1604, 0.2112, 0.2440, 0.2787, 0.4229, 0.2051], device='cuda:6'), in_proj_covar=tensor([0.0291, 0.0246, 0.0226, 0.0314, 0.0216, 0.0231, 0.0229, 0.0186], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:6') 2023-04-27 09:25:10,286 INFO [finetune.py:976] (6/7) Epoch 17, batch 550, loss[loss=0.1717, simple_loss=0.251, pruned_loss=0.04616, over 4806.00 frames. ], tot_loss[loss=0.1762, simple_loss=0.245, pruned_loss=0.05367, over 893696.88 frames. ], batch size: 45, lr: 3.41e-03, grad_scale: 32.0 2023-04-27 09:25:13,459 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=92198.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 09:25:35,053 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7822, 1.3749, 1.9083, 2.3590, 1.9461, 1.8018, 1.8820, 1.8305], device='cuda:6'), covar=tensor([0.4416, 0.6169, 0.6195, 0.5487, 0.5631, 0.7483, 0.7413, 0.7544], device='cuda:6'), in_proj_covar=tensor([0.0420, 0.0408, 0.0496, 0.0507, 0.0449, 0.0472, 0.0478, 0.0483], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 09:25:44,127 INFO [finetune.py:976] (6/7) Epoch 17, batch 600, loss[loss=0.2125, simple_loss=0.2804, pruned_loss=0.07233, over 4831.00 frames. ], tot_loss[loss=0.1771, simple_loss=0.2459, pruned_loss=0.05419, over 907413.74 frames. ], batch size: 49, lr: 3.41e-03, grad_scale: 32.0 2023-04-27 09:25:46,056 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=92246.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 09:25:46,682 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=92247.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:25:49,028 INFO [optim.py:369] (6/7) 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] (6/7) Epoch 17, batch 650, loss[loss=0.1637, simple_loss=0.2505, pruned_loss=0.03841, over 4788.00 frames. ], tot_loss[loss=0.1805, simple_loss=0.2504, pruned_loss=0.05536, over 917948.24 frames. ], batch size: 29, lr: 3.41e-03, grad_scale: 32.0 2023-04-27 09:26:29,964 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=92311.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:26:33,991 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0504, 1.4435, 1.9105, 2.1527, 1.8094, 1.4757, 1.0814, 1.5941], device='cuda:6'), covar=tensor([0.3031, 0.3381, 0.1604, 0.2208, 0.2507, 0.2641, 0.4318, 0.2077], device='cuda:6'), in_proj_covar=tensor([0.0291, 0.0247, 0.0226, 0.0315, 0.0217, 0.0231, 0.0229, 0.0186], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:6') 2023-04-27 09:26:56,499 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=92337.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 09:27:00,061 INFO [finetune.py:976] (6/7) Epoch 17, batch 700, loss[loss=0.1671, simple_loss=0.241, pruned_loss=0.04659, over 4778.00 frames. ], tot_loss[loss=0.1806, simple_loss=0.2514, pruned_loss=0.05494, over 926394.62 frames. ], batch size: 29, lr: 3.41e-03, grad_scale: 32.0 2023-04-27 09:27:10,674 INFO [optim.py:369] (6/7) 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 09:27:38,758 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7730, 1.8087, 0.8259, 1.4123, 1.9772, 1.6387, 1.4992, 1.5985], device='cuda:6'), covar=tensor([0.0485, 0.0362, 0.0335, 0.0546, 0.0253, 0.0518, 0.0503, 0.0565], device='cuda:6'), in_proj_covar=tensor([0.0029, 0.0024, 0.0022, 0.0029, 0.0020, 0.0028, 0.0028, 0.0028], device='cuda:6'), out_proj_covar=tensor([0.0050, 0.0045, 0.0038, 0.0051, 0.0038, 0.0049, 0.0049, 0.0050], device='cuda:6') 2023-04-27 09:27:49,356 INFO [finetune.py:976] (6/7) Epoch 17, batch 750, loss[loss=0.1934, simple_loss=0.2659, pruned_loss=0.06046, over 4744.00 frames. ], tot_loss[loss=0.1807, simple_loss=0.2522, pruned_loss=0.05462, over 934203.21 frames. ], batch size: 54, lr: 3.41e-03, grad_scale: 32.0 2023-04-27 09:28:05,801 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7001, 2.0435, 1.6122, 1.4094, 1.2630, 1.2830, 1.6337, 1.2043], device='cuda:6'), covar=tensor([0.1858, 0.1400, 0.1563, 0.1864, 0.2542, 0.2127, 0.1109, 0.2140], device='cuda:6'), in_proj_covar=tensor([0.0198, 0.0215, 0.0170, 0.0206, 0.0203, 0.0186, 0.0157, 0.0187], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:6') 2023-04-27 09:28:16,733 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.2115, 1.7136, 2.2037, 2.3306, 2.1417, 1.7371, 1.1786, 1.8523], device='cuda:6'), covar=tensor([0.3316, 0.3137, 0.1492, 0.2450, 0.2384, 0.2659, 0.4176, 0.2050], device='cuda:6'), in_proj_covar=tensor([0.0294, 0.0249, 0.0228, 0.0318, 0.0218, 0.0233, 0.0231, 0.0188], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:6') 2023-04-27 09:28:44,566 INFO [finetune.py:976] (6/7) Epoch 17, batch 800, loss[loss=0.1839, simple_loss=0.2464, pruned_loss=0.06069, over 4918.00 frames. ], tot_loss[loss=0.1801, simple_loss=0.2514, pruned_loss=0.05442, over 940874.16 frames. ], batch size: 33, lr: 3.41e-03, grad_scale: 32.0 2023-04-27 09:28:54,788 INFO [optim.py:369] (6/7) 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 09:29:07,043 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.4562, 1.3472, 1.7960, 1.7213, 1.3424, 1.1895, 1.4684, 0.9346], device='cuda:6'), covar=tensor([0.0615, 0.0620, 0.0387, 0.0575, 0.0697, 0.1093, 0.0572, 0.0608], device='cuda:6'), in_proj_covar=tensor([0.0067, 0.0069, 0.0068, 0.0067, 0.0075, 0.0096, 0.0074, 0.0067], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-27 09:29:31,910 INFO [finetune.py:976] (6/7) Epoch 17, batch 850, loss[loss=0.1585, simple_loss=0.2356, pruned_loss=0.04075, over 4726.00 frames. ], tot_loss[loss=0.1789, simple_loss=0.2498, pruned_loss=0.05403, over 940473.64 frames. ], batch size: 54, lr: 3.41e-03, grad_scale: 32.0 2023-04-27 09:29:41,206 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0066, 2.4591, 2.0146, 1.7870, 1.4531, 1.5174, 2.0130, 1.3731], device='cuda:6'), covar=tensor([0.1495, 0.1344, 0.1352, 0.1695, 0.2240, 0.1976, 0.1008, 0.1956], device='cuda:6'), in_proj_covar=tensor([0.0198, 0.0215, 0.0170, 0.0206, 0.0203, 0.0186, 0.0157, 0.0186], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:6') 2023-04-27 09:30:05,438 INFO [finetune.py:976] (6/7) Epoch 17, batch 900, loss[loss=0.1365, simple_loss=0.2021, pruned_loss=0.03541, over 4917.00 frames. ], tot_loss[loss=0.1771, simple_loss=0.2473, pruned_loss=0.05352, over 943763.65 frames. ], batch size: 32, lr: 3.41e-03, grad_scale: 32.0 2023-04-27 09:30:07,969 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=92547.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:30:10,296 INFO [optim.py:369] (6/7) 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 09:30:19,690 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.3453, 1.1980, 1.4471, 1.5586, 1.2504, 1.1251, 1.2900, 0.7860], device='cuda:6'), covar=tensor([0.0393, 0.0522, 0.0426, 0.0369, 0.0467, 0.0879, 0.0390, 0.0549], device='cuda:6'), in_proj_covar=tensor([0.0068, 0.0070, 0.0069, 0.0068, 0.0075, 0.0096, 0.0075, 0.0068], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-27 09:30:26,009 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-04-27 09:30:38,496 INFO [finetune.py:976] (6/7) Epoch 17, batch 950, loss[loss=0.2246, simple_loss=0.2952, pruned_loss=0.07706, over 4734.00 frames. ], tot_loss[loss=0.1766, simple_loss=0.2463, pruned_loss=0.05344, over 944875.08 frames. ], batch size: 59, lr: 3.40e-03, grad_scale: 32.0 2023-04-27 09:30:39,829 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=92595.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:30:49,732 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=92611.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:30:59,802 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=92625.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:31:07,667 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=92637.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 09:31:12,292 INFO [finetune.py:976] (6/7) Epoch 17, batch 1000, loss[loss=0.238, simple_loss=0.3166, pruned_loss=0.07972, over 4803.00 frames. ], tot_loss[loss=0.1781, simple_loss=0.2482, pruned_loss=0.05403, over 943813.93 frames. ], batch size: 51, lr: 3.40e-03, grad_scale: 32.0 2023-04-27 09:31:17,226 INFO [optim.py:369] (6/7) 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 09:31:22,263 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=92659.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:31:40,696 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=92685.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 09:31:41,332 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=92686.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:31:45,966 INFO [finetune.py:976] (6/7) Epoch 17, batch 1050, loss[loss=0.1917, simple_loss=0.2592, pruned_loss=0.06213, over 4912.00 frames. ], tot_loss[loss=0.181, simple_loss=0.2519, pruned_loss=0.05509, over 948626.25 frames. ], batch size: 43, lr: 3.40e-03, grad_scale: 64.0 2023-04-27 09:32:03,095 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=92720.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:32:21,428 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([0.8395, 1.9052, 1.7676, 1.6577, 2.0305, 1.6389, 2.5677, 1.6001], device='cuda:6'), covar=tensor([0.3530, 0.1932, 0.4422, 0.2612, 0.1487, 0.2423, 0.1258, 0.4282], device='cuda:6'), in_proj_covar=tensor([0.0346, 0.0352, 0.0436, 0.0361, 0.0389, 0.0386, 0.0376, 0.0429], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 09:32:30,324 INFO [finetune.py:976] (6/7) Epoch 17, batch 1100, loss[loss=0.1632, simple_loss=0.2418, pruned_loss=0.04227, over 4809.00 frames. ], tot_loss[loss=0.1815, simple_loss=0.2525, pruned_loss=0.05524, over 948701.30 frames. ], batch size: 40, lr: 3.40e-03, grad_scale: 64.0 2023-04-27 09:32:36,188 INFO [optim.py:369] (6/7) 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 09:32:50,457 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2023-04-27 09:33:12,031 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=92781.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:33:32,083 INFO [finetune.py:976] (6/7) Epoch 17, batch 1150, loss[loss=0.231, simple_loss=0.3001, pruned_loss=0.08096, over 4231.00 frames. ], tot_loss[loss=0.1806, simple_loss=0.2517, pruned_loss=0.05476, over 950819.53 frames. ], batch size: 65, lr: 3.40e-03, grad_scale: 32.0 2023-04-27 09:33:45,155 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6028, 1.6839, 0.7732, 1.2387, 1.9512, 1.4636, 1.3469, 1.4316], device='cuda:6'), covar=tensor([0.0524, 0.0379, 0.0378, 0.0581, 0.0275, 0.0527, 0.0548, 0.0570], device='cuda:6'), in_proj_covar=tensor([0.0029, 0.0024, 0.0023, 0.0029, 0.0020, 0.0028, 0.0028, 0.0029], device='cuda:6'), out_proj_covar=tensor([0.0050, 0.0045, 0.0038, 0.0051, 0.0038, 0.0049, 0.0049, 0.0051], device='cuda:6') 2023-04-27 09:34:18,997 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.8713, 2.1020, 2.0996, 2.2443, 2.0132, 2.2156, 2.1406, 2.1024], device='cuda:6'), covar=tensor([0.4287, 0.7178, 0.5590, 0.5057, 0.5837, 0.7522, 0.6826, 0.6163], device='cuda:6'), in_proj_covar=tensor([0.0331, 0.0371, 0.0316, 0.0332, 0.0344, 0.0395, 0.0354, 0.0323], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 09:34:20,046 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=92831.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:34:39,530 INFO [finetune.py:976] (6/7) Epoch 17, batch 1200, loss[loss=0.1672, simple_loss=0.2337, pruned_loss=0.05034, over 4892.00 frames. ], tot_loss[loss=0.1779, simple_loss=0.2487, pruned_loss=0.05358, over 949608.70 frames. ], batch size: 35, lr: 3.40e-03, grad_scale: 32.0 2023-04-27 09:34:50,264 INFO [optim.py:369] (6/7) 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 09:35:24,734 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8084, 1.3301, 4.6030, 4.3316, 4.0288, 4.3763, 4.3293, 4.0208], device='cuda:6'), covar=tensor([0.6688, 0.6364, 0.1064, 0.1753, 0.1058, 0.2554, 0.1180, 0.1509], device='cuda:6'), in_proj_covar=tensor([0.0308, 0.0304, 0.0400, 0.0402, 0.0346, 0.0406, 0.0308, 0.0362], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-27 09:35:25,400 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.3305, 2.9102, 2.0851, 2.3364, 1.6277, 1.6427, 2.2818, 1.4799], device='cuda:6'), covar=tensor([0.1568, 0.1340, 0.1482, 0.1609, 0.2234, 0.1913, 0.0950, 0.2044], device='cuda:6'), in_proj_covar=tensor([0.0196, 0.0212, 0.0168, 0.0204, 0.0201, 0.0184, 0.0156, 0.0185], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:6') 2023-04-27 09:35:44,786 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=92892.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:35:45,252 INFO [finetune.py:976] (6/7) Epoch 17, batch 1250, loss[loss=0.2303, simple_loss=0.2735, pruned_loss=0.0936, over 4885.00 frames. ], tot_loss[loss=0.1766, simple_loss=0.2468, pruned_loss=0.0532, over 952378.33 frames. ], batch size: 35, lr: 3.40e-03, grad_scale: 32.0 2023-04-27 09:35:48,999 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=92899.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:36:25,931 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=92931.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:36:45,901 INFO [finetune.py:976] (6/7) Epoch 17, batch 1300, loss[loss=0.1832, simple_loss=0.2392, pruned_loss=0.0636, over 4901.00 frames. ], tot_loss[loss=0.1752, simple_loss=0.2447, pruned_loss=0.0528, over 953740.17 frames. ], batch size: 32, lr: 3.40e-03, grad_scale: 32.0 2023-04-27 09:36:57,404 INFO [optim.py:369] (6/7) 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 09:36:59,301 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5880, 3.3395, 1.0244, 2.0010, 2.0411, 2.3358, 2.0036, 0.9270], device='cuda:6'), covar=tensor([0.1246, 0.0718, 0.1769, 0.1116, 0.0930, 0.0980, 0.1309, 0.2038], device='cuda:6'), in_proj_covar=tensor([0.0118, 0.0243, 0.0137, 0.0121, 0.0132, 0.0153, 0.0118, 0.0120], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-27 09:37:08,810 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7604, 1.9486, 0.9311, 1.4626, 2.0636, 1.6581, 1.5696, 1.6972], device='cuda:6'), covar=tensor([0.0500, 0.0361, 0.0352, 0.0551, 0.0263, 0.0546, 0.0546, 0.0561], device='cuda:6'), in_proj_covar=tensor([0.0029, 0.0024, 0.0022, 0.0029, 0.0020, 0.0028, 0.0028, 0.0029], device='cuda:6'), out_proj_covar=tensor([0.0050, 0.0045, 0.0038, 0.0051, 0.0038, 0.0049, 0.0049, 0.0051], device='cuda:6') 2023-04-27 09:37:08,836 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=92960.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:37:29,936 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=92975.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:37:30,563 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.2513, 1.7002, 1.5182, 1.9826, 1.8507, 1.9916, 1.4578, 4.0223], device='cuda:6'), covar=tensor([0.0588, 0.0750, 0.0788, 0.1129, 0.0609, 0.0586, 0.0768, 0.0120], device='cuda:6'), in_proj_covar=tensor([0.0037, 0.0038, 0.0039, 0.0043, 0.0039, 0.0038, 0.0038, 0.0056], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:6') 2023-04-27 09:37:33,593 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=92981.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:37:40,807 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 17, batch 1350, loss[loss=0.2275, simple_loss=0.2797, pruned_loss=0.08761, over 4925.00 frames. ], tot_loss[loss=0.1775, simple_loss=0.2464, pruned_loss=0.05431, over 954901.98 frames. ], batch size: 33, lr: 3.40e-03, grad_scale: 32.0 2023-04-27 09:37:54,410 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=93010.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:38:16,123 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=93036.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:38:16,138 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=93036.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:38:22,342 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.29 vs. limit=5.0 2023-04-27 09:38:24,778 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6628, 1.3636, 1.8248, 2.1170, 1.8249, 1.6211, 1.7189, 1.7602], device='cuda:6'), covar=tensor([0.4127, 0.5948, 0.5100, 0.5262, 0.5411, 0.7009, 0.6558, 0.6832], device='cuda:6'), in_proj_covar=tensor([0.0420, 0.0406, 0.0495, 0.0505, 0.0450, 0.0472, 0.0478, 0.0483], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 09:38:26,344 INFO [finetune.py:976] (6/7) Epoch 17, batch 1400, loss[loss=0.1423, simple_loss=0.2128, pruned_loss=0.03596, over 4733.00 frames. ], tot_loss[loss=0.1798, simple_loss=0.2499, pruned_loss=0.05483, over 956115.11 frames. ], batch size: 23, lr: 3.40e-03, grad_scale: 32.0 2023-04-27 09:38:44,539 INFO [optim.py:369] (6/7) 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 09:39:08,824 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=93071.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:39:12,768 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=93076.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:39:28,198 INFO [finetune.py:976] (6/7) Epoch 17, batch 1450, loss[loss=0.2174, simple_loss=0.2908, pruned_loss=0.07202, over 4807.00 frames. ], tot_loss[loss=0.1809, simple_loss=0.2513, pruned_loss=0.0552, over 954183.72 frames. ], batch size: 40, lr: 3.40e-03, grad_scale: 32.0 2023-04-27 09:39:37,775 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=93097.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 09:40:01,744 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=93115.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:40:18,616 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6841, 1.3396, 1.2952, 1.4826, 1.8648, 1.4922, 1.2785, 1.2869], device='cuda:6'), covar=tensor([0.1568, 0.1434, 0.1867, 0.1235, 0.0727, 0.1731, 0.1999, 0.2000], device='cuda:6'), in_proj_covar=tensor([0.0306, 0.0311, 0.0352, 0.0289, 0.0327, 0.0311, 0.0301, 0.0366], device='cuda:6'), out_proj_covar=tensor([6.3197e-05, 6.4926e-05, 7.5184e-05, 5.8880e-05, 6.7966e-05, 6.5624e-05, 6.3496e-05, 7.8246e-05], device='cuda:6') 2023-04-27 09:40:24,594 INFO [finetune.py:976] (6/7) Epoch 17, batch 1500, loss[loss=0.1887, simple_loss=0.2586, pruned_loss=0.05935, over 4918.00 frames. ], tot_loss[loss=0.1802, simple_loss=0.2512, pruned_loss=0.05463, over 954385.40 frames. ], batch size: 38, lr: 3.40e-03, grad_scale: 32.0 2023-04-27 09:40:31,496 INFO [optim.py:369] (6/7) 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 09:40:42,300 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0940, 1.6246, 1.9948, 2.3813, 2.0045, 1.6048, 1.4030, 1.8258], device='cuda:6'), covar=tensor([0.3040, 0.3115, 0.1605, 0.2045, 0.2410, 0.2646, 0.4110, 0.2005], device='cuda:6'), in_proj_covar=tensor([0.0287, 0.0244, 0.0223, 0.0311, 0.0215, 0.0228, 0.0227, 0.0184], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:6') 2023-04-27 09:40:54,002 INFO [zipformer.py:1188] (6/7) 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] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=93187.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:41:15,562 INFO [finetune.py:976] (6/7) Epoch 17, batch 1550, loss[loss=0.1893, simple_loss=0.2572, pruned_loss=0.06072, over 4809.00 frames. ], tot_loss[loss=0.1797, simple_loss=0.2511, pruned_loss=0.05416, over 955529.15 frames. ], batch size: 41, lr: 3.40e-03, grad_scale: 32.0 2023-04-27 09:41:32,289 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.5325, 2.3023, 2.7019, 2.9826, 2.8121, 2.2869, 2.0889, 2.5146], device='cuda:6'), covar=tensor([0.0828, 0.1008, 0.0557, 0.0609, 0.0625, 0.0899, 0.0742, 0.0640], device='cuda:6'), in_proj_covar=tensor([0.0189, 0.0201, 0.0182, 0.0172, 0.0177, 0.0181, 0.0152, 0.0180], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 09:41:49,257 INFO [finetune.py:976] (6/7) Epoch 17, batch 1600, loss[loss=0.1559, simple_loss=0.2297, pruned_loss=0.04111, over 4824.00 frames. ], tot_loss[loss=0.1786, simple_loss=0.2489, pruned_loss=0.05412, over 953402.76 frames. ], batch size: 30, lr: 3.40e-03, grad_scale: 32.0 2023-04-27 09:41:54,716 INFO [optim.py:369] (6/7) 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 09:41:58,104 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=93255.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:42:02,897 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5007, 2.4727, 2.0291, 2.2960, 2.5770, 2.2519, 3.4582, 1.8354], device='cuda:6'), covar=tensor([0.3942, 0.2568, 0.4814, 0.3578, 0.1945, 0.2925, 0.1368, 0.4973], device='cuda:6'), in_proj_covar=tensor([0.0343, 0.0347, 0.0431, 0.0356, 0.0383, 0.0383, 0.0371, 0.0424], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 09:42:15,824 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=93281.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:42:19,490 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=93287.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:42:23,065 INFO [finetune.py:976] (6/7) Epoch 17, batch 1650, loss[loss=0.1591, simple_loss=0.2163, pruned_loss=0.05095, over 4017.00 frames. ], tot_loss[loss=0.1768, simple_loss=0.2466, pruned_loss=0.05351, over 953797.20 frames. ], batch size: 17, lr: 3.40e-03, grad_scale: 32.0 2023-04-27 09:43:03,633 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=93329.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:43:04,886 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=93331.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:43:14,690 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.2703, 1.4986, 1.3037, 1.4468, 1.2828, 1.2727, 1.2596, 1.0852], device='cuda:6'), covar=tensor([0.1471, 0.1181, 0.0938, 0.1066, 0.3275, 0.1074, 0.1484, 0.1936], device='cuda:6'), in_proj_covar=tensor([0.0288, 0.0310, 0.0222, 0.0281, 0.0317, 0.0262, 0.0253, 0.0269], device='cuda:6'), out_proj_covar=tensor([1.1584e-04, 1.2329e-04, 8.8251e-05, 1.1157e-04, 1.2874e-04, 1.0418e-04, 1.0222e-04, 1.0701e-04], device='cuda:6') 2023-04-27 09:43:17,625 INFO [finetune.py:976] (6/7) Epoch 17, batch 1700, loss[loss=0.2046, simple_loss=0.269, pruned_loss=0.07011, over 4904.00 frames. ], tot_loss[loss=0.1747, simple_loss=0.2439, pruned_loss=0.05272, over 955296.43 frames. ], batch size: 35, lr: 3.40e-03, grad_scale: 32.0 2023-04-27 09:43:23,104 INFO [optim.py:369] (6/7) 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 09:43:33,693 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=93366.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:43:36,044 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.3803, 1.7725, 1.5688, 2.2024, 2.3555, 1.9441, 1.8547, 1.7460], device='cuda:6'), covar=tensor([0.1596, 0.1686, 0.2092, 0.1466, 0.1395, 0.2059, 0.2220, 0.2218], device='cuda:6'), in_proj_covar=tensor([0.0307, 0.0312, 0.0352, 0.0290, 0.0329, 0.0311, 0.0301, 0.0367], device='cuda:6'), out_proj_covar=tensor([6.3435e-05, 6.5084e-05, 7.4992e-05, 5.9074e-05, 6.8378e-05, 6.5659e-05, 6.3600e-05, 7.8309e-05], device='cuda:6') 2023-04-27 09:43:41,237 INFO [zipformer.py:1188] (6/7) 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] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=93392.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 09:43:51,576 INFO [finetune.py:976] (6/7) Epoch 17, batch 1750, loss[loss=0.2297, simple_loss=0.3046, pruned_loss=0.07741, over 4941.00 frames. ], tot_loss[loss=0.1783, simple_loss=0.2475, pruned_loss=0.05459, over 955634.17 frames. ], batch size: 33, lr: 3.40e-03, grad_scale: 32.0 2023-04-27 09:44:13,254 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=93424.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:44:36,167 INFO [finetune.py:976] (6/7) Epoch 17, batch 1800, loss[loss=0.2053, simple_loss=0.2664, pruned_loss=0.07211, over 4798.00 frames. ], tot_loss[loss=0.1806, simple_loss=0.2505, pruned_loss=0.05533, over 954837.42 frames. ], batch size: 51, lr: 3.40e-03, grad_scale: 32.0 2023-04-27 09:44:47,782 INFO [optim.py:369] (6/7) 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 09:45:05,915 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=93471.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:45:21,540 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7634, 1.3974, 1.2285, 1.5954, 1.8716, 1.4819, 1.3060, 1.1905], device='cuda:6'), covar=tensor([0.1685, 0.1803, 0.2168, 0.1316, 0.1085, 0.1994, 0.2376, 0.2647], device='cuda:6'), in_proj_covar=tensor([0.0309, 0.0314, 0.0354, 0.0292, 0.0331, 0.0313, 0.0303, 0.0369], device='cuda:6'), out_proj_covar=tensor([6.3826e-05, 6.5503e-05, 7.5519e-05, 5.9375e-05, 6.8808e-05, 6.5961e-05, 6.3957e-05, 7.8800e-05], device='cuda:6') 2023-04-27 09:45:22,734 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=93487.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:45:26,247 INFO [finetune.py:976] (6/7) Epoch 17, batch 1850, loss[loss=0.2088, simple_loss=0.2891, pruned_loss=0.06431, over 4818.00 frames. ], tot_loss[loss=0.1829, simple_loss=0.2528, pruned_loss=0.0565, over 953945.91 frames. ], batch size: 33, lr: 3.40e-03, grad_scale: 32.0 2023-04-27 09:46:20,316 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7855, 1.3618, 1.4226, 1.3758, 1.8763, 1.5110, 1.1868, 1.3451], device='cuda:6'), covar=tensor([0.1518, 0.1296, 0.2025, 0.1287, 0.0793, 0.1417, 0.2181, 0.2113], device='cuda:6'), in_proj_covar=tensor([0.0308, 0.0313, 0.0352, 0.0291, 0.0330, 0.0312, 0.0302, 0.0368], device='cuda:6'), out_proj_covar=tensor([6.3589e-05, 6.5351e-05, 7.5133e-05, 5.9121e-05, 6.8638e-05, 6.5703e-05, 6.3738e-05, 7.8490e-05], device='cuda:6') 2023-04-27 09:46:27,200 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=93535.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:46:37,237 INFO [finetune.py:976] (6/7) Epoch 17, batch 1900, loss[loss=0.2028, simple_loss=0.2709, pruned_loss=0.06734, over 4903.00 frames. ], tot_loss[loss=0.1829, simple_loss=0.2533, pruned_loss=0.05619, over 954959.87 frames. ], batch size: 37, lr: 3.40e-03, grad_scale: 32.0 2023-04-27 09:46:42,802 INFO [optim.py:369] (6/7) 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 09:46:44,113 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([0.4912, 1.3420, 1.3826, 0.9697, 1.3612, 1.0960, 1.7394, 1.3524], device='cuda:6'), covar=tensor([0.3523, 0.1887, 0.5529, 0.2802, 0.1616, 0.2351, 0.1649, 0.5100], device='cuda:6'), in_proj_covar=tensor([0.0342, 0.0346, 0.0432, 0.0357, 0.0383, 0.0383, 0.0371, 0.0424], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 09:46:44,737 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=93555.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:47:06,593 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=93587.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:47:10,127 INFO [finetune.py:976] (6/7) Epoch 17, batch 1950, loss[loss=0.1381, simple_loss=0.2173, pruned_loss=0.02949, over 4744.00 frames. ], tot_loss[loss=0.181, simple_loss=0.2513, pruned_loss=0.05533, over 952999.57 frames. ], batch size: 26, lr: 3.40e-03, grad_scale: 32.0 2023-04-27 09:47:16,347 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=93603.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:47:35,238 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=93631.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:47:37,614 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 17, batch 2000, loss[loss=0.1678, simple_loss=0.2411, pruned_loss=0.04725, over 4930.00 frames. ], tot_loss[loss=0.1795, simple_loss=0.2493, pruned_loss=0.05489, over 953682.26 frames. ], batch size: 33, lr: 3.40e-03, grad_scale: 32.0 2023-04-27 09:47:54,465 INFO [optim.py:369] (6/7) 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 09:47:58,292 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.1305, 0.7108, 0.9142, 0.7599, 1.2515, 1.0072, 0.8467, 0.9646], device='cuda:6'), covar=tensor([0.1845, 0.1603, 0.2255, 0.1617, 0.1144, 0.1544, 0.1832, 0.2349], device='cuda:6'), in_proj_covar=tensor([0.0309, 0.0314, 0.0353, 0.0292, 0.0330, 0.0312, 0.0302, 0.0368], device='cuda:6'), out_proj_covar=tensor([6.3923e-05, 6.5489e-05, 7.5199e-05, 5.9307e-05, 6.8576e-05, 6.5796e-05, 6.3761e-05, 7.8535e-05], device='cuda:6') 2023-04-27 09:48:03,151 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=93666.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:48:12,491 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=93679.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:48:21,378 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=93692.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:48:21,912 INFO [finetune.py:976] (6/7) Epoch 17, batch 2050, loss[loss=0.2057, simple_loss=0.2628, pruned_loss=0.07427, over 4936.00 frames. ], tot_loss[loss=0.1782, simple_loss=0.2472, pruned_loss=0.05456, over 954782.18 frames. ], batch size: 33, lr: 3.40e-03, grad_scale: 32.0 2023-04-27 09:48:35,564 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=93714.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:48:50,244 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0450, 1.6341, 1.8538, 2.2623, 2.2853, 1.8152, 1.6596, 2.0789], device='cuda:6'), covar=tensor([0.0712, 0.1112, 0.0675, 0.0557, 0.0613, 0.0867, 0.0752, 0.0524], device='cuda:6'), in_proj_covar=tensor([0.0189, 0.0201, 0.0183, 0.0172, 0.0176, 0.0181, 0.0153, 0.0180], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 09:48:53,240 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=93740.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:48:55,041 INFO [finetune.py:976] (6/7) Epoch 17, batch 2100, loss[loss=0.1438, simple_loss=0.2141, pruned_loss=0.03672, over 4822.00 frames. ], tot_loss[loss=0.1785, simple_loss=0.2475, pruned_loss=0.05481, over 955095.91 frames. ], batch size: 25, lr: 3.40e-03, grad_scale: 32.0 2023-04-27 09:49:01,813 INFO [optim.py:369] (6/7) 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 09:49:12,997 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=93770.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:49:13,053 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.14 vs. limit=2.0 2023-04-27 09:49:13,577 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=93771.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:49:29,663 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([0.7788, 1.6804, 1.5352, 1.2395, 1.7053, 1.4871, 2.1945, 1.3972], device='cuda:6'), covar=tensor([0.3526, 0.1751, 0.4763, 0.2822, 0.1682, 0.2145, 0.1577, 0.4682], device='cuda:6'), in_proj_covar=tensor([0.0339, 0.0343, 0.0425, 0.0353, 0.0380, 0.0378, 0.0367, 0.0418], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 09:49:33,699 INFO [finetune.py:976] (6/7) Epoch 17, batch 2150, loss[loss=0.242, simple_loss=0.3175, pruned_loss=0.08325, over 4826.00 frames. ], tot_loss[loss=0.1808, simple_loss=0.2499, pruned_loss=0.05587, over 955886.71 frames. ], batch size: 40, lr: 3.39e-03, grad_scale: 32.0 2023-04-27 09:50:13,740 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=93819.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:50:15,015 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.4204, 3.1654, 0.8687, 1.8712, 1.7968, 2.3653, 1.8541, 1.0488], device='cuda:6'), covar=tensor([0.1349, 0.0959, 0.1994, 0.1202, 0.1103, 0.0920, 0.1460, 0.1923], device='cuda:6'), in_proj_covar=tensor([0.0117, 0.0241, 0.0137, 0.0120, 0.0131, 0.0152, 0.0116, 0.0119], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-27 09:50:26,750 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=93831.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:50:47,870 INFO [finetune.py:976] (6/7) Epoch 17, batch 2200, loss[loss=0.191, simple_loss=0.2674, pruned_loss=0.0573, over 4815.00 frames. ], tot_loss[loss=0.1827, simple_loss=0.2527, pruned_loss=0.05638, over 956707.96 frames. ], batch size: 51, lr: 3.39e-03, grad_scale: 32.0 2023-04-27 09:50:59,348 INFO [optim.py:369] (6/7) 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 09:51:02,380 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.2342, 1.2938, 3.8143, 3.5302, 3.3338, 3.6688, 3.6475, 3.3739], device='cuda:6'), covar=tensor([0.7189, 0.5793, 0.1090, 0.1791, 0.1235, 0.1871, 0.1548, 0.1472], device='cuda:6'), in_proj_covar=tensor([0.0307, 0.0304, 0.0398, 0.0400, 0.0345, 0.0404, 0.0308, 0.0361], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-27 09:51:08,690 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.67 vs. limit=2.0 2023-04-27 09:51:40,106 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.4783, 1.6922, 1.3769, 1.1174, 1.1018, 1.1312, 1.3283, 1.0771], device='cuda:6'), covar=tensor([0.1849, 0.1312, 0.1566, 0.1796, 0.2445, 0.2053, 0.1184, 0.2157], device='cuda:6'), in_proj_covar=tensor([0.0197, 0.0213, 0.0169, 0.0206, 0.0201, 0.0185, 0.0157, 0.0185], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:6') 2023-04-27 09:51:41,203 INFO [finetune.py:976] (6/7) Epoch 17, batch 2250, loss[loss=0.1963, simple_loss=0.2607, pruned_loss=0.06593, over 4893.00 frames. ], tot_loss[loss=0.1831, simple_loss=0.2532, pruned_loss=0.05648, over 956396.34 frames. ], batch size: 32, lr: 3.39e-03, grad_scale: 32.0 2023-04-27 09:51:42,994 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5768, 1.3358, 4.5703, 4.2502, 3.9419, 4.3044, 4.2427, 3.9997], device='cuda:6'), covar=tensor([0.7179, 0.5976, 0.1058, 0.1678, 0.1212, 0.1559, 0.1133, 0.1470], device='cuda:6'), in_proj_covar=tensor([0.0307, 0.0304, 0.0398, 0.0400, 0.0345, 0.0404, 0.0308, 0.0362], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-27 09:51:44,336 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-04-27 09:51:51,335 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-04-27 09:52:09,518 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-04-27 09:52:14,468 INFO [finetune.py:976] (6/7) Epoch 17, batch 2300, loss[loss=0.173, simple_loss=0.2473, pruned_loss=0.04933, over 4814.00 frames. ], tot_loss[loss=0.181, simple_loss=0.2519, pruned_loss=0.055, over 958205.12 frames. ], batch size: 41, lr: 3.39e-03, grad_scale: 32.0 2023-04-27 09:52:20,960 INFO [optim.py:369] (6/7) 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 09:52:40,201 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=93981.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:52:47,882 INFO [finetune.py:976] (6/7) Epoch 17, batch 2350, loss[loss=0.1476, simple_loss=0.2136, pruned_loss=0.0408, over 4924.00 frames. ], tot_loss[loss=0.1783, simple_loss=0.249, pruned_loss=0.05383, over 958281.05 frames. ], batch size: 33, lr: 3.39e-03, grad_scale: 32.0 2023-04-27 09:53:48,608 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=94042.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:53:49,095 INFO [finetune.py:976] (6/7) Epoch 17, batch 2400, loss[loss=0.1517, simple_loss=0.2217, pruned_loss=0.0408, over 4894.00 frames. ], tot_loss[loss=0.1764, simple_loss=0.2463, pruned_loss=0.05327, over 959918.06 frames. ], batch size: 32, lr: 3.39e-03, grad_scale: 32.0 2023-04-27 09:53:56,108 INFO [optim.py:369] (6/7) 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 09:54:21,375 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.3411, 1.5998, 1.4160, 1.5580, 1.3084, 1.4211, 1.3730, 1.1941], device='cuda:6'), covar=tensor([0.1649, 0.1170, 0.0848, 0.1034, 0.3438, 0.1011, 0.1677, 0.2063], device='cuda:6'), in_proj_covar=tensor([0.0284, 0.0307, 0.0218, 0.0277, 0.0312, 0.0257, 0.0249, 0.0265], device='cuda:6'), out_proj_covar=tensor([1.1403e-04, 1.2195e-04, 8.6922e-05, 1.1009e-04, 1.2678e-04, 1.0238e-04, 1.0059e-04, 1.0531e-04], device='cuda:6') 2023-04-27 09:54:23,068 INFO [finetune.py:976] (6/7) Epoch 17, batch 2450, loss[loss=0.1878, simple_loss=0.2608, pruned_loss=0.05745, over 4826.00 frames. ], tot_loss[loss=0.1746, simple_loss=0.2443, pruned_loss=0.05246, over 960721.77 frames. ], batch size: 51, lr: 3.39e-03, grad_scale: 32.0 2023-04-27 09:54:46,634 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 17, batch 2500, loss[loss=0.2017, simple_loss=0.278, pruned_loss=0.06268, over 4873.00 frames. ], tot_loss[loss=0.1754, simple_loss=0.2456, pruned_loss=0.05256, over 960453.41 frames. ], batch size: 34, lr: 3.39e-03, grad_scale: 32.0 2023-04-27 09:55:02,069 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6025, 1.0608, 1.3756, 1.2248, 1.7287, 1.4325, 1.1469, 1.3119], device='cuda:6'), covar=tensor([0.1666, 0.1664, 0.1999, 0.1488, 0.0938, 0.1505, 0.1861, 0.2272], device='cuda:6'), in_proj_covar=tensor([0.0307, 0.0312, 0.0350, 0.0289, 0.0328, 0.0310, 0.0300, 0.0366], device='cuda:6'), out_proj_covar=tensor([6.3433e-05, 6.5110e-05, 7.4654e-05, 5.8848e-05, 6.8246e-05, 6.5372e-05, 6.3272e-05, 7.8028e-05], device='cuda:6') 2023-04-27 09:55:03,648 INFO [optim.py:369] (6/7) 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 09:55:31,028 INFO [finetune.py:976] (6/7) Epoch 17, batch 2550, loss[loss=0.2404, simple_loss=0.3083, pruned_loss=0.0863, over 4909.00 frames. ], tot_loss[loss=0.1771, simple_loss=0.2481, pruned_loss=0.05307, over 957884.88 frames. ], batch size: 37, lr: 3.39e-03, grad_scale: 32.0 2023-04-27 09:55:52,786 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=94223.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 09:56:09,850 INFO [finetune.py:976] (6/7) Epoch 17, batch 2600, loss[loss=0.1447, simple_loss=0.2191, pruned_loss=0.03511, over 4757.00 frames. ], tot_loss[loss=0.1787, simple_loss=0.2495, pruned_loss=0.05395, over 957091.32 frames. ], batch size: 28, lr: 3.39e-03, grad_scale: 32.0 2023-04-27 09:56:21,228 INFO [optim.py:369] (6/7) 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 09:57:00,082 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=94284.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 09:57:05,479 INFO [finetune.py:976] (6/7) Epoch 17, batch 2650, loss[loss=0.1966, simple_loss=0.266, pruned_loss=0.06364, over 4902.00 frames. ], tot_loss[loss=0.1794, simple_loss=0.2507, pruned_loss=0.054, over 954899.74 frames. ], batch size: 37, lr: 3.39e-03, grad_scale: 32.0 2023-04-27 09:57:34,900 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=94337.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:57:38,467 INFO [finetune.py:976] (6/7) Epoch 17, batch 2700, loss[loss=0.2303, simple_loss=0.2967, pruned_loss=0.08201, over 4839.00 frames. ], tot_loss[loss=0.178, simple_loss=0.2493, pruned_loss=0.05339, over 955143.85 frames. ], batch size: 47, lr: 3.39e-03, grad_scale: 32.0 2023-04-27 09:57:41,613 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=94348.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:57:43,966 INFO [optim.py:369] (6/7) 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 09:58:17,889 INFO [finetune.py:976] (6/7) Epoch 17, batch 2750, loss[loss=0.1846, simple_loss=0.2445, pruned_loss=0.06236, over 4818.00 frames. ], tot_loss[loss=0.1767, simple_loss=0.2472, pruned_loss=0.05311, over 956623.29 frames. ], batch size: 38, lr: 3.39e-03, grad_scale: 32.0 2023-04-27 09:58:38,905 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=94409.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 09:58:59,210 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.3704, 1.2924, 1.3988, 1.6097, 1.6804, 1.2870, 1.0341, 1.5738], device='cuda:6'), covar=tensor([0.0798, 0.1334, 0.0902, 0.0651, 0.0653, 0.0824, 0.0821, 0.0574], device='cuda:6'), in_proj_covar=tensor([0.0188, 0.0199, 0.0181, 0.0170, 0.0175, 0.0178, 0.0151, 0.0178], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 09:59:01,610 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=94426.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:59:03,387 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7222, 2.0983, 1.7820, 2.0238, 1.6307, 1.8394, 1.7133, 1.4047], device='cuda:6'), covar=tensor([0.1674, 0.1296, 0.0809, 0.1025, 0.2909, 0.0962, 0.1580, 0.2431], device='cuda:6'), in_proj_covar=tensor([0.0286, 0.0309, 0.0220, 0.0280, 0.0313, 0.0260, 0.0251, 0.0266], device='cuda:6'), out_proj_covar=tensor([1.1464e-04, 1.2271e-04, 8.7703e-05, 1.1129e-04, 1.2731e-04, 1.0328e-04, 1.0132e-04, 1.0582e-04], device='cuda:6') 2023-04-27 09:59:24,236 INFO [finetune.py:976] (6/7) Epoch 17, batch 2800, loss[loss=0.183, simple_loss=0.2412, pruned_loss=0.06245, over 4825.00 frames. ], tot_loss[loss=0.1757, simple_loss=0.2452, pruned_loss=0.05304, over 954497.44 frames. ], batch size: 41, lr: 3.39e-03, grad_scale: 32.0 2023-04-27 09:59:35,022 INFO [optim.py:369] (6/7) 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 09:59:51,545 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6567, 2.0876, 1.7400, 1.9947, 1.5352, 1.8068, 1.7199, 1.3352], device='cuda:6'), covar=tensor([0.1809, 0.1056, 0.0878, 0.1037, 0.3095, 0.0988, 0.1653, 0.2312], device='cuda:6'), in_proj_covar=tensor([0.0287, 0.0310, 0.0221, 0.0281, 0.0314, 0.0261, 0.0251, 0.0268], device='cuda:6'), out_proj_covar=tensor([1.1512e-04, 1.2352e-04, 8.8010e-05, 1.1169e-04, 1.2789e-04, 1.0368e-04, 1.0159e-04, 1.0630e-04], device='cuda:6') 2023-04-27 09:59:54,420 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=94474.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 09:59:54,643 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.45 vs. limit=5.0 2023-04-27 09:59:56,908 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.1801, 2.7655, 2.1978, 2.6671, 1.9201, 2.4598, 2.3446, 1.7308], device='cuda:6'), covar=tensor([0.1843, 0.1019, 0.0839, 0.0995, 0.3086, 0.0996, 0.1567, 0.2626], device='cuda:6'), in_proj_covar=tensor([0.0287, 0.0310, 0.0221, 0.0281, 0.0314, 0.0260, 0.0251, 0.0267], device='cuda:6'), out_proj_covar=tensor([1.1505e-04, 1.2346e-04, 8.7915e-05, 1.1158e-04, 1.2783e-04, 1.0359e-04, 1.0148e-04, 1.0616e-04], device='cuda:6') 2023-04-27 10:00:06,869 INFO [finetune.py:976] (6/7) Epoch 17, batch 2850, loss[loss=0.1697, simple_loss=0.234, pruned_loss=0.05265, over 4719.00 frames. ], tot_loss[loss=0.1746, simple_loss=0.2438, pruned_loss=0.05268, over 955389.80 frames. ], batch size: 23, lr: 3.39e-03, grad_scale: 32.0 2023-04-27 10:00:40,994 INFO [finetune.py:976] (6/7) Epoch 17, batch 2900, loss[loss=0.1692, simple_loss=0.258, pruned_loss=0.04018, over 4825.00 frames. ], tot_loss[loss=0.1772, simple_loss=0.2471, pruned_loss=0.05366, over 954259.09 frames. ], batch size: 30, lr: 3.39e-03, grad_scale: 32.0 2023-04-27 10:00:46,388 INFO [optim.py:369] (6/7) 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 10:01:04,342 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=94579.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 10:01:14,242 INFO [finetune.py:976] (6/7) Epoch 17, batch 2950, loss[loss=0.2143, simple_loss=0.2986, pruned_loss=0.06497, over 4049.00 frames. ], tot_loss[loss=0.1785, simple_loss=0.2491, pruned_loss=0.05393, over 953684.52 frames. ], batch size: 65, lr: 3.39e-03, grad_scale: 32.0 2023-04-27 10:02:00,086 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=94637.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:02:04,043 INFO [finetune.py:976] (6/7) Epoch 17, batch 3000, loss[loss=0.1519, simple_loss=0.2275, pruned_loss=0.03812, over 4869.00 frames. ], tot_loss[loss=0.1794, simple_loss=0.2503, pruned_loss=0.05428, over 954648.61 frames. ], batch size: 31, lr: 3.39e-03, grad_scale: 32.0 2023-04-27 10:02:04,044 INFO [finetune.py:1001] (6/7) Computing validation loss 2023-04-27 10:02:27,437 INFO [finetune.py:1010] (6/7) 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] (6/7) Maximum memory allocated so far is 6365MB 2023-04-27 10:02:39,393 INFO [optim.py:369] (6/7) 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 10:03:10,439 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=94685.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:03:15,763 INFO [finetune.py:976] (6/7) Epoch 17, batch 3050, loss[loss=0.1882, simple_loss=0.2528, pruned_loss=0.06181, over 4708.00 frames. ], tot_loss[loss=0.1787, simple_loss=0.25, pruned_loss=0.05375, over 955229.79 frames. ], batch size: 23, lr: 3.39e-03, grad_scale: 32.0 2023-04-27 10:03:24,050 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=94704.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 10:03:24,792 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-04-27 10:03:46,476 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.78 vs. limit=2.0 2023-04-27 10:03:47,899 INFO [finetune.py:976] (6/7) Epoch 17, batch 3100, loss[loss=0.1495, simple_loss=0.2189, pruned_loss=0.04011, over 4789.00 frames. ], tot_loss[loss=0.1768, simple_loss=0.2476, pruned_loss=0.05294, over 955732.57 frames. ], batch size: 26, lr: 3.39e-03, grad_scale: 32.0 2023-04-27 10:03:55,873 INFO [optim.py:369] (6/7) 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 10:04:21,244 INFO [finetune.py:976] (6/7) Epoch 17, batch 3150, loss[loss=0.1732, simple_loss=0.2386, pruned_loss=0.05387, over 4820.00 frames. ], tot_loss[loss=0.1763, simple_loss=0.2464, pruned_loss=0.0531, over 957156.49 frames. ], batch size: 25, lr: 3.39e-03, grad_scale: 64.0 2023-04-27 10:04:26,821 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.3935, 3.3971, 0.7322, 1.9260, 2.0407, 2.3479, 1.9292, 0.9688], device='cuda:6'), covar=tensor([0.1421, 0.0907, 0.2056, 0.1141, 0.0947, 0.1024, 0.1573, 0.2170], device='cuda:6'), in_proj_covar=tensor([0.0118, 0.0244, 0.0138, 0.0121, 0.0133, 0.0154, 0.0118, 0.0121], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-27 10:05:23,147 INFO [finetune.py:976] (6/7) Epoch 17, batch 3200, loss[loss=0.114, simple_loss=0.1955, pruned_loss=0.01623, over 4759.00 frames. ], tot_loss[loss=0.1734, simple_loss=0.2431, pruned_loss=0.05186, over 957927.84 frames. ], batch size: 27, lr: 3.39e-03, grad_scale: 64.0 2023-04-27 10:05:34,599 INFO [optim.py:369] (6/7) 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 10:05:53,665 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=94879.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 10:06:02,200 INFO [finetune.py:976] (6/7) Epoch 17, batch 3250, loss[loss=0.2292, simple_loss=0.2875, pruned_loss=0.08547, over 4837.00 frames. ], tot_loss[loss=0.1752, simple_loss=0.2446, pruned_loss=0.05292, over 957852.29 frames. ], batch size: 33, lr: 3.39e-03, grad_scale: 64.0 2023-04-27 10:06:26,491 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=94927.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 10:06:26,517 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0178, 2.5084, 0.9862, 1.3396, 1.9534, 1.2313, 3.4844, 1.6984], device='cuda:6'), covar=tensor([0.0696, 0.0713, 0.0817, 0.1263, 0.0574, 0.1063, 0.0297, 0.0701], device='cuda:6'), in_proj_covar=tensor([0.0051, 0.0065, 0.0048, 0.0046, 0.0050, 0.0052, 0.0074, 0.0051], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0007, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:6') 2023-04-27 10:06:36,229 INFO [finetune.py:976] (6/7) Epoch 17, batch 3300, loss[loss=0.1689, simple_loss=0.2222, pruned_loss=0.05784, over 4826.00 frames. ], tot_loss[loss=0.1776, simple_loss=0.2483, pruned_loss=0.05345, over 956960.92 frames. ], batch size: 25, lr: 3.38e-03, grad_scale: 64.0 2023-04-27 10:06:47,633 INFO [optim.py:369] (6/7) 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 10:07:30,062 INFO [finetune.py:976] (6/7) Epoch 17, batch 3350, loss[loss=0.1313, simple_loss=0.206, pruned_loss=0.02834, over 4789.00 frames. ], tot_loss[loss=0.1804, simple_loss=0.2518, pruned_loss=0.05452, over 957465.55 frames. ], batch size: 26, lr: 3.38e-03, grad_scale: 64.0 2023-04-27 10:07:33,234 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9783, 2.6093, 2.0146, 2.0799, 1.5921, 1.5336, 2.0701, 1.4998], device='cuda:6'), covar=tensor([0.1311, 0.1338, 0.1236, 0.1504, 0.1942, 0.1610, 0.0831, 0.1752], device='cuda:6'), in_proj_covar=tensor([0.0197, 0.0213, 0.0168, 0.0206, 0.0201, 0.0185, 0.0156, 0.0185], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:6') 2023-04-27 10:07:36,974 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=95004.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:07:55,416 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.1279, 1.4748, 1.3434, 1.7083, 1.5551, 1.8642, 1.3637, 3.3991], device='cuda:6'), covar=tensor([0.0642, 0.0774, 0.0803, 0.1158, 0.0620, 0.0519, 0.0755, 0.0178], device='cuda:6'), in_proj_covar=tensor([0.0038, 0.0039, 0.0040, 0.0044, 0.0040, 0.0038, 0.0039, 0.0056], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:6') 2023-04-27 10:08:02,708 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5929, 2.1051, 1.7309, 1.9616, 1.5088, 1.6374, 1.6724, 1.2957], device='cuda:6'), covar=tensor([0.1980, 0.1252, 0.0924, 0.1091, 0.3527, 0.1301, 0.1807, 0.2458], device='cuda:6'), in_proj_covar=tensor([0.0285, 0.0306, 0.0219, 0.0279, 0.0312, 0.0259, 0.0249, 0.0266], device='cuda:6'), out_proj_covar=tensor([1.1432e-04, 1.2162e-04, 8.7230e-05, 1.1083e-04, 1.2685e-04, 1.0316e-04, 1.0089e-04, 1.0555e-04], device='cuda:6') 2023-04-27 10:08:03,817 INFO [finetune.py:976] (6/7) Epoch 17, batch 3400, loss[loss=0.2185, simple_loss=0.2857, pruned_loss=0.07564, over 4903.00 frames. ], tot_loss[loss=0.1821, simple_loss=0.2533, pruned_loss=0.05546, over 954558.07 frames. ], batch size: 37, lr: 3.38e-03, grad_scale: 64.0 2023-04-27 10:08:09,294 INFO [optim.py:369] (6/7) 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 10:08:09,365 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=95052.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:08:14,391 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-04-27 10:08:21,182 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.2155, 1.1777, 1.4930, 1.4376, 1.1764, 1.0443, 1.2299, 0.8970], device='cuda:6'), covar=tensor([0.0677, 0.0566, 0.0435, 0.0518, 0.0743, 0.1020, 0.0579, 0.0599], device='cuda:6'), in_proj_covar=tensor([0.0069, 0.0070, 0.0068, 0.0068, 0.0075, 0.0096, 0.0075, 0.0067], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-27 10:08:37,249 INFO [finetune.py:976] (6/7) Epoch 17, batch 3450, loss[loss=0.1575, simple_loss=0.2256, pruned_loss=0.04468, over 4739.00 frames. ], tot_loss[loss=0.1811, simple_loss=0.2524, pruned_loss=0.05492, over 953982.96 frames. ], batch size: 54, lr: 3.38e-03, grad_scale: 64.0 2023-04-27 10:08:44,645 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=95105.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 10:09:09,312 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([3.2761, 3.1925, 2.4542, 3.8315, 3.3341, 3.2867, 1.5236, 3.2560], device='cuda:6'), covar=tensor([0.2092, 0.1456, 0.3428, 0.2408, 0.3192, 0.1959, 0.5638, 0.2622], device='cuda:6'), in_proj_covar=tensor([0.0244, 0.0213, 0.0249, 0.0303, 0.0296, 0.0247, 0.0271, 0.0271], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 10:09:11,015 INFO [finetune.py:976] (6/7) Epoch 17, batch 3500, loss[loss=0.1426, simple_loss=0.2056, pruned_loss=0.03977, over 4857.00 frames. ], tot_loss[loss=0.1793, simple_loss=0.25, pruned_loss=0.05431, over 955666.86 frames. ], batch size: 49, lr: 3.38e-03, grad_scale: 64.0 2023-04-27 10:09:16,407 INFO [optim.py:369] (6/7) 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 10:09:17,160 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6841, 2.1490, 1.7765, 1.9862, 1.5813, 1.6883, 1.7245, 1.3322], device='cuda:6'), covar=tensor([0.1739, 0.1201, 0.0916, 0.1080, 0.3234, 0.1130, 0.1662, 0.2592], device='cuda:6'), in_proj_covar=tensor([0.0283, 0.0305, 0.0219, 0.0279, 0.0312, 0.0258, 0.0248, 0.0265], device='cuda:6'), out_proj_covar=tensor([1.1382e-04, 1.2130e-04, 8.7130e-05, 1.1062e-04, 1.2671e-04, 1.0264e-04, 1.0053e-04, 1.0538e-04], device='cuda:6') 2023-04-27 10:09:26,058 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=95166.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 10:09:44,905 INFO [finetune.py:976] (6/7) Epoch 17, batch 3550, loss[loss=0.1577, simple_loss=0.2343, pruned_loss=0.0406, over 4828.00 frames. ], tot_loss[loss=0.1771, simple_loss=0.2472, pruned_loss=0.05355, over 955887.92 frames. ], batch size: 33, lr: 3.38e-03, grad_scale: 64.0 2023-04-27 10:10:01,540 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-04-27 10:10:42,815 INFO [finetune.py:976] (6/7) Epoch 17, batch 3600, loss[loss=0.1501, simple_loss=0.2199, pruned_loss=0.04022, over 4830.00 frames. ], tot_loss[loss=0.1747, simple_loss=0.2442, pruned_loss=0.05258, over 957967.00 frames. ], batch size: 30, lr: 3.38e-03, grad_scale: 64.0 2023-04-27 10:10:54,066 INFO [optim.py:369] (6/7) 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 10:10:57,863 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.64 vs. limit=5.0 2023-04-27 10:10:59,906 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=95260.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:11:10,462 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=95268.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:11:18,112 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-04-27 10:11:34,141 INFO [finetune.py:976] (6/7) Epoch 17, batch 3650, loss[loss=0.21, simple_loss=0.2876, pruned_loss=0.06622, over 4833.00 frames. ], tot_loss[loss=0.1783, simple_loss=0.2475, pruned_loss=0.05458, over 957639.56 frames. ], batch size: 39, lr: 3.38e-03, grad_scale: 64.0 2023-04-27 10:11:34,901 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7163, 2.3157, 1.7762, 1.6849, 1.2632, 1.2918, 1.8722, 1.2141], device='cuda:6'), covar=tensor([0.1682, 0.1380, 0.1393, 0.1739, 0.2297, 0.1975, 0.0896, 0.1960], device='cuda:6'), in_proj_covar=tensor([0.0195, 0.0211, 0.0167, 0.0204, 0.0200, 0.0183, 0.0155, 0.0184], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:6') 2023-04-27 10:11:51,460 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=95321.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:11:57,809 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=95329.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:12:05,330 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=95339.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:12:07,681 INFO [finetune.py:976] (6/7) Epoch 17, batch 3700, loss[loss=0.1968, simple_loss=0.2626, pruned_loss=0.06543, over 4899.00 frames. ], tot_loss[loss=0.1798, simple_loss=0.2501, pruned_loss=0.05475, over 957191.53 frames. ], batch size: 43, lr: 3.38e-03, grad_scale: 64.0 2023-04-27 10:12:13,179 INFO [optim.py:369] (6/7) 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 10:12:24,283 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5775, 1.2431, 1.3226, 1.2435, 1.6912, 1.3524, 1.1439, 1.2889], device='cuda:6'), covar=tensor([0.1522, 0.1309, 0.1928, 0.1385, 0.0817, 0.1480, 0.1731, 0.2324], device='cuda:6'), in_proj_covar=tensor([0.0310, 0.0316, 0.0355, 0.0290, 0.0330, 0.0313, 0.0302, 0.0368], device='cuda:6'), out_proj_covar=tensor([6.4260e-05, 6.6014e-05, 7.5620e-05, 5.8930e-05, 6.8618e-05, 6.5956e-05, 6.3787e-05, 7.8407e-05], device='cuda:6') 2023-04-27 10:12:34,638 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([0.9900, 2.4528, 0.7029, 1.3668, 1.4394, 1.7146, 1.5621, 0.8377], device='cuda:6'), covar=tensor([0.1654, 0.1431, 0.1997, 0.1480, 0.1206, 0.1174, 0.1735, 0.1907], device='cuda:6'), in_proj_covar=tensor([0.0117, 0.0242, 0.0137, 0.0120, 0.0131, 0.0152, 0.0117, 0.0119], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-27 10:12:41,442 INFO [finetune.py:976] (6/7) Epoch 17, batch 3750, loss[loss=0.2009, simple_loss=0.2781, pruned_loss=0.06186, over 4895.00 frames. ], tot_loss[loss=0.1801, simple_loss=0.2506, pruned_loss=0.05473, over 954810.93 frames. ], batch size: 37, lr: 3.38e-03, grad_scale: 32.0 2023-04-27 10:12:45,998 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=95400.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:13:11,620 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([0.5422, 1.3897, 1.3731, 1.0340, 1.4159, 1.1510, 1.7904, 1.2941], device='cuda:6'), covar=tensor([0.3080, 0.1662, 0.5161, 0.2259, 0.1371, 0.1955, 0.1345, 0.4674], device='cuda:6'), in_proj_covar=tensor([0.0338, 0.0344, 0.0427, 0.0353, 0.0381, 0.0379, 0.0370, 0.0421], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 10:13:14,397 INFO [finetune.py:976] (6/7) Epoch 17, batch 3800, loss[loss=0.1494, simple_loss=0.227, pruned_loss=0.03588, over 4786.00 frames. ], tot_loss[loss=0.1809, simple_loss=0.2515, pruned_loss=0.05515, over 952729.25 frames. ], batch size: 51, lr: 3.38e-03, grad_scale: 32.0 2023-04-27 10:13:20,913 INFO [optim.py:369] (6/7) 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 10:13:23,417 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([4.2710, 4.1175, 3.1191, 4.9020, 4.2539, 4.2015, 1.9821, 4.3148], device='cuda:6'), covar=tensor([0.1649, 0.1075, 0.3413, 0.1131, 0.2371, 0.1730, 0.5686, 0.1915], device='cuda:6'), in_proj_covar=tensor([0.0246, 0.0215, 0.0251, 0.0306, 0.0299, 0.0249, 0.0274, 0.0273], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 10:13:25,328 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.1836, 3.0023, 2.4434, 2.7519, 2.0750, 2.3844, 2.6921, 1.8687], device='cuda:6'), covar=tensor([0.2278, 0.1216, 0.0833, 0.1313, 0.3200, 0.1268, 0.1937, 0.2869], device='cuda:6'), in_proj_covar=tensor([0.0284, 0.0305, 0.0218, 0.0278, 0.0312, 0.0258, 0.0248, 0.0265], device='cuda:6'), out_proj_covar=tensor([1.1388e-04, 1.2135e-04, 8.6824e-05, 1.1031e-04, 1.2666e-04, 1.0267e-04, 1.0040e-04, 1.0517e-04], device='cuda:6') 2023-04-27 10:13:25,884 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=95461.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 10:13:26,560 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.5418, 1.9168, 2.4688, 2.8129, 2.3394, 1.8798, 1.6761, 2.2080], device='cuda:6'), covar=tensor([0.3165, 0.3470, 0.1630, 0.2903, 0.3094, 0.2662, 0.4269, 0.2400], device='cuda:6'), in_proj_covar=tensor([0.0292, 0.0247, 0.0227, 0.0316, 0.0218, 0.0231, 0.0229, 0.0186], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:6') 2023-04-27 10:13:48,140 INFO [finetune.py:976] (6/7) Epoch 17, batch 3850, loss[loss=0.1397, simple_loss=0.2126, pruned_loss=0.03345, over 4762.00 frames. ], tot_loss[loss=0.1796, simple_loss=0.2501, pruned_loss=0.05455, over 953015.12 frames. ], batch size: 26, lr: 3.38e-03, grad_scale: 32.0 2023-04-27 10:13:51,718 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([3.9710, 3.8806, 2.7956, 4.5959, 3.9991, 3.9425, 1.7301, 3.9999], device='cuda:6'), covar=tensor([0.1579, 0.1171, 0.3012, 0.1339, 0.2782, 0.1751, 0.5920, 0.2175], device='cuda:6'), in_proj_covar=tensor([0.0246, 0.0214, 0.0250, 0.0305, 0.0298, 0.0249, 0.0273, 0.0273], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 10:14:06,851 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7632, 1.2798, 4.8136, 4.5327, 4.1989, 4.5314, 4.3144, 4.2590], device='cuda:6'), covar=tensor([0.6605, 0.5976, 0.0958, 0.1540, 0.1094, 0.1424, 0.1727, 0.1476], device='cuda:6'), in_proj_covar=tensor([0.0307, 0.0304, 0.0400, 0.0404, 0.0347, 0.0404, 0.0309, 0.0363], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-27 10:14:20,428 INFO [finetune.py:976] (6/7) Epoch 17, batch 3900, loss[loss=0.155, simple_loss=0.2129, pruned_loss=0.04857, over 4822.00 frames. ], tot_loss[loss=0.1776, simple_loss=0.2473, pruned_loss=0.05398, over 955134.62 frames. ], batch size: 30, lr: 3.38e-03, grad_scale: 32.0 2023-04-27 10:14:27,546 INFO [optim.py:369] (6/7) 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 10:14:49,429 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6160, 2.0887, 1.7900, 2.0321, 1.5764, 1.6710, 1.6613, 1.2833], device='cuda:6'), covar=tensor([0.1812, 0.1272, 0.0825, 0.1009, 0.3265, 0.1146, 0.1909, 0.2678], device='cuda:6'), in_proj_covar=tensor([0.0283, 0.0304, 0.0217, 0.0277, 0.0311, 0.0257, 0.0247, 0.0264], device='cuda:6'), out_proj_covar=tensor([1.1355e-04, 1.2081e-04, 8.6363e-05, 1.0982e-04, 1.2623e-04, 1.0231e-04, 1.0011e-04, 1.0497e-04], device='cuda:6') 2023-04-27 10:14:52,354 INFO [finetune.py:976] (6/7) Epoch 17, batch 3950, loss[loss=0.1735, simple_loss=0.2395, pruned_loss=0.05374, over 4835.00 frames. ], tot_loss[loss=0.1748, simple_loss=0.2441, pruned_loss=0.05273, over 955501.53 frames. ], batch size: 49, lr: 3.38e-03, grad_scale: 32.0 2023-04-27 10:15:08,553 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=95616.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:15:09,197 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=95617.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:15:13,455 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=95624.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:15:31,267 INFO [finetune.py:976] (6/7) Epoch 17, batch 4000, loss[loss=0.154, simple_loss=0.2204, pruned_loss=0.04378, over 3977.00 frames. ], tot_loss[loss=0.1745, simple_loss=0.2433, pruned_loss=0.05283, over 956192.19 frames. ], batch size: 17, lr: 3.38e-03, grad_scale: 32.0 2023-04-27 10:15:33,197 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([0.8272, 1.0856, 3.2853, 3.0470, 2.9509, 3.1951, 3.1799, 2.8995], device='cuda:6'), covar=tensor([0.7433, 0.5615, 0.1466, 0.2144, 0.1574, 0.2475, 0.1642, 0.1822], device='cuda:6'), in_proj_covar=tensor([0.0309, 0.0306, 0.0403, 0.0405, 0.0349, 0.0405, 0.0309, 0.0365], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-27 10:15:43,870 INFO [optim.py:369] (6/7) 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 10:16:16,160 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=95678.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:16:36,029 INFO [finetune.py:976] (6/7) Epoch 17, batch 4050, loss[loss=0.1795, simple_loss=0.2547, pruned_loss=0.05213, over 4770.00 frames. ], tot_loss[loss=0.1767, simple_loss=0.2463, pruned_loss=0.05357, over 953858.35 frames. ], batch size: 59, lr: 3.38e-03, grad_scale: 32.0 2023-04-27 10:16:37,794 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=95695.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:17:18,760 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.80 vs. limit=2.0 2023-04-27 10:17:36,637 INFO [finetune.py:976] (6/7) Epoch 17, batch 4100, loss[loss=0.1434, simple_loss=0.2085, pruned_loss=0.0392, over 4738.00 frames. ], tot_loss[loss=0.1781, simple_loss=0.2483, pruned_loss=0.05391, over 953940.02 frames. ], batch size: 23, lr: 3.38e-03, grad_scale: 32.0 2023-04-27 10:17:43,699 INFO [optim.py:369] (6/7) 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 10:17:50,100 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=95761.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 10:18:09,896 INFO [finetune.py:976] (6/7) Epoch 17, batch 4150, loss[loss=0.1402, simple_loss=0.2181, pruned_loss=0.0311, over 4865.00 frames. ], tot_loss[loss=0.18, simple_loss=0.2505, pruned_loss=0.05475, over 954494.09 frames. ], batch size: 31, lr: 3.38e-03, grad_scale: 32.0 2023-04-27 10:18:21,839 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=95809.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 10:18:21,852 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([4.4890, 4.4082, 3.0708, 5.1947, 4.5400, 4.4314, 1.8328, 4.4425], device='cuda:6'), covar=tensor([0.1648, 0.0811, 0.3040, 0.0780, 0.2329, 0.1607, 0.5677, 0.1980], device='cuda:6'), in_proj_covar=tensor([0.0245, 0.0214, 0.0249, 0.0305, 0.0299, 0.0250, 0.0272, 0.0273], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 10:18:24,196 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=95812.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 10:18:43,506 INFO [finetune.py:976] (6/7) Epoch 17, batch 4200, loss[loss=0.1887, simple_loss=0.2556, pruned_loss=0.06089, over 4811.00 frames. ], tot_loss[loss=0.1799, simple_loss=0.251, pruned_loss=0.05441, over 953229.84 frames. ], batch size: 40, lr: 3.38e-03, grad_scale: 32.0 2023-04-27 10:18:46,048 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=95847.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:18:46,675 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.1974, 2.7825, 1.0790, 1.4838, 2.2085, 1.2630, 3.8523, 2.0341], device='cuda:6'), covar=tensor([0.0688, 0.0668, 0.0799, 0.1340, 0.0494, 0.1046, 0.0244, 0.0586], device='cuda:6'), in_proj_covar=tensor([0.0051, 0.0065, 0.0048, 0.0046, 0.0050, 0.0052, 0.0074, 0.0051], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:6') 2023-04-27 10:18:49,669 INFO [optim.py:369] (6/7) 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 10:19:04,202 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=95873.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 10:19:05,411 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9291, 2.5439, 2.1112, 2.3610, 1.7783, 2.0481, 2.1315, 1.5863], device='cuda:6'), covar=tensor([0.2070, 0.1087, 0.0829, 0.0985, 0.3304, 0.1243, 0.1989, 0.2667], device='cuda:6'), in_proj_covar=tensor([0.0283, 0.0305, 0.0218, 0.0278, 0.0311, 0.0258, 0.0248, 0.0264], device='cuda:6'), 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:6') 2023-04-27 10:19:16,329 INFO [finetune.py:976] (6/7) Epoch 17, batch 4250, loss[loss=0.1916, simple_loss=0.2534, pruned_loss=0.0649, over 4792.00 frames. ], tot_loss[loss=0.178, simple_loss=0.249, pruned_loss=0.05348, over 954257.21 frames. ], batch size: 45, lr: 3.38e-03, grad_scale: 32.0 2023-04-27 10:19:19,443 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.2882, 1.3269, 3.8177, 3.5476, 3.3953, 3.6403, 3.6310, 3.3425], device='cuda:6'), covar=tensor([0.6723, 0.5569, 0.1281, 0.1882, 0.1283, 0.1816, 0.1617, 0.1538], device='cuda:6'), in_proj_covar=tensor([0.0309, 0.0306, 0.0401, 0.0405, 0.0347, 0.0405, 0.0309, 0.0364], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-27 10:19:26,043 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=95908.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 10:19:31,945 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=95916.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:19:36,874 INFO [zipformer.py:1188] (6/7) 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] (6/7) attn_weights_entropy = tensor([2.0133, 2.6620, 2.2391, 2.4648, 1.7968, 2.1468, 2.2018, 1.6058], device='cuda:6'), covar=tensor([0.2033, 0.1375, 0.0873, 0.1199, 0.3454, 0.1339, 0.1806, 0.2826], device='cuda:6'), in_proj_covar=tensor([0.0282, 0.0304, 0.0217, 0.0277, 0.0310, 0.0256, 0.0247, 0.0263], device='cuda:6'), 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:6') 2023-04-27 10:19:48,683 INFO [finetune.py:976] (6/7) Epoch 17, batch 4300, loss[loss=0.1792, simple_loss=0.2538, pruned_loss=0.05234, over 4871.00 frames. ], tot_loss[loss=0.1764, simple_loss=0.2466, pruned_loss=0.0531, over 954894.21 frames. ], batch size: 34, lr: 3.38e-03, grad_scale: 32.0 2023-04-27 10:19:54,840 INFO [optim.py:369] (6/7) 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 10:20:03,625 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=95964.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:20:08,435 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-04-27 10:20:09,300 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=95972.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:20:09,902 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=95973.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:20:14,624 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=95980.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:20:22,491 INFO [finetune.py:976] (6/7) Epoch 17, batch 4350, loss[loss=0.1922, simple_loss=0.2506, pruned_loss=0.0669, over 4937.00 frames. ], tot_loss[loss=0.1731, simple_loss=0.2431, pruned_loss=0.05156, over 955650.72 frames. ], batch size: 38, lr: 3.38e-03, grad_scale: 32.0 2023-04-27 10:20:23,773 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=95995.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:20:29,634 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.44 vs. limit=5.0 2023-04-27 10:21:06,651 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.0602, 1.9509, 1.7799, 1.7370, 2.1158, 1.7251, 2.6071, 1.5820], device='cuda:6'), covar=tensor([0.3742, 0.2041, 0.4723, 0.2990, 0.1791, 0.2538, 0.1546, 0.5013], device='cuda:6'), in_proj_covar=tensor([0.0334, 0.0342, 0.0421, 0.0351, 0.0375, 0.0376, 0.0365, 0.0414], device='cuda:6'), 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:6') 2023-04-27 10:21:07,901 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=96041.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:21:08,980 INFO [finetune.py:976] (6/7) Epoch 17, batch 4400, loss[loss=0.1791, simple_loss=0.2594, pruned_loss=0.04944, over 4900.00 frames. ], tot_loss[loss=0.1732, simple_loss=0.2434, pruned_loss=0.05154, over 953824.17 frames. ], batch size: 36, lr: 3.38e-03, grad_scale: 32.0 2023-04-27 10:21:09,043 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=96043.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:21:15,035 INFO [optim.py:369] (6/7) 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 10:21:34,171 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6265, 1.3391, 1.8223, 1.8340, 1.4321, 1.2538, 1.5058, 0.8461], device='cuda:6'), covar=tensor([0.0489, 0.0787, 0.0409, 0.0545, 0.0720, 0.1355, 0.0502, 0.0752], device='cuda:6'), in_proj_covar=tensor([0.0069, 0.0070, 0.0068, 0.0068, 0.0075, 0.0096, 0.0075, 0.0067], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-27 10:21:34,766 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 17, batch 4450, loss[loss=0.1897, simple_loss=0.2636, pruned_loss=0.05792, over 4862.00 frames. ], tot_loss[loss=0.1757, simple_loss=0.2465, pruned_loss=0.05247, over 954035.81 frames. ], batch size: 49, lr: 3.38e-03, grad_scale: 32.0 2023-04-27 10:22:53,808 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 17, batch 4500, loss[loss=0.1577, simple_loss=0.2337, pruned_loss=0.04092, over 4751.00 frames. ], tot_loss[loss=0.1789, simple_loss=0.2498, pruned_loss=0.05402, over 954578.13 frames. ], batch size: 27, lr: 3.37e-03, grad_scale: 32.0 2023-04-27 10:23:02,750 INFO [optim.py:369] (6/7) 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 10:23:16,182 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=96165.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:23:23,319 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=96168.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 10:23:58,146 INFO [finetune.py:976] (6/7) Epoch 17, batch 4550, loss[loss=0.2128, simple_loss=0.2834, pruned_loss=0.07108, over 4896.00 frames. ], tot_loss[loss=0.1805, simple_loss=0.2512, pruned_loss=0.05487, over 954299.32 frames. ], batch size: 32, lr: 3.37e-03, grad_scale: 32.0 2023-04-27 10:24:10,601 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=96203.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 10:24:20,526 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.82 vs. limit=5.0 2023-04-27 10:24:42,285 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 17, batch 4600, loss[loss=0.144, simple_loss=0.2163, pruned_loss=0.03584, over 4820.00 frames. ], tot_loss[loss=0.1797, simple_loss=0.2507, pruned_loss=0.05432, over 954135.45 frames. ], batch size: 25, lr: 3.37e-03, grad_scale: 32.0 2023-04-27 10:24:57,514 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.1841, 1.3301, 5.1840, 4.8960, 4.5518, 4.9868, 4.5977, 4.6213], device='cuda:6'), covar=tensor([0.6247, 0.6146, 0.0999, 0.1652, 0.1043, 0.1050, 0.1045, 0.1302], device='cuda:6'), in_proj_covar=tensor([0.0309, 0.0306, 0.0401, 0.0404, 0.0348, 0.0404, 0.0310, 0.0364], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-27 10:25:01,052 INFO [optim.py:369] (6/7) 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 10:25:13,318 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 17, batch 4650, loss[loss=0.1583, simple_loss=0.2319, pruned_loss=0.04232, over 4818.00 frames. ], tot_loss[loss=0.1776, simple_loss=0.248, pruned_loss=0.0536, over 953410.61 frames. ], batch size: 39, lr: 3.37e-03, grad_scale: 32.0 2023-04-27 10:25:39,950 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.3494, 1.3634, 1.4756, 1.6739, 1.6728, 1.3521, 0.9170, 1.5256], device='cuda:6'), covar=tensor([0.0911, 0.1277, 0.0891, 0.0632, 0.0708, 0.0912, 0.0948, 0.0653], device='cuda:6'), in_proj_covar=tensor([0.0190, 0.0201, 0.0182, 0.0172, 0.0177, 0.0181, 0.0152, 0.0180], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 10:25:45,407 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=96321.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:25:47,259 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=96324.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:25:56,373 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=96336.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:25:57,046 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.2224, 1.6205, 1.4726, 1.8338, 1.7513, 1.8608, 1.4751, 3.4044], device='cuda:6'), covar=tensor([0.0594, 0.0753, 0.0753, 0.1100, 0.0573, 0.0510, 0.0707, 0.0143], device='cuda:6'), in_proj_covar=tensor([0.0037, 0.0039, 0.0040, 0.0043, 0.0040, 0.0038, 0.0038, 0.0056], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:6') 2023-04-27 10:26:01,179 INFO [finetune.py:976] (6/7) Epoch 17, batch 4700, loss[loss=0.1615, simple_loss=0.2302, pruned_loss=0.0464, over 4908.00 frames. ], tot_loss[loss=0.1763, simple_loss=0.2457, pruned_loss=0.05345, over 952683.62 frames. ], batch size: 36, lr: 3.37e-03, grad_scale: 32.0 2023-04-27 10:26:07,579 INFO [optim.py:369] (6/7) 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 10:26:43,955 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=96385.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:26:48,552 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=96390.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:26:55,841 INFO [finetune.py:976] (6/7) Epoch 17, batch 4750, loss[loss=0.1893, simple_loss=0.2537, pruned_loss=0.0625, over 4896.00 frames. ], tot_loss[loss=0.1752, simple_loss=0.244, pruned_loss=0.05315, over 953885.09 frames. ], batch size: 35, lr: 3.37e-03, grad_scale: 32.0 2023-04-27 10:27:29,716 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-04-27 10:27:30,858 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.4250, 1.6526, 1.7751, 1.9066, 1.6582, 1.8032, 1.9002, 1.8232], device='cuda:6'), covar=tensor([0.4244, 0.5897, 0.4644, 0.4355, 0.6171, 0.7777, 0.5439, 0.5234], device='cuda:6'), in_proj_covar=tensor([0.0331, 0.0372, 0.0317, 0.0334, 0.0345, 0.0395, 0.0355, 0.0324], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 10:27:44,228 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.1781, 1.4206, 1.3251, 1.6746, 1.5540, 1.6376, 1.3470, 3.0212], device='cuda:6'), covar=tensor([0.0606, 0.0783, 0.0782, 0.1161, 0.0606, 0.0509, 0.0736, 0.0158], device='cuda:6'), in_proj_covar=tensor([0.0037, 0.0038, 0.0039, 0.0043, 0.0039, 0.0038, 0.0038, 0.0056], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:6') 2023-04-27 10:27:45,319 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=96436.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:27:50,992 INFO [finetune.py:976] (6/7) Epoch 17, batch 4800, loss[loss=0.1378, simple_loss=0.1981, pruned_loss=0.03875, over 4376.00 frames. ], tot_loss[loss=0.1775, simple_loss=0.2468, pruned_loss=0.05408, over 953219.77 frames. ], batch size: 19, lr: 3.37e-03, grad_scale: 32.0 2023-04-27 10:27:56,384 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=96451.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:27:57,501 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.4309, 1.3407, 1.8615, 1.7659, 1.3235, 1.1440, 1.5917, 0.9948], device='cuda:6'), covar=tensor([0.0598, 0.0789, 0.0440, 0.0688, 0.0783, 0.1231, 0.0654, 0.0823], device='cuda:6'), in_proj_covar=tensor([0.0068, 0.0069, 0.0068, 0.0067, 0.0074, 0.0095, 0.0074, 0.0067], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-27 10:27:57,982 INFO [optim.py:369] (6/7) 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 10:28:07,184 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=96468.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 10:28:34,487 INFO [finetune.py:976] (6/7) Epoch 17, batch 4850, loss[loss=0.1774, simple_loss=0.2675, pruned_loss=0.04369, over 4760.00 frames. ], tot_loss[loss=0.1788, simple_loss=0.2493, pruned_loss=0.05409, over 954646.25 frames. ], batch size: 54, lr: 3.37e-03, grad_scale: 32.0 2023-04-27 10:28:41,606 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=96503.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:28:49,916 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=96516.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 10:28:51,848 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.46 vs. limit=5.0 2023-04-27 10:28:52,925 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=96521.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:29:07,107 INFO [finetune.py:976] (6/7) Epoch 17, batch 4900, loss[loss=0.1914, simple_loss=0.2575, pruned_loss=0.06266, over 4858.00 frames. ], tot_loss[loss=0.1808, simple_loss=0.2517, pruned_loss=0.0549, over 957283.03 frames. ], batch size: 31, lr: 3.37e-03, grad_scale: 32.0 2023-04-27 10:29:11,354 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-04-27 10:29:13,971 INFO [zipformer.py:1188] (6/7) 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] (6/7) 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 10:29:39,401 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9018, 2.6294, 1.9930, 1.9333, 1.3878, 1.4070, 2.0724, 1.3427], device='cuda:6'), covar=tensor([0.1598, 0.1231, 0.1342, 0.1646, 0.2294, 0.1948, 0.0952, 0.1935], device='cuda:6'), in_proj_covar=tensor([0.0196, 0.0212, 0.0167, 0.0203, 0.0200, 0.0184, 0.0155, 0.0184], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:6') 2023-04-27 10:29:39,873 INFO [finetune.py:976] (6/7) Epoch 17, batch 4950, loss[loss=0.1821, simple_loss=0.2441, pruned_loss=0.06006, over 4865.00 frames. ], tot_loss[loss=0.181, simple_loss=0.2521, pruned_loss=0.05491, over 957609.85 frames. ], batch size: 34, lr: 3.37e-03, grad_scale: 32.0 2023-04-27 10:29:53,190 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.4348, 1.7413, 1.7990, 1.9487, 1.7595, 1.8312, 1.9072, 1.8199], device='cuda:6'), covar=tensor([0.4205, 0.6099, 0.5112, 0.4481, 0.6175, 0.7320, 0.5851, 0.5625], device='cuda:6'), in_proj_covar=tensor([0.0330, 0.0370, 0.0317, 0.0333, 0.0343, 0.0392, 0.0354, 0.0323], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 10:30:09,389 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=96636.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:30:13,516 INFO [finetune.py:976] (6/7) Epoch 17, batch 5000, loss[loss=0.1345, simple_loss=0.2072, pruned_loss=0.03088, over 4895.00 frames. ], tot_loss[loss=0.179, simple_loss=0.2503, pruned_loss=0.05383, over 956181.86 frames. ], batch size: 35, lr: 3.37e-03, grad_scale: 32.0 2023-04-27 10:30:14,059 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.99 vs. limit=2.0 2023-04-27 10:30:19,794 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.4352, 1.6233, 1.8437, 1.9512, 1.8284, 1.9651, 1.9407, 1.9025], device='cuda:6'), covar=tensor([0.3547, 0.5097, 0.4401, 0.4322, 0.5166, 0.6912, 0.4796, 0.4726], device='cuda:6'), in_proj_covar=tensor([0.0330, 0.0370, 0.0317, 0.0332, 0.0343, 0.0392, 0.0354, 0.0323], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 10:30:21,475 INFO [optim.py:369] (6/7) 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 10:30:38,903 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=96680.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:30:41,300 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=96684.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:30:46,663 INFO [finetune.py:976] (6/7) Epoch 17, batch 5050, loss[loss=0.1439, simple_loss=0.2164, pruned_loss=0.03567, over 4703.00 frames. ], tot_loss[loss=0.1776, simple_loss=0.2482, pruned_loss=0.05348, over 957780.67 frames. ], batch size: 23, lr: 3.37e-03, grad_scale: 32.0 2023-04-27 10:30:46,858 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.81 vs. limit=5.0 2023-04-27 10:30:49,206 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.1213, 1.4377, 1.3476, 1.7930, 1.6415, 1.6234, 1.4242, 2.4625], device='cuda:6'), covar=tensor([0.0623, 0.0795, 0.0814, 0.1141, 0.0582, 0.0442, 0.0719, 0.0212], device='cuda:6'), in_proj_covar=tensor([0.0037, 0.0039, 0.0040, 0.0043, 0.0040, 0.0038, 0.0038, 0.0056], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:6') 2023-04-27 10:31:09,864 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.22 vs. limit=5.0 2023-04-27 10:31:15,663 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 17, batch 5100, loss[loss=0.1586, simple_loss=0.2307, pruned_loss=0.04321, over 4855.00 frames. ], tot_loss[loss=0.1748, simple_loss=0.2446, pruned_loss=0.05252, over 957359.40 frames. ], batch size: 44, lr: 3.37e-03, grad_scale: 32.0 2023-04-27 10:31:21,738 INFO [zipformer.py:1188] (6/7) 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] (6/7) 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 10:31:27,603 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.4371, 1.9842, 2.4225, 3.0026, 2.8665, 2.3213, 1.8932, 2.5106], device='cuda:6'), covar=tensor([0.0883, 0.1203, 0.0699, 0.0595, 0.0576, 0.0858, 0.0806, 0.0639], device='cuda:6'), in_proj_covar=tensor([0.0189, 0.0202, 0.0182, 0.0172, 0.0177, 0.0181, 0.0153, 0.0180], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 10:31:54,165 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=96776.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:32:04,478 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=96784.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:32:14,428 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.67 vs. limit=2.0 2023-04-27 10:32:15,344 INFO [finetune.py:976] (6/7) Epoch 17, batch 5150, loss[loss=0.2437, simple_loss=0.3078, pruned_loss=0.08975, over 4829.00 frames. ], tot_loss[loss=0.1757, simple_loss=0.2451, pruned_loss=0.05317, over 955270.14 frames. ], batch size: 39, lr: 3.37e-03, grad_scale: 32.0 2023-04-27 10:32:46,401 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=96821.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:33:03,928 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=96837.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:33:12,842 INFO [finetune.py:976] (6/7) Epoch 17, batch 5200, loss[loss=0.2199, simple_loss=0.2975, pruned_loss=0.07113, over 4933.00 frames. ], tot_loss[loss=0.1789, simple_loss=0.2491, pruned_loss=0.05439, over 956369.56 frames. ], batch size: 33, lr: 3.37e-03, grad_scale: 32.0 2023-04-27 10:33:24,746 INFO [optim.py:369] (6/7) 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 10:33:47,097 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=96869.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:33:56,891 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8532, 1.5327, 1.3988, 1.7023, 2.1583, 1.6905, 1.4340, 1.2555], device='cuda:6'), covar=tensor([0.1495, 0.1597, 0.2131, 0.1277, 0.0720, 0.1580, 0.2187, 0.2462], device='cuda:6'), in_proj_covar=tensor([0.0306, 0.0311, 0.0350, 0.0287, 0.0326, 0.0308, 0.0298, 0.0365], device='cuda:6'), out_proj_covar=tensor([6.3265e-05, 6.4822e-05, 7.4653e-05, 5.8198e-05, 6.7802e-05, 6.4888e-05, 6.2878e-05, 7.7798e-05], device='cuda:6') 2023-04-27 10:33:59,110 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-04-27 10:34:02,259 INFO [finetune.py:976] (6/7) Epoch 17, batch 5250, loss[loss=0.1681, simple_loss=0.251, pruned_loss=0.04261, over 4885.00 frames. ], tot_loss[loss=0.1801, simple_loss=0.2512, pruned_loss=0.05446, over 957379.39 frames. ], batch size: 35, lr: 3.37e-03, grad_scale: 32.0 2023-04-27 10:34:17,714 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-04-27 10:34:35,558 INFO [finetune.py:976] (6/7) Epoch 17, batch 5300, loss[loss=0.1595, simple_loss=0.2399, pruned_loss=0.03956, over 4730.00 frames. ], tot_loss[loss=0.1801, simple_loss=0.2517, pruned_loss=0.05421, over 958936.45 frames. ], batch size: 27, lr: 3.37e-03, grad_scale: 32.0 2023-04-27 10:34:39,384 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.1751, 1.7971, 2.0316, 2.4501, 2.0730, 1.6443, 1.5288, 1.8992], device='cuda:6'), covar=tensor([0.2395, 0.2604, 0.1401, 0.1792, 0.2342, 0.2354, 0.4251, 0.1827], device='cuda:6'), in_proj_covar=tensor([0.0291, 0.0247, 0.0227, 0.0316, 0.0218, 0.0231, 0.0228, 0.0185], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:6') 2023-04-27 10:34:41,675 INFO [optim.py:369] (6/7) 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 10:34:43,095 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.63 vs. limit=2.0 2023-04-27 10:35:00,619 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 17, batch 5350, loss[loss=0.1843, simple_loss=0.2558, pruned_loss=0.05638, over 4733.00 frames. ], tot_loss[loss=0.1806, simple_loss=0.2521, pruned_loss=0.0545, over 956908.84 frames. ], batch size: 54, lr: 3.37e-03, grad_scale: 32.0 2023-04-27 10:35:13,334 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=97000.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:35:15,231 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-04-27 10:35:33,057 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 17, batch 5400, loss[loss=0.179, simple_loss=0.2476, pruned_loss=0.05517, over 4823.00 frames. ], tot_loss[loss=0.1782, simple_loss=0.2491, pruned_loss=0.05364, over 957155.55 frames. ], batch size: 39, lr: 3.37e-03, grad_scale: 32.0 2023-04-27 10:35:44,602 INFO [zipformer.py:1188] (6/7) 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] (6/7) 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 10:35:53,796 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=97061.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:36:15,420 INFO [finetune.py:976] (6/7) Epoch 17, batch 5450, loss[loss=0.1578, simple_loss=0.2259, pruned_loss=0.04485, over 4695.00 frames. ], tot_loss[loss=0.1769, simple_loss=0.247, pruned_loss=0.05347, over 956959.03 frames. ], batch size: 23, lr: 3.37e-03, grad_scale: 32.0 2023-04-27 10:36:16,112 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=97094.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:36:21,641 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5704, 1.5334, 1.9639, 1.9446, 1.4726, 1.3633, 1.6343, 0.9675], device='cuda:6'), covar=tensor([0.0506, 0.0771, 0.0346, 0.0631, 0.0681, 0.1089, 0.0590, 0.0674], device='cuda:6'), in_proj_covar=tensor([0.0068, 0.0069, 0.0068, 0.0068, 0.0075, 0.0096, 0.0074, 0.0067], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-27 10:36:40,619 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=97132.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:36:41,270 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.2024, 2.5750, 0.9476, 1.5642, 1.6363, 2.0409, 1.6572, 0.9078], device='cuda:6'), covar=tensor([0.1463, 0.1161, 0.1608, 0.1271, 0.1038, 0.0831, 0.1718, 0.1586], device='cuda:6'), in_proj_covar=tensor([0.0117, 0.0243, 0.0138, 0.0121, 0.0133, 0.0153, 0.0119, 0.0121], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-27 10:36:47,755 INFO [finetune.py:976] (6/7) Epoch 17, batch 5500, loss[loss=0.1823, simple_loss=0.2507, pruned_loss=0.05693, over 4851.00 frames. ], tot_loss[loss=0.174, simple_loss=0.2435, pruned_loss=0.05228, over 955673.05 frames. ], batch size: 47, lr: 3.37e-03, grad_scale: 32.0 2023-04-27 10:36:54,343 INFO [optim.py:369] (6/7) 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 10:37:01,113 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.1379, 1.3829, 1.2355, 1.3316, 1.1518, 1.2119, 1.1701, 0.9377], device='cuda:6'), covar=tensor([0.1591, 0.1361, 0.1090, 0.1315, 0.3200, 0.1200, 0.1645, 0.2078], device='cuda:6'), in_proj_covar=tensor([0.0282, 0.0303, 0.0217, 0.0277, 0.0309, 0.0256, 0.0247, 0.0264], device='cuda:6'), out_proj_covar=tensor([1.1335e-04, 1.2038e-04, 8.6215e-05, 1.0994e-04, 1.2546e-04, 1.0190e-04, 9.9984e-05, 1.0469e-04], device='cuda:6') 2023-04-27 10:37:25,738 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8206, 1.4621, 1.7367, 2.1454, 2.1299, 1.7134, 1.4531, 1.9986], device='cuda:6'), covar=tensor([0.0759, 0.1240, 0.0725, 0.0484, 0.0563, 0.0784, 0.0785, 0.0527], device='cuda:6'), in_proj_covar=tensor([0.0188, 0.0200, 0.0180, 0.0170, 0.0176, 0.0179, 0.0151, 0.0178], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 10:37:37,875 INFO [finetune.py:976] (6/7) Epoch 17, batch 5550, loss[loss=0.1901, simple_loss=0.261, pruned_loss=0.05961, over 4828.00 frames. ], tot_loss[loss=0.1744, simple_loss=0.2441, pruned_loss=0.05236, over 955694.24 frames. ], batch size: 39, lr: 3.37e-03, grad_scale: 32.0 2023-04-27 10:38:03,800 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0438, 2.0044, 2.5118, 2.7908, 1.9937, 1.7046, 2.1428, 1.1708], device='cuda:6'), covar=tensor([0.0603, 0.0697, 0.0390, 0.0507, 0.0669, 0.1143, 0.0600, 0.0797], device='cuda:6'), in_proj_covar=tensor([0.0069, 0.0070, 0.0068, 0.0068, 0.0075, 0.0096, 0.0074, 0.0067], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-27 10:38:15,507 INFO [finetune.py:976] (6/7) Epoch 17, batch 5600, loss[loss=0.2027, simple_loss=0.282, pruned_loss=0.06174, over 4803.00 frames. ], tot_loss[loss=0.1767, simple_loss=0.2475, pruned_loss=0.05293, over 955450.23 frames. ], batch size: 51, lr: 3.37e-03, grad_scale: 32.0 2023-04-27 10:38:26,565 INFO [optim.py:369] (6/7) 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 10:38:37,083 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=97260.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:39:14,494 INFO [finetune.py:976] (6/7) Epoch 17, batch 5650, loss[loss=0.1575, simple_loss=0.2494, pruned_loss=0.03281, over 4760.00 frames. ], tot_loss[loss=0.1774, simple_loss=0.2492, pruned_loss=0.05274, over 954318.85 frames. ], batch size: 28, lr: 3.37e-03, grad_scale: 32.0 2023-04-27 10:39:22,751 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7817, 2.2481, 1.7764, 1.5934, 1.3021, 1.3048, 1.8683, 1.2609], device='cuda:6'), covar=tensor([0.1684, 0.1299, 0.1450, 0.1766, 0.2364, 0.1964, 0.0978, 0.2033], device='cuda:6'), in_proj_covar=tensor([0.0197, 0.0213, 0.0168, 0.0205, 0.0201, 0.0184, 0.0156, 0.0186], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:6') 2023-04-27 10:39:52,839 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=97321.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:39:54,291 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.11 vs. limit=2.0 2023-04-27 10:40:16,399 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=97342.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:40:16,916 INFO [finetune.py:976] (6/7) Epoch 17, batch 5700, loss[loss=0.151, simple_loss=0.2108, pruned_loss=0.04561, over 4113.00 frames. ], tot_loss[loss=0.1742, simple_loss=0.2449, pruned_loss=0.05173, over 933005.76 frames. ], batch size: 18, lr: 3.36e-03, grad_scale: 32.0 2023-04-27 10:40:19,380 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=97347.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:40:22,862 INFO [optim.py:369] (6/7) 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 10:40:24,679 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=97356.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:40:46,584 INFO [finetune.py:976] (6/7) Epoch 18, batch 0, loss[loss=0.218, simple_loss=0.2878, pruned_loss=0.07411, over 4818.00 frames. ], tot_loss[loss=0.218, simple_loss=0.2878, pruned_loss=0.07411, over 4818.00 frames. ], batch size: 47, lr: 3.36e-03, grad_scale: 64.0 2023-04-27 10:40:46,584 INFO [finetune.py:1001] (6/7) Computing validation loss 2023-04-27 10:41:03,166 INFO [finetune.py:1010] (6/7) 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] (6/7) Maximum memory allocated so far is 6365MB 2023-04-27 10:41:05,488 INFO [zipformer.py:1188] (6/7) 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] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=97403.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 10:41:32,012 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=97408.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:41:40,832 INFO [finetune.py:976] (6/7) Epoch 18, batch 50, loss[loss=0.1945, simple_loss=0.2616, pruned_loss=0.06374, over 4845.00 frames. ], tot_loss[loss=0.1824, simple_loss=0.2526, pruned_loss=0.05608, over 216060.50 frames. ], batch size: 44, lr: 3.36e-03, grad_scale: 64.0 2023-04-27 10:41:49,629 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=97432.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:41:51,455 INFO [zipformer.py:1188] (6/7) 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] (6/7) 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] (6/7) Epoch 18, batch 100, loss[loss=0.1416, simple_loss=0.2132, pruned_loss=0.03501, over 4816.00 frames. ], tot_loss[loss=0.1734, simple_loss=0.2437, pruned_loss=0.05156, over 381601.50 frames. ], batch size: 39, lr: 3.36e-03, grad_scale: 64.0 2023-04-27 10:42:22,017 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=97480.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:42:23,956 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.69 vs. limit=5.0 2023-04-27 10:42:25,740 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=3.36 vs. limit=5.0 2023-04-27 10:42:30,761 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-04-27 10:42:47,723 INFO [finetune.py:976] (6/7) Epoch 18, batch 150, loss[loss=0.1613, simple_loss=0.2296, pruned_loss=0.04646, over 4910.00 frames. ], tot_loss[loss=0.1714, simple_loss=0.2398, pruned_loss=0.05151, over 510047.03 frames. ], batch size: 37, lr: 3.36e-03, grad_scale: 64.0 2023-04-27 10:43:03,124 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([3.6501, 3.6172, 2.7323, 4.2653, 3.7137, 3.6813, 1.4576, 3.6234], device='cuda:6'), covar=tensor([0.1970, 0.1350, 0.3646, 0.1587, 0.3018, 0.1834, 0.6062, 0.2406], device='cuda:6'), in_proj_covar=tensor([0.0243, 0.0212, 0.0247, 0.0303, 0.0297, 0.0249, 0.0272, 0.0271], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 10:43:08,511 INFO [optim.py:369] (6/7) 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 10:43:09,868 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 18, batch 200, loss[loss=0.1828, simple_loss=0.2481, pruned_loss=0.05878, over 4865.00 frames. ], tot_loss[loss=0.1741, simple_loss=0.2421, pruned_loss=0.05309, over 609674.93 frames. ], batch size: 49, lr: 3.36e-03, grad_scale: 64.0 2023-04-27 10:43:55,466 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=97616.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:43:55,508 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 18, batch 250, loss[loss=0.1703, simple_loss=0.2505, pruned_loss=0.04504, over 4899.00 frames. ], tot_loss[loss=0.177, simple_loss=0.2462, pruned_loss=0.05393, over 686866.57 frames. ], batch size: 43, lr: 3.36e-03, grad_scale: 64.0 2023-04-27 10:44:42,424 INFO [optim.py:369] (6/7) 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 10:44:42,874 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-04-27 10:44:49,661 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=97656.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:44:59,561 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.3223, 1.8693, 2.3204, 2.6144, 2.3082, 1.8104, 1.4311, 2.1115], device='cuda:6'), covar=tensor([0.3536, 0.3251, 0.1733, 0.2515, 0.2672, 0.2758, 0.4247, 0.2046], device='cuda:6'), in_proj_covar=tensor([0.0291, 0.0246, 0.0227, 0.0315, 0.0218, 0.0232, 0.0228, 0.0185], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:6') 2023-04-27 10:45:03,746 INFO [finetune.py:976] (6/7) Epoch 18, batch 300, loss[loss=0.1784, simple_loss=0.2517, pruned_loss=0.05253, over 4772.00 frames. ], tot_loss[loss=0.179, simple_loss=0.2495, pruned_loss=0.05425, over 748514.60 frames. ], batch size: 27, lr: 3.36e-03, grad_scale: 64.0 2023-04-27 10:45:09,980 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=97674.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 10:45:43,988 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=97698.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 10:45:47,054 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=97703.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:45:47,659 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=97704.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:46:07,660 INFO [finetune.py:976] (6/7) Epoch 18, batch 350, loss[loss=0.1855, simple_loss=0.2687, pruned_loss=0.05119, over 4791.00 frames. ], tot_loss[loss=0.1821, simple_loss=0.2528, pruned_loss=0.05569, over 792120.77 frames. ], batch size: 59, lr: 3.36e-03, grad_scale: 64.0 2023-04-27 10:46:18,798 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=97730.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:46:28,151 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=97735.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 10:46:38,870 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.2372, 1.6418, 2.0875, 2.6578, 2.0737, 1.6427, 1.4017, 1.8780], device='cuda:6'), covar=tensor([0.3525, 0.3558, 0.1949, 0.2381, 0.2691, 0.2891, 0.4390, 0.2135], device='cuda:6'), in_proj_covar=tensor([0.0292, 0.0247, 0.0228, 0.0316, 0.0218, 0.0232, 0.0229, 0.0186], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:6') 2023-04-27 10:46:51,633 INFO [optim.py:369] (6/7) 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 10:47:03,653 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.33 vs. limit=5.0 2023-04-27 10:47:08,343 INFO [finetune.py:976] (6/7) Epoch 18, batch 400, loss[loss=0.1534, simple_loss=0.2317, pruned_loss=0.03758, over 4859.00 frames. ], tot_loss[loss=0.1811, simple_loss=0.2528, pruned_loss=0.05468, over 828910.34 frames. ], batch size: 31, lr: 3.36e-03, grad_scale: 64.0 2023-04-27 10:47:17,219 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-04-27 10:47:19,716 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([3.3411, 3.2364, 2.4049, 3.8794, 3.3552, 3.3341, 1.3120, 3.3100], device='cuda:6'), covar=tensor([0.1908, 0.1262, 0.3255, 0.2013, 0.2620, 0.2012, 0.5691, 0.2393], device='cuda:6'), in_proj_covar=tensor([0.0237, 0.0208, 0.0243, 0.0297, 0.0291, 0.0244, 0.0266, 0.0265], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 10:47:42,005 INFO [finetune.py:976] (6/7) Epoch 18, batch 450, loss[loss=0.1584, simple_loss=0.2321, pruned_loss=0.0423, over 4771.00 frames. ], tot_loss[loss=0.1787, simple_loss=0.2504, pruned_loss=0.05348, over 858383.41 frames. ], batch size: 28, lr: 3.36e-03, grad_scale: 32.0 2023-04-27 10:48:05,036 INFO [optim.py:369] (6/7) 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 10:48:15,397 INFO [finetune.py:976] (6/7) Epoch 18, batch 500, loss[loss=0.1791, simple_loss=0.2453, pruned_loss=0.05645, over 4825.00 frames. ], tot_loss[loss=0.1774, simple_loss=0.2484, pruned_loss=0.05319, over 878865.14 frames. ], batch size: 41, lr: 3.36e-03, grad_scale: 32.0 2023-04-27 10:48:22,058 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.1460, 3.1970, 0.9682, 1.5674, 1.5403, 2.2055, 1.8287, 1.1038], device='cuda:6'), covar=tensor([0.1953, 0.1366, 0.2303, 0.1813, 0.1450, 0.1330, 0.1627, 0.2178], device='cuda:6'), in_proj_covar=tensor([0.0116, 0.0241, 0.0136, 0.0120, 0.0132, 0.0152, 0.0117, 0.0119], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-27 10:48:26,733 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=97888.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:48:42,855 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=97911.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:48:45,975 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=97916.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:48:48,919 INFO [finetune.py:976] (6/7) Epoch 18, batch 550, loss[loss=0.1595, simple_loss=0.2344, pruned_loss=0.04233, over 4818.00 frames. ], tot_loss[loss=0.1751, simple_loss=0.2451, pruned_loss=0.05255, over 895307.67 frames. ], batch size: 33, lr: 3.36e-03, grad_scale: 32.0 2023-04-27 10:49:09,142 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=97949.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:49:10,968 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7908, 2.4978, 1.6725, 1.8617, 1.3000, 1.3300, 1.7648, 1.2089], device='cuda:6'), covar=tensor([0.1883, 0.1362, 0.1750, 0.1796, 0.2591, 0.2291, 0.1171, 0.2311], device='cuda:6'), in_proj_covar=tensor([0.0197, 0.0212, 0.0168, 0.0204, 0.0199, 0.0184, 0.0156, 0.0185], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:6') 2023-04-27 10:49:12,032 INFO [optim.py:369] (6/7) 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 10:49:18,170 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=97964.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:49:22,479 INFO [finetune.py:976] (6/7) Epoch 18, batch 600, loss[loss=0.2056, simple_loss=0.272, pruned_loss=0.06957, over 4869.00 frames. ], tot_loss[loss=0.1752, simple_loss=0.2452, pruned_loss=0.05263, over 910356.02 frames. ], batch size: 34, lr: 3.36e-03, grad_scale: 32.0 2023-04-27 10:49:41,026 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=97998.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:49:46,921 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=98003.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:49:52,549 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=5.01 vs. limit=5.0 2023-04-27 10:49:57,857 INFO [finetune.py:976] (6/7) Epoch 18, batch 650, loss[loss=0.1933, simple_loss=0.2593, pruned_loss=0.0636, over 4821.00 frames. ], tot_loss[loss=0.1758, simple_loss=0.2465, pruned_loss=0.05258, over 921209.67 frames. ], batch size: 40, lr: 3.36e-03, grad_scale: 32.0 2023-04-27 10:50:03,546 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=98030.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 10:50:03,559 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=98030.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:50:14,395 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=98046.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:50:18,390 INFO [zipformer.py:1188] (6/7) 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] (6/7) 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 10:50:48,620 INFO [finetune.py:976] (6/7) Epoch 18, batch 700, loss[loss=0.1959, simple_loss=0.2655, pruned_loss=0.06318, over 4817.00 frames. ], tot_loss[loss=0.1775, simple_loss=0.2485, pruned_loss=0.05323, over 929881.05 frames. ], batch size: 33, lr: 3.36e-03, grad_scale: 32.0 2023-04-27 10:50:58,818 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=98078.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:51:11,481 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8830, 2.3454, 1.9102, 2.2500, 1.6733, 1.9514, 1.8971, 1.4153], device='cuda:6'), covar=tensor([0.2030, 0.1284, 0.0939, 0.1345, 0.3321, 0.1379, 0.1972, 0.2999], device='cuda:6'), in_proj_covar=tensor([0.0288, 0.0308, 0.0221, 0.0282, 0.0315, 0.0261, 0.0252, 0.0268], device='cuda:6'), out_proj_covar=tensor([1.1564e-04, 1.2221e-04, 8.7896e-05, 1.1203e-04, 1.2817e-04, 1.0378e-04, 1.0199e-04, 1.0630e-04], device='cuda:6') 2023-04-27 10:51:21,799 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.67 vs. limit=2.0 2023-04-27 10:51:25,170 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0251, 1.7454, 2.2115, 2.5557, 2.0512, 2.0011, 2.0814, 2.0696], device='cuda:6'), covar=tensor([0.5120, 0.7295, 0.7861, 0.6107, 0.6477, 0.9482, 0.9236, 1.0366], device='cuda:6'), in_proj_covar=tensor([0.0420, 0.0407, 0.0496, 0.0503, 0.0450, 0.0477, 0.0483, 0.0486], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 10:51:31,110 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0038, 2.6529, 1.8698, 2.0095, 1.4137, 1.4051, 1.9486, 1.3423], device='cuda:6'), covar=tensor([0.1727, 0.1474, 0.1574, 0.1838, 0.2437, 0.2041, 0.1050, 0.2108], device='cuda:6'), in_proj_covar=tensor([0.0197, 0.0213, 0.0168, 0.0205, 0.0200, 0.0184, 0.0156, 0.0185], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:6') 2023-04-27 10:51:33,948 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=98106.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:51:54,809 INFO [finetune.py:976] (6/7) Epoch 18, batch 750, loss[loss=0.2197, simple_loss=0.3028, pruned_loss=0.06829, over 4885.00 frames. ], tot_loss[loss=0.1785, simple_loss=0.2499, pruned_loss=0.05355, over 934262.16 frames. ], batch size: 43, lr: 3.36e-03, grad_scale: 32.0 2023-04-27 10:52:31,990 INFO [optim.py:369] (6/7) 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 10:52:42,194 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 18, batch 800, loss[loss=0.1667, simple_loss=0.2318, pruned_loss=0.05083, over 4883.00 frames. ], tot_loss[loss=0.1771, simple_loss=0.2487, pruned_loss=0.05275, over 940393.72 frames. ], batch size: 32, lr: 3.36e-03, grad_scale: 32.0 2023-04-27 10:53:00,506 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8452, 1.3971, 1.9019, 2.3324, 1.9328, 1.8462, 1.8656, 1.8249], device='cuda:6'), covar=tensor([0.4444, 0.6605, 0.6255, 0.5439, 0.5823, 0.7222, 0.7695, 0.8611], device='cuda:6'), in_proj_covar=tensor([0.0421, 0.0407, 0.0497, 0.0503, 0.0451, 0.0477, 0.0484, 0.0488], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 10:53:10,127 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.66 vs. limit=2.0 2023-04-27 10:53:10,920 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=98211.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:53:17,965 INFO [finetune.py:976] (6/7) Epoch 18, batch 850, loss[loss=0.155, simple_loss=0.2217, pruned_loss=0.04412, over 4827.00 frames. ], tot_loss[loss=0.1755, simple_loss=0.2466, pruned_loss=0.05218, over 944772.66 frames. ], batch size: 25, lr: 3.35e-03, grad_scale: 32.0 2023-04-27 10:53:32,038 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=98244.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:53:38,451 INFO [optim.py:369] (6/7) 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 10:53:42,078 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=98259.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:53:51,235 INFO [finetune.py:976] (6/7) Epoch 18, batch 900, loss[loss=0.1505, simple_loss=0.2307, pruned_loss=0.03512, over 4881.00 frames. ], tot_loss[loss=0.1735, simple_loss=0.2441, pruned_loss=0.05145, over 947279.02 frames. ], batch size: 34, lr: 3.35e-03, grad_scale: 32.0 2023-04-27 10:54:24,149 INFO [finetune.py:976] (6/7) Epoch 18, batch 950, loss[loss=0.188, simple_loss=0.2593, pruned_loss=0.05832, over 4911.00 frames. ], tot_loss[loss=0.1732, simple_loss=0.2429, pruned_loss=0.05173, over 948340.10 frames. ], batch size: 36, lr: 3.35e-03, grad_scale: 32.0 2023-04-27 10:54:30,300 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=98330.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 10:54:44,940 INFO [optim.py:369] (6/7) 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 10:54:46,282 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.4243, 1.2281, 4.0766, 3.7789, 3.6076, 3.8608, 3.8288, 3.5832], device='cuda:6'), covar=tensor([0.7186, 0.6113, 0.1246, 0.2001, 0.1203, 0.1835, 0.1886, 0.1766], device='cuda:6'), in_proj_covar=tensor([0.0309, 0.0306, 0.0403, 0.0407, 0.0348, 0.0405, 0.0312, 0.0365], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-27 10:54:57,897 INFO [finetune.py:976] (6/7) Epoch 18, batch 1000, loss[loss=0.2153, simple_loss=0.2951, pruned_loss=0.06769, over 4904.00 frames. ], tot_loss[loss=0.1751, simple_loss=0.2453, pruned_loss=0.05247, over 951823.78 frames. ], batch size: 35, lr: 3.35e-03, grad_scale: 32.0 2023-04-27 10:54:58,026 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7818, 2.1285, 1.8374, 2.0458, 1.7616, 1.8965, 1.8200, 1.3959], device='cuda:6'), covar=tensor([0.1571, 0.1185, 0.0814, 0.1007, 0.2926, 0.0995, 0.1521, 0.2217], device='cuda:6'), in_proj_covar=tensor([0.0288, 0.0307, 0.0221, 0.0281, 0.0316, 0.0261, 0.0252, 0.0267], device='cuda:6'), out_proj_covar=tensor([1.1565e-04, 1.2211e-04, 8.7914e-05, 1.1157e-04, 1.2817e-04, 1.0381e-04, 1.0184e-04, 1.0628e-04], device='cuda:6') 2023-04-27 10:55:02,793 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=98378.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 10:55:22,918 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=98410.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:55:28,852 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.0737, 1.9685, 1.7393, 1.6571, 2.1676, 1.7490, 2.5606, 1.5200], device='cuda:6'), covar=tensor([0.3975, 0.2158, 0.5333, 0.3488, 0.1803, 0.2715, 0.1436, 0.5180], device='cuda:6'), in_proj_covar=tensor([0.0341, 0.0349, 0.0430, 0.0357, 0.0383, 0.0382, 0.0373, 0.0422], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 10:55:29,958 INFO [finetune.py:976] (6/7) Epoch 18, batch 1050, loss[loss=0.1758, simple_loss=0.2456, pruned_loss=0.05297, over 4829.00 frames. ], tot_loss[loss=0.1752, simple_loss=0.2465, pruned_loss=0.05193, over 950990.19 frames. ], batch size: 30, lr: 3.35e-03, grad_scale: 32.0 2023-04-27 10:55:49,754 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.2457, 2.1671, 1.8147, 1.8753, 2.2891, 1.8327, 2.7799, 1.5791], device='cuda:6'), covar=tensor([0.3838, 0.2005, 0.4686, 0.3459, 0.1962, 0.2785, 0.1753, 0.4766], device='cuda:6'), in_proj_covar=tensor([0.0338, 0.0345, 0.0427, 0.0354, 0.0380, 0.0379, 0.0370, 0.0418], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 10:55:51,555 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7131, 1.5028, 1.6191, 1.9666, 1.9412, 1.6281, 1.3099, 1.8187], device='cuda:6'), covar=tensor([0.0770, 0.1085, 0.0797, 0.0601, 0.0630, 0.0798, 0.0806, 0.0581], device='cuda:6'), in_proj_covar=tensor([0.0186, 0.0199, 0.0179, 0.0170, 0.0176, 0.0178, 0.0150, 0.0176], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 10:55:52,032 INFO [optim.py:369] (6/7) 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 10:55:56,962 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=98462.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:56:07,900 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.1470, 0.7716, 0.9348, 0.7981, 1.2526, 0.9344, 0.8466, 0.9970], device='cuda:6'), covar=tensor([0.1852, 0.1790, 0.2150, 0.1779, 0.1103, 0.1718, 0.1823, 0.2656], device='cuda:6'), in_proj_covar=tensor([0.0309, 0.0314, 0.0353, 0.0289, 0.0330, 0.0311, 0.0301, 0.0369], device='cuda:6'), out_proj_covar=tensor([6.3965e-05, 6.5415e-05, 7.5074e-05, 5.8671e-05, 6.8634e-05, 6.5530e-05, 6.3408e-05, 7.8561e-05], device='cuda:6') 2023-04-27 10:56:08,986 INFO [finetune.py:976] (6/7) Epoch 18, batch 1100, loss[loss=0.1541, simple_loss=0.2335, pruned_loss=0.03738, over 4865.00 frames. ], tot_loss[loss=0.1766, simple_loss=0.248, pruned_loss=0.0526, over 952100.45 frames. ], batch size: 31, lr: 3.35e-03, grad_scale: 32.0 2023-04-27 10:56:09,100 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=98471.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:56:32,950 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9601, 2.5102, 2.0535, 2.3105, 1.7294, 2.2017, 2.0309, 1.6236], device='cuda:6'), covar=tensor([0.2232, 0.1398, 0.0891, 0.1352, 0.3629, 0.1246, 0.2002, 0.2777], device='cuda:6'), in_proj_covar=tensor([0.0287, 0.0307, 0.0221, 0.0281, 0.0315, 0.0261, 0.0251, 0.0267], device='cuda:6'), out_proj_covar=tensor([1.1516e-04, 1.2192e-04, 8.7788e-05, 1.1134e-04, 1.2796e-04, 1.0357e-04, 1.0144e-04, 1.0616e-04], device='cuda:6') 2023-04-27 10:57:13,274 INFO [finetune.py:976] (6/7) Epoch 18, batch 1150, loss[loss=0.1564, simple_loss=0.2301, pruned_loss=0.04138, over 4726.00 frames. ], tot_loss[loss=0.1771, simple_loss=0.2485, pruned_loss=0.05285, over 951501.38 frames. ], batch size: 23, lr: 3.35e-03, grad_scale: 32.0 2023-04-27 10:57:23,768 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.3655, 1.2644, 1.5838, 1.5991, 1.3199, 1.1601, 1.3832, 0.9487], device='cuda:6'), covar=tensor([0.0566, 0.0562, 0.0431, 0.0460, 0.0636, 0.0877, 0.0509, 0.0586], device='cuda:6'), in_proj_covar=tensor([0.0068, 0.0069, 0.0068, 0.0067, 0.0075, 0.0096, 0.0074, 0.0067], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-27 10:57:45,477 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=98543.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:57:46,072 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=98544.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:57:57,925 INFO [optim.py:369] (6/7) 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 10:58:07,260 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.58 vs. limit=2.0 2023-04-27 10:58:14,187 INFO [finetune.py:976] (6/7) Epoch 18, batch 1200, loss[loss=0.1661, simple_loss=0.2389, pruned_loss=0.04664, over 4799.00 frames. ], tot_loss[loss=0.1767, simple_loss=0.2478, pruned_loss=0.05283, over 951824.71 frames. ], batch size: 51, lr: 3.35e-03, grad_scale: 32.0 2023-04-27 10:58:24,354 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=3.53 vs. limit=5.0 2023-04-27 10:58:28,911 INFO [zipformer.py:1188] (6/7) 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] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=98604.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 10:58:40,357 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.33 vs. limit=5.0 2023-04-27 10:58:46,674 INFO [finetune.py:976] (6/7) Epoch 18, batch 1250, loss[loss=0.1672, simple_loss=0.2279, pruned_loss=0.05318, over 4903.00 frames. ], tot_loss[loss=0.1758, simple_loss=0.2459, pruned_loss=0.05287, over 955032.91 frames. ], batch size: 32, lr: 3.35e-03, grad_scale: 32.0 2023-04-27 10:59:10,194 INFO [optim.py:369] (6/7) 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 10:59:12,459 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.96 vs. limit=2.0 2023-04-27 10:59:20,472 INFO [finetune.py:976] (6/7) Epoch 18, batch 1300, loss[loss=0.1847, simple_loss=0.2555, pruned_loss=0.05692, over 4761.00 frames. ], tot_loss[loss=0.1733, simple_loss=0.2432, pruned_loss=0.05169, over 955082.67 frames. ], batch size: 28, lr: 3.35e-03, grad_scale: 32.0 2023-04-27 10:59:24,906 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-04-27 10:59:41,194 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.50 vs. limit=2.0 2023-04-27 10:59:53,699 INFO [finetune.py:976] (6/7) Epoch 18, batch 1350, loss[loss=0.1501, simple_loss=0.2306, pruned_loss=0.03479, over 4892.00 frames. ], tot_loss[loss=0.1727, simple_loss=0.2429, pruned_loss=0.0513, over 955414.67 frames. ], batch size: 35, lr: 3.35e-03, grad_scale: 32.0 2023-04-27 10:59:54,449 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.3366, 1.7834, 1.5703, 2.1952, 2.4562, 1.9763, 1.8743, 1.6668], device='cuda:6'), covar=tensor([0.1378, 0.1620, 0.1905, 0.1480, 0.1158, 0.1770, 0.2031, 0.2422], device='cuda:6'), in_proj_covar=tensor([0.0308, 0.0313, 0.0352, 0.0289, 0.0330, 0.0310, 0.0300, 0.0367], device='cuda:6'), out_proj_covar=tensor([6.3775e-05, 6.5276e-05, 7.4762e-05, 5.8597e-05, 6.8518e-05, 6.5371e-05, 6.3257e-05, 7.8233e-05], device='cuda:6') 2023-04-27 11:00:00,237 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0831, 2.7595, 2.1111, 2.5911, 1.9950, 2.4192, 2.5140, 1.7130], device='cuda:6'), covar=tensor([0.2302, 0.1299, 0.0914, 0.1465, 0.3055, 0.1261, 0.2009, 0.2799], device='cuda:6'), in_proj_covar=tensor([0.0288, 0.0309, 0.0222, 0.0282, 0.0317, 0.0263, 0.0252, 0.0269], device='cuda:6'), out_proj_covar=tensor([1.1569e-04, 1.2268e-04, 8.8198e-05, 1.1199e-04, 1.2865e-04, 1.0428e-04, 1.0209e-04, 1.0670e-04], device='cuda:6') 2023-04-27 11:00:08,543 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.10 vs. limit=2.0 2023-04-27 11:00:17,035 INFO [optim.py:369] (6/7) 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 11:00:20,244 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9595, 1.9261, 2.2630, 2.4116, 1.8177, 1.6638, 2.0042, 1.1885], device='cuda:6'), covar=tensor([0.0618, 0.0835, 0.0408, 0.0753, 0.0845, 0.1126, 0.0746, 0.0813], device='cuda:6'), in_proj_covar=tensor([0.0068, 0.0069, 0.0068, 0.0067, 0.0075, 0.0095, 0.0074, 0.0067], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-27 11:00:22,051 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=98762.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:00:24,460 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=98766.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:00:25,692 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.0193, 1.2781, 1.2395, 1.5993, 1.4360, 1.4898, 1.2550, 2.4484], device='cuda:6'), covar=tensor([0.0666, 0.0903, 0.0889, 0.1261, 0.0711, 0.0521, 0.0829, 0.0254], device='cuda:6'), in_proj_covar=tensor([0.0037, 0.0039, 0.0040, 0.0044, 0.0040, 0.0038, 0.0038, 0.0056], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:6') 2023-04-27 11:00:27,465 INFO [finetune.py:976] (6/7) Epoch 18, batch 1400, loss[loss=0.1922, simple_loss=0.2756, pruned_loss=0.05438, over 4821.00 frames. ], tot_loss[loss=0.1758, simple_loss=0.2463, pruned_loss=0.05261, over 954671.53 frames. ], batch size: 38, lr: 3.35e-03, grad_scale: 32.0 2023-04-27 11:00:34,597 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.4599, 1.2707, 0.4916, 1.1653, 1.4067, 1.3360, 1.2167, 1.3057], device='cuda:6'), covar=tensor([0.0512, 0.0412, 0.0420, 0.0580, 0.0317, 0.0543, 0.0539, 0.0594], device='cuda:6'), in_proj_covar=tensor([0.0028, 0.0024, 0.0022, 0.0029, 0.0020, 0.0028, 0.0028, 0.0028], device='cuda:6'), out_proj_covar=tensor([0.0050, 0.0045, 0.0038, 0.0051, 0.0038, 0.0049, 0.0049, 0.0050], device='cuda:6') 2023-04-27 11:00:47,569 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.5920, 1.8011, 1.7233, 2.3952, 2.5216, 2.0713, 2.0112, 1.9107], device='cuda:6'), covar=tensor([0.1476, 0.1894, 0.1786, 0.1400, 0.1002, 0.2064, 0.2106, 0.2195], device='cuda:6'), in_proj_covar=tensor([0.0309, 0.0315, 0.0353, 0.0289, 0.0330, 0.0311, 0.0301, 0.0368], device='cuda:6'), out_proj_covar=tensor([6.3973e-05, 6.5522e-05, 7.5074e-05, 5.8575e-05, 6.8596e-05, 6.5450e-05, 6.3492e-05, 7.8434e-05], device='cuda:6') 2023-04-27 11:00:54,325 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=98810.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:01:00,975 INFO [finetune.py:976] (6/7) Epoch 18, batch 1450, loss[loss=0.1734, simple_loss=0.2473, pruned_loss=0.04968, over 4842.00 frames. ], tot_loss[loss=0.1766, simple_loss=0.2474, pruned_loss=0.05291, over 952418.92 frames. ], batch size: 49, lr: 3.35e-03, grad_scale: 32.0 2023-04-27 11:01:22,273 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.2871, 3.5066, 0.9486, 1.5860, 1.5637, 2.3801, 1.9641, 1.0102], device='cuda:6'), covar=tensor([0.2126, 0.1591, 0.2631, 0.2106, 0.1552, 0.1497, 0.1885, 0.2201], device='cuda:6'), in_proj_covar=tensor([0.0118, 0.0244, 0.0138, 0.0122, 0.0134, 0.0154, 0.0119, 0.0121], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-27 11:01:24,435 INFO [optim.py:369] (6/7) 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 11:01:29,504 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.44 vs. limit=5.0 2023-04-27 11:01:34,711 INFO [finetune.py:976] (6/7) Epoch 18, batch 1500, loss[loss=0.1794, simple_loss=0.2451, pruned_loss=0.05689, over 4710.00 frames. ], tot_loss[loss=0.1789, simple_loss=0.2497, pruned_loss=0.05402, over 952828.24 frames. ], batch size: 23, lr: 3.35e-03, grad_scale: 32.0 2023-04-27 11:01:52,826 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=98890.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:02:02,819 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=98895.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:02:11,049 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=98899.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:02:35,629 INFO [finetune.py:976] (6/7) Epoch 18, batch 1550, loss[loss=0.2085, simple_loss=0.2686, pruned_loss=0.07418, over 4888.00 frames. ], tot_loss[loss=0.1787, simple_loss=0.2502, pruned_loss=0.05355, over 953169.53 frames. ], batch size: 32, lr: 3.35e-03, grad_scale: 32.0 2023-04-27 11:03:17,464 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=98951.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:03:19,155 INFO [optim.py:369] (6/7) 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 11:03:25,935 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=98956.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:03:41,684 INFO [finetune.py:976] (6/7) Epoch 18, batch 1600, loss[loss=0.1486, simple_loss=0.2038, pruned_loss=0.04672, over 4925.00 frames. ], tot_loss[loss=0.1769, simple_loss=0.2478, pruned_loss=0.05297, over 954287.31 frames. ], batch size: 33, lr: 3.35e-03, grad_scale: 32.0 2023-04-27 11:03:41,824 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=98971.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:04:11,078 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.2082, 2.6905, 2.2808, 2.5257, 1.9590, 2.4355, 2.4755, 1.9827], device='cuda:6'), covar=tensor([0.1858, 0.1045, 0.0716, 0.1057, 0.2944, 0.0946, 0.1568, 0.2155], device='cuda:6'), in_proj_covar=tensor([0.0289, 0.0308, 0.0222, 0.0282, 0.0316, 0.0262, 0.0251, 0.0269], device='cuda:6'), out_proj_covar=tensor([1.1614e-04, 1.2255e-04, 8.8157e-05, 1.1202e-04, 1.2835e-04, 1.0406e-04, 1.0164e-04, 1.0701e-04], device='cuda:6') 2023-04-27 11:04:20,541 INFO [finetune.py:976] (6/7) Epoch 18, batch 1650, loss[loss=0.1682, simple_loss=0.233, pruned_loss=0.0517, over 3133.00 frames. ], tot_loss[loss=0.1758, simple_loss=0.2458, pruned_loss=0.05287, over 953160.65 frames. ], batch size: 13, lr: 3.35e-03, grad_scale: 32.0 2023-04-27 11:04:27,340 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=99032.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:04:40,845 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=99051.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:04:42,977 INFO [optim.py:369] (6/7) 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 11:04:50,876 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 18, batch 1700, loss[loss=0.1104, simple_loss=0.1855, pruned_loss=0.0177, over 4757.00 frames. ], tot_loss[loss=0.1758, simple_loss=0.2451, pruned_loss=0.05324, over 955505.77 frames. ], batch size: 28, lr: 3.35e-03, grad_scale: 32.0 2023-04-27 11:05:21,701 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=99112.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:05:22,865 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=99114.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:05:27,559 INFO [finetune.py:976] (6/7) Epoch 18, batch 1750, loss[loss=0.1827, simple_loss=0.2489, pruned_loss=0.05822, over 4797.00 frames. ], tot_loss[loss=0.1777, simple_loss=0.2469, pruned_loss=0.05428, over 955299.76 frames. ], batch size: 29, lr: 3.35e-03, grad_scale: 32.0 2023-04-27 11:05:50,008 INFO [optim.py:369] (6/7) 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 11:05:59,372 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.69 vs. limit=2.0 2023-04-27 11:06:01,218 INFO [finetune.py:976] (6/7) Epoch 18, batch 1800, loss[loss=0.1632, simple_loss=0.2442, pruned_loss=0.04114, over 4811.00 frames. ], tot_loss[loss=0.178, simple_loss=0.2486, pruned_loss=0.05375, over 955901.35 frames. ], batch size: 51, lr: 3.35e-03, grad_scale: 32.0 2023-04-27 11:06:07,731 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-04-27 11:06:19,263 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=99199.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:06:22,377 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.3927, 1.9080, 2.2843, 2.9390, 2.2807, 1.8280, 1.8354, 2.2159], device='cuda:6'), covar=tensor([0.2934, 0.2978, 0.1497, 0.2213, 0.2565, 0.2554, 0.3793, 0.2064], device='cuda:6'), in_proj_covar=tensor([0.0291, 0.0246, 0.0227, 0.0315, 0.0218, 0.0231, 0.0228, 0.0185], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:6') 2023-04-27 11:06:34,183 INFO [finetune.py:976] (6/7) Epoch 18, batch 1850, loss[loss=0.1886, simple_loss=0.2553, pruned_loss=0.06091, over 4825.00 frames. ], tot_loss[loss=0.181, simple_loss=0.2513, pruned_loss=0.05537, over 955092.30 frames. ], batch size: 30, lr: 3.35e-03, grad_scale: 32.0 2023-04-27 11:06:39,652 INFO [zipformer.py:1188] (6/7) 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] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=99246.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:06:50,538 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=99247.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:06:53,934 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=99251.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:06:55,189 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.1975, 2.0822, 1.7344, 1.7654, 2.1868, 1.7842, 2.4640, 1.5131], device='cuda:6'), covar=tensor([0.3314, 0.1466, 0.4007, 0.2867, 0.1566, 0.2160, 0.1514, 0.4214], device='cuda:6'), in_proj_covar=tensor([0.0340, 0.0348, 0.0429, 0.0355, 0.0383, 0.0381, 0.0372, 0.0422], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 11:06:55,635 INFO [optim.py:369] (6/7) 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 11:07:06,164 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=99268.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:07:07,879 INFO [finetune.py:976] (6/7) Epoch 18, batch 1900, loss[loss=0.1662, simple_loss=0.2391, pruned_loss=0.04667, over 4774.00 frames. ], tot_loss[loss=0.1808, simple_loss=0.2517, pruned_loss=0.05496, over 953841.06 frames. ], batch size: 26, lr: 3.35e-03, grad_scale: 32.0 2023-04-27 11:07:12,675 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0398, 1.3947, 5.1884, 4.8027, 4.5817, 4.9979, 4.5917, 4.6158], device='cuda:6'), covar=tensor([0.6856, 0.6116, 0.0950, 0.1745, 0.0954, 0.1508, 0.1176, 0.1409], device='cuda:6'), in_proj_covar=tensor([0.0306, 0.0306, 0.0401, 0.0405, 0.0349, 0.0403, 0.0310, 0.0364], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-27 11:07:20,133 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=99290.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:08:01,436 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.2148, 2.5873, 2.1468, 2.3819, 1.8538, 2.2550, 2.2033, 1.7146], device='cuda:6'), covar=tensor([0.1786, 0.0895, 0.0765, 0.1163, 0.3067, 0.1044, 0.1672, 0.2565], device='cuda:6'), in_proj_covar=tensor([0.0289, 0.0308, 0.0221, 0.0282, 0.0315, 0.0262, 0.0251, 0.0269], device='cuda:6'), out_proj_covar=tensor([1.1609e-04, 1.2251e-04, 8.7811e-05, 1.1196e-04, 1.2802e-04, 1.0393e-04, 1.0135e-04, 1.0671e-04], device='cuda:6') 2023-04-27 11:08:03,840 INFO [finetune.py:976] (6/7) Epoch 18, batch 1950, loss[loss=0.1602, simple_loss=0.2321, pruned_loss=0.04412, over 4905.00 frames. ], tot_loss[loss=0.1797, simple_loss=0.2504, pruned_loss=0.05448, over 955085.75 frames. ], batch size: 32, lr: 3.35e-03, grad_scale: 32.0 2023-04-27 11:08:05,846 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=99324.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:08:07,592 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=99327.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:08:09,371 INFO [zipformer.py:1188] (6/7) 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] (6/7) 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 11:08:52,741 INFO [finetune.py:976] (6/7) Epoch 18, batch 2000, loss[loss=0.1429, simple_loss=0.2259, pruned_loss=0.02992, over 4889.00 frames. ], tot_loss[loss=0.1773, simple_loss=0.2474, pruned_loss=0.05359, over 955374.52 frames. ], batch size: 32, lr: 3.34e-03, grad_scale: 32.0 2023-04-27 11:09:13,002 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=99385.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:09:44,510 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=99407.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:09:49,622 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([4.6605, 4.6189, 3.1064, 5.4006, 4.7386, 4.7000, 2.1363, 4.6426], device='cuda:6'), covar=tensor([0.1358, 0.0934, 0.2766, 0.0933, 0.2716, 0.1717, 0.5907, 0.1912], device='cuda:6'), in_proj_covar=tensor([0.0244, 0.0214, 0.0249, 0.0305, 0.0298, 0.0249, 0.0272, 0.0269], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 11:09:54,337 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=99420.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:09:54,372 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8788, 2.2877, 1.9476, 1.7396, 1.4000, 1.4608, 1.9678, 1.3697], device='cuda:6'), covar=tensor([0.1683, 0.1500, 0.1451, 0.1703, 0.2339, 0.1879, 0.0988, 0.2018], device='cuda:6'), in_proj_covar=tensor([0.0198, 0.0214, 0.0169, 0.0206, 0.0201, 0.0185, 0.0157, 0.0187], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:6') 2023-04-27 11:09:54,851 INFO [finetune.py:976] (6/7) Epoch 18, batch 2050, loss[loss=0.1594, simple_loss=0.2293, pruned_loss=0.04478, over 4870.00 frames. ], tot_loss[loss=0.1743, simple_loss=0.2441, pruned_loss=0.05223, over 956731.28 frames. ], batch size: 34, lr: 3.34e-03, grad_scale: 32.0 2023-04-27 11:10:06,991 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.5130, 1.7563, 1.8317, 1.9286, 1.8263, 1.8590, 1.9042, 1.8482], device='cuda:6'), covar=tensor([0.3631, 0.5950, 0.4463, 0.4422, 0.5439, 0.7432, 0.5323, 0.5059], device='cuda:6'), in_proj_covar=tensor([0.0332, 0.0374, 0.0322, 0.0334, 0.0345, 0.0397, 0.0356, 0.0327], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 11:10:15,959 INFO [optim.py:369] (6/7) 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 11:10:18,939 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=2.00 vs. limit=2.0 2023-04-27 11:10:25,985 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([3.5390, 3.4821, 2.5242, 4.1496, 3.6122, 3.5476, 1.6477, 3.5426], device='cuda:6'), covar=tensor([0.2202, 0.1382, 0.3517, 0.2096, 0.3873, 0.2077, 0.6013, 0.2620], device='cuda:6'), in_proj_covar=tensor([0.0244, 0.0213, 0.0248, 0.0305, 0.0297, 0.0248, 0.0272, 0.0269], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 11:10:26,036 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=99468.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 11:10:28,726 INFO [finetune.py:976] (6/7) Epoch 18, batch 2100, loss[loss=0.1847, simple_loss=0.2485, pruned_loss=0.06048, over 4864.00 frames. ], tot_loss[loss=0.1727, simple_loss=0.2425, pruned_loss=0.05147, over 957950.62 frames. ], batch size: 31, lr: 3.34e-03, grad_scale: 32.0 2023-04-27 11:10:35,463 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=99481.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:10:37,279 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9099, 1.1281, 1.4811, 1.6140, 1.5627, 1.6216, 1.5371, 1.5142], device='cuda:6'), covar=tensor([0.3516, 0.4250, 0.3911, 0.3671, 0.4803, 0.6659, 0.3993, 0.4094], device='cuda:6'), in_proj_covar=tensor([0.0330, 0.0372, 0.0320, 0.0331, 0.0342, 0.0393, 0.0354, 0.0324], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 11:11:02,727 INFO [finetune.py:976] (6/7) Epoch 18, batch 2150, loss[loss=0.1537, simple_loss=0.2442, pruned_loss=0.0316, over 4822.00 frames. ], tot_loss[loss=0.1751, simple_loss=0.2455, pruned_loss=0.05233, over 955803.54 frames. ], batch size: 40, lr: 3.34e-03, grad_scale: 32.0 2023-04-27 11:11:08,199 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=99529.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 11:11:18,915 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=99546.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:11:20,756 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.1663, 1.5181, 1.3558, 1.7293, 1.6623, 1.8686, 1.4117, 3.4125], device='cuda:6'), covar=tensor([0.0610, 0.0788, 0.0771, 0.1156, 0.0606, 0.0570, 0.0788, 0.0149], device='cuda:6'), in_proj_covar=tensor([0.0037, 0.0038, 0.0039, 0.0043, 0.0040, 0.0038, 0.0038, 0.0056], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:6') 2023-04-27 11:11:21,995 INFO [zipformer.py:1188] (6/7) 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] (6/7) 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 11:11:32,859 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 18, batch 2200, loss[loss=0.1542, simple_loss=0.2296, pruned_loss=0.03944, over 4743.00 frames. ], tot_loss[loss=0.1771, simple_loss=0.2484, pruned_loss=0.05291, over 956259.78 frames. ], batch size: 59, lr: 3.34e-03, grad_scale: 32.0 2023-04-27 11:11:45,027 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=99585.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:11:47,518 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=99589.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:11:51,022 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=99594.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:11:54,027 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=99599.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:12:08,624 INFO [finetune.py:976] (6/7) Epoch 18, batch 2250, loss[loss=0.1461, simple_loss=0.2188, pruned_loss=0.03675, over 4753.00 frames. ], tot_loss[loss=0.1777, simple_loss=0.2489, pruned_loss=0.05329, over 955765.74 frames. ], batch size: 28, lr: 3.34e-03, grad_scale: 32.0 2023-04-27 11:12:11,475 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=99624.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:12:11,590 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-04-27 11:12:13,832 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=99627.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:12:15,111 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=99629.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:12:28,863 INFO [zipformer.py:1188] (6/7) 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] (6/7) 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 11:12:41,914 INFO [finetune.py:976] (6/7) Epoch 18, batch 2300, loss[loss=0.2003, simple_loss=0.2649, pruned_loss=0.06782, over 4898.00 frames. ], tot_loss[loss=0.1786, simple_loss=0.2501, pruned_loss=0.05359, over 956306.99 frames. ], batch size: 36, lr: 3.34e-03, grad_scale: 32.0 2023-04-27 11:12:45,389 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=99675.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:12:49,371 INFO [zipformer.py:1188] (6/7) 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] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=99680.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:13:01,084 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.07 vs. limit=5.0 2023-04-27 11:13:12,570 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=99707.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:13:26,441 INFO [finetune.py:976] (6/7) Epoch 18, batch 2350, loss[loss=0.1531, simple_loss=0.2207, pruned_loss=0.04278, over 4816.00 frames. ], tot_loss[loss=0.1769, simple_loss=0.2477, pruned_loss=0.05301, over 955503.91 frames. ], batch size: 38, lr: 3.34e-03, grad_scale: 32.0 2023-04-27 11:13:56,985 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=99741.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:14:08,436 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2023-04-27 11:14:08,612 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.1010, 2.5248, 0.9630, 1.3920, 1.9702, 1.2563, 3.0642, 1.6679], device='cuda:6'), covar=tensor([0.0660, 0.0545, 0.0756, 0.1233, 0.0440, 0.0997, 0.0295, 0.0627], device='cuda:6'), in_proj_covar=tensor([0.0051, 0.0065, 0.0048, 0.0046, 0.0049, 0.0052, 0.0073, 0.0051], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0007, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:6') 2023-04-27 11:14:16,632 INFO [optim.py:369] (6/7) 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] (6/7) 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] (6/7) Epoch 18, batch 2400, loss[loss=0.179, simple_loss=0.2536, pruned_loss=0.05225, over 4833.00 frames. ], tot_loss[loss=0.1743, simple_loss=0.2443, pruned_loss=0.05212, over 955363.73 frames. ], batch size: 38, lr: 3.34e-03, grad_scale: 32.0 2023-04-27 11:14:37,394 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=99776.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:15:08,167 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-04-27 11:15:25,051 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.5647, 3.1892, 2.5708, 3.0050, 2.3594, 2.7940, 2.8699, 2.1844], device='cuda:6'), covar=tensor([0.1974, 0.1313, 0.0841, 0.1312, 0.3159, 0.1419, 0.1811, 0.2623], device='cuda:6'), in_proj_covar=tensor([0.0288, 0.0307, 0.0220, 0.0281, 0.0313, 0.0261, 0.0250, 0.0267], device='cuda:6'), 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:6') 2023-04-27 11:15:27,496 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.4652, 2.3635, 2.6633, 3.1359, 2.9035, 2.4754, 2.0814, 2.6518], device='cuda:6'), covar=tensor([0.0846, 0.0901, 0.0569, 0.0462, 0.0524, 0.0757, 0.0719, 0.0539], device='cuda:6'), in_proj_covar=tensor([0.0188, 0.0201, 0.0182, 0.0171, 0.0178, 0.0181, 0.0151, 0.0179], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 11:15:35,837 INFO [finetune.py:976] (6/7) Epoch 18, batch 2450, loss[loss=0.181, simple_loss=0.2533, pruned_loss=0.05435, over 4909.00 frames. ], tot_loss[loss=0.1733, simple_loss=0.2428, pruned_loss=0.05189, over 957687.19 frames. ], batch size: 37, lr: 3.34e-03, grad_scale: 64.0 2023-04-27 11:15:37,170 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.1078, 2.4069, 1.0370, 1.4000, 1.9426, 1.3006, 2.9266, 1.7286], device='cuda:6'), covar=tensor([0.0635, 0.0654, 0.0723, 0.1194, 0.0444, 0.0917, 0.0357, 0.0565], device='cuda:6'), in_proj_covar=tensor([0.0051, 0.0065, 0.0048, 0.0046, 0.0049, 0.0052, 0.0073, 0.0051], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0007, 0.0008, 0.0008, 0.0010, 0.0007], device='cuda:6') 2023-04-27 11:15:37,743 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=99824.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 11:16:08,394 INFO [optim.py:369] (6/7) 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 11:16:19,144 INFO [finetune.py:976] (6/7) Epoch 18, batch 2500, loss[loss=0.172, simple_loss=0.2423, pruned_loss=0.05086, over 4783.00 frames. ], tot_loss[loss=0.1751, simple_loss=0.2448, pruned_loss=0.05276, over 954301.38 frames. ], batch size: 25, lr: 3.34e-03, grad_scale: 64.0 2023-04-27 11:16:28,674 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=99885.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:16:36,205 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([4.7557, 4.7179, 3.1258, 5.5453, 4.8829, 4.7294, 2.2327, 4.8228], device='cuda:6'), covar=tensor([0.1521, 0.0915, 0.3081, 0.0901, 0.3860, 0.1774, 0.6029, 0.2077], device='cuda:6'), in_proj_covar=tensor([0.0248, 0.0217, 0.0253, 0.0309, 0.0301, 0.0252, 0.0276, 0.0275], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 11:16:50,578 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9425, 2.3219, 0.9738, 1.2974, 1.8949, 1.2448, 3.0154, 1.5395], device='cuda:6'), covar=tensor([0.0693, 0.0634, 0.0759, 0.1271, 0.0470, 0.0981, 0.0282, 0.0641], device='cuda:6'), in_proj_covar=tensor([0.0051, 0.0065, 0.0048, 0.0046, 0.0049, 0.0052, 0.0073, 0.0051], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0007], device='cuda:6') 2023-04-27 11:16:52,953 INFO [finetune.py:976] (6/7) Epoch 18, batch 2550, loss[loss=0.1488, simple_loss=0.2223, pruned_loss=0.03759, over 4741.00 frames. ], tot_loss[loss=0.1779, simple_loss=0.2484, pruned_loss=0.05375, over 956252.46 frames. ], batch size: 54, lr: 3.34e-03, grad_scale: 64.0 2023-04-27 11:16:54,882 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=99924.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:16:54,899 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=99924.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:17:00,817 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=99933.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:17:00,860 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.7863, 3.1707, 0.9442, 1.8473, 2.6364, 1.8297, 4.5830, 2.5549], device='cuda:6'), covar=tensor([0.0551, 0.0747, 0.0836, 0.1219, 0.0471, 0.0869, 0.0205, 0.0516], device='cuda:6'), in_proj_covar=tensor([0.0051, 0.0065, 0.0048, 0.0046, 0.0049, 0.0052, 0.0073, 0.0051], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0007, 0.0008, 0.0008, 0.0010, 0.0007], device='cuda:6') 2023-04-27 11:17:10,141 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=99945.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:17:16,097 INFO [optim.py:369] (6/7) 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 11:17:26,891 INFO [finetune.py:976] (6/7) Epoch 18, batch 2600, loss[loss=0.1558, simple_loss=0.2362, pruned_loss=0.03768, over 4764.00 frames. ], tot_loss[loss=0.1789, simple_loss=0.2499, pruned_loss=0.05401, over 954565.23 frames. ], batch size: 28, lr: 3.34e-03, grad_scale: 64.0 2023-04-27 11:17:27,555 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=99972.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:17:28,949 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 2023-04-27 11:17:32,445 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=99980.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:17:36,587 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8043, 2.3348, 2.0435, 1.8271, 1.3164, 1.4050, 2.0968, 1.2700], device='cuda:6'), covar=tensor([0.1528, 0.1422, 0.1178, 0.1541, 0.2161, 0.1715, 0.0811, 0.1988], device='cuda:6'), in_proj_covar=tensor([0.0198, 0.0212, 0.0168, 0.0205, 0.0200, 0.0184, 0.0156, 0.0187], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:6') 2023-04-27 11:17:52,680 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=100007.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 11:18:01,563 INFO [finetune.py:976] (6/7) Epoch 18, batch 2650, loss[loss=0.189, simple_loss=0.2724, pruned_loss=0.05276, over 4898.00 frames. ], tot_loss[loss=0.1802, simple_loss=0.2515, pruned_loss=0.05442, over 954706.20 frames. ], batch size: 37, lr: 3.34e-03, grad_scale: 64.0 2023-04-27 11:18:05,844 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=100028.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:18:10,687 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=100036.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:18:24,399 INFO [optim.py:369] (6/7) 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 11:18:33,087 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=100068.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 11:18:35,203 INFO [finetune.py:976] (6/7) Epoch 18, batch 2700, loss[loss=0.1431, simple_loss=0.2257, pruned_loss=0.03024, over 4752.00 frames. ], tot_loss[loss=0.179, simple_loss=0.2502, pruned_loss=0.05386, over 957411.75 frames. ], batch size: 28, lr: 3.34e-03, grad_scale: 64.0 2023-04-27 11:18:38,370 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=100076.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:19:37,243 INFO [finetune.py:976] (6/7) Epoch 18, batch 2750, loss[loss=0.1959, simple_loss=0.2532, pruned_loss=0.06929, over 4895.00 frames. ], tot_loss[loss=0.1768, simple_loss=0.2472, pruned_loss=0.05316, over 956984.91 frames. ], batch size: 35, lr: 3.34e-03, grad_scale: 64.0 2023-04-27 11:19:39,088 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=100124.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:19:39,115 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=100124.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 11:20:08,703 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.81 vs. limit=2.0 2023-04-27 11:20:11,922 INFO [optim.py:369] (6/7) 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 11:20:28,163 INFO [finetune.py:976] (6/7) Epoch 18, batch 2800, loss[loss=0.1553, simple_loss=0.2218, pruned_loss=0.04442, over 4770.00 frames. ], tot_loss[loss=0.1729, simple_loss=0.2429, pruned_loss=0.05146, over 957982.86 frames. ], batch size: 26, lr: 3.34e-03, grad_scale: 32.0 2023-04-27 11:20:34,238 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=100172.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 11:21:07,287 INFO [finetune.py:976] (6/7) Epoch 18, batch 2850, loss[loss=0.1744, simple_loss=0.2488, pruned_loss=0.05003, over 4913.00 frames. ], tot_loss[loss=0.1734, simple_loss=0.2429, pruned_loss=0.05202, over 955835.99 frames. ], batch size: 36, lr: 3.34e-03, grad_scale: 32.0 2023-04-27 11:21:15,056 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=100224.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:21:29,445 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=100238.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:21:39,955 INFO [zipformer.py:1188] (6/7) 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] (6/7) 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 11:22:14,790 INFO [finetune.py:976] (6/7) Epoch 18, batch 2900, loss[loss=0.1684, simple_loss=0.2325, pruned_loss=0.0521, over 4195.00 frames. ], tot_loss[loss=0.1767, simple_loss=0.246, pruned_loss=0.05364, over 954413.11 frames. ], batch size: 65, lr: 3.34e-03, grad_scale: 32.0 2023-04-27 11:22:15,484 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=100272.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:22:28,744 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=100293.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:22:32,436 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=100299.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:22:42,374 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.75 vs. limit=2.0 2023-04-27 11:22:48,676 INFO [finetune.py:976] (6/7) Epoch 18, batch 2950, loss[loss=0.1674, simple_loss=0.2404, pruned_loss=0.04723, over 4843.00 frames. ], tot_loss[loss=0.1781, simple_loss=0.2482, pruned_loss=0.05396, over 954140.57 frames. ], batch size: 49, lr: 3.34e-03, grad_scale: 32.0 2023-04-27 11:22:52,446 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0828, 1.8034, 2.0517, 2.5752, 2.4604, 2.0017, 1.7067, 2.1675], device='cuda:6'), covar=tensor([0.0846, 0.1169, 0.0720, 0.0518, 0.0616, 0.0819, 0.0813, 0.0563], device='cuda:6'), in_proj_covar=tensor([0.0190, 0.0203, 0.0184, 0.0173, 0.0180, 0.0183, 0.0153, 0.0181], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 11:22:57,896 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=100336.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:23:09,858 INFO [optim.py:369] (6/7) 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 11:23:15,823 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=100363.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 11:23:20,203 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 2023-04-27 11:23:22,043 INFO [finetune.py:976] (6/7) Epoch 18, batch 3000, loss[loss=0.1341, simple_loss=0.1968, pruned_loss=0.03565, over 3801.00 frames. ], tot_loss[loss=0.179, simple_loss=0.2497, pruned_loss=0.05418, over 954146.26 frames. ], batch size: 16, lr: 3.34e-03, grad_scale: 32.0 2023-04-27 11:23:22,043 INFO [finetune.py:1001] (6/7) Computing validation loss 2023-04-27 11:23:32,627 INFO [finetune.py:1010] (6/7) 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] (6/7) Maximum memory allocated so far is 6365MB 2023-04-27 11:23:41,189 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=100384.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:23:41,972 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=5.17 vs. limit=5.0 2023-04-27 11:23:42,517 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.0978, 1.4070, 1.1937, 1.3620, 1.1994, 1.1798, 1.2441, 1.0390], device='cuda:6'), covar=tensor([0.2030, 0.1808, 0.1337, 0.1533, 0.3543, 0.1505, 0.1943, 0.2408], device='cuda:6'), in_proj_covar=tensor([0.0285, 0.0303, 0.0218, 0.0279, 0.0310, 0.0258, 0.0248, 0.0265], device='cuda:6'), out_proj_covar=tensor([1.1449e-04, 1.2044e-04, 8.6461e-05, 1.1068e-04, 1.2584e-04, 1.0237e-04, 1.0040e-04, 1.0527e-04], device='cuda:6') 2023-04-27 11:23:52,496 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=100402.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:24:04,563 INFO [finetune.py:976] (6/7) Epoch 18, batch 3050, loss[loss=0.1608, simple_loss=0.252, pruned_loss=0.03484, over 4789.00 frames. ], tot_loss[loss=0.1787, simple_loss=0.25, pruned_loss=0.05372, over 953452.92 frames. ], batch size: 29, lr: 3.34e-03, grad_scale: 32.0 2023-04-27 11:24:06,114 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-04-27 11:24:12,704 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.4256, 2.0928, 2.2936, 2.9136, 2.3773, 1.9452, 2.0333, 2.2141], device='cuda:6'), covar=tensor([0.2466, 0.2686, 0.1385, 0.2080, 0.2552, 0.2253, 0.3474, 0.2107], device='cuda:6'), in_proj_covar=tensor([0.0290, 0.0246, 0.0226, 0.0316, 0.0219, 0.0231, 0.0229, 0.0184], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:6') 2023-04-27 11:24:43,544 INFO [optim.py:369] (6/7) 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 11:24:55,035 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=100463.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 11:24:59,811 INFO [finetune.py:976] (6/7) Epoch 18, batch 3100, loss[loss=0.1875, simple_loss=0.2572, pruned_loss=0.05896, over 4846.00 frames. ], tot_loss[loss=0.1771, simple_loss=0.2481, pruned_loss=0.05304, over 952652.66 frames. ], batch size: 44, lr: 3.34e-03, grad_scale: 32.0 2023-04-27 11:25:58,534 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9251, 2.4509, 1.9937, 2.3171, 1.7004, 2.1271, 2.1738, 1.6271], device='cuda:6'), covar=tensor([0.1948, 0.1056, 0.0823, 0.1090, 0.2873, 0.1031, 0.1627, 0.2387], device='cuda:6'), in_proj_covar=tensor([0.0287, 0.0305, 0.0219, 0.0280, 0.0311, 0.0259, 0.0249, 0.0267], device='cuda:6'), out_proj_covar=tensor([1.1512e-04, 1.2116e-04, 8.6931e-05, 1.1125e-04, 1.2634e-04, 1.0290e-04, 1.0087e-04, 1.0596e-04], device='cuda:6') 2023-04-27 11:26:01,618 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5516, 1.3555, 0.5208, 1.2945, 1.4724, 1.4629, 1.3441, 1.4317], device='cuda:6'), covar=tensor([0.0456, 0.0372, 0.0373, 0.0529, 0.0269, 0.0464, 0.0446, 0.0536], device='cuda:6'), in_proj_covar=tensor([0.0028, 0.0024, 0.0022, 0.0029, 0.0019, 0.0028, 0.0028, 0.0028], device='cuda:6'), out_proj_covar=tensor([0.0050, 0.0045, 0.0038, 0.0051, 0.0038, 0.0049, 0.0049, 0.0050], device='cuda:6') 2023-04-27 11:26:02,762 INFO [finetune.py:976] (6/7) Epoch 18, batch 3150, loss[loss=0.148, simple_loss=0.2168, pruned_loss=0.03961, over 4920.00 frames. ], tot_loss[loss=0.1756, simple_loss=0.2461, pruned_loss=0.05251, over 955282.10 frames. ], batch size: 37, lr: 3.34e-03, grad_scale: 32.0 2023-04-27 11:26:37,706 INFO [optim.py:369] (6/7) 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 11:26:59,502 INFO [finetune.py:976] (6/7) Epoch 18, batch 3200, loss[loss=0.1674, simple_loss=0.2286, pruned_loss=0.05311, over 4816.00 frames. ], tot_loss[loss=0.1724, simple_loss=0.2425, pruned_loss=0.0511, over 957159.71 frames. ], batch size: 30, lr: 3.33e-03, grad_scale: 32.0 2023-04-27 11:26:59,580 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([3.1459, 3.1113, 2.5078, 3.6872, 3.0767, 3.1689, 1.6611, 3.1386], device='cuda:6'), covar=tensor([0.1903, 0.1462, 0.3922, 0.2247, 0.3455, 0.1835, 0.4864, 0.2767], device='cuda:6'), in_proj_covar=tensor([0.0245, 0.0215, 0.0250, 0.0306, 0.0297, 0.0248, 0.0271, 0.0271], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 11:27:33,049 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=100594.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:27:40,555 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-04-27 11:27:55,347 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.05 vs. limit=5.0 2023-04-27 11:28:02,070 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([3.6016, 3.5430, 2.5982, 4.2211, 3.5894, 3.6007, 1.5300, 3.6125], device='cuda:6'), covar=tensor([0.1719, 0.1270, 0.3574, 0.1602, 0.4367, 0.1735, 0.5699, 0.2472], device='cuda:6'), in_proj_covar=tensor([0.0244, 0.0215, 0.0250, 0.0305, 0.0297, 0.0248, 0.0271, 0.0271], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 11:28:06,111 INFO [finetune.py:976] (6/7) Epoch 18, batch 3250, loss[loss=0.1623, simple_loss=0.2317, pruned_loss=0.04643, over 4760.00 frames. ], tot_loss[loss=0.1737, simple_loss=0.244, pruned_loss=0.05176, over 955113.29 frames. ], batch size: 27, lr: 3.33e-03, grad_scale: 32.0 2023-04-27 11:28:14,718 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.2263, 2.0017, 1.7004, 1.6463, 2.0329, 1.7548, 2.3244, 1.5244], device='cuda:6'), covar=tensor([0.3067, 0.1385, 0.3857, 0.2775, 0.1488, 0.2043, 0.1542, 0.4070], device='cuda:6'), in_proj_covar=tensor([0.0340, 0.0347, 0.0432, 0.0354, 0.0384, 0.0382, 0.0371, 0.0422], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 11:28:30,300 INFO [optim.py:369] (6/7) 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 11:28:35,351 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=100663.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 11:28:40,635 INFO [finetune.py:976] (6/7) Epoch 18, batch 3300, loss[loss=0.1606, simple_loss=0.2363, pruned_loss=0.04241, over 4933.00 frames. ], tot_loss[loss=0.1769, simple_loss=0.2479, pruned_loss=0.05291, over 955127.53 frames. ], batch size: 33, lr: 3.33e-03, grad_scale: 32.0 2023-04-27 11:29:07,785 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=100711.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 11:29:13,777 INFO [finetune.py:976] (6/7) Epoch 18, batch 3350, loss[loss=0.1221, simple_loss=0.1949, pruned_loss=0.02471, over 4744.00 frames. ], tot_loss[loss=0.1782, simple_loss=0.2494, pruned_loss=0.05351, over 954492.36 frames. ], batch size: 23, lr: 3.33e-03, grad_scale: 32.0 2023-04-27 11:29:19,905 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=100730.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 11:29:29,588 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([0.9912, 0.9900, 1.2277, 1.1480, 0.9513, 0.8969, 0.9200, 0.5405], device='cuda:6'), covar=tensor([0.0510, 0.0528, 0.0429, 0.0500, 0.0617, 0.1154, 0.0441, 0.0679], device='cuda:6'), in_proj_covar=tensor([0.0067, 0.0068, 0.0067, 0.0067, 0.0074, 0.0095, 0.0073, 0.0066], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-27 11:29:35,834 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.48 vs. limit=2.0 2023-04-27 11:29:37,234 INFO [optim.py:369] (6/7) 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 11:29:39,127 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=100758.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 11:29:47,009 INFO [finetune.py:976] (6/7) Epoch 18, batch 3400, loss[loss=0.1493, simple_loss=0.2341, pruned_loss=0.03224, over 4915.00 frames. ], tot_loss[loss=0.1788, simple_loss=0.2504, pruned_loss=0.05365, over 954380.38 frames. ], batch size: 41, lr: 3.33e-03, grad_scale: 32.0 2023-04-27 11:29:48,319 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6091, 3.0636, 0.7262, 1.8225, 1.8765, 2.2067, 1.7826, 1.0084], device='cuda:6'), covar=tensor([0.1303, 0.1364, 0.2285, 0.1253, 0.1069, 0.1111, 0.1667, 0.1902], device='cuda:6'), in_proj_covar=tensor([0.0117, 0.0244, 0.0137, 0.0121, 0.0132, 0.0153, 0.0118, 0.0120], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-27 11:30:00,854 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=100791.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 11:30:20,271 INFO [finetune.py:976] (6/7) Epoch 18, batch 3450, loss[loss=0.1754, simple_loss=0.2379, pruned_loss=0.05644, over 4931.00 frames. ], tot_loss[loss=0.1778, simple_loss=0.2497, pruned_loss=0.05297, over 955179.08 frames. ], batch size: 33, lr: 3.33e-03, grad_scale: 32.0 2023-04-27 11:30:21,007 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.4005, 1.3390, 1.7262, 1.6874, 1.2880, 1.1547, 1.3736, 0.7923], device='cuda:6'), covar=tensor([0.0543, 0.0610, 0.0364, 0.0546, 0.0780, 0.1093, 0.0589, 0.0658], device='cuda:6'), in_proj_covar=tensor([0.0067, 0.0068, 0.0067, 0.0067, 0.0074, 0.0095, 0.0073, 0.0066], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-27 11:30:54,161 INFO [optim.py:369] (6/7) 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 11:31:08,667 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=3.66 vs. limit=5.0 2023-04-27 11:31:09,717 INFO [finetune.py:976] (6/7) Epoch 18, batch 3500, loss[loss=0.1454, simple_loss=0.2236, pruned_loss=0.03358, over 4863.00 frames. ], tot_loss[loss=0.1743, simple_loss=0.2458, pruned_loss=0.05143, over 956608.60 frames. ], batch size: 31, lr: 3.33e-03, grad_scale: 32.0 2023-04-27 11:31:31,110 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=100894.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:31:47,779 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5390, 1.6954, 1.3901, 1.1896, 1.2246, 1.1605, 1.3831, 1.1270], device='cuda:6'), covar=tensor([0.1769, 0.1338, 0.1607, 0.1882, 0.2468, 0.2120, 0.1123, 0.2293], device='cuda:6'), in_proj_covar=tensor([0.0199, 0.0215, 0.0170, 0.0206, 0.0201, 0.0186, 0.0157, 0.0188], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:6') 2023-04-27 11:31:49,439 INFO [finetune.py:976] (6/7) Epoch 18, batch 3550, loss[loss=0.152, simple_loss=0.2205, pruned_loss=0.04181, over 4923.00 frames. ], tot_loss[loss=0.1737, simple_loss=0.2442, pruned_loss=0.05159, over 956702.78 frames. ], batch size: 46, lr: 3.33e-03, grad_scale: 32.0 2023-04-27 11:31:59,813 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.2480, 1.8048, 2.1932, 2.6857, 2.1471, 1.7601, 1.4446, 1.9590], device='cuda:6'), covar=tensor([0.3356, 0.3106, 0.1686, 0.2255, 0.2683, 0.2629, 0.4222, 0.2117], device='cuda:6'), in_proj_covar=tensor([0.0287, 0.0245, 0.0224, 0.0312, 0.0217, 0.0230, 0.0227, 0.0183], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:6') 2023-04-27 11:32:02,729 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=100942.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:32:23,481 INFO [optim.py:369] (6/7) 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] (6/7) Epoch 18, batch 3600, loss[loss=0.1708, simple_loss=0.2448, pruned_loss=0.04834, over 4820.00 frames. ], tot_loss[loss=0.172, simple_loss=0.2421, pruned_loss=0.05096, over 957406.69 frames. ], batch size: 33, lr: 3.33e-03, grad_scale: 32.0 2023-04-27 11:33:40,678 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8679, 1.3737, 5.2043, 4.7961, 4.5241, 4.9564, 4.6572, 4.6469], device='cuda:6'), covar=tensor([0.7126, 0.6210, 0.1018, 0.2076, 0.1100, 0.1599, 0.1006, 0.1369], device='cuda:6'), in_proj_covar=tensor([0.0309, 0.0307, 0.0405, 0.0409, 0.0351, 0.0405, 0.0313, 0.0367], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-27 11:33:50,170 INFO [finetune.py:976] (6/7) Epoch 18, batch 3650, loss[loss=0.1636, simple_loss=0.2504, pruned_loss=0.0384, over 4819.00 frames. ], tot_loss[loss=0.1738, simple_loss=0.2445, pruned_loss=0.05157, over 958170.19 frames. ], batch size: 38, lr: 3.33e-03, grad_scale: 32.0 2023-04-27 11:34:31,993 INFO [optim.py:369] (6/7) 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 11:34:34,913 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=101058.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:34:48,426 INFO [finetune.py:976] (6/7) Epoch 18, batch 3700, loss[loss=0.2017, simple_loss=0.2836, pruned_loss=0.05994, over 4849.00 frames. ], tot_loss[loss=0.1745, simple_loss=0.2463, pruned_loss=0.05137, over 956575.01 frames. ], batch size: 44, lr: 3.33e-03, grad_scale: 32.0 2023-04-27 11:34:49,409 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.57 vs. limit=2.0 2023-04-27 11:34:57,629 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=101086.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 11:34:58,249 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.1148, 2.5257, 0.7576, 1.4782, 1.5040, 1.8615, 1.6377, 0.8345], device='cuda:6'), covar=tensor([0.1597, 0.1251, 0.1955, 0.1377, 0.1181, 0.0972, 0.1534, 0.1760], device='cuda:6'), in_proj_covar=tensor([0.0117, 0.0244, 0.0137, 0.0121, 0.0132, 0.0152, 0.0117, 0.0120], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-27 11:35:00,144 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.3599, 1.5858, 1.3981, 1.5843, 1.3345, 1.3579, 1.4167, 1.1225], device='cuda:6'), covar=tensor([0.1602, 0.1206, 0.0875, 0.1181, 0.3372, 0.1182, 0.1706, 0.2086], device='cuda:6'), in_proj_covar=tensor([0.0288, 0.0307, 0.0220, 0.0281, 0.0313, 0.0262, 0.0251, 0.0269], device='cuda:6'), out_proj_covar=tensor([1.1570e-04, 1.2184e-04, 8.7378e-05, 1.1166e-04, 1.2731e-04, 1.0388e-04, 1.0176e-04, 1.0669e-04], device='cuda:6') 2023-04-27 11:35:11,273 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=101106.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:35:11,911 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=101107.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:35:22,248 INFO [finetune.py:976] (6/7) Epoch 18, batch 3750, loss[loss=0.1797, simple_loss=0.2445, pruned_loss=0.05744, over 4865.00 frames. ], tot_loss[loss=0.1771, simple_loss=0.249, pruned_loss=0.05258, over 956679.74 frames. ], batch size: 34, lr: 3.33e-03, grad_scale: 32.0 2023-04-27 11:35:24,191 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.0064, 1.3348, 1.2615, 1.5688, 1.3960, 1.4957, 1.2878, 2.4125], device='cuda:6'), covar=tensor([0.0623, 0.0818, 0.0797, 0.1195, 0.0631, 0.0528, 0.0791, 0.0227], device='cuda:6'), in_proj_covar=tensor([0.0037, 0.0038, 0.0039, 0.0043, 0.0039, 0.0038, 0.0038, 0.0055], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0012, 0.0013, 0.0016], device='cuda:6') 2023-04-27 11:35:24,205 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.3013, 1.2458, 1.3724, 1.5500, 1.6090, 1.2353, 0.9600, 1.4298], device='cuda:6'), covar=tensor([0.0833, 0.1294, 0.0906, 0.0614, 0.0689, 0.0845, 0.0866, 0.0635], device='cuda:6'), in_proj_covar=tensor([0.0190, 0.0203, 0.0184, 0.0172, 0.0179, 0.0183, 0.0152, 0.0180], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 11:35:29,077 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5587, 1.7927, 1.6679, 1.9670, 1.7966, 2.1084, 1.6509, 3.8420], device='cuda:6'), covar=tensor([0.0560, 0.0777, 0.0795, 0.1136, 0.0640, 0.0463, 0.0727, 0.0130], device='cuda:6'), in_proj_covar=tensor([0.0037, 0.0038, 0.0039, 0.0043, 0.0039, 0.0037, 0.0038, 0.0055], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0012, 0.0013, 0.0016], device='cuda:6') 2023-04-27 11:35:43,440 INFO [optim.py:369] (6/7) 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 11:35:54,457 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 18, batch 3800, loss[loss=0.1795, simple_loss=0.2636, pruned_loss=0.04772, over 4901.00 frames. ], tot_loss[loss=0.1774, simple_loss=0.2494, pruned_loss=0.05271, over 954403.65 frames. ], batch size: 36, lr: 3.33e-03, grad_scale: 32.0 2023-04-27 11:36:19,476 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.1844, 1.7626, 2.0085, 2.5591, 2.5283, 2.1564, 1.8445, 2.2618], device='cuda:6'), covar=tensor([0.0837, 0.1169, 0.0723, 0.0553, 0.0562, 0.0827, 0.0705, 0.0590], device='cuda:6'), in_proj_covar=tensor([0.0189, 0.0202, 0.0183, 0.0172, 0.0178, 0.0181, 0.0151, 0.0179], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 11:36:27,243 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9971, 1.7338, 1.9459, 2.3881, 2.3129, 2.0124, 1.5282, 2.1356], device='cuda:6'), covar=tensor([0.0852, 0.1207, 0.0745, 0.0584, 0.0633, 0.0834, 0.0804, 0.0555], device='cuda:6'), in_proj_covar=tensor([0.0189, 0.0202, 0.0183, 0.0172, 0.0178, 0.0181, 0.0151, 0.0179], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 11:36:30,058 INFO [finetune.py:976] (6/7) Epoch 18, batch 3850, loss[loss=0.1388, simple_loss=0.2116, pruned_loss=0.03297, over 4668.00 frames. ], tot_loss[loss=0.1764, simple_loss=0.2481, pruned_loss=0.05233, over 956105.99 frames. ], batch size: 23, lr: 3.33e-03, grad_scale: 32.0 2023-04-27 11:36:46,003 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.65 vs. limit=2.0 2023-04-27 11:36:46,577 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.6569, 2.2057, 2.5239, 3.2883, 2.4664, 2.0627, 1.9440, 2.4932], device='cuda:6'), covar=tensor([0.3368, 0.3114, 0.1656, 0.2246, 0.2928, 0.2638, 0.3779, 0.1993], device='cuda:6'), in_proj_covar=tensor([0.0287, 0.0244, 0.0224, 0.0311, 0.0216, 0.0229, 0.0226, 0.0182], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:6') 2023-04-27 11:36:48,638 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.58 vs. limit=5.0 2023-04-27 11:36:50,613 INFO [optim.py:369] (6/7) 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 11:37:02,693 INFO [finetune.py:976] (6/7) Epoch 18, batch 3900, loss[loss=0.217, simple_loss=0.2752, pruned_loss=0.07936, over 4096.00 frames. ], tot_loss[loss=0.1742, simple_loss=0.2452, pruned_loss=0.05163, over 955726.47 frames. ], batch size: 65, lr: 3.33e-03, grad_scale: 32.0 2023-04-27 11:37:35,480 INFO [finetune.py:976] (6/7) Epoch 18, batch 3950, loss[loss=0.2093, simple_loss=0.2754, pruned_loss=0.07164, over 4887.00 frames. ], tot_loss[loss=0.1708, simple_loss=0.2416, pruned_loss=0.05003, over 956007.44 frames. ], batch size: 35, lr: 3.33e-03, grad_scale: 32.0 2023-04-27 11:38:09,085 INFO [optim.py:369] (6/7) 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 11:38:30,023 INFO [finetune.py:976] (6/7) Epoch 18, batch 4000, loss[loss=0.172, simple_loss=0.2468, pruned_loss=0.04859, over 4817.00 frames. ], tot_loss[loss=0.1719, simple_loss=0.2422, pruned_loss=0.05079, over 954417.74 frames. ], batch size: 38, lr: 3.33e-03, grad_scale: 32.0 2023-04-27 11:38:52,339 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=101386.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 11:39:35,325 INFO [finetune.py:976] (6/7) Epoch 18, batch 4050, loss[loss=0.1849, simple_loss=0.2674, pruned_loss=0.05125, over 4739.00 frames. ], tot_loss[loss=0.1759, simple_loss=0.2468, pruned_loss=0.05256, over 954778.74 frames. ], batch size: 59, lr: 3.33e-03, grad_scale: 32.0 2023-04-27 11:39:55,131 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=101434.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 11:40:07,833 INFO [optim.py:369] (6/7) 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 11:40:10,151 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.57 vs. limit=2.0 2023-04-27 11:40:12,815 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=101463.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:40:18,125 INFO [finetune.py:976] (6/7) Epoch 18, batch 4100, loss[loss=0.1542, simple_loss=0.2294, pruned_loss=0.03952, over 4784.00 frames. ], tot_loss[loss=0.1772, simple_loss=0.2483, pruned_loss=0.05307, over 954222.41 frames. ], batch size: 29, lr: 3.33e-03, grad_scale: 32.0 2023-04-27 11:40:46,537 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([3.9995, 3.9187, 2.7630, 4.6345, 3.9269, 3.9910, 1.8808, 3.9922], device='cuda:6'), covar=tensor([0.1894, 0.1447, 0.3267, 0.1645, 0.3621, 0.2002, 0.5707, 0.2643], device='cuda:6'), in_proj_covar=tensor([0.0249, 0.0218, 0.0254, 0.0308, 0.0303, 0.0252, 0.0277, 0.0275], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 11:40:51,379 INFO [finetune.py:976] (6/7) Epoch 18, batch 4150, loss[loss=0.1603, simple_loss=0.2388, pruned_loss=0.04092, over 4189.00 frames. ], tot_loss[loss=0.1781, simple_loss=0.2493, pruned_loss=0.05344, over 953423.08 frames. ], batch size: 65, lr: 3.33e-03, grad_scale: 32.0 2023-04-27 11:41:11,723 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.30 vs. limit=5.0 2023-04-27 11:41:13,286 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.1307, 1.3444, 1.2696, 1.6094, 1.3840, 1.8385, 1.2769, 3.3190], device='cuda:6'), covar=tensor([0.0725, 0.1089, 0.1047, 0.1379, 0.0860, 0.0702, 0.1031, 0.0223], device='cuda:6'), in_proj_covar=tensor([0.0037, 0.0038, 0.0039, 0.0043, 0.0040, 0.0038, 0.0038, 0.0056], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:6') 2023-04-27 11:41:14,402 INFO [optim.py:369] (6/7) 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 11:41:24,217 INFO [finetune.py:976] (6/7) Epoch 18, batch 4200, loss[loss=0.2009, simple_loss=0.2684, pruned_loss=0.06668, over 4784.00 frames. ], tot_loss[loss=0.1771, simple_loss=0.249, pruned_loss=0.05263, over 955986.50 frames. ], batch size: 51, lr: 3.33e-03, grad_scale: 32.0 2023-04-27 11:41:48,375 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.5221, 3.0226, 1.0079, 1.6603, 2.3990, 1.4789, 4.2063, 2.1472], device='cuda:6'), covar=tensor([0.0630, 0.0697, 0.0891, 0.1281, 0.0527, 0.0988, 0.0216, 0.0602], device='cuda:6'), in_proj_covar=tensor([0.0051, 0.0064, 0.0048, 0.0046, 0.0049, 0.0051, 0.0073, 0.0050], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0009, 0.0007, 0.0007, 0.0008, 0.0008, 0.0010, 0.0007], device='cuda:6') 2023-04-27 11:41:50,830 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5589, 1.5077, 1.9395, 1.9196, 1.4594, 1.2663, 1.6566, 1.0032], device='cuda:6'), covar=tensor([0.0512, 0.0720, 0.0401, 0.0592, 0.0727, 0.1119, 0.0633, 0.0627], device='cuda:6'), in_proj_covar=tensor([0.0067, 0.0068, 0.0067, 0.0066, 0.0074, 0.0095, 0.0073, 0.0066], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-27 11:41:54,536 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.50 vs. limit=5.0 2023-04-27 11:41:56,898 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.3903, 1.7027, 1.6898, 2.0479, 1.8788, 2.0470, 1.6163, 4.3558], device='cuda:6'), covar=tensor([0.0525, 0.0766, 0.0770, 0.1124, 0.0598, 0.0507, 0.0719, 0.0123], device='cuda:6'), in_proj_covar=tensor([0.0037, 0.0038, 0.0039, 0.0043, 0.0039, 0.0037, 0.0038, 0.0055], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0014, 0.0013, 0.0012, 0.0013, 0.0016], device='cuda:6') 2023-04-27 11:41:58,014 INFO [finetune.py:976] (6/7) Epoch 18, batch 4250, loss[loss=0.2281, simple_loss=0.2859, pruned_loss=0.08518, over 4796.00 frames. ], tot_loss[loss=0.1753, simple_loss=0.2467, pruned_loss=0.05194, over 956460.61 frames. ], batch size: 45, lr: 3.33e-03, grad_scale: 32.0 2023-04-27 11:42:21,983 INFO [optim.py:369] (6/7) 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 11:42:25,769 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.02 vs. limit=5.0 2023-04-27 11:42:31,623 INFO [finetune.py:976] (6/7) Epoch 18, batch 4300, loss[loss=0.1399, simple_loss=0.217, pruned_loss=0.03142, over 4782.00 frames. ], tot_loss[loss=0.1725, simple_loss=0.2434, pruned_loss=0.05081, over 955328.17 frames. ], batch size: 29, lr: 3.33e-03, grad_scale: 32.0 2023-04-27 11:42:40,118 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=101684.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:43:04,509 INFO [finetune.py:976] (6/7) Epoch 18, batch 4350, loss[loss=0.1407, simple_loss=0.2121, pruned_loss=0.03463, over 4789.00 frames. ], tot_loss[loss=0.1706, simple_loss=0.2409, pruned_loss=0.05012, over 954751.66 frames. ], batch size: 26, lr: 3.33e-03, grad_scale: 32.0 2023-04-27 11:43:09,555 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.3626, 1.8142, 1.5960, 2.1905, 2.4616, 1.9755, 1.9080, 1.7121], device='cuda:6'), covar=tensor([0.1677, 0.1737, 0.1868, 0.1647, 0.1333, 0.1908, 0.1908, 0.2120], device='cuda:6'), in_proj_covar=tensor([0.0308, 0.0313, 0.0351, 0.0289, 0.0329, 0.0308, 0.0301, 0.0369], device='cuda:6'), out_proj_covar=tensor([6.3653e-05, 6.4978e-05, 7.4518e-05, 5.8647e-05, 6.8466e-05, 6.4721e-05, 6.3216e-05, 7.8552e-05], device='cuda:6') 2023-04-27 11:43:16,453 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.69 vs. limit=2.0 2023-04-27 11:43:20,713 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=101745.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:43:32,960 INFO [optim.py:369] (6/7) 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 11:43:43,357 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 18, batch 4400, loss[loss=0.2188, simple_loss=0.2836, pruned_loss=0.07695, over 4936.00 frames. ], tot_loss[loss=0.1718, simple_loss=0.2422, pruned_loss=0.05065, over 951917.81 frames. ], batch size: 38, lr: 3.32e-03, grad_scale: 32.0 2023-04-27 11:44:37,299 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=101811.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:44:46,120 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([4.1410, 4.0912, 3.1238, 4.8026, 4.1507, 4.1892, 1.9160, 4.0399], device='cuda:6'), covar=tensor([0.1797, 0.1075, 0.2770, 0.1462, 0.2904, 0.1831, 0.5842, 0.2762], device='cuda:6'), in_proj_covar=tensor([0.0246, 0.0217, 0.0252, 0.0307, 0.0301, 0.0251, 0.0275, 0.0275], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 11:44:49,152 INFO [finetune.py:976] (6/7) Epoch 18, batch 4450, loss[loss=0.1507, simple_loss=0.2284, pruned_loss=0.03651, over 4766.00 frames. ], tot_loss[loss=0.1757, simple_loss=0.2463, pruned_loss=0.0525, over 952189.34 frames. ], batch size: 26, lr: 3.32e-03, grad_scale: 32.0 2023-04-27 11:44:59,444 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8399, 1.6328, 1.8223, 2.2021, 2.3049, 1.7281, 1.5121, 1.9100], device='cuda:6'), covar=tensor([0.0865, 0.1024, 0.0760, 0.0576, 0.0523, 0.0879, 0.0784, 0.0593], device='cuda:6'), in_proj_covar=tensor([0.0189, 0.0201, 0.0183, 0.0171, 0.0177, 0.0181, 0.0150, 0.0178], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 11:45:32,216 INFO [optim.py:369] (6/7) 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 11:45:42,464 INFO [finetune.py:976] (6/7) Epoch 18, batch 4500, loss[loss=0.1586, simple_loss=0.21, pruned_loss=0.05363, over 4021.00 frames. ], tot_loss[loss=0.1776, simple_loss=0.2481, pruned_loss=0.05357, over 953001.15 frames. ], batch size: 17, lr: 3.32e-03, grad_scale: 32.0 2023-04-27 11:46:15,944 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=101920.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 11:46:16,438 INFO [finetune.py:976] (6/7) Epoch 18, batch 4550, loss[loss=0.1799, simple_loss=0.2506, pruned_loss=0.0546, over 4813.00 frames. ], tot_loss[loss=0.1782, simple_loss=0.2491, pruned_loss=0.05359, over 952604.44 frames. ], batch size: 33, lr: 3.32e-03, grad_scale: 32.0 2023-04-27 11:46:19,717 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.12 vs. limit=5.0 2023-04-27 11:46:31,206 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7916, 2.3280, 1.8616, 1.7445, 1.3083, 1.3037, 1.9620, 1.2997], device='cuda:6'), covar=tensor([0.1730, 0.1395, 0.1355, 0.1730, 0.2357, 0.1932, 0.0994, 0.2016], device='cuda:6'), in_proj_covar=tensor([0.0198, 0.0213, 0.0168, 0.0206, 0.0201, 0.0185, 0.0156, 0.0187], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:6') 2023-04-27 11:46:38,530 INFO [optim.py:369] (6/7) 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 11:46:49,879 INFO [finetune.py:976] (6/7) Epoch 18, batch 4600, loss[loss=0.1878, simple_loss=0.2587, pruned_loss=0.05843, over 4841.00 frames. ], tot_loss[loss=0.1777, simple_loss=0.2489, pruned_loss=0.05326, over 952052.26 frames. ], batch size: 44, lr: 3.32e-03, grad_scale: 32.0 2023-04-27 11:46:56,102 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=101981.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 11:46:59,697 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=101987.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:47:11,228 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.2194, 2.7814, 2.1995, 2.1112, 1.6496, 1.4913, 2.4609, 1.5764], device='cuda:6'), covar=tensor([0.1680, 0.1470, 0.1427, 0.1769, 0.2330, 0.1972, 0.0917, 0.2058], device='cuda:6'), in_proj_covar=tensor([0.0198, 0.0213, 0.0168, 0.0206, 0.0202, 0.0185, 0.0156, 0.0187], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:6') 2023-04-27 11:47:24,661 INFO [finetune.py:976] (6/7) Epoch 18, batch 4650, loss[loss=0.1585, simple_loss=0.2365, pruned_loss=0.04025, over 4822.00 frames. ], tot_loss[loss=0.1756, simple_loss=0.2461, pruned_loss=0.05257, over 952746.22 frames. ], batch size: 38, lr: 3.32e-03, grad_scale: 32.0 2023-04-27 11:47:36,310 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=102040.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:47:39,424 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5776, 1.6005, 1.9913, 1.9687, 1.4587, 1.2662, 1.7241, 1.3165], device='cuda:6'), covar=tensor([0.0553, 0.0563, 0.0379, 0.0481, 0.0739, 0.1137, 0.0538, 0.0572], device='cuda:6'), in_proj_covar=tensor([0.0067, 0.0068, 0.0067, 0.0066, 0.0074, 0.0094, 0.0073, 0.0065], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-27 11:47:41,228 INFO [zipformer.py:1188] (6/7) 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] (6/7) 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 11:47:58,102 INFO [finetune.py:976] (6/7) Epoch 18, batch 4700, loss[loss=0.1297, simple_loss=0.1999, pruned_loss=0.02974, over 4764.00 frames. ], tot_loss[loss=0.1736, simple_loss=0.2436, pruned_loss=0.05182, over 953484.63 frames. ], batch size: 28, lr: 3.32e-03, grad_scale: 32.0 2023-04-27 11:47:59,973 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.2661, 1.3742, 3.8860, 3.6943, 3.4273, 3.7894, 3.7180, 3.4031], device='cuda:6'), covar=tensor([0.6920, 0.5427, 0.1189, 0.1473, 0.1031, 0.1748, 0.1437, 0.1482], device='cuda:6'), in_proj_covar=tensor([0.0310, 0.0308, 0.0407, 0.0408, 0.0351, 0.0408, 0.0315, 0.0368], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-27 11:48:30,323 INFO [finetune.py:976] (6/7) Epoch 18, batch 4750, loss[loss=0.1852, simple_loss=0.2543, pruned_loss=0.05806, over 4918.00 frames. ], tot_loss[loss=0.1713, simple_loss=0.2411, pruned_loss=0.05078, over 955998.75 frames. ], batch size: 38, lr: 3.32e-03, grad_scale: 32.0 2023-04-27 11:48:34,558 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.65 vs. limit=2.0 2023-04-27 11:48:51,436 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.4494, 1.4823, 4.0799, 3.7857, 3.5485, 3.8529, 3.8088, 3.5501], device='cuda:6'), covar=tensor([0.7108, 0.5789, 0.1216, 0.2039, 0.1352, 0.2344, 0.1780, 0.1804], device='cuda:6'), in_proj_covar=tensor([0.0307, 0.0307, 0.0404, 0.0406, 0.0349, 0.0405, 0.0313, 0.0365], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-27 11:48:51,944 INFO [optim.py:369] (6/7) 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] (6/7) Epoch 18, batch 4800, loss[loss=0.2077, simple_loss=0.2818, pruned_loss=0.06686, over 4900.00 frames. ], tot_loss[loss=0.1735, simple_loss=0.2434, pruned_loss=0.0518, over 955166.83 frames. ], batch size: 35, lr: 3.32e-03, grad_scale: 64.0 2023-04-27 11:50:15,447 INFO [finetune.py:976] (6/7) Epoch 18, batch 4850, loss[loss=0.1608, simple_loss=0.2394, pruned_loss=0.04114, over 4930.00 frames. ], tot_loss[loss=0.1751, simple_loss=0.2464, pruned_loss=0.05195, over 955127.10 frames. ], batch size: 33, lr: 3.32e-03, grad_scale: 64.0 2023-04-27 11:50:27,698 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.2770, 2.6339, 2.2626, 2.5265, 1.8593, 2.2677, 2.2976, 1.7019], device='cuda:6'), covar=tensor([0.1806, 0.1111, 0.0672, 0.1019, 0.3221, 0.1062, 0.1739, 0.2908], device='cuda:6'), in_proj_covar=tensor([0.0285, 0.0304, 0.0217, 0.0279, 0.0310, 0.0258, 0.0248, 0.0265], device='cuda:6'), 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:6') 2023-04-27 11:50:58,317 INFO [optim.py:369] (6/7) 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 11:51:08,949 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0587, 1.4692, 1.8686, 2.2425, 1.8736, 1.4604, 1.1187, 1.5644], device='cuda:6'), covar=tensor([0.2942, 0.3323, 0.1628, 0.1941, 0.2561, 0.2744, 0.4266, 0.2157], device='cuda:6'), in_proj_covar=tensor([0.0289, 0.0246, 0.0226, 0.0314, 0.0218, 0.0230, 0.0227, 0.0184], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:6') 2023-04-27 11:51:19,610 INFO [finetune.py:976] (6/7) Epoch 18, batch 4900, loss[loss=0.1807, simple_loss=0.2539, pruned_loss=0.0537, over 4700.00 frames. ], tot_loss[loss=0.1779, simple_loss=0.2492, pruned_loss=0.05324, over 954173.09 frames. ], batch size: 23, lr: 3.32e-03, grad_scale: 64.0 2023-04-27 11:51:22,006 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([4.0494, 3.9052, 2.7142, 4.6406, 4.0509, 3.9999, 1.7693, 3.9455], device='cuda:6'), covar=tensor([0.1795, 0.1437, 0.3335, 0.1683, 0.3302, 0.1920, 0.5905, 0.2506], device='cuda:6'), in_proj_covar=tensor([0.0243, 0.0214, 0.0249, 0.0303, 0.0297, 0.0249, 0.0272, 0.0272], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 11:51:28,445 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=102276.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 11:51:55,248 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([3.9133, 3.7749, 2.6866, 4.4927, 3.9204, 3.8987, 1.8364, 3.8113], device='cuda:6'), covar=tensor([0.1703, 0.1254, 0.3364, 0.1373, 0.2720, 0.1778, 0.5556, 0.2588], device='cuda:6'), in_proj_covar=tensor([0.0244, 0.0215, 0.0250, 0.0304, 0.0298, 0.0250, 0.0272, 0.0273], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 11:52:16,037 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.81 vs. limit=2.0 2023-04-27 11:52:25,539 INFO [finetune.py:976] (6/7) Epoch 18, batch 4950, loss[loss=0.1398, simple_loss=0.2093, pruned_loss=0.0352, over 4698.00 frames. ], tot_loss[loss=0.1785, simple_loss=0.2501, pruned_loss=0.05344, over 952670.66 frames. ], batch size: 23, lr: 3.32e-03, grad_scale: 64.0 2023-04-27 11:52:35,583 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=6.97 vs. limit=5.0 2023-04-27 11:52:50,485 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=102340.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:52:57,588 INFO [zipformer.py:1188] (6/7) 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] (6/7) 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 11:53:11,039 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-04-27 11:53:12,188 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([3.0538, 2.2288, 2.2030, 2.8582, 2.9479, 2.4650, 2.4596, 2.1386], device='cuda:6'), covar=tensor([0.1096, 0.1471, 0.1470, 0.1186, 0.0984, 0.1622, 0.1706, 0.2078], device='cuda:6'), in_proj_covar=tensor([0.0308, 0.0313, 0.0352, 0.0291, 0.0331, 0.0309, 0.0302, 0.0371], device='cuda:6'), out_proj_covar=tensor([6.3491e-05, 6.4980e-05, 7.4800e-05, 5.8981e-05, 6.8827e-05, 6.5025e-05, 6.3611e-05, 7.9043e-05], device='cuda:6') 2023-04-27 11:53:12,198 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.8354, 3.2131, 2.8222, 3.0173, 2.2622, 2.1765, 3.0251, 2.2764], device='cuda:6'), covar=tensor([0.1312, 0.1406, 0.1056, 0.1230, 0.1898, 0.1527, 0.0793, 0.1573], device='cuda:6'), in_proj_covar=tensor([0.0196, 0.0212, 0.0168, 0.0204, 0.0200, 0.0184, 0.0155, 0.0185], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:6') 2023-04-27 11:53:14,476 INFO [finetune.py:976] (6/7) Epoch 18, batch 5000, loss[loss=0.2006, simple_loss=0.2487, pruned_loss=0.07623, over 4730.00 frames. ], tot_loss[loss=0.1775, simple_loss=0.2483, pruned_loss=0.05337, over 952157.63 frames. ], batch size: 23, lr: 3.32e-03, grad_scale: 64.0 2023-04-27 11:53:20,212 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5125, 1.1176, 1.3125, 1.1683, 1.6821, 1.3361, 1.0490, 1.2471], device='cuda:6'), covar=tensor([0.1289, 0.1230, 0.1686, 0.1296, 0.0785, 0.1444, 0.1661, 0.1997], device='cuda:6'), in_proj_covar=tensor([0.0308, 0.0313, 0.0353, 0.0291, 0.0331, 0.0310, 0.0303, 0.0371], device='cuda:6'), 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:6') 2023-04-27 11:53:28,307 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=102388.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:53:35,011 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6799, 1.3218, 4.4476, 4.2021, 3.8493, 4.1520, 4.0756, 3.9048], device='cuda:6'), covar=tensor([0.7020, 0.6066, 0.1079, 0.1668, 0.1108, 0.1497, 0.1550, 0.1469], device='cuda:6'), in_proj_covar=tensor([0.0307, 0.0307, 0.0406, 0.0408, 0.0350, 0.0407, 0.0314, 0.0365], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-27 11:53:48,402 INFO [finetune.py:976] (6/7) Epoch 18, batch 5050, loss[loss=0.1581, simple_loss=0.2254, pruned_loss=0.04539, over 4913.00 frames. ], tot_loss[loss=0.1751, simple_loss=0.2454, pruned_loss=0.05235, over 952950.72 frames. ], batch size: 36, lr: 3.32e-03, grad_scale: 64.0 2023-04-27 11:54:19,471 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-04-27 11:54:23,901 INFO [optim.py:369] (6/7) 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 11:54:45,117 INFO [finetune.py:976] (6/7) Epoch 18, batch 5100, loss[loss=0.1233, simple_loss=0.1875, pruned_loss=0.02959, over 4703.00 frames. ], tot_loss[loss=0.1723, simple_loss=0.2424, pruned_loss=0.05113, over 954894.37 frames. ], batch size: 23, lr: 3.32e-03, grad_scale: 64.0 2023-04-27 11:54:52,964 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8600, 1.1055, 1.4713, 1.5783, 1.5633, 1.6511, 1.5344, 1.5036], device='cuda:6'), covar=tensor([0.3782, 0.4851, 0.3906, 0.3999, 0.4984, 0.7049, 0.4198, 0.4046], device='cuda:6'), in_proj_covar=tensor([0.0328, 0.0369, 0.0317, 0.0329, 0.0340, 0.0389, 0.0353, 0.0324], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 11:55:18,947 INFO [finetune.py:976] (6/7) Epoch 18, batch 5150, loss[loss=0.2214, simple_loss=0.2896, pruned_loss=0.0766, over 4061.00 frames. ], tot_loss[loss=0.1751, simple_loss=0.2447, pruned_loss=0.05282, over 953972.39 frames. ], batch size: 65, lr: 3.32e-03, grad_scale: 64.0 2023-04-27 11:55:52,831 INFO [optim.py:369] (6/7) 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 11:56:08,075 INFO [finetune.py:976] (6/7) Epoch 18, batch 5200, loss[loss=0.1441, simple_loss=0.2126, pruned_loss=0.03782, over 4250.00 frames. ], tot_loss[loss=0.1778, simple_loss=0.2479, pruned_loss=0.05388, over 951770.54 frames. ], batch size: 18, lr: 3.32e-03, grad_scale: 64.0 2023-04-27 11:56:11,178 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=102576.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 11:56:35,036 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-04-27 11:56:42,077 INFO [finetune.py:976] (6/7) Epoch 18, batch 5250, loss[loss=0.1771, simple_loss=0.2587, pruned_loss=0.04773, over 4818.00 frames. ], tot_loss[loss=0.1776, simple_loss=0.2484, pruned_loss=0.05339, over 948486.06 frames. ], batch size: 51, lr: 3.32e-03, grad_scale: 64.0 2023-04-27 11:56:43,955 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=102624.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 11:56:53,750 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.53 vs. limit=2.0 2023-04-27 11:56:57,435 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=102643.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:56:59,298 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7422, 1.5351, 1.9994, 2.1267, 1.5529, 1.4223, 1.6916, 1.1066], device='cuda:6'), covar=tensor([0.0534, 0.0688, 0.0407, 0.0518, 0.0753, 0.1128, 0.0616, 0.0654], device='cuda:6'), in_proj_covar=tensor([0.0068, 0.0069, 0.0068, 0.0067, 0.0075, 0.0096, 0.0074, 0.0066], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-27 11:57:11,272 INFO [optim.py:369] (6/7) 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 11:57:12,590 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.4543, 1.6900, 1.6071, 1.8459, 1.8178, 2.0509, 1.5826, 3.3680], device='cuda:6'), covar=tensor([0.0560, 0.0667, 0.0684, 0.1042, 0.0543, 0.0679, 0.0690, 0.0147], device='cuda:6'), in_proj_covar=tensor([0.0037, 0.0039, 0.0039, 0.0043, 0.0040, 0.0038, 0.0038, 0.0056], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:6') 2023-04-27 11:57:12,609 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0716, 2.4087, 2.0920, 2.3151, 1.7011, 2.1417, 2.1627, 1.7824], device='cuda:6'), covar=tensor([0.1591, 0.0975, 0.0682, 0.0996, 0.3177, 0.0943, 0.1738, 0.2109], device='cuda:6'), in_proj_covar=tensor([0.0287, 0.0305, 0.0218, 0.0281, 0.0312, 0.0260, 0.0251, 0.0267], device='cuda:6'), out_proj_covar=tensor([1.1544e-04, 1.2116e-04, 8.6724e-05, 1.1139e-04, 1.2692e-04, 1.0304e-04, 1.0136e-04, 1.0572e-04], device='cuda:6') 2023-04-27 11:57:26,877 INFO [finetune.py:976] (6/7) Epoch 18, batch 5300, loss[loss=0.2071, simple_loss=0.2755, pruned_loss=0.06932, over 4915.00 frames. ], tot_loss[loss=0.1789, simple_loss=0.2504, pruned_loss=0.05371, over 949209.78 frames. ], batch size: 38, lr: 3.32e-03, grad_scale: 64.0 2023-04-27 11:57:55,610 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=102691.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 11:58:07,850 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.1118, 2.4560, 2.0617, 2.3011, 1.5016, 2.0592, 2.0634, 1.6063], device='cuda:6'), covar=tensor([0.1727, 0.1058, 0.0777, 0.1053, 0.3969, 0.1165, 0.1788, 0.2555], device='cuda:6'), in_proj_covar=tensor([0.0287, 0.0304, 0.0218, 0.0280, 0.0312, 0.0259, 0.0250, 0.0266], device='cuda:6'), out_proj_covar=tensor([1.1522e-04, 1.2093e-04, 8.6517e-05, 1.1111e-04, 1.2653e-04, 1.0282e-04, 1.0115e-04, 1.0545e-04], device='cuda:6') 2023-04-27 11:58:31,736 INFO [finetune.py:976] (6/7) Epoch 18, batch 5350, loss[loss=0.184, simple_loss=0.2576, pruned_loss=0.05523, over 4889.00 frames. ], tot_loss[loss=0.1773, simple_loss=0.249, pruned_loss=0.0528, over 950999.80 frames. ], batch size: 43, lr: 3.32e-03, grad_scale: 64.0 2023-04-27 11:58:48,872 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.72 vs. limit=2.0 2023-04-27 11:59:22,079 INFO [optim.py:369] (6/7) 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 11:59:37,389 INFO [finetune.py:976] (6/7) Epoch 18, batch 5400, loss[loss=0.1745, simple_loss=0.2421, pruned_loss=0.05344, over 4840.00 frames. ], tot_loss[loss=0.1749, simple_loss=0.2462, pruned_loss=0.05185, over 951297.56 frames. ], batch size: 47, lr: 3.32e-03, grad_scale: 32.0 2023-04-27 12:00:49,559 INFO [finetune.py:976] (6/7) Epoch 18, batch 5450, loss[loss=0.1842, simple_loss=0.2467, pruned_loss=0.0609, over 4929.00 frames. ], tot_loss[loss=0.1723, simple_loss=0.2429, pruned_loss=0.05078, over 951251.07 frames. ], batch size: 33, lr: 3.32e-03, grad_scale: 32.0 2023-04-27 12:01:11,793 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.6484, 2.9838, 0.9278, 1.8161, 2.2602, 1.5891, 4.2460, 2.1976], device='cuda:6'), covar=tensor([0.0550, 0.0712, 0.0908, 0.1213, 0.0539, 0.0970, 0.0180, 0.0585], device='cuda:6'), in_proj_covar=tensor([0.0051, 0.0065, 0.0048, 0.0047, 0.0050, 0.0052, 0.0074, 0.0051], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:6') 2023-04-27 12:01:22,050 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.26 vs. limit=5.0 2023-04-27 12:01:23,541 INFO [zipformer.py:1188] (6/7) 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] (6/7) 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 12:01:44,221 INFO [finetune.py:976] (6/7) Epoch 18, batch 5500, loss[loss=0.155, simple_loss=0.2324, pruned_loss=0.03877, over 4910.00 frames. ], tot_loss[loss=0.1703, simple_loss=0.2407, pruned_loss=0.04999, over 952751.05 frames. ], batch size: 43, lr: 3.32e-03, grad_scale: 32.0 2023-04-27 12:01:59,989 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=102887.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:02:16,264 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7179, 1.4614, 1.7856, 1.8865, 1.4912, 1.3961, 1.4982, 1.0184], device='cuda:6'), covar=tensor([0.0465, 0.0774, 0.0467, 0.0548, 0.0700, 0.1208, 0.0576, 0.0606], device='cuda:6'), in_proj_covar=tensor([0.0068, 0.0068, 0.0068, 0.0067, 0.0075, 0.0095, 0.0073, 0.0066], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-27 12:02:17,483 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 18, batch 5550, loss[loss=0.2132, simple_loss=0.29, pruned_loss=0.06824, over 4844.00 frames. ], tot_loss[loss=0.1721, simple_loss=0.242, pruned_loss=0.05106, over 950784.08 frames. ], batch size: 44, lr: 3.32e-03, grad_scale: 32.0 2023-04-27 12:02:40,500 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=102948.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:02:45,148 INFO [optim.py:369] (6/7) 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] (6/7) Epoch 18, batch 5600, loss[loss=0.1706, simple_loss=0.2446, pruned_loss=0.04829, over 4928.00 frames. ], tot_loss[loss=0.1736, simple_loss=0.2446, pruned_loss=0.05135, over 949537.08 frames. ], batch size: 38, lr: 3.32e-03, grad_scale: 32.0 2023-04-27 12:03:24,718 INFO [finetune.py:976] (6/7) Epoch 18, batch 5650, loss[loss=0.1608, simple_loss=0.2403, pruned_loss=0.04062, over 4804.00 frames. ], tot_loss[loss=0.1764, simple_loss=0.2483, pruned_loss=0.05228, over 952440.07 frames. ], batch size: 51, lr: 3.31e-03, grad_scale: 32.0 2023-04-27 12:03:44,584 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([3.1860, 3.6487, 3.1688, 3.4928, 2.9114, 3.2294, 3.3898, 2.6402], device='cuda:6'), covar=tensor([0.1425, 0.1091, 0.0623, 0.1035, 0.2285, 0.1016, 0.1399, 0.2311], device='cuda:6'), in_proj_covar=tensor([0.0288, 0.0305, 0.0219, 0.0282, 0.0312, 0.0260, 0.0251, 0.0267], device='cuda:6'), out_proj_covar=tensor([1.1586e-04, 1.2147e-04, 8.6822e-05, 1.1195e-04, 1.2691e-04, 1.0323e-04, 1.0146e-04, 1.0604e-04], device='cuda:6') 2023-04-27 12:03:46,254 INFO [optim.py:369] (6/7) 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 12:03:55,215 INFO [finetune.py:976] (6/7) Epoch 18, batch 5700, loss[loss=0.128, simple_loss=0.1948, pruned_loss=0.03065, over 4194.00 frames. ], tot_loss[loss=0.1737, simple_loss=0.244, pruned_loss=0.05168, over 932923.23 frames. ], batch size: 18, lr: 3.31e-03, grad_scale: 32.0 2023-04-27 12:04:24,053 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 19, batch 0, loss[loss=0.2008, simple_loss=0.2837, pruned_loss=0.05896, over 4906.00 frames. ], tot_loss[loss=0.2008, simple_loss=0.2837, pruned_loss=0.05896, over 4906.00 frames. ], batch size: 46, lr: 3.31e-03, grad_scale: 32.0 2023-04-27 12:04:24,532 INFO [finetune.py:1001] (6/7) Computing validation loss 2023-04-27 12:04:35,097 INFO [finetune.py:1010] (6/7) 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] (6/7) Maximum memory allocated so far is 6365MB 2023-04-27 12:04:56,253 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=103132.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:05:02,842 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-04-27 12:05:07,369 INFO [finetune.py:976] (6/7) Epoch 19, batch 50, loss[loss=0.1817, simple_loss=0.246, pruned_loss=0.05864, over 4868.00 frames. ], tot_loss[loss=0.1804, simple_loss=0.2525, pruned_loss=0.05418, over 217531.75 frames. ], batch size: 34, lr: 3.31e-03, grad_scale: 32.0 2023-04-27 12:05:13,227 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-04-27 12:05:13,472 INFO [optim.py:369] (6/7) 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 12:05:15,442 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=103159.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:05:57,697 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=103193.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:06:01,237 INFO [finetune.py:976] (6/7) Epoch 19, batch 100, loss[loss=0.1776, simple_loss=0.2497, pruned_loss=0.05276, over 4871.00 frames. ], tot_loss[loss=0.1758, simple_loss=0.2455, pruned_loss=0.05304, over 382863.08 frames. ], batch size: 34, lr: 3.31e-03, grad_scale: 32.0 2023-04-27 12:06:17,809 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=103207.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:07:01,895 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=103243.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:07:06,072 INFO [finetune.py:976] (6/7) Epoch 19, batch 150, loss[loss=0.1603, simple_loss=0.2378, pruned_loss=0.0414, over 4922.00 frames. ], tot_loss[loss=0.1703, simple_loss=0.2395, pruned_loss=0.05052, over 510990.68 frames. ], batch size: 37, lr: 3.31e-03, grad_scale: 32.0 2023-04-27 12:07:16,416 INFO [optim.py:369] (6/7) 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 12:07:35,080 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.2690, 1.5233, 1.5206, 1.8427, 1.7055, 1.9214, 1.5149, 3.6258], device='cuda:6'), covar=tensor([0.0593, 0.0772, 0.0778, 0.1121, 0.0636, 0.0472, 0.0685, 0.0121], device='cuda:6'), in_proj_covar=tensor([0.0037, 0.0038, 0.0039, 0.0043, 0.0040, 0.0038, 0.0038, 0.0055], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:6') 2023-04-27 12:07:45,051 INFO [finetune.py:976] (6/7) Epoch 19, batch 200, loss[loss=0.1841, simple_loss=0.2456, pruned_loss=0.0613, over 4824.00 frames. ], tot_loss[loss=0.1721, simple_loss=0.2407, pruned_loss=0.05169, over 609527.42 frames. ], batch size: 30, lr: 3.31e-03, grad_scale: 32.0 2023-04-27 12:07:48,064 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.7203, 2.1609, 2.6699, 3.2395, 2.6631, 2.0802, 2.1773, 2.5033], device='cuda:6'), covar=tensor([0.3225, 0.3092, 0.1544, 0.2421, 0.2616, 0.2621, 0.3637, 0.2051], device='cuda:6'), in_proj_covar=tensor([0.0290, 0.0246, 0.0226, 0.0313, 0.0219, 0.0231, 0.0227, 0.0184], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:6') 2023-04-27 12:08:14,337 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9908, 2.6852, 0.9898, 1.4932, 1.8387, 1.3376, 3.3815, 1.9253], device='cuda:6'), covar=tensor([0.0685, 0.0666, 0.0847, 0.1115, 0.0537, 0.0904, 0.0300, 0.0541], device='cuda:6'), in_proj_covar=tensor([0.0050, 0.0064, 0.0047, 0.0046, 0.0049, 0.0052, 0.0073, 0.0050], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0009, 0.0007, 0.0007, 0.0008, 0.0008, 0.0010, 0.0007], device='cuda:6') 2023-04-27 12:08:33,912 INFO [finetune.py:976] (6/7) Epoch 19, batch 250, loss[loss=0.2037, simple_loss=0.2801, pruned_loss=0.06368, over 4813.00 frames. ], tot_loss[loss=0.1739, simple_loss=0.2431, pruned_loss=0.05229, over 688155.01 frames. ], batch size: 45, lr: 3.31e-03, grad_scale: 32.0 2023-04-27 12:08:44,149 INFO [optim.py:369] (6/7) 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 12:08:56,560 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=103366.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 12:09:04,224 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.1461, 1.3500, 1.2881, 1.7488, 1.5227, 1.6751, 1.3148, 3.0692], device='cuda:6'), covar=tensor([0.0729, 0.0971, 0.0956, 0.1288, 0.0767, 0.0534, 0.0913, 0.0228], device='cuda:6'), in_proj_covar=tensor([0.0037, 0.0038, 0.0039, 0.0043, 0.0039, 0.0038, 0.0038, 0.0055], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0014, 0.0013, 0.0012, 0.0013, 0.0016], device='cuda:6') 2023-04-27 12:09:14,427 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.42 vs. limit=5.0 2023-04-27 12:09:22,446 INFO [finetune.py:976] (6/7) Epoch 19, batch 300, loss[loss=0.1884, simple_loss=0.2668, pruned_loss=0.05497, over 4917.00 frames. ], tot_loss[loss=0.1771, simple_loss=0.2474, pruned_loss=0.05342, over 747333.77 frames. ], batch size: 42, lr: 3.31e-03, grad_scale: 32.0 2023-04-27 12:09:42,602 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=103427.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 12:09:55,321 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.4387, 1.4270, 1.7725, 1.8125, 1.4384, 1.2468, 1.5178, 0.9208], device='cuda:6'), covar=tensor([0.0574, 0.0544, 0.0395, 0.0545, 0.0717, 0.1047, 0.0617, 0.0628], device='cuda:6'), in_proj_covar=tensor([0.0068, 0.0068, 0.0067, 0.0067, 0.0075, 0.0094, 0.0073, 0.0065], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-27 12:09:55,812 INFO [finetune.py:976] (6/7) Epoch 19, batch 350, loss[loss=0.1859, simple_loss=0.2522, pruned_loss=0.05979, over 4842.00 frames. ], tot_loss[loss=0.1803, simple_loss=0.2505, pruned_loss=0.05501, over 792995.17 frames. ], batch size: 49, lr: 3.31e-03, grad_scale: 32.0 2023-04-27 12:09:59,434 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=103454.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:10:00,583 INFO [optim.py:369] (6/7) 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 12:10:22,424 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=103488.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:10:23,690 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9753, 1.5027, 1.5623, 1.8101, 2.1768, 1.7775, 1.4772, 1.4324], device='cuda:6'), covar=tensor([0.1595, 0.1521, 0.2003, 0.1246, 0.0954, 0.1529, 0.2276, 0.2457], device='cuda:6'), in_proj_covar=tensor([0.0306, 0.0310, 0.0349, 0.0288, 0.0328, 0.0306, 0.0301, 0.0367], device='cuda:6'), 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:6') 2023-04-27 12:10:24,903 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9151, 1.6489, 2.0465, 2.2071, 1.7151, 1.5572, 1.8126, 1.3167], device='cuda:6'), covar=tensor([0.0428, 0.0766, 0.0434, 0.0437, 0.0697, 0.1017, 0.0524, 0.0593], device='cuda:6'), in_proj_covar=tensor([0.0068, 0.0068, 0.0068, 0.0067, 0.0075, 0.0095, 0.0073, 0.0066], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-27 12:10:29,037 INFO [finetune.py:976] (6/7) Epoch 19, batch 400, loss[loss=0.1499, simple_loss=0.2234, pruned_loss=0.03822, over 4751.00 frames. ], tot_loss[loss=0.1793, simple_loss=0.2504, pruned_loss=0.0541, over 830079.39 frames. ], batch size: 28, lr: 3.31e-03, grad_scale: 32.0 2023-04-27 12:10:35,032 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=103507.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:10:58,907 INFO [zipformer.py:1188] (6/7) 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] (6/7) attn_weights_entropy = tensor([2.3286, 1.6660, 1.5944, 2.2024, 2.3992, 1.9718, 1.9182, 1.6865], device='cuda:6'), covar=tensor([0.1877, 0.1636, 0.2269, 0.1494, 0.1415, 0.1680, 0.1911, 0.2197], device='cuda:6'), in_proj_covar=tensor([0.0306, 0.0311, 0.0349, 0.0289, 0.0329, 0.0306, 0.0301, 0.0368], device='cuda:6'), 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:6') 2023-04-27 12:11:02,472 INFO [finetune.py:976] (6/7) Epoch 19, batch 450, loss[loss=0.2229, simple_loss=0.2834, pruned_loss=0.08124, over 4172.00 frames. ], tot_loss[loss=0.178, simple_loss=0.2494, pruned_loss=0.05334, over 858483.08 frames. ], batch size: 66, lr: 3.31e-03, grad_scale: 32.0 2023-04-27 12:11:06,196 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=103555.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:11:06,728 INFO [optim.py:369] (6/7) 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 12:11:31,862 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9582, 2.3452, 2.1025, 2.2938, 1.6558, 1.9042, 1.9583, 1.4570], device='cuda:6'), covar=tensor([0.1786, 0.1330, 0.0781, 0.1103, 0.3368, 0.1387, 0.1874, 0.2550], device='cuda:6'), in_proj_covar=tensor([0.0286, 0.0303, 0.0218, 0.0279, 0.0311, 0.0259, 0.0251, 0.0265], device='cuda:6'), 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:6') 2023-04-27 12:11:42,152 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=103591.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:11:46,991 INFO [finetune.py:976] (6/7) Epoch 19, batch 500, loss[loss=0.1735, simple_loss=0.2404, pruned_loss=0.05325, over 4762.00 frames. ], tot_loss[loss=0.1762, simple_loss=0.2468, pruned_loss=0.05281, over 877194.89 frames. ], batch size: 28, lr: 3.31e-03, grad_scale: 32.0 2023-04-27 12:11:53,263 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0919, 2.7355, 0.9262, 1.4527, 1.9049, 1.1659, 3.4313, 1.6485], device='cuda:6'), covar=tensor([0.0683, 0.0675, 0.0870, 0.1262, 0.0555, 0.1093, 0.0238, 0.0667], device='cuda:6'), in_proj_covar=tensor([0.0051, 0.0065, 0.0048, 0.0046, 0.0050, 0.0052, 0.0073, 0.0051], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:6') 2023-04-27 12:12:14,440 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9010, 2.3807, 1.9182, 1.8340, 1.3754, 1.3971, 1.9688, 1.2999], device='cuda:6'), covar=tensor([0.1774, 0.1386, 0.1451, 0.1659, 0.2581, 0.2162, 0.1056, 0.2209], device='cuda:6'), in_proj_covar=tensor([0.0197, 0.0213, 0.0168, 0.0205, 0.0200, 0.0185, 0.0156, 0.0186], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:6') 2023-04-27 12:12:30,578 INFO [finetune.py:976] (6/7) Epoch 19, batch 550, loss[loss=0.1944, simple_loss=0.253, pruned_loss=0.06793, over 4726.00 frames. ], tot_loss[loss=0.1734, simple_loss=0.2437, pruned_loss=0.05162, over 896087.27 frames. ], batch size: 54, lr: 3.31e-03, grad_scale: 32.0 2023-04-27 12:12:34,854 INFO [optim.py:369] (6/7) 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 12:13:00,012 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([0.9240, 0.9330, 1.0679, 1.0119, 0.9285, 0.8095, 0.9039, 0.5799], device='cuda:6'), covar=tensor([0.0618, 0.0430, 0.0540, 0.0440, 0.0592, 0.1019, 0.0419, 0.0690], device='cuda:6'), in_proj_covar=tensor([0.0067, 0.0067, 0.0067, 0.0066, 0.0074, 0.0094, 0.0073, 0.0065], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-27 12:13:04,201 INFO [finetune.py:976] (6/7) Epoch 19, batch 600, loss[loss=0.208, simple_loss=0.2832, pruned_loss=0.06644, over 4820.00 frames. ], tot_loss[loss=0.1742, simple_loss=0.2448, pruned_loss=0.05174, over 909900.20 frames. ], batch size: 38, lr: 3.31e-03, grad_scale: 32.0 2023-04-27 12:13:19,610 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=103722.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 12:13:43,208 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=103740.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:13:53,497 INFO [finetune.py:976] (6/7) Epoch 19, batch 650, loss[loss=0.1359, simple_loss=0.2073, pruned_loss=0.03227, over 4809.00 frames. ], tot_loss[loss=0.1755, simple_loss=0.247, pruned_loss=0.05201, over 920950.40 frames. ], batch size: 25, lr: 3.31e-03, grad_scale: 32.0 2023-04-27 12:14:01,785 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=103754.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:14:02,873 INFO [optim.py:369] (6/7) 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 12:14:14,707 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=103765.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:14:30,660 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 19, batch 700, loss[loss=0.1617, simple_loss=0.2385, pruned_loss=0.04248, over 4210.00 frames. ], tot_loss[loss=0.1761, simple_loss=0.2478, pruned_loss=0.05215, over 926385.01 frames. ], batch size: 65, lr: 3.31e-03, grad_scale: 32.0 2023-04-27 12:14:38,723 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=103801.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:14:39,254 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=103802.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:14:53,010 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7900, 2.1597, 2.0000, 2.1713, 1.4692, 1.7991, 2.0158, 1.3962], device='cuda:6'), covar=tensor([0.2217, 0.1694, 0.0998, 0.1401, 0.3749, 0.1444, 0.1822, 0.2460], device='cuda:6'), in_proj_covar=tensor([0.0285, 0.0301, 0.0216, 0.0277, 0.0309, 0.0257, 0.0249, 0.0264], device='cuda:6'), out_proj_covar=tensor([1.1442e-04, 1.1957e-04, 8.5601e-05, 1.0981e-04, 1.2559e-04, 1.0180e-04, 1.0077e-04, 1.0464e-04], device='cuda:6') 2023-04-27 12:14:55,308 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=103826.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:15:02,439 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 19, batch 750, loss[loss=0.1841, simple_loss=0.2513, pruned_loss=0.05846, over 4234.00 frames. ], tot_loss[loss=0.1772, simple_loss=0.249, pruned_loss=0.05272, over 933011.16 frames. ], batch size: 65, lr: 3.30e-03, grad_scale: 32.0 2023-04-27 12:15:15,074 INFO [optim.py:369] (6/7) 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 12:15:21,248 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.3875, 1.9851, 2.3302, 2.7352, 2.7063, 2.2316, 1.9291, 2.4590], device='cuda:6'), covar=tensor([0.0782, 0.1088, 0.0594, 0.0527, 0.0557, 0.0807, 0.0782, 0.0506], device='cuda:6'), in_proj_covar=tensor([0.0184, 0.0196, 0.0177, 0.0167, 0.0173, 0.0177, 0.0148, 0.0173], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 12:15:28,401 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 19, batch 800, loss[loss=0.1784, simple_loss=0.2514, pruned_loss=0.05273, over 4806.00 frames. ], tot_loss[loss=0.1775, simple_loss=0.2491, pruned_loss=0.0529, over 939274.31 frames. ], batch size: 41, lr: 3.30e-03, grad_scale: 32.0 2023-04-27 12:16:09,179 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=103937.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 12:16:17,417 INFO [finetune.py:976] (6/7) Epoch 19, batch 850, loss[loss=0.1819, simple_loss=0.2486, pruned_loss=0.0576, over 4899.00 frames. ], tot_loss[loss=0.1752, simple_loss=0.2462, pruned_loss=0.05207, over 942432.30 frames. ], batch size: 32, lr: 3.30e-03, grad_scale: 32.0 2023-04-27 12:16:21,650 INFO [optim.py:369] (6/7) 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 12:16:27,892 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=3.68 vs. limit=5.0 2023-04-27 12:16:31,219 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-04-27 12:16:53,879 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.90 vs. limit=5.0 2023-04-27 12:16:59,793 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=103990.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:17:09,736 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=103997.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:17:10,832 INFO [finetune.py:976] (6/7) Epoch 19, batch 900, loss[loss=0.179, simple_loss=0.2388, pruned_loss=0.05965, over 4799.00 frames. ], tot_loss[loss=0.1739, simple_loss=0.2441, pruned_loss=0.05183, over 947146.17 frames. ], batch size: 51, lr: 3.30e-03, grad_scale: 32.0 2023-04-27 12:17:43,376 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=104022.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 12:17:57,985 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 2023-04-27 12:18:05,352 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.3122, 1.7277, 2.1636, 2.6382, 2.1724, 1.7552, 1.4586, 1.9089], device='cuda:6'), covar=tensor([0.3110, 0.3167, 0.1764, 0.2275, 0.2677, 0.2774, 0.4185, 0.2257], device='cuda:6'), in_proj_covar=tensor([0.0292, 0.0247, 0.0227, 0.0315, 0.0219, 0.0232, 0.0228, 0.0186], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:6') 2023-04-27 12:18:18,257 INFO [finetune.py:976] (6/7) Epoch 19, batch 950, loss[loss=0.2007, simple_loss=0.2798, pruned_loss=0.06079, over 4935.00 frames. ], tot_loss[loss=0.1721, simple_loss=0.2418, pruned_loss=0.0512, over 948680.04 frames. ], batch size: 33, lr: 3.30e-03, grad_scale: 32.0 2023-04-27 12:18:18,371 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=104049.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:18:19,611 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=104051.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:18:27,842 INFO [optim.py:369] (6/7) 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 12:18:29,221 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=104058.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:18:30,515 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 2023-04-27 12:18:41,643 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=104070.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 12:19:02,256 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.2437, 1.5277, 1.4345, 1.6923, 1.5386, 1.7291, 1.4373, 3.0359], device='cuda:6'), covar=tensor([0.0600, 0.0775, 0.0755, 0.1199, 0.0615, 0.0460, 0.0723, 0.0181], device='cuda:6'), in_proj_covar=tensor([0.0037, 0.0039, 0.0039, 0.0043, 0.0040, 0.0038, 0.0038, 0.0056], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:6') 2023-04-27 12:19:19,037 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=104096.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:19:21,309 INFO [finetune.py:976] (6/7) Epoch 19, batch 1000, loss[loss=0.1954, simple_loss=0.2628, pruned_loss=0.06396, over 4819.00 frames. ], tot_loss[loss=0.1744, simple_loss=0.2441, pruned_loss=0.05234, over 949901.55 frames. ], batch size: 39, lr: 3.30e-03, grad_scale: 32.0 2023-04-27 12:19:33,871 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=104110.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:19:51,327 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=104121.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:19:55,057 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.4336, 1.3531, 1.7413, 1.6336, 1.2947, 1.2043, 1.4246, 0.9737], device='cuda:6'), covar=tensor([0.0544, 0.0628, 0.0394, 0.0616, 0.0817, 0.1068, 0.0635, 0.0607], device='cuda:6'), in_proj_covar=tensor([0.0068, 0.0068, 0.0067, 0.0067, 0.0074, 0.0095, 0.0073, 0.0066], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-27 12:20:05,671 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.1161, 2.6812, 1.2413, 1.5008, 2.1927, 1.3817, 3.6613, 1.9523], device='cuda:6'), covar=tensor([0.0657, 0.0689, 0.0787, 0.1240, 0.0479, 0.0974, 0.0208, 0.0592], device='cuda:6'), in_proj_covar=tensor([0.0050, 0.0065, 0.0048, 0.0046, 0.0050, 0.0052, 0.0074, 0.0051], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0007], device='cuda:6') 2023-04-27 12:20:16,096 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.51 vs. limit=5.0 2023-04-27 12:20:27,137 INFO [finetune.py:976] (6/7) Epoch 19, batch 1050, loss[loss=0.1378, simple_loss=0.2008, pruned_loss=0.03738, over 4326.00 frames. ], tot_loss[loss=0.1752, simple_loss=0.2461, pruned_loss=0.05212, over 951895.39 frames. ], batch size: 19, lr: 3.30e-03, grad_scale: 32.0 2023-04-27 12:20:37,627 INFO [optim.py:369] (6/7) 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 12:20:38,395 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5377, 1.7331, 1.6416, 2.0195, 1.7437, 1.9726, 1.5931, 3.8023], device='cuda:6'), covar=tensor([0.0551, 0.0773, 0.0784, 0.1134, 0.0624, 0.0470, 0.0698, 0.0131], device='cuda:6'), in_proj_covar=tensor([0.0037, 0.0039, 0.0039, 0.0043, 0.0040, 0.0038, 0.0038, 0.0056], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:6') 2023-04-27 12:21:05,398 INFO [finetune.py:976] (6/7) Epoch 19, batch 1100, loss[loss=0.2389, simple_loss=0.2972, pruned_loss=0.09032, over 4227.00 frames. ], tot_loss[loss=0.176, simple_loss=0.2477, pruned_loss=0.05219, over 952641.41 frames. ], batch size: 65, lr: 3.30e-03, grad_scale: 32.0 2023-04-27 12:21:27,607 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=104232.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 12:21:34,652 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9366, 1.2248, 5.0137, 4.7124, 4.3569, 4.8278, 4.5505, 4.4894], device='cuda:6'), covar=tensor([0.6936, 0.6636, 0.1052, 0.2066, 0.1029, 0.1065, 0.1164, 0.1497], device='cuda:6'), in_proj_covar=tensor([0.0306, 0.0307, 0.0403, 0.0407, 0.0350, 0.0407, 0.0312, 0.0366], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-27 12:21:39,371 INFO [finetune.py:976] (6/7) Epoch 19, batch 1150, loss[loss=0.1975, simple_loss=0.2635, pruned_loss=0.06578, over 4823.00 frames. ], tot_loss[loss=0.1764, simple_loss=0.2479, pruned_loss=0.05242, over 953493.82 frames. ], batch size: 30, lr: 3.30e-03, grad_scale: 32.0 2023-04-27 12:21:39,476 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5813, 3.3446, 0.8859, 1.8702, 1.7913, 2.4273, 1.8427, 1.0326], device='cuda:6'), covar=tensor([0.1240, 0.0767, 0.1984, 0.1178, 0.1058, 0.0900, 0.1568, 0.2008], device='cuda:6'), in_proj_covar=tensor([0.0116, 0.0241, 0.0136, 0.0120, 0.0131, 0.0151, 0.0116, 0.0119], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-27 12:21:44,618 INFO [optim.py:369] (6/7) 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 12:22:12,705 INFO [finetune.py:976] (6/7) Epoch 19, batch 1200, loss[loss=0.169, simple_loss=0.237, pruned_loss=0.05053, over 4923.00 frames. ], tot_loss[loss=0.1746, simple_loss=0.2459, pruned_loss=0.05165, over 954614.28 frames. ], batch size: 33, lr: 3.30e-03, grad_scale: 32.0 2023-04-27 12:22:15,767 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0591, 2.7969, 2.0420, 2.1083, 1.5107, 1.5153, 2.0927, 1.4273], device='cuda:6'), covar=tensor([0.1562, 0.1334, 0.1392, 0.1722, 0.2218, 0.1829, 0.0966, 0.2006], device='cuda:6'), in_proj_covar=tensor([0.0197, 0.0212, 0.0168, 0.0206, 0.0201, 0.0185, 0.0156, 0.0186], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:6') 2023-04-27 12:22:29,829 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5663, 1.4210, 1.7744, 1.7897, 1.4059, 1.3261, 1.4569, 1.0695], device='cuda:6'), covar=tensor([0.0560, 0.0673, 0.0454, 0.0565, 0.0816, 0.1230, 0.0684, 0.0608], device='cuda:6'), in_proj_covar=tensor([0.0068, 0.0068, 0.0067, 0.0067, 0.0074, 0.0095, 0.0073, 0.0066], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-27 12:22:38,353 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.2170, 1.9829, 2.4688, 2.7142, 1.9539, 1.8089, 2.0758, 0.9567], device='cuda:6'), covar=tensor([0.0642, 0.0701, 0.0486, 0.0713, 0.0890, 0.1115, 0.0766, 0.0872], device='cuda:6'), in_proj_covar=tensor([0.0068, 0.0068, 0.0067, 0.0067, 0.0075, 0.0095, 0.0074, 0.0066], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-27 12:22:44,718 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 19, batch 1250, loss[loss=0.1671, simple_loss=0.2354, pruned_loss=0.04934, over 4827.00 frames. ], tot_loss[loss=0.1735, simple_loss=0.2443, pruned_loss=0.05134, over 956016.55 frames. ], batch size: 33, lr: 3.30e-03, grad_scale: 32.0 2023-04-27 12:22:49,473 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=104353.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:22:51,236 INFO [optim.py:369] (6/7) 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 12:23:00,953 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([3.3698, 3.2696, 2.5094, 3.9042, 3.4172, 3.3898, 1.4028, 3.3150], device='cuda:6'), covar=tensor([0.2411, 0.1434, 0.3353, 0.2349, 0.3035, 0.2175, 0.6293, 0.2835], device='cuda:6'), in_proj_covar=tensor([0.0244, 0.0215, 0.0249, 0.0308, 0.0299, 0.0248, 0.0272, 0.0274], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 12:23:45,755 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=104396.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:23:47,996 INFO [finetune.py:976] (6/7) Epoch 19, batch 1300, loss[loss=0.2139, simple_loss=0.2758, pruned_loss=0.07605, over 4175.00 frames. ], tot_loss[loss=0.1723, simple_loss=0.2426, pruned_loss=0.05102, over 954754.86 frames. ], batch size: 18, lr: 3.30e-03, grad_scale: 32.0 2023-04-27 12:23:56,155 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=104405.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:24:18,873 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=104421.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:24:44,322 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.80 vs. limit=2.0 2023-04-27 12:24:44,527 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=104444.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:24:53,498 INFO [finetune.py:976] (6/7) Epoch 19, batch 1350, loss[loss=0.2356, simple_loss=0.3065, pruned_loss=0.0823, over 4830.00 frames. ], tot_loss[loss=0.1722, simple_loss=0.2426, pruned_loss=0.05094, over 954567.57 frames. ], batch size: 38, lr: 3.30e-03, grad_scale: 32.0 2023-04-27 12:25:03,940 INFO [optim.py:369] (6/7) 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 12:25:23,558 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=104469.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:25:57,937 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0518, 1.4874, 1.8960, 2.1873, 1.8786, 1.5153, 1.0586, 1.6140], device='cuda:6'), covar=tensor([0.3104, 0.3243, 0.1659, 0.2049, 0.2606, 0.2653, 0.4262, 0.2144], device='cuda:6'), in_proj_covar=tensor([0.0292, 0.0247, 0.0228, 0.0316, 0.0220, 0.0232, 0.0228, 0.0186], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:6') 2023-04-27 12:25:58,403 INFO [finetune.py:976] (6/7) Epoch 19, batch 1400, loss[loss=0.211, simple_loss=0.2881, pruned_loss=0.06698, over 4818.00 frames. ], tot_loss[loss=0.1742, simple_loss=0.2453, pruned_loss=0.05159, over 954362.62 frames. ], batch size: 51, lr: 3.30e-03, grad_scale: 32.0 2023-04-27 12:26:30,548 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=104532.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 12:26:37,396 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.36 vs. limit=5.0 2023-04-27 12:26:41,374 INFO [finetune.py:976] (6/7) Epoch 19, batch 1450, loss[loss=0.1928, simple_loss=0.2762, pruned_loss=0.05466, over 4938.00 frames. ], tot_loss[loss=0.1752, simple_loss=0.2468, pruned_loss=0.05178, over 954297.67 frames. ], batch size: 33, lr: 3.30e-03, grad_scale: 32.0 2023-04-27 12:26:46,133 INFO [optim.py:369] (6/7) 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] (6/7) 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] (6/7) Epoch 19, batch 1500, loss[loss=0.1753, simple_loss=0.2548, pruned_loss=0.04789, over 4839.00 frames. ], tot_loss[loss=0.1769, simple_loss=0.2484, pruned_loss=0.05275, over 953039.90 frames. ], batch size: 47, lr: 3.30e-03, grad_scale: 32.0 2023-04-27 12:27:46,367 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-04-27 12:27:46,871 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=104646.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:27:48,605 INFO [finetune.py:976] (6/7) Epoch 19, batch 1550, loss[loss=0.1663, simple_loss=0.2302, pruned_loss=0.05113, over 4798.00 frames. ], tot_loss[loss=0.1772, simple_loss=0.2485, pruned_loss=0.05296, over 952587.84 frames. ], batch size: 29, lr: 3.30e-03, grad_scale: 32.0 2023-04-27 12:27:51,638 INFO [zipformer.py:1188] (6/7) 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] (6/7) 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 12:27:53,496 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8275, 1.6469, 1.8195, 2.0992, 2.0726, 1.7498, 1.3524, 1.8597], device='cuda:6'), covar=tensor([0.0761, 0.1111, 0.0721, 0.0593, 0.0597, 0.0879, 0.0789, 0.0609], device='cuda:6'), in_proj_covar=tensor([0.0184, 0.0198, 0.0180, 0.0169, 0.0174, 0.0179, 0.0149, 0.0176], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 12:28:22,267 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0556, 1.8181, 2.0721, 2.3641, 2.2997, 2.0136, 1.5480, 2.0866], device='cuda:6'), covar=tensor([0.0779, 0.1076, 0.0636, 0.0548, 0.0624, 0.0834, 0.0793, 0.0548], device='cuda:6'), in_proj_covar=tensor([0.0185, 0.0198, 0.0180, 0.0170, 0.0175, 0.0179, 0.0149, 0.0176], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 12:28:35,130 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=104694.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:28:43,383 INFO [finetune.py:976] (6/7) Epoch 19, batch 1600, loss[loss=0.1824, simple_loss=0.2412, pruned_loss=0.06182, over 3947.00 frames. ], tot_loss[loss=0.1752, simple_loss=0.2462, pruned_loss=0.05212, over 951713.00 frames. ], batch size: 17, lr: 3.30e-03, grad_scale: 32.0 2023-04-27 12:28:44,683 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=104701.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:28:53,299 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=104705.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:29:27,152 INFO [finetune.py:976] (6/7) Epoch 19, batch 1650, loss[loss=0.1611, simple_loss=0.235, pruned_loss=0.0436, over 4902.00 frames. ], tot_loss[loss=0.1736, simple_loss=0.2438, pruned_loss=0.05168, over 952044.32 frames. ], batch size: 35, lr: 3.30e-03, grad_scale: 64.0 2023-04-27 12:29:29,683 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=104753.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:29:31,434 INFO [optim.py:369] (6/7) 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 12:30:01,076 INFO [finetune.py:976] (6/7) Epoch 19, batch 1700, loss[loss=0.127, simple_loss=0.212, pruned_loss=0.02103, over 4769.00 frames. ], tot_loss[loss=0.1722, simple_loss=0.2423, pruned_loss=0.05099, over 952004.04 frames. ], batch size: 28, lr: 3.30e-03, grad_scale: 64.0 2023-04-27 12:30:02,604 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-04-27 12:30:02,951 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.56 vs. limit=2.0 2023-04-27 12:30:48,746 INFO [finetune.py:976] (6/7) Epoch 19, batch 1750, loss[loss=0.1353, simple_loss=0.2232, pruned_loss=0.02368, over 4763.00 frames. ], tot_loss[loss=0.174, simple_loss=0.2444, pruned_loss=0.0518, over 951452.70 frames. ], batch size: 28, lr: 3.30e-03, grad_scale: 64.0 2023-04-27 12:30:53,010 INFO [optim.py:369] (6/7) 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] (6/7) Epoch 19, batch 1800, loss[loss=0.2421, simple_loss=0.3021, pruned_loss=0.09109, over 4840.00 frames. ], tot_loss[loss=0.1764, simple_loss=0.2476, pruned_loss=0.05258, over 951788.59 frames. ], batch size: 49, lr: 3.30e-03, grad_scale: 64.0 2023-04-27 12:32:10,886 INFO [finetune.py:976] (6/7) Epoch 19, batch 1850, loss[loss=0.2061, simple_loss=0.2892, pruned_loss=0.06151, over 4907.00 frames. ], tot_loss[loss=0.1777, simple_loss=0.249, pruned_loss=0.05314, over 951854.65 frames. ], batch size: 43, lr: 3.30e-03, grad_scale: 64.0 2023-04-27 12:32:15,623 INFO [optim.py:369] (6/7) 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 12:32:44,648 INFO [finetune.py:976] (6/7) Epoch 19, batch 1900, loss[loss=0.1928, simple_loss=0.2642, pruned_loss=0.06064, over 4878.00 frames. ], tot_loss[loss=0.1784, simple_loss=0.2499, pruned_loss=0.05342, over 951992.06 frames. ], batch size: 43, lr: 3.30e-03, grad_scale: 64.0 2023-04-27 12:33:18,499 INFO [finetune.py:976] (6/7) Epoch 19, batch 1950, loss[loss=0.1818, simple_loss=0.2537, pruned_loss=0.055, over 4912.00 frames. ], tot_loss[loss=0.1771, simple_loss=0.2484, pruned_loss=0.05292, over 953016.34 frames. ], batch size: 37, lr: 3.30e-03, grad_scale: 64.0 2023-04-27 12:33:22,765 INFO [optim.py:369] (6/7) 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 12:34:13,142 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5092, 1.9138, 1.9667, 2.3289, 2.0357, 2.3647, 1.8888, 4.7781], device='cuda:6'), covar=tensor([0.0515, 0.0723, 0.0719, 0.1111, 0.0605, 0.0459, 0.0695, 0.0104], device='cuda:6'), in_proj_covar=tensor([0.0037, 0.0038, 0.0039, 0.0043, 0.0040, 0.0038, 0.0038, 0.0055], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:6') 2023-04-27 12:34:18,465 INFO [finetune.py:976] (6/7) Epoch 19, batch 2000, loss[loss=0.1797, simple_loss=0.254, pruned_loss=0.05272, over 4863.00 frames. ], tot_loss[loss=0.1754, simple_loss=0.2462, pruned_loss=0.05229, over 953647.02 frames. ], batch size: 34, lr: 3.29e-03, grad_scale: 64.0 2023-04-27 12:35:12,837 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6498, 2.0918, 1.8597, 2.0097, 1.5782, 1.7409, 1.7664, 1.3734], device='cuda:6'), covar=tensor([0.1820, 0.1158, 0.0742, 0.1090, 0.3328, 0.1077, 0.1961, 0.2524], device='cuda:6'), in_proj_covar=tensor([0.0286, 0.0302, 0.0215, 0.0278, 0.0309, 0.0257, 0.0249, 0.0264], device='cuda:6'), out_proj_covar=tensor([1.1499e-04, 1.1981e-04, 8.5392e-05, 1.1048e-04, 1.2546e-04, 1.0190e-04, 1.0060e-04, 1.0465e-04], device='cuda:6') 2023-04-27 12:35:14,513 INFO [finetune.py:976] (6/7) Epoch 19, batch 2050, loss[loss=0.1752, simple_loss=0.2382, pruned_loss=0.05606, over 4827.00 frames. ], tot_loss[loss=0.1739, simple_loss=0.2437, pruned_loss=0.05201, over 951681.98 frames. ], batch size: 40, lr: 3.29e-03, grad_scale: 64.0 2023-04-27 12:35:18,784 INFO [optim.py:369] (6/7) 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 12:35:52,485 INFO [finetune.py:976] (6/7) Epoch 19, batch 2100, loss[loss=0.1501, simple_loss=0.2303, pruned_loss=0.035, over 4907.00 frames. ], tot_loss[loss=0.1737, simple_loss=0.2434, pruned_loss=0.05197, over 953506.59 frames. ], batch size: 36, lr: 3.29e-03, grad_scale: 32.0 2023-04-27 12:36:09,483 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.6196, 2.0666, 2.4236, 3.0714, 2.4392, 1.9929, 1.9533, 2.3812], device='cuda:6'), covar=tensor([0.3006, 0.3233, 0.1635, 0.2515, 0.2742, 0.2700, 0.3719, 0.2066], device='cuda:6'), in_proj_covar=tensor([0.0290, 0.0245, 0.0226, 0.0313, 0.0217, 0.0230, 0.0226, 0.0184], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:6') 2023-04-27 12:36:37,767 INFO [finetune.py:976] (6/7) Epoch 19, batch 2150, loss[loss=0.1382, simple_loss=0.1924, pruned_loss=0.04202, over 3996.00 frames. ], tot_loss[loss=0.1742, simple_loss=0.2445, pruned_loss=0.05199, over 952016.39 frames. ], batch size: 17, lr: 3.29e-03, grad_scale: 32.0 2023-04-27 12:36:48,981 INFO [optim.py:369] (6/7) 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 12:36:53,482 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([0.6889, 1.6740, 1.6942, 1.3407, 1.7595, 1.5936, 2.2892, 1.3992], device='cuda:6'), covar=tensor([0.3565, 0.1909, 0.4584, 0.3029, 0.1679, 0.2157, 0.1462, 0.4717], device='cuda:6'), in_proj_covar=tensor([0.0337, 0.0345, 0.0425, 0.0352, 0.0380, 0.0377, 0.0369, 0.0416], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 12:37:03,652 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9499, 2.5670, 1.1032, 1.2657, 1.8762, 1.1151, 3.3090, 1.6118], device='cuda:6'), covar=tensor([0.0739, 0.0648, 0.0825, 0.1279, 0.0530, 0.1103, 0.0235, 0.0657], device='cuda:6'), in_proj_covar=tensor([0.0051, 0.0065, 0.0048, 0.0047, 0.0050, 0.0052, 0.0074, 0.0051], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:6') 2023-04-27 12:37:22,984 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.5877, 3.0477, 1.0157, 1.6322, 2.4001, 1.6544, 4.4357, 2.2335], device='cuda:6'), covar=tensor([0.0655, 0.0779, 0.0866, 0.1341, 0.0570, 0.1019, 0.0234, 0.0591], device='cuda:6'), in_proj_covar=tensor([0.0051, 0.0065, 0.0048, 0.0046, 0.0050, 0.0052, 0.0073, 0.0051], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:6') 2023-04-27 12:37:36,680 INFO [finetune.py:976] (6/7) Epoch 19, batch 2200, loss[loss=0.1732, simple_loss=0.252, pruned_loss=0.0472, over 4812.00 frames. ], tot_loss[loss=0.1754, simple_loss=0.2463, pruned_loss=0.05225, over 953248.01 frames. ], batch size: 33, lr: 3.29e-03, grad_scale: 32.0 2023-04-27 12:38:29,818 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.33 vs. limit=5.0 2023-04-27 12:38:49,116 INFO [finetune.py:976] (6/7) Epoch 19, batch 2250, loss[loss=0.1971, simple_loss=0.2714, pruned_loss=0.06138, over 4112.00 frames. ], tot_loss[loss=0.1773, simple_loss=0.2484, pruned_loss=0.05306, over 953929.35 frames. ], batch size: 65, lr: 3.29e-03, grad_scale: 32.0 2023-04-27 12:38:59,483 INFO [optim.py:369] (6/7) 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] (6/7) 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] (6/7) Epoch 19, batch 2300, loss[loss=0.1451, simple_loss=0.2255, pruned_loss=0.0324, over 4765.00 frames. ], tot_loss[loss=0.1756, simple_loss=0.2477, pruned_loss=0.05175, over 955745.98 frames. ], batch size: 28, lr: 3.29e-03, grad_scale: 32.0 2023-04-27 12:40:36,545 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 2023-04-27 12:40:36,897 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8698, 2.1950, 0.8168, 1.2203, 1.4808, 1.1510, 2.4218, 1.3684], device='cuda:6'), covar=tensor([0.0674, 0.0653, 0.0665, 0.1156, 0.0467, 0.0925, 0.0314, 0.0653], device='cuda:6'), in_proj_covar=tensor([0.0051, 0.0065, 0.0048, 0.0046, 0.0050, 0.0052, 0.0073, 0.0051], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:6') 2023-04-27 12:40:58,678 INFO [finetune.py:976] (6/7) Epoch 19, batch 2350, loss[loss=0.1465, simple_loss=0.2177, pruned_loss=0.03769, over 4830.00 frames. ], tot_loss[loss=0.1731, simple_loss=0.245, pruned_loss=0.0506, over 955084.13 frames. ], batch size: 30, lr: 3.29e-03, grad_scale: 32.0 2023-04-27 12:40:59,307 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=105449.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 12:41:09,882 INFO [optim.py:369] (6/7) 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 12:41:22,101 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8949, 1.3717, 1.7161, 1.7545, 1.6378, 1.3708, 0.8001, 1.3741], device='cuda:6'), covar=tensor([0.3461, 0.3433, 0.1804, 0.2173, 0.2548, 0.2840, 0.4445, 0.2202], device='cuda:6'), in_proj_covar=tensor([0.0292, 0.0247, 0.0228, 0.0316, 0.0218, 0.0232, 0.0228, 0.0185], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:6') 2023-04-27 12:41:30,597 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0544, 1.9220, 2.4320, 2.6744, 1.9853, 1.6006, 2.0414, 1.1381], device='cuda:6'), covar=tensor([0.0519, 0.0735, 0.0408, 0.0596, 0.0637, 0.1232, 0.0639, 0.0810], device='cuda:6'), in_proj_covar=tensor([0.0068, 0.0068, 0.0067, 0.0067, 0.0074, 0.0095, 0.0073, 0.0066], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-27 12:41:52,224 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=105486.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:41:55,931 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.2136, 3.0205, 1.0367, 1.6500, 1.7851, 2.1103, 1.7880, 0.9644], device='cuda:6'), covar=tensor([0.1556, 0.1069, 0.1785, 0.1254, 0.1072, 0.1029, 0.1541, 0.1793], device='cuda:6'), in_proj_covar=tensor([0.0115, 0.0239, 0.0136, 0.0119, 0.0130, 0.0150, 0.0115, 0.0117], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-27 12:42:06,554 INFO [finetune.py:976] (6/7) Epoch 19, batch 2400, loss[loss=0.1299, simple_loss=0.2028, pruned_loss=0.02851, over 4677.00 frames. ], tot_loss[loss=0.1717, simple_loss=0.243, pruned_loss=0.05024, over 956426.47 frames. ], batch size: 23, lr: 3.29e-03, grad_scale: 32.0 2023-04-27 12:42:09,004 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=105502.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:42:09,598 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.1535, 2.6014, 2.2277, 2.4639, 1.8749, 2.3182, 2.5252, 1.9006], device='cuda:6'), covar=tensor([0.1823, 0.0979, 0.0682, 0.0977, 0.2515, 0.0927, 0.1444, 0.1995], device='cuda:6'), in_proj_covar=tensor([0.0286, 0.0302, 0.0216, 0.0279, 0.0310, 0.0257, 0.0250, 0.0263], device='cuda:6'), out_proj_covar=tensor([1.1471e-04, 1.1993e-04, 8.5507e-05, 1.1064e-04, 1.2598e-04, 1.0196e-04, 1.0087e-04, 1.0452e-04], device='cuda:6') 2023-04-27 12:42:19,312 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-04-27 12:42:39,444 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=105547.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:42:40,537 INFO [finetune.py:976] (6/7) Epoch 19, batch 2450, loss[loss=0.1995, simple_loss=0.2553, pruned_loss=0.07181, over 4103.00 frames. ], tot_loss[loss=0.1694, simple_loss=0.24, pruned_loss=0.04937, over 956361.50 frames. ], batch size: 65, lr: 3.29e-03, grad_scale: 32.0 2023-04-27 12:42:42,450 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.5203, 3.1431, 1.0703, 1.6250, 2.3940, 1.6272, 4.3167, 2.3510], device='cuda:6'), covar=tensor([0.0608, 0.0638, 0.0875, 0.1369, 0.0508, 0.0968, 0.0222, 0.0557], device='cuda:6'), in_proj_covar=tensor([0.0050, 0.0064, 0.0047, 0.0046, 0.0050, 0.0052, 0.0073, 0.0051], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0009, 0.0007, 0.0007, 0.0008, 0.0008, 0.0010, 0.0007], device='cuda:6') 2023-04-27 12:42:45,852 INFO [optim.py:369] (6/7) 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 12:42:50,140 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=105563.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:43:14,473 INFO [finetune.py:976] (6/7) Epoch 19, batch 2500, loss[loss=0.203, simple_loss=0.2793, pruned_loss=0.06337, over 4899.00 frames. ], tot_loss[loss=0.1731, simple_loss=0.2432, pruned_loss=0.0515, over 954753.98 frames. ], batch size: 35, lr: 3.29e-03, grad_scale: 32.0 2023-04-27 12:43:26,031 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8341, 2.1027, 1.1909, 1.6003, 2.2014, 1.7165, 1.6061, 1.8219], device='cuda:6'), covar=tensor([0.0477, 0.0336, 0.0294, 0.0521, 0.0247, 0.0460, 0.0467, 0.0513], device='cuda:6'), in_proj_covar=tensor([0.0028, 0.0024, 0.0022, 0.0029, 0.0020, 0.0028, 0.0028, 0.0028], device='cuda:6'), out_proj_covar=tensor([0.0050, 0.0045, 0.0038, 0.0051, 0.0038, 0.0049, 0.0049, 0.0050], device='cuda:6') 2023-04-27 12:43:29,617 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2023-04-27 12:43:30,676 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0189, 1.4643, 1.8999, 2.2442, 1.8642, 1.4847, 1.1074, 1.6702], device='cuda:6'), covar=tensor([0.3379, 0.3401, 0.1683, 0.2093, 0.2623, 0.2776, 0.4317, 0.1972], device='cuda:6'), in_proj_covar=tensor([0.0292, 0.0247, 0.0228, 0.0317, 0.0219, 0.0233, 0.0228, 0.0186], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:6') 2023-04-27 12:43:31,273 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6679, 1.4117, 1.7918, 1.8975, 1.4875, 1.3210, 1.4833, 1.0155], device='cuda:6'), covar=tensor([0.0436, 0.0727, 0.0434, 0.0538, 0.0701, 0.1122, 0.0688, 0.0608], device='cuda:6'), in_proj_covar=tensor([0.0067, 0.0067, 0.0066, 0.0066, 0.0074, 0.0094, 0.0072, 0.0065], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-27 12:43:47,976 INFO [finetune.py:976] (6/7) Epoch 19, batch 2550, loss[loss=0.1937, simple_loss=0.2588, pruned_loss=0.06436, over 4904.00 frames. ], tot_loss[loss=0.1757, simple_loss=0.2468, pruned_loss=0.05228, over 955001.04 frames. ], batch size: 35, lr: 3.29e-03, grad_scale: 32.0 2023-04-27 12:43:53,309 INFO [optim.py:369] (6/7) 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 12:44:27,707 INFO [finetune.py:976] (6/7) Epoch 19, batch 2600, loss[loss=0.1769, simple_loss=0.2301, pruned_loss=0.06188, over 4341.00 frames. ], tot_loss[loss=0.1768, simple_loss=0.2484, pruned_loss=0.05265, over 953431.03 frames. ], batch size: 19, lr: 3.29e-03, grad_scale: 32.0 2023-04-27 12:44:58,142 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=105744.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 12:44:58,823 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.4936, 1.9675, 2.3553, 2.9863, 2.3169, 1.8514, 1.8596, 2.2768], device='cuda:6'), covar=tensor([0.3343, 0.3212, 0.1699, 0.2498, 0.2804, 0.2610, 0.3857, 0.1942], device='cuda:6'), in_proj_covar=tensor([0.0290, 0.0245, 0.0226, 0.0314, 0.0217, 0.0231, 0.0226, 0.0184], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:6') 2023-04-27 12:45:01,138 INFO [finetune.py:976] (6/7) Epoch 19, batch 2650, loss[loss=0.1547, simple_loss=0.2359, pruned_loss=0.03677, over 4781.00 frames. ], tot_loss[loss=0.1779, simple_loss=0.2496, pruned_loss=0.05307, over 952260.36 frames. ], batch size: 29, lr: 3.29e-03, grad_scale: 32.0 2023-04-27 12:45:06,399 INFO [optim.py:369] (6/7) 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 12:45:26,486 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.2414, 1.6746, 2.1100, 2.4163, 2.0341, 1.6429, 1.2629, 1.8115], device='cuda:6'), covar=tensor([0.3057, 0.3079, 0.1709, 0.2203, 0.2632, 0.2696, 0.4186, 0.2005], device='cuda:6'), in_proj_covar=tensor([0.0291, 0.0245, 0.0227, 0.0315, 0.0218, 0.0231, 0.0227, 0.0185], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:6') 2023-04-27 12:45:30,670 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=105792.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:45:34,812 INFO [finetune.py:976] (6/7) Epoch 19, batch 2700, loss[loss=0.2001, simple_loss=0.2675, pruned_loss=0.0664, over 4801.00 frames. ], tot_loss[loss=0.1777, simple_loss=0.2493, pruned_loss=0.05306, over 952678.34 frames. ], batch size: 41, lr: 3.29e-03, grad_scale: 32.0 2023-04-27 12:45:40,204 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=105799.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:45:44,473 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6013, 1.8736, 1.8090, 2.1276, 1.9961, 2.1511, 1.7831, 3.7759], device='cuda:6'), covar=tensor([0.0528, 0.0653, 0.0735, 0.1021, 0.0553, 0.0477, 0.0697, 0.0173], device='cuda:6'), in_proj_covar=tensor([0.0037, 0.0039, 0.0039, 0.0043, 0.0040, 0.0038, 0.0038, 0.0056], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:6') 2023-04-27 12:46:05,287 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-04-27 12:46:05,635 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.58 vs. limit=2.0 2023-04-27 12:46:36,274 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=105842.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:46:40,483 INFO [finetune.py:976] (6/7) Epoch 19, batch 2750, loss[loss=0.1675, simple_loss=0.2305, pruned_loss=0.0522, over 4788.00 frames. ], tot_loss[loss=0.175, simple_loss=0.2458, pruned_loss=0.05208, over 952016.52 frames. ], batch size: 26, lr: 3.29e-03, grad_scale: 32.0 2023-04-27 12:46:43,048 INFO [zipformer.py:1188] (6/7) 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] (6/7) 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 12:46:46,581 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=105858.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:46:47,895 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=105860.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:46:50,854 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.1630, 2.5734, 2.1365, 2.4966, 1.8341, 2.1554, 2.2115, 1.6544], device='cuda:6'), covar=tensor([0.1934, 0.1289, 0.0854, 0.1118, 0.3054, 0.1168, 0.1869, 0.2698], device='cuda:6'), in_proj_covar=tensor([0.0288, 0.0304, 0.0218, 0.0280, 0.0313, 0.0260, 0.0251, 0.0266], device='cuda:6'), out_proj_covar=tensor([1.1550e-04, 1.2086e-04, 8.6335e-05, 1.1124e-04, 1.2714e-04, 1.0282e-04, 1.0137e-04, 1.0549e-04], device='cuda:6') 2023-04-27 12:47:35,994 INFO [finetune.py:976] (6/7) Epoch 19, batch 2800, loss[loss=0.1744, simple_loss=0.2513, pruned_loss=0.04878, over 4837.00 frames. ], tot_loss[loss=0.1718, simple_loss=0.2424, pruned_loss=0.05064, over 953166.74 frames. ], batch size: 38, lr: 3.29e-03, grad_scale: 32.0 2023-04-27 12:48:43,332 INFO [finetune.py:976] (6/7) Epoch 19, batch 2850, loss[loss=0.1832, simple_loss=0.2636, pruned_loss=0.05139, over 4892.00 frames. ], tot_loss[loss=0.1716, simple_loss=0.2418, pruned_loss=0.05073, over 952793.45 frames. ], batch size: 35, lr: 3.29e-03, grad_scale: 32.0 2023-04-27 12:48:53,756 INFO [optim.py:369] (6/7) 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 12:49:04,266 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.10 vs. limit=2.0 2023-04-27 12:49:24,703 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.4214, 1.0804, 1.2154, 1.1877, 1.6249, 1.3185, 1.0732, 1.1726], device='cuda:6'), covar=tensor([0.1491, 0.1176, 0.1708, 0.1293, 0.0644, 0.1265, 0.1588, 0.2022], device='cuda:6'), in_proj_covar=tensor([0.0310, 0.0314, 0.0353, 0.0293, 0.0330, 0.0311, 0.0303, 0.0372], device='cuda:6'), out_proj_covar=tensor([6.3934e-05, 6.5213e-05, 7.4845e-05, 5.9382e-05, 6.8395e-05, 6.5464e-05, 6.3505e-05, 7.9338e-05], device='cuda:6') 2023-04-27 12:49:49,881 INFO [finetune.py:976] (6/7) Epoch 19, batch 2900, loss[loss=0.2325, simple_loss=0.294, pruned_loss=0.08551, over 4832.00 frames. ], tot_loss[loss=0.1749, simple_loss=0.2453, pruned_loss=0.05226, over 954706.51 frames. ], batch size: 33, lr: 3.29e-03, grad_scale: 32.0 2023-04-27 12:50:24,615 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-04-27 12:50:40,772 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 19, batch 2950, loss[loss=0.1617, simple_loss=0.2379, pruned_loss=0.04273, over 4919.00 frames. ], tot_loss[loss=0.1759, simple_loss=0.2471, pruned_loss=0.05237, over 951884.72 frames. ], batch size: 36, lr: 3.29e-03, grad_scale: 32.0 2023-04-27 12:50:48,564 INFO [optim.py:369] (6/7) 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 12:50:58,104 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9224, 2.4314, 0.9734, 1.2719, 1.7518, 1.1409, 3.0026, 1.5572], device='cuda:6'), covar=tensor([0.0737, 0.0573, 0.0778, 0.1276, 0.0507, 0.1058, 0.0251, 0.0656], device='cuda:6'), in_proj_covar=tensor([0.0051, 0.0066, 0.0048, 0.0047, 0.0051, 0.0052, 0.0074, 0.0052], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:6') 2023-04-27 12:51:01,781 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=106077.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:51:07,095 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6069, 1.7344, 1.5521, 1.1422, 1.2274, 1.1816, 1.5237, 1.1478], device='cuda:6'), covar=tensor([0.1749, 0.1249, 0.1488, 0.1666, 0.2271, 0.2059, 0.0990, 0.2005], device='cuda:6'), in_proj_covar=tensor([0.0196, 0.0210, 0.0168, 0.0203, 0.0198, 0.0183, 0.0155, 0.0184], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:6') 2023-04-27 12:51:12,302 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=106092.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:51:17,511 INFO [finetune.py:976] (6/7) Epoch 19, batch 3000, loss[loss=0.1937, simple_loss=0.2682, pruned_loss=0.05964, over 4820.00 frames. ], tot_loss[loss=0.1772, simple_loss=0.2488, pruned_loss=0.05285, over 952669.28 frames. ], batch size: 39, lr: 3.29e-03, grad_scale: 32.0 2023-04-27 12:51:17,511 INFO [finetune.py:1001] (6/7) Computing validation loss 2023-04-27 12:51:21,915 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.1698, 1.6324, 1.9686, 2.3420, 1.9933, 1.5927, 1.2072, 1.7000], device='cuda:6'), covar=tensor([0.3303, 0.3280, 0.1784, 0.2061, 0.2591, 0.2773, 0.4437, 0.2159], device='cuda:6'), in_proj_covar=tensor([0.0290, 0.0245, 0.0227, 0.0315, 0.0218, 0.0231, 0.0227, 0.0184], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:6') 2023-04-27 12:51:33,614 INFO [finetune.py:1010] (6/7) 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] (6/7) Maximum memory allocated so far is 6365MB 2023-04-27 12:51:34,947 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.2409, 1.2170, 3.8127, 3.5520, 3.4042, 3.6664, 3.6921, 3.4008], device='cuda:6'), covar=tensor([0.7184, 0.5929, 0.1236, 0.1862, 0.1211, 0.1606, 0.1330, 0.1674], device='cuda:6'), in_proj_covar=tensor([0.0303, 0.0301, 0.0400, 0.0402, 0.0346, 0.0401, 0.0309, 0.0362], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-27 12:51:43,199 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-04-27 12:52:24,197 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=106138.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:52:26,617 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=106142.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:52:36,648 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=106148.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:52:37,147 INFO [finetune.py:976] (6/7) Epoch 19, batch 3050, loss[loss=0.1699, simple_loss=0.2332, pruned_loss=0.05327, over 4712.00 frames. ], tot_loss[loss=0.1772, simple_loss=0.2492, pruned_loss=0.05262, over 952735.69 frames. ], batch size: 23, lr: 3.29e-03, grad_scale: 32.0 2023-04-27 12:52:37,415 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.65 vs. limit=2.0 2023-04-27 12:52:46,798 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=106155.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:52:48,482 INFO [optim.py:369] (6/7) 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 12:52:49,722 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=106158.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:52:58,199 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7898, 2.2282, 1.7350, 1.5016, 1.3770, 1.3502, 1.7465, 1.2810], device='cuda:6'), covar=tensor([0.1708, 0.1211, 0.1416, 0.1751, 0.2183, 0.1940, 0.0978, 0.2025], device='cuda:6'), in_proj_covar=tensor([0.0196, 0.0211, 0.0167, 0.0203, 0.0198, 0.0183, 0.0154, 0.0184], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:6') 2023-04-27 12:53:31,155 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=106190.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:53:42,133 INFO [finetune.py:976] (6/7) Epoch 19, batch 3100, loss[loss=0.1614, simple_loss=0.2339, pruned_loss=0.04445, over 4781.00 frames. ], tot_loss[loss=0.1749, simple_loss=0.2465, pruned_loss=0.05166, over 953255.96 frames. ], batch size: 51, lr: 3.29e-03, grad_scale: 32.0 2023-04-27 12:53:51,785 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=106206.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:54:06,136 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.65 vs. limit=5.0 2023-04-27 12:54:35,395 INFO [finetune.py:976] (6/7) Epoch 19, batch 3150, loss[loss=0.1407, simple_loss=0.2146, pruned_loss=0.03343, over 4916.00 frames. ], tot_loss[loss=0.1732, simple_loss=0.244, pruned_loss=0.05123, over 954014.13 frames. ], batch size: 46, lr: 3.29e-03, grad_scale: 32.0 2023-04-27 12:54:36,066 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.9721, 2.1146, 2.0746, 2.3192, 2.0577, 2.1288, 2.2165, 2.1759], device='cuda:6'), covar=tensor([0.3977, 0.6684, 0.5337, 0.4636, 0.5931, 0.7503, 0.6702, 0.6097], device='cuda:6'), in_proj_covar=tensor([0.0337, 0.0374, 0.0321, 0.0335, 0.0345, 0.0396, 0.0359, 0.0329], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 12:54:40,717 INFO [optim.py:369] (6/7) 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 12:54:49,434 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=106268.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:55:23,650 INFO [finetune.py:976] (6/7) Epoch 19, batch 3200, loss[loss=0.1764, simple_loss=0.251, pruned_loss=0.05084, over 4859.00 frames. ], tot_loss[loss=0.1712, simple_loss=0.2413, pruned_loss=0.05049, over 954895.16 frames. ], batch size: 34, lr: 3.28e-03, grad_scale: 32.0 2023-04-27 12:55:26,275 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-04-27 12:55:50,431 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=106329.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:55:54,739 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=106336.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:56:03,600 INFO [finetune.py:976] (6/7) Epoch 19, batch 3250, loss[loss=0.1617, simple_loss=0.2321, pruned_loss=0.04563, over 4179.00 frames. ], tot_loss[loss=0.172, simple_loss=0.2422, pruned_loss=0.05092, over 952917.07 frames. ], batch size: 65, lr: 3.28e-03, grad_scale: 32.0 2023-04-27 12:56:06,821 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.1206, 2.7265, 2.0994, 2.1992, 1.5106, 1.5444, 2.2587, 1.4671], device='cuda:6'), covar=tensor([0.1531, 0.1363, 0.1347, 0.1584, 0.2147, 0.1756, 0.0951, 0.1949], device='cuda:6'), in_proj_covar=tensor([0.0196, 0.0212, 0.0168, 0.0203, 0.0198, 0.0184, 0.0155, 0.0185], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:6') 2023-04-27 12:56:08,529 INFO [optim.py:369] (6/7) 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 12:56:25,636 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.12 vs. limit=5.0 2023-04-27 12:56:35,778 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=106397.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:56:37,426 INFO [finetune.py:976] (6/7) Epoch 19, batch 3300, loss[loss=0.1894, simple_loss=0.2652, pruned_loss=0.05685, over 4739.00 frames. ], tot_loss[loss=0.1751, simple_loss=0.2465, pruned_loss=0.05187, over 955323.87 frames. ], batch size: 59, lr: 3.28e-03, grad_scale: 32.0 2023-04-27 12:56:47,329 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.4408, 3.4476, 0.8687, 1.8156, 1.7728, 2.3982, 1.9588, 0.9382], device='cuda:6'), covar=tensor([0.1454, 0.0990, 0.2121, 0.1326, 0.1214, 0.1051, 0.1540, 0.2083], device='cuda:6'), in_proj_covar=tensor([0.0115, 0.0239, 0.0136, 0.0118, 0.0130, 0.0150, 0.0114, 0.0117], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-27 12:57:00,444 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=106433.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:57:02,924 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.4588, 1.6544, 1.3859, 1.0040, 1.1305, 1.0877, 1.3660, 1.0455], device='cuda:6'), covar=tensor([0.1855, 0.1356, 0.1644, 0.1955, 0.2387, 0.2222, 0.1095, 0.2259], device='cuda:6'), in_proj_covar=tensor([0.0197, 0.0212, 0.0168, 0.0204, 0.0199, 0.0185, 0.0155, 0.0185], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:6') 2023-04-27 12:57:09,677 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 19, batch 3350, loss[loss=0.1854, simple_loss=0.2496, pruned_loss=0.06059, over 4927.00 frames. ], tot_loss[loss=0.1764, simple_loss=0.2485, pruned_loss=0.05215, over 956645.28 frames. ], batch size: 33, lr: 3.28e-03, grad_scale: 32.0 2023-04-27 12:57:14,362 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=106454.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:57:14,942 INFO [zipformer.py:1188] (6/7) 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] (6/7) 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 12:57:42,144 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=106496.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:57:43,904 INFO [finetune.py:976] (6/7) Epoch 19, batch 3400, loss[loss=0.189, simple_loss=0.2617, pruned_loss=0.05809, over 4848.00 frames. ], tot_loss[loss=0.1781, simple_loss=0.2501, pruned_loss=0.05306, over 954465.43 frames. ], batch size: 49, lr: 3.28e-03, grad_scale: 32.0 2023-04-27 12:57:46,915 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=106503.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:57:55,249 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=106515.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:58:10,586 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0609, 1.5329, 1.8557, 2.1614, 1.8610, 1.5048, 1.0863, 1.6958], device='cuda:6'), covar=tensor([0.3074, 0.3024, 0.1708, 0.2212, 0.2337, 0.2523, 0.4177, 0.1843], device='cuda:6'), in_proj_covar=tensor([0.0290, 0.0244, 0.0226, 0.0314, 0.0218, 0.0230, 0.0226, 0.0184], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:6') 2023-04-27 12:58:14,841 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7147, 2.0152, 0.8314, 1.3735, 1.8314, 1.5746, 1.4936, 1.5348], device='cuda:6'), covar=tensor([0.0492, 0.0358, 0.0360, 0.0557, 0.0272, 0.0511, 0.0517, 0.0573], device='cuda:6'), in_proj_covar=tensor([0.0028, 0.0024, 0.0022, 0.0029, 0.0020, 0.0028, 0.0028, 0.0028], device='cuda:6'), out_proj_covar=tensor([0.0050, 0.0045, 0.0038, 0.0052, 0.0038, 0.0049, 0.0049, 0.0051], device='cuda:6') 2023-04-27 12:58:17,777 INFO [finetune.py:976] (6/7) Epoch 19, batch 3450, loss[loss=0.1973, simple_loss=0.2694, pruned_loss=0.06261, over 4895.00 frames. ], tot_loss[loss=0.1767, simple_loss=0.2487, pruned_loss=0.05237, over 949491.23 frames. ], batch size: 43, lr: 3.28e-03, grad_scale: 32.0 2023-04-27 12:58:23,140 INFO [optim.py:369] (6/7) 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 12:58:57,668 INFO [finetune.py:976] (6/7) Epoch 19, batch 3500, loss[loss=0.1828, simple_loss=0.2495, pruned_loss=0.05802, over 4903.00 frames. ], tot_loss[loss=0.1758, simple_loss=0.2472, pruned_loss=0.0522, over 951056.51 frames. ], batch size: 35, lr: 3.28e-03, grad_scale: 32.0 2023-04-27 12:59:14,549 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=106624.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:59:36,352 INFO [finetune.py:976] (6/7) Epoch 19, batch 3550, loss[loss=0.1452, simple_loss=0.2189, pruned_loss=0.03579, over 4787.00 frames. ], tot_loss[loss=0.174, simple_loss=0.2446, pruned_loss=0.05173, over 953494.70 frames. ], batch size: 26, lr: 3.28e-03, grad_scale: 32.0 2023-04-27 12:59:41,160 INFO [optim.py:369] (6/7) 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 13:00:05,461 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=106692.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:00:12,369 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0812, 2.5805, 0.9810, 1.4417, 2.0287, 1.2858, 3.4765, 1.9216], device='cuda:6'), covar=tensor([0.0697, 0.0606, 0.0772, 0.1255, 0.0514, 0.1004, 0.0220, 0.0614], device='cuda:6'), in_proj_covar=tensor([0.0051, 0.0065, 0.0048, 0.0047, 0.0050, 0.0052, 0.0074, 0.0051], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:6') 2023-04-27 13:00:14,734 INFO [finetune.py:976] (6/7) Epoch 19, batch 3600, loss[loss=0.1506, simple_loss=0.2176, pruned_loss=0.04184, over 4772.00 frames. ], tot_loss[loss=0.1722, simple_loss=0.2425, pruned_loss=0.05099, over 955701.88 frames. ], batch size: 28, lr: 3.28e-03, grad_scale: 32.0 2023-04-27 13:00:58,232 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=106733.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:01:14,590 INFO [finetune.py:976] (6/7) Epoch 19, batch 3650, loss[loss=0.1592, simple_loss=0.2413, pruned_loss=0.03853, over 4763.00 frames. ], tot_loss[loss=0.1736, simple_loss=0.2445, pruned_loss=0.05137, over 954672.04 frames. ], batch size: 59, lr: 3.28e-03, grad_scale: 32.0 2023-04-27 13:01:19,423 INFO [optim.py:369] (6/7) 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 13:01:35,804 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=106781.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:01:48,136 INFO [finetune.py:976] (6/7) Epoch 19, batch 3700, loss[loss=0.1634, simple_loss=0.2219, pruned_loss=0.05243, over 3996.00 frames. ], tot_loss[loss=0.1741, simple_loss=0.2457, pruned_loss=0.05121, over 952024.37 frames. ], batch size: 17, lr: 3.28e-03, grad_scale: 32.0 2023-04-27 13:01:53,949 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6762, 2.0348, 1.6848, 1.5325, 1.2555, 1.2598, 1.7228, 1.1899], device='cuda:6'), covar=tensor([0.1778, 0.1409, 0.1507, 0.1739, 0.2440, 0.2110, 0.1058, 0.2121], device='cuda:6'), in_proj_covar=tensor([0.0197, 0.0212, 0.0169, 0.0204, 0.0200, 0.0185, 0.0155, 0.0185], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:6') 2023-04-27 13:01:55,086 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=106810.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:01:55,167 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.4692, 1.8116, 1.7940, 1.9252, 1.8169, 1.8534, 1.8817, 1.8670], device='cuda:6'), covar=tensor([0.4118, 0.5831, 0.4660, 0.4463, 0.5750, 0.7568, 0.5616, 0.5303], device='cuda:6'), in_proj_covar=tensor([0.0336, 0.0373, 0.0320, 0.0335, 0.0344, 0.0395, 0.0359, 0.0328], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 13:02:00,054 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8567, 1.3579, 1.3968, 1.5911, 1.9920, 1.5573, 1.3889, 1.3481], device='cuda:6'), covar=tensor([0.1599, 0.1482, 0.2162, 0.1423, 0.0884, 0.1557, 0.2078, 0.2336], device='cuda:6'), in_proj_covar=tensor([0.0307, 0.0311, 0.0349, 0.0289, 0.0327, 0.0306, 0.0299, 0.0369], device='cuda:6'), out_proj_covar=tensor([6.3145e-05, 6.4549e-05, 7.3929e-05, 5.8497e-05, 6.7824e-05, 6.4322e-05, 6.2653e-05, 7.8611e-05], device='cuda:6') 2023-04-27 13:02:05,294 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7579, 1.2097, 1.8005, 2.2890, 1.8693, 1.7087, 1.7447, 1.7401], device='cuda:6'), covar=tensor([0.4857, 0.6964, 0.6640, 0.5876, 0.5874, 0.8219, 0.8535, 0.8780], device='cuda:6'), in_proj_covar=tensor([0.0427, 0.0409, 0.0502, 0.0505, 0.0454, 0.0482, 0.0488, 0.0494], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 13:02:20,527 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8601, 2.4134, 1.9631, 1.8405, 1.4215, 1.4454, 2.0362, 1.3266], device='cuda:6'), covar=tensor([0.1565, 0.1418, 0.1321, 0.1710, 0.2200, 0.1920, 0.0890, 0.1926], device='cuda:6'), in_proj_covar=tensor([0.0197, 0.0212, 0.0168, 0.0204, 0.0200, 0.0184, 0.0155, 0.0185], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:6') 2023-04-27 13:02:22,252 INFO [finetune.py:976] (6/7) Epoch 19, batch 3750, loss[loss=0.1697, simple_loss=0.2518, pruned_loss=0.04382, over 4898.00 frames. ], tot_loss[loss=0.1764, simple_loss=0.2489, pruned_loss=0.05194, over 952671.20 frames. ], batch size: 43, lr: 3.28e-03, grad_scale: 32.0 2023-04-27 13:02:23,642 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-04-27 13:02:27,097 INFO [optim.py:369] (6/7) 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 13:02:42,211 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-04-27 13:02:55,382 INFO [finetune.py:976] (6/7) Epoch 19, batch 3800, loss[loss=0.1583, simple_loss=0.2429, pruned_loss=0.0369, over 4900.00 frames. ], tot_loss[loss=0.177, simple_loss=0.2496, pruned_loss=0.0522, over 952109.85 frames. ], batch size: 46, lr: 3.28e-03, grad_scale: 32.0 2023-04-27 13:03:10,672 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=106924.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:03:27,657 INFO [finetune.py:976] (6/7) Epoch 19, batch 3850, loss[loss=0.203, simple_loss=0.2711, pruned_loss=0.06749, over 4735.00 frames. ], tot_loss[loss=0.1762, simple_loss=0.2486, pruned_loss=0.05192, over 952656.37 frames. ], batch size: 59, lr: 3.28e-03, grad_scale: 32.0 2023-04-27 13:03:33,089 INFO [optim.py:369] (6/7) 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 13:03:42,315 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=106972.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:03:56,028 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=106992.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:04:00,614 INFO [finetune.py:976] (6/7) Epoch 19, batch 3900, loss[loss=0.1656, simple_loss=0.2293, pruned_loss=0.05098, over 4248.00 frames. ], tot_loss[loss=0.1753, simple_loss=0.2468, pruned_loss=0.05191, over 953333.32 frames. ], batch size: 65, lr: 3.28e-03, grad_scale: 32.0 2023-04-27 13:04:33,288 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=107040.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:04:35,580 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.8842, 2.8229, 2.2807, 3.2940, 2.8378, 2.9306, 1.2171, 2.8394], device='cuda:6'), covar=tensor([0.2178, 0.1864, 0.3403, 0.3117, 0.3912, 0.2071, 0.6163, 0.2874], device='cuda:6'), in_proj_covar=tensor([0.0244, 0.0216, 0.0249, 0.0307, 0.0298, 0.0247, 0.0271, 0.0270], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 13:04:41,522 INFO [finetune.py:976] (6/7) Epoch 19, batch 3950, loss[loss=0.1788, simple_loss=0.2504, pruned_loss=0.05361, over 4914.00 frames. ], tot_loss[loss=0.1726, simple_loss=0.2436, pruned_loss=0.05085, over 954953.70 frames. ], batch size: 36, lr: 3.28e-03, grad_scale: 32.0 2023-04-27 13:04:54,086 INFO [optim.py:369] (6/7) 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 13:05:16,019 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7078, 1.3740, 1.3474, 1.5913, 1.9607, 1.5713, 1.3615, 1.2307], device='cuda:6'), covar=tensor([0.1764, 0.1499, 0.1738, 0.1242, 0.0845, 0.1645, 0.1992, 0.2146], device='cuda:6'), in_proj_covar=tensor([0.0305, 0.0309, 0.0346, 0.0287, 0.0325, 0.0305, 0.0297, 0.0367], device='cuda:6'), out_proj_covar=tensor([6.2792e-05, 6.4043e-05, 7.3327e-05, 5.8110e-05, 6.7381e-05, 6.4023e-05, 6.2274e-05, 7.8256e-05], device='cuda:6') 2023-04-27 13:05:26,567 INFO [finetune.py:976] (6/7) Epoch 19, batch 4000, loss[loss=0.2127, simple_loss=0.2755, pruned_loss=0.07489, over 4195.00 frames. ], tot_loss[loss=0.1743, simple_loss=0.245, pruned_loss=0.05176, over 955154.21 frames. ], batch size: 65, lr: 3.28e-03, grad_scale: 32.0 2023-04-27 13:05:35,380 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=107110.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:05:36,040 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.3652, 1.2662, 1.4307, 1.5669, 1.2972, 1.1817, 1.2686, 0.7958], device='cuda:6'), covar=tensor([0.0585, 0.0662, 0.0422, 0.0644, 0.0737, 0.1286, 0.0563, 0.0654], device='cuda:6'), in_proj_covar=tensor([0.0068, 0.0068, 0.0067, 0.0067, 0.0075, 0.0095, 0.0073, 0.0066], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-27 13:06:15,766 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=107140.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 13:06:16,565 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-04-27 13:06:28,916 INFO [finetune.py:976] (6/7) Epoch 19, batch 4050, loss[loss=0.146, simple_loss=0.214, pruned_loss=0.03901, over 4761.00 frames. ], tot_loss[loss=0.1749, simple_loss=0.246, pruned_loss=0.05192, over 953496.12 frames. ], batch size: 27, lr: 3.28e-03, grad_scale: 32.0 2023-04-27 13:06:30,827 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=107152.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:06:35,808 INFO [optim.py:369] (6/7) 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 13:06:36,996 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=107158.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:06:57,656 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8781, 1.3692, 1.7050, 1.7612, 1.6881, 1.3609, 0.8052, 1.4523], device='cuda:6'), covar=tensor([0.3053, 0.3214, 0.1747, 0.2051, 0.2405, 0.2650, 0.4265, 0.2015], device='cuda:6'), in_proj_covar=tensor([0.0292, 0.0246, 0.0228, 0.0317, 0.0219, 0.0232, 0.0228, 0.0185], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:6') 2023-04-27 13:07:15,576 INFO [finetune.py:976] (6/7) Epoch 19, batch 4100, loss[loss=0.1993, simple_loss=0.2604, pruned_loss=0.06914, over 4899.00 frames. ], tot_loss[loss=0.177, simple_loss=0.2488, pruned_loss=0.05262, over 955840.81 frames. ], batch size: 36, lr: 3.28e-03, grad_scale: 64.0 2023-04-27 13:07:17,372 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=107201.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 13:07:26,165 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=107213.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 13:07:48,304 INFO [finetune.py:976] (6/7) Epoch 19, batch 4150, loss[loss=0.1769, simple_loss=0.2384, pruned_loss=0.05769, over 4795.00 frames. ], tot_loss[loss=0.1783, simple_loss=0.2499, pruned_loss=0.05332, over 955434.76 frames. ], batch size: 25, lr: 3.28e-03, grad_scale: 64.0 2023-04-27 13:07:54,579 INFO [optim.py:369] (6/7) 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] (6/7) Epoch 19, batch 4200, loss[loss=0.147, simple_loss=0.2284, pruned_loss=0.03285, over 4804.00 frames. ], tot_loss[loss=0.1786, simple_loss=0.2503, pruned_loss=0.05344, over 956344.30 frames. ], batch size: 39, lr: 3.28e-03, grad_scale: 64.0 2023-04-27 13:08:55,956 INFO [finetune.py:976] (6/7) Epoch 19, batch 4250, loss[loss=0.1533, simple_loss=0.2205, pruned_loss=0.043, over 4729.00 frames. ], tot_loss[loss=0.1764, simple_loss=0.2476, pruned_loss=0.05264, over 954809.56 frames. ], batch size: 54, lr: 3.28e-03, grad_scale: 64.0 2023-04-27 13:09:01,365 INFO [optim.py:369] (6/7) 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 13:09:05,631 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-04-27 13:09:24,478 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-04-27 13:09:29,660 INFO [finetune.py:976] (6/7) Epoch 19, batch 4300, loss[loss=0.1675, simple_loss=0.2462, pruned_loss=0.04442, over 4789.00 frames. ], tot_loss[loss=0.1756, simple_loss=0.2464, pruned_loss=0.05246, over 955947.51 frames. ], batch size: 29, lr: 3.28e-03, grad_scale: 32.0 2023-04-27 13:10:30,731 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0796, 2.6108, 0.9344, 1.4122, 1.9464, 1.2338, 3.5633, 1.8754], device='cuda:6'), covar=tensor([0.0686, 0.0692, 0.0779, 0.1204, 0.0510, 0.0999, 0.0230, 0.0599], device='cuda:6'), in_proj_covar=tensor([0.0051, 0.0065, 0.0048, 0.0046, 0.0050, 0.0052, 0.0074, 0.0051], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:6') 2023-04-27 13:10:31,835 INFO [finetune.py:976] (6/7) Epoch 19, batch 4350, loss[loss=0.1595, simple_loss=0.2314, pruned_loss=0.0438, over 4900.00 frames. ], tot_loss[loss=0.1729, simple_loss=0.243, pruned_loss=0.05139, over 958330.76 frames. ], batch size: 43, lr: 3.28e-03, grad_scale: 32.0 2023-04-27 13:10:37,329 INFO [optim.py:369] (6/7) 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 13:10:51,765 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.52 vs. limit=2.0 2023-04-27 13:11:03,019 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=107496.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 13:11:04,780 INFO [finetune.py:976] (6/7) Epoch 19, batch 4400, loss[loss=0.2066, simple_loss=0.2792, pruned_loss=0.06701, over 4854.00 frames. ], tot_loss[loss=0.1739, simple_loss=0.2439, pruned_loss=0.05196, over 957231.56 frames. ], batch size: 49, lr: 3.27e-03, grad_scale: 32.0 2023-04-27 13:11:09,829 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.3993, 1.6263, 1.3823, 1.8300, 1.8437, 1.9994, 1.4592, 3.7286], device='cuda:6'), covar=tensor([0.0556, 0.0825, 0.0862, 0.1250, 0.0605, 0.0497, 0.0790, 0.0147], device='cuda:6'), in_proj_covar=tensor([0.0037, 0.0039, 0.0039, 0.0043, 0.0039, 0.0038, 0.0038, 0.0056], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:6') 2023-04-27 13:11:10,414 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=107508.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 13:11:11,657 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=107510.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:11:49,530 INFO [finetune.py:976] (6/7) Epoch 19, batch 4450, loss[loss=0.1966, simple_loss=0.2715, pruned_loss=0.06081, over 4820.00 frames. ], tot_loss[loss=0.1754, simple_loss=0.2463, pruned_loss=0.05226, over 954736.27 frames. ], batch size: 38, lr: 3.27e-03, grad_scale: 32.0 2023-04-27 13:12:00,087 INFO [optim.py:369] (6/7) 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 13:12:19,127 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=107571.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:12:31,457 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.66 vs. limit=2.0 2023-04-27 13:12:43,865 INFO [finetune.py:976] (6/7) Epoch 19, batch 4500, loss[loss=0.1719, simple_loss=0.2534, pruned_loss=0.04518, over 4738.00 frames. ], tot_loss[loss=0.1771, simple_loss=0.2485, pruned_loss=0.05279, over 954346.95 frames. ], batch size: 27, lr: 3.27e-03, grad_scale: 32.0 2023-04-27 13:13:15,911 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.4588, 1.6915, 1.5328, 1.7517, 1.7963, 1.9181, 1.4729, 3.6925], device='cuda:6'), covar=tensor([0.0569, 0.0774, 0.0818, 0.1273, 0.0609, 0.0520, 0.0764, 0.0106], device='cuda:6'), in_proj_covar=tensor([0.0037, 0.0039, 0.0039, 0.0043, 0.0040, 0.0038, 0.0038, 0.0056], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:6') 2023-04-27 13:13:17,095 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7527, 2.0318, 1.9240, 2.2815, 2.1897, 2.3223, 1.8953, 4.7783], device='cuda:6'), covar=tensor([0.0524, 0.0748, 0.0831, 0.1186, 0.0623, 0.0485, 0.0720, 0.0111], device='cuda:6'), in_proj_covar=tensor([0.0037, 0.0039, 0.0039, 0.0043, 0.0040, 0.0038, 0.0038, 0.0056], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:6') 2023-04-27 13:13:17,594 INFO [finetune.py:976] (6/7) Epoch 19, batch 4550, loss[loss=0.1633, simple_loss=0.2266, pruned_loss=0.04998, over 4686.00 frames. ], tot_loss[loss=0.1777, simple_loss=0.2495, pruned_loss=0.05295, over 955088.93 frames. ], batch size: 23, lr: 3.27e-03, grad_scale: 32.0 2023-04-27 13:13:23,033 INFO [optim.py:369] (6/7) 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 13:13:23,748 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=107659.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:13:50,701 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-04-27 13:13:51,125 INFO [finetune.py:976] (6/7) Epoch 19, batch 4600, loss[loss=0.164, simple_loss=0.2432, pruned_loss=0.04242, over 4864.00 frames. ], tot_loss[loss=0.1771, simple_loss=0.2487, pruned_loss=0.05279, over 954944.62 frames. ], batch size: 34, lr: 3.27e-03, grad_scale: 32.0 2023-04-27 13:14:04,030 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=107720.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:14:16,204 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=107737.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:14:24,305 INFO [finetune.py:976] (6/7) Epoch 19, batch 4650, loss[loss=0.1666, simple_loss=0.2288, pruned_loss=0.05226, over 4773.00 frames. ], tot_loss[loss=0.1748, simple_loss=0.2456, pruned_loss=0.05202, over 956633.51 frames. ], batch size: 28, lr: 3.27e-03, grad_scale: 32.0 2023-04-27 13:14:27,494 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.4806, 2.0647, 2.3303, 2.9659, 2.5071, 1.9906, 2.1302, 2.3815], device='cuda:6'), covar=tensor([0.2441, 0.2675, 0.1382, 0.2149, 0.2085, 0.2095, 0.3265, 0.1797], device='cuda:6'), in_proj_covar=tensor([0.0291, 0.0245, 0.0225, 0.0315, 0.0217, 0.0231, 0.0226, 0.0184], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:6') 2023-04-27 13:14:29,767 INFO [optim.py:369] (6/7) 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 13:15:06,670 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=107796.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 13:15:07,910 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=107798.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 13:15:08,399 INFO [finetune.py:976] (6/7) Epoch 19, batch 4700, loss[loss=0.1609, simple_loss=0.2186, pruned_loss=0.05162, over 4218.00 frames. ], tot_loss[loss=0.1712, simple_loss=0.2416, pruned_loss=0.05036, over 955703.57 frames. ], batch size: 65, lr: 3.27e-03, grad_scale: 32.0 2023-04-27 13:15:19,602 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=107808.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:15:23,273 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6576, 3.5474, 0.7949, 1.9226, 1.9695, 2.3701, 1.9978, 1.0581], device='cuda:6'), covar=tensor([0.1288, 0.0867, 0.2024, 0.1201, 0.1070, 0.1124, 0.1417, 0.2023], device='cuda:6'), in_proj_covar=tensor([0.0117, 0.0242, 0.0138, 0.0120, 0.0132, 0.0153, 0.0117, 0.0120], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-27 13:15:27,784 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2023-04-27 13:15:48,349 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 19, batch 4750, loss[loss=0.16, simple_loss=0.2317, pruned_loss=0.04412, over 4833.00 frames. ], tot_loss[loss=0.1693, simple_loss=0.2394, pruned_loss=0.04962, over 956829.63 frames. ], batch size: 33, lr: 3.27e-03, grad_scale: 32.0 2023-04-27 13:15:57,101 INFO [zipformer.py:1188] (6/7) 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] (6/7) 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 13:16:09,254 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=107866.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:16:53,226 INFO [finetune.py:976] (6/7) Epoch 19, batch 4800, loss[loss=0.193, simple_loss=0.2617, pruned_loss=0.06215, over 4813.00 frames. ], tot_loss[loss=0.1726, simple_loss=0.2428, pruned_loss=0.05123, over 954708.63 frames. ], batch size: 38, lr: 3.27e-03, grad_scale: 32.0 2023-04-27 13:17:02,124 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5694, 1.2493, 4.3411, 4.0361, 3.7650, 4.1645, 4.0629, 3.7872], device='cuda:6'), covar=tensor([0.7175, 0.6553, 0.1110, 0.1896, 0.1294, 0.2013, 0.1295, 0.1677], device='cuda:6'), in_proj_covar=tensor([0.0306, 0.0303, 0.0404, 0.0404, 0.0348, 0.0405, 0.0312, 0.0364], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-27 13:17:16,774 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-04-27 13:17:28,991 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.48 vs. limit=2.0 2023-04-27 13:17:31,835 INFO [finetune.py:976] (6/7) Epoch 19, batch 4850, loss[loss=0.1687, simple_loss=0.2504, pruned_loss=0.04348, over 4882.00 frames. ], tot_loss[loss=0.1754, simple_loss=0.2464, pruned_loss=0.05224, over 952856.45 frames. ], batch size: 32, lr: 3.27e-03, grad_scale: 32.0 2023-04-27 13:17:34,082 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-04-27 13:17:37,356 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=3.46 vs. limit=5.0 2023-04-27 13:17:37,757 INFO [optim.py:369] (6/7) 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 13:18:04,091 INFO [finetune.py:976] (6/7) Epoch 19, batch 4900, loss[loss=0.2092, simple_loss=0.2803, pruned_loss=0.06906, over 4812.00 frames. ], tot_loss[loss=0.1767, simple_loss=0.2479, pruned_loss=0.05277, over 952043.97 frames. ], batch size: 41, lr: 3.27e-03, grad_scale: 32.0 2023-04-27 13:18:07,760 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([0.9843, 1.8940, 1.7338, 1.6649, 2.1095, 1.7637, 2.4923, 1.4972], device='cuda:6'), covar=tensor([0.3706, 0.1990, 0.4655, 0.2800, 0.1546, 0.2276, 0.1535, 0.4290], device='cuda:6'), in_proj_covar=tensor([0.0337, 0.0345, 0.0424, 0.0351, 0.0380, 0.0375, 0.0371, 0.0414], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 13:18:16,704 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=108015.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:18:38,325 INFO [finetune.py:976] (6/7) Epoch 19, batch 4950, loss[loss=0.1507, simple_loss=0.2187, pruned_loss=0.04135, over 4779.00 frames. ], tot_loss[loss=0.1765, simple_loss=0.2484, pruned_loss=0.05233, over 953756.99 frames. ], batch size: 26, lr: 3.27e-03, grad_scale: 32.0 2023-04-27 13:18:45,329 INFO [optim.py:369] (6/7) 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 13:19:01,892 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=3.73 vs. limit=5.0 2023-04-27 13:19:07,763 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=108093.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 13:19:11,338 INFO [finetune.py:976] (6/7) Epoch 19, batch 5000, loss[loss=0.1616, simple_loss=0.2277, pruned_loss=0.04775, over 4918.00 frames. ], tot_loss[loss=0.1754, simple_loss=0.2468, pruned_loss=0.05198, over 952429.00 frames. ], batch size: 36, lr: 3.27e-03, grad_scale: 32.0 2023-04-27 13:19:37,029 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.5674, 1.6929, 1.8949, 1.9895, 1.8990, 1.9745, 1.9881, 2.0349], device='cuda:6'), covar=tensor([0.3859, 0.5283, 0.4873, 0.4398, 0.5388, 0.6964, 0.5168, 0.4829], device='cuda:6'), in_proj_covar=tensor([0.0337, 0.0373, 0.0322, 0.0335, 0.0345, 0.0394, 0.0358, 0.0328], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 13:19:44,760 INFO [finetune.py:976] (6/7) Epoch 19, batch 5050, loss[loss=0.1604, simple_loss=0.2335, pruned_loss=0.04368, over 4801.00 frames. ], tot_loss[loss=0.1738, simple_loss=0.2447, pruned_loss=0.0514, over 952659.42 frames. ], batch size: 25, lr: 3.27e-03, grad_scale: 32.0 2023-04-27 13:19:46,096 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7554, 1.3255, 1.3991, 1.5842, 1.9515, 1.5871, 1.3525, 1.3138], device='cuda:6'), covar=tensor([0.1572, 0.1622, 0.1881, 0.1444, 0.0938, 0.1578, 0.2066, 0.2344], device='cuda:6'), in_proj_covar=tensor([0.0310, 0.0313, 0.0352, 0.0291, 0.0329, 0.0309, 0.0300, 0.0372], device='cuda:6'), out_proj_covar=tensor([6.3797e-05, 6.4925e-05, 7.4542e-05, 5.8903e-05, 6.8234e-05, 6.4828e-05, 6.2934e-05, 7.9251e-05], device='cuda:6') 2023-04-27 13:19:51,163 INFO [optim.py:369] (6/7) 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 13:19:57,093 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=108166.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:20:29,632 INFO [finetune.py:976] (6/7) Epoch 19, batch 5100, loss[loss=0.1303, simple_loss=0.2016, pruned_loss=0.02947, over 4828.00 frames. ], tot_loss[loss=0.1711, simple_loss=0.2415, pruned_loss=0.05034, over 955374.96 frames. ], batch size: 39, lr: 3.27e-03, grad_scale: 32.0 2023-04-27 13:20:39,729 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.1896, 1.8202, 2.0806, 2.4425, 2.4029, 1.9391, 1.7303, 2.1377], device='cuda:6'), covar=tensor([0.0749, 0.1047, 0.0635, 0.0588, 0.0606, 0.0931, 0.0733, 0.0581], device='cuda:6'), in_proj_covar=tensor([0.0189, 0.0204, 0.0184, 0.0174, 0.0179, 0.0183, 0.0153, 0.0182], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 13:20:50,617 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=108214.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:20:58,513 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.35 vs. limit=5.0 2023-04-27 13:21:13,516 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.05 vs. limit=5.0 2023-04-27 13:21:34,997 INFO [finetune.py:976] (6/7) Epoch 19, batch 5150, loss[loss=0.1871, simple_loss=0.2532, pruned_loss=0.06044, over 4930.00 frames. ], tot_loss[loss=0.1735, simple_loss=0.2435, pruned_loss=0.05177, over 955868.48 frames. ], batch size: 33, lr: 3.27e-03, grad_scale: 32.0 2023-04-27 13:21:44,948 INFO [optim.py:369] (6/7) 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 13:22:28,077 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.5350, 1.6580, 1.8546, 1.9339, 1.8122, 1.8948, 1.9515, 1.9631], device='cuda:6'), covar=tensor([0.3813, 0.5602, 0.4854, 0.4725, 0.5744, 0.7499, 0.5003, 0.4972], device='cuda:6'), in_proj_covar=tensor([0.0337, 0.0373, 0.0322, 0.0335, 0.0346, 0.0395, 0.0358, 0.0329], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 13:22:29,161 INFO [finetune.py:976] (6/7) Epoch 19, batch 5200, loss[loss=0.1807, simple_loss=0.2582, pruned_loss=0.05159, over 4826.00 frames. ], tot_loss[loss=0.1761, simple_loss=0.2469, pruned_loss=0.05263, over 957062.32 frames. ], batch size: 33, lr: 3.27e-03, grad_scale: 32.0 2023-04-27 13:22:32,977 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([3.0748, 2.2996, 2.2197, 2.3402, 2.2071, 2.2672, 2.3095, 2.2689], device='cuda:6'), covar=tensor([0.3934, 0.6434, 0.5152, 0.5089, 0.5856, 0.7460, 0.6196, 0.5500], device='cuda:6'), in_proj_covar=tensor([0.0337, 0.0373, 0.0322, 0.0335, 0.0345, 0.0395, 0.0358, 0.0329], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 13:22:51,814 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=108315.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:23:05,671 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=108325.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:23:13,350 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=3.98 vs. limit=5.0 2023-04-27 13:23:36,574 INFO [finetune.py:976] (6/7) Epoch 19, batch 5250, loss[loss=0.1659, simple_loss=0.255, pruned_loss=0.03838, over 4823.00 frames. ], tot_loss[loss=0.1766, simple_loss=0.2481, pruned_loss=0.05252, over 956803.94 frames. ], batch size: 51, lr: 3.27e-03, grad_scale: 32.0 2023-04-27 13:23:44,340 INFO [optim.py:369] (6/7) 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 13:23:53,520 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=108363.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:24:16,838 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.50 vs. limit=2.0 2023-04-27 13:24:19,479 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=108386.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:24:23,684 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=108393.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:24:26,044 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8350, 1.5114, 3.9776, 3.7634, 3.5201, 3.6814, 3.5971, 3.5623], device='cuda:6'), covar=tensor([0.6583, 0.5221, 0.1082, 0.1465, 0.1070, 0.1638, 0.2658, 0.1296], device='cuda:6'), in_proj_covar=tensor([0.0307, 0.0303, 0.0406, 0.0404, 0.0349, 0.0406, 0.0312, 0.0364], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-27 13:24:27,186 INFO [finetune.py:976] (6/7) Epoch 19, batch 5300, loss[loss=0.1579, simple_loss=0.2296, pruned_loss=0.04306, over 4825.00 frames. ], tot_loss[loss=0.1781, simple_loss=0.2498, pruned_loss=0.05317, over 955071.53 frames. ], batch size: 30, lr: 3.27e-03, grad_scale: 32.0 2023-04-27 13:24:50,777 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=108432.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:24:56,173 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=108441.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:24:58,739 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6610, 1.7354, 0.7283, 1.3440, 1.8365, 1.5578, 1.4439, 1.5213], device='cuda:6'), covar=tensor([0.0487, 0.0368, 0.0364, 0.0560, 0.0254, 0.0486, 0.0526, 0.0567], device='cuda:6'), in_proj_covar=tensor([0.0028, 0.0024, 0.0022, 0.0029, 0.0019, 0.0028, 0.0028, 0.0028], device='cuda:6'), out_proj_covar=tensor([0.0050, 0.0045, 0.0038, 0.0051, 0.0038, 0.0049, 0.0049, 0.0051], device='cuda:6') 2023-04-27 13:25:01,040 INFO [finetune.py:976] (6/7) Epoch 19, batch 5350, loss[loss=0.1646, simple_loss=0.2278, pruned_loss=0.05073, over 4906.00 frames. ], tot_loss[loss=0.1768, simple_loss=0.2491, pruned_loss=0.05221, over 956667.06 frames. ], batch size: 37, lr: 3.27e-03, grad_scale: 32.0 2023-04-27 13:25:01,188 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.4910, 1.8020, 1.8130, 1.9714, 1.8079, 1.8747, 1.9300, 1.9292], device='cuda:6'), covar=tensor([0.4040, 0.5278, 0.4324, 0.4187, 0.5293, 0.6817, 0.4968, 0.4692], device='cuda:6'), in_proj_covar=tensor([0.0336, 0.0372, 0.0321, 0.0334, 0.0345, 0.0394, 0.0357, 0.0328], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 13:25:06,497 INFO [optim.py:369] (6/7) 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 13:25:31,257 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=108493.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 13:25:34,732 INFO [finetune.py:976] (6/7) Epoch 19, batch 5400, loss[loss=0.1866, simple_loss=0.2604, pruned_loss=0.05641, over 4899.00 frames. ], tot_loss[loss=0.1751, simple_loss=0.2467, pruned_loss=0.05173, over 957602.50 frames. ], batch size: 36, lr: 3.27e-03, grad_scale: 32.0 2023-04-27 13:25:54,689 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.1617, 2.0380, 2.5616, 2.9117, 2.0048, 1.7090, 2.1486, 1.1777], device='cuda:6'), covar=tensor([0.0520, 0.0662, 0.0338, 0.0535, 0.0691, 0.1080, 0.0691, 0.0733], device='cuda:6'), in_proj_covar=tensor([0.0068, 0.0069, 0.0067, 0.0067, 0.0075, 0.0096, 0.0073, 0.0065], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-27 13:26:28,597 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.1797, 2.6826, 1.0502, 1.3840, 2.0662, 1.3834, 3.3881, 1.8465], device='cuda:6'), covar=tensor([0.0631, 0.0512, 0.0713, 0.1244, 0.0486, 0.0957, 0.0280, 0.0613], device='cuda:6'), in_proj_covar=tensor([0.0051, 0.0065, 0.0048, 0.0047, 0.0050, 0.0052, 0.0074, 0.0051], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:6') 2023-04-27 13:26:31,589 INFO [finetune.py:976] (6/7) Epoch 19, batch 5450, loss[loss=0.1376, simple_loss=0.2155, pruned_loss=0.02988, over 4818.00 frames. ], tot_loss[loss=0.1726, simple_loss=0.2435, pruned_loss=0.05081, over 957115.53 frames. ], batch size: 38, lr: 3.27e-03, grad_scale: 32.0 2023-04-27 13:26:42,378 INFO [optim.py:369] (6/7) 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 13:27:10,248 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=108577.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:27:25,453 INFO [finetune.py:976] (6/7) Epoch 19, batch 5500, loss[loss=0.1843, simple_loss=0.2466, pruned_loss=0.06103, over 4847.00 frames. ], tot_loss[loss=0.1699, simple_loss=0.2404, pruned_loss=0.04967, over 956111.96 frames. ], batch size: 49, lr: 3.27e-03, grad_scale: 32.0 2023-04-27 13:27:35,761 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.58 vs. limit=5.0 2023-04-27 13:27:51,518 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=108638.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:27:58,829 INFO [finetune.py:976] (6/7) Epoch 19, batch 5550, loss[loss=0.1426, simple_loss=0.2168, pruned_loss=0.03418, over 4760.00 frames. ], tot_loss[loss=0.1708, simple_loss=0.2416, pruned_loss=0.04999, over 953834.22 frames. ], batch size: 27, lr: 3.27e-03, grad_scale: 32.0 2023-04-27 13:28:09,998 INFO [optim.py:369] (6/7) 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] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=108681.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:29:03,286 INFO [finetune.py:976] (6/7) Epoch 19, batch 5600, loss[loss=0.1238, simple_loss=0.1916, pruned_loss=0.02797, over 4265.00 frames. ], tot_loss[loss=0.174, simple_loss=0.2454, pruned_loss=0.05132, over 952091.01 frames. ], batch size: 18, lr: 3.27e-03, grad_scale: 32.0 2023-04-27 13:29:12,627 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.3432, 1.9594, 2.2573, 2.7548, 2.6425, 2.2571, 1.8373, 2.4137], device='cuda:6'), covar=tensor([0.0866, 0.1141, 0.0636, 0.0488, 0.0632, 0.0829, 0.0766, 0.0564], device='cuda:6'), in_proj_covar=tensor([0.0189, 0.0203, 0.0184, 0.0173, 0.0180, 0.0183, 0.0153, 0.0181], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 13:29:27,145 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-04-27 13:30:00,554 INFO [finetune.py:976] (6/7) Epoch 19, batch 5650, loss[loss=0.228, simple_loss=0.2949, pruned_loss=0.08057, over 4904.00 frames. ], tot_loss[loss=0.1768, simple_loss=0.249, pruned_loss=0.05231, over 952801.52 frames. ], batch size: 43, lr: 3.26e-03, grad_scale: 32.0 2023-04-27 13:30:17,475 INFO [optim.py:369] (6/7) 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 13:30:52,604 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=108788.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 13:30:54,440 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=108791.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:30:55,316 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-04-27 13:31:04,843 INFO [finetune.py:976] (6/7) Epoch 19, batch 5700, loss[loss=0.1263, simple_loss=0.1944, pruned_loss=0.02909, over 4230.00 frames. ], tot_loss[loss=0.1742, simple_loss=0.245, pruned_loss=0.05174, over 932390.47 frames. ], batch size: 18, lr: 3.26e-03, grad_scale: 32.0 2023-04-27 13:31:40,722 INFO [finetune.py:976] (6/7) Epoch 20, batch 0, loss[loss=0.1928, simple_loss=0.262, pruned_loss=0.06175, over 4883.00 frames. ], tot_loss[loss=0.1928, simple_loss=0.262, pruned_loss=0.06175, over 4883.00 frames. ], batch size: 35, lr: 3.26e-03, grad_scale: 32.0 2023-04-27 13:31:40,722 INFO [finetune.py:1001] (6/7) Computing validation loss 2023-04-27 13:31:48,446 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.1470, 2.5765, 1.2242, 1.5697, 1.8838, 1.5123, 2.8570, 1.8343], device='cuda:6'), covar=tensor([0.0531, 0.0590, 0.0632, 0.0928, 0.0375, 0.0714, 0.0233, 0.0461], device='cuda:6'), in_proj_covar=tensor([0.0051, 0.0065, 0.0048, 0.0047, 0.0050, 0.0052, 0.0074, 0.0051], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:6') 2023-04-27 13:31:49,822 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.3420, 1.4974, 1.8338, 1.9829, 1.9128, 2.0655, 1.8615, 1.9206], device='cuda:6'), covar=tensor([0.4121, 0.5487, 0.4649, 0.4130, 0.5649, 0.6661, 0.5303, 0.4692], device='cuda:6'), in_proj_covar=tensor([0.0338, 0.0373, 0.0321, 0.0334, 0.0345, 0.0394, 0.0357, 0.0328], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 13:31:57,317 INFO [finetune.py:1010] (6/7) 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] (6/7) Maximum memory allocated so far is 6365MB 2023-04-27 13:32:13,987 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=108852.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:32:17,458 INFO [optim.py:369] (6/7) 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 13:32:18,756 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-04-27 13:32:30,441 INFO [finetune.py:976] (6/7) Epoch 20, batch 50, loss[loss=0.1448, simple_loss=0.2241, pruned_loss=0.03275, over 4906.00 frames. ], tot_loss[loss=0.1762, simple_loss=0.2484, pruned_loss=0.052, over 217813.72 frames. ], batch size: 37, lr: 3.26e-03, grad_scale: 32.0 2023-04-27 13:32:45,185 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([3.5676, 3.5522, 2.5820, 4.1660, 3.6953, 3.5612, 1.8014, 3.5037], device='cuda:6'), covar=tensor([0.2163, 0.1310, 0.3159, 0.2115, 0.2326, 0.1972, 0.5290, 0.2556], device='cuda:6'), in_proj_covar=tensor([0.0244, 0.0213, 0.0247, 0.0303, 0.0293, 0.0246, 0.0269, 0.0269], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 13:33:03,610 INFO [finetune.py:976] (6/7) Epoch 20, batch 100, loss[loss=0.1496, simple_loss=0.2146, pruned_loss=0.04231, over 4528.00 frames. ], tot_loss[loss=0.1727, simple_loss=0.243, pruned_loss=0.05126, over 379072.27 frames. ], batch size: 19, lr: 3.26e-03, grad_scale: 32.0 2023-04-27 13:33:08,762 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=108933.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:33:22,265 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.2708, 2.6989, 2.2031, 2.6287, 1.9031, 2.2707, 2.2203, 1.8255], device='cuda:6'), covar=tensor([0.1655, 0.0973, 0.0783, 0.0922, 0.3107, 0.1097, 0.1784, 0.2309], device='cuda:6'), in_proj_covar=tensor([0.0288, 0.0303, 0.0218, 0.0279, 0.0314, 0.0260, 0.0251, 0.0265], device='cuda:6'), out_proj_covar=tensor([1.1557e-04, 1.2006e-04, 8.6276e-05, 1.1079e-04, 1.2740e-04, 1.0276e-04, 1.0141e-04, 1.0494e-04], device='cuda:6') 2023-04-27 13:33:23,936 INFO [optim.py:369] (6/7) 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 13:33:36,946 INFO [finetune.py:976] (6/7) Epoch 20, batch 150, loss[loss=0.1997, simple_loss=0.2627, pruned_loss=0.06833, over 4907.00 frames. ], tot_loss[loss=0.1703, simple_loss=0.2399, pruned_loss=0.05034, over 508123.05 frames. ], batch size: 32, lr: 3.26e-03, grad_scale: 32.0 2023-04-27 13:33:39,882 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=108981.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:33:40,996 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5114, 1.4715, 1.8094, 1.8742, 1.3803, 1.3033, 1.5231, 0.9558], device='cuda:6'), covar=tensor([0.0651, 0.0615, 0.0386, 0.0574, 0.0754, 0.1124, 0.0635, 0.0682], device='cuda:6'), in_proj_covar=tensor([0.0068, 0.0068, 0.0067, 0.0067, 0.0074, 0.0096, 0.0073, 0.0065], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-27 13:34:09,757 INFO [finetune.py:976] (6/7) Epoch 20, batch 200, loss[loss=0.2432, simple_loss=0.3049, pruned_loss=0.09078, over 4832.00 frames. ], tot_loss[loss=0.1712, simple_loss=0.2404, pruned_loss=0.051, over 608667.47 frames. ], batch size: 40, lr: 3.26e-03, grad_scale: 32.0 2023-04-27 13:34:11,059 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=109029.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:34:16,943 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=109037.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 13:34:29,138 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0586, 2.4678, 2.0317, 2.4038, 1.6079, 2.1501, 2.1251, 1.6488], device='cuda:6'), covar=tensor([0.1971, 0.1055, 0.0740, 0.1193, 0.3452, 0.1068, 0.1888, 0.2719], device='cuda:6'), in_proj_covar=tensor([0.0286, 0.0301, 0.0216, 0.0278, 0.0311, 0.0257, 0.0249, 0.0263], device='cuda:6'), out_proj_covar=tensor([1.1458e-04, 1.1925e-04, 8.5426e-05, 1.1006e-04, 1.2634e-04, 1.0171e-04, 1.0065e-04, 1.0415e-04], device='cuda:6') 2023-04-27 13:34:29,595 INFO [optim.py:369] (6/7) 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 13:34:42,790 INFO [finetune.py:976] (6/7) Epoch 20, batch 250, loss[loss=0.1295, simple_loss=0.2028, pruned_loss=0.02808, over 4821.00 frames. ], tot_loss[loss=0.1723, simple_loss=0.2421, pruned_loss=0.0513, over 685488.19 frames. ], batch size: 25, lr: 3.26e-03, grad_scale: 32.0 2023-04-27 13:35:02,385 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=109088.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:35:13,176 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=109098.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 13:35:25,460 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.6995, 1.9594, 2.0543, 2.1415, 2.0153, 2.1472, 2.0985, 2.1540], device='cuda:6'), covar=tensor([0.3838, 0.5480, 0.4870, 0.4733, 0.5478, 0.6860, 0.5212, 0.4864], device='cuda:6'), in_proj_covar=tensor([0.0339, 0.0374, 0.0323, 0.0336, 0.0347, 0.0395, 0.0358, 0.0330], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 13:35:31,295 INFO [finetune.py:976] (6/7) Epoch 20, batch 300, loss[loss=0.2095, simple_loss=0.2821, pruned_loss=0.06844, over 4812.00 frames. ], tot_loss[loss=0.1748, simple_loss=0.2457, pruned_loss=0.05196, over 746577.59 frames. ], batch size: 38, lr: 3.26e-03, grad_scale: 32.0 2023-04-27 13:35:49,172 INFO [zipformer.py:1188] (6/7) 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] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=109147.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:36:01,826 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6675, 2.1309, 1.5117, 1.4437, 1.2777, 1.2710, 1.4496, 1.2042], device='cuda:6'), covar=tensor([0.1999, 0.1491, 0.1760, 0.1959, 0.2774, 0.2396, 0.1248, 0.2318], device='cuda:6'), in_proj_covar=tensor([0.0197, 0.0212, 0.0169, 0.0203, 0.0200, 0.0185, 0.0156, 0.0186], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:6') 2023-04-27 13:36:03,479 INFO [optim.py:369] (6/7) 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 13:36:25,686 INFO [finetune.py:976] (6/7) Epoch 20, batch 350, loss[loss=0.1923, simple_loss=0.2728, pruned_loss=0.0559, over 4914.00 frames. ], tot_loss[loss=0.1764, simple_loss=0.2483, pruned_loss=0.05221, over 793809.04 frames. ], batch size: 38, lr: 3.26e-03, grad_scale: 32.0 2023-04-27 13:36:47,072 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=109194.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:36:48,167 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9312, 1.6305, 1.8477, 2.2135, 2.2273, 1.8757, 1.6463, 2.0000], device='cuda:6'), covar=tensor([0.0750, 0.1074, 0.0665, 0.0462, 0.0597, 0.0783, 0.0679, 0.0529], device='cuda:6'), in_proj_covar=tensor([0.0185, 0.0199, 0.0180, 0.0170, 0.0176, 0.0180, 0.0150, 0.0178], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 13:37:01,828 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9093, 1.1182, 1.4916, 1.6086, 1.5597, 1.6580, 1.5230, 1.5688], device='cuda:6'), covar=tensor([0.3402, 0.4510, 0.3769, 0.3847, 0.4626, 0.6291, 0.4034, 0.3782], device='cuda:6'), in_proj_covar=tensor([0.0338, 0.0373, 0.0323, 0.0336, 0.0346, 0.0396, 0.0358, 0.0329], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 13:37:21,407 INFO [finetune.py:976] (6/7) Epoch 20, batch 400, loss[loss=0.1643, simple_loss=0.2412, pruned_loss=0.04367, over 4929.00 frames. ], tot_loss[loss=0.179, simple_loss=0.251, pruned_loss=0.05348, over 831519.98 frames. ], batch size: 41, lr: 3.26e-03, grad_scale: 32.0 2023-04-27 13:37:31,618 INFO [zipformer.py:1188] (6/7) 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] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=109255.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:38:05,570 INFO [optim.py:369] (6/7) 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 13:38:22,933 INFO [finetune.py:976] (6/7) Epoch 20, batch 450, loss[loss=0.1251, simple_loss=0.1904, pruned_loss=0.02994, over 4112.00 frames. ], tot_loss[loss=0.177, simple_loss=0.2487, pruned_loss=0.0526, over 857595.68 frames. ], batch size: 18, lr: 3.26e-03, grad_scale: 32.0 2023-04-27 13:38:25,937 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=109281.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:38:34,977 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=109294.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:38:50,558 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=109318.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:38:54,502 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.79 vs. limit=5.0 2023-04-27 13:38:55,898 INFO [finetune.py:976] (6/7) Epoch 20, batch 500, loss[loss=0.2274, simple_loss=0.2711, pruned_loss=0.0919, over 4719.00 frames. ], tot_loss[loss=0.1753, simple_loss=0.2462, pruned_loss=0.05214, over 879495.42 frames. ], batch size: 54, lr: 3.26e-03, grad_scale: 32.0 2023-04-27 13:39:16,053 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=109355.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:39:17,767 INFO [optim.py:369] (6/7) 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 13:39:24,029 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9136, 1.5896, 1.5448, 1.6799, 2.0918, 1.6837, 1.4201, 1.4263], device='cuda:6'), covar=tensor([0.1416, 0.1367, 0.1776, 0.1229, 0.0733, 0.1423, 0.1766, 0.2068], device='cuda:6'), in_proj_covar=tensor([0.0303, 0.0306, 0.0344, 0.0284, 0.0320, 0.0302, 0.0295, 0.0364], device='cuda:6'), out_proj_covar=tensor([6.2256e-05, 6.3329e-05, 7.2701e-05, 5.7506e-05, 6.6126e-05, 6.3451e-05, 6.1823e-05, 7.7397e-05], device='cuda:6') 2023-04-27 13:39:29,356 INFO [finetune.py:976] (6/7) Epoch 20, batch 550, loss[loss=0.2001, simple_loss=0.265, pruned_loss=0.06759, over 4908.00 frames. ], tot_loss[loss=0.1725, simple_loss=0.2429, pruned_loss=0.05105, over 896862.94 frames. ], batch size: 35, lr: 3.26e-03, grad_scale: 32.0 2023-04-27 13:39:30,089 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9755, 1.8964, 2.4432, 2.5929, 1.8861, 1.6442, 2.0182, 1.2136], device='cuda:6'), covar=tensor([0.0615, 0.0850, 0.0409, 0.0830, 0.0832, 0.1327, 0.0714, 0.0766], device='cuda:6'), in_proj_covar=tensor([0.0068, 0.0068, 0.0067, 0.0067, 0.0074, 0.0096, 0.0073, 0.0065], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-27 13:39:30,724 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=109379.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 13:39:40,662 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=109393.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 13:40:02,927 INFO [finetune.py:976] (6/7) Epoch 20, batch 600, loss[loss=0.1893, simple_loss=0.2704, pruned_loss=0.05411, over 4829.00 frames. ], tot_loss[loss=0.1728, simple_loss=0.2433, pruned_loss=0.05118, over 910664.19 frames. ], batch size: 39, lr: 3.26e-03, grad_scale: 64.0 2023-04-27 13:40:07,332 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=3.85 vs. limit=5.0 2023-04-27 13:40:08,357 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-04-27 13:40:16,628 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=109447.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:40:20,216 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9477, 1.6069, 1.4813, 1.7052, 2.1399, 1.6945, 1.4085, 1.4169], device='cuda:6'), covar=tensor([0.1430, 0.1416, 0.2002, 0.1434, 0.0779, 0.1885, 0.2142, 0.2279], device='cuda:6'), in_proj_covar=tensor([0.0304, 0.0307, 0.0346, 0.0287, 0.0322, 0.0305, 0.0297, 0.0366], device='cuda:6'), out_proj_covar=tensor([6.2646e-05, 6.3675e-05, 7.3143e-05, 5.8003e-05, 6.6640e-05, 6.3943e-05, 6.2197e-05, 7.7932e-05], device='cuda:6') 2023-04-27 13:40:24,267 INFO [optim.py:369] (6/7) 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 13:40:30,343 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7717, 1.9989, 0.9648, 1.4976, 2.1834, 1.6511, 1.5507, 1.5886], device='cuda:6'), covar=tensor([0.0490, 0.0346, 0.0313, 0.0553, 0.0230, 0.0488, 0.0474, 0.0552], device='cuda:6'), in_proj_covar=tensor([0.0028, 0.0024, 0.0022, 0.0029, 0.0019, 0.0028, 0.0028, 0.0028], device='cuda:6'), out_proj_covar=tensor([0.0050, 0.0045, 0.0038, 0.0052, 0.0038, 0.0049, 0.0049, 0.0051], device='cuda:6') 2023-04-27 13:40:36,278 INFO [finetune.py:976] (6/7) Epoch 20, batch 650, loss[loss=0.2053, simple_loss=0.2808, pruned_loss=0.06494, over 4820.00 frames. ], tot_loss[loss=0.1739, simple_loss=0.2454, pruned_loss=0.05122, over 920840.70 frames. ], batch size: 40, lr: 3.26e-03, grad_scale: 64.0 2023-04-27 13:40:38,850 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.3753, 3.1985, 2.5400, 2.8529, 2.2056, 2.6514, 2.7892, 2.1562], device='cuda:6'), covar=tensor([0.1780, 0.0838, 0.0652, 0.1159, 0.2936, 0.1008, 0.1604, 0.2324], device='cuda:6'), in_proj_covar=tensor([0.0292, 0.0306, 0.0220, 0.0282, 0.0316, 0.0262, 0.0254, 0.0268], device='cuda:6'), out_proj_covar=tensor([1.1694e-04, 1.2128e-04, 8.7217e-05, 1.1190e-04, 1.2821e-04, 1.0381e-04, 1.0260e-04, 1.0631e-04], device='cuda:6') 2023-04-27 13:40:48,776 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=109495.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:40:53,114 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.2266, 1.8901, 2.3577, 2.6756, 2.2539, 2.1165, 2.2403, 2.2055], device='cuda:6'), covar=tensor([0.4802, 0.7401, 0.7169, 0.5742, 0.6057, 0.8478, 0.9058, 0.8576], device='cuda:6'), in_proj_covar=tensor([0.0430, 0.0412, 0.0506, 0.0507, 0.0457, 0.0485, 0.0492, 0.0499], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 13:41:10,003 INFO [finetune.py:976] (6/7) Epoch 20, batch 700, loss[loss=0.1253, simple_loss=0.1961, pruned_loss=0.02721, over 4734.00 frames. ], tot_loss[loss=0.1753, simple_loss=0.2469, pruned_loss=0.05183, over 927002.29 frames. ], batch size: 23, lr: 3.26e-03, grad_scale: 64.0 2023-04-27 13:41:10,739 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9010, 1.6217, 1.8159, 2.1817, 2.2216, 1.8479, 1.6201, 1.9970], device='cuda:6'), covar=tensor([0.0693, 0.1059, 0.0612, 0.0472, 0.0510, 0.0727, 0.0640, 0.0497], device='cuda:6'), in_proj_covar=tensor([0.0185, 0.0198, 0.0180, 0.0170, 0.0175, 0.0178, 0.0150, 0.0177], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 13:41:25,581 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=109550.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:41:30,318 INFO [optim.py:369] (6/7) 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 13:41:36,184 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9793, 1.6806, 1.4633, 1.9342, 2.1669, 1.8109, 1.7853, 1.4908], device='cuda:6'), covar=tensor([0.1623, 0.1548, 0.1363, 0.1516, 0.0924, 0.1707, 0.1566, 0.1961], device='cuda:6'), in_proj_covar=tensor([0.0304, 0.0307, 0.0345, 0.0285, 0.0321, 0.0303, 0.0296, 0.0366], device='cuda:6'), out_proj_covar=tensor([6.2583e-05, 6.3569e-05, 7.3073e-05, 5.7642e-05, 6.6440e-05, 6.3682e-05, 6.2068e-05, 7.7813e-05], device='cuda:6') 2023-04-27 13:41:43,810 INFO [finetune.py:976] (6/7) Epoch 20, batch 750, loss[loss=0.1394, simple_loss=0.2172, pruned_loss=0.0308, over 4827.00 frames. ], tot_loss[loss=0.1756, simple_loss=0.2476, pruned_loss=0.05174, over 933735.84 frames. ], batch size: 30, lr: 3.26e-03, grad_scale: 64.0 2023-04-27 13:41:45,086 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=109579.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:42:49,902 INFO [finetune.py:976] (6/7) Epoch 20, batch 800, loss[loss=0.1634, simple_loss=0.2336, pruned_loss=0.04665, over 4927.00 frames. ], tot_loss[loss=0.1759, simple_loss=0.2481, pruned_loss=0.05182, over 939426.58 frames. ], batch size: 33, lr: 3.25e-03, grad_scale: 64.0 2023-04-27 13:43:07,641 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=109640.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:43:19,833 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=109650.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:43:22,876 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.3336, 1.5031, 1.4469, 1.7157, 1.6395, 1.9051, 1.4261, 3.3461], device='cuda:6'), covar=tensor([0.0610, 0.0830, 0.0789, 0.1242, 0.0628, 0.0467, 0.0727, 0.0148], device='cuda:6'), in_proj_covar=tensor([0.0038, 0.0039, 0.0040, 0.0043, 0.0040, 0.0038, 0.0038, 0.0056], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:6') 2023-04-27 13:43:27,913 INFO [optim.py:369] (6/7) 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 13:43:28,672 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=109659.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:43:50,517 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=109674.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 13:43:52,232 INFO [finetune.py:976] (6/7) Epoch 20, batch 850, loss[loss=0.1644, simple_loss=0.2429, pruned_loss=0.043, over 4833.00 frames. ], tot_loss[loss=0.1731, simple_loss=0.2448, pruned_loss=0.05075, over 943161.29 frames. ], batch size: 33, lr: 3.25e-03, grad_scale: 64.0 2023-04-27 13:44:08,063 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=109693.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 13:44:26,932 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 20, batch 900, loss[loss=0.1383, simple_loss=0.2132, pruned_loss=0.03163, over 4808.00 frames. ], tot_loss[loss=0.1724, simple_loss=0.243, pruned_loss=0.05088, over 944721.76 frames. ], batch size: 25, lr: 3.25e-03, grad_scale: 64.0 2023-04-27 13:44:34,798 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.4691, 1.7359, 1.8657, 1.9504, 1.8041, 1.8576, 1.8962, 1.8970], device='cuda:6'), covar=tensor([0.3867, 0.5240, 0.4386, 0.4369, 0.5295, 0.6791, 0.5385, 0.4992], device='cuda:6'), in_proj_covar=tensor([0.0335, 0.0372, 0.0320, 0.0334, 0.0344, 0.0393, 0.0357, 0.0327], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 13:44:40,149 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=109741.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 13:44:45,639 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.4508, 1.6098, 4.2198, 3.9519, 3.6807, 4.0557, 3.9904, 3.7492], device='cuda:6'), covar=tensor([0.6997, 0.5179, 0.1038, 0.1607, 0.1144, 0.1914, 0.1168, 0.1393], device='cuda:6'), in_proj_covar=tensor([0.0309, 0.0303, 0.0403, 0.0404, 0.0348, 0.0406, 0.0312, 0.0364], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-27 13:44:50,920 INFO [optim.py:369] (6/7) 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] (6/7) Epoch 20, batch 950, loss[loss=0.2188, simple_loss=0.2812, pruned_loss=0.07815, over 4826.00 frames. ], tot_loss[loss=0.1727, simple_loss=0.2429, pruned_loss=0.05122, over 948087.87 frames. ], batch size: 39, lr: 3.25e-03, grad_scale: 32.0 2023-04-27 13:45:38,014 INFO [finetune.py:976] (6/7) Epoch 20, batch 1000, loss[loss=0.1969, simple_loss=0.2605, pruned_loss=0.06662, over 4833.00 frames. ], tot_loss[loss=0.1733, simple_loss=0.2435, pruned_loss=0.05151, over 948276.87 frames. ], batch size: 30, lr: 3.25e-03, grad_scale: 32.0 2023-04-27 13:45:39,656 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.80 vs. limit=2.0 2023-04-27 13:45:52,583 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=109850.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:45:58,963 INFO [optim.py:369] (6/7) 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 13:46:01,514 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([4.0468, 3.9770, 2.9116, 4.6389, 4.0442, 4.0211, 1.5317, 3.9598], device='cuda:6'), covar=tensor([0.1588, 0.1047, 0.2567, 0.1405, 0.3247, 0.1571, 0.5970, 0.2080], device='cuda:6'), in_proj_covar=tensor([0.0245, 0.0215, 0.0249, 0.0304, 0.0294, 0.0246, 0.0272, 0.0271], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 13:46:08,743 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6836, 1.7525, 1.0567, 1.3778, 1.9579, 1.5635, 1.4306, 1.6032], device='cuda:6'), covar=tensor([0.0499, 0.0367, 0.0309, 0.0526, 0.0249, 0.0516, 0.0487, 0.0562], device='cuda:6'), in_proj_covar=tensor([0.0028, 0.0024, 0.0022, 0.0029, 0.0019, 0.0028, 0.0028, 0.0028], device='cuda:6'), out_proj_covar=tensor([0.0050, 0.0045, 0.0038, 0.0051, 0.0038, 0.0049, 0.0049, 0.0051], device='cuda:6') 2023-04-27 13:46:11,418 INFO [finetune.py:976] (6/7) Epoch 20, batch 1050, loss[loss=0.1568, simple_loss=0.2287, pruned_loss=0.04248, over 4908.00 frames. ], tot_loss[loss=0.1753, simple_loss=0.2465, pruned_loss=0.05206, over 950243.43 frames. ], batch size: 32, lr: 3.25e-03, grad_scale: 32.0 2023-04-27 13:46:11,629 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-04-27 13:46:25,238 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=109898.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:46:43,731 INFO [finetune.py:976] (6/7) Epoch 20, batch 1100, loss[loss=0.1904, simple_loss=0.2589, pruned_loss=0.06092, over 4903.00 frames. ], tot_loss[loss=0.1766, simple_loss=0.2481, pruned_loss=0.0526, over 951666.09 frames. ], batch size: 35, lr: 3.25e-03, grad_scale: 32.0 2023-04-27 13:46:50,145 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=109935.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:46:59,876 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=109950.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:47:01,449 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-04-27 13:47:05,249 INFO [optim.py:369] (6/7) 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 13:47:14,449 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.47 vs. limit=2.0 2023-04-27 13:47:15,568 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5542, 0.6772, 1.4785, 1.9639, 1.6339, 1.4624, 1.4990, 1.5164], device='cuda:6'), covar=tensor([0.4280, 0.6258, 0.5811, 0.5667, 0.5825, 0.7363, 0.7437, 0.7736], device='cuda:6'), in_proj_covar=tensor([0.0431, 0.0413, 0.0507, 0.0508, 0.0458, 0.0486, 0.0493, 0.0499], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 13:47:16,149 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 20, batch 1150, loss[loss=0.2004, simple_loss=0.2673, pruned_loss=0.06678, over 4887.00 frames. ], tot_loss[loss=0.1776, simple_loss=0.2496, pruned_loss=0.05283, over 950301.98 frames. ], batch size: 32, lr: 3.25e-03, grad_scale: 32.0 2023-04-27 13:47:40,683 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=109998.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:47:52,882 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=110015.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:47:54,544 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=110017.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:47:57,537 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=110022.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:47:57,841 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.78 vs. limit=2.0 2023-04-27 13:48:00,999 INFO [finetune.py:976] (6/7) Epoch 20, batch 1200, loss[loss=0.1439, simple_loss=0.227, pruned_loss=0.03043, over 4754.00 frames. ], tot_loss[loss=0.1749, simple_loss=0.2466, pruned_loss=0.05156, over 949303.58 frames. ], batch size: 27, lr: 3.25e-03, grad_scale: 32.0 2023-04-27 13:48:17,802 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=110044.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:48:26,693 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-04-27 13:48:37,463 INFO [optim.py:369] (6/7) 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 13:49:01,339 INFO [finetune.py:976] (6/7) Epoch 20, batch 1250, loss[loss=0.2057, simple_loss=0.2641, pruned_loss=0.07367, over 4776.00 frames. ], tot_loss[loss=0.1731, simple_loss=0.2442, pruned_loss=0.05102, over 950051.76 frames. ], batch size: 26, lr: 3.25e-03, grad_scale: 32.0 2023-04-27 13:49:02,080 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=110078.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:49:25,652 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=110105.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:49:28,454 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.75 vs. limit=5.0 2023-04-27 13:49:28,690 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=110110.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:49:31,751 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=110115.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:49:40,407 INFO [finetune.py:976] (6/7) Epoch 20, batch 1300, loss[loss=0.1831, simple_loss=0.2469, pruned_loss=0.05961, over 4934.00 frames. ], tot_loss[loss=0.1703, simple_loss=0.2409, pruned_loss=0.04982, over 951452.21 frames. ], batch size: 38, lr: 3.25e-03, grad_scale: 32.0 2023-04-27 13:49:41,217 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.69 vs. limit=2.0 2023-04-27 13:49:49,919 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.1225, 1.8338, 2.0138, 2.4382, 2.4125, 2.0576, 1.7865, 2.1413], device='cuda:6'), covar=tensor([0.0631, 0.0967, 0.0546, 0.0365, 0.0479, 0.0645, 0.0664, 0.0448], device='cuda:6'), in_proj_covar=tensor([0.0188, 0.0201, 0.0182, 0.0172, 0.0178, 0.0181, 0.0152, 0.0179], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 13:50:01,794 INFO [optim.py:369] (6/7) 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 13:50:06,777 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7494, 1.7763, 0.9042, 1.3418, 1.9299, 1.5620, 1.4320, 1.5494], device='cuda:6'), covar=tensor([0.0504, 0.0369, 0.0328, 0.0569, 0.0257, 0.0553, 0.0527, 0.0556], device='cuda:6'), in_proj_covar=tensor([0.0028, 0.0024, 0.0022, 0.0029, 0.0019, 0.0028, 0.0028, 0.0028], device='cuda:6'), out_proj_covar=tensor([0.0050, 0.0045, 0.0038, 0.0052, 0.0038, 0.0049, 0.0050, 0.0051], device='cuda:6') 2023-04-27 13:50:09,733 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=110171.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:50:13,294 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 20, batch 1350, loss[loss=0.1908, simple_loss=0.2546, pruned_loss=0.06353, over 4741.00 frames. ], tot_loss[loss=0.1704, simple_loss=0.2407, pruned_loss=0.05006, over 952276.71 frames. ], batch size: 27, lr: 3.25e-03, grad_scale: 32.0 2023-04-27 13:50:40,220 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5380, 2.5438, 1.9550, 2.2641, 2.6404, 2.2205, 3.4171, 1.6893], device='cuda:6'), covar=tensor([0.4013, 0.2337, 0.5155, 0.3439, 0.1972, 0.2613, 0.1551, 0.5162], device='cuda:6'), in_proj_covar=tensor([0.0336, 0.0344, 0.0422, 0.0349, 0.0378, 0.0371, 0.0368, 0.0415], device='cuda:6'), out_proj_covar=tensor([9.9968e-05, 1.0315e-04, 1.2829e-04, 1.0545e-04, 1.1258e-04, 1.1096e-04, 1.0848e-04, 1.2549e-04], device='cuda:6') 2023-04-27 13:50:47,122 INFO [finetune.py:976] (6/7) Epoch 20, batch 1400, loss[loss=0.1926, simple_loss=0.2575, pruned_loss=0.06383, over 4804.00 frames. ], tot_loss[loss=0.1748, simple_loss=0.2453, pruned_loss=0.05219, over 952572.22 frames. ], batch size: 25, lr: 3.25e-03, grad_scale: 32.0 2023-04-27 13:50:52,581 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=110235.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:51:09,026 INFO [optim.py:369] (6/7) 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 13:51:14,180 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=5.20 vs. limit=5.0 2023-04-27 13:51:19,999 INFO [finetune.py:976] (6/7) Epoch 20, batch 1450, loss[loss=0.1431, simple_loss=0.2156, pruned_loss=0.03528, over 4749.00 frames. ], tot_loss[loss=0.1741, simple_loss=0.2456, pruned_loss=0.05134, over 953000.92 frames. ], batch size: 27, lr: 3.25e-03, grad_scale: 32.0 2023-04-27 13:51:25,118 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=110283.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:51:25,184 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=110283.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:51:38,801 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=110303.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:51:46,531 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=110315.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:51:48,401 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=110318.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:51:51,170 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.47 vs. limit=2.0 2023-04-27 13:51:53,749 INFO [finetune.py:976] (6/7) Epoch 20, batch 1500, loss[loss=0.1448, simple_loss=0.2275, pruned_loss=0.03107, over 4758.00 frames. ], tot_loss[loss=0.1749, simple_loss=0.247, pruned_loss=0.05144, over 955287.02 frames. ], batch size: 28, lr: 3.25e-03, grad_scale: 32.0 2023-04-27 13:52:05,644 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=110344.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:52:16,149 INFO [optim.py:369] (6/7) 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 13:52:19,124 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=110363.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:52:19,787 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=110364.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:52:25,266 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 20, batch 1550, loss[loss=0.2019, simple_loss=0.2667, pruned_loss=0.06857, over 4865.00 frames. ], tot_loss[loss=0.1753, simple_loss=0.2477, pruned_loss=0.05145, over 956459.15 frames. ], batch size: 31, lr: 3.25e-03, grad_scale: 32.0 2023-04-27 13:52:34,535 INFO [zipformer.py:1188] (6/7) 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] (6/7) 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] (6/7) Epoch 20, batch 1600, loss[loss=0.1778, simple_loss=0.2545, pruned_loss=0.05056, over 4753.00 frames. ], tot_loss[loss=0.1738, simple_loss=0.2455, pruned_loss=0.05104, over 957370.27 frames. ], batch size: 26, lr: 3.25e-03, grad_scale: 32.0 2023-04-27 13:54:08,137 INFO [optim.py:369] (6/7) 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] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=110466.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:54:21,383 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=110471.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:54:30,174 INFO [finetune.py:976] (6/7) Epoch 20, batch 1650, loss[loss=0.1497, simple_loss=0.2157, pruned_loss=0.04188, over 4826.00 frames. ], tot_loss[loss=0.1725, simple_loss=0.2434, pruned_loss=0.05079, over 957801.90 frames. ], batch size: 51, lr: 3.25e-03, grad_scale: 32.0 2023-04-27 13:54:48,674 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9624, 1.8857, 2.3620, 2.6559, 1.8948, 1.6903, 1.9671, 1.1672], device='cuda:6'), covar=tensor([0.0595, 0.0849, 0.0418, 0.0751, 0.0822, 0.1126, 0.0774, 0.0777], device='cuda:6'), in_proj_covar=tensor([0.0069, 0.0069, 0.0067, 0.0068, 0.0075, 0.0097, 0.0073, 0.0065], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-27 13:55:09,360 INFO [finetune.py:976] (6/7) Epoch 20, batch 1700, loss[loss=0.1956, simple_loss=0.2789, pruned_loss=0.05613, over 4804.00 frames. ], tot_loss[loss=0.1704, simple_loss=0.2408, pruned_loss=0.04999, over 955340.59 frames. ], batch size: 41, lr: 3.25e-03, grad_scale: 32.0 2023-04-27 13:55:22,338 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9822, 1.7074, 1.8209, 2.1882, 2.2679, 1.9111, 1.5699, 2.0147], device='cuda:6'), covar=tensor([0.0630, 0.0851, 0.0617, 0.0442, 0.0498, 0.0646, 0.0654, 0.0466], device='cuda:6'), in_proj_covar=tensor([0.0187, 0.0200, 0.0183, 0.0173, 0.0178, 0.0181, 0.0152, 0.0180], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 13:55:31,154 INFO [optim.py:369] (6/7) 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 13:55:43,126 INFO [finetune.py:976] (6/7) Epoch 20, batch 1750, loss[loss=0.1759, simple_loss=0.2555, pruned_loss=0.04811, over 4816.00 frames. ], tot_loss[loss=0.1717, simple_loss=0.2422, pruned_loss=0.05063, over 955753.87 frames. ], batch size: 40, lr: 3.25e-03, grad_scale: 32.0 2023-04-27 13:55:54,702 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.0252, 1.2770, 1.1430, 1.5862, 1.3380, 1.5257, 1.2628, 3.0142], device='cuda:6'), covar=tensor([0.0766, 0.1168, 0.1129, 0.1404, 0.0882, 0.0637, 0.1008, 0.0248], device='cuda:6'), in_proj_covar=tensor([0.0038, 0.0039, 0.0040, 0.0043, 0.0040, 0.0038, 0.0038, 0.0056], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0013, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:6') 2023-04-27 13:56:06,621 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=110611.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:56:17,239 INFO [finetune.py:976] (6/7) Epoch 20, batch 1800, loss[loss=0.1967, simple_loss=0.2764, pruned_loss=0.05844, over 4896.00 frames. ], tot_loss[loss=0.1741, simple_loss=0.2459, pruned_loss=0.05116, over 956747.93 frames. ], batch size: 37, lr: 3.25e-03, grad_scale: 32.0 2023-04-27 13:56:24,711 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=110639.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:56:24,829 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-04-27 13:56:38,217 INFO [optim.py:369] (6/7) 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 13:56:38,793 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=110659.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:56:40,068 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.4429, 1.5490, 1.8081, 1.8687, 1.7246, 1.8536, 1.9179, 1.9032], device='cuda:6'), covar=tensor([0.3606, 0.5455, 0.4777, 0.4546, 0.5547, 0.7383, 0.4856, 0.4715], device='cuda:6'), in_proj_covar=tensor([0.0337, 0.0373, 0.0323, 0.0335, 0.0344, 0.0394, 0.0357, 0.0328], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 13:56:47,728 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=110672.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:56:48,307 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=110673.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:56:48,913 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=110674.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:56:50,676 INFO [finetune.py:976] (6/7) Epoch 20, batch 1850, loss[loss=0.1547, simple_loss=0.2294, pruned_loss=0.03996, over 4808.00 frames. ], tot_loss[loss=0.1736, simple_loss=0.2457, pruned_loss=0.05078, over 955459.76 frames. ], batch size: 40, lr: 3.25e-03, grad_scale: 32.0 2023-04-27 13:57:06,441 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=110700.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:57:21,094 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 20, batch 1900, loss[loss=0.1397, simple_loss=0.2266, pruned_loss=0.02645, over 4850.00 frames. ], tot_loss[loss=0.1744, simple_loss=0.2473, pruned_loss=0.05079, over 953874.96 frames. ], batch size: 49, lr: 3.25e-03, grad_scale: 32.0 2023-04-27 13:57:24,845 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.3166, 1.5456, 1.5440, 1.7117, 1.6312, 1.8843, 1.4285, 3.4681], device='cuda:6'), covar=tensor([0.0602, 0.0819, 0.0743, 0.1170, 0.0619, 0.0529, 0.0749, 0.0135], device='cuda:6'), in_proj_covar=tensor([0.0038, 0.0039, 0.0040, 0.0043, 0.0040, 0.0038, 0.0038, 0.0056], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0013, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:6') 2023-04-27 13:57:30,856 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=110736.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:57:35,059 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0914, 1.3426, 4.6740, 4.3694, 4.0881, 4.4438, 4.2029, 4.1028], device='cuda:6'), covar=tensor([0.6256, 0.6015, 0.1011, 0.1741, 0.1017, 0.1382, 0.2247, 0.1495], device='cuda:6'), in_proj_covar=tensor([0.0308, 0.0302, 0.0403, 0.0405, 0.0348, 0.0407, 0.0312, 0.0364], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-27 13:57:38,119 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=110748.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:57:46,355 INFO [optim.py:369] (6/7) 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 13:57:51,447 INFO [zipformer.py:1188] (6/7) 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] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=110771.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:57:59,104 INFO [finetune.py:976] (6/7) Epoch 20, batch 1950, loss[loss=0.133, simple_loss=0.2035, pruned_loss=0.03124, over 4812.00 frames. ], tot_loss[loss=0.1727, simple_loss=0.2455, pruned_loss=0.0499, over 954470.33 frames. ], batch size: 25, lr: 3.25e-03, grad_scale: 32.0 2023-04-27 13:58:08,940 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6684, 1.2396, 1.8042, 2.2200, 1.8000, 1.6633, 1.7371, 1.7211], device='cuda:6'), covar=tensor([0.4393, 0.6283, 0.5779, 0.5476, 0.5374, 0.7332, 0.7360, 0.8047], device='cuda:6'), in_proj_covar=tensor([0.0430, 0.0411, 0.0505, 0.0508, 0.0457, 0.0485, 0.0492, 0.0497], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 13:58:29,616 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=110797.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 13:58:52,669 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=110814.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:59:01,950 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=110819.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:59:11,700 INFO [finetune.py:976] (6/7) Epoch 20, batch 2000, loss[loss=0.1574, simple_loss=0.2255, pruned_loss=0.04468, over 4903.00 frames. ], tot_loss[loss=0.1713, simple_loss=0.2437, pruned_loss=0.04949, over 955531.79 frames. ], batch size: 36, lr: 3.25e-03, grad_scale: 32.0 2023-04-27 13:59:33,915 INFO [optim.py:369] (6/7) 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 13:59:45,780 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=110868.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:59:56,924 INFO [finetune.py:976] (6/7) Epoch 20, batch 2050, loss[loss=0.1619, simple_loss=0.2366, pruned_loss=0.04359, over 4853.00 frames. ], tot_loss[loss=0.1705, simple_loss=0.2424, pruned_loss=0.04931, over 954366.64 frames. ], batch size: 47, lr: 3.24e-03, grad_scale: 32.0 2023-04-27 14:00:10,705 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.4730, 1.2329, 4.0943, 3.8300, 3.6146, 3.9421, 3.9329, 3.6076], device='cuda:6'), covar=tensor([0.6850, 0.6175, 0.1096, 0.1635, 0.1142, 0.1910, 0.1298, 0.1542], device='cuda:6'), in_proj_covar=tensor([0.0307, 0.0302, 0.0403, 0.0405, 0.0347, 0.0407, 0.0311, 0.0363], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-27 14:00:53,302 INFO [finetune.py:976] (6/7) Epoch 20, batch 2100, loss[loss=0.1698, simple_loss=0.2374, pruned_loss=0.05106, over 4898.00 frames. ], tot_loss[loss=0.1708, simple_loss=0.2424, pruned_loss=0.04966, over 955055.58 frames. ], batch size: 32, lr: 3.24e-03, grad_scale: 32.0 2023-04-27 14:00:55,187 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=110929.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:01:01,678 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=110939.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:01:13,874 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=110958.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:01:14,360 INFO [optim.py:369] (6/7) 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 14:01:14,952 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=110959.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:01:20,706 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=110967.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:01:25,024 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=110974.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:01:27,221 INFO [finetune.py:976] (6/7) Epoch 20, batch 2150, loss[loss=0.1607, simple_loss=0.2144, pruned_loss=0.0535, over 4182.00 frames. ], tot_loss[loss=0.1734, simple_loss=0.2453, pruned_loss=0.05075, over 955747.30 frames. ], batch size: 18, lr: 3.24e-03, grad_scale: 32.0 2023-04-27 14:01:34,419 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=110987.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:01:42,177 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8237, 2.3204, 1.9117, 1.6548, 1.3626, 1.3840, 1.9317, 1.2960], device='cuda:6'), covar=tensor([0.1620, 0.1249, 0.1397, 0.1701, 0.2280, 0.1829, 0.0995, 0.1930], device='cuda:6'), in_proj_covar=tensor([0.0195, 0.0210, 0.0168, 0.0202, 0.0198, 0.0183, 0.0154, 0.0184], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:6') 2023-04-27 14:01:47,107 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=111007.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:01:55,844 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=111019.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:01:57,642 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=111022.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:02:00,617 INFO [finetune.py:976] (6/7) Epoch 20, batch 2200, loss[loss=0.1718, simple_loss=0.2387, pruned_loss=0.05244, over 4888.00 frames. ], tot_loss[loss=0.1758, simple_loss=0.2479, pruned_loss=0.05182, over 957109.15 frames. ], batch size: 35, lr: 3.24e-03, grad_scale: 32.0 2023-04-27 14:02:04,617 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.68 vs. limit=5.0 2023-04-27 14:02:22,144 INFO [optim.py:369] (6/7) 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 14:02:33,550 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.3095, 1.6897, 2.1386, 2.6637, 2.1095, 1.7080, 1.4486, 1.9201], device='cuda:6'), covar=tensor([0.3336, 0.3288, 0.1617, 0.2230, 0.2601, 0.2632, 0.4219, 0.2083], device='cuda:6'), in_proj_covar=tensor([0.0294, 0.0246, 0.0228, 0.0317, 0.0219, 0.0234, 0.0229, 0.0186], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:6') 2023-04-27 14:02:34,619 INFO [finetune.py:976] (6/7) Epoch 20, batch 2250, loss[loss=0.1594, simple_loss=0.2387, pruned_loss=0.04002, over 4904.00 frames. ], tot_loss[loss=0.1769, simple_loss=0.2489, pruned_loss=0.05241, over 957679.57 frames. ], batch size: 46, lr: 3.24e-03, grad_scale: 32.0 2023-04-27 14:02:44,876 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=111092.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 14:02:57,103 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.4019, 1.4309, 1.8006, 1.8043, 1.3663, 1.2361, 1.4555, 0.9678], device='cuda:6'), covar=tensor([0.0603, 0.0601, 0.0348, 0.0632, 0.0759, 0.1126, 0.0607, 0.0641], device='cuda:6'), in_proj_covar=tensor([0.0069, 0.0069, 0.0067, 0.0068, 0.0075, 0.0097, 0.0074, 0.0065], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-27 14:03:02,596 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=3.97 vs. limit=5.0 2023-04-27 14:03:08,137 INFO [finetune.py:976] (6/7) Epoch 20, batch 2300, loss[loss=0.1614, simple_loss=0.2471, pruned_loss=0.03791, over 4857.00 frames. ], tot_loss[loss=0.176, simple_loss=0.2485, pruned_loss=0.05173, over 955241.31 frames. ], batch size: 31, lr: 3.24e-03, grad_scale: 32.0 2023-04-27 14:03:29,173 INFO [optim.py:369] (6/7) 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 14:03:41,037 INFO [finetune.py:976] (6/7) Epoch 20, batch 2350, loss[loss=0.1517, simple_loss=0.2339, pruned_loss=0.03475, over 4754.00 frames. ], tot_loss[loss=0.1738, simple_loss=0.2453, pruned_loss=0.05116, over 953722.93 frames. ], batch size: 26, lr: 3.24e-03, grad_scale: 32.0 2023-04-27 14:04:27,660 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=111224.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:04:28,339 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 20, batch 2400, loss[loss=0.1788, simple_loss=0.2488, pruned_loss=0.05444, over 4834.00 frames. ], tot_loss[loss=0.1729, simple_loss=0.2437, pruned_loss=0.05108, over 954746.28 frames. ], batch size: 39, lr: 3.24e-03, grad_scale: 32.0 2023-04-27 14:05:02,139 INFO [optim.py:369] (6/7) 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 14:05:07,603 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=111267.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:05:19,548 INFO [finetune.py:976] (6/7) Epoch 20, batch 2450, loss[loss=0.1975, simple_loss=0.2718, pruned_loss=0.06159, over 4916.00 frames. ], tot_loss[loss=0.1718, simple_loss=0.2422, pruned_loss=0.05072, over 954620.24 frames. ], batch size: 36, lr: 3.24e-03, grad_scale: 32.0 2023-04-27 14:05:36,578 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=111286.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:06:05,502 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=111314.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:06:06,092 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=111315.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:06:13,836 INFO [finetune.py:976] (6/7) Epoch 20, batch 2500, loss[loss=0.1765, simple_loss=0.248, pruned_loss=0.05248, over 4830.00 frames. ], tot_loss[loss=0.1721, simple_loss=0.2425, pruned_loss=0.05088, over 955555.53 frames. ], batch size: 30, lr: 3.24e-03, grad_scale: 32.0 2023-04-27 14:06:22,332 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.1342, 2.6917, 2.2842, 2.4327, 1.8488, 2.3549, 2.3969, 1.8916], device='cuda:6'), covar=tensor([0.2233, 0.1375, 0.0783, 0.1338, 0.3322, 0.1237, 0.2244, 0.2782], device='cuda:6'), in_proj_covar=tensor([0.0288, 0.0302, 0.0218, 0.0279, 0.0313, 0.0259, 0.0250, 0.0263], device='cuda:6'), out_proj_covar=tensor([1.1525e-04, 1.1985e-04, 8.6186e-05, 1.1053e-04, 1.2685e-04, 1.0273e-04, 1.0099e-04, 1.0417e-04], device='cuda:6') 2023-04-27 14:06:35,391 INFO [optim.py:369] (6/7) 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 14:06:46,744 INFO [finetune.py:976] (6/7) Epoch 20, batch 2550, loss[loss=0.2016, simple_loss=0.2837, pruned_loss=0.05969, over 4917.00 frames. ], tot_loss[loss=0.1746, simple_loss=0.2464, pruned_loss=0.05145, over 956104.14 frames. ], batch size: 42, lr: 3.24e-03, grad_scale: 32.0 2023-04-27 14:06:57,256 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=111392.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 14:07:12,886 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0314, 1.4294, 5.0913, 4.8039, 4.3788, 4.8518, 4.5394, 4.4892], device='cuda:6'), covar=tensor([0.6653, 0.5975, 0.0880, 0.1523, 0.1208, 0.1289, 0.1435, 0.1442], device='cuda:6'), in_proj_covar=tensor([0.0311, 0.0306, 0.0409, 0.0409, 0.0354, 0.0410, 0.0315, 0.0368], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-27 14:07:20,615 INFO [finetune.py:976] (6/7) Epoch 20, batch 2600, loss[loss=0.1527, simple_loss=0.2306, pruned_loss=0.03744, over 4749.00 frames. ], tot_loss[loss=0.1744, simple_loss=0.2466, pruned_loss=0.05114, over 956265.49 frames. ], batch size: 27, lr: 3.24e-03, grad_scale: 32.0 2023-04-27 14:07:21,028 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.63 vs. limit=5.0 2023-04-27 14:07:22,130 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=3.53 vs. limit=5.0 2023-04-27 14:07:29,613 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=111440.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:07:42,985 INFO [optim.py:369] (6/7) 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 14:07:48,669 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 20, batch 2650, loss[loss=0.1531, simple_loss=0.2316, pruned_loss=0.03735, over 4802.00 frames. ], tot_loss[loss=0.1744, simple_loss=0.2473, pruned_loss=0.05073, over 958429.67 frames. ], batch size: 51, lr: 3.24e-03, grad_scale: 32.0 2023-04-27 14:08:26,346 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 20, batch 2700, loss[loss=0.1655, simple_loss=0.2281, pruned_loss=0.05145, over 4885.00 frames. ], tot_loss[loss=0.1735, simple_loss=0.2463, pruned_loss=0.05029, over 956956.86 frames. ], batch size: 32, lr: 3.24e-03, grad_scale: 32.0 2023-04-27 14:08:29,407 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=111529.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:08:40,412 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.75 vs. limit=5.0 2023-04-27 14:08:49,485 INFO [optim.py:369] (6/7) 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 14:08:58,400 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=111572.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:09:01,904 INFO [finetune.py:976] (6/7) Epoch 20, batch 2750, loss[loss=0.1891, simple_loss=0.26, pruned_loss=0.0591, over 4926.00 frames. ], tot_loss[loss=0.1723, simple_loss=0.2444, pruned_loss=0.05011, over 956875.59 frames. ], batch size: 33, lr: 3.24e-03, grad_scale: 32.0 2023-04-27 14:09:04,324 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=111581.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:09:06,751 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7039, 1.5582, 1.7111, 2.0748, 2.1739, 1.6964, 1.3441, 1.9471], device='cuda:6'), covar=tensor([0.0752, 0.1196, 0.0811, 0.0512, 0.0512, 0.0808, 0.0784, 0.0501], device='cuda:6'), in_proj_covar=tensor([0.0187, 0.0202, 0.0183, 0.0172, 0.0178, 0.0181, 0.0154, 0.0180], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 14:09:38,139 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=111614.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:09:57,463 INFO [finetune.py:976] (6/7) Epoch 20, batch 2800, loss[loss=0.1192, simple_loss=0.195, pruned_loss=0.02172, over 4755.00 frames. ], tot_loss[loss=0.1696, simple_loss=0.2413, pruned_loss=0.04898, over 959298.74 frames. ], batch size: 26, lr: 3.24e-03, grad_scale: 32.0 2023-04-27 14:10:14,521 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7382, 1.2984, 1.8515, 2.2558, 1.8284, 1.7630, 1.8179, 1.7485], device='cuda:6'), covar=tensor([0.4267, 0.6118, 0.5582, 0.5235, 0.5397, 0.7094, 0.7262, 0.8842], device='cuda:6'), in_proj_covar=tensor([0.0431, 0.0413, 0.0508, 0.0511, 0.0460, 0.0489, 0.0494, 0.0501], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 14:10:24,458 INFO [optim.py:369] (6/7) 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 14:10:26,856 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=111662.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:10:36,929 INFO [finetune.py:976] (6/7) Epoch 20, batch 2850, loss[loss=0.1999, simple_loss=0.2653, pruned_loss=0.06727, over 4869.00 frames. ], tot_loss[loss=0.1703, simple_loss=0.2411, pruned_loss=0.04977, over 957152.74 frames. ], batch size: 34, lr: 3.24e-03, grad_scale: 32.0 2023-04-27 14:10:42,924 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.2776, 2.4831, 1.1341, 1.5202, 2.0615, 1.4730, 3.4193, 1.8913], device='cuda:6'), covar=tensor([0.0597, 0.0663, 0.0725, 0.1190, 0.0474, 0.0953, 0.0338, 0.0614], device='cuda:6'), in_proj_covar=tensor([0.0051, 0.0065, 0.0048, 0.0046, 0.0049, 0.0052, 0.0074, 0.0051], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:6') 2023-04-27 14:11:31,379 INFO [finetune.py:976] (6/7) Epoch 20, batch 2900, loss[loss=0.1634, simple_loss=0.2468, pruned_loss=0.04, over 4880.00 frames. ], tot_loss[loss=0.1727, simple_loss=0.244, pruned_loss=0.05073, over 955878.48 frames. ], batch size: 34, lr: 3.24e-03, grad_scale: 32.0 2023-04-27 14:11:48,006 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.0346, 1.0713, 1.1346, 1.1582, 0.9477, 0.9134, 1.0096, 0.6361], device='cuda:6'), covar=tensor([0.0527, 0.0621, 0.0468, 0.0564, 0.0737, 0.1224, 0.0499, 0.0654], device='cuda:6'), in_proj_covar=tensor([0.0069, 0.0069, 0.0068, 0.0068, 0.0076, 0.0097, 0.0074, 0.0066], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-27 14:12:12,493 INFO [optim.py:369] (6/7) 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 14:12:13,664 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.70 vs. limit=5.0 2023-04-27 14:12:36,595 INFO [finetune.py:976] (6/7) Epoch 20, batch 2950, loss[loss=0.1775, simple_loss=0.2483, pruned_loss=0.05336, over 4821.00 frames. ], tot_loss[loss=0.1729, simple_loss=0.2445, pruned_loss=0.0507, over 951739.83 frames. ], batch size: 40, lr: 3.24e-03, grad_scale: 64.0 2023-04-27 14:12:56,799 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=111791.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 14:13:17,289 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-04-27 14:13:29,946 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=111824.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:13:32,110 INFO [finetune.py:976] (6/7) Epoch 20, batch 3000, loss[loss=0.1809, simple_loss=0.2441, pruned_loss=0.05892, over 4763.00 frames. ], tot_loss[loss=0.1753, simple_loss=0.2469, pruned_loss=0.05182, over 950644.49 frames. ], batch size: 28, lr: 3.24e-03, grad_scale: 64.0 2023-04-27 14:13:32,111 INFO [finetune.py:1001] (6/7) Computing validation loss 2023-04-27 14:13:34,005 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.4148, 1.2933, 1.6339, 1.6615, 1.2963, 1.2245, 1.3110, 0.9108], device='cuda:6'), covar=tensor([0.0525, 0.0639, 0.0389, 0.0523, 0.0800, 0.1226, 0.0542, 0.0539], device='cuda:6'), in_proj_covar=tensor([0.0069, 0.0069, 0.0067, 0.0067, 0.0075, 0.0096, 0.0073, 0.0065], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-27 14:13:35,200 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.3712, 1.3454, 3.8916, 3.5666, 3.5211, 3.7936, 3.8355, 3.4328], device='cuda:6'), covar=tensor([0.7448, 0.5389, 0.1382, 0.2318, 0.1420, 0.1375, 0.0905, 0.2010], device='cuda:6'), in_proj_covar=tensor([0.0312, 0.0309, 0.0409, 0.0409, 0.0354, 0.0411, 0.0317, 0.0368], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-27 14:13:35,978 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.3575, 1.3199, 1.5979, 1.6090, 1.2690, 1.2211, 1.3202, 0.8765], device='cuda:6'), covar=tensor([0.0566, 0.0574, 0.0405, 0.0448, 0.0746, 0.1131, 0.0581, 0.0557], device='cuda:6'), in_proj_covar=tensor([0.0069, 0.0069, 0.0067, 0.0067, 0.0075, 0.0096, 0.0073, 0.0065], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-27 14:13:49,406 INFO [finetune.py:1010] (6/7) Epoch 20, validation: loss=0.1527, simple_loss=0.2229, pruned_loss=0.04123, over 2265189.00 frames. 2023-04-27 14:13:49,407 INFO [finetune.py:1011] (6/7) Maximum memory allocated so far is 6365MB 2023-04-27 14:14:05,966 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5733, 1.6471, 1.6362, 2.0476, 1.9544, 2.0225, 1.7193, 4.3596], device='cuda:6'), covar=tensor([0.0498, 0.0793, 0.0743, 0.1117, 0.0565, 0.0537, 0.0696, 0.0097], device='cuda:6'), in_proj_covar=tensor([0.0038, 0.0039, 0.0040, 0.0043, 0.0040, 0.0038, 0.0038, 0.0056], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0013, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:6') 2023-04-27 14:14:25,667 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=111852.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 14:14:26,933 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7274, 2.9518, 2.3925, 2.6177, 3.0494, 2.5314, 3.8829, 2.3571], device='cuda:6'), covar=tensor([0.3738, 0.2130, 0.4156, 0.3170, 0.1695, 0.2419, 0.1207, 0.3421], device='cuda:6'), in_proj_covar=tensor([0.0336, 0.0348, 0.0423, 0.0353, 0.0380, 0.0373, 0.0372, 0.0415], device='cuda:6'), out_proj_covar=tensor([9.9896e-05, 1.0422e-04, 1.2856e-04, 1.0653e-04, 1.1339e-04, 1.1155e-04, 1.0952e-04, 1.2556e-04], device='cuda:6') 2023-04-27 14:14:29,830 INFO [optim.py:369] (6/7) 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 14:14:38,267 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.1983, 1.4591, 1.3774, 1.6626, 1.5442, 1.7204, 1.3691, 3.2049], device='cuda:6'), covar=tensor([0.0628, 0.0783, 0.0788, 0.1189, 0.0611, 0.0606, 0.0753, 0.0169], device='cuda:6'), in_proj_covar=tensor([0.0038, 0.0039, 0.0040, 0.0043, 0.0040, 0.0038, 0.0038, 0.0056], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0013, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:6') 2023-04-27 14:14:58,617 INFO [finetune.py:976] (6/7) Epoch 20, batch 3050, loss[loss=0.1793, simple_loss=0.2548, pruned_loss=0.05192, over 4765.00 frames. ], tot_loss[loss=0.176, simple_loss=0.2479, pruned_loss=0.05204, over 950418.12 frames. ], batch size: 28, lr: 3.24e-03, grad_scale: 64.0 2023-04-27 14:15:01,121 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=111881.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:15:15,716 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7150, 1.1822, 1.8437, 2.2044, 1.8437, 1.6848, 1.7359, 1.7168], device='cuda:6'), covar=tensor([0.4569, 0.6807, 0.6395, 0.5833, 0.5672, 0.8173, 0.7799, 0.8677], device='cuda:6'), in_proj_covar=tensor([0.0429, 0.0411, 0.0506, 0.0509, 0.0457, 0.0486, 0.0493, 0.0497], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 14:15:37,733 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7157, 1.2968, 4.2859, 4.0249, 3.7498, 4.0248, 3.9211, 3.8046], device='cuda:6'), covar=tensor([0.7093, 0.5948, 0.1075, 0.1609, 0.1162, 0.1885, 0.1962, 0.1422], device='cuda:6'), in_proj_covar=tensor([0.0312, 0.0309, 0.0409, 0.0409, 0.0354, 0.0411, 0.0317, 0.0368], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-27 14:15:39,355 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-04-27 14:15:48,558 INFO [finetune.py:976] (6/7) Epoch 20, batch 3100, loss[loss=0.1858, simple_loss=0.2504, pruned_loss=0.06061, over 4812.00 frames. ], tot_loss[loss=0.1744, simple_loss=0.2462, pruned_loss=0.05127, over 952717.38 frames. ], batch size: 41, lr: 3.24e-03, grad_scale: 32.0 2023-04-27 14:15:50,329 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=111929.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:15:57,727 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5948, 1.4299, 1.2634, 1.5763, 1.8687, 1.6241, 1.4462, 1.1780], device='cuda:6'), covar=tensor([0.1350, 0.1035, 0.1367, 0.1257, 0.0651, 0.1163, 0.1359, 0.1707], device='cuda:6'), in_proj_covar=tensor([0.0308, 0.0309, 0.0349, 0.0288, 0.0325, 0.0306, 0.0299, 0.0369], device='cuda:6'), out_proj_covar=tensor([6.3505e-05, 6.4028e-05, 7.3963e-05, 5.8112e-05, 6.7279e-05, 6.4163e-05, 6.2741e-05, 7.8451e-05], device='cuda:6') 2023-04-27 14:16:12,368 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2023-04-27 14:16:16,416 INFO [optim.py:369] (6/7) 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 14:16:27,780 INFO [finetune.py:976] (6/7) Epoch 20, batch 3150, loss[loss=0.1336, simple_loss=0.1843, pruned_loss=0.04141, over 4033.00 frames. ], tot_loss[loss=0.1723, simple_loss=0.2435, pruned_loss=0.05058, over 953019.50 frames. ], batch size: 17, lr: 3.24e-03, grad_scale: 32.0 2023-04-27 14:17:02,002 INFO [finetune.py:976] (6/7) Epoch 20, batch 3200, loss[loss=0.18, simple_loss=0.2482, pruned_loss=0.05586, over 4845.00 frames. ], tot_loss[loss=0.1706, simple_loss=0.241, pruned_loss=0.05011, over 953790.33 frames. ], batch size: 49, lr: 3.24e-03, grad_scale: 32.0 2023-04-27 14:17:13,005 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.4913, 1.6980, 1.4100, 1.1027, 1.1659, 1.1557, 1.4467, 1.0908], device='cuda:6'), covar=tensor([0.1869, 0.1368, 0.1687, 0.1884, 0.2515, 0.2212, 0.1116, 0.2221], device='cuda:6'), in_proj_covar=tensor([0.0197, 0.0211, 0.0168, 0.0203, 0.0200, 0.0185, 0.0155, 0.0184], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:6') 2023-04-27 14:17:17,092 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0183, 2.5418, 1.9649, 1.9101, 1.5043, 1.5275, 2.1627, 1.4334], device='cuda:6'), covar=tensor([0.1635, 0.1290, 0.1478, 0.1658, 0.2307, 0.1989, 0.0966, 0.2026], device='cuda:6'), in_proj_covar=tensor([0.0197, 0.0211, 0.0168, 0.0203, 0.0200, 0.0185, 0.0155, 0.0185], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:6') 2023-04-27 14:17:25,359 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.5635, 1.7737, 2.0130, 2.0745, 1.9458, 1.9761, 2.0345, 2.0291], device='cuda:6'), covar=tensor([0.4155, 0.5590, 0.4499, 0.4358, 0.5663, 0.7129, 0.5530, 0.5063], device='cuda:6'), in_proj_covar=tensor([0.0336, 0.0373, 0.0321, 0.0335, 0.0345, 0.0393, 0.0357, 0.0327], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 14:17:35,522 INFO [optim.py:369] (6/7) 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 14:17:39,742 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 20, batch 3250, loss[loss=0.2089, simple_loss=0.2797, pruned_loss=0.06905, over 4904.00 frames. ], tot_loss[loss=0.1703, simple_loss=0.241, pruned_loss=0.04982, over 955249.87 frames. ], batch size: 36, lr: 3.23e-03, grad_scale: 32.0 2023-04-27 14:18:01,719 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.0344, 1.0561, 1.2365, 1.2253, 1.0367, 0.9175, 0.9233, 0.4150], device='cuda:6'), covar=tensor([0.0497, 0.0424, 0.0409, 0.0502, 0.0615, 0.1315, 0.0457, 0.0737], device='cuda:6'), in_proj_covar=tensor([0.0070, 0.0069, 0.0068, 0.0068, 0.0076, 0.0097, 0.0074, 0.0066], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-27 14:18:18,440 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=112124.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:18:19,090 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.8143, 2.0352, 2.0688, 2.1796, 1.9605, 2.0412, 2.1367, 2.1148], device='cuda:6'), covar=tensor([0.4404, 0.6285, 0.4772, 0.4638, 0.6084, 0.7294, 0.6253, 0.5437], device='cuda:6'), in_proj_covar=tensor([0.0338, 0.0374, 0.0323, 0.0336, 0.0346, 0.0394, 0.0357, 0.0329], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 14:18:20,138 INFO [finetune.py:976] (6/7) Epoch 20, batch 3300, loss[loss=0.2421, simple_loss=0.3078, pruned_loss=0.08821, over 4728.00 frames. ], tot_loss[loss=0.1728, simple_loss=0.2444, pruned_loss=0.05056, over 954090.09 frames. ], batch size: 59, lr: 3.23e-03, grad_scale: 32.0 2023-04-27 14:18:20,258 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=112127.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:18:44,826 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=112147.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 14:18:46,163 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.66 vs. limit=2.0 2023-04-27 14:18:52,674 INFO [optim.py:369] (6/7) 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 14:18:59,400 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7677, 2.4266, 1.9153, 1.8539, 1.4938, 1.4706, 2.0278, 1.4535], device='cuda:6'), covar=tensor([0.1365, 0.1134, 0.1288, 0.1517, 0.2009, 0.1788, 0.0802, 0.1769], device='cuda:6'), in_proj_covar=tensor([0.0197, 0.0211, 0.0168, 0.0204, 0.0200, 0.0185, 0.0155, 0.0185], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:6') 2023-04-27 14:19:00,557 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=112172.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:19:04,159 INFO [finetune.py:976] (6/7) Epoch 20, batch 3350, loss[loss=0.1694, simple_loss=0.2412, pruned_loss=0.04885, over 4760.00 frames. ], tot_loss[loss=0.1733, simple_loss=0.2454, pruned_loss=0.05059, over 953828.95 frames. ], batch size: 54, lr: 3.23e-03, grad_scale: 32.0 2023-04-27 14:19:37,511 INFO [finetune.py:976] (6/7) Epoch 20, batch 3400, loss[loss=0.2035, simple_loss=0.2674, pruned_loss=0.0698, over 4883.00 frames. ], tot_loss[loss=0.1754, simple_loss=0.2477, pruned_loss=0.05155, over 954336.90 frames. ], batch size: 35, lr: 3.23e-03, grad_scale: 32.0 2023-04-27 14:19:37,688 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-04-27 14:20:15,437 INFO [optim.py:369] (6/7) 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 14:20:37,464 INFO [finetune.py:976] (6/7) Epoch 20, batch 3450, loss[loss=0.1827, simple_loss=0.2445, pruned_loss=0.06039, over 4122.00 frames. ], tot_loss[loss=0.1754, simple_loss=0.2479, pruned_loss=0.05143, over 954888.28 frames. ], batch size: 18, lr: 3.23e-03, grad_scale: 32.0 2023-04-27 14:20:48,939 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.5370, 1.8284, 2.0172, 2.0959, 1.9078, 1.9570, 2.0748, 2.0556], device='cuda:6'), covar=tensor([0.4095, 0.5966, 0.4675, 0.4806, 0.5827, 0.7506, 0.5617, 0.5058], device='cuda:6'), in_proj_covar=tensor([0.0336, 0.0372, 0.0321, 0.0334, 0.0344, 0.0392, 0.0355, 0.0327], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 14:20:50,701 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8861, 2.3938, 1.8664, 1.6231, 1.4197, 1.4649, 1.8435, 1.3315], device='cuda:6'), covar=tensor([0.1711, 0.1384, 0.1388, 0.1778, 0.2370, 0.2041, 0.1039, 0.2027], device='cuda:6'), in_proj_covar=tensor([0.0199, 0.0213, 0.0169, 0.0205, 0.0201, 0.0186, 0.0156, 0.0186], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:6') 2023-04-27 14:21:35,778 INFO [finetune.py:976] (6/7) Epoch 20, batch 3500, loss[loss=0.1484, simple_loss=0.2064, pruned_loss=0.04515, over 4058.00 frames. ], tot_loss[loss=0.1722, simple_loss=0.244, pruned_loss=0.05025, over 954046.58 frames. ], batch size: 17, lr: 3.23e-03, grad_scale: 32.0 2023-04-27 14:21:59,356 INFO [optim.py:369] (6/7) 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 14:22:09,769 INFO [finetune.py:976] (6/7) Epoch 20, batch 3550, loss[loss=0.1547, simple_loss=0.2204, pruned_loss=0.04449, over 4150.00 frames. ], tot_loss[loss=0.1707, simple_loss=0.2418, pruned_loss=0.04977, over 955236.75 frames. ], batch size: 65, lr: 3.23e-03, grad_scale: 32.0 2023-04-27 14:22:18,195 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=112389.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:22:24,835 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.0433, 1.0289, 1.1060, 1.1345, 0.9878, 0.8914, 0.9539, 0.6237], device='cuda:6'), covar=tensor([0.0513, 0.0540, 0.0453, 0.0538, 0.0693, 0.1260, 0.0480, 0.0599], device='cuda:6'), in_proj_covar=tensor([0.0070, 0.0069, 0.0068, 0.0069, 0.0076, 0.0098, 0.0074, 0.0066], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-27 14:22:40,999 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 20, batch 3600, loss[loss=0.1529, simple_loss=0.2155, pruned_loss=0.04511, over 4769.00 frames. ], tot_loss[loss=0.1685, simple_loss=0.239, pruned_loss=0.04896, over 954567.71 frames. ], batch size: 28, lr: 3.23e-03, grad_scale: 32.0 2023-04-27 14:22:57,508 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=112447.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 14:22:59,369 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=112450.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:23:06,297 INFO [optim.py:369] (6/7) 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] (6/7) Epoch 20, batch 3650, loss[loss=0.2185, simple_loss=0.2868, pruned_loss=0.0751, over 4837.00 frames. ], tot_loss[loss=0.1702, simple_loss=0.2411, pruned_loss=0.04963, over 954167.39 frames. ], batch size: 47, lr: 3.23e-03, grad_scale: 32.0 2023-04-27 14:23:30,563 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=112495.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 14:23:58,095 INFO [finetune.py:976] (6/7) Epoch 20, batch 3700, loss[loss=0.1702, simple_loss=0.2456, pruned_loss=0.04736, over 4904.00 frames. ], tot_loss[loss=0.173, simple_loss=0.2445, pruned_loss=0.05071, over 953415.81 frames. ], batch size: 37, lr: 3.23e-03, grad_scale: 32.0 2023-04-27 14:24:41,780 INFO [optim.py:369] (6/7) 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 14:24:48,659 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=112562.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:25:05,135 INFO [finetune.py:976] (6/7) Epoch 20, batch 3750, loss[loss=0.1676, simple_loss=0.2218, pruned_loss=0.05667, over 3982.00 frames. ], tot_loss[loss=0.1739, simple_loss=0.2462, pruned_loss=0.05081, over 953644.30 frames. ], batch size: 17, lr: 3.23e-03, grad_scale: 32.0 2023-04-27 14:25:10,890 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=112577.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:25:26,549 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8910, 2.1345, 0.9248, 1.5197, 2.3731, 1.6394, 1.6175, 1.7104], device='cuda:6'), covar=tensor([0.0484, 0.0355, 0.0300, 0.0543, 0.0210, 0.0523, 0.0492, 0.0559], device='cuda:6'), in_proj_covar=tensor([0.0028, 0.0024, 0.0022, 0.0029, 0.0019, 0.0028, 0.0028, 0.0028], device='cuda:6'), out_proj_covar=tensor([0.0051, 0.0045, 0.0038, 0.0052, 0.0038, 0.0049, 0.0050, 0.0051], device='cuda:6') 2023-04-27 14:26:00,858 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=112617.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:26:11,156 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=112623.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:26:19,269 INFO [finetune.py:976] (6/7) Epoch 20, batch 3800, loss[loss=0.188, simple_loss=0.2549, pruned_loss=0.06049, over 4882.00 frames. ], tot_loss[loss=0.1743, simple_loss=0.247, pruned_loss=0.05083, over 953178.72 frames. ], batch size: 32, lr: 3.23e-03, grad_scale: 32.0 2023-04-27 14:26:31,099 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=112638.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 14:26:42,518 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7683, 2.0620, 1.8299, 1.9997, 1.5402, 1.7357, 1.8838, 1.4790], device='cuda:6'), covar=tensor([0.1659, 0.1372, 0.0876, 0.1179, 0.3891, 0.1306, 0.1610, 0.2483], device='cuda:6'), in_proj_covar=tensor([0.0284, 0.0298, 0.0213, 0.0275, 0.0308, 0.0255, 0.0245, 0.0258], device='cuda:6'), out_proj_covar=tensor([1.1367e-04, 1.1812e-04, 8.4402e-05, 1.0878e-04, 1.2481e-04, 1.0125e-04, 9.9227e-05, 1.0235e-04], device='cuda:6') 2023-04-27 14:27:02,945 INFO [optim.py:369] (6/7) 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 14:27:06,927 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.78 vs. limit=2.0 2023-04-27 14:27:25,229 INFO [finetune.py:976] (6/7) Epoch 20, batch 3850, loss[loss=0.1565, simple_loss=0.2312, pruned_loss=0.04097, over 4905.00 frames. ], tot_loss[loss=0.1734, simple_loss=0.2457, pruned_loss=0.05053, over 954000.58 frames. ], batch size: 43, lr: 3.23e-03, grad_scale: 32.0 2023-04-27 14:27:25,943 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=112678.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 14:28:09,554 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=112722.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:28:13,015 INFO [finetune.py:976] (6/7) Epoch 20, batch 3900, loss[loss=0.1655, simple_loss=0.2323, pruned_loss=0.04935, over 4906.00 frames. ], tot_loss[loss=0.1718, simple_loss=0.243, pruned_loss=0.05026, over 953992.30 frames. ], batch size: 36, lr: 3.23e-03, grad_scale: 32.0 2023-04-27 14:28:25,003 INFO [zipformer.py:1188] (6/7) 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] (6/7) 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 14:28:39,454 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.1724, 1.4067, 1.2774, 1.6641, 1.5797, 1.5520, 1.3351, 2.4299], device='cuda:6'), covar=tensor([0.0611, 0.0868, 0.0873, 0.1288, 0.0659, 0.0476, 0.0788, 0.0223], device='cuda:6'), in_proj_covar=tensor([0.0037, 0.0039, 0.0040, 0.0043, 0.0040, 0.0038, 0.0038, 0.0056], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:6') 2023-04-27 14:28:41,701 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=112770.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:28:45,927 INFO [finetune.py:976] (6/7) Epoch 20, batch 3950, loss[loss=0.181, simple_loss=0.2442, pruned_loss=0.05889, over 4829.00 frames. ], tot_loss[loss=0.1704, simple_loss=0.241, pruned_loss=0.04986, over 954613.64 frames. ], batch size: 30, lr: 3.23e-03, grad_scale: 32.0 2023-04-27 14:28:50,661 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.50 vs. limit=5.0 2023-04-27 14:28:53,827 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-04-27 14:29:15,580 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.43 vs. limit=5.0 2023-04-27 14:29:20,144 INFO [finetune.py:976] (6/7) Epoch 20, batch 4000, loss[loss=0.143, simple_loss=0.2172, pruned_loss=0.03444, over 4771.00 frames. ], tot_loss[loss=0.1713, simple_loss=0.2416, pruned_loss=0.05049, over 952076.46 frames. ], batch size: 28, lr: 3.23e-03, grad_scale: 32.0 2023-04-27 14:29:21,487 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.4955, 1.8695, 2.2726, 2.8949, 2.2767, 1.8717, 1.8554, 2.1492], device='cuda:6'), covar=tensor([0.3101, 0.3298, 0.1727, 0.2352, 0.2907, 0.2627, 0.3739, 0.2294], device='cuda:6'), in_proj_covar=tensor([0.0291, 0.0246, 0.0227, 0.0313, 0.0219, 0.0233, 0.0228, 0.0185], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:6') 2023-04-27 14:29:40,396 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.1386, 1.8888, 2.2904, 2.5578, 2.1076, 1.9981, 2.1112, 2.1373], device='cuda:6'), covar=tensor([0.4854, 0.7632, 0.7657, 0.6209, 0.6486, 0.8575, 0.9598, 0.9739], device='cuda:6'), in_proj_covar=tensor([0.0431, 0.0411, 0.0506, 0.0508, 0.0457, 0.0486, 0.0494, 0.0498], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 14:29:42,039 INFO [optim.py:369] (6/7) 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 14:29:46,122 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.4360, 1.2690, 0.6723, 1.1665, 1.3349, 1.3172, 1.2305, 1.2841], device='cuda:6'), covar=tensor([0.0494, 0.0381, 0.0373, 0.0562, 0.0305, 0.0503, 0.0477, 0.0556], device='cuda:6'), in_proj_covar=tensor([0.0028, 0.0024, 0.0022, 0.0029, 0.0019, 0.0028, 0.0028, 0.0028], device='cuda:6'), out_proj_covar=tensor([0.0051, 0.0045, 0.0038, 0.0052, 0.0038, 0.0050, 0.0050, 0.0051], device='cuda:6') 2023-04-27 14:29:53,759 INFO [finetune.py:976] (6/7) Epoch 20, batch 4050, loss[loss=0.1942, simple_loss=0.2684, pruned_loss=0.06002, over 4735.00 frames. ], tot_loss[loss=0.174, simple_loss=0.2442, pruned_loss=0.05186, over 950653.37 frames. ], batch size: 59, lr: 3.23e-03, grad_scale: 32.0 2023-04-27 14:30:17,054 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9714, 2.4674, 2.1935, 2.2367, 1.7402, 2.1585, 2.0665, 1.7596], device='cuda:6'), covar=tensor([0.1832, 0.1202, 0.0647, 0.1312, 0.3200, 0.1048, 0.1724, 0.2480], device='cuda:6'), in_proj_covar=tensor([0.0281, 0.0294, 0.0211, 0.0272, 0.0304, 0.0252, 0.0243, 0.0255], device='cuda:6'), out_proj_covar=tensor([1.1257e-04, 1.1670e-04, 8.3520e-05, 1.0780e-04, 1.2345e-04, 9.9883e-05, 9.8311e-05, 1.0114e-04], device='cuda:6') 2023-04-27 14:30:21,661 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=112918.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:30:23,572 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 20, batch 4100, loss[loss=0.1919, simple_loss=0.2633, pruned_loss=0.06024, over 4752.00 frames. ], tot_loss[loss=0.1747, simple_loss=0.2459, pruned_loss=0.05177, over 952110.08 frames. ], batch size: 28, lr: 3.23e-03, grad_scale: 32.0 2023-04-27 14:30:31,350 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=112933.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 14:31:11,349 INFO [optim.py:369] (6/7) 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 14:31:12,500 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.79 vs. limit=2.0 2023-04-27 14:31:22,766 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.3188, 1.5302, 1.4395, 1.5163, 1.3198, 1.2849, 1.3260, 1.0783], device='cuda:6'), covar=tensor([0.1710, 0.1294, 0.0843, 0.1187, 0.3587, 0.1253, 0.1721, 0.2297], device='cuda:6'), in_proj_covar=tensor([0.0282, 0.0294, 0.0211, 0.0273, 0.0305, 0.0252, 0.0244, 0.0255], device='cuda:6'), out_proj_covar=tensor([1.1266e-04, 1.1678e-04, 8.3480e-05, 1.0801e-04, 1.2393e-04, 9.9970e-05, 9.8465e-05, 1.0109e-04], device='cuda:6') 2023-04-27 14:31:25,634 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 20, batch 4150, loss[loss=0.1901, simple_loss=0.2646, pruned_loss=0.05782, over 4890.00 frames. ], tot_loss[loss=0.1758, simple_loss=0.2472, pruned_loss=0.05225, over 952723.74 frames. ], batch size: 35, lr: 3.23e-03, grad_scale: 32.0 2023-04-27 14:31:42,870 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=112982.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:32:06,223 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8171, 2.3347, 1.8400, 1.7270, 1.3331, 1.3567, 2.0069, 1.3023], device='cuda:6'), covar=tensor([0.1751, 0.1363, 0.1498, 0.1638, 0.2387, 0.1944, 0.0957, 0.2024], device='cuda:6'), in_proj_covar=tensor([0.0198, 0.0212, 0.0169, 0.0204, 0.0200, 0.0186, 0.0156, 0.0186], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:6') 2023-04-27 14:32:29,694 INFO [finetune.py:976] (6/7) Epoch 20, batch 4200, loss[loss=0.1724, simple_loss=0.2441, pruned_loss=0.05031, over 4900.00 frames. ], tot_loss[loss=0.1761, simple_loss=0.2478, pruned_loss=0.05221, over 953235.84 frames. ], batch size: 36, lr: 3.23e-03, grad_scale: 32.0 2023-04-27 14:32:42,222 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=113045.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:32:42,850 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.1683, 1.8541, 2.1229, 2.5903, 2.5808, 2.1049, 1.7509, 2.2692], device='cuda:6'), covar=tensor([0.0776, 0.1055, 0.0641, 0.0494, 0.0524, 0.0766, 0.0729, 0.0512], device='cuda:6'), in_proj_covar=tensor([0.0187, 0.0202, 0.0184, 0.0172, 0.0177, 0.0182, 0.0153, 0.0180], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 14:32:57,027 INFO [optim.py:369] (6/7) 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 14:33:18,841 INFO [finetune.py:976] (6/7) Epoch 20, batch 4250, loss[loss=0.1655, simple_loss=0.2294, pruned_loss=0.05079, over 4784.00 frames. ], tot_loss[loss=0.175, simple_loss=0.2463, pruned_loss=0.05188, over 954633.67 frames. ], batch size: 51, lr: 3.23e-03, grad_scale: 32.0 2023-04-27 14:33:40,238 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=113093.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:33:42,104 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9450, 1.0962, 1.6824, 1.8199, 1.7179, 1.8265, 1.6938, 1.7259], device='cuda:6'), covar=tensor([0.3898, 0.5046, 0.4070, 0.3878, 0.5074, 0.6837, 0.4423, 0.4201], device='cuda:6'), in_proj_covar=tensor([0.0337, 0.0374, 0.0322, 0.0336, 0.0346, 0.0395, 0.0357, 0.0328], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 14:34:04,771 INFO [finetune.py:976] (6/7) Epoch 20, batch 4300, loss[loss=0.1691, simple_loss=0.2351, pruned_loss=0.05149, over 4867.00 frames. ], tot_loss[loss=0.1716, simple_loss=0.2426, pruned_loss=0.05031, over 953383.89 frames. ], batch size: 34, lr: 3.23e-03, grad_scale: 32.0 2023-04-27 14:34:13,190 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.8580, 2.7513, 2.2245, 3.2712, 2.8574, 2.8045, 1.1803, 2.7448], device='cuda:6'), covar=tensor([0.2187, 0.1802, 0.3720, 0.3063, 0.3191, 0.2335, 0.5928, 0.2930], device='cuda:6'), in_proj_covar=tensor([0.0245, 0.0215, 0.0250, 0.0306, 0.0295, 0.0247, 0.0273, 0.0272], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 14:34:14,459 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.2123, 2.2488, 1.9463, 1.8967, 2.3610, 1.8359, 2.8282, 1.7922], device='cuda:6'), covar=tensor([0.3696, 0.1743, 0.4457, 0.3011, 0.1512, 0.2530, 0.1337, 0.3879], device='cuda:6'), in_proj_covar=tensor([0.0338, 0.0345, 0.0421, 0.0351, 0.0377, 0.0371, 0.0368, 0.0413], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 14:34:27,839 INFO [optim.py:369] (6/7) 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 14:34:30,059 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.70 vs. limit=2.0 2023-04-27 14:34:38,150 INFO [finetune.py:976] (6/7) Epoch 20, batch 4350, loss[loss=0.1596, simple_loss=0.2367, pruned_loss=0.04122, over 4863.00 frames. ], tot_loss[loss=0.1678, simple_loss=0.2384, pruned_loss=0.04861, over 952433.01 frames. ], batch size: 44, lr: 3.23e-03, grad_scale: 32.0 2023-04-27 14:35:06,247 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=113218.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:35:11,650 INFO [finetune.py:976] (6/7) Epoch 20, batch 4400, loss[loss=0.1404, simple_loss=0.2027, pruned_loss=0.03904, over 4039.00 frames. ], tot_loss[loss=0.168, simple_loss=0.2385, pruned_loss=0.04878, over 950548.55 frames. ], batch size: 17, lr: 3.23e-03, grad_scale: 32.0 2023-04-27 14:35:15,447 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=113233.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:35:16,048 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=113234.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:35:34,719 INFO [optim.py:369] (6/7) 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 14:35:38,477 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=113266.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:35:42,697 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=113273.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 14:35:43,940 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.2454, 1.2178, 1.2953, 1.4970, 1.4855, 1.2516, 0.8700, 1.4041], device='cuda:6'), covar=tensor([0.0780, 0.1195, 0.0762, 0.0521, 0.0609, 0.0727, 0.0892, 0.0562], device='cuda:6'), in_proj_covar=tensor([0.0187, 0.0202, 0.0184, 0.0172, 0.0178, 0.0182, 0.0153, 0.0180], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 14:35:45,064 INFO [finetune.py:976] (6/7) Epoch 20, batch 4450, loss[loss=0.2046, simple_loss=0.269, pruned_loss=0.0701, over 4929.00 frames. ], tot_loss[loss=0.1721, simple_loss=0.2432, pruned_loss=0.05051, over 947850.43 frames. ], batch size: 33, lr: 3.23e-03, grad_scale: 32.0 2023-04-27 14:35:45,130 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=113277.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:35:47,574 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=113281.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:35:57,006 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=113295.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:36:15,058 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 20, batch 4500, loss[loss=0.181, simple_loss=0.2498, pruned_loss=0.05616, over 4819.00 frames. ], tot_loss[loss=0.1748, simple_loss=0.2461, pruned_loss=0.05171, over 949822.58 frames. ], batch size: 39, lr: 3.22e-03, grad_scale: 32.0 2023-04-27 14:37:03,577 INFO [optim.py:369] (6/7) 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 14:37:25,451 INFO [finetune.py:976] (6/7) Epoch 20, batch 4550, loss[loss=0.1796, simple_loss=0.2457, pruned_loss=0.0567, over 4816.00 frames. ], tot_loss[loss=0.176, simple_loss=0.2478, pruned_loss=0.05211, over 951041.54 frames. ], batch size: 33, lr: 3.22e-03, grad_scale: 32.0 2023-04-27 14:37:53,069 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.4503, 3.4322, 1.2876, 1.7832, 1.9241, 2.5521, 1.8942, 1.0641], device='cuda:6'), covar=tensor([0.1451, 0.0997, 0.1671, 0.1274, 0.1151, 0.0917, 0.1572, 0.2065], device='cuda:6'), in_proj_covar=tensor([0.0117, 0.0241, 0.0137, 0.0120, 0.0134, 0.0152, 0.0117, 0.0119], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-27 14:37:54,855 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.3283, 1.5800, 1.6923, 1.7907, 1.6965, 1.8257, 1.7507, 1.7928], device='cuda:6'), covar=tensor([0.3871, 0.5014, 0.4100, 0.4238, 0.5513, 0.7289, 0.4727, 0.4585], device='cuda:6'), in_proj_covar=tensor([0.0337, 0.0373, 0.0320, 0.0335, 0.0345, 0.0394, 0.0356, 0.0327], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 14:38:02,715 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8323, 1.5593, 1.4234, 1.7158, 2.0922, 1.7138, 1.5250, 1.3458], device='cuda:6'), covar=tensor([0.1867, 0.1578, 0.2139, 0.1325, 0.1019, 0.1661, 0.2195, 0.2764], device='cuda:6'), in_proj_covar=tensor([0.0312, 0.0314, 0.0352, 0.0290, 0.0328, 0.0308, 0.0302, 0.0375], device='cuda:6'), out_proj_covar=tensor([6.4325e-05, 6.5051e-05, 7.4380e-05, 5.8552e-05, 6.7711e-05, 6.4590e-05, 6.3312e-05, 7.9683e-05], device='cuda:6') 2023-04-27 14:38:05,041 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([3.9392, 3.9360, 2.7571, 4.5170, 3.9655, 3.8939, 1.6066, 3.8844], device='cuda:6'), covar=tensor([0.1828, 0.1196, 0.3410, 0.1639, 0.3065, 0.1759, 0.6108, 0.2602], device='cuda:6'), in_proj_covar=tensor([0.0248, 0.0217, 0.0253, 0.0308, 0.0298, 0.0250, 0.0276, 0.0273], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 14:38:22,947 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6728, 1.1612, 1.8481, 2.1745, 1.7743, 1.6749, 1.7532, 1.7282], device='cuda:6'), covar=tensor([0.4455, 0.6406, 0.5466, 0.5486, 0.5439, 0.6861, 0.7033, 0.7881], device='cuda:6'), in_proj_covar=tensor([0.0429, 0.0410, 0.0503, 0.0504, 0.0457, 0.0484, 0.0493, 0.0498], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 14:38:24,622 INFO [finetune.py:976] (6/7) Epoch 20, batch 4600, loss[loss=0.1709, simple_loss=0.2488, pruned_loss=0.0465, over 4845.00 frames. ], tot_loss[loss=0.1754, simple_loss=0.2475, pruned_loss=0.05167, over 952833.88 frames. ], batch size: 47, lr: 3.22e-03, grad_scale: 32.0 2023-04-27 14:38:52,248 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.78 vs. limit=2.0 2023-04-27 14:39:00,869 INFO [optim.py:369] (6/7) 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 14:39:02,950 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.71 vs. limit=5.0 2023-04-27 14:39:11,491 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7853, 1.1743, 1.8410, 2.2652, 1.8700, 1.7344, 1.8204, 1.7622], device='cuda:6'), covar=tensor([0.4333, 0.6305, 0.5572, 0.5280, 0.5618, 0.7234, 0.7330, 0.7798], device='cuda:6'), in_proj_covar=tensor([0.0430, 0.0411, 0.0505, 0.0505, 0.0458, 0.0485, 0.0493, 0.0499], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 14:39:13,176 INFO [finetune.py:976] (6/7) Epoch 20, batch 4650, loss[loss=0.1546, simple_loss=0.2316, pruned_loss=0.03882, over 4896.00 frames. ], tot_loss[loss=0.1729, simple_loss=0.2445, pruned_loss=0.05072, over 954844.15 frames. ], batch size: 35, lr: 3.22e-03, grad_scale: 32.0 2023-04-27 14:39:40,661 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0857, 1.8820, 2.0043, 2.4352, 2.3888, 1.9367, 1.5652, 2.1550], device='cuda:6'), covar=tensor([0.0740, 0.0995, 0.0677, 0.0449, 0.0527, 0.0707, 0.0732, 0.0508], device='cuda:6'), in_proj_covar=tensor([0.0188, 0.0203, 0.0184, 0.0172, 0.0178, 0.0182, 0.0153, 0.0180], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 14:39:47,028 INFO [finetune.py:976] (6/7) Epoch 20, batch 4700, loss[loss=0.1461, simple_loss=0.2129, pruned_loss=0.03968, over 4837.00 frames. ], tot_loss[loss=0.1705, simple_loss=0.2416, pruned_loss=0.04966, over 955122.21 frames. ], batch size: 47, lr: 3.22e-03, grad_scale: 32.0 2023-04-27 14:40:02,397 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.5946, 2.9285, 0.9390, 1.6041, 2.4054, 1.4855, 4.1737, 2.1162], device='cuda:6'), covar=tensor([0.0583, 0.0763, 0.0875, 0.1248, 0.0461, 0.0994, 0.0273, 0.0601], device='cuda:6'), in_proj_covar=tensor([0.0050, 0.0064, 0.0047, 0.0046, 0.0049, 0.0051, 0.0073, 0.0051], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0007, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:6') 2023-04-27 14:40:08,112 INFO [optim.py:369] (6/7) 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 14:40:19,995 INFO [finetune.py:976] (6/7) Epoch 20, batch 4750, loss[loss=0.1465, simple_loss=0.2157, pruned_loss=0.03866, over 4901.00 frames. ], tot_loss[loss=0.1682, simple_loss=0.2388, pruned_loss=0.04874, over 954197.71 frames. ], batch size: 35, lr: 3.22e-03, grad_scale: 32.0 2023-04-27 14:40:20,622 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=113577.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:40:29,028 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=113590.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:40:36,649 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-04-27 14:40:46,847 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=113617.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:40:52,167 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 20, batch 4800, loss[loss=0.1649, simple_loss=0.2426, pruned_loss=0.04364, over 4772.00 frames. ], tot_loss[loss=0.1725, simple_loss=0.2432, pruned_loss=0.05089, over 954614.74 frames. ], batch size: 26, lr: 3.22e-03, grad_scale: 32.0 2023-04-27 14:41:05,320 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.3236, 1.2310, 1.5441, 1.5897, 1.2079, 1.0804, 1.3365, 0.9109], device='cuda:6'), covar=tensor([0.0536, 0.0548, 0.0397, 0.0557, 0.0812, 0.1101, 0.0564, 0.0586], device='cuda:6'), in_proj_covar=tensor([0.0069, 0.0069, 0.0068, 0.0068, 0.0076, 0.0097, 0.0074, 0.0066], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-27 14:41:06,517 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=113646.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:41:15,406 INFO [optim.py:369] (6/7) 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 14:41:27,082 INFO [finetune.py:976] (6/7) Epoch 20, batch 4850, loss[loss=0.1973, simple_loss=0.2683, pruned_loss=0.06318, over 4819.00 frames. ], tot_loss[loss=0.1746, simple_loss=0.2458, pruned_loss=0.05169, over 953184.79 frames. ], batch size: 40, lr: 3.22e-03, grad_scale: 32.0 2023-04-27 14:41:27,789 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=113678.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:41:46,450 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=113707.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:41:58,266 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.78 vs. limit=2.0 2023-04-27 14:42:00,311 INFO [finetune.py:976] (6/7) Epoch 20, batch 4900, loss[loss=0.1507, simple_loss=0.2352, pruned_loss=0.03309, over 4928.00 frames. ], tot_loss[loss=0.1745, simple_loss=0.2465, pruned_loss=0.05123, over 955610.11 frames. ], batch size: 42, lr: 3.22e-03, grad_scale: 32.0 2023-04-27 14:42:27,122 INFO [optim.py:369] (6/7) 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 14:42:38,977 INFO [finetune.py:976] (6/7) Epoch 20, batch 4950, loss[loss=0.1785, simple_loss=0.2563, pruned_loss=0.05038, over 4917.00 frames. ], tot_loss[loss=0.176, simple_loss=0.248, pruned_loss=0.05198, over 954343.73 frames. ], batch size: 37, lr: 3.22e-03, grad_scale: 32.0 2023-04-27 14:42:49,875 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-04-27 14:43:30,854 INFO [finetune.py:976] (6/7) Epoch 20, batch 5000, loss[loss=0.1485, simple_loss=0.2247, pruned_loss=0.03618, over 4787.00 frames. ], tot_loss[loss=0.1747, simple_loss=0.2466, pruned_loss=0.05136, over 952210.16 frames. ], batch size: 29, lr: 3.22e-03, grad_scale: 32.0 2023-04-27 14:43:41,605 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.1993, 1.6362, 2.0542, 2.3809, 1.9830, 1.6149, 1.2848, 1.7805], device='cuda:6'), covar=tensor([0.3009, 0.2947, 0.1521, 0.2134, 0.2403, 0.2512, 0.4068, 0.1939], device='cuda:6'), in_proj_covar=tensor([0.0292, 0.0246, 0.0226, 0.0313, 0.0219, 0.0233, 0.0226, 0.0184], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:6') 2023-04-27 14:44:14,003 INFO [optim.py:369] (6/7) 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 14:44:32,034 INFO [finetune.py:976] (6/7) Epoch 20, batch 5050, loss[loss=0.1414, simple_loss=0.2176, pruned_loss=0.03261, over 4863.00 frames. ], tot_loss[loss=0.1731, simple_loss=0.2447, pruned_loss=0.05081, over 954350.52 frames. ], batch size: 31, lr: 3.22e-03, grad_scale: 32.0 2023-04-27 14:44:41,960 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.1781, 2.2572, 1.8081, 1.9470, 2.2267, 1.8628, 2.6207, 1.5677], device='cuda:6'), covar=tensor([0.3461, 0.1518, 0.4181, 0.2829, 0.1694, 0.2165, 0.1524, 0.4166], device='cuda:6'), in_proj_covar=tensor([0.0341, 0.0348, 0.0425, 0.0352, 0.0380, 0.0373, 0.0370, 0.0418], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 14:44:53,361 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=113890.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:45:34,209 INFO [finetune.py:976] (6/7) Epoch 20, batch 5100, loss[loss=0.161, simple_loss=0.2268, pruned_loss=0.04755, over 4766.00 frames. ], tot_loss[loss=0.1698, simple_loss=0.2408, pruned_loss=0.04936, over 955237.16 frames. ], batch size: 28, lr: 3.22e-03, grad_scale: 64.0 2023-04-27 14:45:42,072 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=113938.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:45:45,045 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.3306, 1.3889, 1.6753, 1.7791, 1.6573, 1.7444, 1.7223, 1.7573], device='cuda:6'), covar=tensor([0.3886, 0.5189, 0.4199, 0.4451, 0.5436, 0.7149, 0.5049, 0.4581], device='cuda:6'), in_proj_covar=tensor([0.0337, 0.0374, 0.0321, 0.0335, 0.0345, 0.0393, 0.0356, 0.0328], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 14:45:51,555 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.04 vs. limit=5.0 2023-04-27 14:45:57,340 INFO [optim.py:369] (6/7) 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 14:46:05,377 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=113973.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:46:08,236 INFO [finetune.py:976] (6/7) Epoch 20, batch 5150, loss[loss=0.1792, simple_loss=0.2531, pruned_loss=0.05266, over 4903.00 frames. ], tot_loss[loss=0.1717, simple_loss=0.2421, pruned_loss=0.05061, over 954602.05 frames. ], batch size: 43, lr: 3.22e-03, grad_scale: 64.0 2023-04-27 14:46:27,906 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=114002.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:46:27,996 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7489, 1.4063, 1.8861, 2.1745, 1.8275, 1.7604, 1.8585, 1.7657], device='cuda:6'), covar=tensor([0.4451, 0.6485, 0.5860, 0.5423, 0.5708, 0.7361, 0.7476, 0.8393], device='cuda:6'), in_proj_covar=tensor([0.0431, 0.0413, 0.0508, 0.0509, 0.0460, 0.0488, 0.0497, 0.0502], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 14:46:43,543 INFO [finetune.py:976] (6/7) Epoch 20, batch 5200, loss[loss=0.2064, simple_loss=0.2776, pruned_loss=0.06761, over 4770.00 frames. ], tot_loss[loss=0.1755, simple_loss=0.2469, pruned_loss=0.05203, over 957006.20 frames. ], batch size: 59, lr: 3.22e-03, grad_scale: 64.0 2023-04-27 14:47:06,607 INFO [optim.py:369] (6/7) 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 14:47:16,855 INFO [finetune.py:976] (6/7) Epoch 20, batch 5250, loss[loss=0.1567, simple_loss=0.2362, pruned_loss=0.03859, over 4755.00 frames. ], tot_loss[loss=0.1776, simple_loss=0.2491, pruned_loss=0.05304, over 956504.61 frames. ], batch size: 28, lr: 3.22e-03, grad_scale: 64.0 2023-04-27 14:47:27,546 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([0.9848, 2.1871, 1.7818, 1.7084, 1.9025, 1.5895, 2.5443, 1.3806], device='cuda:6'), covar=tensor([0.3521, 0.1478, 0.4486, 0.2571, 0.1843, 0.2546, 0.1221, 0.4876], device='cuda:6'), in_proj_covar=tensor([0.0340, 0.0347, 0.0423, 0.0351, 0.0379, 0.0372, 0.0369, 0.0416], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 14:47:32,981 INFO [scaling.py:679] (6/7) 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] (6/7) Epoch 20, batch 5300, loss[loss=0.165, simple_loss=0.2383, pruned_loss=0.04581, over 4815.00 frames. ], tot_loss[loss=0.1768, simple_loss=0.2485, pruned_loss=0.05252, over 953484.52 frames. ], batch size: 38, lr: 3.22e-03, grad_scale: 64.0 2023-04-27 14:47:50,821 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=114127.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:48:10,459 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0344, 1.7149, 2.2142, 2.5644, 2.0694, 1.9116, 2.0524, 2.0710], device='cuda:6'), covar=tensor([0.5148, 0.7799, 0.7894, 0.6073, 0.6871, 0.9951, 0.9684, 0.9463], device='cuda:6'), in_proj_covar=tensor([0.0430, 0.0413, 0.0506, 0.0507, 0.0458, 0.0487, 0.0495, 0.0500], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 14:48:13,282 INFO [optim.py:369] (6/7) 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 14:48:24,239 INFO [finetune.py:976] (6/7) Epoch 20, batch 5350, loss[loss=0.1705, simple_loss=0.2312, pruned_loss=0.05489, over 4743.00 frames. ], tot_loss[loss=0.1754, simple_loss=0.2476, pruned_loss=0.05163, over 953822.92 frames. ], batch size: 26, lr: 3.22e-03, grad_scale: 64.0 2023-04-27 14:48:31,439 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=114188.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:48:52,727 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.47 vs. limit=5.0 2023-04-27 14:48:58,067 INFO [finetune.py:976] (6/7) Epoch 20, batch 5400, loss[loss=0.1641, simple_loss=0.2221, pruned_loss=0.05307, over 4859.00 frames. ], tot_loss[loss=0.1731, simple_loss=0.2446, pruned_loss=0.05082, over 953853.98 frames. ], batch size: 49, lr: 3.22e-03, grad_scale: 64.0 2023-04-27 14:49:33,870 INFO [optim.py:369] (6/7) 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 14:49:53,932 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=114273.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:49:55,188 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0540, 1.8039, 1.9903, 2.3160, 2.4023, 1.9082, 1.5051, 2.0177], device='cuda:6'), covar=tensor([0.0721, 0.0997, 0.0614, 0.0526, 0.0506, 0.0763, 0.0786, 0.0548], device='cuda:6'), in_proj_covar=tensor([0.0187, 0.0202, 0.0183, 0.0171, 0.0177, 0.0181, 0.0152, 0.0179], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 14:49:56,282 INFO [finetune.py:976] (6/7) Epoch 20, batch 5450, loss[loss=0.1339, simple_loss=0.2104, pruned_loss=0.0287, over 4910.00 frames. ], tot_loss[loss=0.1716, simple_loss=0.2429, pruned_loss=0.05012, over 954730.49 frames. ], batch size: 36, lr: 3.22e-03, grad_scale: 64.0 2023-04-27 14:50:28,202 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=114302.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:50:29,914 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6881, 1.2720, 4.3218, 4.0709, 3.7583, 4.0262, 3.9588, 3.7609], device='cuda:6'), covar=tensor([0.7013, 0.5958, 0.0994, 0.1505, 0.0958, 0.2071, 0.1933, 0.1300], device='cuda:6'), in_proj_covar=tensor([0.0309, 0.0306, 0.0405, 0.0403, 0.0347, 0.0409, 0.0312, 0.0367], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-27 14:50:58,446 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=114321.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:51:02,103 INFO [finetune.py:976] (6/7) Epoch 20, batch 5500, loss[loss=0.1545, simple_loss=0.2212, pruned_loss=0.04394, over 4902.00 frames. ], tot_loss[loss=0.1702, simple_loss=0.2407, pruned_loss=0.04988, over 953810.52 frames. ], batch size: 37, lr: 3.22e-03, grad_scale: 64.0 2023-04-27 14:51:15,065 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.0994, 2.2134, 1.8684, 1.8115, 2.2877, 1.7562, 2.7390, 1.6276], device='cuda:6'), covar=tensor([0.3626, 0.1820, 0.3985, 0.2895, 0.1504, 0.2452, 0.1228, 0.4427], device='cuda:6'), in_proj_covar=tensor([0.0339, 0.0345, 0.0422, 0.0349, 0.0378, 0.0371, 0.0368, 0.0415], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 14:51:22,114 INFO [zipformer.py:1188] (6/7) 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] (6/7) 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 14:51:41,083 INFO [finetune.py:976] (6/7) Epoch 20, batch 5550, loss[loss=0.1803, simple_loss=0.2474, pruned_loss=0.05655, over 4869.00 frames. ], tot_loss[loss=0.1704, simple_loss=0.2412, pruned_loss=0.04982, over 951812.46 frames. ], batch size: 31, lr: 3.22e-03, grad_scale: 64.0 2023-04-27 14:51:47,912 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([3.8185, 3.7314, 2.8840, 4.4605, 3.8214, 3.8310, 1.8416, 3.7624], device='cuda:6'), covar=tensor([0.1728, 0.1193, 0.3713, 0.1345, 0.3176, 0.1801, 0.5343, 0.2383], device='cuda:6'), in_proj_covar=tensor([0.0246, 0.0216, 0.0253, 0.0306, 0.0297, 0.0247, 0.0275, 0.0273], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 14:51:59,024 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-04-27 14:52:13,273 INFO [finetune.py:976] (6/7) Epoch 20, batch 5600, loss[loss=0.1368, simple_loss=0.2211, pruned_loss=0.02626, over 4788.00 frames. ], tot_loss[loss=0.1722, simple_loss=0.244, pruned_loss=0.0502, over 951507.09 frames. ], batch size: 29, lr: 3.22e-03, grad_scale: 64.0 2023-04-27 14:52:14,316 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-04-27 14:52:32,508 INFO [optim.py:369] (6/7) 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 14:52:42,448 INFO [finetune.py:976] (6/7) Epoch 20, batch 5650, loss[loss=0.1954, simple_loss=0.2731, pruned_loss=0.05884, over 4886.00 frames. ], tot_loss[loss=0.1745, simple_loss=0.2468, pruned_loss=0.05106, over 951063.35 frames. ], batch size: 32, lr: 3.22e-03, grad_scale: 64.0 2023-04-27 14:52:42,524 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.3408, 1.6234, 1.4065, 1.7881, 1.7278, 2.0316, 1.4813, 3.8649], device='cuda:6'), covar=tensor([0.0630, 0.0865, 0.0887, 0.1210, 0.0693, 0.0632, 0.0842, 0.0157], device='cuda:6'), in_proj_covar=tensor([0.0038, 0.0039, 0.0040, 0.0043, 0.0040, 0.0038, 0.0038, 0.0056], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0013, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:6') 2023-04-27 14:52:46,428 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=114483.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:52:49,569 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.62 vs. limit=2.0 2023-04-27 14:52:53,049 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8755, 1.2760, 1.8460, 2.3043, 1.9486, 1.7520, 1.8337, 1.8229], device='cuda:6'), covar=tensor([0.4703, 0.7130, 0.6465, 0.6232, 0.6433, 0.8402, 0.8530, 0.8287], device='cuda:6'), in_proj_covar=tensor([0.0429, 0.0412, 0.0505, 0.0506, 0.0457, 0.0486, 0.0495, 0.0498], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 14:52:58,910 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.5848, 2.1889, 2.5556, 3.1466, 2.5080, 2.1499, 2.0490, 2.3712], device='cuda:6'), covar=tensor([0.3248, 0.2989, 0.1606, 0.1922, 0.2730, 0.2690, 0.3393, 0.2020], device='cuda:6'), in_proj_covar=tensor([0.0292, 0.0245, 0.0226, 0.0313, 0.0220, 0.0233, 0.0227, 0.0184], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:6') 2023-04-27 14:53:01,035 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.78 vs. limit=5.0 2023-04-27 14:53:12,743 INFO [finetune.py:976] (6/7) Epoch 20, batch 5700, loss[loss=0.1361, simple_loss=0.2095, pruned_loss=0.03135, over 3940.00 frames. ], tot_loss[loss=0.1709, simple_loss=0.2426, pruned_loss=0.04956, over 937536.90 frames. ], batch size: 17, lr: 3.22e-03, grad_scale: 64.0 2023-04-27 14:53:23,789 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.1879, 2.2466, 2.1470, 1.9432, 2.3666, 1.9733, 2.8748, 1.8965], device='cuda:6'), covar=tensor([0.3417, 0.1455, 0.3661, 0.2431, 0.1374, 0.1989, 0.1099, 0.3728], device='cuda:6'), in_proj_covar=tensor([0.0342, 0.0348, 0.0426, 0.0352, 0.0381, 0.0373, 0.0372, 0.0418], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 14:53:23,855 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.01 vs. limit=5.0 2023-04-27 14:53:39,073 INFO [finetune.py:976] (6/7) Epoch 21, batch 0, loss[loss=0.1245, simple_loss=0.2031, pruned_loss=0.02297, over 4708.00 frames. ], tot_loss[loss=0.1245, simple_loss=0.2031, pruned_loss=0.02297, over 4708.00 frames. ], batch size: 23, lr: 3.21e-03, grad_scale: 64.0 2023-04-27 14:53:39,073 INFO [finetune.py:1001] (6/7) Computing validation loss 2023-04-27 14:53:45,745 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.3876, 1.3838, 3.8431, 3.5868, 3.4712, 3.7027, 3.7976, 3.3768], device='cuda:6'), covar=tensor([0.6956, 0.5352, 0.1329, 0.2045, 0.1222, 0.1257, 0.0700, 0.1626], device='cuda:6'), in_proj_covar=tensor([0.0309, 0.0306, 0.0403, 0.0403, 0.0347, 0.0407, 0.0311, 0.0365], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-27 14:53:56,090 INFO [finetune.py:1010] (6/7) Epoch 21, validation: loss=0.1544, simple_loss=0.2245, pruned_loss=0.04212, over 2265189.00 frames. 2023-04-27 14:53:56,091 INFO [finetune.py:1011] (6/7) Maximum memory allocated so far is 6365MB 2023-04-27 14:54:02,566 INFO [optim.py:369] (6/7) 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 14:54:24,907 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6735, 1.2020, 1.7377, 2.2445, 1.8039, 1.6267, 1.7013, 1.6415], device='cuda:6'), covar=tensor([0.4185, 0.6266, 0.5504, 0.4969, 0.5174, 0.7269, 0.6894, 0.8216], device='cuda:6'), in_proj_covar=tensor([0.0427, 0.0410, 0.0504, 0.0503, 0.0455, 0.0483, 0.0493, 0.0496], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 14:54:37,747 INFO [finetune.py:976] (6/7) Epoch 21, batch 50, loss[loss=0.1637, simple_loss=0.2335, pruned_loss=0.04694, over 4895.00 frames. ], tot_loss[loss=0.1845, simple_loss=0.2559, pruned_loss=0.05654, over 214863.02 frames. ], batch size: 36, lr: 3.21e-03, grad_scale: 64.0 2023-04-27 14:55:05,772 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.2739, 1.4723, 1.6808, 1.8384, 1.7463, 1.8131, 1.7553, 1.7887], device='cuda:6'), covar=tensor([0.3675, 0.4742, 0.4203, 0.4341, 0.5077, 0.6819, 0.4865, 0.4577], device='cuda:6'), in_proj_covar=tensor([0.0340, 0.0376, 0.0322, 0.0337, 0.0348, 0.0395, 0.0358, 0.0329], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 14:55:26,966 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-04-27 14:55:37,893 INFO [finetune.py:976] (6/7) Epoch 21, batch 100, loss[loss=0.2116, simple_loss=0.2707, pruned_loss=0.07625, over 4829.00 frames. ], tot_loss[loss=0.1755, simple_loss=0.2446, pruned_loss=0.05317, over 378624.73 frames. ], batch size: 47, lr: 3.21e-03, grad_scale: 64.0 2023-04-27 14:55:46,627 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-04-27 14:55:48,250 INFO [optim.py:369] (6/7) 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 14:56:44,432 INFO [finetune.py:976] (6/7) Epoch 21, batch 150, loss[loss=0.1473, simple_loss=0.2214, pruned_loss=0.03661, over 4865.00 frames. ], tot_loss[loss=0.1692, simple_loss=0.2388, pruned_loss=0.04984, over 507143.52 frames. ], batch size: 34, lr: 3.21e-03, grad_scale: 64.0 2023-04-27 14:56:57,802 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=114716.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:56:58,766 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.73 vs. limit=2.0 2023-04-27 14:57:12,222 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([0.6693, 1.7619, 1.6232, 1.3094, 1.7023, 1.4436, 2.1576, 1.4451], device='cuda:6'), covar=tensor([0.4009, 0.1744, 0.5284, 0.3440, 0.1949, 0.2443, 0.1515, 0.4969], device='cuda:6'), in_proj_covar=tensor([0.0344, 0.0350, 0.0427, 0.0354, 0.0383, 0.0374, 0.0374, 0.0420], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 14:57:22,717 INFO [finetune.py:976] (6/7) Epoch 21, batch 200, loss[loss=0.1593, simple_loss=0.2348, pruned_loss=0.04184, over 4778.00 frames. ], tot_loss[loss=0.1692, simple_loss=0.2391, pruned_loss=0.04963, over 605490.34 frames. ], batch size: 28, lr: 3.21e-03, grad_scale: 64.0 2023-04-27 14:57:26,735 INFO [optim.py:369] (6/7) 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 14:57:38,699 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=114777.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:57:42,348 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=114783.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:57:56,203 INFO [finetune.py:976] (6/7) Epoch 21, batch 250, loss[loss=0.1755, simple_loss=0.2496, pruned_loss=0.05068, over 4866.00 frames. ], tot_loss[loss=0.1704, simple_loss=0.2408, pruned_loss=0.05003, over 683722.39 frames. ], batch size: 31, lr: 3.21e-03, grad_scale: 32.0 2023-04-27 14:58:15,125 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 21, batch 300, loss[loss=0.172, simple_loss=0.233, pruned_loss=0.05553, over 4763.00 frames. ], tot_loss[loss=0.1726, simple_loss=0.2438, pruned_loss=0.05066, over 743284.91 frames. ], batch size: 26, lr: 3.21e-03, grad_scale: 32.0 2023-04-27 14:58:34,665 INFO [optim.py:369] (6/7) 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 14:58:41,076 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6288, 1.3518, 0.7365, 1.2819, 1.4562, 1.4866, 1.3788, 1.3852], device='cuda:6'), covar=tensor([0.0511, 0.0391, 0.0366, 0.0563, 0.0284, 0.0512, 0.0495, 0.0582], device='cuda:6'), in_proj_covar=tensor([0.0028, 0.0024, 0.0022, 0.0029, 0.0019, 0.0028, 0.0028, 0.0028], device='cuda:6'), out_proj_covar=tensor([0.0051, 0.0045, 0.0038, 0.0052, 0.0038, 0.0050, 0.0050, 0.0051], device='cuda:6') 2023-04-27 14:58:42,260 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=114871.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:58:48,233 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5635, 1.4034, 1.8629, 1.8438, 1.4183, 1.2853, 1.5347, 1.0373], device='cuda:6'), covar=tensor([0.0560, 0.0638, 0.0403, 0.0590, 0.0761, 0.1155, 0.0612, 0.0615], device='cuda:6'), in_proj_covar=tensor([0.0069, 0.0068, 0.0067, 0.0067, 0.0075, 0.0096, 0.0073, 0.0065], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-27 14:58:59,707 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9492, 2.4458, 2.0275, 2.2834, 1.6335, 2.0333, 2.0089, 1.5386], device='cuda:6'), covar=tensor([0.1956, 0.1253, 0.0818, 0.1211, 0.3361, 0.1072, 0.1843, 0.2631], device='cuda:6'), in_proj_covar=tensor([0.0285, 0.0300, 0.0215, 0.0276, 0.0311, 0.0254, 0.0247, 0.0261], device='cuda:6'), out_proj_covar=tensor([1.1405e-04, 1.1935e-04, 8.5258e-05, 1.0914e-04, 1.2608e-04, 1.0050e-04, 1.0002e-04, 1.0355e-04], device='cuda:6') 2023-04-27 14:59:03,088 INFO [finetune.py:976] (6/7) Epoch 21, batch 350, loss[loss=0.2136, simple_loss=0.2841, pruned_loss=0.07159, over 4922.00 frames. ], tot_loss[loss=0.1745, simple_loss=0.246, pruned_loss=0.05156, over 789856.63 frames. ], batch size: 42, lr: 3.21e-03, grad_scale: 32.0 2023-04-27 14:59:05,528 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.7988, 2.0493, 1.9864, 2.1902, 1.9344, 2.0163, 2.0147, 2.0061], device='cuda:6'), covar=tensor([0.4037, 0.6821, 0.5278, 0.4720, 0.6167, 0.7615, 0.6226, 0.6130], device='cuda:6'), in_proj_covar=tensor([0.0339, 0.0376, 0.0323, 0.0337, 0.0348, 0.0395, 0.0358, 0.0330], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 14:59:22,176 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=114932.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:59:36,226 INFO [finetune.py:976] (6/7) Epoch 21, batch 400, loss[loss=0.1591, simple_loss=0.2348, pruned_loss=0.04166, over 4924.00 frames. ], tot_loss[loss=0.1753, simple_loss=0.2473, pruned_loss=0.05165, over 824830.36 frames. ], batch size: 42, lr: 3.21e-03, grad_scale: 32.0 2023-04-27 14:59:40,905 INFO [optim.py:369] (6/7) 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] (6/7) Epoch 21, batch 450, loss[loss=0.1572, simple_loss=0.2303, pruned_loss=0.04201, over 4789.00 frames. ], tot_loss[loss=0.1743, simple_loss=0.2461, pruned_loss=0.05124, over 854544.40 frames. ], batch size: 29, lr: 3.21e-03, grad_scale: 32.0 2023-04-27 15:00:22,692 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.1322, 1.8825, 2.1485, 2.4998, 2.5060, 1.9077, 1.5189, 2.1608], device='cuda:6'), covar=tensor([0.0786, 0.1034, 0.0597, 0.0453, 0.0522, 0.0838, 0.0786, 0.0533], device='cuda:6'), in_proj_covar=tensor([0.0189, 0.0204, 0.0185, 0.0173, 0.0179, 0.0182, 0.0153, 0.0180], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 15:00:59,033 INFO [finetune.py:976] (6/7) Epoch 21, batch 500, loss[loss=0.1696, simple_loss=0.242, pruned_loss=0.04866, over 4806.00 frames. ], tot_loss[loss=0.1728, simple_loss=0.2442, pruned_loss=0.05069, over 877958.18 frames. ], batch size: 45, lr: 3.21e-03, grad_scale: 32.0 2023-04-27 15:01:09,614 INFO [optim.py:369] (6/7) 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 15:01:16,905 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=115072.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:01:32,714 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=115086.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:01:54,518 INFO [finetune.py:976] (6/7) Epoch 21, batch 550, loss[loss=0.2019, simple_loss=0.2681, pruned_loss=0.06783, over 4860.00 frames. ], tot_loss[loss=0.1707, simple_loss=0.2416, pruned_loss=0.04986, over 897465.39 frames. ], batch size: 44, lr: 3.21e-03, grad_scale: 32.0 2023-04-27 15:02:53,400 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=115147.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:03:02,664 INFO [finetune.py:976] (6/7) Epoch 21, batch 600, loss[loss=0.1519, simple_loss=0.2238, pruned_loss=0.04002, over 4836.00 frames. ], tot_loss[loss=0.1694, simple_loss=0.2403, pruned_loss=0.04924, over 910940.57 frames. ], batch size: 25, lr: 3.21e-03, grad_scale: 32.0 2023-04-27 15:03:12,535 INFO [optim.py:369] (6/7) 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 15:03:44,108 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.10 vs. limit=2.0 2023-04-27 15:03:45,086 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.4053, 1.3471, 1.7245, 1.7186, 1.3151, 1.1251, 1.2504, 0.7352], device='cuda:6'), covar=tensor([0.0521, 0.0570, 0.0353, 0.0565, 0.0691, 0.1303, 0.0716, 0.0726], device='cuda:6'), in_proj_covar=tensor([0.0069, 0.0068, 0.0067, 0.0067, 0.0075, 0.0096, 0.0073, 0.0065], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-27 15:04:02,676 INFO [finetune.py:976] (6/7) Epoch 21, batch 650, loss[loss=0.1434, simple_loss=0.2214, pruned_loss=0.03267, over 4773.00 frames. ], tot_loss[loss=0.1722, simple_loss=0.2437, pruned_loss=0.05032, over 920677.72 frames. ], batch size: 27, lr: 3.21e-03, grad_scale: 32.0 2023-04-27 15:04:14,828 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=3.86 vs. limit=5.0 2023-04-27 15:04:17,781 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=115227.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:04:36,595 INFO [finetune.py:976] (6/7) Epoch 21, batch 700, loss[loss=0.1614, simple_loss=0.236, pruned_loss=0.04335, over 4764.00 frames. ], tot_loss[loss=0.1743, simple_loss=0.2465, pruned_loss=0.05104, over 929823.63 frames. ], batch size: 28, lr: 3.21e-03, grad_scale: 32.0 2023-04-27 15:04:40,863 INFO [optim.py:369] (6/7) 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 15:05:10,549 INFO [finetune.py:976] (6/7) Epoch 21, batch 750, loss[loss=0.1555, simple_loss=0.2198, pruned_loss=0.04557, over 4712.00 frames. ], tot_loss[loss=0.1747, simple_loss=0.2469, pruned_loss=0.05131, over 933630.66 frames. ], batch size: 23, lr: 3.21e-03, grad_scale: 32.0 2023-04-27 15:05:44,407 INFO [finetune.py:976] (6/7) Epoch 21, batch 800, loss[loss=0.1808, simple_loss=0.2478, pruned_loss=0.05696, over 4888.00 frames. ], tot_loss[loss=0.1743, simple_loss=0.2468, pruned_loss=0.05097, over 939361.20 frames. ], batch size: 32, lr: 3.21e-03, grad_scale: 32.0 2023-04-27 15:05:47,510 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6123, 1.5739, 0.8021, 1.3187, 1.7049, 1.5253, 1.4042, 1.4930], device='cuda:6'), covar=tensor([0.0492, 0.0365, 0.0331, 0.0538, 0.0273, 0.0504, 0.0495, 0.0562], device='cuda:6'), in_proj_covar=tensor([0.0028, 0.0024, 0.0022, 0.0029, 0.0019, 0.0028, 0.0028, 0.0028], device='cuda:6'), out_proj_covar=tensor([0.0051, 0.0045, 0.0038, 0.0052, 0.0038, 0.0050, 0.0050, 0.0051], device='cuda:6') 2023-04-27 15:05:48,601 INFO [optim.py:369] (6/7) 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 15:05:54,631 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=115370.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:05:55,236 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=115371.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:05:55,828 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=115372.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:06:16,443 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.5190, 2.4452, 2.7301, 3.0005, 2.9558, 2.3752, 1.9404, 2.5973], device='cuda:6'), covar=tensor([0.0861, 0.0891, 0.0522, 0.0520, 0.0580, 0.0881, 0.0811, 0.0554], device='cuda:6'), in_proj_covar=tensor([0.0188, 0.0202, 0.0184, 0.0172, 0.0178, 0.0181, 0.0153, 0.0179], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 15:06:18,147 INFO [finetune.py:976] (6/7) Epoch 21, batch 850, loss[loss=0.142, simple_loss=0.2184, pruned_loss=0.03279, over 4790.00 frames. ], tot_loss[loss=0.1736, simple_loss=0.2452, pruned_loss=0.05094, over 943781.34 frames. ], batch size: 29, lr: 3.21e-03, grad_scale: 32.0 2023-04-27 15:06:28,418 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=115420.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:06:35,556 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=115431.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:06:36,170 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=115432.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:06:43,131 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=115442.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:07:01,735 INFO [finetune.py:976] (6/7) Epoch 21, batch 900, loss[loss=0.1453, simple_loss=0.2191, pruned_loss=0.03573, over 4946.00 frames. ], tot_loss[loss=0.1712, simple_loss=0.2423, pruned_loss=0.05004, over 945956.54 frames. ], batch size: 33, lr: 3.20e-03, grad_scale: 32.0 2023-04-27 15:07:06,010 INFO [optim.py:369] (6/7) 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 15:07:34,518 INFO [finetune.py:976] (6/7) Epoch 21, batch 950, loss[loss=0.1755, simple_loss=0.2442, pruned_loss=0.05345, over 4824.00 frames. ], tot_loss[loss=0.1702, simple_loss=0.241, pruned_loss=0.04969, over 947779.36 frames. ], batch size: 30, lr: 3.20e-03, grad_scale: 32.0 2023-04-27 15:08:06,786 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=115527.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:08:09,244 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6809, 1.4027, 4.1920, 3.9363, 3.6264, 3.9334, 3.8595, 3.7198], device='cuda:6'), covar=tensor([0.6773, 0.5661, 0.1033, 0.1602, 0.1213, 0.1615, 0.1900, 0.1398], device='cuda:6'), in_proj_covar=tensor([0.0305, 0.0303, 0.0399, 0.0399, 0.0342, 0.0403, 0.0307, 0.0362], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-27 15:08:28,453 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.4129, 1.6594, 1.4841, 2.0720, 1.7600, 1.9578, 1.5053, 4.2339], device='cuda:6'), covar=tensor([0.0526, 0.0795, 0.0778, 0.1117, 0.0651, 0.0615, 0.0761, 0.0112], device='cuda:6'), in_proj_covar=tensor([0.0037, 0.0038, 0.0039, 0.0043, 0.0040, 0.0038, 0.0038, 0.0056], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:6') 2023-04-27 15:08:39,547 INFO [finetune.py:976] (6/7) Epoch 21, batch 1000, loss[loss=0.189, simple_loss=0.2619, pruned_loss=0.05804, over 4822.00 frames. ], tot_loss[loss=0.171, simple_loss=0.2425, pruned_loss=0.04973, over 950100.65 frames. ], batch size: 39, lr: 3.20e-03, grad_scale: 32.0 2023-04-27 15:08:49,080 INFO [optim.py:369] (6/7) 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 15:09:08,299 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=115575.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:09:34,383 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.66 vs. limit=5.0 2023-04-27 15:09:45,486 INFO [finetune.py:976] (6/7) Epoch 21, batch 1050, loss[loss=0.1861, simple_loss=0.255, pruned_loss=0.05854, over 4904.00 frames. ], tot_loss[loss=0.1722, simple_loss=0.2446, pruned_loss=0.04986, over 951776.38 frames. ], batch size: 37, lr: 3.20e-03, grad_scale: 32.0 2023-04-27 15:09:54,245 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 2023-04-27 15:10:18,955 INFO [finetune.py:976] (6/7) Epoch 21, batch 1100, loss[loss=0.1567, simple_loss=0.2136, pruned_loss=0.04991, over 4233.00 frames. ], tot_loss[loss=0.1729, simple_loss=0.2453, pruned_loss=0.05025, over 952879.92 frames. ], batch size: 18, lr: 3.20e-03, grad_scale: 32.0 2023-04-27 15:10:23,752 INFO [optim.py:369] (6/7) 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 15:10:33,059 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.82 vs. limit=2.0 2023-04-27 15:10:52,442 INFO [finetune.py:976] (6/7) Epoch 21, batch 1150, loss[loss=0.1629, simple_loss=0.2201, pruned_loss=0.0529, over 4867.00 frames. ], tot_loss[loss=0.1724, simple_loss=0.2453, pruned_loss=0.04973, over 954516.50 frames. ], batch size: 31, lr: 3.20e-03, grad_scale: 32.0 2023-04-27 15:10:53,760 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.2869, 1.2871, 1.3643, 1.5857, 1.5947, 1.2402, 0.9636, 1.4989], device='cuda:6'), covar=tensor([0.0865, 0.1258, 0.0927, 0.0615, 0.0707, 0.0911, 0.0874, 0.0607], device='cuda:6'), in_proj_covar=tensor([0.0186, 0.0199, 0.0181, 0.0170, 0.0175, 0.0178, 0.0150, 0.0177], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 15:11:07,917 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=115726.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:11:08,530 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=115727.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:11:18,226 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 21, batch 1200, loss[loss=0.1554, simple_loss=0.2319, pruned_loss=0.03946, over 4821.00 frames. ], tot_loss[loss=0.1721, simple_loss=0.2444, pruned_loss=0.04992, over 955055.81 frames. ], batch size: 30, lr: 3.20e-03, grad_scale: 32.0 2023-04-27 15:11:27,188 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=3.90 vs. limit=5.0 2023-04-27 15:11:31,122 INFO [optim.py:369] (6/7) 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 15:11:49,080 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7429, 2.3779, 1.2374, 1.4856, 2.2840, 1.5892, 1.5892, 1.6671], device='cuda:6'), covar=tensor([0.0591, 0.0299, 0.0282, 0.0605, 0.0240, 0.0636, 0.0612, 0.0620], device='cuda:6'), in_proj_covar=tensor([0.0028, 0.0024, 0.0022, 0.0029, 0.0019, 0.0028, 0.0028, 0.0028], device='cuda:6'), out_proj_covar=tensor([0.0051, 0.0045, 0.0038, 0.0052, 0.0038, 0.0050, 0.0050, 0.0051], device='cuda:6') 2023-04-27 15:11:50,225 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=115790.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:11:54,994 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6996, 1.7357, 0.8215, 1.3702, 1.7324, 1.6033, 1.5026, 1.5218], device='cuda:6'), covar=tensor([0.0489, 0.0360, 0.0330, 0.0546, 0.0270, 0.0490, 0.0452, 0.0548], device='cuda:6'), in_proj_covar=tensor([0.0028, 0.0024, 0.0022, 0.0029, 0.0019, 0.0028, 0.0028, 0.0028], device='cuda:6'), out_proj_covar=tensor([0.0051, 0.0045, 0.0038, 0.0052, 0.0038, 0.0050, 0.0050, 0.0051], device='cuda:6') 2023-04-27 15:11:56,657 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.9083, 2.7998, 2.1680, 3.3270, 2.8686, 2.8966, 1.3379, 2.8568], device='cuda:6'), covar=tensor([0.2171, 0.1897, 0.3485, 0.3057, 0.3381, 0.2246, 0.5667, 0.3121], device='cuda:6'), in_proj_covar=tensor([0.0246, 0.0215, 0.0252, 0.0307, 0.0297, 0.0246, 0.0274, 0.0273], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 15:11:59,754 INFO [finetune.py:976] (6/7) Epoch 21, batch 1250, loss[loss=0.1635, simple_loss=0.2323, pruned_loss=0.04733, over 4829.00 frames. ], tot_loss[loss=0.1701, simple_loss=0.242, pruned_loss=0.04905, over 955979.06 frames. ], batch size: 39, lr: 3.20e-03, grad_scale: 32.0 2023-04-27 15:12:55,528 INFO [finetune.py:976] (6/7) Epoch 21, batch 1300, loss[loss=0.1246, simple_loss=0.1992, pruned_loss=0.02497, over 4903.00 frames. ], tot_loss[loss=0.1665, simple_loss=0.2382, pruned_loss=0.0474, over 956361.56 frames. ], batch size: 35, lr: 3.20e-03, grad_scale: 32.0 2023-04-27 15:12:59,790 INFO [optim.py:369] (6/7) 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 15:13:03,416 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6257, 1.5095, 1.6289, 2.0136, 2.0206, 1.5140, 1.2940, 1.7767], device='cuda:6'), covar=tensor([0.0755, 0.1041, 0.0703, 0.0507, 0.0593, 0.0882, 0.0756, 0.0521], device='cuda:6'), in_proj_covar=tensor([0.0187, 0.0200, 0.0182, 0.0171, 0.0176, 0.0179, 0.0151, 0.0178], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 15:13:10,419 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5459, 2.5326, 1.8761, 2.2635, 2.5274, 2.0832, 3.3070, 1.6916], device='cuda:6'), covar=tensor([0.3945, 0.2254, 0.4731, 0.3513, 0.1862, 0.2674, 0.1668, 0.4658], device='cuda:6'), in_proj_covar=tensor([0.0346, 0.0351, 0.0429, 0.0356, 0.0383, 0.0377, 0.0374, 0.0422], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 15:13:21,256 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=115892.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 15:13:29,449 INFO [finetune.py:976] (6/7) Epoch 21, batch 1350, loss[loss=0.1704, simple_loss=0.2429, pruned_loss=0.04898, over 4897.00 frames. ], tot_loss[loss=0.1667, simple_loss=0.238, pruned_loss=0.04769, over 957707.00 frames. ], batch size: 35, lr: 3.20e-03, grad_scale: 32.0 2023-04-27 15:13:47,363 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.74 vs. limit=2.0 2023-04-27 15:14:34,428 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=115953.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 15:14:34,901 INFO [finetune.py:976] (6/7) Epoch 21, batch 1400, loss[loss=0.1649, simple_loss=0.2504, pruned_loss=0.03973, over 4836.00 frames. ], tot_loss[loss=0.1703, simple_loss=0.2423, pruned_loss=0.04915, over 957000.38 frames. ], batch size: 30, lr: 3.20e-03, grad_scale: 32.0 2023-04-27 15:14:44,907 INFO [optim.py:369] (6/7) 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 15:15:26,318 INFO [finetune.py:976] (6/7) Epoch 21, batch 1450, loss[loss=0.1773, simple_loss=0.2519, pruned_loss=0.05134, over 4799.00 frames. ], tot_loss[loss=0.172, simple_loss=0.245, pruned_loss=0.04954, over 956410.82 frames. ], batch size: 25, lr: 3.20e-03, grad_scale: 32.0 2023-04-27 15:15:33,452 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=116014.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:15:42,222 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=116026.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:15:42,818 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 21, batch 1500, loss[loss=0.2462, simple_loss=0.3028, pruned_loss=0.09477, over 4735.00 frames. ], tot_loss[loss=0.174, simple_loss=0.2467, pruned_loss=0.05066, over 954903.88 frames. ], batch size: 59, lr: 3.20e-03, grad_scale: 32.0 2023-04-27 15:16:05,189 INFO [optim.py:369] (6/7) 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 15:16:13,682 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=116074.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:16:14,788 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=116075.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:16:14,834 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=116075.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 15:16:33,608 INFO [finetune.py:976] (6/7) Epoch 21, batch 1550, loss[loss=0.1473, simple_loss=0.2098, pruned_loss=0.04242, over 4350.00 frames. ], tot_loss[loss=0.1735, simple_loss=0.246, pruned_loss=0.05052, over 955667.13 frames. ], batch size: 18, lr: 3.20e-03, grad_scale: 32.0 2023-04-27 15:16:34,346 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=116105.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:16:41,981 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=116116.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:17:06,680 INFO [finetune.py:976] (6/7) Epoch 21, batch 1600, loss[loss=0.1493, simple_loss=0.2203, pruned_loss=0.03916, over 4858.00 frames. ], tot_loss[loss=0.1714, simple_loss=0.2432, pruned_loss=0.04978, over 954641.65 frames. ], batch size: 31, lr: 3.20e-03, grad_scale: 32.0 2023-04-27 15:17:10,952 INFO [optim.py:369] (6/7) 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 15:17:14,627 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=116166.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:17:16,585 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.28 vs. limit=5.0 2023-04-27 15:17:21,329 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 21, batch 1650, loss[loss=0.2041, simple_loss=0.2637, pruned_loss=0.07224, over 4716.00 frames. ], tot_loss[loss=0.1693, simple_loss=0.2407, pruned_loss=0.04897, over 954219.11 frames. ], batch size: 23, lr: 3.20e-03, grad_scale: 32.0 2023-04-27 15:18:14,859 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=116248.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 15:18:18,419 INFO [finetune.py:976] (6/7) Epoch 21, batch 1700, loss[loss=0.1751, simple_loss=0.2505, pruned_loss=0.04986, over 4907.00 frames. ], tot_loss[loss=0.1681, simple_loss=0.239, pruned_loss=0.04856, over 957358.81 frames. ], batch size: 43, lr: 3.20e-03, grad_scale: 32.0 2023-04-27 15:18:28,121 INFO [optim.py:369] (6/7) 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 15:18:49,730 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-04-27 15:19:31,551 INFO [finetune.py:976] (6/7) Epoch 21, batch 1750, loss[loss=0.2525, simple_loss=0.3268, pruned_loss=0.08915, over 4261.00 frames. ], tot_loss[loss=0.1716, simple_loss=0.2426, pruned_loss=0.05029, over 954534.70 frames. ], batch size: 66, lr: 3.20e-03, grad_scale: 32.0 2023-04-27 15:19:55,848 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=116323.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:20:33,795 INFO [finetune.py:976] (6/7) Epoch 21, batch 1800, loss[loss=0.1448, simple_loss=0.211, pruned_loss=0.03928, over 4202.00 frames. ], tot_loss[loss=0.1731, simple_loss=0.2444, pruned_loss=0.05092, over 950243.80 frames. ], batch size: 18, lr: 3.20e-03, grad_scale: 32.0 2023-04-27 15:20:38,074 INFO [optim.py:369] (6/7) 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 15:20:44,119 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=116370.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 15:20:50,223 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.2365, 2.2879, 1.8603, 1.9821, 2.3818, 1.8990, 2.9709, 1.6182], device='cuda:6'), covar=tensor([0.3942, 0.1990, 0.4726, 0.3225, 0.1739, 0.2463, 0.1348, 0.4760], device='cuda:6'), in_proj_covar=tensor([0.0344, 0.0349, 0.0428, 0.0354, 0.0382, 0.0377, 0.0373, 0.0419], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 15:20:53,267 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=116384.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:21:07,565 INFO [finetune.py:976] (6/7) Epoch 21, batch 1850, loss[loss=0.1676, simple_loss=0.2344, pruned_loss=0.05036, over 4781.00 frames. ], tot_loss[loss=0.1757, simple_loss=0.2472, pruned_loss=0.05214, over 952000.18 frames. ], batch size: 25, lr: 3.20e-03, grad_scale: 32.0 2023-04-27 15:21:14,459 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7591, 3.4946, 1.1742, 2.1222, 2.1512, 2.7081, 2.1885, 1.1478], device='cuda:6'), covar=tensor([0.1158, 0.0856, 0.1742, 0.1064, 0.0929, 0.0829, 0.1290, 0.2101], device='cuda:6'), in_proj_covar=tensor([0.0116, 0.0239, 0.0136, 0.0119, 0.0132, 0.0151, 0.0116, 0.0118], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-27 15:21:35,308 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 21, batch 1900, loss[loss=0.1292, simple_loss=0.2118, pruned_loss=0.02334, over 4780.00 frames. ], tot_loss[loss=0.176, simple_loss=0.2478, pruned_loss=0.05209, over 952635.64 frames. ], batch size: 25, lr: 3.20e-03, grad_scale: 32.0 2023-04-27 15:21:45,219 INFO [optim.py:369] (6/7) 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 15:21:45,308 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=116461.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:21:52,095 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 21, batch 1950, loss[loss=0.18, simple_loss=0.2473, pruned_loss=0.05637, over 4817.00 frames. ], tot_loss[loss=0.1751, simple_loss=0.2466, pruned_loss=0.05178, over 951428.77 frames. ], batch size: 41, lr: 3.20e-03, grad_scale: 32.0 2023-04-27 15:22:14,298 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=116505.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:22:16,002 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=116507.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 15:22:21,330 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.6647, 1.7234, 1.9723, 2.8799, 2.7785, 2.2991, 1.8164, 2.5225], device='cuda:6'), covar=tensor([0.0681, 0.1360, 0.0923, 0.0511, 0.0510, 0.0959, 0.0811, 0.0550], device='cuda:6'), in_proj_covar=tensor([0.0186, 0.0199, 0.0183, 0.0171, 0.0176, 0.0179, 0.0151, 0.0177], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 15:22:43,339 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=116548.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 15:22:46,839 INFO [finetune.py:976] (6/7) Epoch 21, batch 2000, loss[loss=0.1737, simple_loss=0.2446, pruned_loss=0.05143, over 4909.00 frames. ], tot_loss[loss=0.1727, simple_loss=0.2437, pruned_loss=0.05092, over 950493.88 frames. ], batch size: 37, lr: 3.20e-03, grad_scale: 32.0 2023-04-27 15:22:51,557 INFO [optim.py:369] (6/7) 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 15:22:55,223 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=116566.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:22:55,310 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-04-27 15:22:57,025 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.4070, 2.2342, 2.5005, 2.9408, 2.8634, 2.2395, 1.7407, 2.2802], device='cuda:6'), covar=tensor([0.0865, 0.0936, 0.0566, 0.0521, 0.0598, 0.0913, 0.0884, 0.0655], device='cuda:6'), in_proj_covar=tensor([0.0187, 0.0200, 0.0183, 0.0172, 0.0177, 0.0180, 0.0152, 0.0178], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 15:23:14,415 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=116596.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 15:23:20,569 INFO [finetune.py:976] (6/7) Epoch 21, batch 2050, loss[loss=0.1907, simple_loss=0.2532, pruned_loss=0.06411, over 4872.00 frames. ], tot_loss[loss=0.1698, simple_loss=0.2403, pruned_loss=0.04971, over 952654.54 frames. ], batch size: 31, lr: 3.20e-03, grad_scale: 32.0 2023-04-27 15:23:30,111 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=116618.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:23:30,308 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.14 vs. limit=5.0 2023-04-27 15:23:51,598 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.50 vs. limit=2.0 2023-04-27 15:23:59,068 INFO [finetune.py:976] (6/7) Epoch 21, batch 2100, loss[loss=0.2052, simple_loss=0.2691, pruned_loss=0.07064, over 4932.00 frames. ], tot_loss[loss=0.1696, simple_loss=0.2402, pruned_loss=0.04949, over 953436.47 frames. ], batch size: 33, lr: 3.20e-03, grad_scale: 32.0 2023-04-27 15:24:02,290 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.26 vs. limit=5.0 2023-04-27 15:24:03,921 INFO [optim.py:369] (6/7) 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 15:24:21,176 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=116670.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 15:24:32,900 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=116679.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:24:32,953 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=116679.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:24:45,813 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-04-27 15:25:00,267 INFO [finetune.py:976] (6/7) Epoch 21, batch 2150, loss[loss=0.151, simple_loss=0.2352, pruned_loss=0.03337, over 4822.00 frames. ], tot_loss[loss=0.1722, simple_loss=0.2435, pruned_loss=0.05042, over 954316.39 frames. ], batch size: 33, lr: 3.20e-03, grad_scale: 32.0 2023-04-27 15:25:05,363 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=116704.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:25:21,300 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=116718.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:25:31,344 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.56 vs. limit=2.0 2023-04-27 15:26:00,892 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5472, 1.6810, 0.8214, 1.3221, 1.7902, 1.4113, 1.3674, 1.4230], device='cuda:6'), covar=tensor([0.0466, 0.0355, 0.0337, 0.0510, 0.0271, 0.0481, 0.0473, 0.0530], device='cuda:6'), in_proj_covar=tensor([0.0028, 0.0024, 0.0022, 0.0029, 0.0019, 0.0028, 0.0028, 0.0028], device='cuda:6'), out_proj_covar=tensor([0.0051, 0.0045, 0.0038, 0.0052, 0.0038, 0.0050, 0.0050, 0.0051], device='cuda:6') 2023-04-27 15:26:05,620 INFO [finetune.py:976] (6/7) Epoch 21, batch 2200, loss[loss=0.1921, simple_loss=0.2645, pruned_loss=0.05988, over 4224.00 frames. ], tot_loss[loss=0.1733, simple_loss=0.2455, pruned_loss=0.05051, over 954532.42 frames. ], batch size: 65, lr: 3.19e-03, grad_scale: 32.0 2023-04-27 15:26:10,851 INFO [optim.py:369] (6/7) 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 15:26:10,950 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=116761.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:26:13,914 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=116765.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:26:18,722 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=116772.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:26:38,519 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=116802.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 15:26:38,548 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 21, batch 2250, loss[loss=0.1744, simple_loss=0.2548, pruned_loss=0.04699, over 4807.00 frames. ], tot_loss[loss=0.1739, simple_loss=0.2462, pruned_loss=0.05087, over 954734.00 frames. ], batch size: 40, lr: 3.19e-03, grad_scale: 64.0 2023-04-27 15:26:42,838 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=116809.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:26:50,612 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=116820.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:27:13,185 INFO [finetune.py:976] (6/7) Epoch 21, batch 2300, loss[loss=0.2048, simple_loss=0.2731, pruned_loss=0.06823, over 4868.00 frames. ], tot_loss[loss=0.1741, simple_loss=0.2465, pruned_loss=0.05084, over 955237.86 frames. ], batch size: 31, lr: 3.19e-03, grad_scale: 64.0 2023-04-27 15:27:17,444 INFO [optim.py:369] (6/7) 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 15:27:17,533 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=116861.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:27:18,813 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=116863.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:27:24,189 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-04-27 15:27:43,657 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.1073, 2.4993, 2.1197, 2.4053, 1.7275, 2.1518, 2.1082, 1.6721], device='cuda:6'), covar=tensor([0.1693, 0.0922, 0.0698, 0.1101, 0.2934, 0.0908, 0.1694, 0.2228], device='cuda:6'), in_proj_covar=tensor([0.0286, 0.0300, 0.0216, 0.0277, 0.0311, 0.0254, 0.0247, 0.0261], device='cuda:6'), out_proj_covar=tensor([1.1444e-04, 1.1908e-04, 8.5442e-05, 1.0963e-04, 1.2603e-04, 1.0050e-04, 9.9664e-05, 1.0365e-04], device='cuda:6') 2023-04-27 15:27:46,963 INFO [finetune.py:976] (6/7) Epoch 21, batch 2350, loss[loss=0.1811, simple_loss=0.2606, pruned_loss=0.05082, over 4911.00 frames. ], tot_loss[loss=0.1724, simple_loss=0.2441, pruned_loss=0.05033, over 954705.15 frames. ], batch size: 46, lr: 3.19e-03, grad_scale: 64.0 2023-04-27 15:27:50,089 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.3400, 3.1895, 1.0496, 1.8332, 1.7420, 2.3948, 1.8190, 1.0070], device='cuda:6'), covar=tensor([0.1421, 0.0941, 0.1804, 0.1148, 0.1081, 0.0926, 0.1470, 0.1891], device='cuda:6'), in_proj_covar=tensor([0.0116, 0.0237, 0.0134, 0.0118, 0.0131, 0.0150, 0.0115, 0.0117], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-27 15:28:19,462 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.46 vs. limit=5.0 2023-04-27 15:28:20,894 INFO [finetune.py:976] (6/7) Epoch 21, batch 2400, loss[loss=0.163, simple_loss=0.2312, pruned_loss=0.04737, over 4871.00 frames. ], tot_loss[loss=0.1711, simple_loss=0.2425, pruned_loss=0.04991, over 956738.85 frames. ], batch size: 49, lr: 3.19e-03, grad_scale: 64.0 2023-04-27 15:28:20,983 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0911, 2.3956, 0.9372, 1.2197, 1.8576, 1.2647, 3.0323, 1.5137], device='cuda:6'), covar=tensor([0.0651, 0.0634, 0.0781, 0.1312, 0.0484, 0.0971, 0.0272, 0.0701], device='cuda:6'), in_proj_covar=tensor([0.0050, 0.0065, 0.0047, 0.0046, 0.0049, 0.0052, 0.0073, 0.0051], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:6') 2023-04-27 15:28:25,119 INFO [optim.py:369] (6/7) 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 15:28:34,744 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=116974.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:28:37,803 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=116979.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:28:54,619 INFO [finetune.py:976] (6/7) Epoch 21, batch 2450, loss[loss=0.1762, simple_loss=0.2426, pruned_loss=0.05486, over 4747.00 frames. ], tot_loss[loss=0.169, simple_loss=0.2393, pruned_loss=0.04931, over 953741.76 frames. ], batch size: 59, lr: 3.19e-03, grad_scale: 64.0 2023-04-27 15:29:10,689 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=117027.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:29:13,892 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.80 vs. limit=2.0 2023-04-27 15:29:28,082 INFO [finetune.py:976] (6/7) Epoch 21, batch 2500, loss[loss=0.2071, simple_loss=0.2755, pruned_loss=0.06938, over 4812.00 frames. ], tot_loss[loss=0.1699, simple_loss=0.2406, pruned_loss=0.04964, over 955881.26 frames. ], batch size: 41, lr: 3.19e-03, grad_scale: 64.0 2023-04-27 15:29:32,776 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=117060.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:29:33,283 INFO [optim.py:369] (6/7) 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 15:30:15,219 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.59 vs. limit=2.0 2023-04-27 15:30:16,238 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.1832, 1.6088, 1.9874, 2.1443, 1.9782, 1.5852, 1.0856, 1.6198], device='cuda:6'), covar=tensor([0.3048, 0.3165, 0.1693, 0.2145, 0.2497, 0.2560, 0.4394, 0.2029], device='cuda:6'), in_proj_covar=tensor([0.0291, 0.0244, 0.0225, 0.0312, 0.0218, 0.0231, 0.0227, 0.0183], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:6') 2023-04-27 15:30:29,030 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=117102.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:30:30,122 INFO [finetune.py:976] (6/7) Epoch 21, batch 2550, loss[loss=0.1988, simple_loss=0.2666, pruned_loss=0.06553, over 4754.00 frames. ], tot_loss[loss=0.172, simple_loss=0.2436, pruned_loss=0.05024, over 956990.01 frames. ], batch size: 28, lr: 3.19e-03, grad_scale: 64.0 2023-04-27 15:30:48,401 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.1878, 1.7778, 1.9955, 2.5101, 2.5120, 1.9654, 1.7075, 2.1564], device='cuda:6'), covar=tensor([0.0800, 0.1052, 0.0689, 0.0519, 0.0560, 0.0858, 0.0783, 0.0551], device='cuda:6'), in_proj_covar=tensor([0.0187, 0.0201, 0.0184, 0.0173, 0.0177, 0.0180, 0.0152, 0.0178], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 15:30:59,164 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=3.56 vs. limit=5.0 2023-04-27 15:31:08,023 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.4316, 1.0860, 0.4196, 1.1266, 1.1455, 1.3113, 1.1810, 1.2477], device='cuda:6'), covar=tensor([0.0522, 0.0405, 0.0386, 0.0578, 0.0282, 0.0502, 0.0518, 0.0585], device='cuda:6'), in_proj_covar=tensor([0.0028, 0.0024, 0.0022, 0.0029, 0.0019, 0.0028, 0.0028, 0.0028], device='cuda:6'), out_proj_covar=tensor([0.0051, 0.0045, 0.0038, 0.0052, 0.0038, 0.0050, 0.0050, 0.0051], device='cuda:6') 2023-04-27 15:31:10,649 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2023-04-27 15:31:33,712 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 21, batch 2600, loss[loss=0.1699, simple_loss=0.2436, pruned_loss=0.04812, over 4924.00 frames. ], tot_loss[loss=0.1728, simple_loss=0.2447, pruned_loss=0.05047, over 957339.57 frames. ], batch size: 33, lr: 3.19e-03, grad_scale: 64.0 2023-04-27 15:31:44,427 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=117158.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:31:52,556 INFO [optim.py:369] (6/7) 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 15:31:52,676 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=117161.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:32:15,498 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7754, 1.3191, 1.9117, 2.2401, 1.8475, 1.7413, 1.8199, 1.8109], device='cuda:6'), covar=tensor([0.4733, 0.7063, 0.6800, 0.5800, 0.5955, 0.8165, 0.8657, 0.8939], device='cuda:6'), in_proj_covar=tensor([0.0427, 0.0410, 0.0505, 0.0503, 0.0456, 0.0485, 0.0493, 0.0498], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 15:32:20,830 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7247, 1.5120, 2.0107, 2.0104, 1.5842, 1.4150, 1.5789, 0.9780], device='cuda:6'), covar=tensor([0.0577, 0.0674, 0.0361, 0.0592, 0.0770, 0.1194, 0.0701, 0.0777], device='cuda:6'), in_proj_covar=tensor([0.0068, 0.0067, 0.0066, 0.0067, 0.0074, 0.0095, 0.0072, 0.0064], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-27 15:32:24,647 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.2278, 1.5815, 1.4235, 1.7447, 1.6952, 1.9362, 1.3978, 3.3742], device='cuda:6'), covar=tensor([0.0582, 0.0778, 0.0771, 0.1141, 0.0604, 0.0528, 0.0751, 0.0142], device='cuda:6'), in_proj_covar=tensor([0.0037, 0.0038, 0.0039, 0.0043, 0.0040, 0.0038, 0.0038, 0.0056], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:6') 2023-04-27 15:32:27,550 INFO [finetune.py:976] (6/7) Epoch 21, batch 2650, loss[loss=0.1682, simple_loss=0.2439, pruned_loss=0.04628, over 4907.00 frames. ], tot_loss[loss=0.1732, simple_loss=0.2453, pruned_loss=0.05059, over 955771.02 frames. ], batch size: 37, lr: 3.19e-03, grad_scale: 64.0 2023-04-27 15:32:30,682 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=117209.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:32:39,174 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=117221.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:33:01,273 INFO [finetune.py:976] (6/7) Epoch 21, batch 2700, loss[loss=0.1457, simple_loss=0.2154, pruned_loss=0.03798, over 4782.00 frames. ], tot_loss[loss=0.1727, simple_loss=0.2452, pruned_loss=0.05007, over 958306.76 frames. ], batch size: 29, lr: 3.19e-03, grad_scale: 64.0 2023-04-27 15:33:06,002 INFO [optim.py:369] (6/7) 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 15:33:15,090 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=117274.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:33:16,962 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5378, 2.5169, 2.0274, 2.3137, 2.5421, 2.0027, 3.2515, 1.8762], device='cuda:6'), covar=tensor([0.4079, 0.2375, 0.4494, 0.3701, 0.1989, 0.3090, 0.2045, 0.4840], device='cuda:6'), in_proj_covar=tensor([0.0342, 0.0349, 0.0424, 0.0353, 0.0382, 0.0375, 0.0370, 0.0419], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 15:33:19,983 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 21, batch 2750, loss[loss=0.1731, simple_loss=0.2418, pruned_loss=0.05224, over 4895.00 frames. ], tot_loss[loss=0.1706, simple_loss=0.2427, pruned_loss=0.04929, over 956661.52 frames. ], batch size: 32, lr: 3.19e-03, grad_scale: 64.0 2023-04-27 15:33:39,449 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.2061, 2.6600, 2.2358, 2.5296, 1.9407, 2.2124, 2.2700, 1.7981], device='cuda:6'), covar=tensor([0.1743, 0.1227, 0.0687, 0.1226, 0.2824, 0.1155, 0.1688, 0.2193], device='cuda:6'), in_proj_covar=tensor([0.0284, 0.0297, 0.0214, 0.0275, 0.0307, 0.0251, 0.0244, 0.0259], device='cuda:6'), out_proj_covar=tensor([1.1370e-04, 1.1760e-04, 8.4839e-05, 1.0885e-04, 1.2461e-04, 9.9397e-05, 9.8485e-05, 1.0282e-04], device='cuda:6') 2023-04-27 15:33:40,050 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7962, 1.9283, 0.8623, 1.4747, 1.7984, 1.6925, 1.5823, 1.6267], device='cuda:6'), covar=tensor([0.0467, 0.0324, 0.0333, 0.0512, 0.0261, 0.0469, 0.0452, 0.0537], device='cuda:6'), in_proj_covar=tensor([0.0028, 0.0024, 0.0022, 0.0028, 0.0019, 0.0028, 0.0028, 0.0028], device='cuda:6'), out_proj_covar=tensor([0.0051, 0.0045, 0.0038, 0.0052, 0.0038, 0.0050, 0.0050, 0.0051], device='cuda:6') 2023-04-27 15:33:47,709 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=117322.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:33:52,063 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 21, batch 2800, loss[loss=0.154, simple_loss=0.2234, pruned_loss=0.04223, over 4824.00 frames. ], tot_loss[loss=0.1693, simple_loss=0.2406, pruned_loss=0.04898, over 957162.98 frames. ], batch size: 40, lr: 3.19e-03, grad_scale: 64.0 2023-04-27 15:34:12,938 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=117360.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:34:13,415 INFO [optim.py:369] (6/7) 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 15:34:31,553 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=117390.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:34:40,561 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=3.51 vs. limit=5.0 2023-04-27 15:34:41,634 INFO [finetune.py:976] (6/7) Epoch 21, batch 2850, loss[loss=0.156, simple_loss=0.2261, pruned_loss=0.04293, over 4362.00 frames. ], tot_loss[loss=0.1686, simple_loss=0.2395, pruned_loss=0.04884, over 956358.00 frames. ], batch size: 19, lr: 3.19e-03, grad_scale: 64.0 2023-04-27 15:34:44,560 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=117408.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:35:06,914 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=3.59 vs. limit=5.0 2023-04-27 15:35:11,863 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7640, 2.1423, 1.7341, 2.0918, 1.5063, 1.8071, 1.8348, 1.3932], device='cuda:6'), covar=tensor([0.1931, 0.1143, 0.0860, 0.1121, 0.3230, 0.1136, 0.1700, 0.2509], device='cuda:6'), in_proj_covar=tensor([0.0287, 0.0300, 0.0217, 0.0278, 0.0311, 0.0254, 0.0247, 0.0262], device='cuda:6'), out_proj_covar=tensor([1.1492e-04, 1.1884e-04, 8.5922e-05, 1.0997e-04, 1.2611e-04, 1.0054e-04, 9.9857e-05, 1.0418e-04], device='cuda:6') 2023-04-27 15:35:14,767 INFO [finetune.py:976] (6/7) Epoch 21, batch 2900, loss[loss=0.154, simple_loss=0.2416, pruned_loss=0.03319, over 4794.00 frames. ], tot_loss[loss=0.1718, simple_loss=0.2429, pruned_loss=0.0503, over 951853.69 frames. ], batch size: 54, lr: 3.19e-03, grad_scale: 64.0 2023-04-27 15:35:17,827 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=117458.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:35:19,566 INFO [optim.py:369] (6/7) 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 15:35:52,473 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6344, 1.6024, 1.9576, 2.0099, 1.6115, 1.2399, 1.7017, 1.0335], device='cuda:6'), covar=tensor([0.0558, 0.0507, 0.0418, 0.0690, 0.0667, 0.1008, 0.0589, 0.0719], device='cuda:6'), in_proj_covar=tensor([0.0068, 0.0068, 0.0066, 0.0067, 0.0074, 0.0095, 0.0072, 0.0064], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-27 15:36:03,551 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.2694, 1.8145, 2.0992, 2.6831, 2.2531, 1.7907, 1.8658, 2.0417], device='cuda:6'), covar=tensor([0.2657, 0.2717, 0.1522, 0.2040, 0.2295, 0.2241, 0.3664, 0.2022], device='cuda:6'), in_proj_covar=tensor([0.0291, 0.0244, 0.0226, 0.0313, 0.0218, 0.0231, 0.0227, 0.0183], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:6') 2023-04-27 15:36:16,932 INFO [finetune.py:976] (6/7) Epoch 21, batch 2950, loss[loss=0.1964, simple_loss=0.2609, pruned_loss=0.0659, over 4818.00 frames. ], tot_loss[loss=0.1731, simple_loss=0.245, pruned_loss=0.05058, over 950733.19 frames. ], batch size: 33, lr: 3.19e-03, grad_scale: 64.0 2023-04-27 15:36:18,222 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=117506.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:37:23,420 INFO [finetune.py:976] (6/7) Epoch 21, batch 3000, loss[loss=0.1511, simple_loss=0.2289, pruned_loss=0.03664, over 4778.00 frames. ], tot_loss[loss=0.1733, simple_loss=0.2456, pruned_loss=0.0505, over 952828.76 frames. ], batch size: 28, lr: 3.19e-03, grad_scale: 64.0 2023-04-27 15:37:23,420 INFO [finetune.py:1001] (6/7) Computing validation loss 2023-04-27 15:37:33,896 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0950, 2.4982, 1.1619, 1.4204, 2.0019, 1.4652, 2.8503, 1.6644], device='cuda:6'), covar=tensor([0.0515, 0.0576, 0.0677, 0.1004, 0.0345, 0.0734, 0.0235, 0.0515], device='cuda:6'), in_proj_covar=tensor([0.0050, 0.0065, 0.0047, 0.0046, 0.0049, 0.0052, 0.0073, 0.0051], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:6') 2023-04-27 15:37:43,672 INFO [finetune.py:1010] (6/7) Epoch 21, validation: loss=0.1531, simple_loss=0.2228, pruned_loss=0.04164, over 2265189.00 frames. 2023-04-27 15:37:43,672 INFO [finetune.py:1011] (6/7) Maximum memory allocated so far is 6365MB 2023-04-27 15:37:54,336 INFO [optim.py:369] (6/7) 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 15:38:15,523 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=117577.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:38:18,802 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2023-04-27 15:38:49,964 INFO [finetune.py:976] (6/7) Epoch 21, batch 3050, loss[loss=0.1473, simple_loss=0.2408, pruned_loss=0.02684, over 4717.00 frames. ], tot_loss[loss=0.1748, simple_loss=0.2473, pruned_loss=0.05111, over 954830.61 frames. ], batch size: 59, lr: 3.19e-03, grad_scale: 64.0 2023-04-27 15:39:22,526 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7597, 1.7089, 2.0006, 2.2462, 1.7328, 1.4692, 1.7577, 0.9308], device='cuda:6'), covar=tensor([0.0524, 0.0692, 0.0477, 0.0733, 0.0652, 0.1031, 0.0649, 0.0680], device='cuda:6'), in_proj_covar=tensor([0.0068, 0.0068, 0.0066, 0.0067, 0.0074, 0.0095, 0.0072, 0.0065], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-27 15:39:23,414 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.13 vs. limit=2.0 2023-04-27 15:39:27,901 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.4673, 3.1797, 2.6366, 2.9334, 2.1804, 2.7792, 2.7697, 2.2749], device='cuda:6'), covar=tensor([0.2195, 0.1169, 0.0771, 0.1224, 0.3223, 0.1088, 0.2088, 0.2709], device='cuda:6'), in_proj_covar=tensor([0.0285, 0.0299, 0.0215, 0.0277, 0.0309, 0.0252, 0.0246, 0.0262], device='cuda:6'), out_proj_covar=tensor([1.1408e-04, 1.1826e-04, 8.5032e-05, 1.0941e-04, 1.2542e-04, 9.9750e-05, 9.9229e-05, 1.0389e-04], device='cuda:6') 2023-04-27 15:39:28,402 INFO [finetune.py:976] (6/7) Epoch 21, batch 3100, loss[loss=0.1518, simple_loss=0.2129, pruned_loss=0.04535, over 4819.00 frames. ], tot_loss[loss=0.1727, simple_loss=0.2451, pruned_loss=0.05013, over 954807.50 frames. ], batch size: 30, lr: 3.19e-03, grad_scale: 64.0 2023-04-27 15:39:33,614 INFO [optim.py:369] (6/7) 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 15:39:49,476 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=117685.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:39:58,482 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6691, 2.0711, 1.7841, 2.0858, 1.5292, 1.7926, 1.7247, 1.3707], device='cuda:6'), covar=tensor([0.1737, 0.1080, 0.0848, 0.0973, 0.3799, 0.1167, 0.1853, 0.2418], device='cuda:6'), in_proj_covar=tensor([0.0284, 0.0297, 0.0214, 0.0276, 0.0308, 0.0251, 0.0245, 0.0260], device='cuda:6'), out_proj_covar=tensor([1.1380e-04, 1.1778e-04, 8.4718e-05, 1.0906e-04, 1.2497e-04, 9.9551e-05, 9.8919e-05, 1.0339e-04], device='cuda:6') 2023-04-27 15:40:01,888 INFO [finetune.py:976] (6/7) Epoch 21, batch 3150, loss[loss=0.1187, simple_loss=0.2037, pruned_loss=0.01681, over 4822.00 frames. ], tot_loss[loss=0.1708, simple_loss=0.2426, pruned_loss=0.04949, over 956610.72 frames. ], batch size: 51, lr: 3.19e-03, grad_scale: 64.0 2023-04-27 15:40:09,674 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([0.7478, 1.7779, 1.7258, 1.4786, 1.9269, 1.5652, 2.2921, 1.4589], device='cuda:6'), covar=tensor([0.3367, 0.1604, 0.4896, 0.2336, 0.1294, 0.2189, 0.1307, 0.4903], device='cuda:6'), in_proj_covar=tensor([0.0342, 0.0348, 0.0424, 0.0353, 0.0381, 0.0375, 0.0369, 0.0420], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 15:40:27,795 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9455, 2.1876, 1.8857, 2.2318, 1.5179, 1.8711, 1.9639, 1.5001], device='cuda:6'), covar=tensor([0.1736, 0.1288, 0.0831, 0.1090, 0.3579, 0.1202, 0.1615, 0.2420], device='cuda:6'), in_proj_covar=tensor([0.0286, 0.0299, 0.0215, 0.0277, 0.0309, 0.0252, 0.0246, 0.0262], device='cuda:6'), out_proj_covar=tensor([1.1434e-04, 1.1846e-04, 8.5023e-05, 1.0957e-04, 1.2547e-04, 9.9881e-05, 9.9343e-05, 1.0391e-04], device='cuda:6') 2023-04-27 15:40:34,886 INFO [finetune.py:976] (6/7) Epoch 21, batch 3200, loss[loss=0.1706, simple_loss=0.244, pruned_loss=0.04863, over 4811.00 frames. ], tot_loss[loss=0.1686, simple_loss=0.2397, pruned_loss=0.0487, over 954972.83 frames. ], batch size: 38, lr: 3.19e-03, grad_scale: 32.0 2023-04-27 15:40:40,631 INFO [optim.py:369] (6/7) 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 15:41:09,128 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.4688, 1.9661, 2.4286, 2.8698, 2.3823, 1.9103, 1.9301, 2.3952], device='cuda:6'), covar=tensor([0.3114, 0.3166, 0.1650, 0.2577, 0.2682, 0.2734, 0.3634, 0.1953], device='cuda:6'), in_proj_covar=tensor([0.0292, 0.0244, 0.0226, 0.0314, 0.0219, 0.0232, 0.0228, 0.0184], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:6') 2023-04-27 15:41:18,698 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.3277, 1.6383, 1.5130, 1.8720, 1.8245, 2.0801, 1.4803, 3.8797], device='cuda:6'), covar=tensor([0.0576, 0.0777, 0.0780, 0.1097, 0.0581, 0.0646, 0.0753, 0.0131], device='cuda:6'), in_proj_covar=tensor([0.0037, 0.0038, 0.0039, 0.0043, 0.0040, 0.0038, 0.0038, 0.0056], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:6') 2023-04-27 15:41:31,205 INFO [finetune.py:976] (6/7) Epoch 21, batch 3250, loss[loss=0.1742, simple_loss=0.2527, pruned_loss=0.04783, over 4829.00 frames. ], tot_loss[loss=0.1688, simple_loss=0.2398, pruned_loss=0.04893, over 955758.51 frames. ], batch size: 39, lr: 3.19e-03, grad_scale: 32.0 2023-04-27 15:42:09,334 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-04-27 15:42:33,314 INFO [finetune.py:976] (6/7) Epoch 21, batch 3300, loss[loss=0.2297, simple_loss=0.3056, pruned_loss=0.07689, over 4849.00 frames. ], tot_loss[loss=0.171, simple_loss=0.2434, pruned_loss=0.04926, over 957157.64 frames. ], batch size: 44, lr: 3.19e-03, grad_scale: 32.0 2023-04-27 15:42:45,022 INFO [optim.py:369] (6/7) 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 15:43:01,374 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=117877.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:43:18,019 INFO [finetune.py:976] (6/7) Epoch 21, batch 3350, loss[loss=0.1758, simple_loss=0.2455, pruned_loss=0.05301, over 4761.00 frames. ], tot_loss[loss=0.1729, simple_loss=0.2457, pruned_loss=0.05007, over 955736.08 frames. ], batch size: 27, lr: 3.19e-03, grad_scale: 32.0 2023-04-27 15:43:38,349 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=117925.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:44:02,872 INFO [finetune.py:976] (6/7) Epoch 21, batch 3400, loss[loss=0.1872, simple_loss=0.2673, pruned_loss=0.05354, over 4858.00 frames. ], tot_loss[loss=0.1738, simple_loss=0.2461, pruned_loss=0.0508, over 954581.05 frames. ], batch size: 44, lr: 3.19e-03, grad_scale: 32.0 2023-04-27 15:44:13,574 INFO [optim.py:369] (6/7) 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 15:44:26,783 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-04-27 15:44:35,013 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-04-27 15:44:44,682 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 21, batch 3450, loss[loss=0.1809, simple_loss=0.2408, pruned_loss=0.06047, over 4785.00 frames. ], tot_loss[loss=0.1731, simple_loss=0.2453, pruned_loss=0.05044, over 953933.37 frames. ], batch size: 29, lr: 3.18e-03, grad_scale: 32.0 2023-04-27 15:45:18,120 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=118033.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:45:22,445 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=118040.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:45:31,329 INFO [finetune.py:976] (6/7) Epoch 21, batch 3500, loss[loss=0.1642, simple_loss=0.231, pruned_loss=0.04876, over 4812.00 frames. ], tot_loss[loss=0.1714, simple_loss=0.2428, pruned_loss=0.05, over 954368.44 frames. ], batch size: 51, lr: 3.18e-03, grad_scale: 32.0 2023-04-27 15:45:36,184 INFO [optim.py:369] (6/7) 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 15:46:03,423 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 21, batch 3550, loss[loss=0.1441, simple_loss=0.2212, pruned_loss=0.03354, over 4794.00 frames. ], tot_loss[loss=0.1701, simple_loss=0.2408, pruned_loss=0.0497, over 954064.59 frames. ], batch size: 29, lr: 3.18e-03, grad_scale: 32.0 2023-04-27 15:46:19,641 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.0861, 2.0909, 2.0891, 1.7705, 2.2523, 1.8145, 2.6604, 1.8019], device='cuda:6'), covar=tensor([0.2956, 0.1596, 0.3610, 0.2228, 0.1162, 0.2123, 0.1294, 0.3703], device='cuda:6'), in_proj_covar=tensor([0.0339, 0.0348, 0.0424, 0.0352, 0.0379, 0.0374, 0.0368, 0.0418], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 15:46:39,043 INFO [finetune.py:976] (6/7) Epoch 21, batch 3600, loss[loss=0.1896, simple_loss=0.2536, pruned_loss=0.06281, over 4823.00 frames. ], tot_loss[loss=0.1698, simple_loss=0.2395, pruned_loss=0.05004, over 954851.77 frames. ], batch size: 41, lr: 3.18e-03, grad_scale: 32.0 2023-04-27 15:46:43,924 INFO [optim.py:369] (6/7) 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 15:46:44,736 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.57 vs. limit=2.0 2023-04-27 15:46:47,635 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=118168.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:47:39,452 INFO [finetune.py:976] (6/7) Epoch 21, batch 3650, loss[loss=0.2055, simple_loss=0.2799, pruned_loss=0.06555, over 4805.00 frames. ], tot_loss[loss=0.1718, simple_loss=0.242, pruned_loss=0.05074, over 955040.63 frames. ], batch size: 41, lr: 3.18e-03, grad_scale: 32.0 2023-04-27 15:48:00,470 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-04-27 15:48:01,536 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.7051, 2.0008, 2.1381, 2.1897, 2.1491, 2.2598, 2.2063, 2.2170], device='cuda:6'), covar=tensor([0.3569, 0.5506, 0.4909, 0.4622, 0.5162, 0.6460, 0.5351, 0.4829], device='cuda:6'), in_proj_covar=tensor([0.0338, 0.0374, 0.0323, 0.0336, 0.0348, 0.0394, 0.0357, 0.0329], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 15:48:09,962 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=118229.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 15:48:43,769 INFO [finetune.py:976] (6/7) Epoch 21, batch 3700, loss[loss=0.1703, simple_loss=0.2586, pruned_loss=0.04098, over 4820.00 frames. ], tot_loss[loss=0.1727, simple_loss=0.2438, pruned_loss=0.05079, over 951345.21 frames. ], batch size: 40, lr: 3.18e-03, grad_scale: 32.0 2023-04-27 15:48:54,003 INFO [optim.py:369] (6/7) 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 15:48:54,228 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.52 vs. limit=2.0 2023-04-27 15:49:02,696 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=118269.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:49:10,616 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5101, 1.4047, 1.8513, 1.8252, 1.3863, 1.2656, 1.4905, 0.8953], device='cuda:6'), covar=tensor([0.0529, 0.0779, 0.0383, 0.0564, 0.0757, 0.1100, 0.0676, 0.0649], device='cuda:6'), in_proj_covar=tensor([0.0069, 0.0068, 0.0067, 0.0067, 0.0075, 0.0096, 0.0073, 0.0065], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-27 15:49:16,491 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=118289.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:49:27,412 INFO [finetune.py:976] (6/7) Epoch 21, batch 3750, loss[loss=0.1918, simple_loss=0.2598, pruned_loss=0.06189, over 4824.00 frames. ], tot_loss[loss=0.1731, simple_loss=0.245, pruned_loss=0.05061, over 951248.09 frames. ], batch size: 33, lr: 3.18e-03, grad_scale: 32.0 2023-04-27 15:49:43,207 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=118329.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:49:43,810 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=118330.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:50:10,283 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 21, batch 3800, loss[loss=0.1751, simple_loss=0.2591, pruned_loss=0.04551, over 4817.00 frames. ], tot_loss[loss=0.1734, simple_loss=0.2458, pruned_loss=0.05047, over 951463.24 frames. ], batch size: 39, lr: 3.18e-03, grad_scale: 32.0 2023-04-27 15:50:23,127 INFO [optim.py:369] (6/7) 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 15:50:33,543 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-04-27 15:50:52,605 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=118390.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:50:56,202 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=118396.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:51:02,179 INFO [finetune.py:976] (6/7) Epoch 21, batch 3850, loss[loss=0.1742, simple_loss=0.2463, pruned_loss=0.05104, over 4905.00 frames. ], tot_loss[loss=0.1732, simple_loss=0.2456, pruned_loss=0.05039, over 953072.44 frames. ], batch size: 43, lr: 3.18e-03, grad_scale: 32.0 2023-04-27 15:51:35,150 INFO [finetune.py:976] (6/7) Epoch 21, batch 3900, loss[loss=0.1482, simple_loss=0.2161, pruned_loss=0.04016, over 4770.00 frames. ], tot_loss[loss=0.1701, simple_loss=0.242, pruned_loss=0.0491, over 953139.79 frames. ], batch size: 23, lr: 3.18e-03, grad_scale: 32.0 2023-04-27 15:51:40,430 INFO [optim.py:369] (6/7) 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 15:51:56,332 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.3763, 1.2578, 1.5588, 1.5450, 1.3142, 1.1953, 1.2948, 0.8152], device='cuda:6'), covar=tensor([0.0545, 0.0599, 0.0402, 0.0499, 0.0666, 0.0996, 0.0519, 0.0534], device='cuda:6'), in_proj_covar=tensor([0.0068, 0.0068, 0.0066, 0.0067, 0.0074, 0.0095, 0.0072, 0.0064], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-27 15:52:07,427 INFO [finetune.py:976] (6/7) Epoch 21, batch 3950, loss[loss=0.1878, simple_loss=0.2553, pruned_loss=0.06009, over 4915.00 frames. ], tot_loss[loss=0.1686, simple_loss=0.2397, pruned_loss=0.04875, over 956512.29 frames. ], batch size: 37, lr: 3.18e-03, grad_scale: 32.0 2023-04-27 15:52:19,355 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8106, 1.5412, 1.4302, 1.6415, 2.0277, 1.6721, 1.4219, 1.3614], device='cuda:6'), covar=tensor([0.1525, 0.1402, 0.1813, 0.1344, 0.0819, 0.1667, 0.1905, 0.2302], device='cuda:6'), in_proj_covar=tensor([0.0309, 0.0310, 0.0350, 0.0288, 0.0325, 0.0307, 0.0300, 0.0369], device='cuda:6'), out_proj_covar=tensor([6.3456e-05, 6.4114e-05, 7.3826e-05, 5.8088e-05, 6.7056e-05, 6.4355e-05, 6.2949e-05, 7.8264e-05], device='cuda:6') 2023-04-27 15:52:21,111 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=118524.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 15:52:21,134 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9783, 2.4076, 0.8842, 1.1672, 1.7274, 1.1221, 3.2176, 1.4294], device='cuda:6'), covar=tensor([0.0849, 0.0762, 0.0979, 0.1667, 0.0681, 0.1416, 0.0422, 0.0978], device='cuda:6'), in_proj_covar=tensor([0.0050, 0.0065, 0.0047, 0.0046, 0.0049, 0.0051, 0.0073, 0.0051], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:6') 2023-04-27 15:52:40,872 INFO [finetune.py:976] (6/7) Epoch 21, batch 4000, loss[loss=0.2195, simple_loss=0.2979, pruned_loss=0.07055, over 4855.00 frames. ], tot_loss[loss=0.1672, simple_loss=0.238, pruned_loss=0.04818, over 954638.79 frames. ], batch size: 44, lr: 3.18e-03, grad_scale: 32.0 2023-04-27 15:52:47,269 INFO [optim.py:369] (6/7) 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 15:52:49,738 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.70 vs. limit=2.0 2023-04-27 15:52:53,396 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8466, 2.0474, 1.7506, 1.6322, 1.3676, 1.3674, 1.7228, 1.3765], device='cuda:6'), covar=tensor([0.1471, 0.1342, 0.1295, 0.1524, 0.2021, 0.1745, 0.0908, 0.1807], device='cuda:6'), in_proj_covar=tensor([0.0198, 0.0212, 0.0170, 0.0205, 0.0200, 0.0185, 0.0157, 0.0186], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:6') 2023-04-27 15:53:30,868 INFO [finetune.py:976] (6/7) Epoch 21, batch 4050, loss[loss=0.1897, simple_loss=0.2697, pruned_loss=0.05488, over 4814.00 frames. ], tot_loss[loss=0.1705, simple_loss=0.2415, pruned_loss=0.04974, over 954728.67 frames. ], batch size: 39, lr: 3.18e-03, grad_scale: 32.0 2023-04-27 15:53:56,789 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=118625.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:54:06,613 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=118631.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:54:19,890 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9994, 2.4852, 2.1251, 2.3485, 1.7119, 2.1718, 2.0811, 1.5633], device='cuda:6'), covar=tensor([0.1873, 0.1045, 0.0773, 0.1180, 0.3066, 0.1016, 0.1739, 0.2469], device='cuda:6'), in_proj_covar=tensor([0.0284, 0.0300, 0.0215, 0.0277, 0.0311, 0.0252, 0.0246, 0.0263], device='cuda:6'), out_proj_covar=tensor([1.1385e-04, 1.1874e-04, 8.4972e-05, 1.0983e-04, 1.2598e-04, 9.9957e-05, 9.9079e-05, 1.0409e-04], device='cuda:6') 2023-04-27 15:54:26,640 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=118645.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:54:27,384 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=3.59 vs. limit=5.0 2023-04-27 15:54:33,134 INFO [finetune.py:976] (6/7) Epoch 21, batch 4100, loss[loss=0.1444, simple_loss=0.2185, pruned_loss=0.03513, over 4783.00 frames. ], tot_loss[loss=0.1728, simple_loss=0.2444, pruned_loss=0.05064, over 951752.68 frames. ], batch size: 26, lr: 3.18e-03, grad_scale: 32.0 2023-04-27 15:54:38,507 INFO [optim.py:369] (6/7) 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 15:54:39,747 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.2635, 1.2535, 3.7939, 3.4970, 3.3463, 3.6289, 3.6198, 3.3401], device='cuda:6'), covar=tensor([0.7421, 0.5863, 0.1216, 0.1980, 0.1322, 0.1822, 0.1755, 0.1648], device='cuda:6'), in_proj_covar=tensor([0.0313, 0.0309, 0.0408, 0.0408, 0.0350, 0.0412, 0.0314, 0.0370], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-27 15:54:54,352 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=118685.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:54:58,624 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=118692.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:55:00,960 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=118696.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:55:06,697 INFO [finetune.py:976] (6/7) Epoch 21, batch 4150, loss[loss=0.1615, simple_loss=0.237, pruned_loss=0.04301, over 4917.00 frames. ], tot_loss[loss=0.1742, simple_loss=0.246, pruned_loss=0.05116, over 952016.10 frames. ], batch size: 38, lr: 3.18e-03, grad_scale: 32.0 2023-04-27 15:55:37,041 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.4612, 2.5156, 1.9324, 2.0946, 2.5097, 2.0237, 3.3333, 1.7866], device='cuda:6'), covar=tensor([0.4125, 0.2406, 0.4900, 0.3828, 0.2002, 0.2971, 0.1785, 0.4740], device='cuda:6'), in_proj_covar=tensor([0.0340, 0.0348, 0.0424, 0.0353, 0.0379, 0.0374, 0.0369, 0.0417], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 15:55:56,015 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=118744.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:56:07,491 INFO [finetune.py:976] (6/7) Epoch 21, batch 4200, loss[loss=0.1672, simple_loss=0.2518, pruned_loss=0.04129, over 4879.00 frames. ], tot_loss[loss=0.1727, simple_loss=0.2455, pruned_loss=0.04995, over 953947.42 frames. ], batch size: 43, lr: 3.18e-03, grad_scale: 32.0 2023-04-27 15:56:19,457 INFO [optim.py:369] (6/7) 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 15:56:29,999 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.31 vs. limit=5.0 2023-04-27 15:56:47,380 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.11 vs. limit=2.0 2023-04-27 15:56:49,525 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8489, 1.3782, 5.2738, 4.9067, 4.5464, 5.0572, 4.6748, 4.6447], device='cuda:6'), covar=tensor([0.7183, 0.6010, 0.0815, 0.1527, 0.1064, 0.1048, 0.0933, 0.1442], device='cuda:6'), in_proj_covar=tensor([0.0313, 0.0308, 0.0408, 0.0407, 0.0349, 0.0411, 0.0313, 0.0369], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-27 15:56:58,126 INFO [finetune.py:976] (6/7) Epoch 21, batch 4250, loss[loss=0.1113, simple_loss=0.178, pruned_loss=0.02232, over 4226.00 frames. ], tot_loss[loss=0.1709, simple_loss=0.2434, pruned_loss=0.04916, over 953070.71 frames. ], batch size: 18, lr: 3.18e-03, grad_scale: 32.0 2023-04-27 15:57:13,412 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=118824.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:57:32,155 INFO [finetune.py:976] (6/7) Epoch 21, batch 4300, loss[loss=0.1387, simple_loss=0.2009, pruned_loss=0.03823, over 4209.00 frames. ], tot_loss[loss=0.1687, simple_loss=0.2405, pruned_loss=0.04848, over 952997.82 frames. ], batch size: 18, lr: 3.18e-03, grad_scale: 32.0 2023-04-27 15:57:37,516 INFO [optim.py:369] (6/7) 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 15:57:44,623 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=118872.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:58:06,090 INFO [finetune.py:976] (6/7) Epoch 21, batch 4350, loss[loss=0.1411, simple_loss=0.2027, pruned_loss=0.03976, over 4191.00 frames. ], tot_loss[loss=0.1661, simple_loss=0.2373, pruned_loss=0.04743, over 954134.07 frames. ], batch size: 18, lr: 3.18e-03, grad_scale: 32.0 2023-04-27 15:58:20,462 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=118925.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:58:39,669 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=118945.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:58:51,399 INFO [finetune.py:976] (6/7) Epoch 21, batch 4400, loss[loss=0.1991, simple_loss=0.2653, pruned_loss=0.06647, over 4888.00 frames. ], tot_loss[loss=0.1696, simple_loss=0.2407, pruned_loss=0.04929, over 954284.60 frames. ], batch size: 32, lr: 3.18e-03, grad_scale: 32.0 2023-04-27 15:58:59,648 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.1387, 1.4906, 1.3082, 1.7109, 1.6300, 1.9520, 1.3530, 3.4208], device='cuda:6'), covar=tensor([0.0574, 0.0765, 0.0764, 0.1130, 0.0569, 0.0530, 0.0701, 0.0167], device='cuda:6'), in_proj_covar=tensor([0.0037, 0.0038, 0.0039, 0.0043, 0.0040, 0.0038, 0.0038, 0.0056], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:6') 2023-04-27 15:59:00,773 INFO [optim.py:369] (6/7) 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 15:59:01,884 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.81 vs. limit=2.0 2023-04-27 15:59:20,464 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=118973.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:59:20,506 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.4823, 1.7404, 1.5770, 1.9563, 1.9384, 2.0783, 1.6205, 3.6596], device='cuda:6'), covar=tensor([0.0518, 0.0720, 0.0703, 0.1016, 0.0532, 0.0700, 0.0724, 0.0191], device='cuda:6'), in_proj_covar=tensor([0.0037, 0.0038, 0.0039, 0.0043, 0.0040, 0.0038, 0.0038, 0.0056], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:6') 2023-04-27 15:59:35,152 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=118985.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:59:36,319 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=118987.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:59:45,378 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=118993.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:59:47,309 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8186, 2.3225, 1.8955, 2.2012, 1.6359, 1.9051, 2.0209, 1.4538], device='cuda:6'), covar=tensor([0.1937, 0.1195, 0.0816, 0.1145, 0.3073, 0.1071, 0.1721, 0.2486], device='cuda:6'), in_proj_covar=tensor([0.0282, 0.0296, 0.0214, 0.0275, 0.0310, 0.0251, 0.0244, 0.0261], device='cuda:6'), out_proj_covar=tensor([1.1292e-04, 1.1741e-04, 8.4507e-05, 1.0893e-04, 1.2546e-04, 9.9575e-05, 9.8624e-05, 1.0338e-04], device='cuda:6') 2023-04-27 15:59:52,146 INFO [finetune.py:976] (6/7) Epoch 21, batch 4450, loss[loss=0.1878, simple_loss=0.2598, pruned_loss=0.05785, over 4735.00 frames. ], tot_loss[loss=0.1716, simple_loss=0.2433, pruned_loss=0.04994, over 954845.14 frames. ], batch size: 54, lr: 3.18e-03, grad_scale: 32.0 2023-04-27 16:00:12,481 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 21, batch 4500, loss[loss=0.1656, simple_loss=0.2327, pruned_loss=0.04925, over 4925.00 frames. ], tot_loss[loss=0.1729, simple_loss=0.2448, pruned_loss=0.05049, over 952855.88 frames. ], batch size: 38, lr: 3.18e-03, grad_scale: 32.0 2023-04-27 16:00:26,461 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.1550, 2.8083, 2.1914, 2.3881, 1.5109, 1.5371, 2.3492, 1.5302], device='cuda:6'), covar=tensor([0.1671, 0.1444, 0.1376, 0.1483, 0.2348, 0.2022, 0.0943, 0.2068], device='cuda:6'), in_proj_covar=tensor([0.0198, 0.0212, 0.0170, 0.0205, 0.0200, 0.0186, 0.0156, 0.0186], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:6') 2023-04-27 16:00:30,591 INFO [optim.py:369] (6/7) 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 16:00:40,044 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=119075.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:00:58,656 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=119093.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:01:15,261 INFO [finetune.py:976] (6/7) Epoch 21, batch 4550, loss[loss=0.1665, simple_loss=0.2399, pruned_loss=0.04653, over 4863.00 frames. ], tot_loss[loss=0.1743, simple_loss=0.2461, pruned_loss=0.05126, over 952218.85 frames. ], batch size: 31, lr: 3.18e-03, grad_scale: 32.0 2023-04-27 16:01:42,301 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-04-27 16:01:59,495 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=119136.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:02:16,874 INFO [finetune.py:976] (6/7) Epoch 21, batch 4600, loss[loss=0.2007, simple_loss=0.2372, pruned_loss=0.08211, over 4023.00 frames. ], tot_loss[loss=0.1733, simple_loss=0.2452, pruned_loss=0.0507, over 952264.33 frames. ], batch size: 17, lr: 3.18e-03, grad_scale: 32.0 2023-04-27 16:02:16,994 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=119154.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 16:02:20,086 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.8776, 2.5182, 2.8315, 3.2059, 3.1416, 2.7215, 2.1842, 2.9805], device='cuda:6'), covar=tensor([0.0794, 0.0904, 0.0567, 0.0589, 0.0501, 0.0778, 0.0765, 0.0510], device='cuda:6'), in_proj_covar=tensor([0.0188, 0.0203, 0.0185, 0.0175, 0.0179, 0.0181, 0.0153, 0.0181], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 16:02:21,815 INFO [optim.py:369] (6/7) 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 16:02:24,365 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9407, 1.4677, 4.9044, 4.6360, 4.2353, 4.5920, 4.3565, 4.3487], device='cuda:6'), covar=tensor([0.6741, 0.5493, 0.1063, 0.1737, 0.1089, 0.1466, 0.1490, 0.1500], device='cuda:6'), in_proj_covar=tensor([0.0310, 0.0307, 0.0405, 0.0405, 0.0347, 0.0409, 0.0311, 0.0367], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-27 16:02:31,508 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.2948, 2.9624, 0.8270, 1.7005, 1.7023, 2.1908, 1.7064, 0.9833], device='cuda:6'), covar=tensor([0.1394, 0.0937, 0.1930, 0.1189, 0.1111, 0.0951, 0.1520, 0.1873], device='cuda:6'), in_proj_covar=tensor([0.0117, 0.0239, 0.0136, 0.0119, 0.0133, 0.0152, 0.0115, 0.0118], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-27 16:02:41,994 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.4810, 2.4856, 1.9211, 2.1309, 2.5840, 1.9552, 3.2537, 1.8047], device='cuda:6'), covar=tensor([0.3801, 0.2255, 0.4569, 0.3854, 0.1809, 0.2934, 0.1785, 0.4495], device='cuda:6'), in_proj_covar=tensor([0.0343, 0.0352, 0.0427, 0.0355, 0.0383, 0.0376, 0.0371, 0.0421], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 16:02:50,903 INFO [finetune.py:976] (6/7) Epoch 21, batch 4650, loss[loss=0.1569, simple_loss=0.2231, pruned_loss=0.04536, over 4924.00 frames. ], tot_loss[loss=0.1711, simple_loss=0.2422, pruned_loss=0.05005, over 953562.77 frames. ], batch size: 38, lr: 3.18e-03, grad_scale: 32.0 2023-04-27 16:03:11,317 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 2023-04-27 16:03:24,688 INFO [finetune.py:976] (6/7) Epoch 21, batch 4700, loss[loss=0.1354, simple_loss=0.2056, pruned_loss=0.03257, over 4790.00 frames. ], tot_loss[loss=0.1689, simple_loss=0.2394, pruned_loss=0.04919, over 954585.61 frames. ], batch size: 29, lr: 3.18e-03, grad_scale: 32.0 2023-04-27 16:03:29,615 INFO [optim.py:369] (6/7) 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 16:03:33,429 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.2301, 2.7027, 2.3189, 2.5614, 1.8659, 2.2812, 2.3011, 1.7137], device='cuda:6'), covar=tensor([0.1872, 0.1159, 0.0750, 0.1204, 0.3206, 0.1252, 0.1919, 0.2688], device='cuda:6'), in_proj_covar=tensor([0.0282, 0.0295, 0.0212, 0.0274, 0.0308, 0.0251, 0.0245, 0.0261], device='cuda:6'), out_proj_covar=tensor([1.1282e-04, 1.1692e-04, 8.3952e-05, 1.0837e-04, 1.2499e-04, 9.9443e-05, 9.8635e-05, 1.0338e-04], device='cuda:6') 2023-04-27 16:03:33,770 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2023-04-27 16:03:45,980 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=119287.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:03:58,706 INFO [finetune.py:976] (6/7) Epoch 21, batch 4750, loss[loss=0.1326, simple_loss=0.1992, pruned_loss=0.03294, over 4742.00 frames. ], tot_loss[loss=0.1675, simple_loss=0.2378, pruned_loss=0.04861, over 955985.10 frames. ], batch size: 27, lr: 3.17e-03, grad_scale: 32.0 2023-04-27 16:04:05,931 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.4235, 1.8794, 2.2913, 2.8973, 2.3734, 1.8236, 1.8327, 2.1416], device='cuda:6'), covar=tensor([0.3053, 0.2920, 0.1477, 0.2219, 0.2562, 0.2411, 0.3629, 0.2042], device='cuda:6'), in_proj_covar=tensor([0.0295, 0.0246, 0.0229, 0.0317, 0.0221, 0.0234, 0.0230, 0.0185], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:6') 2023-04-27 16:04:08,394 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.4397, 3.5306, 0.9609, 1.8331, 1.9871, 2.4519, 1.8463, 1.0652], device='cuda:6'), covar=tensor([0.1560, 0.0989, 0.2150, 0.1322, 0.1179, 0.1141, 0.1774, 0.1995], device='cuda:6'), in_proj_covar=tensor([0.0117, 0.0239, 0.0136, 0.0119, 0.0133, 0.0151, 0.0115, 0.0118], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-27 16:04:39,497 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 21, batch 4800, loss[loss=0.2105, simple_loss=0.2856, pruned_loss=0.06765, over 4923.00 frames. ], tot_loss[loss=0.1697, simple_loss=0.2404, pruned_loss=0.04947, over 955414.72 frames. ], batch size: 38, lr: 3.17e-03, grad_scale: 32.0 2023-04-27 16:04:59,372 INFO [optim.py:369] (6/7) 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 16:05:26,792 INFO [finetune.py:976] (6/7) Epoch 21, batch 4850, loss[loss=0.1446, simple_loss=0.2197, pruned_loss=0.03478, over 4879.00 frames. ], tot_loss[loss=0.1731, simple_loss=0.2446, pruned_loss=0.05082, over 957109.25 frames. ], batch size: 32, lr: 3.17e-03, grad_scale: 32.0 2023-04-27 16:05:36,749 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.38 vs. limit=5.0 2023-04-27 16:05:41,871 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.2914, 1.7052, 2.1549, 2.6815, 2.1566, 1.6740, 1.4416, 1.9449], device='cuda:6'), covar=tensor([0.3387, 0.3327, 0.1755, 0.2281, 0.2566, 0.2742, 0.4151, 0.2121], device='cuda:6'), in_proj_covar=tensor([0.0293, 0.0245, 0.0227, 0.0315, 0.0219, 0.0233, 0.0228, 0.0184], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:6') 2023-04-27 16:05:43,556 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=119431.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:05:55,524 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=119449.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 16:05:58,993 INFO [finetune.py:976] (6/7) Epoch 21, batch 4900, loss[loss=0.209, simple_loss=0.2797, pruned_loss=0.06908, over 4859.00 frames. ], tot_loss[loss=0.1731, simple_loss=0.2453, pruned_loss=0.05048, over 957226.83 frames. ], batch size: 34, lr: 3.17e-03, grad_scale: 32.0 2023-04-27 16:06:04,798 INFO [optim.py:369] (6/7) 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 16:06:35,842 INFO [finetune.py:976] (6/7) Epoch 21, batch 4950, loss[loss=0.164, simple_loss=0.2197, pruned_loss=0.05414, over 4254.00 frames. ], tot_loss[loss=0.1737, simple_loss=0.2465, pruned_loss=0.05049, over 956666.73 frames. ], batch size: 18, lr: 3.17e-03, grad_scale: 16.0 2023-04-27 16:06:46,291 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 2023-04-27 16:06:47,771 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7400, 1.5390, 1.3908, 1.5467, 1.9551, 1.6094, 1.3325, 1.3015], device='cuda:6'), covar=tensor([0.1591, 0.1265, 0.1826, 0.1358, 0.0811, 0.1576, 0.1832, 0.2324], device='cuda:6'), in_proj_covar=tensor([0.0311, 0.0310, 0.0351, 0.0288, 0.0326, 0.0307, 0.0300, 0.0370], device='cuda:6'), 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:6') 2023-04-27 16:06:48,361 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9387, 2.5649, 0.9196, 1.2583, 1.7179, 1.1081, 3.4210, 1.6174], device='cuda:6'), covar=tensor([0.0922, 0.0795, 0.0941, 0.1757, 0.0705, 0.1396, 0.0377, 0.0908], device='cuda:6'), in_proj_covar=tensor([0.0051, 0.0065, 0.0047, 0.0046, 0.0050, 0.0052, 0.0073, 0.0050], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0007], device='cuda:6') 2023-04-27 16:06:49,606 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6643, 1.8643, 0.6799, 1.3672, 1.8158, 1.5070, 1.4400, 1.5966], device='cuda:6'), covar=tensor([0.0493, 0.0359, 0.0359, 0.0556, 0.0277, 0.0495, 0.0500, 0.0558], device='cuda:6'), in_proj_covar=tensor([0.0028, 0.0024, 0.0022, 0.0029, 0.0019, 0.0028, 0.0028, 0.0028], device='cuda:6'), out_proj_covar=tensor([0.0051, 0.0045, 0.0038, 0.0052, 0.0038, 0.0050, 0.0050, 0.0051], device='cuda:6') 2023-04-27 16:06:49,622 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=119514.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:07:42,890 INFO [finetune.py:976] (6/7) Epoch 21, batch 5000, loss[loss=0.1419, simple_loss=0.2224, pruned_loss=0.0307, over 4906.00 frames. ], tot_loss[loss=0.1731, simple_loss=0.2457, pruned_loss=0.05032, over 958058.70 frames. ], batch size: 36, lr: 3.17e-03, grad_scale: 16.0 2023-04-27 16:07:56,237 INFO [optim.py:369] (6/7) 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 16:08:15,242 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=119575.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:08:27,012 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-04-27 16:08:44,479 INFO [finetune.py:976] (6/7) Epoch 21, batch 5050, loss[loss=0.1513, simple_loss=0.2236, pruned_loss=0.03952, over 4823.00 frames. ], tot_loss[loss=0.1706, simple_loss=0.2421, pruned_loss=0.04952, over 956305.65 frames. ], batch size: 41, lr: 3.17e-03, grad_scale: 16.0 2023-04-27 16:09:07,163 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.3220, 1.8343, 2.2569, 2.6392, 2.2048, 1.7869, 1.5618, 1.9507], device='cuda:6'), covar=tensor([0.3015, 0.3051, 0.1528, 0.2332, 0.2619, 0.2585, 0.4057, 0.2102], device='cuda:6'), in_proj_covar=tensor([0.0292, 0.0245, 0.0226, 0.0315, 0.0219, 0.0233, 0.0229, 0.0184], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:6') 2023-04-27 16:09:18,761 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.1106, 2.6522, 2.2186, 2.5087, 1.8166, 2.2483, 2.2498, 1.6538], device='cuda:6'), covar=tensor([0.1986, 0.1003, 0.0740, 0.1143, 0.3284, 0.1138, 0.1747, 0.2663], device='cuda:6'), in_proj_covar=tensor([0.0285, 0.0297, 0.0214, 0.0276, 0.0312, 0.0254, 0.0248, 0.0262], device='cuda:6'), out_proj_covar=tensor([1.1402e-04, 1.1766e-04, 8.4564e-05, 1.0929e-04, 1.2640e-04, 1.0065e-04, 9.9840e-05, 1.0392e-04], device='cuda:6') 2023-04-27 16:09:28,916 INFO [finetune.py:976] (6/7) Epoch 21, batch 5100, loss[loss=0.1458, simple_loss=0.2135, pruned_loss=0.03907, over 4751.00 frames. ], tot_loss[loss=0.1672, simple_loss=0.2382, pruned_loss=0.04806, over 956775.14 frames. ], batch size: 27, lr: 3.17e-03, grad_scale: 16.0 2023-04-27 16:09:41,673 INFO [optim.py:369] (6/7) 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 16:09:42,658 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 2023-04-27 16:09:43,052 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=119665.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:10:31,235 INFO [finetune.py:976] (6/7) Epoch 21, batch 5150, loss[loss=0.2374, simple_loss=0.3074, pruned_loss=0.08368, over 4773.00 frames. ], tot_loss[loss=0.1673, simple_loss=0.2383, pruned_loss=0.04815, over 957085.07 frames. ], batch size: 59, lr: 3.17e-03, grad_scale: 16.0 2023-04-27 16:11:05,040 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=119726.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:11:08,012 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=119731.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:11:16,585 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-04-27 16:11:30,485 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=119749.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 16:11:33,423 INFO [finetune.py:976] (6/7) Epoch 21, batch 5200, loss[loss=0.1936, simple_loss=0.2687, pruned_loss=0.05926, over 4898.00 frames. ], tot_loss[loss=0.1711, simple_loss=0.2429, pruned_loss=0.04971, over 954705.57 frames. ], batch size: 35, lr: 3.17e-03, grad_scale: 16.0 2023-04-27 16:11:39,368 INFO [optim.py:369] (6/7) 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 16:11:51,105 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=119779.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:12:02,215 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=119797.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:12:06,571 INFO [finetune.py:976] (6/7) Epoch 21, batch 5250, loss[loss=0.1273, simple_loss=0.21, pruned_loss=0.02234, over 4755.00 frames. ], tot_loss[loss=0.1725, simple_loss=0.245, pruned_loss=0.05004, over 955681.67 frames. ], batch size: 27, lr: 3.17e-03, grad_scale: 16.0 2023-04-27 16:12:17,449 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.10 vs. limit=2.0 2023-04-27 16:12:21,736 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([3.8084, 3.8565, 2.8609, 4.4362, 3.9079, 3.8037, 1.7552, 3.9411], device='cuda:6'), covar=tensor([0.1744, 0.1107, 0.3049, 0.1583, 0.2992, 0.1917, 0.5775, 0.2060], device='cuda:6'), in_proj_covar=tensor([0.0244, 0.0214, 0.0250, 0.0303, 0.0294, 0.0245, 0.0275, 0.0271], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 16:12:51,200 INFO [finetune.py:976] (6/7) Epoch 21, batch 5300, loss[loss=0.2113, simple_loss=0.2704, pruned_loss=0.07606, over 4915.00 frames. ], tot_loss[loss=0.1744, simple_loss=0.2469, pruned_loss=0.05093, over 955543.22 frames. ], batch size: 41, lr: 3.17e-03, grad_scale: 16.0 2023-04-27 16:13:02,332 INFO [optim.py:369] (6/7) 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 16:13:13,803 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=119870.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:13:53,927 INFO [finetune.py:976] (6/7) Epoch 21, batch 5350, loss[loss=0.142, simple_loss=0.2116, pruned_loss=0.0362, over 4854.00 frames. ], tot_loss[loss=0.1737, simple_loss=0.2463, pruned_loss=0.05048, over 955623.05 frames. ], batch size: 31, lr: 3.17e-03, grad_scale: 16.0 2023-04-27 16:14:05,457 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.1210, 2.7414, 1.0831, 1.3914, 2.1461, 1.2950, 3.4424, 1.8112], device='cuda:6'), covar=tensor([0.0636, 0.0692, 0.0824, 0.1253, 0.0465, 0.1006, 0.0228, 0.0595], device='cuda:6'), in_proj_covar=tensor([0.0050, 0.0064, 0.0047, 0.0046, 0.0049, 0.0051, 0.0073, 0.0050], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0007], device='cuda:6') 2023-04-27 16:14:25,760 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.1922, 2.6950, 2.2002, 2.4626, 1.8958, 2.3503, 2.2034, 1.6510], device='cuda:6'), covar=tensor([0.2007, 0.1218, 0.0814, 0.1316, 0.2998, 0.1219, 0.2084, 0.2937], device='cuda:6'), in_proj_covar=tensor([0.0284, 0.0296, 0.0213, 0.0274, 0.0310, 0.0253, 0.0247, 0.0261], device='cuda:6'), out_proj_covar=tensor([1.1377e-04, 1.1697e-04, 8.4387e-05, 1.0845e-04, 1.2586e-04, 1.0029e-04, 9.9777e-05, 1.0356e-04], device='cuda:6') 2023-04-27 16:14:25,775 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0103, 1.6753, 1.5670, 1.7828, 2.1795, 1.8284, 1.5442, 1.4669], device='cuda:6'), covar=tensor([0.1469, 0.1525, 0.1823, 0.1164, 0.0928, 0.1369, 0.2123, 0.2502], device='cuda:6'), in_proj_covar=tensor([0.0311, 0.0311, 0.0350, 0.0288, 0.0327, 0.0308, 0.0301, 0.0371], device='cuda:6'), out_proj_covar=tensor([6.3854e-05, 6.4286e-05, 7.3908e-05, 5.8037e-05, 6.7455e-05, 6.4476e-05, 6.3116e-05, 7.8746e-05], device='cuda:6') 2023-04-27 16:14:43,495 INFO [finetune.py:976] (6/7) Epoch 21, batch 5400, loss[loss=0.1436, simple_loss=0.2234, pruned_loss=0.03185, over 4839.00 frames. ], tot_loss[loss=0.1718, simple_loss=0.2437, pruned_loss=0.04998, over 955525.86 frames. ], batch size: 30, lr: 3.17e-03, grad_scale: 16.0 2023-04-27 16:14:48,951 INFO [optim.py:369] (6/7) 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 16:15:18,158 INFO [finetune.py:976] (6/7) Epoch 21, batch 5450, loss[loss=0.1444, simple_loss=0.2176, pruned_loss=0.03562, over 4936.00 frames. ], tot_loss[loss=0.17, simple_loss=0.2411, pruned_loss=0.04946, over 957193.27 frames. ], batch size: 33, lr: 3.17e-03, grad_scale: 16.0 2023-04-27 16:15:34,275 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=120021.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:15:49,996 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.3835, 1.4092, 1.4286, 1.6710, 1.7039, 1.3870, 0.8759, 1.5696], device='cuda:6'), covar=tensor([0.0937, 0.1280, 0.0928, 0.0666, 0.0695, 0.0842, 0.0956, 0.0623], device='cuda:6'), in_proj_covar=tensor([0.0186, 0.0200, 0.0183, 0.0173, 0.0176, 0.0180, 0.0151, 0.0177], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 16:16:03,189 INFO [finetune.py:976] (6/7) Epoch 21, batch 5500, loss[loss=0.1283, simple_loss=0.1973, pruned_loss=0.02967, over 4761.00 frames. ], tot_loss[loss=0.1665, simple_loss=0.2374, pruned_loss=0.04781, over 956604.05 frames. ], batch size: 27, lr: 3.17e-03, grad_scale: 16.0 2023-04-27 16:16:14,181 INFO [optim.py:369] (6/7) 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 16:16:26,392 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.8315, 2.0813, 2.0817, 2.2049, 1.9929, 2.1309, 2.1074, 2.0773], device='cuda:6'), covar=tensor([0.3568, 0.6441, 0.4965, 0.4458, 0.5721, 0.7023, 0.6348, 0.5696], device='cuda:6'), in_proj_covar=tensor([0.0339, 0.0375, 0.0324, 0.0338, 0.0349, 0.0396, 0.0358, 0.0330], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 16:17:07,730 INFO [finetune.py:976] (6/7) Epoch 21, batch 5550, loss[loss=0.1884, simple_loss=0.2593, pruned_loss=0.05877, over 4905.00 frames. ], tot_loss[loss=0.1675, simple_loss=0.2385, pruned_loss=0.04825, over 955047.41 frames. ], batch size: 36, lr: 3.17e-03, grad_scale: 16.0 2023-04-27 16:17:29,469 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5361, 3.0839, 0.9992, 1.8469, 1.8760, 2.2779, 1.8970, 1.1169], device='cuda:6'), covar=tensor([0.1404, 0.1015, 0.1969, 0.1135, 0.1114, 0.1011, 0.1714, 0.1792], device='cuda:6'), in_proj_covar=tensor([0.0118, 0.0242, 0.0138, 0.0120, 0.0134, 0.0153, 0.0117, 0.0119], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-27 16:17:37,971 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5353, 1.4363, 1.8366, 1.8084, 1.4209, 1.3179, 1.6142, 0.8648], device='cuda:6'), covar=tensor([0.0591, 0.0683, 0.0395, 0.0687, 0.0898, 0.1264, 0.0576, 0.0731], device='cuda:6'), in_proj_covar=tensor([0.0069, 0.0068, 0.0067, 0.0068, 0.0075, 0.0097, 0.0073, 0.0065], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-27 16:18:05,779 INFO [finetune.py:976] (6/7) Epoch 21, batch 5600, loss[loss=0.1899, simple_loss=0.2538, pruned_loss=0.06299, over 4213.00 frames. ], tot_loss[loss=0.1687, simple_loss=0.2409, pruned_loss=0.04821, over 952725.87 frames. ], batch size: 65, lr: 3.17e-03, grad_scale: 16.0 2023-04-27 16:18:09,071 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-04-27 16:18:11,014 INFO [optim.py:369] (6/7) 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 16:18:13,458 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.3704, 1.4330, 1.3125, 1.6889, 1.5958, 1.8771, 1.3451, 3.7369], device='cuda:6'), covar=tensor([0.0567, 0.0815, 0.0841, 0.1233, 0.0654, 0.0527, 0.0790, 0.0125], device='cuda:6'), in_proj_covar=tensor([0.0037, 0.0038, 0.0039, 0.0043, 0.0040, 0.0038, 0.0038, 0.0055], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:6') 2023-04-27 16:18:15,223 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=120170.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:18:30,394 INFO [scaling.py:679] (6/7) 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] (6/7) Epoch 21, batch 5650, loss[loss=0.185, simple_loss=0.251, pruned_loss=0.05948, over 4801.00 frames. ], tot_loss[loss=0.1718, simple_loss=0.2447, pruned_loss=0.04941, over 953043.85 frames. ], batch size: 41, lr: 3.17e-03, grad_scale: 16.0 2023-04-27 16:18:45,335 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=120218.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:19:08,057 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.4699, 2.2141, 2.6437, 3.0937, 2.2226, 1.9983, 2.4094, 1.4772], device='cuda:6'), covar=tensor([0.0406, 0.0607, 0.0372, 0.0489, 0.0512, 0.1006, 0.0492, 0.0642], device='cuda:6'), in_proj_covar=tensor([0.0069, 0.0068, 0.0067, 0.0068, 0.0075, 0.0097, 0.0073, 0.0065], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-27 16:19:09,210 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9326, 2.3627, 1.2926, 1.6486, 2.1678, 1.7622, 1.7346, 1.7791], device='cuda:6'), covar=tensor([0.0537, 0.0277, 0.0307, 0.0568, 0.0244, 0.0637, 0.0563, 0.0590], device='cuda:6'), in_proj_covar=tensor([0.0028, 0.0024, 0.0022, 0.0029, 0.0019, 0.0028, 0.0028, 0.0028], device='cuda:6'), out_proj_covar=tensor([0.0051, 0.0045, 0.0038, 0.0052, 0.0038, 0.0050, 0.0050, 0.0051], device='cuda:6') 2023-04-27 16:19:30,421 INFO [finetune.py:976] (6/7) Epoch 21, batch 5700, loss[loss=0.1884, simple_loss=0.2525, pruned_loss=0.06215, over 4226.00 frames. ], tot_loss[loss=0.1698, simple_loss=0.2414, pruned_loss=0.04915, over 934003.17 frames. ], batch size: 18, lr: 3.17e-03, grad_scale: 16.0 2023-04-27 16:19:39,338 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.3757, 2.6385, 1.2899, 1.6438, 2.1974, 1.5385, 3.6261, 2.1552], device='cuda:6'), covar=tensor([0.0593, 0.0675, 0.0685, 0.1077, 0.0431, 0.0859, 0.0214, 0.0468], device='cuda:6'), in_proj_covar=tensor([0.0051, 0.0064, 0.0047, 0.0046, 0.0049, 0.0052, 0.0073, 0.0050], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0007], device='cuda:6') 2023-04-27 16:19:39,347 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5486, 1.8544, 1.8000, 2.1315, 1.9652, 2.1203, 1.7173, 3.7052], device='cuda:6'), covar=tensor([0.0497, 0.0656, 0.0670, 0.1012, 0.0539, 0.0438, 0.0632, 0.0153], device='cuda:6'), in_proj_covar=tensor([0.0037, 0.0038, 0.0039, 0.0043, 0.0040, 0.0038, 0.0038, 0.0055], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:6') 2023-04-27 16:19:41,634 INFO [optim.py:369] (6/7) 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] (6/7) Epoch 22, batch 0, loss[loss=0.1952, simple_loss=0.273, pruned_loss=0.05873, over 4904.00 frames. ], tot_loss[loss=0.1952, simple_loss=0.273, pruned_loss=0.05873, over 4904.00 frames. ], batch size: 36, lr: 3.16e-03, grad_scale: 16.0 2023-04-27 16:20:17,302 INFO [finetune.py:1001] (6/7) Computing validation loss 2023-04-27 16:20:33,672 INFO [finetune.py:1010] (6/7) 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] (6/7) Maximum memory allocated so far is 6365MB 2023-04-27 16:20:37,880 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9958, 1.7811, 1.9513, 2.3473, 2.3372, 1.9352, 1.6679, 2.0844], device='cuda:6'), covar=tensor([0.0715, 0.0999, 0.0612, 0.0453, 0.0569, 0.0799, 0.0774, 0.0537], device='cuda:6'), in_proj_covar=tensor([0.0187, 0.0201, 0.0183, 0.0174, 0.0177, 0.0180, 0.0152, 0.0178], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 16:20:59,172 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=120321.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:21:06,362 INFO [finetune.py:976] (6/7) Epoch 22, batch 50, loss[loss=0.186, simple_loss=0.2573, pruned_loss=0.05736, over 4921.00 frames. ], tot_loss[loss=0.1735, simple_loss=0.247, pruned_loss=0.04999, over 215773.66 frames. ], batch size: 33, lr: 3.16e-03, grad_scale: 16.0 2023-04-27 16:21:27,312 INFO [optim.py:369] (6/7) 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 16:21:31,537 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=120369.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:21:51,385 INFO [finetune.py:976] (6/7) Epoch 22, batch 100, loss[loss=0.1905, simple_loss=0.2569, pruned_loss=0.06205, over 4817.00 frames. ], tot_loss[loss=0.1732, simple_loss=0.2444, pruned_loss=0.05096, over 379837.18 frames. ], batch size: 39, lr: 3.16e-03, grad_scale: 16.0 2023-04-27 16:22:05,272 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.09 vs. limit=2.0 2023-04-27 16:22:55,498 INFO [finetune.py:976] (6/7) Epoch 22, batch 150, loss[loss=0.1303, simple_loss=0.2069, pruned_loss=0.02685, over 4740.00 frames. ], tot_loss[loss=0.1708, simple_loss=0.2401, pruned_loss=0.05076, over 507041.92 frames. ], batch size: 54, lr: 3.16e-03, grad_scale: 16.0 2023-04-27 16:23:14,428 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 2023-04-27 16:23:29,003 INFO [optim.py:369] (6/7) 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 16:23:40,080 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-04-27 16:23:58,242 INFO [finetune.py:976] (6/7) Epoch 22, batch 200, loss[loss=0.1996, simple_loss=0.2793, pruned_loss=0.05999, over 4814.00 frames. ], tot_loss[loss=0.1669, simple_loss=0.2368, pruned_loss=0.04853, over 608250.45 frames. ], batch size: 41, lr: 3.16e-03, grad_scale: 16.0 2023-04-27 16:24:08,578 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5053, 1.3547, 4.1915, 3.9126, 3.6516, 3.9672, 3.9073, 3.6636], device='cuda:6'), covar=tensor([0.7041, 0.5867, 0.1175, 0.1817, 0.1141, 0.1795, 0.1707, 0.1619], device='cuda:6'), in_proj_covar=tensor([0.0313, 0.0308, 0.0410, 0.0409, 0.0350, 0.0413, 0.0314, 0.0369], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-27 16:24:20,190 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6535, 1.5885, 0.7178, 1.3006, 1.7682, 1.5082, 1.4014, 1.4953], device='cuda:6'), covar=tensor([0.0488, 0.0371, 0.0357, 0.0551, 0.0266, 0.0492, 0.0464, 0.0565], device='cuda:6'), in_proj_covar=tensor([0.0028, 0.0024, 0.0022, 0.0029, 0.0019, 0.0028, 0.0028, 0.0028], device='cuda:6'), out_proj_covar=tensor([0.0051, 0.0045, 0.0038, 0.0052, 0.0038, 0.0050, 0.0050, 0.0051], device='cuda:6') 2023-04-27 16:25:06,006 INFO [finetune.py:976] (6/7) Epoch 22, batch 250, loss[loss=0.143, simple_loss=0.2188, pruned_loss=0.03355, over 4826.00 frames. ], tot_loss[loss=0.1689, simple_loss=0.2399, pruned_loss=0.049, over 685615.57 frames. ], batch size: 30, lr: 3.16e-03, grad_scale: 16.0 2023-04-27 16:25:49,440 INFO [optim.py:369] (6/7) 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 16:25:51,980 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([0.4160, 1.3769, 1.4527, 1.0599, 1.3664, 1.2214, 1.7414, 1.4126], device='cuda:6'), covar=tensor([0.3507, 0.1955, 0.4601, 0.2563, 0.1446, 0.2140, 0.1425, 0.4400], device='cuda:6'), in_proj_covar=tensor([0.0339, 0.0349, 0.0423, 0.0351, 0.0379, 0.0372, 0.0366, 0.0416], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 16:26:12,262 INFO [finetune.py:976] (6/7) Epoch 22, batch 300, loss[loss=0.1831, simple_loss=0.2588, pruned_loss=0.05369, over 4924.00 frames. ], tot_loss[loss=0.1735, simple_loss=0.2458, pruned_loss=0.05067, over 746562.14 frames. ], batch size: 33, lr: 3.16e-03, grad_scale: 16.0 2023-04-27 16:27:19,463 INFO [finetune.py:976] (6/7) Epoch 22, batch 350, loss[loss=0.231, simple_loss=0.2948, pruned_loss=0.08364, over 4809.00 frames. ], tot_loss[loss=0.1756, simple_loss=0.2482, pruned_loss=0.05153, over 794324.15 frames. ], batch size: 39, lr: 3.16e-03, grad_scale: 16.0 2023-04-27 16:28:01,877 INFO [optim.py:369] (6/7) 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 16:28:20,609 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=120676.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:28:24,209 INFO [finetune.py:976] (6/7) Epoch 22, batch 400, loss[loss=0.1677, simple_loss=0.2448, pruned_loss=0.04525, over 4896.00 frames. ], tot_loss[loss=0.1738, simple_loss=0.2471, pruned_loss=0.05032, over 830329.75 frames. ], batch size: 37, lr: 3.16e-03, grad_scale: 16.0 2023-04-27 16:29:32,132 INFO [finetune.py:976] (6/7) Epoch 22, batch 450, loss[loss=0.166, simple_loss=0.2507, pruned_loss=0.04065, over 4811.00 frames. ], tot_loss[loss=0.1746, simple_loss=0.2472, pruned_loss=0.05098, over 859327.45 frames. ], batch size: 40, lr: 3.16e-03, grad_scale: 16.0 2023-04-27 16:29:40,503 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=120737.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:30:03,658 INFO [optim.py:369] (6/7) 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 16:30:15,270 INFO [finetune.py:976] (6/7) Epoch 22, batch 500, loss[loss=0.1247, simple_loss=0.1911, pruned_loss=0.02917, over 4853.00 frames. ], tot_loss[loss=0.173, simple_loss=0.2444, pruned_loss=0.05079, over 881800.77 frames. ], batch size: 49, lr: 3.16e-03, grad_scale: 16.0 2023-04-27 16:30:49,307 INFO [finetune.py:976] (6/7) Epoch 22, batch 550, loss[loss=0.1219, simple_loss=0.2039, pruned_loss=0.01995, over 4766.00 frames. ], tot_loss[loss=0.1704, simple_loss=0.2416, pruned_loss=0.04958, over 898777.66 frames. ], batch size: 28, lr: 3.16e-03, grad_scale: 16.0 2023-04-27 16:30:51,306 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.1622, 2.7938, 2.1093, 2.1002, 1.4946, 1.5379, 2.2282, 1.5095], device='cuda:6'), covar=tensor([0.1599, 0.1380, 0.1424, 0.1641, 0.2329, 0.1899, 0.0964, 0.2015], device='cuda:6'), in_proj_covar=tensor([0.0196, 0.0210, 0.0168, 0.0203, 0.0199, 0.0185, 0.0155, 0.0185], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:6') 2023-04-27 16:30:54,608 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-04-27 16:30:58,474 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.1603, 1.9367, 2.2788, 2.4033, 2.2413, 1.9960, 2.1269, 2.0813], device='cuda:6'), covar=tensor([0.4188, 0.6447, 0.6236, 0.5034, 0.5487, 0.8513, 0.8507, 0.8868], device='cuda:6'), in_proj_covar=tensor([0.0428, 0.0412, 0.0502, 0.0503, 0.0456, 0.0485, 0.0492, 0.0499], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 16:31:15,724 INFO [optim.py:369] (6/7) 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 16:31:25,100 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.2638, 1.4589, 1.6894, 1.8205, 1.7281, 1.8759, 1.7185, 1.7407], device='cuda:6'), covar=tensor([0.3769, 0.5089, 0.4142, 0.4119, 0.5282, 0.6781, 0.4822, 0.4538], device='cuda:6'), in_proj_covar=tensor([0.0339, 0.0373, 0.0324, 0.0338, 0.0347, 0.0395, 0.0358, 0.0329], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 16:31:38,951 INFO [finetune.py:976] (6/7) Epoch 22, batch 600, loss[loss=0.1465, simple_loss=0.2277, pruned_loss=0.03268, over 4817.00 frames. ], tot_loss[loss=0.1702, simple_loss=0.2412, pruned_loss=0.04956, over 911617.10 frames. ], batch size: 39, lr: 3.16e-03, grad_scale: 16.0 2023-04-27 16:32:45,428 INFO [finetune.py:976] (6/7) Epoch 22, batch 650, loss[loss=0.2446, simple_loss=0.2983, pruned_loss=0.09547, over 4831.00 frames. ], tot_loss[loss=0.1743, simple_loss=0.246, pruned_loss=0.05132, over 922548.12 frames. ], batch size: 49, lr: 3.16e-03, grad_scale: 16.0 2023-04-27 16:33:17,912 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.62 vs. limit=2.0 2023-04-27 16:33:26,707 INFO [optim.py:369] (6/7) 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 16:33:52,003 INFO [finetune.py:976] (6/7) Epoch 22, batch 700, loss[loss=0.1411, simple_loss=0.2141, pruned_loss=0.03407, over 4932.00 frames. ], tot_loss[loss=0.176, simple_loss=0.2481, pruned_loss=0.05199, over 929557.10 frames. ], batch size: 38, lr: 3.16e-03, grad_scale: 16.0 2023-04-27 16:34:21,528 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=121003.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:34:45,578 INFO [finetune.py:976] (6/7) Epoch 22, batch 750, loss[loss=0.1954, simple_loss=0.2684, pruned_loss=0.06124, over 4810.00 frames. ], tot_loss[loss=0.1762, simple_loss=0.2484, pruned_loss=0.05198, over 935375.36 frames. ], batch size: 39, lr: 3.16e-03, grad_scale: 16.0 2023-04-27 16:34:45,675 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=121032.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:35:04,851 INFO [optim.py:369] (6/7) 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 16:35:05,568 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=121064.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:35:06,188 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8031, 1.6700, 2.1720, 2.2720, 1.6500, 1.4204, 1.9004, 0.9751], device='cuda:6'), covar=tensor([0.0691, 0.0752, 0.0426, 0.0590, 0.0776, 0.1205, 0.0658, 0.0714], device='cuda:6'), in_proj_covar=tensor([0.0069, 0.0068, 0.0067, 0.0068, 0.0075, 0.0096, 0.0072, 0.0064], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-27 16:35:19,167 INFO [finetune.py:976] (6/7) Epoch 22, batch 800, loss[loss=0.1793, simple_loss=0.2532, pruned_loss=0.05272, over 4721.00 frames. ], tot_loss[loss=0.1742, simple_loss=0.2466, pruned_loss=0.05088, over 938714.09 frames. ], batch size: 54, lr: 3.16e-03, grad_scale: 16.0 2023-04-27 16:35:25,420 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.0502, 0.8196, 0.9157, 0.7661, 1.1904, 0.9969, 0.8752, 0.9575], device='cuda:6'), covar=tensor([0.1678, 0.1479, 0.1870, 0.1716, 0.0964, 0.1374, 0.1708, 0.2260], device='cuda:6'), in_proj_covar=tensor([0.0313, 0.0312, 0.0352, 0.0289, 0.0326, 0.0310, 0.0302, 0.0374], device='cuda:6'), out_proj_covar=tensor([6.4295e-05, 6.4602e-05, 7.4316e-05, 5.8174e-05, 6.7164e-05, 6.4866e-05, 6.3169e-05, 7.9379e-05], device='cuda:6') 2023-04-27 16:35:37,480 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.2426, 1.2038, 3.8317, 3.5499, 3.3641, 3.6437, 3.6723, 3.4126], device='cuda:6'), covar=tensor([0.6889, 0.5906, 0.1258, 0.1966, 0.1346, 0.1776, 0.1432, 0.1732], device='cuda:6'), in_proj_covar=tensor([0.0311, 0.0305, 0.0407, 0.0407, 0.0350, 0.0410, 0.0312, 0.0367], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-27 16:35:52,466 INFO [finetune.py:976] (6/7) Epoch 22, batch 850, loss[loss=0.1393, simple_loss=0.2061, pruned_loss=0.03623, over 4791.00 frames. ], tot_loss[loss=0.1723, simple_loss=0.2445, pruned_loss=0.05002, over 941773.64 frames. ], batch size: 25, lr: 3.16e-03, grad_scale: 16.0 2023-04-27 16:36:07,618 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.3935, 1.5930, 1.5422, 1.9284, 1.8791, 2.0592, 1.5503, 4.0597], device='cuda:6'), covar=tensor([0.0562, 0.0797, 0.0750, 0.1187, 0.0587, 0.0560, 0.0733, 0.0101], device='cuda:6'), in_proj_covar=tensor([0.0037, 0.0038, 0.0039, 0.0043, 0.0040, 0.0038, 0.0038, 0.0055], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:6') 2023-04-27 16:36:11,704 INFO [optim.py:369] (6/7) 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 16:36:12,462 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=121164.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:36:25,803 INFO [finetune.py:976] (6/7) Epoch 22, batch 900, loss[loss=0.1365, simple_loss=0.2195, pruned_loss=0.02676, over 4864.00 frames. ], tot_loss[loss=0.1692, simple_loss=0.2411, pruned_loss=0.04868, over 945500.01 frames. ], batch size: 34, lr: 3.16e-03, grad_scale: 16.0 2023-04-27 16:36:32,593 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7398, 2.1588, 1.7837, 1.5387, 1.2954, 1.3195, 1.9428, 1.2681], device='cuda:6'), covar=tensor([0.1706, 0.1419, 0.1494, 0.1895, 0.2509, 0.2075, 0.0960, 0.2173], device='cuda:6'), in_proj_covar=tensor([0.0195, 0.0209, 0.0167, 0.0201, 0.0198, 0.0184, 0.0154, 0.0184], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:6') 2023-04-27 16:36:52,987 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=121225.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:36:59,097 INFO [finetune.py:976] (6/7) Epoch 22, batch 950, loss[loss=0.1876, simple_loss=0.262, pruned_loss=0.05664, over 4928.00 frames. ], tot_loss[loss=0.1689, simple_loss=0.2398, pruned_loss=0.04901, over 947414.27 frames. ], batch size: 38, lr: 3.16e-03, grad_scale: 16.0 2023-04-27 16:37:35,641 INFO [optim.py:369] (6/7) 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 16:38:01,690 INFO [finetune.py:976] (6/7) Epoch 22, batch 1000, loss[loss=0.1698, simple_loss=0.2391, pruned_loss=0.05026, over 4896.00 frames. ], tot_loss[loss=0.1718, simple_loss=0.2426, pruned_loss=0.0505, over 949433.50 frames. ], batch size: 32, lr: 3.16e-03, grad_scale: 16.0 2023-04-27 16:38:30,426 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7785, 2.2333, 1.9812, 1.6546, 1.3097, 1.3824, 2.0731, 1.2418], device='cuda:6'), covar=tensor([0.1881, 0.1649, 0.1428, 0.1889, 0.2469, 0.2077, 0.0886, 0.2253], device='cuda:6'), in_proj_covar=tensor([0.0195, 0.0209, 0.0168, 0.0202, 0.0198, 0.0184, 0.0154, 0.0185], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:6') 2023-04-27 16:38:41,537 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=121311.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 16:39:01,159 INFO [finetune.py:976] (6/7) Epoch 22, batch 1050, loss[loss=0.148, simple_loss=0.2331, pruned_loss=0.03139, over 4764.00 frames. ], tot_loss[loss=0.1732, simple_loss=0.2449, pruned_loss=0.05075, over 951207.48 frames. ], batch size: 28, lr: 3.16e-03, grad_scale: 16.0 2023-04-27 16:39:01,744 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=121332.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:39:18,518 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=121359.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:39:20,859 INFO [optim.py:369] (6/7) 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 16:39:26,944 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=121372.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 16:39:31,744 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 22, batch 1100, loss[loss=0.1667, simple_loss=0.2398, pruned_loss=0.04679, over 4907.00 frames. ], tot_loss[loss=0.1749, simple_loss=0.2467, pruned_loss=0.05158, over 952324.47 frames. ], batch size: 43, lr: 3.16e-03, grad_scale: 16.0 2023-04-27 16:39:41,056 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-04-27 16:39:50,921 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=121407.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:40:05,807 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([4.9223, 4.7003, 3.2026, 5.5740, 4.9125, 4.8519, 2.2302, 4.8104], device='cuda:6'), covar=tensor([0.1225, 0.0871, 0.2892, 0.0779, 0.2551, 0.1511, 0.5136, 0.2073], device='cuda:6'), in_proj_covar=tensor([0.0243, 0.0215, 0.0251, 0.0304, 0.0294, 0.0245, 0.0273, 0.0272], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 16:40:06,343 INFO [finetune.py:976] (6/7) Epoch 22, batch 1150, loss[loss=0.1494, simple_loss=0.2241, pruned_loss=0.0373, over 4804.00 frames. ], tot_loss[loss=0.1748, simple_loss=0.247, pruned_loss=0.05132, over 953751.71 frames. ], batch size: 25, lr: 3.16e-03, grad_scale: 16.0 2023-04-27 16:40:27,720 INFO [optim.py:369] (6/7) 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 16:40:30,881 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=121468.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:40:39,781 INFO [finetune.py:976] (6/7) Epoch 22, batch 1200, loss[loss=0.1564, simple_loss=0.2249, pruned_loss=0.04393, over 4811.00 frames. ], tot_loss[loss=0.1728, simple_loss=0.2447, pruned_loss=0.05045, over 955373.68 frames. ], batch size: 39, lr: 3.15e-03, grad_scale: 16.0 2023-04-27 16:40:57,510 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.81 vs. limit=5.0 2023-04-27 16:41:05,077 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=121520.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:41:12,820 INFO [finetune.py:976] (6/7) Epoch 22, batch 1250, loss[loss=0.1593, simple_loss=0.237, pruned_loss=0.04081, over 4929.00 frames. ], tot_loss[loss=0.1704, simple_loss=0.2419, pruned_loss=0.04944, over 954694.18 frames. ], batch size: 38, lr: 3.15e-03, grad_scale: 32.0 2023-04-27 16:41:14,166 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.1084, 3.2336, 2.7665, 2.9130, 3.3175, 2.8444, 4.0824, 2.6212], device='cuda:6'), covar=tensor([0.3508, 0.1631, 0.3689, 0.2643, 0.1398, 0.2313, 0.0899, 0.3431], device='cuda:6'), in_proj_covar=tensor([0.0339, 0.0349, 0.0423, 0.0352, 0.0380, 0.0374, 0.0366, 0.0417], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 16:41:34,046 INFO [optim.py:369] (6/7) 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 16:41:46,168 INFO [finetune.py:976] (6/7) Epoch 22, batch 1300, loss[loss=0.1611, simple_loss=0.2305, pruned_loss=0.0459, over 4929.00 frames. ], tot_loss[loss=0.1681, simple_loss=0.2395, pruned_loss=0.04839, over 956170.18 frames. ], batch size: 33, lr: 3.15e-03, grad_scale: 32.0 2023-04-27 16:41:52,486 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-04-27 16:42:19,075 INFO [finetune.py:976] (6/7) Epoch 22, batch 1350, loss[loss=0.1841, simple_loss=0.2489, pruned_loss=0.05966, over 4821.00 frames. ], tot_loss[loss=0.1701, simple_loss=0.2413, pruned_loss=0.04947, over 956267.79 frames. ], batch size: 33, lr: 3.15e-03, grad_scale: 32.0 2023-04-27 16:42:32,456 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([3.4044, 3.3118, 2.5851, 3.8834, 3.3111, 3.3774, 1.4253, 3.3552], device='cuda:6'), covar=tensor([0.1788, 0.1378, 0.3130, 0.2326, 0.3977, 0.1811, 0.5711, 0.2362], device='cuda:6'), in_proj_covar=tensor([0.0242, 0.0214, 0.0249, 0.0303, 0.0292, 0.0244, 0.0272, 0.0270], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 16:42:55,955 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=121659.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:42:58,250 INFO [optim.py:369] (6/7) 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 16:42:58,800 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.49 vs. limit=2.0 2023-04-27 16:43:04,729 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=121667.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 16:43:19,353 INFO [finetune.py:976] (6/7) Epoch 22, batch 1400, loss[loss=0.1606, simple_loss=0.2296, pruned_loss=0.04581, over 4919.00 frames. ], tot_loss[loss=0.172, simple_loss=0.2439, pruned_loss=0.05003, over 958383.02 frames. ], batch size: 36, lr: 3.15e-03, grad_scale: 32.0 2023-04-27 16:43:59,859 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=121707.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:44:03,288 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 2023-04-27 16:44:24,400 INFO [finetune.py:976] (6/7) Epoch 22, batch 1450, loss[loss=0.2457, simple_loss=0.3002, pruned_loss=0.0956, over 4877.00 frames. ], tot_loss[loss=0.174, simple_loss=0.2463, pruned_loss=0.05087, over 958890.93 frames. ], batch size: 34, lr: 3.15e-03, grad_scale: 32.0 2023-04-27 16:45:04,294 INFO [optim.py:369] (6/7) 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] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=121763.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:45:24,654 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-04-27 16:45:27,550 INFO [finetune.py:976] (6/7) Epoch 22, batch 1500, loss[loss=0.1803, simple_loss=0.249, pruned_loss=0.05585, over 4888.00 frames. ], tot_loss[loss=0.175, simple_loss=0.2475, pruned_loss=0.05121, over 959105.41 frames. ], batch size: 35, lr: 3.15e-03, grad_scale: 32.0 2023-04-27 16:45:58,969 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=121820.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:46:06,251 INFO [finetune.py:976] (6/7) Epoch 22, batch 1550, loss[loss=0.2006, simple_loss=0.2669, pruned_loss=0.06718, over 4905.00 frames. ], tot_loss[loss=0.175, simple_loss=0.2476, pruned_loss=0.05114, over 958807.74 frames. ], batch size: 36, lr: 3.15e-03, grad_scale: 32.0 2023-04-27 16:46:28,045 INFO [optim.py:369] (6/7) 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 16:46:31,189 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=121868.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:46:34,388 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.6434, 2.1209, 2.5453, 3.0845, 2.4397, 2.0314, 1.9867, 2.4507], device='cuda:6'), covar=tensor([0.3120, 0.3204, 0.1686, 0.2502, 0.2893, 0.2729, 0.3773, 0.2146], device='cuda:6'), in_proj_covar=tensor([0.0290, 0.0242, 0.0224, 0.0311, 0.0219, 0.0231, 0.0226, 0.0182], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:6') 2023-04-27 16:46:39,811 INFO [finetune.py:976] (6/7) Epoch 22, batch 1600, loss[loss=0.183, simple_loss=0.2509, pruned_loss=0.05756, over 4875.00 frames. ], tot_loss[loss=0.1736, simple_loss=0.2459, pruned_loss=0.05067, over 958752.33 frames. ], batch size: 31, lr: 3.15e-03, grad_scale: 32.0 2023-04-27 16:47:13,761 INFO [finetune.py:976] (6/7) Epoch 22, batch 1650, loss[loss=0.1318, simple_loss=0.2068, pruned_loss=0.02838, over 4823.00 frames. ], tot_loss[loss=0.1718, simple_loss=0.2434, pruned_loss=0.05008, over 959121.55 frames. ], batch size: 41, lr: 3.15e-03, grad_scale: 32.0 2023-04-27 16:47:21,525 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-04-27 16:47:24,843 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=121950.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:47:33,997 INFO [optim.py:369] (6/7) 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 16:47:38,073 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=121967.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 16:47:47,187 INFO [finetune.py:976] (6/7) Epoch 22, batch 1700, loss[loss=0.1725, simple_loss=0.2346, pruned_loss=0.05518, over 4832.00 frames. ], tot_loss[loss=0.1693, simple_loss=0.2404, pruned_loss=0.04915, over 957951.27 frames. ], batch size: 33, lr: 3.15e-03, grad_scale: 32.0 2023-04-27 16:47:47,329 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.1968, 1.6325, 2.0426, 2.1393, 2.0068, 1.6439, 1.2115, 1.6801], device='cuda:6'), covar=tensor([0.3214, 0.3077, 0.1593, 0.2225, 0.2403, 0.2540, 0.4167, 0.1870], device='cuda:6'), in_proj_covar=tensor([0.0292, 0.0243, 0.0226, 0.0313, 0.0220, 0.0232, 0.0227, 0.0183], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:6') 2023-04-27 16:48:06,897 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=122011.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:48:16,070 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=122015.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 16:48:37,956 INFO [finetune.py:976] (6/7) Epoch 22, batch 1750, loss[loss=0.2217, simple_loss=0.3175, pruned_loss=0.06294, over 4860.00 frames. ], tot_loss[loss=0.1708, simple_loss=0.2419, pruned_loss=0.04987, over 956466.85 frames. ], batch size: 44, lr: 3.15e-03, grad_scale: 32.0 2023-04-27 16:48:46,854 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.3005, 1.0956, 4.0022, 3.7487, 3.5371, 3.8285, 3.7688, 3.5386], device='cuda:6'), covar=tensor([0.7499, 0.6634, 0.1172, 0.1718, 0.1161, 0.1715, 0.1721, 0.1526], device='cuda:6'), in_proj_covar=tensor([0.0310, 0.0308, 0.0408, 0.0407, 0.0348, 0.0410, 0.0313, 0.0366], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-27 16:49:13,876 INFO [optim.py:369] (6/7) 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 16:49:19,205 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=122063.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:49:43,796 INFO [finetune.py:976] (6/7) Epoch 22, batch 1800, loss[loss=0.1231, simple_loss=0.1894, pruned_loss=0.02838, over 4699.00 frames. ], tot_loss[loss=0.1722, simple_loss=0.2439, pruned_loss=0.05027, over 956169.28 frames. ], batch size: 23, lr: 3.15e-03, grad_scale: 32.0 2023-04-27 16:50:18,108 INFO [zipformer.py:1188] (6/7) 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] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=122117.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:50:48,519 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.99 vs. limit=2.0 2023-04-27 16:50:49,860 INFO [finetune.py:976] (6/7) Epoch 22, batch 1850, loss[loss=0.1842, simple_loss=0.262, pruned_loss=0.05323, over 4844.00 frames. ], tot_loss[loss=0.1738, simple_loss=0.2459, pruned_loss=0.05086, over 955180.83 frames. ], batch size: 47, lr: 3.15e-03, grad_scale: 32.0 2023-04-27 16:51:14,577 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-04-27 16:51:22,412 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.2533, 1.5203, 1.4219, 1.7177, 1.6404, 1.9353, 1.3545, 3.4439], device='cuda:6'), covar=tensor([0.0604, 0.0847, 0.0779, 0.1210, 0.0653, 0.0544, 0.0810, 0.0142], device='cuda:6'), in_proj_covar=tensor([0.0038, 0.0039, 0.0039, 0.0043, 0.0040, 0.0038, 0.0038, 0.0055], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0013, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:6') 2023-04-27 16:51:27,104 INFO [optim.py:369] (6/7) 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 16:51:32,498 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9405, 2.3624, 0.9871, 1.2877, 1.8761, 1.1137, 2.8792, 1.4388], device='cuda:6'), covar=tensor([0.0710, 0.0662, 0.0750, 0.1228, 0.0454, 0.1037, 0.0259, 0.0673], device='cuda:6'), in_proj_covar=tensor([0.0051, 0.0064, 0.0047, 0.0046, 0.0049, 0.0052, 0.0073, 0.0050], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:6') 2023-04-27 16:51:53,492 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=122178.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 16:51:56,334 INFO [finetune.py:976] (6/7) Epoch 22, batch 1900, loss[loss=0.1634, simple_loss=0.2364, pruned_loss=0.04522, over 4906.00 frames. ], tot_loss[loss=0.1741, simple_loss=0.2469, pruned_loss=0.05066, over 954580.33 frames. ], batch size: 38, lr: 3.15e-03, grad_scale: 32.0 2023-04-27 16:52:14,289 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=122195.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:52:20,614 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.2892, 1.5145, 1.3531, 1.5804, 1.2629, 1.3448, 1.4562, 1.0241], device='cuda:6'), covar=tensor([0.1614, 0.1247, 0.0858, 0.1020, 0.3698, 0.1117, 0.1523, 0.2268], device='cuda:6'), in_proj_covar=tensor([0.0290, 0.0306, 0.0220, 0.0281, 0.0319, 0.0260, 0.0253, 0.0267], device='cuda:6'), 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:6') 2023-04-27 16:52:27,913 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.14 vs. limit=2.0 2023-04-27 16:52:40,530 INFO [finetune.py:976] (6/7) Epoch 22, batch 1950, loss[loss=0.1534, simple_loss=0.2182, pruned_loss=0.04432, over 4806.00 frames. ], tot_loss[loss=0.1724, simple_loss=0.2455, pruned_loss=0.04963, over 956073.36 frames. ], batch size: 25, lr: 3.15e-03, grad_scale: 32.0 2023-04-27 16:52:49,565 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-04-27 16:52:55,275 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=122256.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:52:58,857 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9130, 2.4313, 0.9419, 1.2891, 1.9341, 1.1668, 2.9160, 1.4811], device='cuda:6'), covar=tensor([0.0909, 0.0641, 0.0875, 0.1625, 0.0566, 0.1264, 0.0352, 0.0885], device='cuda:6'), in_proj_covar=tensor([0.0050, 0.0064, 0.0047, 0.0046, 0.0049, 0.0051, 0.0073, 0.0050], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0009, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0007], device='cuda:6') 2023-04-27 16:52:59,364 INFO [optim.py:369] (6/7) 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 16:53:13,156 INFO [finetune.py:976] (6/7) Epoch 22, batch 2000, loss[loss=0.1632, simple_loss=0.2451, pruned_loss=0.04066, over 4792.00 frames. ], tot_loss[loss=0.1703, simple_loss=0.2424, pruned_loss=0.04915, over 955962.08 frames. ], batch size: 29, lr: 3.15e-03, grad_scale: 32.0 2023-04-27 16:53:29,027 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=122306.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:53:32,211 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.52 vs. limit=2.0 2023-04-27 16:53:37,541 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 22, batch 2050, loss[loss=0.1144, simple_loss=0.1917, pruned_loss=0.01858, over 4777.00 frames. ], tot_loss[loss=0.1673, simple_loss=0.2388, pruned_loss=0.04794, over 956595.04 frames. ], batch size: 23, lr: 3.15e-03, grad_scale: 32.0 2023-04-27 16:53:52,829 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7807, 1.5354, 1.7099, 2.1775, 2.0850, 1.7610, 1.4165, 1.8970], device='cuda:6'), covar=tensor([0.0701, 0.1150, 0.0736, 0.0495, 0.0534, 0.0775, 0.0780, 0.0523], device='cuda:6'), in_proj_covar=tensor([0.0185, 0.0200, 0.0182, 0.0173, 0.0176, 0.0178, 0.0151, 0.0176], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 16:54:02,535 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.1321, 1.7322, 1.9554, 2.5449, 2.3537, 2.1361, 1.8809, 2.1671], device='cuda:6'), covar=tensor([0.0778, 0.1164, 0.0809, 0.0528, 0.0643, 0.0865, 0.0730, 0.0570], device='cuda:6'), in_proj_covar=tensor([0.0185, 0.0200, 0.0182, 0.0173, 0.0176, 0.0178, 0.0150, 0.0176], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 16:54:12,098 INFO [optim.py:369] (6/7) 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 16:54:21,802 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.2232, 1.4200, 1.2812, 1.6446, 1.4904, 1.7001, 1.2895, 2.9981], device='cuda:6'), covar=tensor([0.0606, 0.0825, 0.0813, 0.1240, 0.0642, 0.0520, 0.0787, 0.0182], device='cuda:6'), in_proj_covar=tensor([0.0038, 0.0039, 0.0039, 0.0043, 0.0040, 0.0038, 0.0038, 0.0055], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0013, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:6') 2023-04-27 16:54:23,026 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.2137, 1.4175, 1.3086, 1.6737, 1.4923, 1.6388, 1.3350, 3.0318], device='cuda:6'), covar=tensor([0.0599, 0.0829, 0.0807, 0.1237, 0.0651, 0.0570, 0.0792, 0.0171], device='cuda:6'), in_proj_covar=tensor([0.0038, 0.0039, 0.0039, 0.0043, 0.0040, 0.0038, 0.0038, 0.0055], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0013, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:6') 2023-04-27 16:54:35,060 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=122380.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:54:36,180 INFO [finetune.py:976] (6/7) Epoch 22, batch 2100, loss[loss=0.1577, simple_loss=0.2388, pruned_loss=0.0383, over 4763.00 frames. ], tot_loss[loss=0.1691, simple_loss=0.2403, pruned_loss=0.04892, over 957635.04 frames. ], batch size: 26, lr: 3.15e-03, grad_scale: 32.0 2023-04-27 16:55:40,953 INFO [finetune.py:976] (6/7) Epoch 22, batch 2150, loss[loss=0.2011, simple_loss=0.2744, pruned_loss=0.06391, over 4894.00 frames. ], tot_loss[loss=0.1709, simple_loss=0.2424, pruned_loss=0.0497, over 955379.80 frames. ], batch size: 35, lr: 3.15e-03, grad_scale: 32.0 2023-04-27 16:56:18,248 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.3208, 1.7971, 2.1926, 2.6767, 2.3108, 1.7382, 1.7414, 1.9469], device='cuda:6'), covar=tensor([0.3392, 0.3950, 0.1950, 0.2960, 0.2945, 0.2975, 0.4413, 0.2591], device='cuda:6'), in_proj_covar=tensor([0.0293, 0.0244, 0.0227, 0.0313, 0.0220, 0.0233, 0.0228, 0.0183], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:6') 2023-04-27 16:56:19,330 INFO [optim.py:369] (6/7) 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 16:56:27,940 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=122468.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:56:30,962 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=122473.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 16:56:42,040 INFO [finetune.py:976] (6/7) Epoch 22, batch 2200, loss[loss=0.1702, simple_loss=0.251, pruned_loss=0.04475, over 4810.00 frames. ], tot_loss[loss=0.173, simple_loss=0.2447, pruned_loss=0.05065, over 956273.80 frames. ], batch size: 45, lr: 3.15e-03, grad_scale: 32.0 2023-04-27 16:57:14,000 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.2428, 1.4547, 1.3820, 1.6921, 1.5280, 1.8179, 1.3640, 3.4212], device='cuda:6'), covar=tensor([0.0609, 0.0831, 0.0801, 0.1225, 0.0650, 0.0596, 0.0768, 0.0142], device='cuda:6'), in_proj_covar=tensor([0.0038, 0.0039, 0.0039, 0.0043, 0.0040, 0.0038, 0.0038, 0.0055], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0013, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:6') 2023-04-27 16:57:48,019 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=122529.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:57:49,743 INFO [finetune.py:976] (6/7) Epoch 22, batch 2250, loss[loss=0.1738, simple_loss=0.2606, pruned_loss=0.04353, over 4827.00 frames. ], tot_loss[loss=0.173, simple_loss=0.2454, pruned_loss=0.05036, over 956030.42 frames. ], batch size: 33, lr: 3.15e-03, grad_scale: 32.0 2023-04-27 16:58:20,295 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=122551.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:58:34,013 INFO [optim.py:369] (6/7) 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 16:58:56,474 INFO [finetune.py:976] (6/7) Epoch 22, batch 2300, loss[loss=0.1469, simple_loss=0.2232, pruned_loss=0.03534, over 4783.00 frames. ], tot_loss[loss=0.1723, simple_loss=0.2449, pruned_loss=0.04987, over 954709.30 frames. ], batch size: 29, lr: 3.15e-03, grad_scale: 32.0 2023-04-27 16:59:36,007 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=122606.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:00:08,779 INFO [finetune.py:976] (6/7) Epoch 22, batch 2350, loss[loss=0.1487, simple_loss=0.2248, pruned_loss=0.03627, over 4779.00 frames. ], tot_loss[loss=0.1714, simple_loss=0.2435, pruned_loss=0.04961, over 955453.05 frames. ], batch size: 29, lr: 3.15e-03, grad_scale: 32.0 2023-04-27 17:00:19,536 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0072, 1.8270, 2.0598, 2.4605, 2.3848, 1.9973, 1.6307, 2.0340], device='cuda:6'), covar=tensor([0.0734, 0.1031, 0.0623, 0.0476, 0.0512, 0.0724, 0.0773, 0.0569], device='cuda:6'), in_proj_covar=tensor([0.0186, 0.0201, 0.0182, 0.0174, 0.0176, 0.0178, 0.0152, 0.0177], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 17:00:40,760 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=122654.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:00:45,807 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.59 vs. limit=2.0 2023-04-27 17:00:46,226 INFO [optim.py:369] (6/7) 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 17:00:56,220 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-04-27 17:01:02,395 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.4840, 2.0730, 2.4001, 3.0175, 2.3787, 1.9478, 1.9834, 2.2603], device='cuda:6'), covar=tensor([0.2894, 0.3065, 0.1658, 0.1992, 0.2430, 0.2336, 0.3518, 0.2005], device='cuda:6'), in_proj_covar=tensor([0.0294, 0.0245, 0.0226, 0.0314, 0.0220, 0.0233, 0.0228, 0.0183], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:6') 2023-04-27 17:01:04,135 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=122675.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:01:14,120 INFO [finetune.py:976] (6/7) Epoch 22, batch 2400, loss[loss=0.1707, simple_loss=0.2368, pruned_loss=0.0523, over 4777.00 frames. ], tot_loss[loss=0.1694, simple_loss=0.2411, pruned_loss=0.04884, over 955254.35 frames. ], batch size: 26, lr: 3.15e-03, grad_scale: 32.0 2023-04-27 17:02:21,392 INFO [finetune.py:976] (6/7) Epoch 22, batch 2450, loss[loss=0.1714, simple_loss=0.2331, pruned_loss=0.05483, over 4836.00 frames. ], tot_loss[loss=0.1663, simple_loss=0.2376, pruned_loss=0.04751, over 954168.90 frames. ], batch size: 33, lr: 3.14e-03, grad_scale: 32.0 2023-04-27 17:02:28,220 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6394, 1.1634, 4.4917, 4.1993, 3.9061, 4.2079, 4.1187, 3.9209], device='cuda:6'), covar=tensor([0.6938, 0.6169, 0.0992, 0.1655, 0.0993, 0.1405, 0.1378, 0.1589], device='cuda:6'), in_proj_covar=tensor([0.0308, 0.0306, 0.0405, 0.0404, 0.0346, 0.0409, 0.0311, 0.0364], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-27 17:02:40,022 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7798, 1.5673, 2.0901, 2.0954, 1.5833, 1.4534, 1.7472, 1.0932], device='cuda:6'), covar=tensor([0.0433, 0.0618, 0.0401, 0.0603, 0.0674, 0.1067, 0.0568, 0.0612], device='cuda:6'), in_proj_covar=tensor([0.0069, 0.0068, 0.0066, 0.0068, 0.0075, 0.0095, 0.0073, 0.0065], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-27 17:02:42,826 INFO [optim.py:369] (6/7) 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 17:02:48,962 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=122773.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 17:02:54,348 INFO [finetune.py:976] (6/7) Epoch 22, batch 2500, loss[loss=0.2277, simple_loss=0.2957, pruned_loss=0.07991, over 4829.00 frames. ], tot_loss[loss=0.167, simple_loss=0.2379, pruned_loss=0.04806, over 953259.65 frames. ], batch size: 38, lr: 3.14e-03, grad_scale: 32.0 2023-04-27 17:03:00,572 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6290, 1.5767, 1.7328, 2.0243, 2.0568, 1.6111, 1.2924, 1.7822], device='cuda:6'), covar=tensor([0.0820, 0.1058, 0.0746, 0.0598, 0.0612, 0.0819, 0.0791, 0.0597], device='cuda:6'), in_proj_covar=tensor([0.0188, 0.0202, 0.0183, 0.0175, 0.0177, 0.0180, 0.0153, 0.0178], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 17:03:19,988 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7751, 1.1581, 1.7917, 2.1760, 1.8135, 1.7168, 1.7859, 1.7229], device='cuda:6'), covar=tensor([0.4335, 0.6674, 0.6387, 0.5505, 0.5690, 0.7459, 0.7231, 0.8954], device='cuda:6'), in_proj_covar=tensor([0.0429, 0.0413, 0.0506, 0.0505, 0.0457, 0.0488, 0.0497, 0.0504], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 17:03:21,721 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=122821.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:03:23,548 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=122824.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:03:28,389 INFO [finetune.py:976] (6/7) Epoch 22, batch 2550, loss[loss=0.1399, simple_loss=0.2129, pruned_loss=0.03342, over 4801.00 frames. ], tot_loss[loss=0.1704, simple_loss=0.2417, pruned_loss=0.04953, over 952110.01 frames. ], batch size: 25, lr: 3.14e-03, grad_scale: 32.0 2023-04-27 17:03:40,596 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=122851.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:03:48,778 INFO [optim.py:369] (6/7) 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 17:04:01,400 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=122872.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:04:13,219 INFO [finetune.py:976] (6/7) Epoch 22, batch 2600, loss[loss=0.2357, simple_loss=0.3008, pruned_loss=0.08532, over 4897.00 frames. ], tot_loss[loss=0.1724, simple_loss=0.2441, pruned_loss=0.05031, over 951436.35 frames. ], batch size: 35, lr: 3.14e-03, grad_scale: 32.0 2023-04-27 17:04:16,394 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.3039, 1.5714, 1.4602, 1.7483, 1.6371, 1.9227, 1.4162, 3.5521], device='cuda:6'), covar=tensor([0.0556, 0.0776, 0.0752, 0.1146, 0.0617, 0.0592, 0.0763, 0.0122], device='cuda:6'), in_proj_covar=tensor([0.0037, 0.0038, 0.0039, 0.0043, 0.0039, 0.0038, 0.0037, 0.0055], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:6') 2023-04-27 17:04:34,934 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=122899.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:05:19,792 INFO [finetune.py:976] (6/7) Epoch 22, batch 2650, loss[loss=0.2103, simple_loss=0.2767, pruned_loss=0.07196, over 4888.00 frames. ], tot_loss[loss=0.1726, simple_loss=0.2449, pruned_loss=0.05016, over 952191.11 frames. ], batch size: 32, lr: 3.14e-03, grad_scale: 32.0 2023-04-27 17:05:20,539 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=122933.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:05:32,281 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.1756, 1.3919, 1.6952, 1.8068, 1.7189, 1.7823, 1.7188, 1.7417], device='cuda:6'), covar=tensor([0.3783, 0.5063, 0.4180, 0.3953, 0.4850, 0.6332, 0.4466, 0.4516], device='cuda:6'), in_proj_covar=tensor([0.0339, 0.0372, 0.0324, 0.0338, 0.0346, 0.0395, 0.0356, 0.0330], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 17:06:00,529 INFO [optim.py:369] (6/7) 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 17:06:15,897 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=122975.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:06:26,131 INFO [finetune.py:976] (6/7) Epoch 22, batch 2700, loss[loss=0.1284, simple_loss=0.2136, pruned_loss=0.02162, over 4883.00 frames. ], tot_loss[loss=0.1713, simple_loss=0.2441, pruned_loss=0.04927, over 952674.13 frames. ], batch size: 43, lr: 3.14e-03, grad_scale: 32.0 2023-04-27 17:06:33,085 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8697, 2.2725, 1.9310, 2.3418, 1.7282, 1.9685, 2.0364, 1.4864], device='cuda:6'), covar=tensor([0.1933, 0.1250, 0.0926, 0.1237, 0.3238, 0.1236, 0.1874, 0.2690], device='cuda:6'), in_proj_covar=tensor([0.0285, 0.0303, 0.0217, 0.0278, 0.0315, 0.0256, 0.0249, 0.0263], device='cuda:6'), out_proj_covar=tensor([1.1387e-04, 1.1999e-04, 8.5841e-05, 1.0968e-04, 1.2764e-04, 1.0107e-04, 1.0033e-04, 1.0402e-04], device='cuda:6') 2023-04-27 17:06:33,960 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.10 vs. limit=2.0 2023-04-27 17:06:56,566 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=123023.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:07:02,522 INFO [finetune.py:976] (6/7) Epoch 22, batch 2750, loss[loss=0.2043, simple_loss=0.2734, pruned_loss=0.06757, over 4849.00 frames. ], tot_loss[loss=0.17, simple_loss=0.2419, pruned_loss=0.04903, over 954399.68 frames. ], batch size: 47, lr: 3.14e-03, grad_scale: 32.0 2023-04-27 17:07:28,109 INFO [optim.py:369] (6/7) 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 17:07:53,161 INFO [finetune.py:976] (6/7) Epoch 22, batch 2800, loss[loss=0.1518, simple_loss=0.2265, pruned_loss=0.03861, over 4918.00 frames. ], tot_loss[loss=0.167, simple_loss=0.2385, pruned_loss=0.04781, over 954013.07 frames. ], batch size: 36, lr: 3.14e-03, grad_scale: 32.0 2023-04-27 17:08:39,218 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.2689, 1.3867, 3.7633, 3.5204, 3.3138, 3.5866, 3.5490, 3.3211], device='cuda:6'), covar=tensor([0.7134, 0.5324, 0.1154, 0.1858, 0.1205, 0.2353, 0.2033, 0.1493], device='cuda:6'), in_proj_covar=tensor([0.0309, 0.0306, 0.0407, 0.0406, 0.0348, 0.0410, 0.0313, 0.0365], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-27 17:08:48,178 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=123124.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:08:49,475 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8071, 1.2607, 1.8639, 2.2306, 1.8921, 1.7566, 1.8019, 1.7720], device='cuda:6'), covar=tensor([0.4166, 0.6550, 0.5575, 0.5415, 0.5278, 0.7241, 0.7086, 0.9148], device='cuda:6'), in_proj_covar=tensor([0.0431, 0.0416, 0.0509, 0.0509, 0.0459, 0.0491, 0.0499, 0.0508], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 17:08:53,533 INFO [finetune.py:976] (6/7) Epoch 22, batch 2850, loss[loss=0.1848, simple_loss=0.2562, pruned_loss=0.05667, over 4760.00 frames. ], tot_loss[loss=0.167, simple_loss=0.238, pruned_loss=0.048, over 954991.56 frames. ], batch size: 26, lr: 3.14e-03, grad_scale: 32.0 2023-04-27 17:09:12,897 INFO [optim.py:369] (6/7) 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 17:09:19,864 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=123172.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:09:27,497 INFO [finetune.py:976] (6/7) Epoch 22, batch 2900, loss[loss=0.2174, simple_loss=0.288, pruned_loss=0.07344, over 4834.00 frames. ], tot_loss[loss=0.1698, simple_loss=0.2411, pruned_loss=0.04926, over 954732.75 frames. ], batch size: 51, lr: 3.14e-03, grad_scale: 32.0 2023-04-27 17:09:29,231 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-04-27 17:09:48,016 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.9135, 2.1746, 2.1769, 2.2943, 2.0446, 2.1524, 2.2678, 2.2181], device='cuda:6'), covar=tensor([0.4113, 0.6008, 0.4527, 0.4718, 0.5907, 0.7176, 0.5854, 0.5200], device='cuda:6'), in_proj_covar=tensor([0.0340, 0.0373, 0.0326, 0.0340, 0.0348, 0.0396, 0.0357, 0.0331], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 17:10:09,348 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=123228.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:10:11,741 INFO [finetune.py:976] (6/7) Epoch 22, batch 2950, loss[loss=0.1819, simple_loss=0.2523, pruned_loss=0.05579, over 4793.00 frames. ], tot_loss[loss=0.1717, simple_loss=0.2437, pruned_loss=0.04983, over 955212.40 frames. ], batch size: 25, lr: 3.14e-03, grad_scale: 32.0 2023-04-27 17:10:55,130 INFO [optim.py:369] (6/7) 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 17:10:56,484 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.4448, 2.8168, 0.9659, 1.6116, 2.4669, 1.4580, 4.0870, 2.1110], device='cuda:6'), covar=tensor([0.0644, 0.0753, 0.0924, 0.1280, 0.0491, 0.1030, 0.0189, 0.0583], device='cuda:6'), in_proj_covar=tensor([0.0050, 0.0064, 0.0048, 0.0046, 0.0050, 0.0052, 0.0074, 0.0051], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:6') 2023-04-27 17:11:18,373 INFO [finetune.py:976] (6/7) Epoch 22, batch 3000, loss[loss=0.1988, simple_loss=0.2712, pruned_loss=0.06323, over 4852.00 frames. ], tot_loss[loss=0.1739, simple_loss=0.2461, pruned_loss=0.05087, over 953209.94 frames. ], batch size: 44, lr: 3.14e-03, grad_scale: 32.0 2023-04-27 17:11:18,373 INFO [finetune.py:1001] (6/7) Computing validation loss 2023-04-27 17:11:29,116 INFO [finetune.py:1010] (6/7) 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] (6/7) Maximum memory allocated so far is 6365MB 2023-04-27 17:11:49,337 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.3472, 1.7485, 2.1663, 2.6968, 2.1844, 1.7122, 1.4867, 1.9758], device='cuda:6'), covar=tensor([0.2936, 0.3036, 0.1545, 0.1948, 0.2536, 0.2535, 0.3988, 0.2128], device='cuda:6'), in_proj_covar=tensor([0.0293, 0.0245, 0.0226, 0.0314, 0.0220, 0.0232, 0.0228, 0.0184], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:6') 2023-04-27 17:11:52,036 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.91 vs. limit=5.0 2023-04-27 17:12:11,394 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=123314.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:12:33,514 INFO [finetune.py:976] (6/7) Epoch 22, batch 3050, loss[loss=0.1622, simple_loss=0.2472, pruned_loss=0.0386, over 4735.00 frames. ], tot_loss[loss=0.175, simple_loss=0.2475, pruned_loss=0.05122, over 954482.76 frames. ], batch size: 54, lr: 3.14e-03, grad_scale: 32.0 2023-04-27 17:12:51,668 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5150, 1.6450, 1.4236, 1.1143, 1.1810, 1.1738, 1.3951, 1.1190], device='cuda:6'), covar=tensor([0.1545, 0.1248, 0.1283, 0.1574, 0.2090, 0.1746, 0.0922, 0.1838], device='cuda:6'), in_proj_covar=tensor([0.0198, 0.0213, 0.0170, 0.0205, 0.0201, 0.0187, 0.0156, 0.0187], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:6') 2023-04-27 17:13:01,591 INFO [optim.py:369] (6/7) 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 17:13:03,077 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.52 vs. limit=5.0 2023-04-27 17:13:14,322 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=123375.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:13:24,032 INFO [finetune.py:976] (6/7) Epoch 22, batch 3100, loss[loss=0.1507, simple_loss=0.2269, pruned_loss=0.03726, over 4751.00 frames. ], tot_loss[loss=0.1738, simple_loss=0.246, pruned_loss=0.05074, over 954545.14 frames. ], batch size: 27, lr: 3.14e-03, grad_scale: 32.0 2023-04-27 17:14:29,437 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=123428.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:14:31,785 INFO [finetune.py:976] (6/7) Epoch 22, batch 3150, loss[loss=0.1483, simple_loss=0.2228, pruned_loss=0.0369, over 4796.00 frames. ], tot_loss[loss=0.1713, simple_loss=0.2428, pruned_loss=0.04988, over 954634.68 frames. ], batch size: 29, lr: 3.14e-03, grad_scale: 32.0 2023-04-27 17:15:16,034 INFO [optim.py:369] (6/7) 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] (6/7) Epoch 22, batch 3200, loss[loss=0.183, simple_loss=0.253, pruned_loss=0.05647, over 4822.00 frames. ], tot_loss[loss=0.1687, simple_loss=0.2398, pruned_loss=0.04882, over 953097.01 frames. ], batch size: 40, lr: 3.14e-03, grad_scale: 32.0 2023-04-27 17:15:48,889 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=123489.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:15:56,366 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([3.6161, 3.5779, 2.6882, 4.1965, 3.6189, 3.5921, 1.5327, 3.6133], device='cuda:6'), covar=tensor([0.1898, 0.1416, 0.3851, 0.1720, 0.3513, 0.1861, 0.6050, 0.2375], device='cuda:6'), in_proj_covar=tensor([0.0245, 0.0218, 0.0254, 0.0308, 0.0299, 0.0248, 0.0279, 0.0273], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 17:16:43,112 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=123528.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:16:51,254 INFO [finetune.py:976] (6/7) Epoch 22, batch 3250, loss[loss=0.1748, simple_loss=0.2488, pruned_loss=0.05041, over 4902.00 frames. ], tot_loss[loss=0.1691, simple_loss=0.2405, pruned_loss=0.04883, over 953692.77 frames. ], batch size: 35, lr: 3.14e-03, grad_scale: 64.0 2023-04-27 17:17:26,394 INFO [optim.py:369] (6/7) 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 17:17:40,419 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=123576.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:17:42,944 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9148, 1.3929, 1.5387, 1.5072, 2.0468, 1.6511, 1.3612, 1.4738], device='cuda:6'), covar=tensor([0.1632, 0.1721, 0.2066, 0.1661, 0.0934, 0.1513, 0.2049, 0.2224], device='cuda:6'), in_proj_covar=tensor([0.0314, 0.0312, 0.0352, 0.0289, 0.0327, 0.0309, 0.0302, 0.0374], device='cuda:6'), out_proj_covar=tensor([6.4550e-05, 6.4554e-05, 7.4192e-05, 5.8248e-05, 6.7343e-05, 6.4769e-05, 6.3102e-05, 7.9450e-05], device='cuda:6') 2023-04-27 17:17:44,046 INFO [finetune.py:976] (6/7) Epoch 22, batch 3300, loss[loss=0.1796, simple_loss=0.2489, pruned_loss=0.05515, over 4905.00 frames. ], tot_loss[loss=0.1727, simple_loss=0.2447, pruned_loss=0.05034, over 952960.57 frames. ], batch size: 37, lr: 3.14e-03, grad_scale: 64.0 2023-04-27 17:18:05,724 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([3.7439, 3.7244, 2.8195, 4.3238, 3.8153, 3.6635, 1.7402, 3.7925], device='cuda:6'), covar=tensor([0.1598, 0.1185, 0.3352, 0.1486, 0.2726, 0.1697, 0.5417, 0.2140], device='cuda:6'), in_proj_covar=tensor([0.0244, 0.0217, 0.0253, 0.0307, 0.0297, 0.0246, 0.0277, 0.0272], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 17:18:17,759 INFO [finetune.py:976] (6/7) Epoch 22, batch 3350, loss[loss=0.1335, simple_loss=0.2005, pruned_loss=0.03322, over 4708.00 frames. ], tot_loss[loss=0.1724, simple_loss=0.2455, pruned_loss=0.04971, over 954242.71 frames. ], batch size: 23, lr: 3.14e-03, grad_scale: 64.0 2023-04-27 17:18:40,220 INFO [optim.py:369] (6/7) 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 17:18:43,916 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=123670.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:18:48,328 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.11 vs. limit=2.0 2023-04-27 17:18:51,586 INFO [finetune.py:976] (6/7) Epoch 22, batch 3400, loss[loss=0.1491, simple_loss=0.2284, pruned_loss=0.03491, over 4801.00 frames. ], tot_loss[loss=0.1738, simple_loss=0.2468, pruned_loss=0.05042, over 954524.18 frames. ], batch size: 51, lr: 3.14e-03, grad_scale: 32.0 2023-04-27 17:18:56,550 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0969, 4.3850, 0.8145, 2.1934, 2.5872, 2.8454, 2.5646, 1.0373], device='cuda:6'), covar=tensor([0.1261, 0.1012, 0.2252, 0.1312, 0.0932, 0.1113, 0.1410, 0.2041], device='cuda:6'), in_proj_covar=tensor([0.0117, 0.0242, 0.0139, 0.0121, 0.0133, 0.0153, 0.0117, 0.0120], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-27 17:19:24,043 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8040, 2.3129, 1.8123, 1.7187, 1.3512, 1.3629, 1.8935, 1.3433], device='cuda:6'), covar=tensor([0.1637, 0.1320, 0.1372, 0.1681, 0.2377, 0.2022, 0.0996, 0.2033], device='cuda:6'), in_proj_covar=tensor([0.0197, 0.0212, 0.0169, 0.0204, 0.0200, 0.0186, 0.0156, 0.0186], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:6') 2023-04-27 17:19:25,128 INFO [finetune.py:976] (6/7) Epoch 22, batch 3450, loss[loss=0.1611, simple_loss=0.2139, pruned_loss=0.05417, over 4386.00 frames. ], tot_loss[loss=0.173, simple_loss=0.2459, pruned_loss=0.05006, over 954741.40 frames. ], batch size: 19, lr: 3.14e-03, grad_scale: 32.0 2023-04-27 17:19:28,420 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-04-27 17:19:57,042 INFO [optim.py:369] (6/7) 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 17:20:21,187 INFO [finetune.py:976] (6/7) Epoch 22, batch 3500, loss[loss=0.1603, simple_loss=0.2278, pruned_loss=0.04642, over 4863.00 frames. ], tot_loss[loss=0.1713, simple_loss=0.2438, pruned_loss=0.0494, over 952887.05 frames. ], batch size: 31, lr: 3.14e-03, grad_scale: 32.0 2023-04-27 17:20:28,350 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=123784.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:20:30,864 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.2752, 1.5407, 1.8144, 1.9315, 1.8355, 1.9187, 1.8433, 1.8759], device='cuda:6'), covar=tensor([0.3872, 0.5337, 0.4509, 0.4334, 0.5474, 0.6763, 0.4815, 0.4348], device='cuda:6'), in_proj_covar=tensor([0.0340, 0.0376, 0.0327, 0.0340, 0.0349, 0.0397, 0.0359, 0.0332], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 17:21:22,866 INFO [finetune.py:976] (6/7) Epoch 22, batch 3550, loss[loss=0.1712, simple_loss=0.2319, pruned_loss=0.05522, over 4819.00 frames. ], tot_loss[loss=0.1687, simple_loss=0.2407, pruned_loss=0.04833, over 953395.33 frames. ], batch size: 39, lr: 3.14e-03, grad_scale: 32.0 2023-04-27 17:21:43,412 INFO [optim.py:369] (6/7) 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 17:21:49,908 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=123872.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:21:56,779 INFO [finetune.py:976] (6/7) Epoch 22, batch 3600, loss[loss=0.1576, simple_loss=0.2183, pruned_loss=0.04841, over 4744.00 frames. ], tot_loss[loss=0.1675, simple_loss=0.2384, pruned_loss=0.04833, over 953573.99 frames. ], batch size: 23, lr: 3.14e-03, grad_scale: 32.0 2023-04-27 17:22:10,847 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.64 vs. limit=2.0 2023-04-27 17:22:13,025 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.49 vs. limit=2.0 2023-04-27 17:22:30,655 INFO [finetune.py:976] (6/7) Epoch 22, batch 3650, loss[loss=0.1775, simple_loss=0.2607, pruned_loss=0.04717, over 4817.00 frames. ], tot_loss[loss=0.1693, simple_loss=0.2404, pruned_loss=0.04911, over 953070.98 frames. ], batch size: 40, lr: 3.14e-03, grad_scale: 32.0 2023-04-27 17:22:31,426 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=123933.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:22:53,914 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.57 vs. limit=2.0 2023-04-27 17:23:01,062 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.2429, 2.8843, 1.0036, 1.6128, 2.3351, 1.2697, 3.8807, 1.9476], device='cuda:6'), covar=tensor([0.0690, 0.0794, 0.0896, 0.1172, 0.0475, 0.1034, 0.0173, 0.0604], device='cuda:6'), in_proj_covar=tensor([0.0050, 0.0064, 0.0047, 0.0046, 0.0049, 0.0051, 0.0073, 0.0050], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0009, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0007], device='cuda:6') 2023-04-27 17:23:14,431 INFO [optim.py:369] (6/7) 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 17:23:23,376 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=123970.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:23:37,658 INFO [finetune.py:976] (6/7) Epoch 22, batch 3700, loss[loss=0.1618, simple_loss=0.2297, pruned_loss=0.04692, over 4170.00 frames. ], tot_loss[loss=0.1717, simple_loss=0.2437, pruned_loss=0.04988, over 954450.28 frames. ], batch size: 18, lr: 3.14e-03, grad_scale: 32.0 2023-04-27 17:24:22,558 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=124018.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:24:41,985 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6406, 0.7370, 1.5361, 1.9801, 1.6712, 1.5267, 1.5799, 1.5772], device='cuda:6'), covar=tensor([0.4084, 0.6150, 0.5881, 0.5519, 0.5677, 0.6722, 0.7170, 0.7649], device='cuda:6'), in_proj_covar=tensor([0.0430, 0.0413, 0.0509, 0.0505, 0.0460, 0.0489, 0.0495, 0.0505], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 17:24:44,187 INFO [finetune.py:976] (6/7) Epoch 22, batch 3750, loss[loss=0.1393, simple_loss=0.2169, pruned_loss=0.0308, over 4759.00 frames. ], tot_loss[loss=0.1721, simple_loss=0.2445, pruned_loss=0.04987, over 953295.59 frames. ], batch size: 28, lr: 3.13e-03, grad_scale: 32.0 2023-04-27 17:24:52,246 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=124036.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:25:08,756 INFO [zipformer.py:1188] (6/7) 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] (6/7) 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 17:25:22,232 INFO [finetune.py:976] (6/7) Epoch 22, batch 3800, loss[loss=0.1501, simple_loss=0.2192, pruned_loss=0.04054, over 4923.00 frames. ], tot_loss[loss=0.1735, simple_loss=0.2461, pruned_loss=0.05044, over 953210.51 frames. ], batch size: 33, lr: 3.13e-03, grad_scale: 32.0 2023-04-27 17:25:24,008 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=124084.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:25:32,955 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=124097.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:25:49,425 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=124123.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:25:56,295 INFO [finetune.py:976] (6/7) Epoch 22, batch 3850, loss[loss=0.1407, simple_loss=0.2249, pruned_loss=0.02828, over 4898.00 frames. ], tot_loss[loss=0.1712, simple_loss=0.2439, pruned_loss=0.04925, over 954988.81 frames. ], batch size: 37, lr: 3.13e-03, grad_scale: 32.0 2023-04-27 17:25:56,356 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=124132.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:26:03,403 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5033, 1.4118, 1.7808, 1.8496, 1.4223, 1.2198, 1.6044, 0.9233], device='cuda:6'), covar=tensor([0.0624, 0.0616, 0.0454, 0.0532, 0.0758, 0.1066, 0.0545, 0.0646], device='cuda:6'), in_proj_covar=tensor([0.0069, 0.0068, 0.0067, 0.0068, 0.0075, 0.0096, 0.0073, 0.0065], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-27 17:26:17,750 INFO [optim.py:369] (6/7) 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 17:26:35,209 INFO [finetune.py:976] (6/7) Epoch 22, batch 3900, loss[loss=0.1401, simple_loss=0.2167, pruned_loss=0.0318, over 4905.00 frames. ], tot_loss[loss=0.1694, simple_loss=0.2413, pruned_loss=0.04875, over 955544.20 frames. ], batch size: 43, lr: 3.13e-03, grad_scale: 32.0 2023-04-27 17:26:42,958 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.0899, 1.4186, 1.3663, 1.7381, 1.6132, 1.5542, 1.3550, 2.4432], device='cuda:6'), covar=tensor([0.0610, 0.0777, 0.0746, 0.1140, 0.0578, 0.0431, 0.0710, 0.0231], device='cuda:6'), in_proj_covar=tensor([0.0037, 0.0038, 0.0039, 0.0043, 0.0040, 0.0037, 0.0038, 0.0055], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:6') 2023-04-27 17:27:26,669 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.4827, 1.0993, 1.2125, 1.1778, 1.6143, 1.3294, 1.0951, 1.1610], device='cuda:6'), covar=tensor([0.1357, 0.1217, 0.1660, 0.1270, 0.0739, 0.1328, 0.1805, 0.2289], device='cuda:6'), in_proj_covar=tensor([0.0315, 0.0313, 0.0353, 0.0290, 0.0327, 0.0308, 0.0302, 0.0374], device='cuda:6'), out_proj_covar=tensor([6.4712e-05, 6.4776e-05, 7.4318e-05, 5.8352e-05, 6.7371e-05, 6.4659e-05, 6.3087e-05, 7.9301e-05], device='cuda:6') 2023-04-27 17:27:38,100 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=124228.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:27:40,955 INFO [finetune.py:976] (6/7) Epoch 22, batch 3950, loss[loss=0.1432, simple_loss=0.2142, pruned_loss=0.03603, over 4800.00 frames. ], tot_loss[loss=0.167, simple_loss=0.2384, pruned_loss=0.04781, over 953248.33 frames. ], batch size: 29, lr: 3.13e-03, grad_scale: 32.0 2023-04-27 17:28:07,729 INFO [optim.py:369] (6/7) 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 17:28:13,200 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=124272.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:28:19,117 INFO [finetune.py:976] (6/7) Epoch 22, batch 4000, loss[loss=0.1665, simple_loss=0.245, pruned_loss=0.04395, over 4803.00 frames. ], tot_loss[loss=0.1671, simple_loss=0.238, pruned_loss=0.04806, over 952871.74 frames. ], batch size: 45, lr: 3.13e-03, grad_scale: 32.0 2023-04-27 17:29:09,167 INFO [finetune.py:976] (6/7) Epoch 22, batch 4050, loss[loss=0.1737, simple_loss=0.2368, pruned_loss=0.05534, over 4711.00 frames. ], tot_loss[loss=0.1708, simple_loss=0.2424, pruned_loss=0.04964, over 954210.21 frames. ], batch size: 23, lr: 3.13e-03, grad_scale: 32.0 2023-04-27 17:29:09,935 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=124333.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:29:38,968 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=124348.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 17:29:39,009 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.2714, 1.7004, 2.1163, 2.6145, 2.1914, 1.6759, 1.4547, 2.0606], device='cuda:6'), covar=tensor([0.3221, 0.3323, 0.1805, 0.2393, 0.2626, 0.2817, 0.4185, 0.1882], device='cuda:6'), in_proj_covar=tensor([0.0293, 0.0244, 0.0226, 0.0314, 0.0221, 0.0233, 0.0228, 0.0184], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:6') 2023-04-27 17:29:54,262 INFO [optim.py:369] (6/7) 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 17:30:17,169 INFO [finetune.py:976] (6/7) Epoch 22, batch 4100, loss[loss=0.1315, simple_loss=0.1961, pruned_loss=0.03347, over 3927.00 frames. ], tot_loss[loss=0.1707, simple_loss=0.2429, pruned_loss=0.04923, over 954288.31 frames. ], batch size: 17, lr: 3.13e-03, grad_scale: 32.0 2023-04-27 17:30:23,879 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-04-27 17:30:34,615 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=124392.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:30:36,209 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.34 vs. limit=5.0 2023-04-27 17:30:58,108 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=124409.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 17:31:09,013 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=124418.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:31:27,632 INFO [finetune.py:976] (6/7) Epoch 22, batch 4150, loss[loss=0.2025, simple_loss=0.2697, pruned_loss=0.06762, over 4802.00 frames. ], tot_loss[loss=0.1719, simple_loss=0.2449, pruned_loss=0.04946, over 955541.09 frames. ], batch size: 45, lr: 3.13e-03, grad_scale: 32.0 2023-04-27 17:31:27,749 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=124432.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:31:29,621 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.1409, 2.7455, 1.0555, 1.5760, 2.1206, 1.3370, 3.4931, 1.9582], device='cuda:6'), covar=tensor([0.0661, 0.0577, 0.0766, 0.1237, 0.0491, 0.0976, 0.0301, 0.0550], device='cuda:6'), in_proj_covar=tensor([0.0050, 0.0064, 0.0047, 0.0046, 0.0049, 0.0051, 0.0073, 0.0050], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0007], device='cuda:6') 2023-04-27 17:32:00,112 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.49 vs. limit=2.0 2023-04-27 17:32:10,278 INFO [optim.py:369] (6/7) 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 17:32:33,359 INFO [finetune.py:976] (6/7) Epoch 22, batch 4200, loss[loss=0.1483, simple_loss=0.2304, pruned_loss=0.03307, over 4928.00 frames. ], tot_loss[loss=0.1718, simple_loss=0.245, pruned_loss=0.04924, over 954408.85 frames. ], batch size: 33, lr: 3.13e-03, grad_scale: 32.0 2023-04-27 17:32:36,517 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([3.5046, 3.5722, 2.4995, 4.1188, 3.6076, 3.5584, 1.4267, 3.4907], device='cuda:6'), covar=tensor([0.1959, 0.1293, 0.3595, 0.2032, 0.2915, 0.1865, 0.6309, 0.2503], device='cuda:6'), in_proj_covar=tensor([0.0244, 0.0219, 0.0252, 0.0307, 0.0298, 0.0247, 0.0277, 0.0272], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 17:32:45,614 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=124493.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:33:09,258 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=124528.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:33:11,585 INFO [finetune.py:976] (6/7) Epoch 22, batch 4250, loss[loss=0.1661, simple_loss=0.2194, pruned_loss=0.05644, over 4895.00 frames. ], tot_loss[loss=0.1705, simple_loss=0.2434, pruned_loss=0.04877, over 956356.89 frames. ], batch size: 35, lr: 3.13e-03, grad_scale: 32.0 2023-04-27 17:33:33,184 INFO [optim.py:369] (6/7) 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 17:33:37,051 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.1320, 2.5959, 2.1819, 2.5187, 1.8772, 2.1750, 2.1858, 1.6652], device='cuda:6'), covar=tensor([0.1794, 0.0949, 0.0808, 0.1044, 0.2928, 0.1148, 0.1846, 0.2584], device='cuda:6'), in_proj_covar=tensor([0.0287, 0.0303, 0.0219, 0.0279, 0.0318, 0.0259, 0.0252, 0.0265], device='cuda:6'), out_proj_covar=tensor([1.1479e-04, 1.1995e-04, 8.6577e-05, 1.1018e-04, 1.2863e-04, 1.0230e-04, 1.0147e-04, 1.0454e-04], device='cuda:6') 2023-04-27 17:33:40,672 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=124576.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:33:44,719 INFO [finetune.py:976] (6/7) Epoch 22, batch 4300, loss[loss=0.1941, simple_loss=0.2614, pruned_loss=0.06338, over 4900.00 frames. ], tot_loss[loss=0.1695, simple_loss=0.2421, pruned_loss=0.04842, over 958103.74 frames. ], batch size: 36, lr: 3.13e-03, grad_scale: 32.0 2023-04-27 17:34:10,659 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8864, 2.3547, 1.9122, 1.7678, 1.3305, 1.4349, 2.0004, 1.3308], device='cuda:6'), covar=tensor([0.1694, 0.1433, 0.1423, 0.1810, 0.2358, 0.2030, 0.0997, 0.2097], device='cuda:6'), in_proj_covar=tensor([0.0198, 0.0212, 0.0169, 0.0205, 0.0201, 0.0186, 0.0156, 0.0187], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:6') 2023-04-27 17:34:15,439 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=124628.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:34:17,830 INFO [finetune.py:976] (6/7) Epoch 22, batch 4350, loss[loss=0.1561, simple_loss=0.2218, pruned_loss=0.04516, over 4788.00 frames. ], tot_loss[loss=0.1666, simple_loss=0.2383, pruned_loss=0.04747, over 958697.07 frames. ], batch size: 29, lr: 3.13e-03, grad_scale: 32.0 2023-04-27 17:34:22,056 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=124638.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:34:38,877 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=124663.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:34:39,363 INFO [optim.py:369] (6/7) 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 17:34:46,931 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.49 vs. limit=2.0 2023-04-27 17:34:51,255 INFO [finetune.py:976] (6/7) Epoch 22, batch 4400, loss[loss=0.1825, simple_loss=0.2645, pruned_loss=0.05026, over 4912.00 frames. ], tot_loss[loss=0.167, simple_loss=0.2389, pruned_loss=0.0476, over 958342.43 frames. ], batch size: 37, lr: 3.13e-03, grad_scale: 32.0 2023-04-27 17:34:56,787 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5957, 1.4068, 4.5225, 4.2348, 3.9981, 4.4018, 4.2104, 3.9829], device='cuda:6'), covar=tensor([0.7128, 0.6562, 0.1104, 0.1774, 0.1189, 0.2420, 0.1392, 0.1739], device='cuda:6'), in_proj_covar=tensor([0.0308, 0.0304, 0.0406, 0.0406, 0.0347, 0.0408, 0.0311, 0.0364], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-27 17:34:58,060 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=124692.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:35:02,327 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=124699.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:35:11,387 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=124704.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 17:35:33,058 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=124718.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:35:33,795 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-04-27 17:35:42,951 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=124724.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:35:53,108 INFO [finetune.py:976] (6/7) Epoch 22, batch 4450, loss[loss=0.1732, simple_loss=0.2471, pruned_loss=0.04967, over 4922.00 frames. ], tot_loss[loss=0.1691, simple_loss=0.2418, pruned_loss=0.04818, over 958251.87 frames. ], batch size: 38, lr: 3.13e-03, grad_scale: 32.0 2023-04-27 17:36:03,832 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=124740.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:36:30,398 INFO [optim.py:369] (6/7) 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 17:36:38,228 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=124766.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:37:00,154 INFO [finetune.py:976] (6/7) Epoch 22, batch 4500, loss[loss=0.1349, simple_loss=0.2008, pruned_loss=0.03455, over 4751.00 frames. ], tot_loss[loss=0.1706, simple_loss=0.2435, pruned_loss=0.04884, over 957306.68 frames. ], batch size: 23, lr: 3.13e-03, grad_scale: 32.0 2023-04-27 17:37:03,944 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=124788.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:37:47,351 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=124820.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:38:07,814 INFO [finetune.py:976] (6/7) Epoch 22, batch 4550, loss[loss=0.1624, simple_loss=0.2473, pruned_loss=0.03875, over 4772.00 frames. ], tot_loss[loss=0.1717, simple_loss=0.2451, pruned_loss=0.0492, over 957565.27 frames. ], batch size: 28, lr: 3.13e-03, grad_scale: 32.0 2023-04-27 17:38:40,616 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=3.88 vs. limit=5.0 2023-04-27 17:38:49,605 INFO [optim.py:369] (6/7) 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 17:39:13,965 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=124881.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:39:14,476 INFO [finetune.py:976] (6/7) Epoch 22, batch 4600, loss[loss=0.1717, simple_loss=0.2379, pruned_loss=0.05276, over 4824.00 frames. ], tot_loss[loss=0.1714, simple_loss=0.2445, pruned_loss=0.04915, over 955919.44 frames. ], batch size: 30, lr: 3.13e-03, grad_scale: 32.0 2023-04-27 17:39:25,727 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2023-04-27 17:39:57,020 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9091, 1.4201, 1.7355, 1.7386, 1.7189, 1.4242, 0.9024, 1.4323], device='cuda:6'), covar=tensor([0.3470, 0.3317, 0.1877, 0.2316, 0.2744, 0.2692, 0.4132, 0.2159], device='cuda:6'), in_proj_covar=tensor([0.0293, 0.0244, 0.0225, 0.0314, 0.0221, 0.0233, 0.0227, 0.0184], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:6') 2023-04-27 17:39:58,092 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0942, 2.6543, 1.0384, 1.4050, 1.9302, 1.2362, 3.0661, 1.5918], device='cuda:6'), covar=tensor([0.0711, 0.0531, 0.0691, 0.1168, 0.0479, 0.0996, 0.0229, 0.0650], device='cuda:6'), in_proj_covar=tensor([0.0050, 0.0064, 0.0047, 0.0046, 0.0050, 0.0051, 0.0073, 0.0050], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0007], device='cuda:6') 2023-04-27 17:40:00,432 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=124928.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:40:03,270 INFO [finetune.py:976] (6/7) Epoch 22, batch 4650, loss[loss=0.1971, simple_loss=0.2601, pruned_loss=0.06709, over 4923.00 frames. ], tot_loss[loss=0.1699, simple_loss=0.2426, pruned_loss=0.0486, over 956393.50 frames. ], batch size: 38, lr: 3.13e-03, grad_scale: 32.0 2023-04-27 17:40:23,402 INFO [optim.py:369] (6/7) 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 17:40:31,673 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 22, batch 4700, loss[loss=0.1433, simple_loss=0.2236, pruned_loss=0.03145, over 4837.00 frames. ], tot_loss[loss=0.168, simple_loss=0.2399, pruned_loss=0.04809, over 956946.46 frames. ], batch size: 33, lr: 3.13e-03, grad_scale: 32.0 2023-04-27 17:40:44,823 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=124994.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:40:51,151 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=125004.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 17:40:57,851 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.2780, 1.7298, 2.1774, 2.7025, 2.1409, 1.7674, 1.6069, 1.9566], device='cuda:6'), covar=tensor([0.2809, 0.3105, 0.1502, 0.2125, 0.2593, 0.2512, 0.3839, 0.1878], device='cuda:6'), in_proj_covar=tensor([0.0292, 0.0243, 0.0225, 0.0312, 0.0220, 0.0232, 0.0226, 0.0184], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:6') 2023-04-27 17:41:00,859 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=125019.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:41:10,029 INFO [finetune.py:976] (6/7) Epoch 22, batch 4750, loss[loss=0.1468, simple_loss=0.2168, pruned_loss=0.03838, over 4798.00 frames. ], tot_loss[loss=0.1678, simple_loss=0.2389, pruned_loss=0.0484, over 957078.39 frames. ], batch size: 29, lr: 3.13e-03, grad_scale: 32.0 2023-04-27 17:41:23,892 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=125052.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 17:41:31,621 INFO [optim.py:369] (6/7) 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 17:41:43,189 INFO [finetune.py:976] (6/7) Epoch 22, batch 4800, loss[loss=0.1557, simple_loss=0.2306, pruned_loss=0.04044, over 4751.00 frames. ], tot_loss[loss=0.1695, simple_loss=0.2412, pruned_loss=0.0489, over 956933.72 frames. ], batch size: 27, lr: 3.13e-03, grad_scale: 32.0 2023-04-27 17:41:48,388 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=125088.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:42:14,790 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=125128.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:42:17,145 INFO [finetune.py:976] (6/7) Epoch 22, batch 4850, loss[loss=0.1766, simple_loss=0.2486, pruned_loss=0.05236, over 4906.00 frames. ], tot_loss[loss=0.1716, simple_loss=0.2443, pruned_loss=0.04952, over 954862.69 frames. ], batch size: 36, lr: 3.13e-03, grad_scale: 32.0 2023-04-27 17:42:17,231 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([0.9275, 1.1120, 3.2307, 2.9657, 2.9064, 3.2037, 3.1532, 2.8101], device='cuda:6'), covar=tensor([0.7708, 0.5974, 0.1543, 0.2364, 0.1603, 0.2083, 0.1906, 0.1890], device='cuda:6'), in_proj_covar=tensor([0.0308, 0.0304, 0.0404, 0.0406, 0.0346, 0.0408, 0.0312, 0.0364], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-27 17:42:20,148 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=125136.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:42:44,289 INFO [optim.py:369] (6/7) 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 17:43:03,307 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=125176.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:43:13,297 INFO [finetune.py:976] (6/7) Epoch 22, batch 4900, loss[loss=0.1447, simple_loss=0.2353, pruned_loss=0.02706, over 4827.00 frames. ], tot_loss[loss=0.173, simple_loss=0.2461, pruned_loss=0.04993, over 955047.90 frames. ], batch size: 38, lr: 3.13e-03, grad_scale: 32.0 2023-04-27 17:43:23,301 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=125189.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:44:14,012 INFO [finetune.py:976] (6/7) Epoch 22, batch 4950, loss[loss=0.2023, simple_loss=0.2684, pruned_loss=0.06813, over 4852.00 frames. ], tot_loss[loss=0.1725, simple_loss=0.2463, pruned_loss=0.04938, over 954577.67 frames. ], batch size: 44, lr: 3.13e-03, grad_scale: 32.0 2023-04-27 17:44:27,576 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-04-27 17:44:50,836 INFO [optim.py:369] (6/7) 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] (6/7) Epoch 22, batch 5000, loss[loss=0.1544, simple_loss=0.2232, pruned_loss=0.04281, over 4783.00 frames. ], tot_loss[loss=0.1709, simple_loss=0.244, pruned_loss=0.04888, over 954111.79 frames. ], batch size: 26, lr: 3.13e-03, grad_scale: 16.0 2023-04-27 17:45:33,324 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=125294.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:45:33,379 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([0.9017, 0.6751, 0.7558, 0.7028, 1.0482, 0.8259, 0.7967, 0.7982], device='cuda:6'), covar=tensor([0.1698, 0.1414, 0.1985, 0.1618, 0.1118, 0.1308, 0.1553, 0.2191], device='cuda:6'), in_proj_covar=tensor([0.0310, 0.0308, 0.0348, 0.0285, 0.0323, 0.0304, 0.0297, 0.0370], device='cuda:6'), out_proj_covar=tensor([6.3642e-05, 6.3625e-05, 7.3283e-05, 5.7328e-05, 6.6611e-05, 6.3685e-05, 6.2061e-05, 7.8465e-05], device='cuda:6') 2023-04-27 17:46:05,805 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=125319.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:46:07,027 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5250, 1.0815, 0.2272, 1.2321, 1.1282, 1.3908, 1.3066, 1.3517], device='cuda:6'), covar=tensor([0.0487, 0.0375, 0.0413, 0.0526, 0.0288, 0.0487, 0.0465, 0.0530], device='cuda:6'), in_proj_covar=tensor([0.0028, 0.0024, 0.0022, 0.0028, 0.0019, 0.0028, 0.0027, 0.0028], device='cuda:6'), out_proj_covar=tensor([0.0051, 0.0045, 0.0038, 0.0052, 0.0038, 0.0050, 0.0050, 0.0051], device='cuda:6') 2023-04-27 17:46:18,221 INFO [finetune.py:976] (6/7) Epoch 22, batch 5050, loss[loss=0.1686, simple_loss=0.2353, pruned_loss=0.05091, over 4820.00 frames. ], tot_loss[loss=0.1696, simple_loss=0.2421, pruned_loss=0.04849, over 952854.91 frames. ], batch size: 41, lr: 3.13e-03, grad_scale: 16.0 2023-04-27 17:46:32,852 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=125342.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:46:43,468 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=125356.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:46:48,791 INFO [optim.py:369] (6/7) 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 17:46:50,093 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=125367.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:46:55,671 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=125376.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:46:59,443 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.48 vs. limit=5.0 2023-04-27 17:46:59,779 INFO [finetune.py:976] (6/7) Epoch 22, batch 5100, loss[loss=0.1495, simple_loss=0.2139, pruned_loss=0.04257, over 4793.00 frames. ], tot_loss[loss=0.1676, simple_loss=0.2392, pruned_loss=0.04805, over 953720.32 frames. ], batch size: 51, lr: 3.12e-03, grad_scale: 16.0 2023-04-27 17:47:00,745 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-04-27 17:47:24,386 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=125417.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 17:47:26,784 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=125421.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:47:29,863 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8645, 2.3914, 1.9555, 1.7898, 1.3891, 1.3996, 1.9903, 1.3795], device='cuda:6'), covar=tensor([0.1676, 0.1376, 0.1444, 0.1724, 0.2399, 0.1994, 0.1011, 0.2200], device='cuda:6'), in_proj_covar=tensor([0.0197, 0.0211, 0.0169, 0.0203, 0.0200, 0.0185, 0.0156, 0.0186], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:6') 2023-04-27 17:47:33,408 INFO [finetune.py:976] (6/7) Epoch 22, batch 5150, loss[loss=0.1341, simple_loss=0.1992, pruned_loss=0.03453, over 4723.00 frames. ], tot_loss[loss=0.1687, simple_loss=0.2399, pruned_loss=0.04871, over 954317.12 frames. ], batch size: 23, lr: 3.12e-03, grad_scale: 16.0 2023-04-27 17:47:37,059 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=125437.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:48:02,908 INFO [optim.py:369] (6/7) 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 17:48:09,684 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=125476.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:48:13,157 INFO [finetune.py:976] (6/7) Epoch 22, batch 5200, loss[loss=0.1946, simple_loss=0.2787, pruned_loss=0.05523, over 4895.00 frames. ], tot_loss[loss=0.1712, simple_loss=0.2433, pruned_loss=0.04957, over 953366.23 frames. ], batch size: 35, lr: 3.12e-03, grad_scale: 16.0 2023-04-27 17:48:13,278 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=125482.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:48:14,967 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=125484.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:48:42,437 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=125524.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:48:47,214 INFO [finetune.py:976] (6/7) Epoch 22, batch 5250, loss[loss=0.1731, simple_loss=0.2454, pruned_loss=0.05042, over 4874.00 frames. ], tot_loss[loss=0.1726, simple_loss=0.2452, pruned_loss=0.04999, over 951766.07 frames. ], batch size: 31, lr: 3.12e-03, grad_scale: 16.0 2023-04-27 17:49:09,790 INFO [optim.py:369] (6/7) 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 17:49:12,800 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9296, 1.0624, 1.5679, 1.6827, 1.5704, 1.6562, 1.5701, 1.5639], device='cuda:6'), covar=tensor([0.3699, 0.5200, 0.3871, 0.3934, 0.5435, 0.6682, 0.4393, 0.4164], device='cuda:6'), in_proj_covar=tensor([0.0337, 0.0371, 0.0323, 0.0338, 0.0345, 0.0390, 0.0355, 0.0328], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 17:49:20,593 INFO [finetune.py:976] (6/7) Epoch 22, batch 5300, loss[loss=0.2375, simple_loss=0.3088, pruned_loss=0.08307, over 4905.00 frames. ], tot_loss[loss=0.1748, simple_loss=0.2474, pruned_loss=0.05116, over 951584.22 frames. ], batch size: 37, lr: 3.12e-03, grad_scale: 16.0 2023-04-27 17:49:30,968 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5392, 1.4133, 1.2457, 1.4840, 1.8030, 1.5002, 1.3612, 1.1983], device='cuda:6'), covar=tensor([0.1399, 0.1098, 0.1460, 0.0996, 0.0550, 0.1379, 0.1577, 0.1823], device='cuda:6'), in_proj_covar=tensor([0.0310, 0.0308, 0.0348, 0.0285, 0.0323, 0.0304, 0.0297, 0.0370], device='cuda:6'), out_proj_covar=tensor([6.3628e-05, 6.3670e-05, 7.3406e-05, 5.7258e-05, 6.6514e-05, 6.3747e-05, 6.2049e-05, 7.8404e-05], device='cuda:6') 2023-04-27 17:49:34,598 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=125603.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:49:45,831 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0041, 2.6308, 2.1195, 1.9807, 1.4748, 1.4881, 2.2862, 1.4359], device='cuda:6'), covar=tensor([0.1747, 0.1514, 0.1516, 0.1878, 0.2362, 0.1986, 0.0987, 0.2203], device='cuda:6'), in_proj_covar=tensor([0.0196, 0.0210, 0.0168, 0.0203, 0.0199, 0.0184, 0.0155, 0.0185], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:6') 2023-04-27 17:49:50,889 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-04-27 17:49:54,072 INFO [finetune.py:976] (6/7) Epoch 22, batch 5350, loss[loss=0.187, simple_loss=0.2627, pruned_loss=0.0556, over 4834.00 frames. ], tot_loss[loss=0.1741, simple_loss=0.2468, pruned_loss=0.05075, over 950498.68 frames. ], batch size: 47, lr: 3.12e-03, grad_scale: 16.0 2023-04-27 17:50:09,858 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6085, 1.3194, 1.3526, 1.3966, 1.7910, 1.4492, 1.2102, 1.2239], device='cuda:6'), covar=tensor([0.2070, 0.1339, 0.1856, 0.1359, 0.0886, 0.1704, 0.1957, 0.2419], device='cuda:6'), in_proj_covar=tensor([0.0312, 0.0310, 0.0350, 0.0287, 0.0325, 0.0306, 0.0299, 0.0371], device='cuda:6'), out_proj_covar=tensor([6.4028e-05, 6.4058e-05, 7.3901e-05, 5.7748e-05, 6.7011e-05, 6.4170e-05, 6.2511e-05, 7.8787e-05], device='cuda:6') 2023-04-27 17:50:15,644 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=125664.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:50:16,101 INFO [optim.py:369] (6/7) 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 17:50:24,496 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-04-27 17:50:38,573 INFO [finetune.py:976] (6/7) Epoch 22, batch 5400, loss[loss=0.2031, simple_loss=0.2596, pruned_loss=0.07328, over 4721.00 frames. ], tot_loss[loss=0.1724, simple_loss=0.2442, pruned_loss=0.05027, over 950530.38 frames. ], batch size: 54, lr: 3.12e-03, grad_scale: 16.0 2023-04-27 17:51:19,944 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=125712.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 17:51:40,593 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=125725.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:51:44,689 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6031, 1.4698, 4.4425, 4.2515, 3.9078, 4.2409, 4.1228, 3.9290], device='cuda:6'), covar=tensor([0.6760, 0.5280, 0.1109, 0.1579, 0.0921, 0.1442, 0.1357, 0.1407], device='cuda:6'), in_proj_covar=tensor([0.0309, 0.0305, 0.0406, 0.0406, 0.0347, 0.0409, 0.0314, 0.0366], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-27 17:51:45,247 INFO [finetune.py:976] (6/7) Epoch 22, batch 5450, loss[loss=0.1457, simple_loss=0.2152, pruned_loss=0.03806, over 4889.00 frames. ], tot_loss[loss=0.17, simple_loss=0.2414, pruned_loss=0.04931, over 952860.51 frames. ], batch size: 32, lr: 3.12e-03, grad_scale: 16.0 2023-04-27 17:51:45,326 INFO [zipformer.py:1188] (6/7) 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] (6/7) 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 17:52:47,950 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=125777.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:52:51,409 INFO [finetune.py:976] (6/7) Epoch 22, batch 5500, loss[loss=0.2047, simple_loss=0.2568, pruned_loss=0.07628, over 4093.00 frames. ], tot_loss[loss=0.1683, simple_loss=0.239, pruned_loss=0.04874, over 953530.82 frames. ], batch size: 65, lr: 3.12e-03, grad_scale: 16.0 2023-04-27 17:52:58,430 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=125784.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:52:58,441 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.4407, 2.9467, 2.5015, 2.8225, 2.0973, 2.6656, 2.6840, 2.0222], device='cuda:6'), covar=tensor([0.1928, 0.1115, 0.0700, 0.1278, 0.3280, 0.1143, 0.2005, 0.2657], device='cuda:6'), in_proj_covar=tensor([0.0287, 0.0302, 0.0218, 0.0279, 0.0318, 0.0258, 0.0251, 0.0264], device='cuda:6'), out_proj_covar=tensor([1.1477e-04, 1.1944e-04, 8.5946e-05, 1.1002e-04, 1.2856e-04, 1.0191e-04, 1.0110e-04, 1.0429e-04], device='cuda:6') 2023-04-27 17:52:59,642 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=125786.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:53:57,320 INFO [finetune.py:976] (6/7) Epoch 22, batch 5550, loss[loss=0.1822, simple_loss=0.2513, pruned_loss=0.0565, over 4886.00 frames. ], tot_loss[loss=0.1704, simple_loss=0.241, pruned_loss=0.04993, over 951368.88 frames. ], batch size: 32, lr: 3.12e-03, grad_scale: 16.0 2023-04-27 17:54:03,349 INFO [zipformer.py:1188] (6/7) 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] (6/7) 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 17:55:02,079 INFO [finetune.py:976] (6/7) Epoch 22, batch 5600, loss[loss=0.2011, simple_loss=0.2865, pruned_loss=0.05787, over 4837.00 frames. ], tot_loss[loss=0.1726, simple_loss=0.2445, pruned_loss=0.0504, over 948787.40 frames. ], batch size: 40, lr: 3.12e-03, grad_scale: 16.0 2023-04-27 17:55:11,764 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.4060, 1.8700, 2.3565, 2.8127, 2.3349, 1.8473, 1.7859, 2.1839], device='cuda:6'), covar=tensor([0.3374, 0.3269, 0.1710, 0.2437, 0.2815, 0.2751, 0.3816, 0.2103], device='cuda:6'), in_proj_covar=tensor([0.0293, 0.0244, 0.0227, 0.0314, 0.0221, 0.0234, 0.0228, 0.0184], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:6') 2023-04-27 17:55:20,382 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5290, 2.6275, 1.9833, 2.1758, 2.2867, 1.9256, 3.2784, 1.5725], device='cuda:6'), covar=tensor([0.4015, 0.2005, 0.4622, 0.3439, 0.2447, 0.3245, 0.1599, 0.5251], device='cuda:6'), in_proj_covar=tensor([0.0335, 0.0347, 0.0422, 0.0349, 0.0377, 0.0372, 0.0365, 0.0416], device='cuda:6'), out_proj_covar=tensor([9.9323e-05, 1.0380e-04, 1.2804e-04, 1.0502e-04, 1.1226e-04, 1.1090e-04, 1.0741e-04, 1.2540e-04], device='cuda:6') 2023-04-27 17:55:24,431 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6603, 2.4148, 1.5754, 1.7263, 1.2738, 1.2649, 1.6716, 1.2023], device='cuda:6'), covar=tensor([0.1968, 0.1285, 0.1810, 0.1885, 0.2657, 0.2476, 0.1199, 0.2328], device='cuda:6'), in_proj_covar=tensor([0.0198, 0.0212, 0.0170, 0.0205, 0.0201, 0.0186, 0.0156, 0.0187], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:6') 2023-04-27 17:56:05,133 INFO [finetune.py:976] (6/7) Epoch 22, batch 5650, loss[loss=0.1627, simple_loss=0.2368, pruned_loss=0.04429, over 4762.00 frames. ], tot_loss[loss=0.1728, simple_loss=0.2458, pruned_loss=0.04995, over 949911.64 frames. ], batch size: 27, lr: 3.12e-03, grad_scale: 16.0 2023-04-27 17:56:31,894 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=125959.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:56:35,184 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2023-04-27 17:56:35,410 INFO [optim.py:369] (6/7) 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 17:56:36,116 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.4886, 2.7837, 2.4382, 2.8754, 2.2364, 2.4970, 2.5218, 2.0740], device='cuda:6'), covar=tensor([0.1384, 0.1157, 0.0694, 0.0910, 0.2775, 0.0953, 0.1541, 0.2140], device='cuda:6'), in_proj_covar=tensor([0.0286, 0.0301, 0.0217, 0.0278, 0.0316, 0.0257, 0.0250, 0.0263], device='cuda:6'), out_proj_covar=tensor([1.1414e-04, 1.1914e-04, 8.5611e-05, 1.0957e-04, 1.2804e-04, 1.0145e-04, 1.0070e-04, 1.0380e-04], device='cuda:6') 2023-04-27 17:56:45,407 INFO [finetune.py:976] (6/7) Epoch 22, batch 5700, loss[loss=0.1446, simple_loss=0.2053, pruned_loss=0.04195, over 4216.00 frames. ], tot_loss[loss=0.1699, simple_loss=0.2415, pruned_loss=0.04915, over 933623.11 frames. ], batch size: 18, lr: 3.12e-03, grad_scale: 16.0 2023-04-27 17:56:54,917 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=125998.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:57:25,280 INFO [finetune.py:976] (6/7) Epoch 23, batch 0, loss[loss=0.1882, simple_loss=0.2692, pruned_loss=0.0536, over 4824.00 frames. ], tot_loss[loss=0.1882, simple_loss=0.2692, pruned_loss=0.0536, over 4824.00 frames. ], batch size: 47, lr: 3.12e-03, grad_scale: 16.0 2023-04-27 17:57:25,280 INFO [finetune.py:1001] (6/7) Computing validation loss 2023-04-27 17:57:36,730 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5125, 2.9954, 0.9182, 1.8150, 1.8819, 2.1757, 1.8362, 1.0179], device='cuda:6'), covar=tensor([0.1263, 0.1101, 0.1860, 0.1167, 0.0973, 0.0929, 0.1541, 0.1740], device='cuda:6'), in_proj_covar=tensor([0.0118, 0.0240, 0.0138, 0.0120, 0.0132, 0.0151, 0.0118, 0.0119], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-27 17:57:37,736 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8090, 1.6050, 1.7596, 2.1351, 2.0644, 1.6699, 1.5436, 1.8983], device='cuda:6'), covar=tensor([0.0876, 0.1217, 0.0803, 0.0603, 0.0689, 0.0967, 0.0762, 0.0594], device='cuda:6'), in_proj_covar=tensor([0.0187, 0.0202, 0.0184, 0.0175, 0.0177, 0.0181, 0.0152, 0.0178], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 17:57:41,122 INFO [finetune.py:1010] (6/7) Epoch 23, validation: loss=0.1552, simple_loss=0.2246, pruned_loss=0.04292, over 2265189.00 frames. 2023-04-27 17:57:41,123 INFO [finetune.py:1011] (6/7) Maximum memory allocated so far is 6435MB 2023-04-27 17:57:48,914 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=126012.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 17:58:05,566 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=126032.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:58:21,096 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-04-27 17:58:23,638 INFO [finetune.py:976] (6/7) Epoch 23, batch 50, loss[loss=0.1607, simple_loss=0.24, pruned_loss=0.04071, over 4900.00 frames. ], tot_loss[loss=0.1737, simple_loss=0.2466, pruned_loss=0.05036, over 217236.22 frames. ], batch size: 43, lr: 3.12e-03, grad_scale: 16.0 2023-04-27 17:58:24,226 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=126059.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:58:25,268 INFO [zipformer.py:1188] (6/7) 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] (6/7) 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 17:58:32,066 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2023-04-27 17:58:41,179 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=126077.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:58:42,979 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=126080.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:58:48,843 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=126081.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:59:08,952 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-04-27 17:59:21,044 INFO [finetune.py:976] (6/7) Epoch 23, batch 100, loss[loss=0.1989, simple_loss=0.267, pruned_loss=0.06544, over 4823.00 frames. ], tot_loss[loss=0.1684, simple_loss=0.2403, pruned_loss=0.04824, over 381557.62 frames. ], batch size: 38, lr: 3.12e-03, grad_scale: 16.0 2023-04-27 17:59:42,930 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=126125.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:00:14,331 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8348, 2.1978, 1.1525, 1.5839, 2.2261, 1.7000, 1.6830, 1.7464], device='cuda:6'), covar=tensor([0.0457, 0.0315, 0.0268, 0.0509, 0.0222, 0.0474, 0.0441, 0.0534], device='cuda:6'), in_proj_covar=tensor([0.0028, 0.0024, 0.0022, 0.0028, 0.0019, 0.0028, 0.0027, 0.0028], device='cuda:6'), out_proj_covar=tensor([0.0051, 0.0046, 0.0038, 0.0052, 0.0038, 0.0050, 0.0050, 0.0051], device='cuda:6') 2023-04-27 18:00:28,539 INFO [finetune.py:976] (6/7) Epoch 23, batch 150, loss[loss=0.1688, simple_loss=0.2404, pruned_loss=0.04858, over 4836.00 frames. ], tot_loss[loss=0.1666, simple_loss=0.2367, pruned_loss=0.04823, over 508388.29 frames. ], batch size: 33, lr: 3.12e-03, grad_scale: 16.0 2023-04-27 18:00:38,442 INFO [optim.py:369] (6/7) 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 18:00:48,008 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.6101, 1.8464, 1.9334, 2.0508, 1.8685, 1.9560, 1.9582, 1.9273], device='cuda:6'), covar=tensor([0.3709, 0.5998, 0.4731, 0.4685, 0.5694, 0.7176, 0.5547, 0.5252], device='cuda:6'), in_proj_covar=tensor([0.0338, 0.0372, 0.0324, 0.0339, 0.0346, 0.0393, 0.0356, 0.0330], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 18:01:00,009 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6091, 2.6303, 2.1633, 2.3271, 2.6360, 2.1667, 3.4801, 2.0181], device='cuda:6'), covar=tensor([0.3693, 0.2337, 0.4440, 0.3325, 0.1857, 0.2561, 0.1704, 0.4231], device='cuda:6'), in_proj_covar=tensor([0.0336, 0.0349, 0.0424, 0.0352, 0.0379, 0.0372, 0.0368, 0.0419], device='cuda:6'), out_proj_covar=tensor([9.9772e-05, 1.0439e-04, 1.2871e-04, 1.0571e-04, 1.1265e-04, 1.1098e-04, 1.0816e-04, 1.2623e-04], device='cuda:6') 2023-04-27 18:01:10,275 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.5930, 3.0968, 2.6967, 3.1256, 2.2844, 2.8508, 2.8605, 2.2851], device='cuda:6'), covar=tensor([0.1808, 0.1189, 0.0751, 0.1070, 0.3096, 0.0965, 0.1867, 0.2516], device='cuda:6'), in_proj_covar=tensor([0.0285, 0.0301, 0.0217, 0.0277, 0.0315, 0.0256, 0.0249, 0.0262], device='cuda:6'), out_proj_covar=tensor([1.1388e-04, 1.1899e-04, 8.5465e-05, 1.0941e-04, 1.2747e-04, 1.0123e-04, 1.0061e-04, 1.0358e-04], device='cuda:6') 2023-04-27 18:01:34,971 INFO [finetune.py:976] (6/7) Epoch 23, batch 200, loss[loss=0.134, simple_loss=0.2048, pruned_loss=0.03155, over 4824.00 frames. ], tot_loss[loss=0.166, simple_loss=0.2356, pruned_loss=0.04824, over 605625.45 frames. ], batch size: 38, lr: 3.12e-03, grad_scale: 16.0 2023-04-27 18:02:00,367 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.62 vs. limit=2.0 2023-04-27 18:02:09,563 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.81 vs. limit=2.0 2023-04-27 18:02:18,948 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 2023-04-27 18:02:19,988 INFO [finetune.py:976] (6/7) Epoch 23, batch 250, loss[loss=0.1407, simple_loss=0.1993, pruned_loss=0.0411, over 4037.00 frames. ], tot_loss[loss=0.171, simple_loss=0.2408, pruned_loss=0.05059, over 681366.68 frames. ], batch size: 17, lr: 3.11e-03, grad_scale: 16.0 2023-04-27 18:02:20,607 INFO [zipformer.py:1188] (6/7) 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] (6/7) 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 18:02:32,902 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.46 vs. limit=5.0 2023-04-27 18:02:51,719 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=126307.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:02:53,382 INFO [finetune.py:976] (6/7) Epoch 23, batch 300, loss[loss=0.2128, simple_loss=0.2843, pruned_loss=0.07067, over 4892.00 frames. ], tot_loss[loss=0.1749, simple_loss=0.2466, pruned_loss=0.05164, over 743639.93 frames. ], batch size: 35, lr: 3.11e-03, grad_scale: 16.0 2023-04-27 18:02:55,368 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.3437, 1.1981, 1.6418, 1.5318, 1.2512, 1.1497, 1.2439, 0.7727], device='cuda:6'), covar=tensor([0.0508, 0.0596, 0.0334, 0.0497, 0.0710, 0.1080, 0.0494, 0.0553], device='cuda:6'), in_proj_covar=tensor([0.0069, 0.0068, 0.0066, 0.0068, 0.0074, 0.0095, 0.0073, 0.0064], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-27 18:03:00,990 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.78 vs. limit=5.0 2023-04-27 18:03:01,204 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.6445, 3.2375, 2.6718, 3.1746, 2.3272, 2.8030, 2.8583, 2.1619], device='cuda:6'), covar=tensor([0.1871, 0.1123, 0.0793, 0.1027, 0.3094, 0.1141, 0.1756, 0.2434], device='cuda:6'), in_proj_covar=tensor([0.0285, 0.0300, 0.0216, 0.0277, 0.0315, 0.0256, 0.0249, 0.0262], device='cuda:6'), out_proj_covar=tensor([1.1381e-04, 1.1867e-04, 8.5319e-05, 1.0927e-04, 1.2727e-04, 1.0111e-04, 1.0044e-04, 1.0356e-04], device='cuda:6') 2023-04-27 18:03:22,883 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=126354.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:03:26,337 INFO [finetune.py:976] (6/7) Epoch 23, batch 350, loss[loss=0.1996, simple_loss=0.2686, pruned_loss=0.0653, over 4806.00 frames. ], tot_loss[loss=0.1758, simple_loss=0.2476, pruned_loss=0.05193, over 788960.23 frames. ], batch size: 40, lr: 3.11e-03, grad_scale: 16.0 2023-04-27 18:03:30,385 INFO [optim.py:369] (6/7) 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 18:03:41,541 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=126380.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:03:42,125 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=126381.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:03:51,909 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.0686, 0.7619, 0.9124, 0.7997, 1.2016, 1.0101, 0.8893, 0.8987], device='cuda:6'), covar=tensor([0.1765, 0.1416, 0.1783, 0.1589, 0.0929, 0.1385, 0.1558, 0.2319], device='cuda:6'), in_proj_covar=tensor([0.0312, 0.0309, 0.0350, 0.0285, 0.0324, 0.0306, 0.0296, 0.0369], device='cuda:6'), out_proj_covar=tensor([6.3942e-05, 6.3985e-05, 7.3903e-05, 5.7298e-05, 6.6657e-05, 6.4098e-05, 6.1849e-05, 7.8191e-05], device='cuda:6') 2023-04-27 18:03:59,739 INFO [finetune.py:976] (6/7) Epoch 23, batch 400, loss[loss=0.1631, simple_loss=0.2365, pruned_loss=0.04489, over 4928.00 frames. ], tot_loss[loss=0.1734, simple_loss=0.2462, pruned_loss=0.0503, over 825892.50 frames. ], batch size: 33, lr: 3.11e-03, grad_scale: 16.0 2023-04-27 18:04:30,797 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=126429.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:04:42,870 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=126441.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:04:48,380 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=126450.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:04:50,881 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7620, 1.0301, 1.7294, 2.1962, 1.8488, 1.6801, 1.7187, 1.6738], device='cuda:6'), covar=tensor([0.4369, 0.6729, 0.5965, 0.5764, 0.5580, 0.7510, 0.7224, 0.8688], device='cuda:6'), in_proj_covar=tensor([0.0433, 0.0415, 0.0508, 0.0507, 0.0462, 0.0491, 0.0498, 0.0509], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 18:04:54,264 INFO [finetune.py:976] (6/7) Epoch 23, batch 450, loss[loss=0.2118, simple_loss=0.2666, pruned_loss=0.07847, over 4827.00 frames. ], tot_loss[loss=0.1721, simple_loss=0.245, pruned_loss=0.0496, over 853612.97 frames. ], batch size: 40, lr: 3.11e-03, grad_scale: 16.0 2023-04-27 18:04:58,396 INFO [optim.py:369] (6/7) 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 18:05:24,777 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.5153, 1.7624, 1.8325, 1.9702, 1.8244, 1.8802, 1.9178, 1.8840], device='cuda:6'), covar=tensor([0.4290, 0.5565, 0.4640, 0.4446, 0.5682, 0.7599, 0.5195, 0.4940], device='cuda:6'), in_proj_covar=tensor([0.0341, 0.0375, 0.0325, 0.0341, 0.0348, 0.0397, 0.0358, 0.0332], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 18:05:27,629 INFO [finetune.py:976] (6/7) Epoch 23, batch 500, loss[loss=0.1442, simple_loss=0.2226, pruned_loss=0.03291, over 4784.00 frames. ], tot_loss[loss=0.1697, simple_loss=0.2419, pruned_loss=0.04877, over 878238.46 frames. ], batch size: 51, lr: 3.11e-03, grad_scale: 16.0 2023-04-27 18:05:29,454 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=126511.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:06:06,869 INFO [finetune.py:976] (6/7) Epoch 23, batch 550, loss[loss=0.1719, simple_loss=0.2465, pruned_loss=0.04866, over 4823.00 frames. ], tot_loss[loss=0.1676, simple_loss=0.239, pruned_loss=0.04815, over 896568.70 frames. ], batch size: 38, lr: 3.11e-03, grad_scale: 16.0 2023-04-27 18:06:16,306 INFO [optim.py:369] (6/7) 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 18:06:38,376 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0979, 1.5569, 1.9443, 2.3550, 2.0069, 1.5705, 1.4290, 1.7841], device='cuda:6'), covar=tensor([0.2772, 0.2975, 0.1634, 0.2218, 0.2395, 0.2610, 0.3968, 0.1870], device='cuda:6'), in_proj_covar=tensor([0.0293, 0.0244, 0.0227, 0.0315, 0.0220, 0.0234, 0.0228, 0.0185], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:6') 2023-04-27 18:07:12,073 INFO [finetune.py:976] (6/7) Epoch 23, batch 600, loss[loss=0.2037, simple_loss=0.272, pruned_loss=0.06771, over 4018.00 frames. ], tot_loss[loss=0.1687, simple_loss=0.2397, pruned_loss=0.04884, over 907769.09 frames. ], batch size: 65, lr: 3.11e-03, grad_scale: 16.0 2023-04-27 18:08:06,581 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.76 vs. limit=5.0 2023-04-27 18:08:14,786 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=126654.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:08:15,406 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9022, 2.4373, 0.9871, 1.3325, 1.8490, 1.1623, 2.9883, 1.5525], device='cuda:6'), covar=tensor([0.0746, 0.0561, 0.0779, 0.1270, 0.0509, 0.1079, 0.0261, 0.0675], device='cuda:6'), in_proj_covar=tensor([0.0051, 0.0064, 0.0047, 0.0046, 0.0050, 0.0051, 0.0073, 0.0051], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:6') 2023-04-27 18:08:16,643 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0603, 2.6675, 1.0419, 1.5093, 2.0770, 1.3236, 3.3546, 1.7668], device='cuda:6'), covar=tensor([0.0665, 0.0650, 0.0831, 0.1178, 0.0485, 0.0940, 0.0254, 0.0611], device='cuda:6'), in_proj_covar=tensor([0.0050, 0.0064, 0.0047, 0.0046, 0.0050, 0.0051, 0.0073, 0.0051], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:6') 2023-04-27 18:08:17,791 INFO [finetune.py:976] (6/7) Epoch 23, batch 650, loss[loss=0.1939, simple_loss=0.2688, pruned_loss=0.05949, over 4908.00 frames. ], tot_loss[loss=0.1726, simple_loss=0.2439, pruned_loss=0.05068, over 918150.76 frames. ], batch size: 36, lr: 3.11e-03, grad_scale: 16.0 2023-04-27 18:08:26,640 INFO [optim.py:369] (6/7) 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 18:09:20,360 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 23, batch 700, loss[loss=0.1916, simple_loss=0.269, pruned_loss=0.05712, over 4903.00 frames. ], tot_loss[loss=0.1724, simple_loss=0.2443, pruned_loss=0.05025, over 924853.73 frames. ], batch size: 35, lr: 3.11e-03, grad_scale: 16.0 2023-04-27 18:10:03,685 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9384, 1.3414, 4.9567, 4.6417, 4.3260, 4.7290, 4.4244, 4.3441], device='cuda:6'), covar=tensor([0.7101, 0.5941, 0.1077, 0.1860, 0.1178, 0.1140, 0.1488, 0.1688], device='cuda:6'), in_proj_covar=tensor([0.0308, 0.0305, 0.0404, 0.0405, 0.0347, 0.0409, 0.0313, 0.0363], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-27 18:10:03,687 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=126736.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:10:13,729 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2023-04-27 18:10:19,572 INFO [finetune.py:976] (6/7) Epoch 23, batch 750, loss[loss=0.1486, simple_loss=0.2282, pruned_loss=0.03452, over 4889.00 frames. ], tot_loss[loss=0.1728, simple_loss=0.245, pruned_loss=0.05034, over 929790.92 frames. ], batch size: 32, lr: 3.11e-03, grad_scale: 16.0 2023-04-27 18:10:23,176 INFO [optim.py:369] (6/7) 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] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=126806.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:10:53,514 INFO [finetune.py:976] (6/7) Epoch 23, batch 800, loss[loss=0.1518, simple_loss=0.234, pruned_loss=0.03478, over 4834.00 frames. ], tot_loss[loss=0.1725, simple_loss=0.2448, pruned_loss=0.05007, over 936685.30 frames. ], batch size: 47, lr: 3.11e-03, grad_scale: 16.0 2023-04-27 18:10:57,966 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=126816.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:11:27,333 INFO [finetune.py:976] (6/7) Epoch 23, batch 850, loss[loss=0.1767, simple_loss=0.2443, pruned_loss=0.05461, over 4896.00 frames. ], tot_loss[loss=0.1704, simple_loss=0.2428, pruned_loss=0.04896, over 942649.98 frames. ], batch size: 32, lr: 3.11e-03, grad_scale: 16.0 2023-04-27 18:11:30,956 INFO [optim.py:369] (6/7) 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 18:11:44,270 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=126877.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:12:28,957 INFO [finetune.py:976] (6/7) Epoch 23, batch 900, loss[loss=0.1611, simple_loss=0.2375, pruned_loss=0.04237, over 4783.00 frames. ], tot_loss[loss=0.1681, simple_loss=0.2399, pruned_loss=0.04812, over 946568.82 frames. ], batch size: 28, lr: 3.11e-03, grad_scale: 16.0 2023-04-27 18:13:36,985 INFO [finetune.py:976] (6/7) Epoch 23, batch 950, loss[loss=0.2016, simple_loss=0.263, pruned_loss=0.07004, over 4935.00 frames. ], tot_loss[loss=0.1672, simple_loss=0.2382, pruned_loss=0.04805, over 947951.26 frames. ], batch size: 33, lr: 3.11e-03, grad_scale: 16.0 2023-04-27 18:13:40,660 INFO [optim.py:369] (6/7) 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] (6/7) Epoch 23, batch 1000, loss[loss=0.1942, simple_loss=0.2701, pruned_loss=0.05915, over 4919.00 frames. ], tot_loss[loss=0.17, simple_loss=0.2416, pruned_loss=0.04922, over 949157.01 frames. ], batch size: 38, lr: 3.11e-03, grad_scale: 16.0 2023-04-27 18:14:14,243 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([0.9096, 1.1446, 3.2900, 3.0259, 2.9428, 3.2403, 3.2283, 2.9135], device='cuda:6'), covar=tensor([0.7511, 0.5625, 0.1504, 0.2130, 0.1453, 0.2489, 0.1346, 0.1753], device='cuda:6'), in_proj_covar=tensor([0.0310, 0.0306, 0.0404, 0.0407, 0.0348, 0.0410, 0.0314, 0.0363], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-27 18:14:26,500 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 2023-04-27 18:14:28,623 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=127036.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:14:45,073 INFO [finetune.py:976] (6/7) Epoch 23, batch 1050, loss[loss=0.1385, simple_loss=0.2165, pruned_loss=0.03027, over 4744.00 frames. ], tot_loss[loss=0.1712, simple_loss=0.2437, pruned_loss=0.04935, over 948106.44 frames. ], batch size: 54, lr: 3.11e-03, grad_scale: 16.0 2023-04-27 18:14:48,720 INFO [optim.py:369] (6/7) 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 18:14:49,478 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.3782, 1.6600, 1.7646, 1.8737, 1.7511, 1.7682, 1.8156, 1.8003], device='cuda:6'), covar=tensor([0.3743, 0.5205, 0.4456, 0.4043, 0.5240, 0.6792, 0.4956, 0.4861], device='cuda:6'), in_proj_covar=tensor([0.0336, 0.0371, 0.0321, 0.0336, 0.0343, 0.0390, 0.0353, 0.0328], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 18:15:00,297 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=127084.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:15:16,209 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=127106.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:15:18,457 INFO [finetune.py:976] (6/7) Epoch 23, batch 1100, loss[loss=0.198, simple_loss=0.2815, pruned_loss=0.05729, over 4850.00 frames. ], tot_loss[loss=0.1721, simple_loss=0.2454, pruned_loss=0.04939, over 950898.08 frames. ], batch size: 44, lr: 3.11e-03, grad_scale: 16.0 2023-04-27 18:15:30,669 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-04-27 18:15:47,875 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 23, batch 1150, loss[loss=0.1767, simple_loss=0.2484, pruned_loss=0.05245, over 4920.00 frames. ], tot_loss[loss=0.173, simple_loss=0.2466, pruned_loss=0.04971, over 951421.48 frames. ], batch size: 38, lr: 3.11e-03, grad_scale: 16.0 2023-04-27 18:15:56,436 INFO [optim.py:369] (6/7) 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 18:16:00,254 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([3.0645, 2.3365, 2.2195, 2.3645, 2.2321, 2.3430, 2.2776, 2.2845], device='cuda:6'), covar=tensor([0.4154, 0.5941, 0.5281, 0.4991, 0.5613, 0.6983, 0.6265, 0.5459], device='cuda:6'), in_proj_covar=tensor([0.0338, 0.0372, 0.0323, 0.0338, 0.0345, 0.0393, 0.0354, 0.0329], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 18:16:00,784 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=127172.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:16:25,336 INFO [finetune.py:976] (6/7) Epoch 23, batch 1200, loss[loss=0.1833, simple_loss=0.2592, pruned_loss=0.05372, over 4703.00 frames. ], tot_loss[loss=0.1719, simple_loss=0.2452, pruned_loss=0.04928, over 952618.89 frames. ], batch size: 54, lr: 3.11e-03, grad_scale: 16.0 2023-04-27 18:16:50,890 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.1346, 1.6665, 2.2228, 2.5977, 2.2317, 2.0648, 2.1214, 2.0260], device='cuda:6'), covar=tensor([0.4702, 0.7077, 0.6264, 0.5089, 0.6237, 0.7585, 0.7853, 0.7758], device='cuda:6'), in_proj_covar=tensor([0.0434, 0.0416, 0.0509, 0.0508, 0.0464, 0.0493, 0.0498, 0.0510], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 18:16:52,555 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=127249.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:17:03,695 INFO [finetune.py:976] (6/7) Epoch 23, batch 1250, loss[loss=0.1565, simple_loss=0.2459, pruned_loss=0.03355, over 4913.00 frames. ], tot_loss[loss=0.1702, simple_loss=0.2428, pruned_loss=0.04876, over 953583.94 frames. ], batch size: 36, lr: 3.11e-03, grad_scale: 16.0 2023-04-27 18:17:14,002 INFO [optim.py:369] (6/7) 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 18:17:33,725 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-04-27 18:18:08,738 INFO [finetune.py:976] (6/7) Epoch 23, batch 1300, loss[loss=0.1555, simple_loss=0.2211, pruned_loss=0.04497, over 4823.00 frames. ], tot_loss[loss=0.167, simple_loss=0.239, pruned_loss=0.04749, over 952825.30 frames. ], batch size: 30, lr: 3.11e-03, grad_scale: 32.0 2023-04-27 18:18:09,490 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=127310.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:19:03,377 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.3714, 3.0076, 0.8445, 1.6831, 1.7827, 2.2308, 1.7762, 0.9635], device='cuda:6'), covar=tensor([0.1321, 0.0930, 0.1887, 0.1232, 0.1023, 0.0977, 0.1395, 0.1771], device='cuda:6'), in_proj_covar=tensor([0.0116, 0.0237, 0.0137, 0.0118, 0.0131, 0.0150, 0.0116, 0.0118], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-27 18:19:13,548 INFO [finetune.py:976] (6/7) Epoch 23, batch 1350, loss[loss=0.1544, simple_loss=0.2321, pruned_loss=0.0383, over 4815.00 frames. ], tot_loss[loss=0.168, simple_loss=0.2399, pruned_loss=0.04807, over 955133.56 frames. ], batch size: 51, lr: 3.11e-03, grad_scale: 32.0 2023-04-27 18:19:21,695 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9031, 1.4410, 2.0241, 2.4304, 2.0014, 1.8575, 1.9147, 1.9053], device='cuda:6'), covar=tensor([0.4618, 0.7029, 0.6356, 0.5819, 0.6056, 0.8343, 0.8272, 0.9072], device='cuda:6'), in_proj_covar=tensor([0.0435, 0.0418, 0.0510, 0.0509, 0.0465, 0.0495, 0.0500, 0.0511], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 18:19:23,391 INFO [optim.py:369] (6/7) 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 18:20:19,980 INFO [finetune.py:976] (6/7) Epoch 23, batch 1400, loss[loss=0.1892, simple_loss=0.2721, pruned_loss=0.05309, over 4836.00 frames. ], tot_loss[loss=0.1699, simple_loss=0.2424, pruned_loss=0.0487, over 955639.69 frames. ], batch size: 47, lr: 3.11e-03, grad_scale: 32.0 2023-04-27 18:20:43,122 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=127427.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 18:21:15,052 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-04-27 18:21:24,258 INFO [finetune.py:976] (6/7) Epoch 23, batch 1450, loss[loss=0.1836, simple_loss=0.2622, pruned_loss=0.05257, over 4857.00 frames. ], tot_loss[loss=0.172, simple_loss=0.2453, pruned_loss=0.04933, over 956304.88 frames. ], batch size: 31, lr: 3.11e-03, grad_scale: 32.0 2023-04-27 18:21:34,077 INFO [optim.py:369] (6/7) 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 18:21:43,925 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=127472.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:21:48,026 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.1464, 1.6497, 2.0373, 2.3715, 2.0637, 1.6058, 1.3356, 1.7945], device='cuda:6'), covar=tensor([0.3345, 0.3038, 0.1545, 0.2111, 0.2539, 0.2656, 0.3801, 0.1944], device='cuda:6'), in_proj_covar=tensor([0.0293, 0.0246, 0.0228, 0.0317, 0.0221, 0.0235, 0.0228, 0.0186], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:6') 2023-04-27 18:22:06,040 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=127488.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 18:22:07,244 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([0.9539, 1.8743, 1.7835, 1.5891, 2.0969, 1.7347, 2.5598, 1.6052], device='cuda:6'), covar=tensor([0.3298, 0.1848, 0.4382, 0.2574, 0.1478, 0.2247, 0.1234, 0.4233], device='cuda:6'), in_proj_covar=tensor([0.0335, 0.0345, 0.0424, 0.0349, 0.0374, 0.0372, 0.0367, 0.0416], device='cuda:6'), out_proj_covar=tensor([9.9394e-05, 1.0313e-04, 1.2858e-04, 1.0485e-04, 1.1139e-04, 1.1094e-04, 1.0773e-04, 1.2550e-04], device='cuda:6') 2023-04-27 18:22:29,897 INFO [finetune.py:976] (6/7) Epoch 23, batch 1500, loss[loss=0.1872, simple_loss=0.2613, pruned_loss=0.05648, over 4762.00 frames. ], tot_loss[loss=0.1736, simple_loss=0.2473, pruned_loss=0.04994, over 956778.54 frames. ], batch size: 28, lr: 3.11e-03, grad_scale: 32.0 2023-04-27 18:22:37,638 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=127520.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:22:43,189 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=127529.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:22:43,283 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-04-27 18:23:18,063 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.3767, 3.0941, 1.1347, 1.6913, 2.4841, 1.4314, 4.1577, 1.9797], device='cuda:6'), covar=tensor([0.0637, 0.0712, 0.0864, 0.1322, 0.0456, 0.1016, 0.0202, 0.0615], device='cuda:6'), in_proj_covar=tensor([0.0050, 0.0065, 0.0047, 0.0046, 0.0050, 0.0051, 0.0073, 0.0051], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:6') 2023-04-27 18:23:19,694 INFO [finetune.py:976] (6/7) Epoch 23, batch 1550, loss[loss=0.1638, simple_loss=0.2494, pruned_loss=0.03915, over 4758.00 frames. ], tot_loss[loss=0.1717, simple_loss=0.2455, pruned_loss=0.04897, over 956893.42 frames. ], batch size: 26, lr: 3.10e-03, grad_scale: 32.0 2023-04-27 18:23:23,842 INFO [optim.py:369] (6/7) 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 18:23:24,571 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.3151, 1.8518, 5.4468, 5.1305, 4.7468, 5.1948, 4.6467, 4.7913], device='cuda:6'), covar=tensor([0.6206, 0.4914, 0.0798, 0.1434, 0.0940, 0.1392, 0.1021, 0.1370], device='cuda:6'), in_proj_covar=tensor([0.0310, 0.0307, 0.0405, 0.0408, 0.0348, 0.0410, 0.0314, 0.0365], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-27 18:23:41,767 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=127590.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:23:50,932 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=127605.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:23:52,738 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6554, 3.6560, 0.9153, 1.8320, 2.0730, 2.5233, 2.0367, 1.0611], device='cuda:6'), covar=tensor([0.1383, 0.0824, 0.2092, 0.1317, 0.1102, 0.1057, 0.1436, 0.1854], device='cuda:6'), in_proj_covar=tensor([0.0116, 0.0236, 0.0136, 0.0118, 0.0131, 0.0150, 0.0115, 0.0117], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-27 18:23:53,258 INFO [finetune.py:976] (6/7) Epoch 23, batch 1600, loss[loss=0.1601, simple_loss=0.2283, pruned_loss=0.04593, over 4823.00 frames. ], tot_loss[loss=0.1702, simple_loss=0.2436, pruned_loss=0.04843, over 957048.52 frames. ], batch size: 30, lr: 3.10e-03, grad_scale: 32.0 2023-04-27 18:24:05,759 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.2726, 2.9258, 2.4647, 2.7870, 2.1095, 2.5412, 2.6605, 1.9006], device='cuda:6'), covar=tensor([0.2147, 0.1316, 0.0749, 0.1231, 0.3008, 0.1094, 0.1965, 0.2736], device='cuda:6'), in_proj_covar=tensor([0.0282, 0.0297, 0.0213, 0.0274, 0.0311, 0.0253, 0.0246, 0.0260], device='cuda:6'), out_proj_covar=tensor([1.1291e-04, 1.1749e-04, 8.3898e-05, 1.0836e-04, 1.2589e-04, 1.0015e-04, 9.9227e-05, 1.0280e-04], device='cuda:6') 2023-04-27 18:24:13,419 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-04-27 18:24:19,733 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5943, 1.5178, 1.9296, 1.9399, 1.4176, 1.3134, 1.5558, 0.9257], device='cuda:6'), covar=tensor([0.0496, 0.0537, 0.0322, 0.0487, 0.0695, 0.1054, 0.0528, 0.0598], device='cuda:6'), in_proj_covar=tensor([0.0069, 0.0068, 0.0067, 0.0068, 0.0075, 0.0096, 0.0073, 0.0065], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-27 18:24:26,894 INFO [finetune.py:976] (6/7) Epoch 23, batch 1650, loss[loss=0.185, simple_loss=0.2656, pruned_loss=0.05221, over 4758.00 frames. ], tot_loss[loss=0.1684, simple_loss=0.2411, pruned_loss=0.0478, over 956291.00 frames. ], batch size: 54, lr: 3.10e-03, grad_scale: 32.0 2023-04-27 18:24:27,607 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5143, 3.3124, 1.0130, 1.8580, 1.9489, 2.5800, 1.9104, 1.1258], device='cuda:6'), covar=tensor([0.1236, 0.0873, 0.1802, 0.1145, 0.0969, 0.0875, 0.1395, 0.1763], device='cuda:6'), in_proj_covar=tensor([0.0117, 0.0238, 0.0137, 0.0119, 0.0131, 0.0151, 0.0116, 0.0118], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-27 18:24:28,839 INFO [zipformer.py:1188] (6/7) 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] (6/7) 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 18:25:00,768 INFO [finetune.py:976] (6/7) Epoch 23, batch 1700, loss[loss=0.241, simple_loss=0.2949, pruned_loss=0.09351, over 4839.00 frames. ], tot_loss[loss=0.1679, simple_loss=0.2398, pruned_loss=0.04795, over 956314.60 frames. ], batch size: 49, lr: 3.10e-03, grad_scale: 32.0 2023-04-27 18:25:10,334 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=127723.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:25:20,498 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.0551, 2.4891, 0.7556, 1.5096, 1.4575, 1.8984, 1.5646, 0.8983], device='cuda:6'), covar=tensor([0.1406, 0.1163, 0.1726, 0.1202, 0.1096, 0.0911, 0.1520, 0.1633], device='cuda:6'), in_proj_covar=tensor([0.0117, 0.0238, 0.0137, 0.0119, 0.0131, 0.0151, 0.0116, 0.0117], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-27 18:25:34,585 INFO [finetune.py:976] (6/7) Epoch 23, batch 1750, loss[loss=0.2534, simple_loss=0.3185, pruned_loss=0.09419, over 4859.00 frames. ], tot_loss[loss=0.1713, simple_loss=0.2433, pruned_loss=0.04963, over 957739.82 frames. ], batch size: 49, lr: 3.10e-03, grad_scale: 32.0 2023-04-27 18:25:38,231 INFO [optim.py:369] (6/7) 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 18:25:48,427 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=127779.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:25:50,835 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=127783.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 18:26:14,168 INFO [finetune.py:976] (6/7) Epoch 23, batch 1800, loss[loss=0.1704, simple_loss=0.2465, pruned_loss=0.04717, over 4863.00 frames. ], tot_loss[loss=0.1734, simple_loss=0.2461, pruned_loss=0.05033, over 957519.14 frames. ], batch size: 34, lr: 3.10e-03, grad_scale: 32.0 2023-04-27 18:26:44,887 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.58 vs. limit=5.0 2023-04-27 18:26:57,479 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=127840.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:27:09,581 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 23, batch 1850, loss[loss=0.204, simple_loss=0.2743, pruned_loss=0.0668, over 4793.00 frames. ], tot_loss[loss=0.1751, simple_loss=0.2481, pruned_loss=0.05101, over 958457.11 frames. ], batch size: 51, lr: 3.10e-03, grad_scale: 32.0 2023-04-27 18:27:29,678 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.3065, 1.9696, 2.1882, 2.5220, 2.5535, 2.1179, 1.8782, 2.2667], device='cuda:6'), covar=tensor([0.0672, 0.0956, 0.0590, 0.0498, 0.0601, 0.0739, 0.0763, 0.0509], device='cuda:6'), in_proj_covar=tensor([0.0186, 0.0202, 0.0185, 0.0175, 0.0177, 0.0182, 0.0152, 0.0178], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 18:27:31,374 INFO [optim.py:369] (6/7) 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 18:27:40,526 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=127872.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:27:49,851 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=127885.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:28:04,001 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=127905.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:28:06,377 INFO [finetune.py:976] (6/7) Epoch 23, batch 1900, loss[loss=0.1823, simple_loss=0.265, pruned_loss=0.04976, over 4917.00 frames. ], tot_loss[loss=0.1754, simple_loss=0.2485, pruned_loss=0.05114, over 957435.14 frames. ], batch size: 42, lr: 3.10e-03, grad_scale: 32.0 2023-04-27 18:28:07,692 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=127911.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:28:14,059 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.32 vs. limit=5.0 2023-04-27 18:28:32,867 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=127933.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:28:57,592 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 23, batch 1950, loss[loss=0.1544, simple_loss=0.2284, pruned_loss=0.04019, over 4816.00 frames. ], tot_loss[loss=0.174, simple_loss=0.2469, pruned_loss=0.05054, over 955215.14 frames. ], batch size: 38, lr: 3.10e-03, grad_scale: 32.0 2023-04-27 18:29:16,660 INFO [optim.py:369] (6/7) 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 18:30:02,325 INFO [finetune.py:976] (6/7) Epoch 23, batch 2000, loss[loss=0.1568, simple_loss=0.2231, pruned_loss=0.04527, over 4825.00 frames. ], tot_loss[loss=0.1712, simple_loss=0.2438, pruned_loss=0.04931, over 954670.21 frames. ], batch size: 33, lr: 3.10e-03, grad_scale: 32.0 2023-04-27 18:30:07,958 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=128018.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:30:13,588 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.2337, 1.6745, 2.0104, 2.2497, 2.0118, 1.5876, 1.2069, 1.7348], device='cuda:6'), covar=tensor([0.2858, 0.3123, 0.1619, 0.1996, 0.2477, 0.2576, 0.3853, 0.1868], device='cuda:6'), in_proj_covar=tensor([0.0296, 0.0247, 0.0230, 0.0319, 0.0223, 0.0237, 0.0230, 0.0187], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:6') 2023-04-27 18:30:27,425 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7216, 0.7660, 1.5655, 2.0694, 1.7815, 1.6134, 1.6375, 1.6366], device='cuda:6'), covar=tensor([0.4557, 0.6637, 0.6090, 0.6335, 0.6142, 0.7417, 0.7346, 0.8230], device='cuda:6'), in_proj_covar=tensor([0.0434, 0.0418, 0.0509, 0.0506, 0.0464, 0.0493, 0.0498, 0.0509], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 18:30:35,022 INFO [finetune.py:976] (6/7) Epoch 23, batch 2050, loss[loss=0.1888, simple_loss=0.2696, pruned_loss=0.05404, over 4804.00 frames. ], tot_loss[loss=0.1698, simple_loss=0.2417, pruned_loss=0.04898, over 954775.61 frames. ], batch size: 51, lr: 3.10e-03, grad_scale: 32.0 2023-04-27 18:30:39,641 INFO [optim.py:369] (6/7) 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 18:30:50,719 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=128083.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 18:30:59,530 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.57 vs. limit=2.0 2023-04-27 18:31:08,826 INFO [finetune.py:976] (6/7) Epoch 23, batch 2100, loss[loss=0.1994, simple_loss=0.2739, pruned_loss=0.06245, over 4843.00 frames. ], tot_loss[loss=0.1702, simple_loss=0.2418, pruned_loss=0.04932, over 953170.51 frames. ], batch size: 47, lr: 3.10e-03, grad_scale: 32.0 2023-04-27 18:31:16,183 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.4753, 1.9033, 1.7917, 2.1981, 2.1777, 2.2744, 1.7534, 4.7121], device='cuda:6'), covar=tensor([0.0561, 0.0802, 0.0804, 0.1176, 0.0609, 0.0462, 0.0727, 0.0093], device='cuda:6'), in_proj_covar=tensor([0.0038, 0.0039, 0.0039, 0.0043, 0.0040, 0.0038, 0.0038, 0.0055], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0013, 0.0013, 0.0015, 0.0014, 0.0013, 0.0013, 0.0016], device='cuda:6') 2023-04-27 18:31:22,862 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=128131.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 18:31:25,338 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=128135.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:31:26,530 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.6889, 2.2473, 2.5293, 3.1381, 2.4905, 2.0005, 1.9668, 2.5487], device='cuda:6'), covar=tensor([0.2997, 0.2876, 0.1574, 0.2472, 0.2495, 0.2459, 0.3565, 0.1849], device='cuda:6'), in_proj_covar=tensor([0.0294, 0.0246, 0.0229, 0.0317, 0.0221, 0.0235, 0.0228, 0.0186], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:6') 2023-04-27 18:31:40,698 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.0228, 1.2874, 1.2039, 1.5651, 1.4165, 1.5305, 1.2383, 2.5080], device='cuda:6'), covar=tensor([0.0666, 0.0865, 0.0859, 0.1267, 0.0690, 0.0512, 0.0771, 0.0222], device='cuda:6'), in_proj_covar=tensor([0.0038, 0.0039, 0.0039, 0.0043, 0.0040, 0.0038, 0.0038, 0.0055], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0013, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:6') 2023-04-27 18:31:42,235 INFO [finetune.py:976] (6/7) Epoch 23, batch 2150, loss[loss=0.1721, simple_loss=0.2463, pruned_loss=0.04898, over 4757.00 frames. ], tot_loss[loss=0.1718, simple_loss=0.2439, pruned_loss=0.04984, over 952595.13 frames. ], batch size: 28, lr: 3.10e-03, grad_scale: 32.0 2023-04-27 18:31:44,542 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.76 vs. limit=5.0 2023-04-27 18:31:46,839 INFO [optim.py:369] (6/7) 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 18:31:52,483 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-04-27 18:31:59,138 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=128185.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:32:11,731 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.4706, 3.3827, 0.7943, 1.8817, 1.8392, 2.4563, 1.7831, 1.0380], device='cuda:6'), covar=tensor([0.1376, 0.0903, 0.2177, 0.1278, 0.1161, 0.1007, 0.1647, 0.2016], device='cuda:6'), in_proj_covar=tensor([0.0117, 0.0239, 0.0137, 0.0119, 0.0131, 0.0151, 0.0117, 0.0118], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-27 18:32:13,545 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=128206.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:32:15,341 INFO [finetune.py:976] (6/7) Epoch 23, batch 2200, loss[loss=0.223, simple_loss=0.2865, pruned_loss=0.07973, over 4908.00 frames. ], tot_loss[loss=0.1742, simple_loss=0.2466, pruned_loss=0.0509, over 952528.27 frames. ], batch size: 37, lr: 3.10e-03, grad_scale: 32.0 2023-04-27 18:32:39,410 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=128228.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:32:42,389 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=128233.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:33:12,836 INFO [finetune.py:976] (6/7) Epoch 23, batch 2250, loss[loss=0.1897, simple_loss=0.2679, pruned_loss=0.0557, over 4890.00 frames. ], tot_loss[loss=0.175, simple_loss=0.2476, pruned_loss=0.05118, over 952938.92 frames. ], batch size: 32, lr: 3.10e-03, grad_scale: 32.0 2023-04-27 18:33:22,484 INFO [optim.py:369] (6/7) 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 18:34:26,340 INFO [finetune.py:976] (6/7) Epoch 23, batch 2300, loss[loss=0.2091, simple_loss=0.2775, pruned_loss=0.07036, over 4140.00 frames. ], tot_loss[loss=0.1738, simple_loss=0.2471, pruned_loss=0.05028, over 951609.62 frames. ], batch size: 65, lr: 3.10e-03, grad_scale: 32.0 2023-04-27 18:34:37,294 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=128318.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:34:39,049 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=128320.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:35:34,257 INFO [finetune.py:976] (6/7) Epoch 23, batch 2350, loss[loss=0.1713, simple_loss=0.2391, pruned_loss=0.05172, over 4821.00 frames. ], tot_loss[loss=0.1725, simple_loss=0.2451, pruned_loss=0.04995, over 951619.13 frames. ], batch size: 39, lr: 3.10e-03, grad_scale: 32.0 2023-04-27 18:35:37,973 INFO [optim.py:369] (6/7) 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 18:35:39,775 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=128366.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:36:01,143 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=128381.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:36:02,952 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.1369, 1.4973, 1.5713, 2.0834, 2.2272, 1.8104, 1.8210, 1.5316], device='cuda:6'), covar=tensor([0.2000, 0.1799, 0.1729, 0.1442, 0.1068, 0.1969, 0.2030, 0.2375], device='cuda:6'), in_proj_covar=tensor([0.0316, 0.0312, 0.0352, 0.0289, 0.0328, 0.0311, 0.0300, 0.0374], device='cuda:6'), out_proj_covar=tensor([6.4864e-05, 6.4589e-05, 7.4266e-05, 5.8243e-05, 6.7598e-05, 6.5134e-05, 6.2562e-05, 7.9317e-05], device='cuda:6') 2023-04-27 18:36:31,442 INFO [finetune.py:976] (6/7) Epoch 23, batch 2400, loss[loss=0.1306, simple_loss=0.204, pruned_loss=0.02859, over 4827.00 frames. ], tot_loss[loss=0.1695, simple_loss=0.2417, pruned_loss=0.04867, over 953753.83 frames. ], batch size: 39, lr: 3.10e-03, grad_scale: 32.0 2023-04-27 18:36:34,506 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=128413.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:37:01,421 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=128435.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:37:05,973 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-04-27 18:37:21,181 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.3709, 3.2262, 0.9022, 1.7416, 1.8359, 2.2889, 1.8238, 1.0096], device='cuda:6'), covar=tensor([0.1373, 0.1002, 0.1904, 0.1274, 0.1038, 0.1073, 0.1475, 0.1954], device='cuda:6'), in_proj_covar=tensor([0.0117, 0.0239, 0.0137, 0.0119, 0.0131, 0.0151, 0.0117, 0.0118], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-27 18:37:29,234 INFO [finetune.py:976] (6/7) Epoch 23, batch 2450, loss[loss=0.2247, simple_loss=0.2858, pruned_loss=0.08185, over 4792.00 frames. ], tot_loss[loss=0.1686, simple_loss=0.2399, pruned_loss=0.04863, over 953855.94 frames. ], batch size: 29, lr: 3.10e-03, grad_scale: 32.0 2023-04-27 18:37:35,178 INFO [optim.py:369] (6/7) 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 18:37:46,710 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=128474.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:37:56,946 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=128483.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:37:58,861 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.57 vs. limit=2.0 2023-04-27 18:38:27,903 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 23, batch 2500, loss[loss=0.1597, simple_loss=0.2297, pruned_loss=0.04482, over 4910.00 frames. ], tot_loss[loss=0.1691, simple_loss=0.2405, pruned_loss=0.04882, over 955576.28 frames. ], batch size: 35, lr: 3.10e-03, grad_scale: 32.0 2023-04-27 18:38:53,767 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=128528.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:39:23,548 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.6136, 2.3934, 2.7674, 3.0535, 2.6397, 2.5260, 2.6951, 2.5874], device='cuda:6'), covar=tensor([0.4524, 0.6410, 0.6665, 0.5081, 0.5656, 0.7539, 0.7461, 0.7431], device='cuda:6'), in_proj_covar=tensor([0.0435, 0.0420, 0.0511, 0.0507, 0.0465, 0.0495, 0.0500, 0.0511], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 18:39:29,775 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=128554.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:39:32,696 INFO [finetune.py:976] (6/7) Epoch 23, batch 2550, loss[loss=0.1901, simple_loss=0.261, pruned_loss=0.05957, over 4934.00 frames. ], tot_loss[loss=0.1706, simple_loss=0.2426, pruned_loss=0.0493, over 955554.13 frames. ], batch size: 33, lr: 3.10e-03, grad_scale: 32.0 2023-04-27 18:39:42,965 INFO [optim.py:369] (6/7) 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 18:39:55,802 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=128576.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:39:55,942 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.59 vs. limit=2.0 2023-04-27 18:40:40,369 INFO [finetune.py:976] (6/7) Epoch 23, batch 2600, loss[loss=0.1652, simple_loss=0.2414, pruned_loss=0.04454, over 4752.00 frames. ], tot_loss[loss=0.1703, simple_loss=0.2434, pruned_loss=0.04863, over 956583.57 frames. ], batch size: 27, lr: 3.10e-03, grad_scale: 32.0 2023-04-27 18:40:58,862 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.7134, 2.1975, 2.4171, 3.1346, 2.6031, 2.0047, 2.0392, 2.5309], device='cuda:6'), covar=tensor([0.3237, 0.3078, 0.1751, 0.2412, 0.2577, 0.2747, 0.3553, 0.1826], device='cuda:6'), in_proj_covar=tensor([0.0292, 0.0245, 0.0227, 0.0314, 0.0219, 0.0234, 0.0227, 0.0184], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:6') 2023-04-27 18:41:01,275 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.1159, 2.4577, 2.1456, 2.5212, 1.7603, 2.1537, 2.1288, 1.6764], device='cuda:6'), covar=tensor([0.1866, 0.1260, 0.0754, 0.0988, 0.3110, 0.1042, 0.1849, 0.2404], device='cuda:6'), in_proj_covar=tensor([0.0283, 0.0299, 0.0213, 0.0274, 0.0312, 0.0254, 0.0246, 0.0261], device='cuda:6'), out_proj_covar=tensor([1.1291e-04, 1.1814e-04, 8.4049e-05, 1.0808e-04, 1.2617e-04, 1.0046e-04, 9.9457e-05, 1.0309e-04], device='cuda:6') 2023-04-27 18:41:07,474 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8996, 1.5483, 1.4488, 1.8023, 2.1123, 1.6825, 1.5413, 1.3445], device='cuda:6'), covar=tensor([0.1705, 0.1510, 0.1938, 0.1080, 0.0830, 0.1670, 0.2154, 0.2406], device='cuda:6'), in_proj_covar=tensor([0.0314, 0.0310, 0.0352, 0.0286, 0.0325, 0.0308, 0.0299, 0.0371], device='cuda:6'), out_proj_covar=tensor([6.4473e-05, 6.4125e-05, 7.4115e-05, 5.7568e-05, 6.7022e-05, 6.4678e-05, 6.2322e-05, 7.8812e-05], device='cuda:6') 2023-04-27 18:41:44,293 INFO [finetune.py:976] (6/7) Epoch 23, batch 2650, loss[loss=0.1411, simple_loss=0.2144, pruned_loss=0.03393, over 4727.00 frames. ], tot_loss[loss=0.1716, simple_loss=0.245, pruned_loss=0.04912, over 956805.61 frames. ], batch size: 23, lr: 3.10e-03, grad_scale: 32.0 2023-04-27 18:41:53,486 INFO [optim.py:369] (6/7) 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 18:42:03,683 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0623, 1.7581, 2.2469, 2.5624, 2.0944, 2.0021, 2.1300, 2.0569], device='cuda:6'), covar=tensor([0.4859, 0.7510, 0.7758, 0.5549, 0.6137, 0.9558, 0.9028, 0.9690], device='cuda:6'), in_proj_covar=tensor([0.0434, 0.0420, 0.0511, 0.0507, 0.0464, 0.0495, 0.0500, 0.0510], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 18:42:11,799 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=128676.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:42:55,233 INFO [finetune.py:976] (6/7) Epoch 23, batch 2700, loss[loss=0.1883, simple_loss=0.2497, pruned_loss=0.0635, over 4833.00 frames. ], tot_loss[loss=0.169, simple_loss=0.2429, pruned_loss=0.04752, over 958218.34 frames. ], batch size: 49, lr: 3.10e-03, grad_scale: 32.0 2023-04-27 18:43:18,355 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6492, 3.2165, 1.1829, 1.9286, 1.9202, 2.4812, 1.9613, 1.2399], device='cuda:6'), covar=tensor([0.1200, 0.0853, 0.1707, 0.1134, 0.0969, 0.0884, 0.1348, 0.1953], device='cuda:6'), in_proj_covar=tensor([0.0117, 0.0238, 0.0137, 0.0119, 0.0131, 0.0151, 0.0117, 0.0118], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-27 18:43:20,830 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.4375, 1.3582, 1.7673, 1.7377, 1.3565, 1.2145, 1.5223, 0.9121], device='cuda:6'), covar=tensor([0.0503, 0.0611, 0.0342, 0.0563, 0.0716, 0.1062, 0.0536, 0.0606], device='cuda:6'), in_proj_covar=tensor([0.0069, 0.0067, 0.0066, 0.0067, 0.0074, 0.0094, 0.0072, 0.0063], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-27 18:44:00,968 INFO [finetune.py:976] (6/7) Epoch 23, batch 2750, loss[loss=0.1392, simple_loss=0.2232, pruned_loss=0.02764, over 4791.00 frames. ], tot_loss[loss=0.1682, simple_loss=0.2414, pruned_loss=0.04748, over 957434.14 frames. ], batch size: 29, lr: 3.10e-03, grad_scale: 32.0 2023-04-27 18:44:04,650 INFO [optim.py:369] (6/7) 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] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=128769.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:44:13,457 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([3.3030, 3.2565, 2.4636, 3.7784, 3.2210, 3.2435, 1.4953, 3.2624], device='cuda:6'), covar=tensor([0.2025, 0.1463, 0.3698, 0.2604, 0.3039, 0.2010, 0.5699, 0.2571], device='cuda:6'), in_proj_covar=tensor([0.0246, 0.0219, 0.0254, 0.0307, 0.0297, 0.0247, 0.0275, 0.0274], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 18:44:26,269 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-04-27 18:44:32,561 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=128781.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:45:09,449 INFO [finetune.py:976] (6/7) Epoch 23, batch 2800, loss[loss=0.1519, simple_loss=0.2289, pruned_loss=0.03741, over 4826.00 frames. ], tot_loss[loss=0.1667, simple_loss=0.2392, pruned_loss=0.04712, over 956143.16 frames. ], batch size: 39, lr: 3.10e-03, grad_scale: 32.0 2023-04-27 18:45:34,247 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9437, 2.0862, 1.0612, 1.6169, 2.2298, 1.7678, 1.7387, 1.8188], device='cuda:6'), covar=tensor([0.0460, 0.0343, 0.0287, 0.0536, 0.0227, 0.0485, 0.0488, 0.0552], device='cuda:6'), in_proj_covar=tensor([0.0028, 0.0024, 0.0022, 0.0028, 0.0019, 0.0028, 0.0027, 0.0028], device='cuda:6'), out_proj_covar=tensor([0.0051, 0.0046, 0.0038, 0.0052, 0.0039, 0.0050, 0.0050, 0.0052], device='cuda:6') 2023-04-27 18:45:55,106 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=128842.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:46:17,710 INFO [finetune.py:976] (6/7) Epoch 23, batch 2850, loss[loss=0.1197, simple_loss=0.2004, pruned_loss=0.01943, over 4766.00 frames. ], tot_loss[loss=0.1666, simple_loss=0.2384, pruned_loss=0.04739, over 954765.96 frames. ], batch size: 26, lr: 3.10e-03, grad_scale: 32.0 2023-04-27 18:46:22,646 INFO [optim.py:369] (6/7) 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 18:46:33,850 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.8592, 2.8671, 2.2575, 3.2500, 2.8711, 2.8217, 1.2295, 2.7701], device='cuda:6'), covar=tensor([0.2362, 0.1790, 0.3515, 0.3095, 0.4588, 0.2295, 0.6104, 0.3080], device='cuda:6'), in_proj_covar=tensor([0.0243, 0.0217, 0.0251, 0.0304, 0.0294, 0.0245, 0.0272, 0.0271], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 18:46:48,999 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.4799, 1.3831, 1.8261, 1.8264, 1.3880, 1.2873, 1.5679, 0.9389], device='cuda:6'), covar=tensor([0.0563, 0.0818, 0.0331, 0.0604, 0.0730, 0.1049, 0.0549, 0.0591], device='cuda:6'), in_proj_covar=tensor([0.0068, 0.0067, 0.0065, 0.0067, 0.0074, 0.0094, 0.0072, 0.0063], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-27 18:46:50,223 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 23, batch 2900, loss[loss=0.1692, simple_loss=0.2466, pruned_loss=0.04592, over 4813.00 frames. ], tot_loss[loss=0.1695, simple_loss=0.2421, pruned_loss=0.04849, over 956028.39 frames. ], batch size: 38, lr: 3.09e-03, grad_scale: 32.0 2023-04-27 18:47:28,319 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.4393, 1.7556, 1.8919, 1.9723, 1.8521, 1.8438, 1.9533, 1.9177], device='cuda:6'), covar=tensor([0.4840, 0.6000, 0.4629, 0.4457, 0.5516, 0.7089, 0.5221, 0.4730], device='cuda:6'), in_proj_covar=tensor([0.0337, 0.0372, 0.0324, 0.0336, 0.0346, 0.0392, 0.0355, 0.0328], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 18:47:56,538 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=128951.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 18:48:04,749 INFO [finetune.py:976] (6/7) Epoch 23, batch 2950, loss[loss=0.1254, simple_loss=0.1916, pruned_loss=0.02959, over 4138.00 frames. ], tot_loss[loss=0.1712, simple_loss=0.2441, pruned_loss=0.04911, over 953119.18 frames. ], batch size: 18, lr: 3.09e-03, grad_scale: 32.0 2023-04-27 18:48:10,256 INFO [optim.py:369] (6/7) 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 18:48:22,828 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=128976.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:49:01,106 INFO [finetune.py:976] (6/7) Epoch 23, batch 3000, loss[loss=0.2164, simple_loss=0.2874, pruned_loss=0.07272, over 4926.00 frames. ], tot_loss[loss=0.1745, simple_loss=0.2477, pruned_loss=0.05064, over 955310.89 frames. ], batch size: 33, lr: 3.09e-03, grad_scale: 32.0 2023-04-27 18:49:01,107 INFO [finetune.py:1001] (6/7) Computing validation loss 2023-04-27 18:49:21,212 INFO [finetune.py:1010] (6/7) Epoch 23, validation: loss=0.1527, simple_loss=0.2222, pruned_loss=0.04158, over 2265189.00 frames. 2023-04-27 18:49:21,212 INFO [finetune.py:1011] (6/7) Maximum memory allocated so far is 6435MB 2023-04-27 18:49:31,993 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=129024.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:50:14,494 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([0.9135, 1.3059, 3.2608, 3.0224, 2.9292, 3.2176, 3.1775, 2.8579], device='cuda:6'), covar=tensor([0.6966, 0.4986, 0.1408, 0.2100, 0.1329, 0.1807, 0.1962, 0.1792], device='cuda:6'), in_proj_covar=tensor([0.0308, 0.0306, 0.0406, 0.0410, 0.0349, 0.0411, 0.0315, 0.0367], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-27 18:50:16,854 INFO [finetune.py:976] (6/7) Epoch 23, batch 3050, loss[loss=0.1522, simple_loss=0.2234, pruned_loss=0.04051, over 4920.00 frames. ], tot_loss[loss=0.1742, simple_loss=0.2474, pruned_loss=0.05046, over 956053.10 frames. ], batch size: 33, lr: 3.09e-03, grad_scale: 32.0 2023-04-27 18:50:25,141 INFO [optim.py:369] (6/7) 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 18:50:26,383 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7121, 1.2906, 1.8343, 2.2059, 1.7838, 1.7419, 1.8485, 1.7352], device='cuda:6'), covar=tensor([0.4640, 0.6390, 0.6276, 0.5633, 0.6041, 0.7795, 0.7295, 0.9138], device='cuda:6'), in_proj_covar=tensor([0.0435, 0.0419, 0.0512, 0.0508, 0.0465, 0.0496, 0.0499, 0.0511], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 18:50:34,476 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=129069.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:51:23,716 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.11 vs. limit=2.0 2023-04-27 18:51:30,729 INFO [finetune.py:976] (6/7) Epoch 23, batch 3100, loss[loss=0.166, simple_loss=0.2443, pruned_loss=0.0438, over 4830.00 frames. ], tot_loss[loss=0.1719, simple_loss=0.2453, pruned_loss=0.04929, over 957793.80 frames. ], batch size: 33, lr: 3.09e-03, grad_scale: 16.0 2023-04-27 18:51:41,687 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=129117.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:52:07,898 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=129137.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:52:29,955 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5521, 3.1719, 1.0537, 1.9178, 1.8802, 2.4155, 1.9522, 1.1870], device='cuda:6'), covar=tensor([0.1217, 0.0753, 0.1778, 0.1126, 0.0969, 0.0892, 0.1245, 0.1860], device='cuda:6'), in_proj_covar=tensor([0.0116, 0.0235, 0.0135, 0.0118, 0.0130, 0.0149, 0.0116, 0.0117], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-27 18:52:39,819 INFO [finetune.py:976] (6/7) Epoch 23, batch 3150, loss[loss=0.1575, simple_loss=0.2285, pruned_loss=0.04325, over 4834.00 frames. ], tot_loss[loss=0.1711, simple_loss=0.2431, pruned_loss=0.04956, over 955408.44 frames. ], batch size: 30, lr: 3.09e-03, grad_scale: 16.0 2023-04-27 18:52:50,086 INFO [optim.py:369] (6/7) 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 18:53:46,919 INFO [finetune.py:976] (6/7) Epoch 23, batch 3200, loss[loss=0.1486, simple_loss=0.2191, pruned_loss=0.03905, over 4908.00 frames. ], tot_loss[loss=0.1679, simple_loss=0.2396, pruned_loss=0.04807, over 956880.45 frames. ], batch size: 43, lr: 3.09e-03, grad_scale: 16.0 2023-04-27 18:54:11,210 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-04-27 18:54:20,956 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([0.6385, 1.6708, 1.7583, 1.3390, 1.7640, 1.4636, 2.2390, 1.5642], device='cuda:6'), covar=tensor([0.3238, 0.1682, 0.4075, 0.2494, 0.1520, 0.2138, 0.1276, 0.3906], device='cuda:6'), in_proj_covar=tensor([0.0337, 0.0347, 0.0424, 0.0348, 0.0375, 0.0373, 0.0365, 0.0417], device='cuda:6'), out_proj_covar=tensor([9.9903e-05, 1.0372e-04, 1.2866e-04, 1.0482e-04, 1.1148e-04, 1.1116e-04, 1.0720e-04, 1.2570e-04], device='cuda:6') 2023-04-27 18:54:42,094 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=129246.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 18:54:44,203 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 2023-04-27 18:54:55,918 INFO [finetune.py:976] (6/7) Epoch 23, batch 3250, loss[loss=0.1947, simple_loss=0.2644, pruned_loss=0.06255, over 4816.00 frames. ], tot_loss[loss=0.169, simple_loss=0.2404, pruned_loss=0.04878, over 957738.44 frames. ], batch size: 33, lr: 3.09e-03, grad_scale: 16.0 2023-04-27 18:55:06,951 INFO [optim.py:369] (6/7) 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 18:55:39,627 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([3.7455, 3.7896, 2.8369, 4.3548, 3.8691, 3.6998, 1.6455, 3.7083], device='cuda:6'), covar=tensor([0.1853, 0.1366, 0.3073, 0.1696, 0.4627, 0.1858, 0.6575, 0.2756], device='cuda:6'), in_proj_covar=tensor([0.0245, 0.0219, 0.0253, 0.0305, 0.0296, 0.0246, 0.0273, 0.0273], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 18:55:51,326 INFO [finetune.py:976] (6/7) Epoch 23, batch 3300, loss[loss=0.1594, simple_loss=0.2502, pruned_loss=0.03431, over 4766.00 frames. ], tot_loss[loss=0.1707, simple_loss=0.2433, pruned_loss=0.04904, over 956558.28 frames. ], batch size: 28, lr: 3.09e-03, grad_scale: 16.0 2023-04-27 18:56:51,618 INFO [finetune.py:976] (6/7) Epoch 23, batch 3350, loss[loss=0.2075, simple_loss=0.2927, pruned_loss=0.06111, over 4900.00 frames. ], tot_loss[loss=0.172, simple_loss=0.2452, pruned_loss=0.04938, over 957589.53 frames. ], batch size: 43, lr: 3.09e-03, grad_scale: 16.0 2023-04-27 18:57:02,544 INFO [optim.py:369] (6/7) 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 18:57:58,594 INFO [finetune.py:976] (6/7) Epoch 23, batch 3400, loss[loss=0.1339, simple_loss=0.2131, pruned_loss=0.02733, over 4737.00 frames. ], tot_loss[loss=0.1725, simple_loss=0.2459, pruned_loss=0.04952, over 957425.72 frames. ], batch size: 27, lr: 3.09e-03, grad_scale: 16.0 2023-04-27 18:58:06,882 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=129413.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 18:58:40,740 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=129437.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:58:41,843 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0732, 2.3615, 1.0688, 1.3645, 1.8105, 1.1843, 3.0998, 1.6479], device='cuda:6'), covar=tensor([0.0691, 0.0715, 0.0725, 0.1221, 0.0497, 0.1041, 0.0269, 0.0657], device='cuda:6'), in_proj_covar=tensor([0.0050, 0.0063, 0.0047, 0.0045, 0.0049, 0.0050, 0.0071, 0.0050], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0009, 0.0007, 0.0007, 0.0008, 0.0008, 0.0010, 0.0007], device='cuda:6') 2023-04-27 18:59:06,103 INFO [finetune.py:976] (6/7) Epoch 23, batch 3450, loss[loss=0.1095, simple_loss=0.1753, pruned_loss=0.02183, over 4737.00 frames. ], tot_loss[loss=0.1724, simple_loss=0.2459, pruned_loss=0.04943, over 955580.23 frames. ], batch size: 23, lr: 3.09e-03, grad_scale: 16.0 2023-04-27 18:59:15,876 INFO [optim.py:369] (6/7) 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 18:59:26,978 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=129474.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 18:59:44,263 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=129485.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:00:06,157 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-04-27 19:00:07,511 INFO [finetune.py:976] (6/7) Epoch 23, batch 3500, loss[loss=0.1789, simple_loss=0.2559, pruned_loss=0.05094, over 4765.00 frames. ], tot_loss[loss=0.1709, simple_loss=0.244, pruned_loss=0.04889, over 955589.52 frames. ], batch size: 26, lr: 3.09e-03, grad_scale: 16.0 2023-04-27 19:00:40,774 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.2935, 1.5000, 1.8369, 1.9605, 1.8326, 1.9605, 1.8630, 1.9014], device='cuda:6'), covar=tensor([0.3698, 0.5011, 0.4292, 0.3810, 0.5187, 0.6810, 0.4699, 0.4361], device='cuda:6'), in_proj_covar=tensor([0.0341, 0.0377, 0.0329, 0.0340, 0.0349, 0.0397, 0.0359, 0.0332], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 19:00:46,479 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=129546.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 19:00:47,242 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 2023-04-27 19:01:00,752 INFO [finetune.py:976] (6/7) Epoch 23, batch 3550, loss[loss=0.1318, simple_loss=0.1984, pruned_loss=0.03256, over 4923.00 frames. ], tot_loss[loss=0.1683, simple_loss=0.2409, pruned_loss=0.04786, over 955811.97 frames. ], batch size: 33, lr: 3.09e-03, grad_scale: 16.0 2023-04-27 19:01:05,541 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=3.80 vs. limit=5.0 2023-04-27 19:01:08,388 INFO [optim.py:369] (6/7) 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 19:01:30,956 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-04-27 19:01:38,378 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=129594.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 19:01:57,768 INFO [finetune.py:976] (6/7) Epoch 23, batch 3600, loss[loss=0.1657, simple_loss=0.2303, pruned_loss=0.05053, over 4755.00 frames. ], tot_loss[loss=0.1666, simple_loss=0.2385, pruned_loss=0.0474, over 955992.35 frames. ], batch size: 28, lr: 3.09e-03, grad_scale: 16.0 2023-04-27 19:01:59,681 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=129612.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:02:47,465 INFO [finetune.py:976] (6/7) Epoch 23, batch 3650, loss[loss=0.2066, simple_loss=0.2791, pruned_loss=0.06708, over 4827.00 frames. ], tot_loss[loss=0.1681, simple_loss=0.2401, pruned_loss=0.04808, over 954874.48 frames. ], batch size: 40, lr: 3.09e-03, grad_scale: 16.0 2023-04-27 19:02:51,891 INFO [optim.py:369] (6/7) 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 19:02:52,040 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([4.0069, 3.9830, 2.8902, 4.6484, 4.0864, 3.9840, 1.8219, 3.9952], device='cuda:6'), covar=tensor([0.1650, 0.1241, 0.3198, 0.1492, 0.4062, 0.1731, 0.5978, 0.2282], device='cuda:6'), in_proj_covar=tensor([0.0245, 0.0218, 0.0253, 0.0306, 0.0296, 0.0246, 0.0274, 0.0274], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 19:03:00,815 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.5736, 1.9621, 2.0274, 2.1225, 1.9976, 2.0931, 2.0878, 2.0442], device='cuda:6'), covar=tensor([0.3912, 0.5194, 0.4367, 0.4394, 0.5446, 0.6620, 0.5220, 0.5058], device='cuda:6'), in_proj_covar=tensor([0.0339, 0.0374, 0.0327, 0.0339, 0.0347, 0.0394, 0.0357, 0.0330], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 19:03:02,085 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=129673.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:03:55,978 INFO [finetune.py:976] (6/7) Epoch 23, batch 3700, loss[loss=0.1727, simple_loss=0.2493, pruned_loss=0.048, over 4864.00 frames. ], tot_loss[loss=0.1698, simple_loss=0.2425, pruned_loss=0.04857, over 952763.35 frames. ], batch size: 44, lr: 3.09e-03, grad_scale: 16.0 2023-04-27 19:04:41,832 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0102, 1.1738, 5.2312, 4.8861, 4.6139, 5.0125, 4.6752, 4.6186], device='cuda:6'), covar=tensor([0.6981, 0.6825, 0.0948, 0.1963, 0.1145, 0.1299, 0.1255, 0.1494], device='cuda:6'), in_proj_covar=tensor([0.0309, 0.0307, 0.0407, 0.0406, 0.0347, 0.0409, 0.0315, 0.0366], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-27 19:05:03,363 INFO [finetune.py:976] (6/7) Epoch 23, batch 3750, loss[loss=0.1523, simple_loss=0.2354, pruned_loss=0.03458, over 4816.00 frames. ], tot_loss[loss=0.1727, simple_loss=0.2456, pruned_loss=0.04993, over 952784.45 frames. ], batch size: 40, lr: 3.09e-03, grad_scale: 16.0 2023-04-27 19:05:12,865 INFO [optim.py:369] (6/7) 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] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=129769.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 19:05:23,611 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.3091, 1.9140, 2.3509, 2.7012, 2.3192, 2.1059, 2.2454, 2.1319], device='cuda:6'), covar=tensor([0.4076, 0.6108, 0.6028, 0.5359, 0.5746, 0.8082, 0.8489, 0.8166], device='cuda:6'), in_proj_covar=tensor([0.0433, 0.0417, 0.0511, 0.0507, 0.0462, 0.0495, 0.0499, 0.0509], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 19:05:45,765 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=129791.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:06:08,152 INFO [finetune.py:976] (6/7) Epoch 23, batch 3800, loss[loss=0.1412, simple_loss=0.2245, pruned_loss=0.02892, over 4803.00 frames. ], tot_loss[loss=0.1719, simple_loss=0.2453, pruned_loss=0.04929, over 952430.97 frames. ], batch size: 25, lr: 3.09e-03, grad_scale: 16.0 2023-04-27 19:06:19,800 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9597, 2.3631, 2.0381, 2.3721, 1.7720, 2.0799, 1.9820, 1.7579], device='cuda:6'), covar=tensor([0.1632, 0.1076, 0.0693, 0.0970, 0.2908, 0.0996, 0.1717, 0.2096], device='cuda:6'), in_proj_covar=tensor([0.0281, 0.0298, 0.0212, 0.0273, 0.0311, 0.0253, 0.0245, 0.0260], device='cuda:6'), out_proj_covar=tensor([1.1242e-04, 1.1788e-04, 8.3659e-05, 1.0764e-04, 1.2551e-04, 1.0003e-04, 9.8816e-05, 1.0236e-04], device='cuda:6') 2023-04-27 19:07:04,484 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=129852.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:07:13,929 INFO [finetune.py:976] (6/7) Epoch 23, batch 3850, loss[loss=0.1601, simple_loss=0.2281, pruned_loss=0.04601, over 4831.00 frames. ], tot_loss[loss=0.1708, simple_loss=0.2439, pruned_loss=0.04886, over 954022.43 frames. ], batch size: 25, lr: 3.09e-03, grad_scale: 16.0 2023-04-27 19:07:24,864 INFO [optim.py:369] (6/7) 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 19:07:57,987 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9889, 1.9719, 4.6036, 4.3692, 4.1338, 4.4072, 4.2699, 4.1418], device='cuda:6'), covar=tensor([0.5720, 0.4535, 0.0917, 0.1336, 0.0887, 0.1667, 0.0954, 0.1279], device='cuda:6'), in_proj_covar=tensor([0.0308, 0.0305, 0.0404, 0.0404, 0.0345, 0.0406, 0.0313, 0.0364], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-27 19:08:15,872 INFO [finetune.py:976] (6/7) Epoch 23, batch 3900, loss[loss=0.1608, simple_loss=0.2351, pruned_loss=0.04327, over 4271.00 frames. ], tot_loss[loss=0.17, simple_loss=0.2422, pruned_loss=0.04887, over 951923.65 frames. ], batch size: 65, lr: 3.09e-03, grad_scale: 16.0 2023-04-27 19:09:17,307 INFO [finetune.py:976] (6/7) Epoch 23, batch 3950, loss[loss=0.1521, simple_loss=0.2193, pruned_loss=0.04248, over 4898.00 frames. ], tot_loss[loss=0.1671, simple_loss=0.2385, pruned_loss=0.04787, over 949135.00 frames. ], batch size: 32, lr: 3.09e-03, grad_scale: 16.0 2023-04-27 19:09:27,940 INFO [optim.py:369] (6/7) 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 19:09:35,336 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=129968.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:10:16,066 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=130000.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:10:22,640 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=130002.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:10:33,623 INFO [finetune.py:976] (6/7) Epoch 23, batch 4000, loss[loss=0.1478, simple_loss=0.2202, pruned_loss=0.03774, over 4903.00 frames. ], tot_loss[loss=0.1665, simple_loss=0.2375, pruned_loss=0.04773, over 951425.53 frames. ], batch size: 35, lr: 3.09e-03, grad_scale: 16.0 2023-04-27 19:11:43,267 INFO [finetune.py:976] (6/7) Epoch 23, batch 4050, loss[loss=0.1963, simple_loss=0.28, pruned_loss=0.05635, over 4835.00 frames. ], tot_loss[loss=0.1701, simple_loss=0.2417, pruned_loss=0.04923, over 953151.37 frames. ], batch size: 47, lr: 3.09e-03, grad_scale: 16.0 2023-04-27 19:11:44,628 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=130061.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:11:44,635 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0047, 2.6302, 2.0481, 2.3087, 1.4549, 1.4324, 2.0321, 1.3830], device='cuda:6'), covar=tensor([0.1590, 0.1427, 0.1288, 0.1474, 0.2117, 0.1891, 0.0978, 0.1988], device='cuda:6'), in_proj_covar=tensor([0.0198, 0.0212, 0.0170, 0.0205, 0.0200, 0.0187, 0.0157, 0.0187], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:6') 2023-04-27 19:11:45,893 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=130063.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 19:11:53,120 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.3877, 1.0388, 0.3717, 1.1004, 1.0735, 1.2750, 1.1735, 1.1743], device='cuda:6'), covar=tensor([0.0499, 0.0404, 0.0454, 0.0565, 0.0340, 0.0510, 0.0498, 0.0584], device='cuda:6'), in_proj_covar=tensor([0.0028, 0.0023, 0.0022, 0.0028, 0.0019, 0.0027, 0.0027, 0.0028], device='cuda:6'), out_proj_covar=tensor([0.0051, 0.0045, 0.0038, 0.0052, 0.0038, 0.0050, 0.0050, 0.0051], device='cuda:6') 2023-04-27 19:11:53,611 INFO [optim.py:369] (6/7) 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 19:11:56,092 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=130069.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 19:12:19,141 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([3.3854, 3.4773, 2.7075, 3.9818, 3.4187, 3.3951, 1.5849, 3.4627], device='cuda:6'), covar=tensor([0.1885, 0.1277, 0.3564, 0.1911, 0.3051, 0.1929, 0.5562, 0.2425], device='cuda:6'), in_proj_covar=tensor([0.0246, 0.0219, 0.0253, 0.0306, 0.0297, 0.0246, 0.0274, 0.0274], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 19:12:39,044 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.4182, 1.2750, 1.7069, 1.6942, 1.3234, 1.2216, 1.4284, 0.9283], device='cuda:6'), covar=tensor([0.0528, 0.0927, 0.0413, 0.0754, 0.0805, 0.1133, 0.0661, 0.0649], device='cuda:6'), in_proj_covar=tensor([0.0070, 0.0068, 0.0067, 0.0068, 0.0075, 0.0096, 0.0073, 0.0064], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-27 19:12:52,034 INFO [finetune.py:976] (6/7) Epoch 23, batch 4100, loss[loss=0.2673, simple_loss=0.326, pruned_loss=0.1043, over 4219.00 frames. ], tot_loss[loss=0.1712, simple_loss=0.2436, pruned_loss=0.04942, over 951320.92 frames. ], batch size: 65, lr: 3.09e-03, grad_scale: 16.0 2023-04-27 19:13:01,825 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=130117.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 19:13:32,158 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=130147.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:13:51,931 INFO [finetune.py:976] (6/7) Epoch 23, batch 4150, loss[loss=0.1767, simple_loss=0.2573, pruned_loss=0.04806, over 4759.00 frames. ], tot_loss[loss=0.1716, simple_loss=0.2445, pruned_loss=0.0494, over 953657.65 frames. ], batch size: 26, lr: 3.09e-03, grad_scale: 16.0 2023-04-27 19:14:02,700 INFO [optim.py:369] (6/7) 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 19:14:40,282 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.8345, 2.8718, 2.2241, 3.2571, 2.8694, 2.8603, 1.2192, 2.7684], device='cuda:6'), covar=tensor([0.2349, 0.1687, 0.4127, 0.3555, 0.3445, 0.2243, 0.5675, 0.3286], device='cuda:6'), in_proj_covar=tensor([0.0248, 0.0220, 0.0255, 0.0308, 0.0300, 0.0248, 0.0277, 0.0276], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 19:14:48,093 INFO [finetune.py:976] (6/7) Epoch 23, batch 4200, loss[loss=0.1652, simple_loss=0.2356, pruned_loss=0.04741, over 4901.00 frames. ], tot_loss[loss=0.1707, simple_loss=0.2438, pruned_loss=0.04874, over 954487.35 frames. ], batch size: 46, lr: 3.09e-03, grad_scale: 16.0 2023-04-27 19:14:51,301 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-04-27 19:15:10,237 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7241, 1.2280, 1.7880, 2.1852, 1.7914, 1.7023, 1.7584, 1.7070], device='cuda:6'), covar=tensor([0.4246, 0.6107, 0.5560, 0.5365, 0.5633, 0.7713, 0.7381, 0.8349], device='cuda:6'), in_proj_covar=tensor([0.0433, 0.0417, 0.0510, 0.0506, 0.0462, 0.0495, 0.0499, 0.0510], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 19:15:18,779 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.2888, 1.3488, 1.3891, 1.5806, 1.6653, 1.3201, 0.9322, 1.4682], device='cuda:6'), covar=tensor([0.0836, 0.1152, 0.0870, 0.0599, 0.0635, 0.0831, 0.0798, 0.0627], device='cuda:6'), in_proj_covar=tensor([0.0187, 0.0203, 0.0184, 0.0174, 0.0177, 0.0180, 0.0150, 0.0179], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 19:15:51,651 INFO [finetune.py:976] (6/7) Epoch 23, batch 4250, loss[loss=0.1567, simple_loss=0.225, pruned_loss=0.0442, over 4821.00 frames. ], tot_loss[loss=0.1699, simple_loss=0.2424, pruned_loss=0.04874, over 954070.04 frames. ], batch size: 30, lr: 3.08e-03, grad_scale: 16.0 2023-04-27 19:16:01,962 INFO [optim.py:369] (6/7) 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 19:16:03,336 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=130268.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:16:55,315 INFO [finetune.py:976] (6/7) Epoch 23, batch 4300, loss[loss=0.1344, simple_loss=0.2136, pruned_loss=0.02758, over 4763.00 frames. ], tot_loss[loss=0.1676, simple_loss=0.2396, pruned_loss=0.04782, over 955844.22 frames. ], batch size: 28, lr: 3.08e-03, grad_scale: 16.0 2023-04-27 19:17:05,156 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=130316.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:17:58,687 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=130356.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:18:05,094 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=130358.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 19:18:05,662 INFO [finetune.py:976] (6/7) Epoch 23, batch 4350, loss[loss=0.167, simple_loss=0.2352, pruned_loss=0.04934, over 4693.00 frames. ], tot_loss[loss=0.1665, simple_loss=0.238, pruned_loss=0.04748, over 954967.52 frames. ], batch size: 23, lr: 3.08e-03, grad_scale: 16.0 2023-04-27 19:18:10,118 INFO [optim.py:369] (6/7) 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 19:18:12,527 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-04-27 19:18:29,088 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=130384.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:19:01,539 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.4269, 1.9094, 2.3253, 2.9881, 2.3714, 1.8292, 1.7953, 2.4192], device='cuda:6'), covar=tensor([0.3242, 0.3120, 0.1705, 0.2338, 0.2496, 0.2791, 0.3832, 0.1915], device='cuda:6'), in_proj_covar=tensor([0.0292, 0.0245, 0.0228, 0.0313, 0.0219, 0.0234, 0.0226, 0.0184], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:6') 2023-04-27 19:19:02,659 INFO [finetune.py:976] (6/7) Epoch 23, batch 4400, loss[loss=0.1825, simple_loss=0.2476, pruned_loss=0.05866, over 4926.00 frames. ], tot_loss[loss=0.168, simple_loss=0.2395, pruned_loss=0.04822, over 956194.09 frames. ], batch size: 38, lr: 3.08e-03, grad_scale: 16.0 2023-04-27 19:19:34,818 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=130432.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:19:45,153 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=130438.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:19:55,325 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=130445.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:19:56,558 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=130447.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:20:05,992 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2023-04-27 19:20:16,251 INFO [finetune.py:976] (6/7) Epoch 23, batch 4450, loss[loss=0.1928, simple_loss=0.2691, pruned_loss=0.05828, over 4912.00 frames. ], tot_loss[loss=0.1702, simple_loss=0.2425, pruned_loss=0.04892, over 953113.18 frames. ], batch size: 37, lr: 3.08e-03, grad_scale: 16.0 2023-04-27 19:20:16,421 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.0141, 2.0164, 1.7109, 1.7068, 2.0964, 1.6639, 2.5560, 1.5198], device='cuda:6'), covar=tensor([0.3929, 0.2133, 0.5286, 0.3014, 0.1902, 0.2654, 0.1465, 0.4844], device='cuda:6'), in_proj_covar=tensor([0.0338, 0.0349, 0.0425, 0.0350, 0.0377, 0.0373, 0.0368, 0.0417], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 19:20:25,829 INFO [optim.py:369] (6/7) 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 19:21:01,268 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=130493.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:21:02,960 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=130495.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:21:11,832 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=130499.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:21:23,655 INFO [finetune.py:976] (6/7) Epoch 23, batch 4500, loss[loss=0.1748, simple_loss=0.2411, pruned_loss=0.05421, over 4195.00 frames. ], tot_loss[loss=0.1702, simple_loss=0.2435, pruned_loss=0.04849, over 954028.47 frames. ], batch size: 65, lr: 3.08e-03, grad_scale: 16.0 2023-04-27 19:22:17,955 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=130548.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:22:31,481 INFO [finetune.py:976] (6/7) Epoch 23, batch 4550, loss[loss=0.147, simple_loss=0.2178, pruned_loss=0.03816, over 4803.00 frames. ], tot_loss[loss=0.1724, simple_loss=0.2458, pruned_loss=0.04954, over 955234.75 frames. ], batch size: 25, lr: 3.08e-03, grad_scale: 16.0 2023-04-27 19:22:42,288 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=130565.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:22:42,755 INFO [optim.py:369] (6/7) 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 19:23:36,955 INFO [finetune.py:976] (6/7) Epoch 23, batch 4600, loss[loss=0.1426, simple_loss=0.2198, pruned_loss=0.03265, over 4782.00 frames. ], tot_loss[loss=0.1711, simple_loss=0.2448, pruned_loss=0.0487, over 957094.12 frames. ], batch size: 29, lr: 3.08e-03, grad_scale: 16.0 2023-04-27 19:23:37,110 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=130609.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 19:23:56,113 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=130626.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:24:36,999 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=130656.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:24:38,735 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=130658.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 19:24:39,265 INFO [finetune.py:976] (6/7) Epoch 23, batch 4650, loss[loss=0.207, simple_loss=0.2736, pruned_loss=0.07023, over 4800.00 frames. ], tot_loss[loss=0.169, simple_loss=0.242, pruned_loss=0.04799, over 955444.06 frames. ], batch size: 51, lr: 3.08e-03, grad_scale: 16.0 2023-04-27 19:24:48,331 INFO [optim.py:369] (6/7) 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 19:25:39,304 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=130704.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:25:40,551 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 23, batch 4700, loss[loss=0.1703, simple_loss=0.2346, pruned_loss=0.05299, over 4833.00 frames. ], tot_loss[loss=0.1675, simple_loss=0.2396, pruned_loss=0.04772, over 956944.04 frames. ], batch size: 30, lr: 3.08e-03, grad_scale: 16.0 2023-04-27 19:25:49,719 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([0.7175, 1.7908, 1.8093, 1.4488, 1.8516, 1.5269, 2.3424, 1.6095], device='cuda:6'), covar=tensor([0.3217, 0.1766, 0.3941, 0.2305, 0.1425, 0.2092, 0.1362, 0.3892], device='cuda:6'), in_proj_covar=tensor([0.0337, 0.0348, 0.0425, 0.0350, 0.0377, 0.0374, 0.0367, 0.0418], device='cuda:6'), out_proj_covar=tensor([9.9931e-05, 1.0394e-04, 1.2876e-04, 1.0514e-04, 1.1208e-04, 1.1143e-04, 1.0785e-04, 1.2585e-04], device='cuda:6') 2023-04-27 19:26:22,717 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=130740.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:26:50,253 INFO [finetune.py:976] (6/7) Epoch 23, batch 4750, loss[loss=0.2159, simple_loss=0.28, pruned_loss=0.07593, over 4081.00 frames. ], tot_loss[loss=0.1657, simple_loss=0.2374, pruned_loss=0.04702, over 955853.83 frames. ], batch size: 65, lr: 3.08e-03, grad_scale: 16.0 2023-04-27 19:27:00,091 INFO [optim.py:369] (6/7) 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 19:27:24,060 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=130788.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:27:33,676 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 23, batch 4800, loss[loss=0.1764, simple_loss=0.257, pruned_loss=0.04784, over 4785.00 frames. ], tot_loss[loss=0.1687, simple_loss=0.2406, pruned_loss=0.04842, over 956001.96 frames. ], batch size: 29, lr: 3.08e-03, grad_scale: 16.0 2023-04-27 19:28:05,974 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9972, 1.8234, 2.1107, 2.5451, 2.0602, 1.9612, 2.0446, 1.9849], device='cuda:6'), covar=tensor([0.4900, 0.7170, 0.7617, 0.5429, 0.6251, 0.8596, 0.8749, 1.0439], device='cuda:6'), in_proj_covar=tensor([0.0436, 0.0418, 0.0512, 0.0506, 0.0464, 0.0497, 0.0501, 0.0512], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 19:28:13,768 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7784, 1.4731, 1.7719, 2.2919, 1.9053, 1.7234, 1.7807, 1.7878], device='cuda:6'), covar=tensor([0.3675, 0.5282, 0.5178, 0.4226, 0.4980, 0.5923, 0.5987, 0.6297], device='cuda:6'), in_proj_covar=tensor([0.0436, 0.0418, 0.0512, 0.0507, 0.0464, 0.0497, 0.0501, 0.0513], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 19:28:24,062 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.1138, 0.7306, 0.9060, 0.7667, 1.2403, 0.9754, 0.8295, 0.9624], device='cuda:6'), covar=tensor([0.1769, 0.1478, 0.2269, 0.1677, 0.1058, 0.1571, 0.1665, 0.2809], device='cuda:6'), in_proj_covar=tensor([0.0312, 0.0309, 0.0351, 0.0285, 0.0326, 0.0307, 0.0300, 0.0372], device='cuda:6'), out_proj_covar=tensor([6.3918e-05, 6.3841e-05, 7.3878e-05, 5.7310e-05, 6.7086e-05, 6.4265e-05, 6.2492e-05, 7.8988e-05], device='cuda:6') 2023-04-27 19:28:54,603 INFO [finetune.py:976] (6/7) Epoch 23, batch 4850, loss[loss=0.1125, simple_loss=0.1889, pruned_loss=0.01804, over 4748.00 frames. ], tot_loss[loss=0.1716, simple_loss=0.2441, pruned_loss=0.04958, over 955309.90 frames. ], batch size: 26, lr: 3.08e-03, grad_scale: 16.0 2023-04-27 19:29:05,298 INFO [optim.py:369] (6/7) 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 19:29:52,781 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-04-27 19:30:00,425 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=130904.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 19:30:03,972 INFO [finetune.py:976] (6/7) Epoch 23, batch 4900, loss[loss=0.1363, simple_loss=0.1992, pruned_loss=0.03668, over 4246.00 frames. ], tot_loss[loss=0.1722, simple_loss=0.2449, pruned_loss=0.04969, over 953770.28 frames. ], batch size: 18, lr: 3.08e-03, grad_scale: 16.0 2023-04-27 19:30:23,696 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=130921.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:31:11,899 INFO [finetune.py:976] (6/7) Epoch 23, batch 4950, loss[loss=0.1698, simple_loss=0.2436, pruned_loss=0.04801, over 4874.00 frames. ], tot_loss[loss=0.1717, simple_loss=0.2449, pruned_loss=0.04928, over 953130.01 frames. ], batch size: 34, lr: 3.08e-03, grad_scale: 16.0 2023-04-27 19:31:21,966 INFO [optim.py:369] (6/7) 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 19:31:30,410 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.45 vs. limit=5.0 2023-04-27 19:31:33,178 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([3.4006, 3.2815, 2.5431, 3.9126, 3.3154, 3.3909, 1.3291, 3.3170], device='cuda:6'), covar=tensor([0.1777, 0.1642, 0.3271, 0.2399, 0.3615, 0.2019, 0.6064, 0.2870], device='cuda:6'), in_proj_covar=tensor([0.0247, 0.0221, 0.0255, 0.0307, 0.0299, 0.0249, 0.0276, 0.0275], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 19:32:17,695 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6378, 2.6029, 2.0394, 2.3028, 2.5848, 2.0997, 3.2204, 1.7853], device='cuda:6'), covar=tensor([0.3382, 0.2134, 0.4855, 0.3219, 0.1805, 0.2695, 0.1918, 0.4416], device='cuda:6'), in_proj_covar=tensor([0.0336, 0.0347, 0.0424, 0.0349, 0.0375, 0.0373, 0.0367, 0.0417], device='cuda:6'), out_proj_covar=tensor([9.9527e-05, 1.0349e-04, 1.2848e-04, 1.0488e-04, 1.1138e-04, 1.1112e-04, 1.0793e-04, 1.2566e-04], device='cuda:6') 2023-04-27 19:32:26,076 INFO [finetune.py:976] (6/7) Epoch 23, batch 5000, loss[loss=0.1667, simple_loss=0.2353, pruned_loss=0.04908, over 4903.00 frames. ], tot_loss[loss=0.1706, simple_loss=0.2434, pruned_loss=0.04888, over 952612.70 frames. ], batch size: 46, lr: 3.08e-03, grad_scale: 16.0 2023-04-27 19:33:10,272 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=131040.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:33:35,486 INFO [finetune.py:976] (6/7) Epoch 23, batch 5050, loss[loss=0.14, simple_loss=0.2063, pruned_loss=0.03688, over 4872.00 frames. ], tot_loss[loss=0.1691, simple_loss=0.241, pruned_loss=0.04862, over 953723.66 frames. ], batch size: 31, lr: 3.08e-03, grad_scale: 16.0 2023-04-27 19:33:41,158 INFO [optim.py:369] (6/7) 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 19:34:14,405 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=131088.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:34:14,454 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=131088.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:34:18,298 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=131094.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:34:19,030 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-04-27 19:34:38,867 INFO [finetune.py:976] (6/7) Epoch 23, batch 5100, loss[loss=0.1256, simple_loss=0.2068, pruned_loss=0.0222, over 4756.00 frames. ], tot_loss[loss=0.167, simple_loss=0.2381, pruned_loss=0.04792, over 954321.07 frames. ], batch size: 26, lr: 3.08e-03, grad_scale: 32.0 2023-04-27 19:35:17,045 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=131136.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:35:21,313 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=131142.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:35:40,105 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.61 vs. limit=5.0 2023-04-27 19:35:41,499 INFO [finetune.py:976] (6/7) Epoch 23, batch 5150, loss[loss=0.1531, simple_loss=0.2356, pruned_loss=0.03528, over 4791.00 frames. ], tot_loss[loss=0.1662, simple_loss=0.2375, pruned_loss=0.04745, over 954727.00 frames. ], batch size: 29, lr: 3.08e-03, grad_scale: 32.0 2023-04-27 19:35:51,237 INFO [optim.py:369] (6/7) 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 19:36:02,270 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=131174.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:36:34,243 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.7587, 2.0363, 2.0593, 2.2224, 1.9891, 2.0187, 2.1472, 2.0681], device='cuda:6'), covar=tensor([0.4005, 0.6330, 0.4997, 0.4675, 0.5843, 0.7438, 0.6280, 0.5510], device='cuda:6'), in_proj_covar=tensor([0.0340, 0.0375, 0.0327, 0.0339, 0.0350, 0.0393, 0.0357, 0.0329], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 19:36:42,387 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=131204.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 19:36:45,363 INFO [finetune.py:976] (6/7) Epoch 23, batch 5200, loss[loss=0.1757, simple_loss=0.2486, pruned_loss=0.05136, over 4860.00 frames. ], tot_loss[loss=0.1694, simple_loss=0.2419, pruned_loss=0.04842, over 955416.15 frames. ], batch size: 31, lr: 3.08e-03, grad_scale: 32.0 2023-04-27 19:37:04,430 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=131221.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:37:05,651 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7007, 3.9220, 1.1732, 1.9598, 2.1383, 2.7922, 2.1801, 0.9471], device='cuda:6'), covar=tensor([0.1357, 0.0988, 0.1945, 0.1300, 0.1027, 0.0974, 0.1497, 0.2162], device='cuda:6'), in_proj_covar=tensor([0.0117, 0.0239, 0.0137, 0.0121, 0.0132, 0.0152, 0.0117, 0.0119], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-27 19:37:22,633 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=131235.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 19:37:43,653 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=131252.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:37:52,700 INFO [finetune.py:976] (6/7) Epoch 23, batch 5250, loss[loss=0.1258, simple_loss=0.2116, pruned_loss=0.02005, over 4764.00 frames. ], tot_loss[loss=0.1705, simple_loss=0.2434, pruned_loss=0.04874, over 954408.78 frames. ], batch size: 28, lr: 3.08e-03, grad_scale: 32.0 2023-04-27 19:37:57,551 INFO [optim.py:369] (6/7) 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 19:38:04,768 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=131269.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:38:44,517 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.1594, 2.5376, 1.0405, 1.4822, 2.0781, 1.3671, 3.4099, 1.8545], device='cuda:6'), covar=tensor([0.0634, 0.0688, 0.0830, 0.1231, 0.0521, 0.0981, 0.0213, 0.0637], device='cuda:6'), in_proj_covar=tensor([0.0050, 0.0064, 0.0047, 0.0046, 0.0050, 0.0051, 0.0073, 0.0050], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:6') 2023-04-27 19:38:50,181 INFO [finetune.py:976] (6/7) Epoch 23, batch 5300, loss[loss=0.2439, simple_loss=0.3147, pruned_loss=0.0865, over 4197.00 frames. ], tot_loss[loss=0.172, simple_loss=0.2447, pruned_loss=0.04962, over 954263.64 frames. ], batch size: 65, lr: 3.08e-03, grad_scale: 32.0 2023-04-27 19:39:32,270 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.6299, 2.8566, 1.4728, 1.9407, 2.4465, 1.7511, 3.8272, 2.2547], device='cuda:6'), covar=tensor([0.0527, 0.0649, 0.0645, 0.1032, 0.0421, 0.0803, 0.0341, 0.0504], device='cuda:6'), in_proj_covar=tensor([0.0050, 0.0064, 0.0047, 0.0046, 0.0050, 0.0051, 0.0073, 0.0050], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:6') 2023-04-27 19:39:56,069 INFO [finetune.py:976] (6/7) Epoch 23, batch 5350, loss[loss=0.1767, simple_loss=0.2437, pruned_loss=0.05485, over 4803.00 frames. ], tot_loss[loss=0.1726, simple_loss=0.2454, pruned_loss=0.04992, over 954124.55 frames. ], batch size: 40, lr: 3.08e-03, grad_scale: 32.0 2023-04-27 19:40:04,894 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-04-27 19:40:07,008 INFO [optim.py:369] (6/7) 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 19:41:03,593 INFO [finetune.py:976] (6/7) Epoch 23, batch 5400, loss[loss=0.13, simple_loss=0.2011, pruned_loss=0.02942, over 4824.00 frames. ], tot_loss[loss=0.1707, simple_loss=0.243, pruned_loss=0.04914, over 953829.79 frames. ], batch size: 30, lr: 3.08e-03, grad_scale: 32.0 2023-04-27 19:41:03,706 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=131409.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:41:10,913 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.5004, 2.8898, 2.4077, 2.7190, 2.0664, 2.4877, 2.4462, 2.0571], device='cuda:6'), covar=tensor([0.1494, 0.1047, 0.0785, 0.1003, 0.3124, 0.1039, 0.1686, 0.2481], device='cuda:6'), in_proj_covar=tensor([0.0283, 0.0298, 0.0213, 0.0274, 0.0313, 0.0255, 0.0246, 0.0261], device='cuda:6'), out_proj_covar=tensor([1.1301e-04, 1.1785e-04, 8.3984e-05, 1.0782e-04, 1.2626e-04, 1.0059e-04, 9.9253e-05, 1.0282e-04], device='cuda:6') 2023-04-27 19:41:12,781 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=131416.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:42:12,939 INFO [finetune.py:976] (6/7) Epoch 23, batch 5450, loss[loss=0.1342, simple_loss=0.2104, pruned_loss=0.02896, over 4774.00 frames. ], tot_loss[loss=0.1687, simple_loss=0.2403, pruned_loss=0.04859, over 953101.64 frames. ], batch size: 26, lr: 3.08e-03, grad_scale: 32.0 2023-04-27 19:42:22,592 INFO [optim.py:369] (6/7) 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 19:42:31,966 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=131470.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:42:43,251 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=131477.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:43:05,097 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=131492.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:43:22,052 INFO [finetune.py:976] (6/7) Epoch 23, batch 5500, loss[loss=0.112, simple_loss=0.1848, pruned_loss=0.01955, over 4760.00 frames. ], tot_loss[loss=0.1662, simple_loss=0.2374, pruned_loss=0.04752, over 951876.22 frames. ], batch size: 26, lr: 3.08e-03, grad_scale: 32.0 2023-04-27 19:43:52,947 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=131530.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 19:44:17,485 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 23, batch 5550, loss[loss=0.1528, simple_loss=0.2344, pruned_loss=0.03563, over 4767.00 frames. ], tot_loss[loss=0.1684, simple_loss=0.2399, pruned_loss=0.04848, over 951968.27 frames. ], batch size: 54, lr: 3.08e-03, grad_scale: 32.0 2023-04-27 19:44:32,388 INFO [optim.py:369] (6/7) 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 19:45:01,469 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.2998, 1.5093, 1.3329, 1.4516, 1.2978, 1.2404, 1.2042, 1.0057], device='cuda:6'), covar=tensor([0.1581, 0.1252, 0.0866, 0.1130, 0.3594, 0.1224, 0.1720, 0.2256], device='cuda:6'), in_proj_covar=tensor([0.0283, 0.0299, 0.0213, 0.0273, 0.0312, 0.0255, 0.0247, 0.0261], device='cuda:6'), out_proj_covar=tensor([1.1309e-04, 1.1811e-04, 8.3929e-05, 1.0780e-04, 1.2607e-04, 1.0069e-04, 9.9539e-05, 1.0292e-04], device='cuda:6') 2023-04-27 19:45:32,467 INFO [finetune.py:976] (6/7) Epoch 23, batch 5600, loss[loss=0.1639, simple_loss=0.2349, pruned_loss=0.04641, over 4921.00 frames. ], tot_loss[loss=0.1709, simple_loss=0.243, pruned_loss=0.04939, over 952986.96 frames. ], batch size: 38, lr: 3.07e-03, grad_scale: 32.0 2023-04-27 19:46:28,599 INFO [finetune.py:976] (6/7) Epoch 23, batch 5650, loss[loss=0.17, simple_loss=0.2501, pruned_loss=0.04498, over 4857.00 frames. ], tot_loss[loss=0.173, simple_loss=0.2455, pruned_loss=0.0502, over 951585.27 frames. ], batch size: 44, lr: 3.07e-03, grad_scale: 32.0 2023-04-27 19:46:38,007 INFO [optim.py:369] (6/7) 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 19:47:11,380 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.3445, 2.2293, 2.1218, 1.9961, 2.3172, 1.9301, 2.8482, 1.8877], device='cuda:6'), covar=tensor([0.3181, 0.1486, 0.3848, 0.2526, 0.1528, 0.2454, 0.1451, 0.3676], device='cuda:6'), in_proj_covar=tensor([0.0342, 0.0355, 0.0432, 0.0356, 0.0382, 0.0380, 0.0374, 0.0427], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 19:47:29,409 INFO [finetune.py:976] (6/7) Epoch 23, batch 5700, loss[loss=0.198, simple_loss=0.2628, pruned_loss=0.06654, over 4284.00 frames. ], tot_loss[loss=0.1712, simple_loss=0.2434, pruned_loss=0.04951, over 938165.64 frames. ], batch size: 18, lr: 3.07e-03, grad_scale: 32.0 2023-04-27 19:48:21,855 INFO [finetune.py:976] (6/7) Epoch 24, batch 0, loss[loss=0.1473, simple_loss=0.2186, pruned_loss=0.03802, over 4198.00 frames. ], tot_loss[loss=0.1473, simple_loss=0.2186, pruned_loss=0.03802, over 4198.00 frames. ], batch size: 66, lr: 3.07e-03, grad_scale: 32.0 2023-04-27 19:48:21,856 INFO [finetune.py:1001] (6/7) Computing validation loss 2023-04-27 19:48:25,200 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.3581, 1.4946, 1.8882, 2.0161, 1.9404, 2.0422, 1.8909, 1.9667], device='cuda:6'), covar=tensor([0.3954, 0.5422, 0.4672, 0.4748, 0.5644, 0.7544, 0.5179, 0.4864], device='cuda:6'), in_proj_covar=tensor([0.0340, 0.0375, 0.0328, 0.0339, 0.0349, 0.0394, 0.0357, 0.0329], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 19:48:31,485 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.2990, 1.5202, 1.8279, 1.9470, 1.9153, 1.9751, 1.8460, 1.8904], device='cuda:6'), covar=tensor([0.3704, 0.5380, 0.4670, 0.4843, 0.5385, 0.7133, 0.5465, 0.4685], device='cuda:6'), in_proj_covar=tensor([0.0340, 0.0375, 0.0328, 0.0339, 0.0349, 0.0394, 0.0357, 0.0329], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 19:48:37,425 INFO [finetune.py:1010] (6/7) Epoch 24, validation: loss=0.1552, simple_loss=0.2243, pruned_loss=0.04308, over 2265189.00 frames. 2023-04-27 19:48:37,426 INFO [finetune.py:1011] (6/7) Maximum memory allocated so far is 6435MB 2023-04-27 19:48:45,627 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.1391, 1.4824, 1.3197, 1.7479, 1.6620, 1.7764, 1.3734, 3.0612], device='cuda:6'), covar=tensor([0.0638, 0.0825, 0.0808, 0.1202, 0.0637, 0.0487, 0.0757, 0.0168], device='cuda:6'), in_proj_covar=tensor([0.0037, 0.0038, 0.0039, 0.0043, 0.0040, 0.0038, 0.0038, 0.0055], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0013, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:6') 2023-04-27 19:49:12,803 INFO [zipformer.py:1188] (6/7) 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] (6/7) 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 19:49:22,000 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=131772.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:49:37,726 INFO [finetune.py:976] (6/7) Epoch 24, batch 50, loss[loss=0.1523, simple_loss=0.2476, pruned_loss=0.02854, over 4923.00 frames. ], tot_loss[loss=0.1748, simple_loss=0.2482, pruned_loss=0.05071, over 217759.79 frames. ], batch size: 42, lr: 3.07e-03, grad_scale: 32.0 2023-04-27 19:50:39,507 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=131830.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:50:48,616 INFO [finetune.py:976] (6/7) Epoch 24, batch 100, loss[loss=0.1519, simple_loss=0.2204, pruned_loss=0.0417, over 4825.00 frames. ], tot_loss[loss=0.1689, simple_loss=0.2402, pruned_loss=0.04878, over 380703.04 frames. ], batch size: 33, lr: 3.07e-03, grad_scale: 32.0 2023-04-27 19:50:56,378 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=131848.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:50:58,849 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0722, 1.7867, 2.1004, 2.5207, 2.6290, 2.0949, 1.7977, 2.2729], device='cuda:6'), covar=tensor([0.0746, 0.1201, 0.0696, 0.0494, 0.0507, 0.0716, 0.0672, 0.0492], device='cuda:6'), in_proj_covar=tensor([0.0185, 0.0201, 0.0182, 0.0171, 0.0174, 0.0177, 0.0147, 0.0176], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 19:51:01,829 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=131857.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:51:07,126 INFO [optim.py:369] (6/7) 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 19:51:23,457 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=131878.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:51:35,580 INFO [finetune.py:976] (6/7) Epoch 24, batch 150, loss[loss=0.1656, simple_loss=0.2415, pruned_loss=0.04481, over 4762.00 frames. ], tot_loss[loss=0.1665, simple_loss=0.2368, pruned_loss=0.04807, over 508092.78 frames. ], batch size: 27, lr: 3.07e-03, grad_scale: 32.0 2023-04-27 19:52:18,780 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=131918.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:52:44,204 INFO [finetune.py:976] (6/7) Epoch 24, batch 200, loss[loss=0.1632, simple_loss=0.2312, pruned_loss=0.04756, over 4817.00 frames. ], tot_loss[loss=0.1649, simple_loss=0.2353, pruned_loss=0.04721, over 608224.51 frames. ], batch size: 25, lr: 3.07e-03, grad_scale: 32.0 2023-04-27 19:53:12,914 INFO [optim.py:369] (6/7) 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] (6/7) Epoch 24, batch 250, loss[loss=0.1682, simple_loss=0.2406, pruned_loss=0.04788, over 4762.00 frames. ], tot_loss[loss=0.1672, simple_loss=0.2392, pruned_loss=0.04761, over 687349.24 frames. ], batch size: 26, lr: 3.07e-03, grad_scale: 32.0 2023-04-27 19:53:57,447 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=132031.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:54:01,451 INFO [finetune.py:976] (6/7) Epoch 24, batch 300, loss[loss=0.232, simple_loss=0.2987, pruned_loss=0.08263, over 4914.00 frames. ], tot_loss[loss=0.1723, simple_loss=0.2447, pruned_loss=0.04989, over 745589.97 frames. ], batch size: 36, lr: 3.07e-03, grad_scale: 32.0 2023-04-27 19:54:24,658 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=132054.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:54:25,252 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6403, 1.7453, 0.8077, 1.3360, 1.8767, 1.5062, 1.4092, 1.4635], device='cuda:6'), covar=tensor([0.0478, 0.0366, 0.0337, 0.0548, 0.0271, 0.0516, 0.0481, 0.0583], device='cuda:6'), in_proj_covar=tensor([0.0028, 0.0023, 0.0022, 0.0028, 0.0019, 0.0028, 0.0027, 0.0028], device='cuda:6'), out_proj_covar=tensor([0.0051, 0.0045, 0.0038, 0.0052, 0.0038, 0.0050, 0.0050, 0.0051], device='cuda:6') 2023-04-27 19:54:36,462 INFO [zipformer.py:1188] (6/7) 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] (6/7) 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 19:54:46,375 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=132072.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:55:00,437 INFO [finetune.py:976] (6/7) Epoch 24, batch 350, loss[loss=0.1765, simple_loss=0.2564, pruned_loss=0.04826, over 4798.00 frames. ], tot_loss[loss=0.1726, simple_loss=0.2458, pruned_loss=0.04971, over 790754.51 frames. ], batch size: 45, lr: 3.07e-03, grad_scale: 32.0 2023-04-27 19:55:10,379 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=132092.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:55:34,251 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8311, 2.3145, 1.8092, 1.7453, 1.3202, 1.3539, 1.8959, 1.2860], device='cuda:6'), covar=tensor([0.1685, 0.1341, 0.1465, 0.1677, 0.2297, 0.2077, 0.1002, 0.2073], device='cuda:6'), in_proj_covar=tensor([0.0196, 0.0209, 0.0168, 0.0204, 0.0199, 0.0185, 0.0155, 0.0184], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:6') 2023-04-27 19:55:40,640 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=132113.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:55:41,957 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=132115.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 19:55:44,970 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=132120.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:56:06,851 INFO [finetune.py:976] (6/7) Epoch 24, batch 400, loss[loss=0.1783, simple_loss=0.2459, pruned_loss=0.05532, over 4873.00 frames. ], tot_loss[loss=0.1734, simple_loss=0.2469, pruned_loss=0.04998, over 826836.06 frames. ], batch size: 34, lr: 3.07e-03, grad_scale: 32.0 2023-04-27 19:56:25,709 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=132148.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:56:38,340 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.0897, 2.5463, 0.8432, 1.5593, 1.4855, 1.8319, 1.6323, 0.8661], device='cuda:6'), covar=tensor([0.1435, 0.1270, 0.1729, 0.1223, 0.1083, 0.0918, 0.1453, 0.1663], device='cuda:6'), in_proj_covar=tensor([0.0117, 0.0237, 0.0137, 0.0120, 0.0132, 0.0151, 0.0116, 0.0118], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-27 19:56:49,861 INFO [optim.py:369] (6/7) 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 19:57:08,717 INFO [finetune.py:976] (6/7) Epoch 24, batch 450, loss[loss=0.1602, simple_loss=0.2439, pruned_loss=0.03825, over 4777.00 frames. ], tot_loss[loss=0.1726, simple_loss=0.246, pruned_loss=0.04955, over 855565.20 frames. ], batch size: 26, lr: 3.07e-03, grad_scale: 32.0 2023-04-27 19:57:15,196 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=132196.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:57:18,045 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.0163, 1.0281, 1.1536, 1.1541, 1.0042, 0.9515, 0.9711, 0.5313], device='cuda:6'), covar=tensor([0.0500, 0.0527, 0.0415, 0.0532, 0.0645, 0.1110, 0.0406, 0.0592], device='cuda:6'), in_proj_covar=tensor([0.0069, 0.0067, 0.0066, 0.0067, 0.0074, 0.0094, 0.0072, 0.0063], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-27 19:57:27,639 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=132213.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:57:42,206 INFO [finetune.py:976] (6/7) Epoch 24, batch 500, loss[loss=0.1859, simple_loss=0.2554, pruned_loss=0.05822, over 4895.00 frames. ], tot_loss[loss=0.171, simple_loss=0.2437, pruned_loss=0.04914, over 879868.84 frames. ], batch size: 43, lr: 3.07e-03, grad_scale: 32.0 2023-04-27 19:57:55,782 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-04-27 19:58:03,235 INFO [optim.py:369] (6/7) 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] (6/7) Epoch 24, batch 550, loss[loss=0.1375, simple_loss=0.2082, pruned_loss=0.03336, over 4866.00 frames. ], tot_loss[loss=0.1687, simple_loss=0.2406, pruned_loss=0.04837, over 897687.17 frames. ], batch size: 31, lr: 3.07e-03, grad_scale: 32.0 2023-04-27 19:58:48,465 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-04-27 19:58:49,841 INFO [finetune.py:976] (6/7) Epoch 24, batch 600, loss[loss=0.2041, simple_loss=0.2796, pruned_loss=0.06428, over 4837.00 frames. ], tot_loss[loss=0.1699, simple_loss=0.2421, pruned_loss=0.04887, over 910527.20 frames. ], batch size: 49, lr: 3.07e-03, grad_scale: 32.0 2023-04-27 19:58:51,161 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0314, 2.4567, 1.1759, 1.3872, 2.1331, 1.1444, 3.2295, 1.5663], device='cuda:6'), covar=tensor([0.0675, 0.0773, 0.0786, 0.1155, 0.0437, 0.1002, 0.0219, 0.0639], device='cuda:6'), in_proj_covar=tensor([0.0050, 0.0064, 0.0047, 0.0046, 0.0050, 0.0051, 0.0073, 0.0051], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:6') 2023-04-27 19:59:10,242 INFO [optim.py:369] (6/7) 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] (6/7) Epoch 24, batch 650, loss[loss=0.1722, simple_loss=0.2578, pruned_loss=0.0433, over 4809.00 frames. ], tot_loss[loss=0.1733, simple_loss=0.246, pruned_loss=0.05027, over 921664.57 frames. ], batch size: 38, lr: 3.07e-03, grad_scale: 32.0 2023-04-27 19:59:23,064 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=132387.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:59:39,630 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=132410.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 20:00:08,100 INFO [finetune.py:976] (6/7) Epoch 24, batch 700, loss[loss=0.1798, simple_loss=0.2505, pruned_loss=0.05458, over 4908.00 frames. ], tot_loss[loss=0.1734, simple_loss=0.2466, pruned_loss=0.0501, over 929725.66 frames. ], batch size: 32, lr: 3.07e-03, grad_scale: 32.0 2023-04-27 20:00:50,356 INFO [optim.py:369] (6/7) 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 20:01:15,826 INFO [finetune.py:976] (6/7) Epoch 24, batch 750, loss[loss=0.2211, simple_loss=0.2821, pruned_loss=0.08008, over 4817.00 frames. ], tot_loss[loss=0.1745, simple_loss=0.2482, pruned_loss=0.0504, over 935435.44 frames. ], batch size: 33, lr: 3.06e-03, grad_scale: 32.0 2023-04-27 20:01:48,581 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=132513.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:02:21,500 INFO [finetune.py:976] (6/7) Epoch 24, batch 800, loss[loss=0.1909, simple_loss=0.2614, pruned_loss=0.06023, over 4905.00 frames. ], tot_loss[loss=0.1734, simple_loss=0.2468, pruned_loss=0.04995, over 939763.92 frames. ], batch size: 37, lr: 3.06e-03, grad_scale: 32.0 2023-04-27 20:02:53,600 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=132561.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:03:01,811 INFO [optim.py:369] (6/7) 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 20:03:27,813 INFO [finetune.py:976] (6/7) Epoch 24, batch 850, loss[loss=0.142, simple_loss=0.2197, pruned_loss=0.03218, over 4758.00 frames. ], tot_loss[loss=0.1712, simple_loss=0.2444, pruned_loss=0.04898, over 945942.93 frames. ], batch size: 26, lr: 3.06e-03, grad_scale: 16.0 2023-04-27 20:04:34,501 INFO [finetune.py:976] (6/7) Epoch 24, batch 900, loss[loss=0.1474, simple_loss=0.2207, pruned_loss=0.03705, over 4806.00 frames. ], tot_loss[loss=0.1683, simple_loss=0.2408, pruned_loss=0.04792, over 947670.41 frames. ], batch size: 29, lr: 3.06e-03, grad_scale: 16.0 2023-04-27 20:05:02,127 INFO [zipformer.py:1188] (6/7) 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] (6/7) 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 20:05:14,168 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.1889, 1.6168, 2.0399, 2.2483, 2.0293, 1.6452, 1.2006, 1.7188], device='cuda:6'), covar=tensor([0.2918, 0.2918, 0.1509, 0.2030, 0.2301, 0.2477, 0.3988, 0.1752], device='cuda:6'), in_proj_covar=tensor([0.0296, 0.0247, 0.0229, 0.0315, 0.0222, 0.0236, 0.0228, 0.0187], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:6') 2023-04-27 20:05:25,965 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5520, 1.3260, 1.2785, 1.3035, 1.7415, 1.4301, 1.2094, 1.2263], device='cuda:6'), covar=tensor([0.1586, 0.1090, 0.1668, 0.1188, 0.0623, 0.1435, 0.1533, 0.1845], device='cuda:6'), in_proj_covar=tensor([0.0313, 0.0310, 0.0352, 0.0286, 0.0329, 0.0309, 0.0301, 0.0375], device='cuda:6'), out_proj_covar=tensor([6.4157e-05, 6.4095e-05, 7.4189e-05, 5.7363e-05, 6.7692e-05, 6.4689e-05, 6.2765e-05, 7.9577e-05], device='cuda:6') 2023-04-27 20:05:44,311 INFO [finetune.py:976] (6/7) Epoch 24, batch 950, loss[loss=0.1632, simple_loss=0.2461, pruned_loss=0.0402, over 4747.00 frames. ], tot_loss[loss=0.1669, simple_loss=0.2387, pruned_loss=0.04749, over 949418.74 frames. ], batch size: 54, lr: 3.06e-03, grad_scale: 16.0 2023-04-27 20:05:44,428 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=132687.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:06:08,256 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=132710.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 20:06:27,935 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=132721.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:06:43,788 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=132735.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:06:49,486 INFO [finetune.py:976] (6/7) Epoch 24, batch 1000, loss[loss=0.1908, simple_loss=0.2665, pruned_loss=0.05754, over 4732.00 frames. ], tot_loss[loss=0.1682, simple_loss=0.24, pruned_loss=0.04816, over 949707.25 frames. ], batch size: 54, lr: 3.06e-03, grad_scale: 16.0 2023-04-27 20:06:50,879 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6225, 0.6768, 1.5374, 1.9907, 1.7266, 1.5703, 1.5958, 1.5625], device='cuda:6'), covar=tensor([0.3816, 0.5875, 0.4984, 0.5113, 0.4952, 0.6165, 0.6338, 0.7122], device='cuda:6'), in_proj_covar=tensor([0.0434, 0.0418, 0.0512, 0.0506, 0.0464, 0.0496, 0.0501, 0.0512], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 20:06:55,169 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=132746.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:06:56,392 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8977, 1.8511, 1.9969, 2.3717, 2.3802, 1.7394, 1.5764, 1.9318], device='cuda:6'), covar=tensor([0.0964, 0.1112, 0.0779, 0.0696, 0.0678, 0.0977, 0.0837, 0.0684], device='cuda:6'), in_proj_covar=tensor([0.0186, 0.0203, 0.0185, 0.0172, 0.0177, 0.0179, 0.0149, 0.0178], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 20:07:02,349 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=132758.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:07:07,258 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.1623, 2.5780, 0.8699, 1.4902, 1.5648, 1.9177, 1.6719, 0.8314], device='cuda:6'), covar=tensor([0.1417, 0.1085, 0.1821, 0.1302, 0.1086, 0.0902, 0.1490, 0.1719], device='cuda:6'), in_proj_covar=tensor([0.0117, 0.0239, 0.0137, 0.0121, 0.0133, 0.0152, 0.0117, 0.0119], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-27 20:07:07,767 INFO [optim.py:369] (6/7) 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 20:07:22,249 INFO [finetune.py:976] (6/7) Epoch 24, batch 1050, loss[loss=0.195, simple_loss=0.2649, pruned_loss=0.06252, over 4802.00 frames. ], tot_loss[loss=0.1696, simple_loss=0.2427, pruned_loss=0.04827, over 953007.76 frames. ], batch size: 51, lr: 3.06e-03, grad_scale: 16.0 2023-04-27 20:07:35,161 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=132807.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:07:52,665 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6081, 1.7639, 1.4814, 1.0912, 1.2232, 1.2079, 1.4661, 1.1945], device='cuda:6'), covar=tensor([0.1639, 0.1240, 0.1426, 0.1630, 0.2285, 0.1936, 0.1031, 0.1996], device='cuda:6'), in_proj_covar=tensor([0.0196, 0.0209, 0.0167, 0.0203, 0.0199, 0.0185, 0.0155, 0.0184], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:6') 2023-04-27 20:07:56,051 INFO [finetune.py:976] (6/7) Epoch 24, batch 1100, loss[loss=0.1464, simple_loss=0.2158, pruned_loss=0.03846, over 4492.00 frames. ], tot_loss[loss=0.1714, simple_loss=0.2448, pruned_loss=0.04899, over 952635.73 frames. ], batch size: 20, lr: 3.06e-03, grad_scale: 16.0 2023-04-27 20:08:12,101 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2023-04-27 20:08:14,647 INFO [optim.py:369] (6/7) 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 20:08:28,715 INFO [finetune.py:976] (6/7) Epoch 24, batch 1150, loss[loss=0.1578, simple_loss=0.2347, pruned_loss=0.04047, over 4900.00 frames. ], tot_loss[loss=0.1723, simple_loss=0.2455, pruned_loss=0.04956, over 952741.99 frames. ], batch size: 37, lr: 3.06e-03, grad_scale: 16.0 2023-04-27 20:09:11,141 INFO [finetune.py:976] (6/7) Epoch 24, batch 1200, loss[loss=0.1486, simple_loss=0.2186, pruned_loss=0.03933, over 4819.00 frames. ], tot_loss[loss=0.1715, simple_loss=0.2442, pruned_loss=0.04938, over 954158.39 frames. ], batch size: 33, lr: 3.06e-03, grad_scale: 16.0 2023-04-27 20:09:45,986 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.2313, 1.2456, 3.7723, 3.5466, 3.3523, 3.6750, 3.6238, 3.3063], device='cuda:6'), covar=tensor([0.7045, 0.5854, 0.1309, 0.1886, 0.1284, 0.1417, 0.1630, 0.1625], device='cuda:6'), in_proj_covar=tensor([0.0309, 0.0307, 0.0405, 0.0407, 0.0347, 0.0409, 0.0316, 0.0365], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-27 20:09:48,368 INFO [optim.py:369] (6/7) 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 20:09:56,248 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.6856, 2.2408, 2.5616, 3.2845, 2.4863, 2.0125, 2.0519, 2.5227], device='cuda:6'), covar=tensor([0.3305, 0.3018, 0.1565, 0.2054, 0.2619, 0.2626, 0.3394, 0.1915], device='cuda:6'), in_proj_covar=tensor([0.0292, 0.0243, 0.0225, 0.0311, 0.0220, 0.0232, 0.0225, 0.0184], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:6') 2023-04-27 20:10:17,853 INFO [finetune.py:976] (6/7) Epoch 24, batch 1250, loss[loss=0.1368, simple_loss=0.2003, pruned_loss=0.03671, over 4764.00 frames. ], tot_loss[loss=0.17, simple_loss=0.2423, pruned_loss=0.04884, over 954607.84 frames. ], batch size: 26, lr: 3.06e-03, grad_scale: 16.0 2023-04-27 20:10:19,400 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.50 vs. limit=5.0 2023-04-27 20:10:53,170 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=133016.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:11:03,102 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.3664, 1.8337, 2.2434, 2.7403, 2.2451, 1.7575, 1.5314, 2.0121], device='cuda:6'), covar=tensor([0.2588, 0.2837, 0.1415, 0.1958, 0.2544, 0.2422, 0.3932, 0.1947], device='cuda:6'), in_proj_covar=tensor([0.0292, 0.0243, 0.0226, 0.0311, 0.0220, 0.0233, 0.0225, 0.0184], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:6') 2023-04-27 20:11:17,185 INFO [finetune.py:976] (6/7) Epoch 24, batch 1300, loss[loss=0.183, simple_loss=0.249, pruned_loss=0.05847, over 4752.00 frames. ], tot_loss[loss=0.1673, simple_loss=0.2394, pruned_loss=0.0476, over 957398.48 frames. ], batch size: 59, lr: 3.06e-03, grad_scale: 16.0 2023-04-27 20:11:37,887 INFO [optim.py:369] (6/7) 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 20:11:50,601 INFO [finetune.py:976] (6/7) Epoch 24, batch 1350, loss[loss=0.1613, simple_loss=0.2411, pruned_loss=0.04071, over 4873.00 frames. ], tot_loss[loss=0.1676, simple_loss=0.2393, pruned_loss=0.04793, over 959115.61 frames. ], batch size: 34, lr: 3.06e-03, grad_scale: 16.0 2023-04-27 20:12:07,622 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=133102.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:12:28,948 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.77 vs. limit=5.0 2023-04-27 20:12:52,452 INFO [finetune.py:976] (6/7) Epoch 24, batch 1400, loss[loss=0.2376, simple_loss=0.3144, pruned_loss=0.0804, over 4127.00 frames. ], tot_loss[loss=0.1702, simple_loss=0.2425, pruned_loss=0.049, over 958708.57 frames. ], batch size: 65, lr: 3.06e-03, grad_scale: 16.0 2023-04-27 20:13:02,840 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5024, 1.5955, 0.6074, 1.2310, 1.5224, 1.3334, 1.2409, 1.3844], device='cuda:6'), covar=tensor([0.0507, 0.0348, 0.0390, 0.0553, 0.0296, 0.0521, 0.0501, 0.0549], device='cuda:6'), in_proj_covar=tensor([0.0028, 0.0023, 0.0022, 0.0028, 0.0019, 0.0027, 0.0027, 0.0028], device='cuda:6'), out_proj_covar=tensor([0.0051, 0.0045, 0.0038, 0.0052, 0.0038, 0.0050, 0.0050, 0.0051], device='cuda:6') 2023-04-27 20:13:29,243 INFO [optim.py:369] (6/7) 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 20:13:41,458 INFO [finetune.py:976] (6/7) Epoch 24, batch 1450, loss[loss=0.1728, simple_loss=0.2577, pruned_loss=0.04395, over 4908.00 frames. ], tot_loss[loss=0.1713, simple_loss=0.2441, pruned_loss=0.04929, over 956029.56 frames. ], batch size: 37, lr: 3.06e-03, grad_scale: 16.0 2023-04-27 20:13:55,454 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.6314, 3.2952, 2.8160, 3.2141, 2.2106, 2.8844, 2.9806, 2.2305], device='cuda:6'), covar=tensor([0.1913, 0.1126, 0.0714, 0.1004, 0.3020, 0.1007, 0.1700, 0.2366], device='cuda:6'), in_proj_covar=tensor([0.0290, 0.0305, 0.0218, 0.0278, 0.0318, 0.0259, 0.0251, 0.0267], device='cuda:6'), out_proj_covar=tensor([1.1571e-04, 1.2047e-04, 8.6004e-05, 1.0954e-04, 1.2842e-04, 1.0220e-04, 1.0141e-04, 1.0549e-04], device='cuda:6') 2023-04-27 20:14:31,503 INFO [finetune.py:976] (6/7) Epoch 24, batch 1500, loss[loss=0.1986, simple_loss=0.2726, pruned_loss=0.0623, over 4837.00 frames. ], tot_loss[loss=0.1721, simple_loss=0.2449, pruned_loss=0.04967, over 956761.82 frames. ], batch size: 44, lr: 3.06e-03, grad_scale: 16.0 2023-04-27 20:14:37,105 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7413, 1.3261, 4.5943, 4.2776, 3.9715, 4.3515, 4.1898, 4.0524], device='cuda:6'), covar=tensor([0.7139, 0.6316, 0.1041, 0.1753, 0.1226, 0.1783, 0.1189, 0.1724], device='cuda:6'), in_proj_covar=tensor([0.0310, 0.0307, 0.0405, 0.0408, 0.0349, 0.0409, 0.0317, 0.0367], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-27 20:14:37,340 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.12 vs. limit=5.0 2023-04-27 20:15:14,562 INFO [optim.py:369] (6/7) 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 20:15:32,666 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.3529, 1.9166, 2.1714, 2.8508, 2.3051, 1.8026, 1.9772, 2.2317], device='cuda:6'), covar=tensor([0.2717, 0.2855, 0.1496, 0.1992, 0.2398, 0.2387, 0.3386, 0.1954], device='cuda:6'), in_proj_covar=tensor([0.0294, 0.0244, 0.0227, 0.0313, 0.0220, 0.0234, 0.0226, 0.0185], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:6') 2023-04-27 20:15:44,449 INFO [finetune.py:976] (6/7) Epoch 24, batch 1550, loss[loss=0.1723, simple_loss=0.2459, pruned_loss=0.04938, over 4805.00 frames. ], tot_loss[loss=0.1716, simple_loss=0.2449, pruned_loss=0.0491, over 954519.54 frames. ], batch size: 40, lr: 3.06e-03, grad_scale: 16.0 2023-04-27 20:16:10,947 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=133316.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:16:34,613 INFO [finetune.py:976] (6/7) Epoch 24, batch 1600, loss[loss=0.1708, simple_loss=0.2409, pruned_loss=0.05028, over 4873.00 frames. ], tot_loss[loss=0.1704, simple_loss=0.2432, pruned_loss=0.0488, over 954144.57 frames. ], batch size: 34, lr: 3.06e-03, grad_scale: 16.0 2023-04-27 20:17:17,459 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=133364.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:17:19,226 INFO [optim.py:369] (6/7) 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 20:17:21,169 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5738, 1.2040, 1.3517, 1.2482, 1.6701, 1.3404, 1.1421, 1.2899], device='cuda:6'), covar=tensor([0.1708, 0.1198, 0.1809, 0.1333, 0.0949, 0.1569, 0.1713, 0.2187], device='cuda:6'), in_proj_covar=tensor([0.0313, 0.0308, 0.0351, 0.0285, 0.0328, 0.0307, 0.0300, 0.0374], device='cuda:6'), out_proj_covar=tensor([6.4072e-05, 6.3518e-05, 7.3807e-05, 5.7175e-05, 6.7553e-05, 6.4337e-05, 6.2528e-05, 7.9295e-05], device='cuda:6') 2023-04-27 20:17:29,328 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=133374.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 20:17:30,015 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=3.33 vs. limit=5.0 2023-04-27 20:17:30,577 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 24, batch 1650, loss[loss=0.1301, simple_loss=0.2131, pruned_loss=0.02352, over 4940.00 frames. ], tot_loss[loss=0.1682, simple_loss=0.2406, pruned_loss=0.04784, over 955243.39 frames. ], batch size: 33, lr: 3.06e-03, grad_scale: 16.0 2023-04-27 20:18:02,202 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=133402.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:18:48,096 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=133435.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 20:18:49,165 INFO [finetune.py:976] (6/7) Epoch 24, batch 1700, loss[loss=0.2314, simple_loss=0.2891, pruned_loss=0.0869, over 4845.00 frames. ], tot_loss[loss=0.1657, simple_loss=0.2377, pruned_loss=0.04686, over 954795.79 frames. ], batch size: 47, lr: 3.06e-03, grad_scale: 16.0 2023-04-27 20:18:49,284 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=133437.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:18:55,757 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.4209, 3.3613, 0.9432, 1.8365, 2.0228, 2.4892, 2.0077, 0.9435], device='cuda:6'), covar=tensor([0.1402, 0.0874, 0.1955, 0.1215, 0.1067, 0.0934, 0.1381, 0.2039], device='cuda:6'), in_proj_covar=tensor([0.0117, 0.0239, 0.0137, 0.0121, 0.0133, 0.0152, 0.0117, 0.0119], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-27 20:19:08,016 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=133450.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:19:31,794 INFO [optim.py:369] (6/7) 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 20:19:40,019 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([0.8970, 1.2488, 3.2793, 3.0552, 2.8970, 3.1827, 3.1540, 2.8694], device='cuda:6'), covar=tensor([0.7359, 0.5138, 0.1429, 0.2030, 0.1504, 0.1977, 0.1735, 0.1931], device='cuda:6'), in_proj_covar=tensor([0.0309, 0.0306, 0.0404, 0.0408, 0.0349, 0.0409, 0.0315, 0.0370], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-27 20:19:41,277 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 24, batch 1750, loss[loss=0.1614, simple_loss=0.2192, pruned_loss=0.05184, over 4013.00 frames. ], tot_loss[loss=0.1674, simple_loss=0.2395, pruned_loss=0.04766, over 955676.25 frames. ], batch size: 17, lr: 3.06e-03, grad_scale: 16.0 2023-04-27 20:20:39,258 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.2544, 2.8695, 1.0847, 1.6448, 2.4339, 1.2684, 3.8606, 1.7600], device='cuda:6'), covar=tensor([0.0679, 0.0722, 0.0839, 0.1278, 0.0473, 0.0995, 0.0277, 0.0616], device='cuda:6'), in_proj_covar=tensor([0.0050, 0.0063, 0.0047, 0.0046, 0.0049, 0.0050, 0.0072, 0.0050], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0009, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0007], device='cuda:6') 2023-04-27 20:21:01,625 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 24, batch 1800, loss[loss=0.2142, simple_loss=0.2882, pruned_loss=0.07012, over 4913.00 frames. ], tot_loss[loss=0.1694, simple_loss=0.2428, pruned_loss=0.04795, over 957788.34 frames. ], batch size: 36, lr: 3.06e-03, grad_scale: 16.0 2023-04-27 20:21:44,191 INFO [optim.py:369] (6/7) 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 20:21:54,930 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9367, 1.6199, 2.0919, 2.2589, 1.9706, 1.8705, 1.9562, 1.9783], device='cuda:6'), covar=tensor([0.5785, 0.8617, 0.7557, 0.7998, 0.7879, 0.9555, 0.9528, 1.0159], device='cuda:6'), in_proj_covar=tensor([0.0435, 0.0418, 0.0511, 0.0507, 0.0466, 0.0496, 0.0502, 0.0511], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 20:22:08,897 INFO [finetune.py:976] (6/7) Epoch 24, batch 1850, loss[loss=0.1864, simple_loss=0.2697, pruned_loss=0.05152, over 4919.00 frames. ], tot_loss[loss=0.1706, simple_loss=0.2442, pruned_loss=0.04844, over 957081.12 frames. ], batch size: 33, lr: 3.06e-03, grad_scale: 16.0 2023-04-27 20:22:29,809 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.49 vs. limit=2.0 2023-04-27 20:22:35,187 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6255, 1.4617, 4.3680, 4.1222, 3.7876, 4.1338, 3.9851, 3.8175], device='cuda:6'), covar=tensor([0.6620, 0.4930, 0.0812, 0.1255, 0.1035, 0.1239, 0.1574, 0.1446], device='cuda:6'), in_proj_covar=tensor([0.0307, 0.0304, 0.0402, 0.0405, 0.0346, 0.0406, 0.0314, 0.0367], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-27 20:22:38,033 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8004, 1.1048, 4.2502, 3.6762, 3.7233, 3.9275, 3.8696, 3.5482], device='cuda:6'), covar=tensor([0.8374, 0.8666, 0.1367, 0.2793, 0.1985, 0.2738, 0.2465, 0.3145], device='cuda:6'), in_proj_covar=tensor([0.0307, 0.0304, 0.0402, 0.0405, 0.0346, 0.0406, 0.0313, 0.0367], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-27 20:22:42,124 INFO [finetune.py:976] (6/7) Epoch 24, batch 1900, loss[loss=0.1971, simple_loss=0.2796, pruned_loss=0.05724, over 4822.00 frames. ], tot_loss[loss=0.172, simple_loss=0.246, pruned_loss=0.04902, over 954943.18 frames. ], batch size: 33, lr: 3.06e-03, grad_scale: 16.0 2023-04-27 20:22:51,325 INFO [zipformer.py:1188] (6/7) 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] (6/7) 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 20:23:18,587 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7340, 1.5336, 2.0226, 2.1526, 1.5254, 1.4211, 1.7220, 1.0766], device='cuda:6'), covar=tensor([0.0519, 0.0654, 0.0364, 0.0543, 0.0765, 0.1016, 0.0567, 0.0609], device='cuda:6'), in_proj_covar=tensor([0.0070, 0.0068, 0.0066, 0.0068, 0.0074, 0.0095, 0.0073, 0.0064], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-27 20:23:24,539 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9606, 1.8869, 2.3291, 2.5509, 1.7920, 1.6452, 1.9747, 1.0079], device='cuda:6'), covar=tensor([0.0580, 0.0601, 0.0389, 0.0603, 0.0710, 0.1060, 0.0548, 0.0717], device='cuda:6'), in_proj_covar=tensor([0.0070, 0.0068, 0.0066, 0.0068, 0.0074, 0.0095, 0.0073, 0.0064], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-27 20:23:26,869 INFO [finetune.py:976] (6/7) Epoch 24, batch 1950, loss[loss=0.135, simple_loss=0.2179, pruned_loss=0.02605, over 4826.00 frames. ], tot_loss[loss=0.1687, simple_loss=0.2429, pruned_loss=0.04724, over 954552.29 frames. ], batch size: 33, lr: 3.06e-03, grad_scale: 16.0 2023-04-27 20:23:43,326 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=133713.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:23:56,167 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=133730.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 20:23:57,386 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=133732.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:24:00,817 INFO [finetune.py:976] (6/7) Epoch 24, batch 2000, loss[loss=0.1645, simple_loss=0.2363, pruned_loss=0.04633, over 4344.00 frames. ], tot_loss[loss=0.1677, simple_loss=0.2409, pruned_loss=0.04723, over 955303.86 frames. ], batch size: 65, lr: 3.06e-03, grad_scale: 16.0 2023-04-27 20:24:25,087 INFO [optim.py:369] (6/7) 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 20:24:55,311 INFO [finetune.py:976] (6/7) Epoch 24, batch 2050, loss[loss=0.1663, simple_loss=0.2308, pruned_loss=0.05093, over 4868.00 frames. ], tot_loss[loss=0.1648, simple_loss=0.2374, pruned_loss=0.04606, over 955131.47 frames. ], batch size: 34, lr: 3.06e-03, grad_scale: 16.0 2023-04-27 20:25:48,594 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=133830.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:25:58,915 INFO [finetune.py:976] (6/7) Epoch 24, batch 2100, loss[loss=0.2009, simple_loss=0.261, pruned_loss=0.07042, over 4714.00 frames. ], tot_loss[loss=0.1653, simple_loss=0.2374, pruned_loss=0.04658, over 955135.88 frames. ], batch size: 59, lr: 3.06e-03, grad_scale: 16.0 2023-04-27 20:26:10,282 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.4530, 0.9774, 0.3187, 1.1743, 1.1744, 1.3365, 1.2515, 1.2149], device='cuda:6'), covar=tensor([0.0493, 0.0403, 0.0410, 0.0532, 0.0289, 0.0488, 0.0465, 0.0559], device='cuda:6'), in_proj_covar=tensor([0.0028, 0.0023, 0.0022, 0.0028, 0.0019, 0.0027, 0.0027, 0.0028], device='cuda:6'), out_proj_covar=tensor([0.0051, 0.0045, 0.0038, 0.0052, 0.0038, 0.0050, 0.0050, 0.0051], device='cuda:6') 2023-04-27 20:26:22,909 INFO [optim.py:369] (6/7) 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 20:26:33,012 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.3004, 2.0838, 2.3415, 2.8047, 2.7803, 2.0499, 1.9586, 2.2013], device='cuda:6'), covar=tensor([0.0821, 0.1083, 0.0725, 0.0547, 0.0613, 0.1000, 0.0765, 0.0656], device='cuda:6'), in_proj_covar=tensor([0.0189, 0.0208, 0.0189, 0.0175, 0.0181, 0.0183, 0.0153, 0.0182], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 20:26:36,415 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.3453, 3.4224, 0.8702, 1.8460, 1.7894, 2.4375, 1.9335, 0.9018], device='cuda:6'), covar=tensor([0.1595, 0.1144, 0.2313, 0.1484, 0.1220, 0.1217, 0.1779, 0.2293], device='cuda:6'), in_proj_covar=tensor([0.0118, 0.0241, 0.0137, 0.0122, 0.0134, 0.0153, 0.0117, 0.0120], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-27 20:26:37,510 INFO [finetune.py:976] (6/7) Epoch 24, batch 2150, loss[loss=0.1414, simple_loss=0.2189, pruned_loss=0.03197, over 4757.00 frames. ], tot_loss[loss=0.1696, simple_loss=0.2418, pruned_loss=0.04874, over 955673.24 frames. ], batch size: 27, lr: 3.05e-03, grad_scale: 16.0 2023-04-27 20:27:00,376 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5850, 1.2137, 4.4641, 4.1785, 3.9041, 4.2575, 4.0809, 3.8572], device='cuda:6'), covar=tensor([0.7001, 0.5916, 0.0921, 0.1480, 0.1087, 0.1652, 0.1569, 0.1644], device='cuda:6'), in_proj_covar=tensor([0.0307, 0.0304, 0.0402, 0.0405, 0.0348, 0.0406, 0.0315, 0.0366], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-27 20:27:00,430 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7743, 2.4756, 1.8863, 1.8323, 1.2811, 1.3043, 1.9933, 1.2531], device='cuda:6'), covar=tensor([0.1815, 0.1408, 0.1540, 0.1851, 0.2385, 0.2092, 0.0982, 0.2199], device='cuda:6'), in_proj_covar=tensor([0.0198, 0.0211, 0.0169, 0.0206, 0.0200, 0.0186, 0.0156, 0.0185], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:6') 2023-04-27 20:27:01,045 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9784, 2.7015, 1.9828, 2.2238, 1.5286, 1.5348, 2.1052, 1.4243], device='cuda:6'), covar=tensor([0.1400, 0.1388, 0.1273, 0.1489, 0.2034, 0.1709, 0.0874, 0.1860], device='cuda:6'), in_proj_covar=tensor([0.0198, 0.0211, 0.0169, 0.0206, 0.0200, 0.0186, 0.0156, 0.0185], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:6') 2023-04-27 20:27:02,873 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7796, 2.2734, 1.9173, 2.2035, 1.7438, 1.9753, 1.8983, 1.5611], device='cuda:6'), covar=tensor([0.1797, 0.1065, 0.0869, 0.1138, 0.3093, 0.1104, 0.1762, 0.2558], device='cuda:6'), in_proj_covar=tensor([0.0287, 0.0302, 0.0217, 0.0278, 0.0315, 0.0257, 0.0250, 0.0266], device='cuda:6'), out_proj_covar=tensor([1.1483e-04, 1.1955e-04, 8.5266e-05, 1.0946e-04, 1.2700e-04, 1.0134e-04, 1.0108e-04, 1.0502e-04], device='cuda:6') 2023-04-27 20:27:09,768 INFO [finetune.py:976] (6/7) Epoch 24, batch 2200, loss[loss=0.2194, simple_loss=0.2855, pruned_loss=0.07663, over 4902.00 frames. ], tot_loss[loss=0.1715, simple_loss=0.2442, pruned_loss=0.04945, over 956174.01 frames. ], batch size: 46, lr: 3.05e-03, grad_scale: 16.0 2023-04-27 20:27:48,467 INFO [optim.py:369] (6/7) 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] (6/7) Epoch 24, batch 2250, loss[loss=0.2255, simple_loss=0.2915, pruned_loss=0.07969, over 4685.00 frames. ], tot_loss[loss=0.1724, simple_loss=0.2455, pruned_loss=0.04963, over 956027.72 frames. ], batch size: 59, lr: 3.05e-03, grad_scale: 16.0 2023-04-27 20:28:14,989 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=133997.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:28:23,009 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=134008.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:28:31,659 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-04-27 20:28:36,988 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=134030.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 20:28:38,163 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=134032.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:28:41,585 INFO [finetune.py:976] (6/7) Epoch 24, batch 2300, loss[loss=0.1656, simple_loss=0.2352, pruned_loss=0.04795, over 4778.00 frames. ], tot_loss[loss=0.1718, simple_loss=0.2456, pruned_loss=0.04896, over 954994.66 frames. ], batch size: 26, lr: 3.05e-03, grad_scale: 16.0 2023-04-27 20:28:55,893 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=134058.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:29:01,258 INFO [optim.py:369] (6/7) 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 20:29:08,091 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=134078.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 20:29:09,791 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=134080.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:29:11,424 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.83 vs. limit=5.0 2023-04-27 20:29:14,453 INFO [finetune.py:976] (6/7) Epoch 24, batch 2350, loss[loss=0.1415, simple_loss=0.2169, pruned_loss=0.03308, over 4755.00 frames. ], tot_loss[loss=0.17, simple_loss=0.2432, pruned_loss=0.04845, over 954048.25 frames. ], batch size: 28, lr: 3.05e-03, grad_scale: 16.0 2023-04-27 20:29:42,024 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=134130.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:29:46,680 INFO [finetune.py:976] (6/7) Epoch 24, batch 2400, loss[loss=0.1651, simple_loss=0.2308, pruned_loss=0.04971, over 4822.00 frames. ], tot_loss[loss=0.1684, simple_loss=0.2405, pruned_loss=0.04812, over 953790.71 frames. ], batch size: 30, lr: 3.05e-03, grad_scale: 16.0 2023-04-27 20:30:06,808 INFO [optim.py:369] (6/7) 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] (6/7) 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] (6/7) Epoch 24, batch 2450, loss[loss=0.1267, simple_loss=0.2009, pruned_loss=0.02629, over 4814.00 frames. ], tot_loss[loss=0.1658, simple_loss=0.2374, pruned_loss=0.04711, over 956426.62 frames. ], batch size: 38, lr: 3.05e-03, grad_scale: 16.0 2023-04-27 20:31:07,207 INFO [finetune.py:976] (6/7) Epoch 24, batch 2500, loss[loss=0.2216, simple_loss=0.2947, pruned_loss=0.07423, over 4813.00 frames. ], tot_loss[loss=0.1679, simple_loss=0.2396, pruned_loss=0.04812, over 955121.61 frames. ], batch size: 51, lr: 3.05e-03, grad_scale: 16.0 2023-04-27 20:31:31,881 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-04-27 20:31:39,320 INFO [optim.py:369] (6/7) 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 20:32:02,533 INFO [finetune.py:976] (6/7) Epoch 24, batch 2550, loss[loss=0.1623, simple_loss=0.2371, pruned_loss=0.0438, over 4899.00 frames. ], tot_loss[loss=0.1695, simple_loss=0.2423, pruned_loss=0.04831, over 955594.26 frames. ], batch size: 35, lr: 3.05e-03, grad_scale: 16.0 2023-04-27 20:32:32,156 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=134304.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:32:34,609 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=134308.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:32:35,218 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=134309.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:32:57,004 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.0264, 3.0599, 0.8124, 1.4520, 1.5394, 2.2905, 1.7689, 1.0787], device='cuda:6'), covar=tensor([0.2113, 0.1493, 0.2580, 0.1914, 0.1498, 0.1276, 0.1802, 0.2149], device='cuda:6'), in_proj_covar=tensor([0.0118, 0.0240, 0.0137, 0.0121, 0.0134, 0.0153, 0.0117, 0.0120], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-27 20:33:08,804 INFO [finetune.py:976] (6/7) Epoch 24, batch 2600, loss[loss=0.1563, simple_loss=0.2317, pruned_loss=0.04046, over 4741.00 frames. ], tot_loss[loss=0.1701, simple_loss=0.2435, pruned_loss=0.04833, over 954937.21 frames. ], batch size: 27, lr: 3.05e-03, grad_scale: 16.0 2023-04-27 20:33:27,035 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=134353.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:33:30,513 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=134356.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:33:32,959 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=3.92 vs. limit=5.0 2023-04-27 20:33:42,129 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=134365.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:33:43,229 INFO [optim.py:369] (6/7) 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 20:33:45,764 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=134370.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:33:56,653 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.12 vs. limit=2.0 2023-04-27 20:34:07,321 INFO [finetune.py:976] (6/7) Epoch 24, batch 2650, loss[loss=0.1764, simple_loss=0.2396, pruned_loss=0.05664, over 4721.00 frames. ], tot_loss[loss=0.1694, simple_loss=0.2434, pruned_loss=0.04768, over 955580.36 frames. ], batch size: 54, lr: 3.05e-03, grad_scale: 16.0 2023-04-27 20:34:38,315 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=3.89 vs. limit=5.0 2023-04-27 20:35:02,816 INFO [finetune.py:976] (6/7) Epoch 24, batch 2700, loss[loss=0.1547, simple_loss=0.2306, pruned_loss=0.03941, over 4763.00 frames. ], tot_loss[loss=0.1696, simple_loss=0.2434, pruned_loss=0.04783, over 954958.08 frames. ], batch size: 26, lr: 3.05e-03, grad_scale: 16.0 2023-04-27 20:35:16,010 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-04-27 20:35:22,782 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.14 vs. limit=2.0 2023-04-27 20:35:23,803 INFO [optim.py:369] (6/7) 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 20:35:36,504 INFO [finetune.py:976] (6/7) Epoch 24, batch 2750, loss[loss=0.1358, simple_loss=0.1951, pruned_loss=0.03823, over 3994.00 frames. ], tot_loss[loss=0.1668, simple_loss=0.24, pruned_loss=0.04685, over 955323.24 frames. ], batch size: 17, lr: 3.05e-03, grad_scale: 16.0 2023-04-27 20:36:04,980 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0250, 2.4621, 0.9720, 1.2863, 1.9083, 1.1455, 3.0988, 1.4266], device='cuda:6'), covar=tensor([0.0725, 0.0698, 0.0842, 0.1405, 0.0541, 0.1113, 0.0308, 0.0740], device='cuda:6'), in_proj_covar=tensor([0.0051, 0.0064, 0.0048, 0.0046, 0.0049, 0.0051, 0.0073, 0.0051], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:6') 2023-04-27 20:36:43,196 INFO [finetune.py:976] (6/7) Epoch 24, batch 2800, loss[loss=0.1506, simple_loss=0.2175, pruned_loss=0.04185, over 4814.00 frames. ], tot_loss[loss=0.1644, simple_loss=0.2368, pruned_loss=0.04605, over 956339.23 frames. ], batch size: 30, lr: 3.05e-03, grad_scale: 16.0 2023-04-27 20:37:23,285 INFO [optim.py:369] (6/7) 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 20:37:46,575 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0749, 1.7455, 2.0515, 2.4598, 2.3912, 1.8991, 1.7005, 2.1005], device='cuda:6'), covar=tensor([0.0826, 0.1156, 0.0733, 0.0555, 0.0613, 0.0867, 0.0767, 0.0595], device='cuda:6'), in_proj_covar=tensor([0.0187, 0.0205, 0.0186, 0.0173, 0.0179, 0.0180, 0.0151, 0.0179], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 20:37:48,300 INFO [finetune.py:976] (6/7) Epoch 24, batch 2850, loss[loss=0.1746, simple_loss=0.2501, pruned_loss=0.04955, over 4835.00 frames. ], tot_loss[loss=0.1651, simple_loss=0.2364, pruned_loss=0.04691, over 956905.99 frames. ], batch size: 47, lr: 3.05e-03, grad_scale: 32.0 2023-04-27 20:38:49,114 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8782, 1.1092, 1.5179, 1.6058, 1.5907, 1.6451, 1.4990, 1.4970], device='cuda:6'), covar=tensor([0.3575, 0.4625, 0.3925, 0.3744, 0.4773, 0.6595, 0.4223, 0.4131], device='cuda:6'), in_proj_covar=tensor([0.0339, 0.0374, 0.0327, 0.0338, 0.0346, 0.0393, 0.0357, 0.0330], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 20:38:59,967 INFO [finetune.py:976] (6/7) Epoch 24, batch 2900, loss[loss=0.1704, simple_loss=0.2527, pruned_loss=0.04402, over 4889.00 frames. ], tot_loss[loss=0.1678, simple_loss=0.2401, pruned_loss=0.04776, over 957321.25 frames. ], batch size: 32, lr: 3.05e-03, grad_scale: 32.0 2023-04-27 20:39:10,100 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.4522, 3.1075, 2.6599, 2.8497, 2.1121, 2.6411, 2.7690, 2.2299], device='cuda:6'), covar=tensor([0.1898, 0.1128, 0.0737, 0.1250, 0.3024, 0.1099, 0.1917, 0.2519], device='cuda:6'), in_proj_covar=tensor([0.0284, 0.0299, 0.0214, 0.0275, 0.0312, 0.0254, 0.0249, 0.0264], device='cuda:6'), out_proj_covar=tensor([1.1367e-04, 1.1802e-04, 8.4364e-05, 1.0867e-04, 1.2603e-04, 1.0003e-04, 1.0058e-04, 1.0415e-04], device='cuda:6') 2023-04-27 20:39:15,066 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=134653.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:39:24,552 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=134660.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:39:33,042 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=134665.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:39:34,171 INFO [optim.py:369] (6/7) 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 20:39:52,964 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-04-27 20:40:04,278 INFO [finetune.py:976] (6/7) Epoch 24, batch 2950, loss[loss=0.1753, simple_loss=0.2477, pruned_loss=0.05146, over 4915.00 frames. ], tot_loss[loss=0.1693, simple_loss=0.2425, pruned_loss=0.04807, over 955920.59 frames. ], batch size: 42, lr: 3.05e-03, grad_scale: 32.0 2023-04-27 20:40:07,435 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.3930, 1.7770, 1.7821, 1.8671, 1.7178, 1.8180, 1.8491, 1.7671], device='cuda:6'), covar=tensor([0.3804, 0.5678, 0.4622, 0.4532, 0.5959, 0.7340, 0.5366, 0.5261], device='cuda:6'), in_proj_covar=tensor([0.0341, 0.0375, 0.0328, 0.0339, 0.0348, 0.0394, 0.0358, 0.0331], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 20:40:18,491 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=134701.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:40:25,849 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.6488, 2.3480, 2.5440, 3.0899, 2.8917, 2.2872, 2.0968, 2.5508], device='cuda:6'), covar=tensor([0.0766, 0.0944, 0.0609, 0.0535, 0.0597, 0.0961, 0.0754, 0.0544], device='cuda:6'), in_proj_covar=tensor([0.0187, 0.0205, 0.0187, 0.0173, 0.0179, 0.0180, 0.0151, 0.0179], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 20:40:47,531 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-04-27 20:40:49,709 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0017, 2.3890, 0.9669, 1.3747, 1.8617, 1.1696, 2.9306, 1.5670], device='cuda:6'), covar=tensor([0.0657, 0.0550, 0.0724, 0.1186, 0.0453, 0.0973, 0.0223, 0.0564], device='cuda:6'), in_proj_covar=tensor([0.0050, 0.0064, 0.0047, 0.0046, 0.0049, 0.0051, 0.0073, 0.0050], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:6') 2023-04-27 20:41:09,549 INFO [finetune.py:976] (6/7) Epoch 24, batch 3000, loss[loss=0.1776, simple_loss=0.258, pruned_loss=0.0486, over 4784.00 frames. ], tot_loss[loss=0.1707, simple_loss=0.2442, pruned_loss=0.04856, over 956700.13 frames. ], batch size: 29, lr: 3.05e-03, grad_scale: 32.0 2023-04-27 20:41:09,549 INFO [finetune.py:1001] (6/7) Computing validation loss 2023-04-27 20:41:19,919 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7459, 1.1120, 1.7031, 2.2451, 1.8620, 1.6833, 1.6868, 1.6778], device='cuda:6'), covar=tensor([0.4368, 0.6450, 0.6009, 0.5564, 0.5809, 0.7791, 0.7652, 0.8198], device='cuda:6'), in_proj_covar=tensor([0.0434, 0.0417, 0.0509, 0.0505, 0.0465, 0.0496, 0.0501, 0.0511], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 20:41:24,192 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7282, 1.0457, 1.6594, 2.1981, 1.8311, 1.6677, 1.6433, 1.6509], device='cuda:6'), covar=tensor([0.4476, 0.7076, 0.6341, 0.5929, 0.6286, 0.7588, 0.8218, 0.8696], device='cuda:6'), in_proj_covar=tensor([0.0434, 0.0417, 0.0509, 0.0505, 0.0465, 0.0496, 0.0501, 0.0511], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 20:41:25,510 INFO [finetune.py:1010] (6/7) Epoch 24, validation: loss=0.1526, simple_loss=0.2221, pruned_loss=0.04154, over 2265189.00 frames. 2023-04-27 20:41:25,511 INFO [finetune.py:1011] (6/7) Maximum memory allocated so far is 6435MB 2023-04-27 20:41:44,218 INFO [optim.py:369] (6/7) 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 20:41:57,440 INFO [finetune.py:976] (6/7) Epoch 24, batch 3050, loss[loss=0.1474, simple_loss=0.2098, pruned_loss=0.04248, over 3985.00 frames. ], tot_loss[loss=0.1707, simple_loss=0.2444, pruned_loss=0.04853, over 955300.36 frames. ], batch size: 17, lr: 3.05e-03, grad_scale: 32.0 2023-04-27 20:42:30,086 INFO [finetune.py:976] (6/7) Epoch 24, batch 3100, loss[loss=0.182, simple_loss=0.2552, pruned_loss=0.05444, over 4889.00 frames. ], tot_loss[loss=0.1707, simple_loss=0.2433, pruned_loss=0.04906, over 955841.50 frames. ], batch size: 32, lr: 3.05e-03, grad_scale: 32.0 2023-04-27 20:42:47,441 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=134862.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:42:50,372 INFO [optim.py:369] (6/7) 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 20:42:58,650 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.05 vs. limit=5.0 2023-04-27 20:43:02,621 INFO [finetune.py:976] (6/7) Epoch 24, batch 3150, loss[loss=0.122, simple_loss=0.1921, pruned_loss=0.02596, over 3937.00 frames. ], tot_loss[loss=0.1693, simple_loss=0.2413, pruned_loss=0.0487, over 954159.45 frames. ], batch size: 17, lr: 3.05e-03, grad_scale: 32.0 2023-04-27 20:43:26,867 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.3929, 1.2804, 1.6502, 1.6635, 1.3002, 1.2343, 1.3171, 0.9001], device='cuda:6'), covar=tensor([0.0491, 0.0628, 0.0408, 0.0515, 0.0662, 0.1202, 0.0535, 0.0575], device='cuda:6'), in_proj_covar=tensor([0.0070, 0.0067, 0.0066, 0.0068, 0.0074, 0.0095, 0.0073, 0.0063], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-27 20:43:28,091 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=134923.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:43:36,610 INFO [finetune.py:976] (6/7) Epoch 24, batch 3200, loss[loss=0.1892, simple_loss=0.2481, pruned_loss=0.0652, over 4828.00 frames. ], tot_loss[loss=0.1663, simple_loss=0.238, pruned_loss=0.04729, over 952926.04 frames. ], batch size: 33, lr: 3.05e-03, grad_scale: 32.0 2023-04-27 20:43:41,457 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0474, 2.6094, 1.0367, 1.4410, 2.1364, 1.2574, 3.3121, 1.7265], device='cuda:6'), covar=tensor([0.0663, 0.0678, 0.0823, 0.1220, 0.0477, 0.0986, 0.0265, 0.0613], device='cuda:6'), in_proj_covar=tensor([0.0050, 0.0063, 0.0047, 0.0046, 0.0049, 0.0050, 0.0072, 0.0050], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0009, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0007], device='cuda:6') 2023-04-27 20:43:53,555 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=134960.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:43:56,583 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=134965.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:43:57,656 INFO [optim.py:369] (6/7) 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 20:44:10,010 INFO [finetune.py:976] (6/7) Epoch 24, batch 3250, loss[loss=0.162, simple_loss=0.2256, pruned_loss=0.04924, over 4766.00 frames. ], tot_loss[loss=0.1661, simple_loss=0.2378, pruned_loss=0.04719, over 950464.18 frames. ], batch size: 54, lr: 3.05e-03, grad_scale: 32.0 2023-04-27 20:44:25,317 INFO [zipformer.py:1188] (6/7) 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] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=135013.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:44:39,220 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6751, 1.9957, 1.0750, 1.3441, 2.0339, 1.5054, 1.4221, 1.5144], device='cuda:6'), covar=tensor([0.0479, 0.0349, 0.0299, 0.0542, 0.0257, 0.0490, 0.0483, 0.0568], device='cuda:6'), in_proj_covar=tensor([0.0028, 0.0023, 0.0022, 0.0028, 0.0019, 0.0028, 0.0027, 0.0028], device='cuda:6'), out_proj_covar=tensor([0.0051, 0.0046, 0.0038, 0.0052, 0.0039, 0.0050, 0.0050, 0.0052], device='cuda:6') 2023-04-27 20:44:43,394 INFO [finetune.py:976] (6/7) Epoch 24, batch 3300, loss[loss=0.1603, simple_loss=0.2455, pruned_loss=0.03751, over 4891.00 frames. ], tot_loss[loss=0.1694, simple_loss=0.2419, pruned_loss=0.04848, over 949057.45 frames. ], batch size: 32, lr: 3.05e-03, grad_scale: 32.0 2023-04-27 20:44:53,099 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.80 vs. limit=5.0 2023-04-27 20:45:15,630 INFO [optim.py:369] (6/7) 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 20:45:44,556 INFO [finetune.py:976] (6/7) Epoch 24, batch 3350, loss[loss=0.1973, simple_loss=0.2572, pruned_loss=0.06869, over 4845.00 frames. ], tot_loss[loss=0.1708, simple_loss=0.2437, pruned_loss=0.04899, over 948965.88 frames. ], batch size: 31, lr: 3.05e-03, grad_scale: 32.0 2023-04-27 20:46:49,138 INFO [finetune.py:976] (6/7) Epoch 24, batch 3400, loss[loss=0.1866, simple_loss=0.2548, pruned_loss=0.05918, over 4813.00 frames. ], tot_loss[loss=0.1736, simple_loss=0.2466, pruned_loss=0.0503, over 951857.32 frames. ], batch size: 38, lr: 3.05e-03, grad_scale: 32.0 2023-04-27 20:46:51,726 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.4939, 3.3309, 0.9494, 1.8204, 1.8902, 2.4689, 1.9328, 1.0069], device='cuda:6'), covar=tensor([0.1470, 0.1186, 0.2124, 0.1319, 0.1121, 0.1030, 0.1665, 0.2127], device='cuda:6'), in_proj_covar=tensor([0.0116, 0.0236, 0.0134, 0.0119, 0.0131, 0.0151, 0.0116, 0.0118], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-27 20:47:25,526 INFO [optim.py:369] (6/7) 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 20:47:33,937 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=135171.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:47:53,650 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0950, 2.5544, 2.1773, 2.4249, 1.5480, 2.2792, 2.2935, 1.6914], device='cuda:6'), covar=tensor([0.1715, 0.1029, 0.0768, 0.1086, 0.3821, 0.1002, 0.1555, 0.2365], device='cuda:6'), in_proj_covar=tensor([0.0282, 0.0297, 0.0214, 0.0275, 0.0311, 0.0253, 0.0248, 0.0263], device='cuda:6'), out_proj_covar=tensor([1.1276e-04, 1.1754e-04, 8.4068e-05, 1.0838e-04, 1.2552e-04, 9.9620e-05, 1.0014e-04, 1.0372e-04], device='cuda:6') 2023-04-27 20:47:54,155 INFO [finetune.py:976] (6/7) Epoch 24, batch 3450, loss[loss=0.1951, simple_loss=0.2686, pruned_loss=0.0608, over 4909.00 frames. ], tot_loss[loss=0.1734, simple_loss=0.2465, pruned_loss=0.0501, over 951813.16 frames. ], batch size: 37, lr: 3.05e-03, grad_scale: 32.0 2023-04-27 20:48:06,434 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=3.23 vs. limit=5.0 2023-04-27 20:48:27,448 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=135218.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:48:35,268 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8614, 1.3306, 1.5603, 2.1610, 2.1882, 1.7433, 1.3580, 1.9524], device='cuda:6'), covar=tensor([0.0833, 0.1619, 0.1090, 0.0563, 0.0577, 0.0923, 0.0932, 0.0583], device='cuda:6'), in_proj_covar=tensor([0.0187, 0.0205, 0.0186, 0.0174, 0.0179, 0.0179, 0.0151, 0.0179], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 20:48:36,530 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=135232.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:48:39,440 INFO [finetune.py:976] (6/7) Epoch 24, batch 3500, loss[loss=0.1342, simple_loss=0.2058, pruned_loss=0.03136, over 4907.00 frames. ], tot_loss[loss=0.1704, simple_loss=0.243, pruned_loss=0.04888, over 952212.11 frames. ], batch size: 32, lr: 3.04e-03, grad_scale: 32.0 2023-04-27 20:48:59,253 INFO [optim.py:369] (6/7) 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] (6/7) Epoch 24, batch 3550, loss[loss=0.1783, simple_loss=0.2441, pruned_loss=0.05623, over 4904.00 frames. ], tot_loss[loss=0.1673, simple_loss=0.2394, pruned_loss=0.04757, over 954399.02 frames. ], batch size: 43, lr: 3.04e-03, grad_scale: 32.0 2023-04-27 20:49:47,282 INFO [finetune.py:976] (6/7) Epoch 24, batch 3600, loss[loss=0.1629, simple_loss=0.2323, pruned_loss=0.04678, over 4823.00 frames. ], tot_loss[loss=0.1636, simple_loss=0.2357, pruned_loss=0.04575, over 955878.77 frames. ], batch size: 51, lr: 3.04e-03, grad_scale: 32.0 2023-04-27 20:49:55,958 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0635, 1.9040, 2.2494, 2.4674, 2.1319, 1.9375, 2.0659, 2.0528], device='cuda:6'), covar=tensor([0.4858, 0.6990, 0.7366, 0.6078, 0.6487, 0.9464, 0.9458, 1.0096], device='cuda:6'), in_proj_covar=tensor([0.0436, 0.0419, 0.0512, 0.0506, 0.0465, 0.0498, 0.0503, 0.0513], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 20:50:02,946 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 2023-04-27 20:50:05,963 INFO [optim.py:369] (6/7) 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] (6/7) Epoch 24, batch 3650, loss[loss=0.1452, simple_loss=0.2311, pruned_loss=0.02968, over 4857.00 frames. ], tot_loss[loss=0.1656, simple_loss=0.2379, pruned_loss=0.04669, over 954322.58 frames. ], batch size: 44, lr: 3.04e-03, grad_scale: 32.0 2023-04-27 20:50:52,243 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8377, 2.3337, 1.9478, 2.2475, 1.5542, 2.0359, 1.9521, 1.4934], device='cuda:6'), covar=tensor([0.1992, 0.1016, 0.0793, 0.1274, 0.3374, 0.1019, 0.1992, 0.2680], device='cuda:6'), in_proj_covar=tensor([0.0286, 0.0300, 0.0216, 0.0277, 0.0314, 0.0255, 0.0250, 0.0265], device='cuda:6'), out_proj_covar=tensor([1.1418e-04, 1.1863e-04, 8.4779e-05, 1.0953e-04, 1.2670e-04, 1.0052e-04, 1.0114e-04, 1.0475e-04], device='cuda:6') 2023-04-27 20:50:53,967 INFO [finetune.py:976] (6/7) Epoch 24, batch 3700, loss[loss=0.1692, simple_loss=0.2527, pruned_loss=0.04284, over 4752.00 frames. ], tot_loss[loss=0.1686, simple_loss=0.2415, pruned_loss=0.04785, over 955562.93 frames. ], batch size: 59, lr: 3.04e-03, grad_scale: 32.0 2023-04-27 20:51:12,482 INFO [optim.py:369] (6/7) 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 20:51:18,472 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7000, 1.2950, 1.4061, 1.4127, 1.7756, 1.5310, 1.2542, 1.3180], device='cuda:6'), covar=tensor([0.1464, 0.1211, 0.1714, 0.1169, 0.0797, 0.1299, 0.1714, 0.2050], device='cuda:6'), in_proj_covar=tensor([0.0311, 0.0306, 0.0349, 0.0283, 0.0326, 0.0304, 0.0297, 0.0370], device='cuda:6'), out_proj_covar=tensor([6.3726e-05, 6.3098e-05, 7.3515e-05, 5.6752e-05, 6.7035e-05, 6.3790e-05, 6.1779e-05, 7.8409e-05], device='cuda:6') 2023-04-27 20:51:23,031 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5566, 1.2306, 4.2141, 3.9656, 3.6744, 3.9921, 3.9792, 3.7449], device='cuda:6'), covar=tensor([0.6861, 0.5999, 0.1116, 0.1743, 0.1103, 0.1456, 0.1292, 0.1578], device='cuda:6'), in_proj_covar=tensor([0.0309, 0.0304, 0.0404, 0.0407, 0.0347, 0.0407, 0.0317, 0.0367], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-27 20:51:27,088 INFO [finetune.py:976] (6/7) Epoch 24, batch 3750, loss[loss=0.1712, simple_loss=0.2464, pruned_loss=0.04794, over 4790.00 frames. ], tot_loss[loss=0.1705, simple_loss=0.2435, pruned_loss=0.04879, over 954961.13 frames. ], batch size: 25, lr: 3.04e-03, grad_scale: 32.0 2023-04-27 20:51:35,862 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-04-27 20:52:05,348 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=135518.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:52:16,496 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=135527.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:52:29,079 INFO [finetune.py:976] (6/7) Epoch 24, batch 3800, loss[loss=0.1494, simple_loss=0.2169, pruned_loss=0.04097, over 4669.00 frames. ], tot_loss[loss=0.1719, simple_loss=0.2451, pruned_loss=0.04933, over 954705.84 frames. ], batch size: 23, lr: 3.04e-03, grad_scale: 32.0 2023-04-27 20:53:08,378 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=135566.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:53:08,917 INFO [optim.py:369] (6/7) 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 20:53:20,332 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-04-27 20:53:32,878 INFO [finetune.py:976] (6/7) Epoch 24, batch 3850, loss[loss=0.1492, simple_loss=0.2204, pruned_loss=0.03898, over 4781.00 frames. ], tot_loss[loss=0.1703, simple_loss=0.2434, pruned_loss=0.04855, over 955512.29 frames. ], batch size: 28, lr: 3.04e-03, grad_scale: 32.0 2023-04-27 20:54:08,510 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.3420, 2.0612, 1.7427, 1.7371, 2.1436, 1.8397, 2.4070, 1.6102], device='cuda:6'), covar=tensor([0.2729, 0.1335, 0.3676, 0.2289, 0.1128, 0.1767, 0.1182, 0.3586], device='cuda:6'), in_proj_covar=tensor([0.0338, 0.0352, 0.0424, 0.0351, 0.0378, 0.0376, 0.0368, 0.0421], device='cuda:6'), out_proj_covar=tensor([9.9941e-05, 1.0504e-04, 1.2868e-04, 1.0544e-04, 1.1230e-04, 1.1203e-04, 1.0801e-04, 1.2675e-04], device='cuda:6') 2023-04-27 20:54:28,822 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=135630.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:54:33,841 INFO [finetune.py:976] (6/7) Epoch 24, batch 3900, loss[loss=0.1878, simple_loss=0.2588, pruned_loss=0.05843, over 4899.00 frames. ], tot_loss[loss=0.1693, simple_loss=0.2416, pruned_loss=0.04857, over 954478.96 frames. ], batch size: 36, lr: 3.04e-03, grad_scale: 32.0 2023-04-27 20:55:15,857 INFO [optim.py:369] (6/7) 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 20:55:44,884 INFO [finetune.py:976] (6/7) Epoch 24, batch 3950, loss[loss=0.13, simple_loss=0.2012, pruned_loss=0.02937, over 4824.00 frames. ], tot_loss[loss=0.1669, simple_loss=0.2387, pruned_loss=0.04754, over 957105.03 frames. ], batch size: 25, lr: 3.04e-03, grad_scale: 32.0 2023-04-27 20:55:47,438 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=135691.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:56:19,341 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0417, 2.4045, 2.0605, 2.3948, 1.5499, 2.0775, 2.0823, 1.6396], device='cuda:6'), covar=tensor([0.1793, 0.1019, 0.0857, 0.1027, 0.3636, 0.1140, 0.1707, 0.2613], device='cuda:6'), in_proj_covar=tensor([0.0284, 0.0299, 0.0214, 0.0276, 0.0312, 0.0254, 0.0249, 0.0263], device='cuda:6'), out_proj_covar=tensor([1.1337e-04, 1.1783e-04, 8.4362e-05, 1.0875e-04, 1.2594e-04, 9.9947e-05, 1.0053e-04, 1.0400e-04], device='cuda:6') 2023-04-27 20:56:44,336 INFO [finetune.py:976] (6/7) Epoch 24, batch 4000, loss[loss=0.1726, simple_loss=0.2417, pruned_loss=0.05171, over 4736.00 frames. ], tot_loss[loss=0.1679, simple_loss=0.2393, pruned_loss=0.04821, over 955647.71 frames. ], batch size: 59, lr: 3.04e-03, grad_scale: 32.0 2023-04-27 20:57:22,849 INFO [optim.py:369] (6/7) 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] (6/7) Epoch 24, batch 4050, loss[loss=0.1774, simple_loss=0.2679, pruned_loss=0.04344, over 4898.00 frames. ], tot_loss[loss=0.171, simple_loss=0.243, pruned_loss=0.04948, over 955740.36 frames. ], batch size: 35, lr: 3.04e-03, grad_scale: 32.0 2023-04-27 20:58:02,509 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([0.9709, 1.9609, 1.8655, 1.6555, 2.1281, 1.8234, 2.6207, 1.6380], device='cuda:6'), covar=tensor([0.3338, 0.1805, 0.4853, 0.2597, 0.1554, 0.2171, 0.1442, 0.4225], device='cuda:6'), in_proj_covar=tensor([0.0339, 0.0354, 0.0428, 0.0352, 0.0379, 0.0378, 0.0369, 0.0424], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 20:58:05,486 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.4390, 1.3527, 1.7609, 1.7456, 1.3052, 1.1817, 1.4004, 0.8106], device='cuda:6'), covar=tensor([0.0530, 0.0668, 0.0332, 0.0505, 0.0713, 0.1063, 0.0505, 0.0578], device='cuda:6'), in_proj_covar=tensor([0.0070, 0.0067, 0.0066, 0.0068, 0.0074, 0.0095, 0.0073, 0.0063], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-27 20:58:26,676 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.2477, 2.5979, 2.1835, 2.6743, 1.8314, 2.3091, 2.2200, 1.8189], device='cuda:6'), covar=tensor([0.1811, 0.1251, 0.0851, 0.1015, 0.3387, 0.1004, 0.2186, 0.2771], device='cuda:6'), in_proj_covar=tensor([0.0283, 0.0299, 0.0215, 0.0276, 0.0312, 0.0254, 0.0249, 0.0263], device='cuda:6'), out_proj_covar=tensor([1.1317e-04, 1.1821e-04, 8.4523e-05, 1.0871e-04, 1.2603e-04, 9.9961e-05, 1.0046e-04, 1.0397e-04], device='cuda:6') 2023-04-27 20:58:39,904 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=135827.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:58:47,828 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8369, 1.4280, 1.4360, 1.6261, 1.9413, 1.5883, 1.3825, 1.3764], device='cuda:6'), covar=tensor([0.1704, 0.1598, 0.1863, 0.1350, 0.0945, 0.1854, 0.2110, 0.2186], device='cuda:6'), in_proj_covar=tensor([0.0308, 0.0304, 0.0345, 0.0280, 0.0322, 0.0301, 0.0294, 0.0366], device='cuda:6'), out_proj_covar=tensor([6.3035e-05, 6.2545e-05, 7.2528e-05, 5.6198e-05, 6.6187e-05, 6.2992e-05, 6.1169e-05, 7.7659e-05], device='cuda:6') 2023-04-27 20:58:52,538 INFO [finetune.py:976] (6/7) Epoch 24, batch 4100, loss[loss=0.1683, simple_loss=0.2445, pruned_loss=0.04607, over 4890.00 frames. ], tot_loss[loss=0.1711, simple_loss=0.244, pruned_loss=0.04903, over 957414.38 frames. ], batch size: 32, lr: 3.04e-03, grad_scale: 32.0 2023-04-27 20:59:20,844 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=3.94 vs. limit=5.0 2023-04-27 20:59:21,943 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-04-27 20:59:24,284 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8485, 1.5064, 1.9410, 2.3314, 1.9748, 1.8051, 1.8860, 1.8426], device='cuda:6'), covar=tensor([0.4580, 0.6628, 0.6421, 0.5461, 0.6140, 0.7975, 0.7826, 0.8680], device='cuda:6'), in_proj_covar=tensor([0.0434, 0.0417, 0.0509, 0.0506, 0.0465, 0.0497, 0.0501, 0.0513], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 20:59:28,242 INFO [optim.py:369] (6/7) 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 20:59:33,114 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=135875.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:59:40,399 INFO [finetune.py:976] (6/7) Epoch 24, batch 4150, loss[loss=0.1514, simple_loss=0.2393, pruned_loss=0.03176, over 4806.00 frames. ], tot_loss[loss=0.1704, simple_loss=0.2434, pruned_loss=0.04868, over 954758.10 frames. ], batch size: 38, lr: 3.04e-03, grad_scale: 32.0 2023-04-27 20:59:46,786 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-04-27 20:59:54,943 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.5902, 1.8463, 1.7528, 2.3450, 2.5253, 2.0931, 2.0637, 1.7806], device='cuda:6'), covar=tensor([0.1674, 0.1842, 0.1853, 0.1727, 0.1139, 0.1756, 0.1926, 0.2311], device='cuda:6'), in_proj_covar=tensor([0.0308, 0.0303, 0.0345, 0.0279, 0.0321, 0.0300, 0.0293, 0.0366], device='cuda:6'), out_proj_covar=tensor([6.2972e-05, 6.2362e-05, 7.2536e-05, 5.5936e-05, 6.6052e-05, 6.2847e-05, 6.1019e-05, 7.7620e-05], device='cuda:6') 2023-04-27 21:00:07,656 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9575, 1.7398, 2.1716, 2.3456, 2.0284, 1.9067, 2.0493, 1.9332], device='cuda:6'), covar=tensor([0.4711, 0.7002, 0.6551, 0.5877, 0.6165, 0.8184, 0.8907, 1.0399], device='cuda:6'), in_proj_covar=tensor([0.0436, 0.0420, 0.0511, 0.0508, 0.0468, 0.0499, 0.0503, 0.0516], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 21:00:14,128 INFO [finetune.py:976] (6/7) Epoch 24, batch 4200, loss[loss=0.1456, simple_loss=0.2331, pruned_loss=0.02908, over 4921.00 frames. ], tot_loss[loss=0.1698, simple_loss=0.2434, pruned_loss=0.04813, over 954020.49 frames. ], batch size: 41, lr: 3.04e-03, grad_scale: 32.0 2023-04-27 21:00:35,175 INFO [optim.py:369] (6/7) 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 21:00:44,749 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=135982.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:00:47,167 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=135986.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:00:47,699 INFO [finetune.py:976] (6/7) Epoch 24, batch 4250, loss[loss=0.2098, simple_loss=0.2663, pruned_loss=0.07664, over 4902.00 frames. ], tot_loss[loss=0.1687, simple_loss=0.242, pruned_loss=0.04775, over 954735.78 frames. ], batch size: 46, lr: 3.04e-03, grad_scale: 32.0 2023-04-27 21:00:52,109 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 24, batch 4300, loss[loss=0.1522, simple_loss=0.2245, pruned_loss=0.03996, over 4818.00 frames. ], tot_loss[loss=0.167, simple_loss=0.2398, pruned_loss=0.04712, over 957193.66 frames. ], batch size: 38, lr: 3.04e-03, grad_scale: 32.0 2023-04-27 21:01:24,563 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-04-27 21:01:25,996 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=136043.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:01:34,849 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=136055.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:01:42,942 INFO [optim.py:369] (6/7) 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 21:01:45,476 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5047, 3.3215, 0.9359, 1.7429, 2.0145, 2.3456, 1.9567, 1.0190], device='cuda:6'), covar=tensor([0.1378, 0.0926, 0.1891, 0.1255, 0.0996, 0.1015, 0.1472, 0.1892], device='cuda:6'), in_proj_covar=tensor([0.0117, 0.0238, 0.0135, 0.0120, 0.0131, 0.0151, 0.0117, 0.0118], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-27 21:01:55,531 INFO [finetune.py:976] (6/7) Epoch 24, batch 4350, loss[loss=0.1633, simple_loss=0.2278, pruned_loss=0.04937, over 4716.00 frames. ], tot_loss[loss=0.1648, simple_loss=0.2369, pruned_loss=0.04638, over 957512.69 frames. ], batch size: 23, lr: 3.04e-03, grad_scale: 32.0 2023-04-27 21:01:59,337 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8809, 2.2753, 2.0107, 2.2741, 1.5899, 1.9408, 1.8642, 1.5542], device='cuda:6'), covar=tensor([0.1638, 0.1078, 0.0727, 0.0932, 0.3278, 0.1018, 0.1736, 0.2206], device='cuda:6'), in_proj_covar=tensor([0.0285, 0.0299, 0.0216, 0.0277, 0.0313, 0.0255, 0.0249, 0.0264], device='cuda:6'), out_proj_covar=tensor([1.1374e-04, 1.1810e-04, 8.4849e-05, 1.0908e-04, 1.2624e-04, 1.0061e-04, 1.0067e-04, 1.0420e-04], device='cuda:6') 2023-04-27 21:02:17,101 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-04-27 21:02:33,974 INFO [finetune.py:976] (6/7) Epoch 24, batch 4400, loss[loss=0.2145, simple_loss=0.2709, pruned_loss=0.07902, over 3954.00 frames. ], tot_loss[loss=0.1675, simple_loss=0.239, pruned_loss=0.048, over 956281.98 frames. ], batch size: 65, lr: 3.04e-03, grad_scale: 32.0 2023-04-27 21:02:34,714 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.2348, 2.9116, 1.0316, 1.7007, 2.3467, 1.3968, 3.8185, 2.0573], device='cuda:6'), covar=tensor([0.0661, 0.0790, 0.0812, 0.1196, 0.0463, 0.0959, 0.0176, 0.0570], device='cuda:6'), in_proj_covar=tensor([0.0050, 0.0064, 0.0047, 0.0046, 0.0048, 0.0051, 0.0072, 0.0050], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:6') 2023-04-27 21:02:42,608 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=136142.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:03:16,193 INFO [optim.py:369] (6/7) 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 21:03:40,890 INFO [finetune.py:976] (6/7) Epoch 24, batch 4450, loss[loss=0.2132, simple_loss=0.2845, pruned_loss=0.07097, over 4926.00 frames. ], tot_loss[loss=0.1704, simple_loss=0.2429, pruned_loss=0.04891, over 958105.79 frames. ], batch size: 38, lr: 3.04e-03, grad_scale: 32.0 2023-04-27 21:03:59,665 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5280, 2.4057, 1.8736, 2.1644, 2.5039, 2.0691, 3.0504, 1.7129], device='cuda:6'), covar=tensor([0.3720, 0.2225, 0.4682, 0.3278, 0.1735, 0.2668, 0.2083, 0.4472], device='cuda:6'), in_proj_covar=tensor([0.0334, 0.0348, 0.0419, 0.0346, 0.0374, 0.0371, 0.0364, 0.0416], device='cuda:6'), out_proj_covar=tensor([9.8762e-05, 1.0396e-04, 1.2713e-04, 1.0400e-04, 1.1087e-04, 1.1037e-04, 1.0690e-04, 1.2521e-04], device='cuda:6') 2023-04-27 21:04:02,273 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=136203.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:04:53,812 INFO [finetune.py:976] (6/7) Epoch 24, batch 4500, loss[loss=0.1686, simple_loss=0.2401, pruned_loss=0.04857, over 4921.00 frames. ], tot_loss[loss=0.1728, simple_loss=0.2456, pruned_loss=0.05005, over 959865.16 frames. ], batch size: 33, lr: 3.04e-03, grad_scale: 32.0 2023-04-27 21:04:54,001 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-04-27 21:05:29,628 INFO [optim.py:369] (6/7) 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 21:05:58,021 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=136286.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:05:58,525 INFO [finetune.py:976] (6/7) Epoch 24, batch 4550, loss[loss=0.1957, simple_loss=0.2645, pruned_loss=0.0634, over 4729.00 frames. ], tot_loss[loss=0.1718, simple_loss=0.2447, pruned_loss=0.04942, over 957975.24 frames. ], batch size: 59, lr: 3.04e-03, grad_scale: 32.0 2023-04-27 21:06:56,734 INFO [zipformer.py:1188] (6/7) 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] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=136335.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:07:04,274 INFO [finetune.py:976] (6/7) Epoch 24, batch 4600, loss[loss=0.2157, simple_loss=0.269, pruned_loss=0.08122, over 4798.00 frames. ], tot_loss[loss=0.1704, simple_loss=0.2434, pruned_loss=0.04873, over 956300.07 frames. ], batch size: 45, lr: 3.04e-03, grad_scale: 32.0 2023-04-27 21:07:04,985 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=136338.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:07:18,412 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=136350.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:07:40,419 INFO [optim.py:369] (6/7) 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 21:08:03,900 INFO [finetune.py:976] (6/7) Epoch 24, batch 4650, loss[loss=0.2058, simple_loss=0.2702, pruned_loss=0.07071, over 4182.00 frames. ], tot_loss[loss=0.1688, simple_loss=0.2413, pruned_loss=0.04814, over 956334.75 frames. ], batch size: 65, lr: 3.04e-03, grad_scale: 32.0 2023-04-27 21:08:20,810 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=136396.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:09:06,182 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.1813, 1.8658, 2.1233, 2.4726, 2.4339, 2.1062, 1.8898, 2.3476], device='cuda:6'), covar=tensor([0.0765, 0.1083, 0.0600, 0.0526, 0.0599, 0.0793, 0.0706, 0.0470], device='cuda:6'), in_proj_covar=tensor([0.0188, 0.0204, 0.0187, 0.0175, 0.0179, 0.0180, 0.0152, 0.0180], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 21:09:13,811 INFO [finetune.py:976] (6/7) Epoch 24, batch 4700, loss[loss=0.1654, simple_loss=0.2337, pruned_loss=0.0486, over 4906.00 frames. ], tot_loss[loss=0.1675, simple_loss=0.2394, pruned_loss=0.04786, over 957538.84 frames. ], batch size: 35, lr: 3.04e-03, grad_scale: 32.0 2023-04-27 21:09:15,799 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.2593, 1.7266, 1.5492, 2.1317, 2.2984, 1.9690, 1.8759, 1.5716], device='cuda:6'), covar=tensor([0.2066, 0.1815, 0.1855, 0.1534, 0.1105, 0.2395, 0.2212, 0.2447], device='cuda:6'), in_proj_covar=tensor([0.0312, 0.0307, 0.0350, 0.0285, 0.0326, 0.0304, 0.0297, 0.0372], device='cuda:6'), out_proj_covar=tensor([6.3878e-05, 6.3265e-05, 7.3656e-05, 5.7101e-05, 6.7014e-05, 6.3735e-05, 6.1743e-05, 7.8861e-05], device='cuda:6') 2023-04-27 21:09:45,097 INFO [optim.py:369] (6/7) 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 21:09:50,609 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.3155, 1.6502, 1.4801, 1.7763, 1.6925, 1.9819, 1.4634, 3.4504], device='cuda:6'), covar=tensor([0.0577, 0.0729, 0.0715, 0.1104, 0.0582, 0.0535, 0.0690, 0.0148], device='cuda:6'), in_proj_covar=tensor([0.0038, 0.0039, 0.0039, 0.0043, 0.0040, 0.0038, 0.0038, 0.0055], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0013, 0.0013, 0.0015, 0.0014, 0.0013, 0.0013, 0.0016], device='cuda:6') 2023-04-27 21:09:58,844 INFO [finetune.py:976] (6/7) Epoch 24, batch 4750, loss[loss=0.1706, simple_loss=0.2588, pruned_loss=0.04117, over 4741.00 frames. ], tot_loss[loss=0.1673, simple_loss=0.2385, pruned_loss=0.04805, over 958229.92 frames. ], batch size: 59, lr: 3.04e-03, grad_scale: 32.0 2023-04-27 21:10:07,139 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=136498.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:10:10,208 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([3.4789, 2.9744, 2.8055, 2.9414, 2.7207, 2.9062, 2.8858, 2.7789], device='cuda:6'), covar=tensor([0.3286, 0.4388, 0.3768, 0.4038, 0.4830, 0.5561, 0.4837, 0.4264], device='cuda:6'), in_proj_covar=tensor([0.0340, 0.0372, 0.0325, 0.0339, 0.0347, 0.0394, 0.0356, 0.0330], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 21:10:11,424 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0771, 1.9612, 2.5541, 2.7513, 1.8426, 1.5948, 2.0283, 1.1068], device='cuda:6'), covar=tensor([0.0544, 0.0717, 0.0382, 0.0468, 0.0687, 0.1077, 0.0643, 0.0761], device='cuda:6'), in_proj_covar=tensor([0.0070, 0.0067, 0.0066, 0.0067, 0.0074, 0.0094, 0.0072, 0.0063], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-27 21:10:47,891 INFO [finetune.py:976] (6/7) Epoch 24, batch 4800, loss[loss=0.1879, simple_loss=0.262, pruned_loss=0.05692, over 4897.00 frames. ], tot_loss[loss=0.1695, simple_loss=0.2411, pruned_loss=0.04899, over 956167.82 frames. ], batch size: 35, lr: 3.04e-03, grad_scale: 32.0 2023-04-27 21:10:58,095 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8464, 4.2015, 0.8940, 2.2382, 2.3256, 2.9674, 2.4710, 1.1284], device='cuda:6'), covar=tensor([0.1323, 0.0810, 0.2104, 0.1179, 0.1058, 0.0943, 0.1363, 0.2003], device='cuda:6'), in_proj_covar=tensor([0.0116, 0.0237, 0.0134, 0.0120, 0.0131, 0.0151, 0.0116, 0.0118], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-27 21:11:11,212 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7769, 1.4266, 1.9166, 2.1405, 1.8456, 1.7567, 1.8843, 1.8225], device='cuda:6'), covar=tensor([0.4341, 0.6833, 0.6227, 0.5891, 0.5617, 0.7727, 0.7595, 0.9118], device='cuda:6'), in_proj_covar=tensor([0.0436, 0.0420, 0.0512, 0.0507, 0.0467, 0.0499, 0.0502, 0.0517], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 21:11:29,605 INFO [optim.py:369] (6/7) 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 21:11:53,502 INFO [finetune.py:976] (6/7) Epoch 24, batch 4850, loss[loss=0.1881, simple_loss=0.2345, pruned_loss=0.07084, over 4247.00 frames. ], tot_loss[loss=0.1719, simple_loss=0.2439, pruned_loss=0.04989, over 954569.99 frames. ], batch size: 18, lr: 3.04e-03, grad_scale: 64.0 2023-04-27 21:12:00,712 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.3741, 1.2377, 1.6638, 1.5603, 1.2843, 1.1833, 1.2924, 0.7928], device='cuda:6'), covar=tensor([0.0524, 0.0549, 0.0335, 0.0495, 0.0740, 0.1126, 0.0470, 0.0530], device='cuda:6'), in_proj_covar=tensor([0.0070, 0.0067, 0.0066, 0.0068, 0.0074, 0.0095, 0.0073, 0.0063], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-27 21:12:12,270 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.2040, 2.6938, 1.0702, 1.5663, 2.1840, 1.3682, 3.6644, 1.9767], device='cuda:6'), covar=tensor([0.0660, 0.0650, 0.0836, 0.1246, 0.0515, 0.1001, 0.0203, 0.0612], device='cuda:6'), in_proj_covar=tensor([0.0050, 0.0064, 0.0047, 0.0046, 0.0049, 0.0051, 0.0073, 0.0051], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:6') 2023-04-27 21:12:26,526 INFO [finetune.py:976] (6/7) Epoch 24, batch 4900, loss[loss=0.1794, simple_loss=0.2672, pruned_loss=0.04584, over 4738.00 frames. ], tot_loss[loss=0.1713, simple_loss=0.2442, pruned_loss=0.04924, over 955044.85 frames. ], batch size: 54, lr: 3.03e-03, grad_scale: 32.0 2023-04-27 21:12:27,222 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6121, 1.1591, 4.2831, 4.0210, 3.7674, 4.0430, 3.9472, 3.7379], device='cuda:6'), covar=tensor([0.7365, 0.6009, 0.1038, 0.1749, 0.1113, 0.1278, 0.2037, 0.1693], device='cuda:6'), in_proj_covar=tensor([0.0312, 0.0305, 0.0406, 0.0408, 0.0351, 0.0409, 0.0320, 0.0368], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-27 21:12:27,242 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=136638.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:12:30,264 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=136643.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:12:35,591 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=136650.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:12:46,889 INFO [optim.py:369] (6/7) 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 21:12:50,095 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7419, 1.3059, 1.7523, 2.1667, 1.8131, 1.7042, 1.7711, 1.6757], device='cuda:6'), covar=tensor([0.4319, 0.6738, 0.6069, 0.5616, 0.5467, 0.7685, 0.7520, 0.8903], device='cuda:6'), in_proj_covar=tensor([0.0437, 0.0421, 0.0512, 0.0509, 0.0467, 0.0500, 0.0503, 0.0518], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 21:12:58,274 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 24, batch 4950, loss[loss=0.1881, simple_loss=0.2608, pruned_loss=0.0577, over 4313.00 frames. ], tot_loss[loss=0.1718, simple_loss=0.2452, pruned_loss=0.04921, over 954444.20 frames. ], batch size: 66, lr: 3.03e-03, grad_scale: 32.0 2023-04-27 21:13:02,152 INFO [zipformer.py:1188] (6/7) 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] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=136698.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:13:11,675 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=136704.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:13:32,453 INFO [finetune.py:976] (6/7) Epoch 24, batch 5000, loss[loss=0.1246, simple_loss=0.1886, pruned_loss=0.03028, over 4200.00 frames. ], tot_loss[loss=0.1704, simple_loss=0.2427, pruned_loss=0.04909, over 951875.64 frames. ], batch size: 18, lr: 3.03e-03, grad_scale: 32.0 2023-04-27 21:13:43,468 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7288, 1.9714, 1.2161, 1.5223, 1.9703, 1.5811, 1.5491, 1.6388], device='cuda:6'), covar=tensor([0.0411, 0.0293, 0.0318, 0.0453, 0.0267, 0.0407, 0.0396, 0.0469], device='cuda:6'), in_proj_covar=tensor([0.0028, 0.0023, 0.0022, 0.0028, 0.0019, 0.0028, 0.0027, 0.0028], device='cuda:6'), out_proj_covar=tensor([0.0051, 0.0046, 0.0038, 0.0053, 0.0039, 0.0050, 0.0050, 0.0052], device='cuda:6') 2023-04-27 21:13:53,647 INFO [optim.py:369] (6/7) 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 21:14:05,879 INFO [finetune.py:976] (6/7) Epoch 24, batch 5050, loss[loss=0.1819, simple_loss=0.2513, pruned_loss=0.0563, over 4910.00 frames. ], tot_loss[loss=0.1684, simple_loss=0.2401, pruned_loss=0.04835, over 947562.31 frames. ], batch size: 46, lr: 3.03e-03, grad_scale: 32.0 2023-04-27 21:14:16,392 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8368, 2.2621, 0.9617, 1.1751, 1.5668, 1.1022, 2.4509, 1.2905], device='cuda:6'), covar=tensor([0.0656, 0.0531, 0.0634, 0.1268, 0.0459, 0.0979, 0.0322, 0.0740], device='cuda:6'), in_proj_covar=tensor([0.0050, 0.0064, 0.0047, 0.0046, 0.0048, 0.0051, 0.0072, 0.0050], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0009, 0.0007, 0.0007, 0.0007, 0.0008, 0.0010, 0.0007], device='cuda:6') 2023-04-27 21:14:18,797 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=136798.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:14:29,845 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=136806.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:14:31,073 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7952, 2.0280, 1.1648, 1.4291, 2.0090, 1.6182, 1.5116, 1.5622], device='cuda:6'), covar=tensor([0.0491, 0.0345, 0.0309, 0.0570, 0.0264, 0.0521, 0.0515, 0.0588], device='cuda:6'), in_proj_covar=tensor([0.0028, 0.0023, 0.0022, 0.0028, 0.0019, 0.0028, 0.0027, 0.0028], device='cuda:6'), out_proj_covar=tensor([0.0051, 0.0046, 0.0038, 0.0052, 0.0039, 0.0050, 0.0050, 0.0052], device='cuda:6') 2023-04-27 21:15:01,717 INFO [finetune.py:976] (6/7) Epoch 24, batch 5100, loss[loss=0.1541, simple_loss=0.2114, pruned_loss=0.0484, over 4389.00 frames. ], tot_loss[loss=0.1666, simple_loss=0.2377, pruned_loss=0.04776, over 948433.30 frames. ], batch size: 19, lr: 3.03e-03, grad_scale: 32.0 2023-04-27 21:15:07,746 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=136846.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:15:22,630 INFO [zipformer.py:1188] (6/7) 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] (6/7) 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 21:15:28,009 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5248, 1.3144, 4.3287, 4.0308, 3.8311, 4.1779, 4.1177, 3.7733], device='cuda:6'), covar=tensor([0.6927, 0.5848, 0.1011, 0.1588, 0.1131, 0.1840, 0.1256, 0.1963], device='cuda:6'), in_proj_covar=tensor([0.0313, 0.0307, 0.0406, 0.0409, 0.0350, 0.0410, 0.0320, 0.0369], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-27 21:15:34,951 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.78 vs. limit=2.0 2023-04-27 21:15:35,188 INFO [finetune.py:976] (6/7) Epoch 24, batch 5150, loss[loss=0.1655, simple_loss=0.2356, pruned_loss=0.04774, over 4775.00 frames. ], tot_loss[loss=0.1666, simple_loss=0.2376, pruned_loss=0.0478, over 949031.90 frames. ], batch size: 28, lr: 3.03e-03, grad_scale: 32.0 2023-04-27 21:15:40,732 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.3964, 1.7531, 1.8284, 1.8995, 1.7710, 1.7952, 1.8483, 1.8631], device='cuda:6'), covar=tensor([0.4188, 0.4971, 0.4092, 0.4063, 0.5437, 0.7013, 0.5099, 0.4515], device='cuda:6'), in_proj_covar=tensor([0.0340, 0.0372, 0.0324, 0.0339, 0.0347, 0.0393, 0.0357, 0.0330], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 21:16:29,343 INFO [finetune.py:976] (6/7) Epoch 24, batch 5200, loss[loss=0.1559, simple_loss=0.2327, pruned_loss=0.03955, over 4791.00 frames. ], tot_loss[loss=0.1678, simple_loss=0.2398, pruned_loss=0.04786, over 949527.66 frames. ], batch size: 29, lr: 3.03e-03, grad_scale: 32.0 2023-04-27 21:16:59,706 INFO [optim.py:369] (6/7) 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 21:17:22,717 INFO [finetune.py:976] (6/7) Epoch 24, batch 5250, loss[loss=0.2078, simple_loss=0.2862, pruned_loss=0.06466, over 4917.00 frames. ], tot_loss[loss=0.1693, simple_loss=0.2417, pruned_loss=0.04847, over 949990.43 frames. ], batch size: 42, lr: 3.03e-03, grad_scale: 32.0 2023-04-27 21:17:25,252 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=136991.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:17:36,059 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 24, batch 5300, loss[loss=0.2016, simple_loss=0.2826, pruned_loss=0.06032, over 4798.00 frames. ], tot_loss[loss=0.1711, simple_loss=0.2441, pruned_loss=0.04903, over 952725.04 frames. ], batch size: 45, lr: 3.03e-03, grad_scale: 32.0 2023-04-27 21:18:11,868 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=137039.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:18:30,932 INFO [optim.py:369] (6/7) 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 21:18:44,061 INFO [finetune.py:976] (6/7) Epoch 24, batch 5350, loss[loss=0.1341, simple_loss=0.2067, pruned_loss=0.03074, over 4665.00 frames. ], tot_loss[loss=0.1717, simple_loss=0.2447, pruned_loss=0.04928, over 954751.41 frames. ], batch size: 23, lr: 3.03e-03, grad_scale: 32.0 2023-04-27 21:19:04,286 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.2171, 2.5283, 1.1288, 1.5202, 2.0062, 1.3168, 3.3904, 1.7861], device='cuda:6'), covar=tensor([0.0621, 0.0590, 0.0709, 0.1248, 0.0453, 0.0955, 0.0286, 0.0558], device='cuda:6'), in_proj_covar=tensor([0.0050, 0.0064, 0.0047, 0.0046, 0.0048, 0.0051, 0.0072, 0.0050], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:6') 2023-04-27 21:19:09,856 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.48 vs. limit=2.0 2023-04-27 21:19:14,239 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.83 vs. limit=5.0 2023-04-27 21:19:16,913 INFO [finetune.py:976] (6/7) Epoch 24, batch 5400, loss[loss=0.1945, simple_loss=0.2538, pruned_loss=0.06756, over 4275.00 frames. ], tot_loss[loss=0.1702, simple_loss=0.243, pruned_loss=0.04868, over 956253.55 frames. ], batch size: 65, lr: 3.03e-03, grad_scale: 32.0 2023-04-27 21:19:33,099 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=137162.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:19:35,491 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.6037, 1.7789, 1.8607, 1.9678, 1.8006, 1.9197, 1.9400, 1.9623], device='cuda:6'), covar=tensor([0.3686, 0.5820, 0.5047, 0.4842, 0.6012, 0.7680, 0.5571, 0.5210], device='cuda:6'), in_proj_covar=tensor([0.0339, 0.0371, 0.0324, 0.0339, 0.0347, 0.0392, 0.0357, 0.0331], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 21:19:36,507 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7918, 1.8478, 1.1861, 1.5916, 1.9359, 1.6340, 1.5983, 1.6601], device='cuda:6'), covar=tensor([0.0406, 0.0298, 0.0328, 0.0440, 0.0281, 0.0396, 0.0404, 0.0464], device='cuda:6'), in_proj_covar=tensor([0.0028, 0.0023, 0.0022, 0.0028, 0.0019, 0.0028, 0.0027, 0.0028], device='cuda:6'), out_proj_covar=tensor([0.0051, 0.0046, 0.0038, 0.0052, 0.0039, 0.0050, 0.0050, 0.0052], device='cuda:6') 2023-04-27 21:19:37,583 INFO [optim.py:369] (6/7) 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 21:19:41,106 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-04-27 21:19:50,717 INFO [finetune.py:976] (6/7) Epoch 24, batch 5450, loss[loss=0.1736, simple_loss=0.2447, pruned_loss=0.0513, over 4861.00 frames. ], tot_loss[loss=0.1679, simple_loss=0.2401, pruned_loss=0.0479, over 954783.31 frames. ], batch size: 44, lr: 3.03e-03, grad_scale: 32.0 2023-04-27 21:20:24,724 INFO [finetune.py:976] (6/7) Epoch 24, batch 5500, loss[loss=0.1851, simple_loss=0.2599, pruned_loss=0.05513, over 4905.00 frames. ], tot_loss[loss=0.1655, simple_loss=0.2373, pruned_loss=0.04685, over 954110.53 frames. ], batch size: 37, lr: 3.03e-03, grad_scale: 32.0 2023-04-27 21:20:44,122 INFO [optim.py:369] (6/7) 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 21:20:57,636 INFO [finetune.py:976] (6/7) Epoch 24, batch 5550, loss[loss=0.1446, simple_loss=0.2233, pruned_loss=0.0329, over 4747.00 frames. ], tot_loss[loss=0.1667, simple_loss=0.2385, pruned_loss=0.04747, over 953463.81 frames. ], batch size: 23, lr: 3.03e-03, grad_scale: 32.0 2023-04-27 21:21:05,065 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=137299.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:21:34,763 INFO [finetune.py:976] (6/7) Epoch 24, batch 5600, loss[loss=0.1593, simple_loss=0.2423, pruned_loss=0.0381, over 4868.00 frames. ], tot_loss[loss=0.1692, simple_loss=0.2417, pruned_loss=0.04834, over 953178.14 frames. ], batch size: 34, lr: 3.03e-03, grad_scale: 32.0 2023-04-27 21:21:46,039 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=137347.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:22:14,764 INFO [optim.py:369] (6/7) 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 21:22:37,497 INFO [finetune.py:976] (6/7) Epoch 24, batch 5650, loss[loss=0.1849, simple_loss=0.2555, pruned_loss=0.05713, over 4932.00 frames. ], tot_loss[loss=0.1703, simple_loss=0.2436, pruned_loss=0.04852, over 952612.76 frames. ], batch size: 33, lr: 3.03e-03, grad_scale: 32.0 2023-04-27 21:23:22,991 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([0.7693, 0.8650, 1.0616, 1.0064, 0.8699, 0.7154, 0.8431, 0.4725], device='cuda:6'), covar=tensor([0.0499, 0.0507, 0.0459, 0.0426, 0.0524, 0.1025, 0.0402, 0.0575], device='cuda:6'), in_proj_covar=tensor([0.0070, 0.0068, 0.0066, 0.0068, 0.0075, 0.0095, 0.0073, 0.0064], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-27 21:23:26,501 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.5860, 2.5831, 1.9458, 2.9682, 2.6071, 2.6155, 1.2479, 2.6189], device='cuda:6'), covar=tensor([0.1268, 0.0779, 0.2229, 0.1410, 0.2006, 0.1400, 0.3630, 0.1524], device='cuda:6'), in_proj_covar=tensor([0.0245, 0.0220, 0.0253, 0.0307, 0.0298, 0.0248, 0.0277, 0.0273], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 21:23:28,790 INFO [finetune.py:976] (6/7) Epoch 24, batch 5700, loss[loss=0.1223, simple_loss=0.1862, pruned_loss=0.02918, over 3944.00 frames. ], tot_loss[loss=0.1687, simple_loss=0.2412, pruned_loss=0.04807, over 937240.79 frames. ], batch size: 17, lr: 3.03e-03, grad_scale: 32.0 2023-04-27 21:23:43,833 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=137462.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:23:57,747 INFO [finetune.py:976] (6/7) Epoch 25, batch 0, loss[loss=0.166, simple_loss=0.2499, pruned_loss=0.04106, over 4833.00 frames. ], tot_loss[loss=0.166, simple_loss=0.2499, pruned_loss=0.04106, over 4833.00 frames. ], batch size: 47, lr: 3.03e-03, grad_scale: 32.0 2023-04-27 21:23:57,747 INFO [finetune.py:1001] (6/7) Computing validation loss 2023-04-27 21:24:03,969 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([3.2366, 3.2913, 2.5876, 3.6960, 3.2958, 3.2709, 1.7161, 3.3243], device='cuda:6'), covar=tensor([0.1699, 0.1343, 0.3023, 0.2228, 0.2410, 0.1789, 0.4547, 0.2064], device='cuda:6'), in_proj_covar=tensor([0.0245, 0.0220, 0.0253, 0.0308, 0.0299, 0.0248, 0.0277, 0.0273], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 21:24:08,111 INFO [finetune.py:1010] (6/7) 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] (6/7) Maximum memory allocated so far is 6435MB 2023-04-27 21:24:09,920 INFO [optim.py:369] (6/7) 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 21:24:37,396 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=137510.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:24:40,448 INFO [finetune.py:976] (6/7) Epoch 25, batch 50, loss[loss=0.1787, simple_loss=0.234, pruned_loss=0.06172, over 4878.00 frames. ], tot_loss[loss=0.1648, simple_loss=0.2398, pruned_loss=0.04493, over 216682.12 frames. ], batch size: 35, lr: 3.02e-03, grad_scale: 32.0 2023-04-27 21:24:42,248 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7623, 1.6970, 0.7353, 1.4647, 1.7051, 1.5882, 1.5238, 1.5872], device='cuda:6'), covar=tensor([0.0468, 0.0357, 0.0343, 0.0521, 0.0261, 0.0485, 0.0486, 0.0510], device='cuda:6'), in_proj_covar=tensor([0.0028, 0.0023, 0.0022, 0.0028, 0.0019, 0.0028, 0.0027, 0.0028], device='cuda:6'), out_proj_covar=tensor([0.0051, 0.0046, 0.0038, 0.0052, 0.0038, 0.0050, 0.0050, 0.0052], device='cuda:6') 2023-04-27 21:24:44,296 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.23 vs. limit=5.0 2023-04-27 21:24:55,461 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.2851, 1.4093, 1.3134, 1.7110, 1.4577, 1.7308, 1.2798, 3.4507], device='cuda:6'), covar=tensor([0.0698, 0.1080, 0.1049, 0.1317, 0.0858, 0.0772, 0.0988, 0.0240], device='cuda:6'), in_proj_covar=tensor([0.0037, 0.0038, 0.0039, 0.0042, 0.0040, 0.0038, 0.0037, 0.0055], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:6') 2023-04-27 21:24:57,771 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5712, 3.9444, 0.8905, 2.2429, 2.2129, 2.8032, 2.3391, 1.0747], device='cuda:6'), covar=tensor([0.1585, 0.1412, 0.2313, 0.1365, 0.1139, 0.1138, 0.1550, 0.2339], device='cuda:6'), in_proj_covar=tensor([0.0116, 0.0236, 0.0135, 0.0119, 0.0131, 0.0151, 0.0117, 0.0117], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-27 21:24:59,601 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8873, 1.4848, 1.5174, 1.7100, 2.0209, 1.7761, 1.5610, 1.3933], device='cuda:6'), covar=tensor([0.1668, 0.1673, 0.1930, 0.1391, 0.1015, 0.1521, 0.1942, 0.2626], device='cuda:6'), in_proj_covar=tensor([0.0310, 0.0304, 0.0348, 0.0282, 0.0324, 0.0302, 0.0296, 0.0369], device='cuda:6'), 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:6') 2023-04-27 21:25:13,440 INFO [finetune.py:976] (6/7) Epoch 25, batch 100, loss[loss=0.195, simple_loss=0.2541, pruned_loss=0.06792, over 4905.00 frames. ], tot_loss[loss=0.1637, simple_loss=0.236, pruned_loss=0.04565, over 382134.14 frames. ], batch size: 46, lr: 3.02e-03, grad_scale: 32.0 2023-04-27 21:25:15,239 INFO [optim.py:369] (6/7) 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 21:25:46,403 INFO [finetune.py:976] (6/7) Epoch 25, batch 150, loss[loss=0.1232, simple_loss=0.204, pruned_loss=0.02119, over 4751.00 frames. ], tot_loss[loss=0.1616, simple_loss=0.2325, pruned_loss=0.04529, over 509290.24 frames. ], batch size: 28, lr: 3.02e-03, grad_scale: 32.0 2023-04-27 21:25:55,296 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-04-27 21:26:20,275 INFO [finetune.py:976] (6/7) Epoch 25, batch 200, loss[loss=0.1765, simple_loss=0.2497, pruned_loss=0.05169, over 4903.00 frames. ], tot_loss[loss=0.1617, simple_loss=0.2325, pruned_loss=0.0454, over 607792.58 frames. ], batch size: 43, lr: 3.02e-03, grad_scale: 32.0 2023-04-27 21:26:22,073 INFO [optim.py:369] (6/7) 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 21:26:33,166 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.11 vs. limit=2.0 2023-04-27 21:27:09,385 INFO [finetune.py:976] (6/7) Epoch 25, batch 250, loss[loss=0.1609, simple_loss=0.2457, pruned_loss=0.03809, over 4753.00 frames. ], tot_loss[loss=0.1676, simple_loss=0.2391, pruned_loss=0.04799, over 684678.96 frames. ], batch size: 54, lr: 3.02e-03, grad_scale: 32.0 2023-04-27 21:27:19,176 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([3.4326, 2.9396, 2.6230, 2.7799, 2.5672, 2.7988, 2.7428, 2.6545], device='cuda:6'), covar=tensor([0.2959, 0.4828, 0.4354, 0.4023, 0.5101, 0.5862, 0.5416, 0.4921], device='cuda:6'), in_proj_covar=tensor([0.0339, 0.0372, 0.0324, 0.0338, 0.0348, 0.0392, 0.0356, 0.0329], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 21:28:04,802 INFO [finetune.py:976] (6/7) Epoch 25, batch 300, loss[loss=0.1784, simple_loss=0.2529, pruned_loss=0.05195, over 4735.00 frames. ], tot_loss[loss=0.1697, simple_loss=0.2425, pruned_loss=0.04848, over 745823.95 frames. ], batch size: 59, lr: 3.02e-03, grad_scale: 32.0 2023-04-27 21:28:05,564 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0201, 1.7829, 2.2570, 2.5063, 2.0824, 1.9449, 2.0863, 2.0280], device='cuda:6'), covar=tensor([0.4387, 0.7300, 0.6751, 0.5375, 0.6221, 0.8786, 0.8522, 1.0379], device='cuda:6'), in_proj_covar=tensor([0.0436, 0.0420, 0.0512, 0.0508, 0.0466, 0.0500, 0.0503, 0.0516], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 21:28:06,622 INFO [optim.py:369] (6/7) 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 21:28:30,275 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.69 vs. limit=2.0 2023-04-27 21:28:43,572 INFO [finetune.py:976] (6/7) Epoch 25, batch 350, loss[loss=0.2022, simple_loss=0.2791, pruned_loss=0.06264, over 4866.00 frames. ], tot_loss[loss=0.1722, simple_loss=0.2454, pruned_loss=0.04953, over 792869.04 frames. ], batch size: 34, lr: 3.02e-03, grad_scale: 32.0 2023-04-27 21:28:51,459 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([3.6905, 3.5468, 2.6159, 4.2053, 3.6528, 3.6136, 1.6512, 3.6826], device='cuda:6'), covar=tensor([0.1692, 0.1481, 0.3253, 0.1993, 0.3279, 0.1914, 0.5922, 0.2773], device='cuda:6'), in_proj_covar=tensor([0.0244, 0.0221, 0.0254, 0.0308, 0.0300, 0.0249, 0.0276, 0.0275], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 21:29:01,932 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-04-27 21:29:21,925 INFO [finetune.py:976] (6/7) Epoch 25, batch 400, loss[loss=0.167, simple_loss=0.2444, pruned_loss=0.04475, over 4912.00 frames. ], tot_loss[loss=0.1709, simple_loss=0.2443, pruned_loss=0.04873, over 827600.16 frames. ], batch size: 37, lr: 3.02e-03, grad_scale: 32.0 2023-04-27 21:29:29,042 INFO [optim.py:369] (6/7) 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 21:30:01,141 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=137892.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:30:02,955 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=137895.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:30:12,569 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.2738, 1.6997, 1.6703, 2.0398, 1.9760, 2.1680, 1.6203, 4.3647], device='cuda:6'), covar=tensor([0.0542, 0.0757, 0.0739, 0.1113, 0.0601, 0.0495, 0.0683, 0.0110], device='cuda:6'), in_proj_covar=tensor([0.0038, 0.0038, 0.0039, 0.0043, 0.0040, 0.0038, 0.0038, 0.0055], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0013, 0.0013, 0.0015, 0.0014, 0.0013, 0.0013, 0.0016], device='cuda:6') 2023-04-27 21:30:27,175 INFO [finetune.py:976] (6/7) Epoch 25, batch 450, loss[loss=0.181, simple_loss=0.2527, pruned_loss=0.05466, over 4925.00 frames. ], tot_loss[loss=0.1701, simple_loss=0.2432, pruned_loss=0.04847, over 855641.75 frames. ], batch size: 38, lr: 3.02e-03, grad_scale: 32.0 2023-04-27 21:31:25,543 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-04-27 21:31:27,307 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=137953.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:31:33,998 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 25, batch 500, loss[loss=0.1748, simple_loss=0.2537, pruned_loss=0.04794, over 4900.00 frames. ], tot_loss[loss=0.1685, simple_loss=0.2412, pruned_loss=0.04792, over 879529.97 frames. ], batch size: 32, lr: 3.02e-03, grad_scale: 32.0 2023-04-27 21:31:46,603 INFO [optim.py:369] (6/7) 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 21:32:13,182 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.3313, 2.9957, 1.1121, 1.6592, 2.2672, 1.3360, 3.8268, 1.7130], device='cuda:6'), covar=tensor([0.0665, 0.0719, 0.0817, 0.1214, 0.0469, 0.0965, 0.0206, 0.0658], device='cuda:6'), in_proj_covar=tensor([0.0050, 0.0064, 0.0047, 0.0046, 0.0048, 0.0051, 0.0072, 0.0050], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0007, 0.0007, 0.0008, 0.0010, 0.0008], device='cuda:6') 2023-04-27 21:32:30,092 INFO [finetune.py:976] (6/7) Epoch 25, batch 550, loss[loss=0.1564, simple_loss=0.2277, pruned_loss=0.0425, over 4779.00 frames. ], tot_loss[loss=0.166, simple_loss=0.2378, pruned_loss=0.04712, over 896364.68 frames. ], batch size: 28, lr: 3.02e-03, grad_scale: 32.0 2023-04-27 21:33:09,158 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0803, 2.0054, 2.5744, 2.6893, 1.9619, 1.7051, 2.0624, 1.2061], device='cuda:6'), covar=tensor([0.0503, 0.0568, 0.0347, 0.0444, 0.0560, 0.1045, 0.0513, 0.0629], device='cuda:6'), in_proj_covar=tensor([0.0070, 0.0067, 0.0066, 0.0068, 0.0074, 0.0095, 0.0073, 0.0063], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-27 21:33:11,281 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-04-27 21:33:20,381 INFO [finetune.py:976] (6/7) Epoch 25, batch 600, loss[loss=0.2126, simple_loss=0.2851, pruned_loss=0.07009, over 4764.00 frames. ], tot_loss[loss=0.1676, simple_loss=0.2396, pruned_loss=0.04782, over 912270.15 frames. ], batch size: 54, lr: 3.02e-03, grad_scale: 32.0 2023-04-27 21:33:22,209 INFO [optim.py:369] (6/7) 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 21:33:31,195 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.4572, 1.6477, 1.8659, 1.9926, 1.8940, 1.9662, 1.9354, 1.8867], device='cuda:6'), covar=tensor([0.3411, 0.5320, 0.4407, 0.4330, 0.5235, 0.6718, 0.4828, 0.4634], device='cuda:6'), in_proj_covar=tensor([0.0341, 0.0374, 0.0326, 0.0339, 0.0349, 0.0394, 0.0357, 0.0331], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 21:33:31,764 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.3981, 1.3373, 1.7612, 1.7016, 1.3275, 1.2112, 1.3641, 0.9186], device='cuda:6'), covar=tensor([0.0467, 0.0557, 0.0350, 0.0441, 0.0620, 0.0929, 0.0482, 0.0472], device='cuda:6'), in_proj_covar=tensor([0.0070, 0.0068, 0.0066, 0.0068, 0.0075, 0.0096, 0.0073, 0.0064], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-27 21:33:49,671 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6124, 1.4110, 0.7493, 1.2836, 1.4262, 1.4890, 1.3569, 1.3668], device='cuda:6'), covar=tensor([0.0499, 0.0404, 0.0369, 0.0571, 0.0302, 0.0537, 0.0538, 0.0575], device='cuda:6'), in_proj_covar=tensor([0.0028, 0.0023, 0.0022, 0.0028, 0.0019, 0.0028, 0.0027, 0.0028], device='cuda:6'), out_proj_covar=tensor([0.0051, 0.0046, 0.0038, 0.0052, 0.0039, 0.0050, 0.0050, 0.0052], device='cuda:6') 2023-04-27 21:33:53,154 INFO [finetune.py:976] (6/7) Epoch 25, batch 650, loss[loss=0.1605, simple_loss=0.2486, pruned_loss=0.03622, over 4917.00 frames. ], tot_loss[loss=0.1694, simple_loss=0.2423, pruned_loss=0.04826, over 922196.67 frames. ], batch size: 38, lr: 3.02e-03, grad_scale: 32.0 2023-04-27 21:34:16,273 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6961, 1.4615, 4.4789, 4.2222, 3.9480, 4.2877, 4.1840, 3.9337], device='cuda:6'), covar=tensor([0.7004, 0.6206, 0.1047, 0.1590, 0.1115, 0.1689, 0.1244, 0.1643], device='cuda:6'), in_proj_covar=tensor([0.0310, 0.0307, 0.0406, 0.0408, 0.0348, 0.0409, 0.0319, 0.0366], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-27 21:34:26,499 INFO [finetune.py:976] (6/7) Epoch 25, batch 700, loss[loss=0.2161, simple_loss=0.2904, pruned_loss=0.07091, over 4807.00 frames. ], tot_loss[loss=0.1699, simple_loss=0.2433, pruned_loss=0.04825, over 930073.22 frames. ], batch size: 40, lr: 3.02e-03, grad_scale: 32.0 2023-04-27 21:34:28,312 INFO [optim.py:369] (6/7) 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 21:34:53,522 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([0.9313, 1.2544, 3.3103, 3.0604, 2.9512, 3.2518, 3.2432, 2.9114], device='cuda:6'), covar=tensor([0.7654, 0.5531, 0.1641, 0.2388, 0.1460, 0.2218, 0.1453, 0.1763], device='cuda:6'), in_proj_covar=tensor([0.0312, 0.0309, 0.0409, 0.0411, 0.0350, 0.0412, 0.0321, 0.0368], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-27 21:35:26,625 INFO [finetune.py:976] (6/7) Epoch 25, batch 750, loss[loss=0.1431, simple_loss=0.2308, pruned_loss=0.02765, over 4818.00 frames. ], tot_loss[loss=0.1707, simple_loss=0.2443, pruned_loss=0.0486, over 934154.77 frames. ], batch size: 39, lr: 3.02e-03, grad_scale: 32.0 2023-04-27 21:36:09,244 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=138248.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:36:16,844 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 25, batch 800, loss[loss=0.1681, simple_loss=0.2422, pruned_loss=0.04702, over 4789.00 frames. ], tot_loss[loss=0.1698, simple_loss=0.2434, pruned_loss=0.04806, over 939264.17 frames. ], batch size: 29, lr: 3.02e-03, grad_scale: 32.0 2023-04-27 21:36:27,288 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2023-04-27 21:36:27,566 INFO [optim.py:369] (6/7) 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 21:36:28,898 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=138270.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:37:09,966 INFO [finetune.py:976] (6/7) Epoch 25, batch 850, loss[loss=0.1419, simple_loss=0.2137, pruned_loss=0.03503, over 4911.00 frames. ], tot_loss[loss=0.1691, simple_loss=0.2419, pruned_loss=0.0482, over 941763.65 frames. ], batch size: 36, lr: 3.02e-03, grad_scale: 32.0 2023-04-27 21:37:14,322 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=138322.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:37:25,794 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 25, batch 900, loss[loss=0.1701, simple_loss=0.2305, pruned_loss=0.05485, over 4803.00 frames. ], tot_loss[loss=0.1655, simple_loss=0.2381, pruned_loss=0.04645, over 947240.27 frames. ], batch size: 45, lr: 3.02e-03, grad_scale: 32.0 2023-04-27 21:38:02,475 INFO [optim.py:369] (6/7) 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 21:38:10,248 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=3.07 vs. limit=5.0 2023-04-27 21:38:22,010 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=138383.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:38:44,708 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.2354, 1.8341, 2.1196, 2.2648, 2.1282, 1.7901, 1.2631, 1.8539], device='cuda:6'), covar=tensor([0.3121, 0.2957, 0.1679, 0.2232, 0.2536, 0.2544, 0.3893, 0.1912], device='cuda:6'), in_proj_covar=tensor([0.0294, 0.0246, 0.0228, 0.0314, 0.0222, 0.0234, 0.0228, 0.0186], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:6') 2023-04-27 21:39:01,963 INFO [finetune.py:976] (6/7) Epoch 25, batch 950, loss[loss=0.1834, simple_loss=0.2407, pruned_loss=0.06302, over 4768.00 frames. ], tot_loss[loss=0.1646, simple_loss=0.2369, pruned_loss=0.04619, over 949409.18 frames. ], batch size: 26, lr: 3.02e-03, grad_scale: 32.0 2023-04-27 21:40:06,147 INFO [finetune.py:976] (6/7) Epoch 25, batch 1000, loss[loss=0.1359, simple_loss=0.2142, pruned_loss=0.02877, over 4769.00 frames. ], tot_loss[loss=0.1673, simple_loss=0.2401, pruned_loss=0.04725, over 952801.52 frames. ], batch size: 28, lr: 3.02e-03, grad_scale: 32.0 2023-04-27 21:40:07,976 INFO [optim.py:369] (6/7) 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 21:41:00,027 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.1650, 1.4124, 1.2629, 1.6247, 1.5287, 1.8266, 1.3241, 3.3143], device='cuda:6'), covar=tensor([0.0635, 0.0816, 0.0856, 0.1284, 0.0667, 0.0510, 0.0774, 0.0147], device='cuda:6'), in_proj_covar=tensor([0.0037, 0.0039, 0.0039, 0.0043, 0.0040, 0.0038, 0.0038, 0.0055], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0013, 0.0013, 0.0015, 0.0014, 0.0013, 0.0013, 0.0016], device='cuda:6') 2023-04-27 21:41:09,819 INFO [finetune.py:976] (6/7) Epoch 25, batch 1050, loss[loss=0.1662, simple_loss=0.2512, pruned_loss=0.04058, over 4806.00 frames. ], tot_loss[loss=0.1694, simple_loss=0.2431, pruned_loss=0.04785, over 953166.52 frames. ], batch size: 41, lr: 3.02e-03, grad_scale: 32.0 2023-04-27 21:41:19,912 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.14 vs. limit=2.0 2023-04-27 21:41:30,051 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.4595, 1.3181, 0.4656, 1.1832, 1.3456, 1.3215, 1.2382, 1.3071], device='cuda:6'), covar=tensor([0.0512, 0.0396, 0.0412, 0.0578, 0.0310, 0.0514, 0.0514, 0.0578], device='cuda:6'), in_proj_covar=tensor([0.0028, 0.0023, 0.0022, 0.0028, 0.0019, 0.0028, 0.0027, 0.0028], device='cuda:6'), out_proj_covar=tensor([0.0052, 0.0046, 0.0039, 0.0053, 0.0039, 0.0050, 0.0050, 0.0052], device='cuda:6') 2023-04-27 21:41:52,770 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=138548.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:41:54,997 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=138551.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:42:14,666 INFO [finetune.py:976] (6/7) Epoch 25, batch 1100, loss[loss=0.1281, simple_loss=0.2069, pruned_loss=0.02467, over 4781.00 frames. ], tot_loss[loss=0.1704, simple_loss=0.2444, pruned_loss=0.04823, over 954268.13 frames. ], batch size: 29, lr: 3.02e-03, grad_scale: 32.0 2023-04-27 21:42:16,468 INFO [optim.py:369] (6/7) 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 21:42:55,897 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=138596.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:42:58,181 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=138599.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:43:18,593 INFO [finetune.py:976] (6/7) Epoch 25, batch 1150, loss[loss=0.2217, simple_loss=0.2748, pruned_loss=0.08434, over 4844.00 frames. ], tot_loss[loss=0.1708, simple_loss=0.2448, pruned_loss=0.04836, over 955244.94 frames. ], batch size: 49, lr: 3.02e-03, grad_scale: 32.0 2023-04-27 21:43:38,295 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=138626.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:43:52,179 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-04-27 21:43:53,133 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9099, 1.4131, 1.7472, 1.7038, 1.7297, 1.3865, 0.7784, 1.4128], device='cuda:6'), covar=tensor([0.3113, 0.2978, 0.1721, 0.2056, 0.2341, 0.2516, 0.4015, 0.1911], device='cuda:6'), in_proj_covar=tensor([0.0293, 0.0244, 0.0227, 0.0312, 0.0221, 0.0233, 0.0227, 0.0185], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:6') 2023-04-27 21:44:02,587 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.51 vs. limit=2.0 2023-04-27 21:44:23,368 INFO [finetune.py:976] (6/7) Epoch 25, batch 1200, loss[loss=0.1665, simple_loss=0.241, pruned_loss=0.04597, over 4912.00 frames. ], tot_loss[loss=0.1685, simple_loss=0.2426, pruned_loss=0.04725, over 956440.06 frames. ], batch size: 37, lr: 3.02e-03, grad_scale: 64.0 2023-04-27 21:44:26,077 INFO [optim.py:369] (6/7) 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 21:44:33,928 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0837, 1.9055, 2.1755, 2.4457, 2.1028, 1.9492, 2.1037, 2.0691], device='cuda:6'), covar=tensor([0.4647, 0.7106, 0.7103, 0.5717, 0.5960, 0.8836, 0.9235, 1.0369], device='cuda:6'), in_proj_covar=tensor([0.0436, 0.0420, 0.0511, 0.0506, 0.0465, 0.0500, 0.0501, 0.0515], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 21:44:43,181 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=138678.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:44:44,987 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=138681.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:44:53,113 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=138686.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:45:05,135 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=138697.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:45:29,242 INFO [finetune.py:976] (6/7) Epoch 25, batch 1250, loss[loss=0.1869, simple_loss=0.2579, pruned_loss=0.05792, over 4767.00 frames. ], tot_loss[loss=0.1686, simple_loss=0.2417, pruned_loss=0.04774, over 955549.60 frames. ], batch size: 26, lr: 3.02e-03, grad_scale: 64.0 2023-04-27 21:45:46,785 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9853, 2.5065, 0.9816, 1.2113, 1.7106, 1.0419, 3.1203, 1.4244], device='cuda:6'), covar=tensor([0.0867, 0.0704, 0.0990, 0.1740, 0.0715, 0.1498, 0.0401, 0.1012], device='cuda:6'), in_proj_covar=tensor([0.0050, 0.0064, 0.0047, 0.0046, 0.0049, 0.0051, 0.0072, 0.0050], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0007, 0.0008, 0.0008, 0.0010, 0.0007], device='cuda:6') 2023-04-27 21:46:07,980 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=138742.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:46:11,062 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=138747.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:46:28,620 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 25, batch 1300, loss[loss=0.1366, simple_loss=0.2151, pruned_loss=0.02903, over 4826.00 frames. ], tot_loss[loss=0.166, simple_loss=0.2384, pruned_loss=0.0468, over 954145.49 frames. ], batch size: 39, lr: 3.02e-03, grad_scale: 64.0 2023-04-27 21:46:40,293 INFO [optim.py:369] (6/7) 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 21:47:44,175 INFO [finetune.py:976] (6/7) Epoch 25, batch 1350, loss[loss=0.1514, simple_loss=0.2249, pruned_loss=0.03899, over 4900.00 frames. ], tot_loss[loss=0.1666, simple_loss=0.2385, pruned_loss=0.04732, over 954485.90 frames. ], batch size: 36, lr: 3.02e-03, grad_scale: 64.0 2023-04-27 21:48:48,891 INFO [finetune.py:976] (6/7) Epoch 25, batch 1400, loss[loss=0.2358, simple_loss=0.2948, pruned_loss=0.08836, over 4136.00 frames. ], tot_loss[loss=0.1689, simple_loss=0.2415, pruned_loss=0.04814, over 952979.44 frames. ], batch size: 65, lr: 3.02e-03, grad_scale: 64.0 2023-04-27 21:48:50,720 INFO [optim.py:369] (6/7) 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 21:48:51,508 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5479, 1.6874, 0.5857, 1.2753, 1.9605, 1.3938, 1.3157, 1.5144], device='cuda:6'), covar=tensor([0.0508, 0.0387, 0.0361, 0.0563, 0.0256, 0.0515, 0.0530, 0.0558], device='cuda:6'), in_proj_covar=tensor([0.0028, 0.0023, 0.0022, 0.0028, 0.0019, 0.0028, 0.0027, 0.0028], device='cuda:6'), out_proj_covar=tensor([0.0051, 0.0046, 0.0038, 0.0053, 0.0039, 0.0050, 0.0050, 0.0052], device='cuda:6') 2023-04-27 21:49:14,464 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.6212, 1.9161, 1.9451, 2.0987, 1.9750, 1.9706, 2.0051, 2.0064], device='cuda:6'), covar=tensor([0.3907, 0.5836, 0.5051, 0.4793, 0.5724, 0.7826, 0.6324, 0.5472], device='cuda:6'), in_proj_covar=tensor([0.0341, 0.0375, 0.0328, 0.0342, 0.0350, 0.0394, 0.0359, 0.0333], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 21:49:44,979 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 25, batch 1450, loss[loss=0.1607, simple_loss=0.2413, pruned_loss=0.04001, over 4927.00 frames. ], tot_loss[loss=0.1693, simple_loss=0.2425, pruned_loss=0.04807, over 952706.20 frames. ], batch size: 33, lr: 3.01e-03, grad_scale: 64.0 2023-04-27 21:50:04,880 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6947, 2.0027, 1.0051, 1.3422, 2.0816, 1.4744, 1.4381, 1.5404], device='cuda:6'), covar=tensor([0.0614, 0.0337, 0.0324, 0.0637, 0.0250, 0.0654, 0.0659, 0.0652], device='cuda:6'), in_proj_covar=tensor([0.0028, 0.0023, 0.0022, 0.0028, 0.0019, 0.0028, 0.0027, 0.0028], device='cuda:6'), out_proj_covar=tensor([0.0051, 0.0046, 0.0038, 0.0053, 0.0039, 0.0050, 0.0050, 0.0052], device='cuda:6') 2023-04-27 21:50:07,893 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=138926.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:50:38,962 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.55 vs. limit=2.0 2023-04-27 21:50:53,069 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.1542, 1.4455, 1.3906, 1.6476, 1.5903, 1.6582, 1.3520, 3.0143], device='cuda:6'), covar=tensor([0.0630, 0.0776, 0.0748, 0.1224, 0.0605, 0.0510, 0.0716, 0.0179], device='cuda:6'), in_proj_covar=tensor([0.0038, 0.0039, 0.0039, 0.0043, 0.0040, 0.0038, 0.0038, 0.0055], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0013, 0.0013, 0.0015, 0.0014, 0.0013, 0.0013, 0.0016], device='cuda:6') 2023-04-27 21:51:01,102 INFO [finetune.py:976] (6/7) Epoch 25, batch 1500, loss[loss=0.1674, simple_loss=0.23, pruned_loss=0.05243, over 4806.00 frames. ], tot_loss[loss=0.1693, simple_loss=0.2427, pruned_loss=0.04798, over 954118.38 frames. ], batch size: 25, lr: 3.01e-03, grad_scale: 64.0 2023-04-27 21:51:03,879 INFO [optim.py:369] (6/7) 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 21:51:10,078 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=138970.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 21:51:12,536 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=138974.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:51:19,558 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=138978.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:51:21,392 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=3.64 vs. limit=5.0 2023-04-27 21:52:02,852 INFO [finetune.py:976] (6/7) Epoch 25, batch 1550, loss[loss=0.1715, simple_loss=0.2459, pruned_loss=0.04857, over 4811.00 frames. ], tot_loss[loss=0.1683, simple_loss=0.242, pruned_loss=0.04733, over 954150.15 frames. ], batch size: 39, lr: 3.01e-03, grad_scale: 64.0 2023-04-27 21:52:11,164 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=139026.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:52:11,849 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.4060, 0.9557, 0.5155, 1.1444, 1.1047, 1.2896, 1.2299, 1.2536], device='cuda:6'), covar=tensor([0.0435, 0.0342, 0.0351, 0.0488, 0.0277, 0.0441, 0.0432, 0.0462], device='cuda:6'), in_proj_covar=tensor([0.0028, 0.0023, 0.0022, 0.0028, 0.0019, 0.0028, 0.0027, 0.0028], device='cuda:6'), out_proj_covar=tensor([0.0051, 0.0046, 0.0038, 0.0053, 0.0039, 0.0050, 0.0050, 0.0052], device='cuda:6') 2023-04-27 21:52:17,674 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.4685, 3.0347, 0.8940, 1.6488, 1.8072, 2.2033, 1.8056, 1.0727], device='cuda:6'), covar=tensor([0.1388, 0.1223, 0.2097, 0.1389, 0.1081, 0.1127, 0.1726, 0.1788], device='cuda:6'), in_proj_covar=tensor([0.0117, 0.0239, 0.0137, 0.0121, 0.0132, 0.0153, 0.0117, 0.0120], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-27 21:52:18,855 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=139037.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:52:22,337 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=139042.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:52:23,942 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-04-27 21:52:29,084 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=139053.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:52:41,929 INFO [finetune.py:976] (6/7) Epoch 25, batch 1600, loss[loss=0.1828, simple_loss=0.2493, pruned_loss=0.05813, over 4872.00 frames. ], tot_loss[loss=0.1677, simple_loss=0.2404, pruned_loss=0.04746, over 954804.07 frames. ], batch size: 34, lr: 3.01e-03, grad_scale: 64.0 2023-04-27 21:52:44,234 INFO [optim.py:369] (6/7) 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 21:53:24,802 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.2442, 1.7865, 2.0951, 2.4440, 2.1469, 1.6801, 1.2984, 1.8988], device='cuda:6'), covar=tensor([0.2829, 0.2669, 0.1514, 0.2002, 0.2358, 0.2482, 0.3931, 0.1735], device='cuda:6'), in_proj_covar=tensor([0.0293, 0.0244, 0.0227, 0.0311, 0.0221, 0.0234, 0.0226, 0.0185], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:6') 2023-04-27 21:53:47,320 INFO [finetune.py:976] (6/7) Epoch 25, batch 1650, loss[loss=0.1381, simple_loss=0.2067, pruned_loss=0.03479, over 4788.00 frames. ], tot_loss[loss=0.166, simple_loss=0.2383, pruned_loss=0.04691, over 954180.56 frames. ], batch size: 29, lr: 3.01e-03, grad_scale: 64.0 2023-04-27 21:54:33,487 INFO [finetune.py:976] (6/7) Epoch 25, batch 1700, loss[loss=0.1828, simple_loss=0.2481, pruned_loss=0.0588, over 4912.00 frames. ], tot_loss[loss=0.1653, simple_loss=0.2373, pruned_loss=0.04671, over 953700.86 frames. ], batch size: 37, lr: 3.01e-03, grad_scale: 64.0 2023-04-27 21:54:35,334 INFO [optim.py:369] (6/7) 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 21:54:40,454 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-04-27 21:55:07,049 INFO [finetune.py:976] (6/7) Epoch 25, batch 1750, loss[loss=0.1697, simple_loss=0.2468, pruned_loss=0.04636, over 4925.00 frames. ], tot_loss[loss=0.1681, simple_loss=0.2401, pruned_loss=0.04806, over 953703.30 frames. ], batch size: 33, lr: 3.01e-03, grad_scale: 64.0 2023-04-27 21:55:23,348 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=139238.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:55:29,612 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-04-27 21:55:41,182 INFO [finetune.py:976] (6/7) Epoch 25, batch 1800, loss[loss=0.1708, simple_loss=0.2451, pruned_loss=0.04821, over 4910.00 frames. ], tot_loss[loss=0.1696, simple_loss=0.2425, pruned_loss=0.04838, over 952767.46 frames. ], batch size: 37, lr: 3.01e-03, grad_scale: 64.0 2023-04-27 21:55:41,252 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=139265.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 21:55:42,981 INFO [optim.py:369] (6/7) 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] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=139276.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:55:57,486 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=139289.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:56:04,607 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=139299.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:56:07,553 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.4831, 2.9213, 0.9370, 1.7011, 2.0868, 1.5942, 3.8076, 1.9642], device='cuda:6'), covar=tensor([0.0587, 0.0844, 0.0933, 0.1193, 0.0491, 0.0869, 0.0183, 0.0542], device='cuda:6'), in_proj_covar=tensor([0.0050, 0.0064, 0.0047, 0.0046, 0.0049, 0.0051, 0.0072, 0.0050], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0007], device='cuda:6') 2023-04-27 21:56:14,567 INFO [finetune.py:976] (6/7) Epoch 25, batch 1850, loss[loss=0.1551, simple_loss=0.2277, pruned_loss=0.04126, over 4926.00 frames. ], tot_loss[loss=0.1694, simple_loss=0.2426, pruned_loss=0.04814, over 952969.71 frames. ], batch size: 38, lr: 3.01e-03, grad_scale: 64.0 2023-04-27 21:56:25,769 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5169, 1.4797, 3.6319, 3.3714, 3.2469, 3.3897, 3.3502, 3.2264], device='cuda:6'), covar=tensor([0.6934, 0.5167, 0.1190, 0.1694, 0.1145, 0.2014, 0.4082, 0.1489], device='cuda:6'), in_proj_covar=tensor([0.0314, 0.0309, 0.0408, 0.0412, 0.0351, 0.0414, 0.0321, 0.0370], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-27 21:56:29,730 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=139337.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:56:29,776 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=139337.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 21:56:33,191 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=139342.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:56:44,347 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=139350.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:56:51,966 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=139353.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:57:05,213 INFO [finetune.py:976] (6/7) Epoch 25, batch 1900, loss[loss=0.1636, simple_loss=0.2352, pruned_loss=0.046, over 4794.00 frames. ], tot_loss[loss=0.1697, simple_loss=0.2436, pruned_loss=0.04791, over 954818.26 frames. ], batch size: 25, lr: 3.01e-03, grad_scale: 32.0 2023-04-27 21:57:07,685 INFO [optim.py:369] (6/7) 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 21:57:13,411 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8214, 1.3451, 1.3880, 1.5989, 1.9385, 1.5545, 1.3372, 1.3527], device='cuda:6'), covar=tensor([0.1386, 0.1386, 0.1844, 0.1083, 0.0881, 0.1431, 0.1927, 0.2171], device='cuda:6'), in_proj_covar=tensor([0.0312, 0.0308, 0.0350, 0.0286, 0.0327, 0.0306, 0.0299, 0.0372], device='cuda:6'), out_proj_covar=tensor([6.3913e-05, 6.3401e-05, 7.3633e-05, 5.7281e-05, 6.7239e-05, 6.4070e-05, 6.2088e-05, 7.8941e-05], device='cuda:6') 2023-04-27 21:57:28,904 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=139385.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:57:36,788 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=139390.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:57:49,511 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=139401.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:57:56,055 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([4.0042, 3.9957, 2.9413, 4.6626, 3.9820, 4.0309, 1.7433, 4.0316], device='cuda:6'), covar=tensor([0.1767, 0.1125, 0.3108, 0.1438, 0.3148, 0.1794, 0.6074, 0.2312], device='cuda:6'), in_proj_covar=tensor([0.0242, 0.0217, 0.0249, 0.0303, 0.0295, 0.0246, 0.0272, 0.0271], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 21:58:09,751 INFO [finetune.py:976] (6/7) Epoch 25, batch 1950, loss[loss=0.1647, simple_loss=0.2343, pruned_loss=0.04753, over 4813.00 frames. ], tot_loss[loss=0.1699, simple_loss=0.2434, pruned_loss=0.04817, over 954542.21 frames. ], batch size: 25, lr: 3.01e-03, grad_scale: 32.0 2023-04-27 21:59:13,361 INFO [finetune.py:976] (6/7) Epoch 25, batch 2000, loss[loss=0.1376, simple_loss=0.2108, pruned_loss=0.0322, over 4821.00 frames. ], tot_loss[loss=0.1687, simple_loss=0.2416, pruned_loss=0.0479, over 954412.96 frames. ], batch size: 38, lr: 3.01e-03, grad_scale: 32.0 2023-04-27 21:59:13,469 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.1837, 1.5617, 1.4210, 1.7635, 1.7019, 1.7785, 1.4206, 3.0729], device='cuda:6'), covar=tensor([0.0608, 0.0741, 0.0727, 0.1090, 0.0564, 0.0470, 0.0683, 0.0160], device='cuda:6'), in_proj_covar=tensor([0.0038, 0.0039, 0.0039, 0.0043, 0.0040, 0.0038, 0.0038, 0.0055], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0013, 0.0013, 0.0015, 0.0014, 0.0013, 0.0013, 0.0016], device='cuda:6') 2023-04-27 21:59:15,794 INFO [optim.py:369] (6/7) 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 21:59:25,644 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.68 vs. limit=2.0 2023-04-27 21:59:56,671 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2023-04-27 22:00:17,502 INFO [finetune.py:976] (6/7) Epoch 25, batch 2050, loss[loss=0.1824, simple_loss=0.2508, pruned_loss=0.05701, over 4845.00 frames. ], tot_loss[loss=0.166, simple_loss=0.2382, pruned_loss=0.04693, over 954331.38 frames. ], batch size: 44, lr: 3.01e-03, grad_scale: 32.0 2023-04-27 22:00:50,416 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-04-27 22:01:10,317 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-04-27 22:01:21,674 INFO [finetune.py:976] (6/7) Epoch 25, batch 2100, loss[loss=0.1606, simple_loss=0.2289, pruned_loss=0.04612, over 4765.00 frames. ], tot_loss[loss=0.166, simple_loss=0.2375, pruned_loss=0.04723, over 955391.65 frames. ], batch size: 26, lr: 3.01e-03, grad_scale: 32.0 2023-04-27 22:01:21,788 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=139565.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 22:01:24,122 INFO [optim.py:369] (6/7) 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 22:01:30,516 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=139571.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:01:37,249 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=139582.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:01:44,578 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=139594.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:01:57,942 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=139613.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:01:59,117 INFO [finetune.py:976] (6/7) Epoch 25, batch 2150, loss[loss=0.1885, simple_loss=0.2688, pruned_loss=0.05415, over 4203.00 frames. ], tot_loss[loss=0.1678, simple_loss=0.2403, pruned_loss=0.04764, over 954782.79 frames. ], batch size: 65, lr: 3.01e-03, grad_scale: 32.0 2023-04-27 22:02:10,549 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=139632.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 22:02:10,580 INFO [zipformer.py:1188] (6/7) 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] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=139643.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:02:18,460 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=139645.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:02:27,363 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9472, 1.7765, 1.9866, 2.2599, 2.3617, 1.8310, 1.5761, 2.0698], device='cuda:6'), covar=tensor([0.0843, 0.1142, 0.0710, 0.0568, 0.0626, 0.0891, 0.0744, 0.0545], device='cuda:6'), in_proj_covar=tensor([0.0187, 0.0204, 0.0188, 0.0174, 0.0180, 0.0179, 0.0152, 0.0180], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 22:02:31,933 INFO [finetune.py:976] (6/7) Epoch 25, batch 2200, loss[loss=0.2137, simple_loss=0.2745, pruned_loss=0.07641, over 4828.00 frames. ], tot_loss[loss=0.17, simple_loss=0.2434, pruned_loss=0.04834, over 956519.07 frames. ], batch size: 47, lr: 3.01e-03, grad_scale: 32.0 2023-04-27 22:02:34,820 INFO [optim.py:369] (6/7) 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 22:02:43,413 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-04-27 22:02:45,213 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.1774, 2.6070, 2.2622, 2.5653, 1.8743, 2.0899, 2.2863, 1.6070], device='cuda:6'), covar=tensor([0.1958, 0.1077, 0.0769, 0.1031, 0.3061, 0.1271, 0.1839, 0.2535], device='cuda:6'), in_proj_covar=tensor([0.0286, 0.0301, 0.0214, 0.0277, 0.0313, 0.0256, 0.0249, 0.0264], device='cuda:6'), out_proj_covar=tensor([1.1425e-04, 1.1905e-04, 8.4313e-05, 1.0939e-04, 1.2612e-04, 1.0098e-04, 1.0034e-04, 1.0428e-04], device='cuda:6') 2023-04-27 22:03:01,822 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7367, 2.0893, 1.2241, 1.4952, 2.1797, 1.5552, 1.5221, 1.6855], device='cuda:6'), covar=tensor([0.0485, 0.0334, 0.0273, 0.0545, 0.0241, 0.0482, 0.0465, 0.0549], device='cuda:6'), in_proj_covar=tensor([0.0028, 0.0023, 0.0022, 0.0028, 0.0019, 0.0028, 0.0027, 0.0028], device='cuda:6'), out_proj_covar=tensor([0.0051, 0.0046, 0.0038, 0.0053, 0.0039, 0.0050, 0.0050, 0.0052], device='cuda:6') 2023-04-27 22:03:04,782 INFO [finetune.py:976] (6/7) Epoch 25, batch 2250, loss[loss=0.1855, simple_loss=0.2546, pruned_loss=0.05823, over 4749.00 frames. ], tot_loss[loss=0.1717, simple_loss=0.2449, pruned_loss=0.04928, over 955609.85 frames. ], batch size: 54, lr: 3.01e-03, grad_scale: 32.0 2023-04-27 22:03:38,678 INFO [finetune.py:976] (6/7) Epoch 25, batch 2300, loss[loss=0.1694, simple_loss=0.233, pruned_loss=0.05293, over 4814.00 frames. ], tot_loss[loss=0.1715, simple_loss=0.2449, pruned_loss=0.04903, over 956934.40 frames. ], batch size: 38, lr: 3.01e-03, grad_scale: 32.0 2023-04-27 22:03:41,541 INFO [optim.py:369] (6/7) 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 22:03:41,681 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=139769.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:03:54,762 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=139789.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:04:03,989 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=139804.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:04:10,818 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=139806.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:04:23,018 INFO [finetune.py:976] (6/7) Epoch 25, batch 2350, loss[loss=0.1591, simple_loss=0.2154, pruned_loss=0.05138, over 4696.00 frames. ], tot_loss[loss=0.1693, simple_loss=0.2421, pruned_loss=0.04825, over 957416.75 frames. ], batch size: 23, lr: 3.01e-03, grad_scale: 32.0 2023-04-27 22:04:44,831 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=139830.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:05:06,887 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=139850.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:05:06,896 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=139850.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:05:09,119 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-04-27 22:05:17,203 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.4859, 2.9999, 2.5916, 2.8527, 2.1659, 2.4610, 2.6968, 2.0053], device='cuda:6'), covar=tensor([0.1940, 0.1142, 0.0707, 0.1117, 0.2704, 0.1070, 0.1953, 0.2730], device='cuda:6'), in_proj_covar=tensor([0.0284, 0.0300, 0.0213, 0.0277, 0.0311, 0.0255, 0.0248, 0.0263], device='cuda:6'), out_proj_covar=tensor([1.1344e-04, 1.1850e-04, 8.4008e-05, 1.0923e-04, 1.2556e-04, 1.0058e-04, 1.0005e-04, 1.0379e-04], device='cuda:6') 2023-04-27 22:05:17,689 INFO [finetune.py:976] (6/7) Epoch 25, batch 2400, loss[loss=0.1356, simple_loss=0.2072, pruned_loss=0.03199, over 4902.00 frames. ], tot_loss[loss=0.1674, simple_loss=0.2393, pruned_loss=0.0477, over 955130.27 frames. ], batch size: 36, lr: 3.01e-03, grad_scale: 32.0 2023-04-27 22:05:17,809 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=139865.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:05:19,538 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=139867.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:05:20,605 INFO [optim.py:369] (6/7) 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 22:05:23,157 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.5564, 2.5161, 2.7471, 3.2292, 3.1181, 2.4008, 2.1630, 2.8302], device='cuda:6'), covar=tensor([0.0935, 0.0976, 0.0662, 0.0608, 0.0562, 0.0991, 0.0773, 0.0559], device='cuda:6'), in_proj_covar=tensor([0.0188, 0.0206, 0.0189, 0.0175, 0.0181, 0.0180, 0.0153, 0.0181], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 22:05:37,259 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=139894.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:05:53,351 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=139911.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:05:55,625 INFO [finetune.py:976] (6/7) Epoch 25, batch 2450, loss[loss=0.113, simple_loss=0.1894, pruned_loss=0.0183, over 4765.00 frames. ], tot_loss[loss=0.1642, simple_loss=0.236, pruned_loss=0.04617, over 954128.04 frames. ], batch size: 28, lr: 3.01e-03, grad_scale: 16.0 2023-04-27 22:06:05,169 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-04-27 22:06:15,192 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=139927.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:06:18,763 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=139932.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 22:06:27,883 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=139938.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:06:35,938 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=139942.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:06:37,757 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=139945.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:06:59,672 INFO [finetune.py:976] (6/7) Epoch 25, batch 2500, loss[loss=0.1729, simple_loss=0.2373, pruned_loss=0.05421, over 4749.00 frames. ], tot_loss[loss=0.1664, simple_loss=0.2381, pruned_loss=0.04737, over 953033.42 frames. ], batch size: 23, lr: 3.01e-03, grad_scale: 16.0 2023-04-27 22:07:01,388 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-04-27 22:07:04,983 INFO [optim.py:369] (6/7) 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 22:07:11,621 INFO [zipformer.py:1188] (6/7) 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] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=139993.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:07:35,750 INFO [finetune.py:976] (6/7) Epoch 25, batch 2550, loss[loss=0.2464, simple_loss=0.3197, pruned_loss=0.08652, over 4835.00 frames. ], tot_loss[loss=0.1687, simple_loss=0.2409, pruned_loss=0.0482, over 951640.72 frames. ], batch size: 47, lr: 3.01e-03, grad_scale: 16.0 2023-04-27 22:07:49,717 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6455, 1.7280, 0.8705, 1.3324, 1.8832, 1.4639, 1.3734, 1.4506], device='cuda:6'), covar=tensor([0.0460, 0.0338, 0.0317, 0.0524, 0.0250, 0.0506, 0.0484, 0.0536], device='cuda:6'), in_proj_covar=tensor([0.0028, 0.0023, 0.0022, 0.0028, 0.0019, 0.0028, 0.0027, 0.0028], device='cuda:6'), out_proj_covar=tensor([0.0051, 0.0045, 0.0038, 0.0052, 0.0039, 0.0050, 0.0050, 0.0052], device='cuda:6') 2023-04-27 22:08:09,497 INFO [finetune.py:976] (6/7) Epoch 25, batch 2600, loss[loss=0.1562, simple_loss=0.2207, pruned_loss=0.04587, over 4906.00 frames. ], tot_loss[loss=0.169, simple_loss=0.2415, pruned_loss=0.04823, over 952372.18 frames. ], batch size: 37, lr: 3.01e-03, grad_scale: 16.0 2023-04-27 22:08:09,631 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9212, 1.3431, 1.5391, 1.6142, 2.0808, 1.6917, 1.4452, 1.5242], device='cuda:6'), covar=tensor([0.1739, 0.1814, 0.2008, 0.1668, 0.1004, 0.1707, 0.1941, 0.2438], device='cuda:6'), in_proj_covar=tensor([0.0312, 0.0309, 0.0351, 0.0286, 0.0329, 0.0306, 0.0299, 0.0374], device='cuda:6'), out_proj_covar=tensor([6.3815e-05, 6.3671e-05, 7.3925e-05, 5.7355e-05, 6.7640e-05, 6.4118e-05, 6.1933e-05, 7.9229e-05], device='cuda:6') 2023-04-27 22:08:12,526 INFO [optim.py:369] (6/7) 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 22:08:13,363 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-04-27 22:08:33,243 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0958, 1.7929, 2.2655, 2.4756, 2.0870, 2.0182, 2.1686, 2.1241], device='cuda:6'), covar=tensor([0.4670, 0.7277, 0.7221, 0.5729, 0.6203, 0.8765, 0.9094, 0.9520], device='cuda:6'), in_proj_covar=tensor([0.0438, 0.0421, 0.0512, 0.0509, 0.0466, 0.0499, 0.0504, 0.0518], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 22:08:35,644 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=140103.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:08:39,912 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 25, batch 2650, loss[loss=0.1476, simple_loss=0.2073, pruned_loss=0.04391, over 4102.00 frames. ], tot_loss[loss=0.1702, simple_loss=0.2436, pruned_loss=0.04846, over 954120.43 frames. ], batch size: 17, lr: 3.01e-03, grad_scale: 16.0 2023-04-27 22:08:49,910 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=140125.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:09:04,026 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=140145.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:09:13,156 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=140160.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:09:14,363 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=140162.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:09:15,622 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=140164.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:09:16,094 INFO [finetune.py:976] (6/7) Epoch 25, batch 2700, loss[loss=0.1523, simple_loss=0.2216, pruned_loss=0.04154, over 4828.00 frames. ], tot_loss[loss=0.1698, simple_loss=0.243, pruned_loss=0.04826, over 953678.69 frames. ], batch size: 49, lr: 3.01e-03, grad_scale: 16.0 2023-04-27 22:09:19,142 INFO [optim.py:369] (6/7) 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 22:09:19,891 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=140171.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:09:55,083 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 25, batch 2750, loss[loss=0.17, simple_loss=0.2417, pruned_loss=0.04916, over 4913.00 frames. ], tot_loss[loss=0.1681, simple_loss=0.2407, pruned_loss=0.04769, over 956367.70 frames. ], batch size: 36, lr: 3.01e-03, grad_scale: 16.0 2023-04-27 22:10:18,554 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=140227.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:10:32,149 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=140238.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:11:02,670 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8248, 2.1214, 1.8081, 1.9955, 1.6487, 1.7683, 1.7183, 1.4803], device='cuda:6'), covar=tensor([0.1523, 0.1170, 0.0815, 0.1127, 0.3125, 0.1042, 0.1629, 0.2087], device='cuda:6'), in_proj_covar=tensor([0.0283, 0.0299, 0.0212, 0.0275, 0.0310, 0.0253, 0.0247, 0.0262], device='cuda:6'), out_proj_covar=tensor([1.1324e-04, 1.1813e-04, 8.3532e-05, 1.0851e-04, 1.2495e-04, 9.9874e-05, 9.9577e-05, 1.0353e-04], device='cuda:6') 2023-04-27 22:11:03,258 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 25, batch 2800, loss[loss=0.1762, simple_loss=0.2426, pruned_loss=0.05484, over 4300.00 frames. ], tot_loss[loss=0.1653, simple_loss=0.2376, pruned_loss=0.04655, over 955457.45 frames. ], batch size: 65, lr: 3.01e-03, grad_scale: 16.0 2023-04-27 22:11:15,318 INFO [optim.py:369] (6/7) 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 22:11:20,366 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=140275.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:11:21,690 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9760, 1.4493, 1.6089, 1.7370, 2.1369, 1.7129, 1.4763, 1.4866], device='cuda:6'), covar=tensor([0.1726, 0.1775, 0.1761, 0.1330, 0.0922, 0.1961, 0.2411, 0.2566], device='cuda:6'), in_proj_covar=tensor([0.0314, 0.0310, 0.0353, 0.0288, 0.0330, 0.0308, 0.0300, 0.0375], device='cuda:6'), out_proj_covar=tensor([6.4163e-05, 6.3754e-05, 7.4252e-05, 5.7767e-05, 6.7786e-05, 6.4487e-05, 6.2280e-05, 7.9517e-05], device='cuda:6') 2023-04-27 22:11:32,994 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 25, batch 2850, loss[loss=0.1889, simple_loss=0.2732, pruned_loss=0.05233, over 4864.00 frames. ], tot_loss[loss=0.1646, simple_loss=0.2366, pruned_loss=0.04628, over 955693.98 frames. ], batch size: 49, lr: 3.00e-03, grad_scale: 16.0 2023-04-27 22:12:17,263 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=140320.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:12:48,658 INFO [finetune.py:976] (6/7) Epoch 25, batch 2900, loss[loss=0.1371, simple_loss=0.2183, pruned_loss=0.02791, over 4759.00 frames. ], tot_loss[loss=0.1679, simple_loss=0.2403, pruned_loss=0.04776, over 954116.48 frames. ], batch size: 28, lr: 3.00e-03, grad_scale: 16.0 2023-04-27 22:12:51,723 INFO [optim.py:369] (6/7) 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 22:13:00,442 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.4921, 1.8233, 1.8864, 1.9530, 1.8193, 1.8487, 1.9489, 1.8583], device='cuda:6'), covar=tensor([0.3596, 0.5514, 0.4898, 0.4450, 0.5781, 0.7494, 0.5196, 0.5380], device='cuda:6'), in_proj_covar=tensor([0.0340, 0.0374, 0.0326, 0.0339, 0.0349, 0.0393, 0.0358, 0.0332], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 22:13:03,189 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=140387.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:13:20,921 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6877, 2.2813, 1.7500, 1.6836, 1.2879, 1.2946, 1.7300, 1.2020], device='cuda:6'), covar=tensor([0.1717, 0.1313, 0.1415, 0.1640, 0.2309, 0.1915, 0.0999, 0.2074], device='cuda:6'), in_proj_covar=tensor([0.0198, 0.0211, 0.0169, 0.0204, 0.0200, 0.0187, 0.0157, 0.0187], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:6') 2023-04-27 22:13:22,623 INFO [finetune.py:976] (6/7) Epoch 25, batch 2950, loss[loss=0.2022, simple_loss=0.281, pruned_loss=0.06166, over 4848.00 frames. ], tot_loss[loss=0.1693, simple_loss=0.2425, pruned_loss=0.04802, over 954098.92 frames. ], batch size: 47, lr: 3.00e-03, grad_scale: 16.0 2023-04-27 22:13:28,749 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=140425.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:13:42,315 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=140445.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:13:44,122 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=140448.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:13:52,224 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=140459.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:13:52,866 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=140460.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:13:53,455 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([4.1205, 4.1201, 3.0762, 4.7725, 4.2327, 4.1551, 2.0180, 4.1875], device='cuda:6'), covar=tensor([0.1946, 0.1108, 0.2572, 0.1301, 0.2607, 0.1801, 0.5082, 0.2142], device='cuda:6'), in_proj_covar=tensor([0.0245, 0.0218, 0.0251, 0.0306, 0.0298, 0.0248, 0.0273, 0.0273], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 22:13:54,551 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=140462.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:13:56,275 INFO [finetune.py:976] (6/7) Epoch 25, batch 3000, loss[loss=0.1912, simple_loss=0.2505, pruned_loss=0.06595, over 4923.00 frames. ], tot_loss[loss=0.1711, simple_loss=0.244, pruned_loss=0.04911, over 954965.61 frames. ], batch size: 33, lr: 3.00e-03, grad_scale: 16.0 2023-04-27 22:13:56,275 INFO [finetune.py:1001] (6/7) Computing validation loss 2023-04-27 22:14:07,208 INFO [finetune.py:1010] (6/7) 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] (6/7) Maximum memory allocated so far is 6435MB 2023-04-27 22:14:07,901 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=140466.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:14:08,027 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.61 vs. limit=2.0 2023-04-27 22:14:10,188 INFO [optim.py:369] (6/7) 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 22:14:12,097 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=140473.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:14:21,886 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=140489.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:14:24,277 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=140493.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:14:27,586 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.14 vs. limit=2.0 2023-04-27 22:14:33,158 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=140506.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:14:34,803 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=140508.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:14:35,487 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=140509.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:14:36,051 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=140510.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:14:39,577 INFO [finetune.py:976] (6/7) Epoch 25, batch 3050, loss[loss=0.1563, simple_loss=0.238, pruned_loss=0.03729, over 4927.00 frames. ], tot_loss[loss=0.1722, simple_loss=0.2453, pruned_loss=0.0495, over 953769.67 frames. ], batch size: 33, lr: 3.00e-03, grad_scale: 16.0 2023-04-27 22:15:02,561 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=140550.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:15:05,404 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=140554.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:15:12,553 INFO [finetune.py:976] (6/7) Epoch 25, batch 3100, loss[loss=0.1502, simple_loss=0.2119, pruned_loss=0.04421, over 4163.00 frames. ], tot_loss[loss=0.1695, simple_loss=0.2423, pruned_loss=0.04837, over 953971.53 frames. ], batch size: 18, lr: 3.00e-03, grad_scale: 16.0 2023-04-27 22:15:15,560 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8095, 1.8956, 1.1211, 1.5449, 1.8874, 1.6329, 1.5819, 1.6290], device='cuda:6'), covar=tensor([0.0502, 0.0289, 0.0288, 0.0516, 0.0253, 0.0529, 0.0547, 0.0527], device='cuda:6'), in_proj_covar=tensor([0.0028, 0.0023, 0.0022, 0.0028, 0.0019, 0.0027, 0.0027, 0.0028], device='cuda:6'), out_proj_covar=tensor([0.0051, 0.0046, 0.0038, 0.0052, 0.0039, 0.0050, 0.0050, 0.0052], device='cuda:6') 2023-04-27 22:15:16,056 INFO [optim.py:369] (6/7) 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 22:15:16,227 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=140570.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:15:21,483 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.4837, 1.3641, 4.2910, 4.0108, 3.7325, 4.0967, 4.0109, 3.7442], device='cuda:6'), covar=tensor([0.6876, 0.5889, 0.0985, 0.1544, 0.1114, 0.1651, 0.1312, 0.1539], device='cuda:6'), in_proj_covar=tensor([0.0311, 0.0308, 0.0408, 0.0411, 0.0348, 0.0413, 0.0319, 0.0369], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-27 22:15:56,619 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.64 vs. limit=2.0 2023-04-27 22:16:07,370 INFO [finetune.py:976] (6/7) Epoch 25, batch 3150, loss[loss=0.1359, simple_loss=0.2189, pruned_loss=0.02645, over 4897.00 frames. ], tot_loss[loss=0.168, simple_loss=0.2401, pruned_loss=0.04793, over 954174.22 frames. ], batch size: 32, lr: 3.00e-03, grad_scale: 16.0 2023-04-27 22:16:07,467 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=140615.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:16:37,112 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8257, 2.4803, 1.7364, 1.9528, 1.4091, 1.4036, 1.7285, 1.2834], device='cuda:6'), covar=tensor([0.1791, 0.1244, 0.1642, 0.1633, 0.2350, 0.2195, 0.1084, 0.2158], device='cuda:6'), in_proj_covar=tensor([0.0198, 0.0210, 0.0169, 0.0204, 0.0200, 0.0186, 0.0157, 0.0187], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:6') 2023-04-27 22:17:13,347 INFO [finetune.py:976] (6/7) Epoch 25, batch 3200, loss[loss=0.1171, simple_loss=0.1983, pruned_loss=0.01797, over 4794.00 frames. ], tot_loss[loss=0.1647, simple_loss=0.2365, pruned_loss=0.04649, over 955327.74 frames. ], batch size: 29, lr: 3.00e-03, grad_scale: 16.0 2023-04-27 22:17:21,871 INFO [optim.py:369] (6/7) 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 22:18:19,768 INFO [finetune.py:976] (6/7) Epoch 25, batch 3250, loss[loss=0.2047, simple_loss=0.2788, pruned_loss=0.06527, over 4905.00 frames. ], tot_loss[loss=0.1671, simple_loss=0.2386, pruned_loss=0.04782, over 955466.91 frames. ], batch size: 43, lr: 3.00e-03, grad_scale: 16.0 2023-04-27 22:18:26,958 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=140723.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:18:55,998 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=140743.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:19:16,149 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=140759.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:19:20,225 INFO [finetune.py:976] (6/7) Epoch 25, batch 3300, loss[loss=0.133, simple_loss=0.2076, pruned_loss=0.02923, over 4729.00 frames. ], tot_loss[loss=0.168, simple_loss=0.2401, pruned_loss=0.04798, over 953163.70 frames. ], batch size: 23, lr: 3.00e-03, grad_scale: 16.0 2023-04-27 22:19:21,438 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=140766.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:19:27,807 INFO [optim.py:369] (6/7) 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 22:19:48,012 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=140784.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:20:19,457 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=140807.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:20:20,754 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=140809.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:20:23,780 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=140814.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:20:24,324 INFO [finetune.py:976] (6/7) Epoch 25, batch 3350, loss[loss=0.1596, simple_loss=0.2409, pruned_loss=0.03915, over 4856.00 frames. ], tot_loss[loss=0.1695, simple_loss=0.2421, pruned_loss=0.04844, over 951279.43 frames. ], batch size: 31, lr: 3.00e-03, grad_scale: 16.0 2023-04-27 22:20:46,743 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=140845.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:20:59,013 INFO [finetune.py:976] (6/7) Epoch 25, batch 3400, loss[loss=0.1817, simple_loss=0.2621, pruned_loss=0.05065, over 4196.00 frames. ], tot_loss[loss=0.1705, simple_loss=0.2431, pruned_loss=0.04898, over 952570.38 frames. ], batch size: 65, lr: 3.00e-03, grad_scale: 16.0 2023-04-27 22:20:59,085 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=140865.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:21:02,011 INFO [optim.py:369] (6/7) 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 22:21:02,130 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=140870.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:21:32,401 INFO [finetune.py:976] (6/7) Epoch 25, batch 3450, loss[loss=0.1415, simple_loss=0.2131, pruned_loss=0.03493, over 4873.00 frames. ], tot_loss[loss=0.1694, simple_loss=0.2426, pruned_loss=0.04813, over 953592.29 frames. ], batch size: 31, lr: 3.00e-03, grad_scale: 16.0 2023-04-27 22:21:32,514 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=140915.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:21:33,723 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=140917.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:21:55,370 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=140947.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:22:00,946 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-04-27 22:22:05,106 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 25, batch 3500, loss[loss=0.1502, simple_loss=0.2145, pruned_loss=0.04294, over 4810.00 frames. ], tot_loss[loss=0.1685, simple_loss=0.241, pruned_loss=0.04798, over 953617.63 frames. ], batch size: 33, lr: 3.00e-03, grad_scale: 16.0 2023-04-27 22:22:09,320 INFO [optim.py:369] (6/7) 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] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=140978.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:22:35,377 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 25, batch 3550, loss[loss=0.2279, simple_loss=0.2681, pruned_loss=0.0939, over 4068.00 frames. ], tot_loss[loss=0.1663, simple_loss=0.238, pruned_loss=0.04733, over 953293.94 frames. ], batch size: 65, lr: 3.00e-03, grad_scale: 16.0 2023-04-27 22:22:58,168 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=141043.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:23:13,442 INFO [finetune.py:976] (6/7) Epoch 25, batch 3600, loss[loss=0.1733, simple_loss=0.2587, pruned_loss=0.04393, over 4910.00 frames. ], tot_loss[loss=0.1644, simple_loss=0.2356, pruned_loss=0.04661, over 953437.51 frames. ], batch size: 36, lr: 3.00e-03, grad_scale: 16.0 2023-04-27 22:23:13,613 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.66 vs. limit=2.0 2023-04-27 22:23:16,478 INFO [optim.py:369] (6/7) 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 22:23:22,260 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.2153, 1.4141, 1.3406, 1.6758, 1.5882, 1.8309, 1.2909, 3.2926], device='cuda:6'), covar=tensor([0.0582, 0.0799, 0.0735, 0.1130, 0.0586, 0.0514, 0.0727, 0.0146], device='cuda:6'), in_proj_covar=tensor([0.0037, 0.0039, 0.0039, 0.0042, 0.0040, 0.0038, 0.0038, 0.0055], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0013, 0.0013, 0.0015, 0.0014, 0.0013, 0.0013, 0.0016], device='cuda:6') 2023-04-27 22:23:32,938 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=141079.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:23:46,971 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=141091.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:24:19,422 INFO [finetune.py:976] (6/7) Epoch 25, batch 3650, loss[loss=0.2048, simple_loss=0.2708, pruned_loss=0.06933, over 4733.00 frames. ], tot_loss[loss=0.1669, simple_loss=0.2384, pruned_loss=0.04768, over 954249.40 frames. ], batch size: 54, lr: 3.00e-03, grad_scale: 16.0 2023-04-27 22:24:42,719 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=141143.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:24:43,930 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=141145.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:24:46,344 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([3.5329, 3.5213, 2.8313, 4.1650, 3.5767, 3.6096, 1.5291, 3.5445], device='cuda:6'), covar=tensor([0.2057, 0.1224, 0.3789, 0.1474, 0.3043, 0.1844, 0.6090, 0.2511], device='cuda:6'), in_proj_covar=tensor([0.0248, 0.0221, 0.0255, 0.0309, 0.0301, 0.0252, 0.0277, 0.0276], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 22:24:57,936 INFO [finetune.py:976] (6/7) Epoch 25, batch 3700, loss[loss=0.1646, simple_loss=0.2414, pruned_loss=0.04385, over 4888.00 frames. ], tot_loss[loss=0.1698, simple_loss=0.2418, pruned_loss=0.04893, over 951234.84 frames. ], batch size: 32, lr: 3.00e-03, grad_scale: 16.0 2023-04-27 22:24:58,013 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=141165.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:24:58,030 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=141165.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:25:00,979 INFO [optim.py:369] (6/7) 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 22:25:06,340 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.81 vs. limit=2.0 2023-04-27 22:25:14,086 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.3747, 1.5901, 1.3817, 1.5617, 1.2312, 1.3161, 1.3648, 1.1004], device='cuda:6'), covar=tensor([0.1559, 0.1273, 0.0810, 0.1055, 0.3544, 0.1189, 0.1682, 0.2323], device='cuda:6'), in_proj_covar=tensor([0.0289, 0.0305, 0.0216, 0.0279, 0.0316, 0.0259, 0.0251, 0.0268], device='cuda:6'), out_proj_covar=tensor([1.1536e-04, 1.2025e-04, 8.5175e-05, 1.0998e-04, 1.2730e-04, 1.0200e-04, 1.0120e-04, 1.0569e-04], device='cuda:6') 2023-04-27 22:25:15,813 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=141193.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:25:22,648 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=141204.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:25:30,031 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 25, batch 3750, loss[loss=0.1779, simple_loss=0.2565, pruned_loss=0.04966, over 4921.00 frames. ], tot_loss[loss=0.1718, simple_loss=0.2442, pruned_loss=0.04971, over 952440.93 frames. ], batch size: 38, lr: 3.00e-03, grad_scale: 16.0 2023-04-27 22:26:00,473 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-04-27 22:26:23,955 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.2070, 1.4775, 2.0334, 2.4445, 2.1558, 1.6826, 1.4579, 1.8728], device='cuda:6'), covar=tensor([0.3506, 0.4181, 0.1963, 0.2977, 0.2959, 0.3056, 0.4709, 0.2333], device='cuda:6'), in_proj_covar=tensor([0.0292, 0.0246, 0.0228, 0.0312, 0.0220, 0.0234, 0.0227, 0.0186], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:6') 2023-04-27 22:26:35,279 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9428, 1.7235, 1.9412, 2.3345, 2.3269, 1.9923, 1.5313, 2.0762], device='cuda:6'), covar=tensor([0.0816, 0.1155, 0.0731, 0.0539, 0.0619, 0.0738, 0.0771, 0.0538], device='cuda:6'), in_proj_covar=tensor([0.0182, 0.0200, 0.0184, 0.0170, 0.0176, 0.0175, 0.0148, 0.0176], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 22:26:37,437 INFO [finetune.py:976] (6/7) Epoch 25, batch 3800, loss[loss=0.2012, simple_loss=0.2789, pruned_loss=0.06177, over 4825.00 frames. ], tot_loss[loss=0.1708, simple_loss=0.2441, pruned_loss=0.0488, over 953959.49 frames. ], batch size: 47, lr: 3.00e-03, grad_scale: 16.0 2023-04-27 22:26:45,872 INFO [optim.py:369] (6/7) 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 22:26:47,830 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=141273.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:27:28,698 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=141303.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:27:42,128 INFO [finetune.py:976] (6/7) Epoch 25, batch 3850, loss[loss=0.2029, simple_loss=0.2719, pruned_loss=0.06692, over 4913.00 frames. ], tot_loss[loss=0.1695, simple_loss=0.2424, pruned_loss=0.04831, over 954641.04 frames. ], batch size: 36, lr: 3.00e-03, grad_scale: 16.0 2023-04-27 22:28:02,215 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=141329.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:28:15,657 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.1297, 2.6427, 1.0156, 1.5093, 2.0202, 1.1698, 3.6074, 1.8201], device='cuda:6'), covar=tensor([0.0648, 0.0668, 0.0802, 0.1164, 0.0522, 0.1013, 0.0193, 0.0607], device='cuda:6'), in_proj_covar=tensor([0.0050, 0.0064, 0.0047, 0.0046, 0.0049, 0.0051, 0.0073, 0.0051], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:6') 2023-04-27 22:28:15,719 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7881, 1.4468, 1.9874, 2.3063, 1.9135, 1.8084, 1.8830, 1.8021], device='cuda:6'), covar=tensor([0.4753, 0.6976, 0.6120, 0.5601, 0.6076, 0.7925, 0.7941, 0.9066], device='cuda:6'), in_proj_covar=tensor([0.0438, 0.0420, 0.0512, 0.0506, 0.0466, 0.0500, 0.0503, 0.0516], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 22:28:27,556 INFO [finetune.py:976] (6/7) Epoch 25, batch 3900, loss[loss=0.1336, simple_loss=0.2095, pruned_loss=0.02887, over 4859.00 frames. ], tot_loss[loss=0.1675, simple_loss=0.2403, pruned_loss=0.04738, over 955750.35 frames. ], batch size: 34, lr: 3.00e-03, grad_scale: 16.0 2023-04-27 22:28:31,460 INFO [optim.py:369] (6/7) 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 22:28:37,532 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=141379.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:28:44,168 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=141390.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:29:00,434 INFO [finetune.py:976] (6/7) Epoch 25, batch 3950, loss[loss=0.1653, simple_loss=0.2456, pruned_loss=0.04249, over 4847.00 frames. ], tot_loss[loss=0.1664, simple_loss=0.2385, pruned_loss=0.04713, over 957326.57 frames. ], batch size: 47, lr: 3.00e-03, grad_scale: 16.0 2023-04-27 22:29:01,807 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=141417.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 22:29:09,298 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=141427.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:29:10,551 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.1460, 2.7260, 2.3182, 2.6595, 1.9735, 2.2911, 2.2649, 1.8610], device='cuda:6'), covar=tensor([0.1917, 0.1125, 0.0704, 0.1037, 0.3123, 0.1072, 0.1805, 0.2539], device='cuda:6'), in_proj_covar=tensor([0.0286, 0.0302, 0.0215, 0.0277, 0.0313, 0.0256, 0.0249, 0.0266], device='cuda:6'), 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:6') 2023-04-27 22:29:16,024 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6517, 1.2330, 1.2330, 1.4056, 1.8013, 1.4749, 1.3332, 1.1450], device='cuda:6'), covar=tensor([0.1574, 0.1525, 0.1839, 0.1358, 0.0857, 0.1526, 0.1871, 0.2465], device='cuda:6'), in_proj_covar=tensor([0.0311, 0.0306, 0.0349, 0.0284, 0.0326, 0.0303, 0.0297, 0.0370], device='cuda:6'), out_proj_covar=tensor([6.3669e-05, 6.2944e-05, 7.3328e-05, 5.7024e-05, 6.7010e-05, 6.3451e-05, 6.1506e-05, 7.8430e-05], device='cuda:6') 2023-04-27 22:29:17,827 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=141441.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:29:33,822 INFO [finetune.py:976] (6/7) Epoch 25, batch 4000, loss[loss=0.2067, simple_loss=0.2696, pruned_loss=0.0719, over 4935.00 frames. ], tot_loss[loss=0.1669, simple_loss=0.2388, pruned_loss=0.0475, over 956325.16 frames. ], batch size: 38, lr: 3.00e-03, grad_scale: 16.0 2023-04-27 22:29:33,905 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=141465.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:29:36,896 INFO [optim.py:369] (6/7) 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 22:29:43,328 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=141478.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 22:29:54,792 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.2724, 1.4535, 1.4272, 1.6991, 1.6079, 1.8382, 1.3480, 3.3584], device='cuda:6'), covar=tensor([0.0591, 0.0775, 0.0766, 0.1163, 0.0627, 0.0507, 0.0732, 0.0133], device='cuda:6'), in_proj_covar=tensor([0.0037, 0.0038, 0.0039, 0.0042, 0.0040, 0.0038, 0.0038, 0.0055], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0013, 0.0013, 0.0015, 0.0014, 0.0013, 0.0013, 0.0016], device='cuda:6') 2023-04-27 22:29:55,960 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=141499.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:29:57,868 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=141502.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:30:05,461 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 25, batch 4050, loss[loss=0.1943, simple_loss=0.2816, pruned_loss=0.05352, over 4817.00 frames. ], tot_loss[loss=0.168, simple_loss=0.2406, pruned_loss=0.04766, over 955635.97 frames. ], batch size: 40, lr: 3.00e-03, grad_scale: 16.0 2023-04-27 22:30:35,441 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7465, 2.5341, 2.0442, 1.9002, 1.2805, 1.4017, 2.1092, 1.2760], device='cuda:6'), covar=tensor([0.1845, 0.1410, 0.1351, 0.1732, 0.2327, 0.2027, 0.0901, 0.2202], device='cuda:6'), in_proj_covar=tensor([0.0197, 0.0209, 0.0167, 0.0203, 0.0198, 0.0185, 0.0156, 0.0187], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:6') 2023-04-27 22:30:36,128 INFO [scaling.py:679] (6/7) 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] (6/7) Epoch 25, batch 4100, loss[loss=0.134, simple_loss=0.2198, pruned_loss=0.02408, over 4808.00 frames. ], tot_loss[loss=0.1707, simple_loss=0.2438, pruned_loss=0.04879, over 953608.69 frames. ], batch size: 29, lr: 3.00e-03, grad_scale: 16.0 2023-04-27 22:30:40,821 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.2970, 1.7626, 2.2416, 2.7243, 2.2064, 1.6951, 1.5631, 2.0369], device='cuda:6'), covar=tensor([0.3066, 0.2983, 0.1541, 0.2018, 0.2431, 0.2684, 0.3751, 0.1811], device='cuda:6'), in_proj_covar=tensor([0.0290, 0.0243, 0.0226, 0.0311, 0.0219, 0.0232, 0.0225, 0.0184], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:6') 2023-04-27 22:30:42,489 INFO [optim.py:369] (6/7) 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 22:30:44,327 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=141573.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:31:04,720 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 25, batch 4150, loss[loss=0.1837, simple_loss=0.2719, pruned_loss=0.04773, over 4820.00 frames. ], tot_loss[loss=0.171, simple_loss=0.2444, pruned_loss=0.04878, over 955256.98 frames. ], batch size: 38, lr: 3.00e-03, grad_scale: 16.0 2023-04-27 22:31:21,628 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=141621.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:31:53,091 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-04-27 22:32:02,964 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=141651.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:32:13,601 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-04-27 22:32:23,212 INFO [finetune.py:976] (6/7) Epoch 25, batch 4200, loss[loss=0.1916, simple_loss=0.2541, pruned_loss=0.06452, over 4925.00 frames. ], tot_loss[loss=0.1705, simple_loss=0.2446, pruned_loss=0.04823, over 954824.14 frames. ], batch size: 33, lr: 3.00e-03, grad_scale: 16.0 2023-04-27 22:32:26,265 INFO [optim.py:369] (6/7) 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 22:32:47,766 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=141685.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:33:28,595 INFO [finetune.py:976] (6/7) Epoch 25, batch 4250, loss[loss=0.1869, simple_loss=0.2465, pruned_loss=0.06366, over 4820.00 frames. ], tot_loss[loss=0.1683, simple_loss=0.2417, pruned_loss=0.04748, over 953093.03 frames. ], batch size: 38, lr: 3.00e-03, grad_scale: 16.0 2023-04-27 22:34:30,331 INFO [finetune.py:976] (6/7) Epoch 25, batch 4300, loss[loss=0.1458, simple_loss=0.2153, pruned_loss=0.0382, over 4741.00 frames. ], tot_loss[loss=0.1677, simple_loss=0.2403, pruned_loss=0.04756, over 954077.04 frames. ], batch size: 59, lr: 2.99e-03, grad_scale: 16.0 2023-04-27 22:34:39,434 INFO [optim.py:369] (6/7) 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 22:34:41,377 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=141773.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 22:35:15,326 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=141797.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:35:16,572 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=141799.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:35:23,738 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7803, 1.7360, 0.8526, 1.4650, 1.6855, 1.5926, 1.5591, 1.6382], device='cuda:6'), covar=tensor([0.0465, 0.0345, 0.0324, 0.0522, 0.0268, 0.0486, 0.0456, 0.0516], device='cuda:6'), in_proj_covar=tensor([0.0028, 0.0023, 0.0022, 0.0028, 0.0019, 0.0027, 0.0027, 0.0028], device='cuda:6'), out_proj_covar=tensor([0.0051, 0.0046, 0.0038, 0.0052, 0.0039, 0.0050, 0.0050, 0.0052], device='cuda:6') 2023-04-27 22:35:36,801 INFO [finetune.py:976] (6/7) Epoch 25, batch 4350, loss[loss=0.1437, simple_loss=0.2188, pruned_loss=0.03432, over 4817.00 frames. ], tot_loss[loss=0.165, simple_loss=0.2372, pruned_loss=0.0464, over 954661.33 frames. ], batch size: 25, lr: 2.99e-03, grad_scale: 16.0 2023-04-27 22:36:10,388 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5977, 1.3095, 4.4155, 4.1176, 3.7896, 4.2504, 4.1066, 3.8718], device='cuda:6'), covar=tensor([0.7110, 0.5969, 0.1136, 0.1802, 0.1214, 0.1379, 0.1418, 0.1595], device='cuda:6'), in_proj_covar=tensor([0.0314, 0.0310, 0.0412, 0.0412, 0.0352, 0.0415, 0.0321, 0.0371], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-27 22:36:20,233 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 25, batch 4400, loss[loss=0.1566, simple_loss=0.2373, pruned_loss=0.03795, over 4750.00 frames. ], tot_loss[loss=0.1666, simple_loss=0.2381, pruned_loss=0.04753, over 953099.46 frames. ], batch size: 27, lr: 2.99e-03, grad_scale: 16.0 2023-04-27 22:36:50,566 INFO [optim.py:369] (6/7) 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 22:37:12,493 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7854, 2.3924, 1.9224, 1.8441, 1.3412, 1.3445, 1.9689, 1.3734], device='cuda:6'), covar=tensor([0.1726, 0.1302, 0.1342, 0.1637, 0.2339, 0.2031, 0.0974, 0.1912], device='cuda:6'), in_proj_covar=tensor([0.0198, 0.0209, 0.0168, 0.0203, 0.0199, 0.0185, 0.0156, 0.0187], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:6') 2023-04-27 22:37:14,302 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9244, 2.4624, 1.1089, 1.4420, 1.9446, 1.1507, 3.1207, 1.5561], device='cuda:6'), covar=tensor([0.0693, 0.0645, 0.0741, 0.1115, 0.0479, 0.0963, 0.0216, 0.0603], device='cuda:6'), in_proj_covar=tensor([0.0050, 0.0064, 0.0047, 0.0046, 0.0049, 0.0051, 0.0072, 0.0051], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:6') 2023-04-27 22:37:44,401 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5829, 2.0473, 1.5880, 1.4839, 1.1879, 1.1906, 1.6200, 1.1356], device='cuda:6'), covar=tensor([0.1756, 0.1240, 0.1478, 0.1689, 0.2307, 0.2011, 0.1018, 0.2061], device='cuda:6'), in_proj_covar=tensor([0.0198, 0.0210, 0.0168, 0.0203, 0.0199, 0.0185, 0.0156, 0.0187], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:6') 2023-04-27 22:37:46,060 INFO [finetune.py:976] (6/7) Epoch 25, batch 4450, loss[loss=0.1445, simple_loss=0.2234, pruned_loss=0.03279, over 4755.00 frames. ], tot_loss[loss=0.1701, simple_loss=0.2426, pruned_loss=0.04881, over 951985.07 frames. ], batch size: 27, lr: 2.99e-03, grad_scale: 32.0 2023-04-27 22:37:46,812 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=141916.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:38:05,387 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=141928.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:38:51,129 INFO [finetune.py:976] (6/7) Epoch 25, batch 4500, loss[loss=0.1778, simple_loss=0.2561, pruned_loss=0.04976, over 4893.00 frames. ], tot_loss[loss=0.1713, simple_loss=0.2441, pruned_loss=0.04922, over 952799.99 frames. ], batch size: 43, lr: 2.99e-03, grad_scale: 32.0 2023-04-27 22:38:59,404 INFO [optim.py:369] (6/7) 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 22:39:09,411 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=141977.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:39:21,461 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=141985.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:39:23,987 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=141989.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:39:57,184 INFO [finetune.py:976] (6/7) Epoch 25, batch 4550, loss[loss=0.17, simple_loss=0.2417, pruned_loss=0.04914, over 4830.00 frames. ], tot_loss[loss=0.1712, simple_loss=0.2447, pruned_loss=0.0488, over 952963.62 frames. ], batch size: 49, lr: 2.99e-03, grad_scale: 32.0 2023-04-27 22:40:19,390 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=142033.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:40:27,356 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.3575, 1.1960, 4.0827, 3.8118, 3.5687, 3.9544, 3.8131, 3.5110], device='cuda:6'), covar=tensor([0.7544, 0.6117, 0.1029, 0.1798, 0.1228, 0.1415, 0.1601, 0.1714], device='cuda:6'), in_proj_covar=tensor([0.0314, 0.0309, 0.0410, 0.0411, 0.0350, 0.0415, 0.0320, 0.0370], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-27 22:41:01,246 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([4.4887, 4.3640, 3.1170, 5.1345, 4.4308, 4.4550, 1.7595, 4.4458], device='cuda:6'), covar=tensor([0.1402, 0.0880, 0.3545, 0.0948, 0.3063, 0.1876, 0.5794, 0.1990], device='cuda:6'), in_proj_covar=tensor([0.0247, 0.0220, 0.0253, 0.0308, 0.0302, 0.0251, 0.0276, 0.0276], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 22:41:01,791 INFO [finetune.py:976] (6/7) Epoch 25, batch 4600, loss[loss=0.1376, simple_loss=0.2208, pruned_loss=0.02716, over 4809.00 frames. ], tot_loss[loss=0.1701, simple_loss=0.2441, pruned_loss=0.048, over 951354.02 frames. ], batch size: 40, lr: 2.99e-03, grad_scale: 32.0 2023-04-27 22:41:10,138 INFO [optim.py:369] (6/7) 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 22:41:12,054 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=142073.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 22:41:43,194 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=142097.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:42:06,627 INFO [finetune.py:976] (6/7) Epoch 25, batch 4650, loss[loss=0.142, simple_loss=0.212, pruned_loss=0.03599, over 4935.00 frames. ], tot_loss[loss=0.1688, simple_loss=0.2424, pruned_loss=0.04763, over 952782.24 frames. ], batch size: 33, lr: 2.99e-03, grad_scale: 32.0 2023-04-27 22:42:15,790 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=142121.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 22:42:28,942 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-04-27 22:42:47,440 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=142145.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:43:11,600 INFO [finetune.py:976] (6/7) Epoch 25, batch 4700, loss[loss=0.1155, simple_loss=0.1919, pruned_loss=0.01955, over 4802.00 frames. ], tot_loss[loss=0.1657, simple_loss=0.2387, pruned_loss=0.04638, over 954942.12 frames. ], batch size: 25, lr: 2.99e-03, grad_scale: 32.0 2023-04-27 22:43:19,817 INFO [optim.py:369] (6/7) 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 22:44:06,134 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.3326, 2.5347, 0.7890, 1.5322, 1.5455, 1.9792, 1.5885, 0.7850], device='cuda:6'), covar=tensor([0.1367, 0.1218, 0.1846, 0.1281, 0.1165, 0.0884, 0.1547, 0.1713], device='cuda:6'), in_proj_covar=tensor([0.0119, 0.0242, 0.0138, 0.0122, 0.0133, 0.0154, 0.0119, 0.0120], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-27 22:44:17,217 INFO [finetune.py:976] (6/7) Epoch 25, batch 4750, loss[loss=0.1597, simple_loss=0.2412, pruned_loss=0.03906, over 4944.00 frames. ], tot_loss[loss=0.1649, simple_loss=0.2373, pruned_loss=0.04622, over 953856.09 frames. ], batch size: 33, lr: 2.99e-03, grad_scale: 32.0 2023-04-27 22:44:49,933 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.71 vs. limit=2.0 2023-04-27 22:45:29,866 INFO [finetune.py:976] (6/7) Epoch 25, batch 4800, loss[loss=0.1468, simple_loss=0.2195, pruned_loss=0.03701, over 4803.00 frames. ], tot_loss[loss=0.1662, simple_loss=0.2386, pruned_loss=0.04695, over 951544.46 frames. ], batch size: 25, lr: 2.99e-03, grad_scale: 32.0 2023-04-27 22:45:34,011 INFO [optim.py:369] (6/7) 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 22:45:40,884 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=142272.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:45:42,528 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0039, 1.4631, 1.6263, 1.6665, 2.0775, 1.7514, 1.4933, 1.5193], device='cuda:6'), covar=tensor([0.1533, 0.1560, 0.1791, 0.1439, 0.0800, 0.1509, 0.1956, 0.2233], device='cuda:6'), in_proj_covar=tensor([0.0314, 0.0309, 0.0353, 0.0287, 0.0329, 0.0306, 0.0299, 0.0374], device='cuda:6'), out_proj_covar=tensor([6.4298e-05, 6.3526e-05, 7.4292e-05, 5.7570e-05, 6.7503e-05, 6.3933e-05, 6.2028e-05, 7.9313e-05], device='cuda:6') 2023-04-27 22:45:54,144 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=142284.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:46:36,613 INFO [finetune.py:976] (6/7) Epoch 25, batch 4850, loss[loss=0.2033, simple_loss=0.2912, pruned_loss=0.05771, over 4843.00 frames. ], tot_loss[loss=0.1675, simple_loss=0.2404, pruned_loss=0.04725, over 950485.27 frames. ], batch size: 49, lr: 2.99e-03, grad_scale: 32.0 2023-04-27 22:47:42,701 INFO [finetune.py:976] (6/7) Epoch 25, batch 4900, loss[loss=0.1566, simple_loss=0.2345, pruned_loss=0.03935, over 4853.00 frames. ], tot_loss[loss=0.1701, simple_loss=0.243, pruned_loss=0.04862, over 951798.80 frames. ], batch size: 31, lr: 2.99e-03, grad_scale: 32.0 2023-04-27 22:47:51,886 INFO [optim.py:369] (6/7) 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 22:48:49,025 INFO [finetune.py:976] (6/7) Epoch 25, batch 4950, loss[loss=0.1733, simple_loss=0.2488, pruned_loss=0.04896, over 4763.00 frames. ], tot_loss[loss=0.1708, simple_loss=0.244, pruned_loss=0.04881, over 949638.62 frames. ], batch size: 54, lr: 2.99e-03, grad_scale: 32.0 2023-04-27 22:49:58,430 INFO [finetune.py:976] (6/7) Epoch 25, batch 5000, loss[loss=0.1838, simple_loss=0.2411, pruned_loss=0.06325, over 4929.00 frames. ], tot_loss[loss=0.1705, simple_loss=0.2432, pruned_loss=0.04888, over 952782.00 frames. ], batch size: 33, lr: 2.99e-03, grad_scale: 32.0 2023-04-27 22:50:01,484 INFO [optim.py:369] (6/7) 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 22:50:10,360 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([0.7835, 1.6968, 1.8167, 1.4923, 1.8923, 1.5528, 2.2896, 1.6173], device='cuda:6'), covar=tensor([0.3353, 0.1803, 0.4298, 0.2370, 0.1447, 0.2218, 0.1446, 0.4105], device='cuda:6'), in_proj_covar=tensor([0.0339, 0.0353, 0.0427, 0.0352, 0.0382, 0.0376, 0.0371, 0.0424], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 22:51:03,146 INFO [finetune.py:976] (6/7) Epoch 25, batch 5050, loss[loss=0.1453, simple_loss=0.2206, pruned_loss=0.03499, over 4764.00 frames. ], tot_loss[loss=0.1687, simple_loss=0.241, pruned_loss=0.0482, over 951721.23 frames. ], batch size: 28, lr: 2.99e-03, grad_scale: 32.0 2023-04-27 22:51:14,503 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.3734, 2.4111, 2.7451, 2.9862, 2.9550, 2.2753, 2.0958, 2.5777], device='cuda:6'), covar=tensor([0.0852, 0.0894, 0.0549, 0.0530, 0.0545, 0.0860, 0.0697, 0.0537], device='cuda:6'), in_proj_covar=tensor([0.0186, 0.0203, 0.0186, 0.0173, 0.0178, 0.0179, 0.0151, 0.0178], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 22:51:48,458 INFO [finetune.py:976] (6/7) Epoch 25, batch 5100, loss[loss=0.1761, simple_loss=0.2418, pruned_loss=0.05516, over 4933.00 frames. ], tot_loss[loss=0.1663, simple_loss=0.2383, pruned_loss=0.04718, over 951427.32 frames. ], batch size: 33, lr: 2.99e-03, grad_scale: 32.0 2023-04-27 22:51:51,978 INFO [optim.py:369] (6/7) 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 22:51:53,304 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=142572.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:51:53,929 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=142573.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:52:01,678 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 25, batch 5150, loss[loss=0.2006, simple_loss=0.2645, pruned_loss=0.06834, over 4819.00 frames. ], tot_loss[loss=0.166, simple_loss=0.238, pruned_loss=0.04702, over 953037.72 frames. ], batch size: 47, lr: 2.99e-03, grad_scale: 32.0 2023-04-27 22:52:26,241 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=142620.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:52:33,590 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=142632.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:52:35,344 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=142634.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:52:42,820 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.76 vs. limit=2.0 2023-04-27 22:52:56,103 INFO [finetune.py:976] (6/7) Epoch 25, batch 5200, loss[loss=0.1519, simple_loss=0.2357, pruned_loss=0.03411, over 4759.00 frames. ], tot_loss[loss=0.1683, simple_loss=0.2414, pruned_loss=0.04757, over 953140.64 frames. ], batch size: 27, lr: 2.99e-03, grad_scale: 32.0 2023-04-27 22:53:00,176 INFO [optim.py:369] (6/7) 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 22:53:29,661 INFO [finetune.py:976] (6/7) Epoch 25, batch 5250, loss[loss=0.1602, simple_loss=0.2284, pruned_loss=0.04606, over 4833.00 frames. ], tot_loss[loss=0.1692, simple_loss=0.2433, pruned_loss=0.0476, over 952598.10 frames. ], batch size: 47, lr: 2.99e-03, grad_scale: 32.0 2023-04-27 22:53:36,241 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=142724.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:53:59,290 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([4.6869, 4.7401, 3.2871, 5.4552, 4.7083, 4.7925, 1.9371, 4.6732], device='cuda:6'), covar=tensor([0.1552, 0.0905, 0.2667, 0.0764, 0.3380, 0.1476, 0.5631, 0.1908], device='cuda:6'), in_proj_covar=tensor([0.0245, 0.0218, 0.0250, 0.0304, 0.0298, 0.0248, 0.0273, 0.0273], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 22:54:19,562 INFO [finetune.py:976] (6/7) Epoch 25, batch 5300, loss[loss=0.1609, simple_loss=0.2213, pruned_loss=0.05028, over 4579.00 frames. ], tot_loss[loss=0.1712, simple_loss=0.245, pruned_loss=0.04866, over 953538.66 frames. ], batch size: 20, lr: 2.99e-03, grad_scale: 32.0 2023-04-27 22:54:22,611 INFO [optim.py:369] (6/7) 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 22:54:44,805 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=142785.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:55:27,672 INFO [finetune.py:976] (6/7) Epoch 25, batch 5350, loss[loss=0.115, simple_loss=0.1954, pruned_loss=0.01731, over 4721.00 frames. ], tot_loss[loss=0.1698, simple_loss=0.2438, pruned_loss=0.0479, over 952191.22 frames. ], batch size: 59, lr: 2.99e-03, grad_scale: 32.0 2023-04-27 22:56:34,015 INFO [finetune.py:976] (6/7) Epoch 25, batch 5400, loss[loss=0.1862, simple_loss=0.244, pruned_loss=0.06424, over 4767.00 frames. ], tot_loss[loss=0.1674, simple_loss=0.2405, pruned_loss=0.04715, over 949623.39 frames. ], batch size: 26, lr: 2.99e-03, grad_scale: 32.0 2023-04-27 22:56:42,414 INFO [optim.py:369] (6/7) 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 22:57:18,642 INFO [finetune.py:976] (6/7) Epoch 25, batch 5450, loss[loss=0.1418, simple_loss=0.196, pruned_loss=0.04383, over 4055.00 frames. ], tot_loss[loss=0.1656, simple_loss=0.2376, pruned_loss=0.04675, over 948976.59 frames. ], batch size: 17, lr: 2.99e-03, grad_scale: 32.0 2023-04-27 22:57:27,223 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=142929.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:57:49,773 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.4833, 1.2932, 4.1143, 3.8537, 3.6294, 3.9137, 3.9247, 3.6686], device='cuda:6'), covar=tensor([0.7053, 0.5844, 0.1173, 0.1749, 0.1129, 0.1751, 0.1509, 0.1597], device='cuda:6'), in_proj_covar=tensor([0.0309, 0.0306, 0.0406, 0.0407, 0.0347, 0.0412, 0.0317, 0.0365], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-27 22:57:52,212 INFO [finetune.py:976] (6/7) Epoch 25, batch 5500, loss[loss=0.1275, simple_loss=0.2006, pruned_loss=0.02722, over 4745.00 frames. ], tot_loss[loss=0.1627, simple_loss=0.2341, pruned_loss=0.04558, over 950094.92 frames. ], batch size: 28, lr: 2.99e-03, grad_scale: 32.0 2023-04-27 22:57:55,644 INFO [optim.py:369] (6/7) 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 22:58:26,145 INFO [finetune.py:976] (6/7) Epoch 25, batch 5550, loss[loss=0.1384, simple_loss=0.2157, pruned_loss=0.03051, over 4903.00 frames. ], tot_loss[loss=0.1632, simple_loss=0.235, pruned_loss=0.04572, over 952730.00 frames. ], batch size: 32, lr: 2.99e-03, grad_scale: 32.0 2023-04-27 22:58:57,671 INFO [finetune.py:976] (6/7) Epoch 25, batch 5600, loss[loss=0.1982, simple_loss=0.267, pruned_loss=0.06469, over 4908.00 frames. ], tot_loss[loss=0.1675, simple_loss=0.2406, pruned_loss=0.04722, over 953194.56 frames. ], batch size: 36, lr: 2.99e-03, grad_scale: 32.0 2023-04-27 22:59:00,544 INFO [optim.py:369] (6/7) 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 22:59:06,470 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=143080.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:59:24,966 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.2012, 2.7276, 2.3769, 2.5926, 2.0567, 2.3793, 2.5241, 1.8265], device='cuda:6'), covar=tensor([0.2211, 0.1440, 0.0831, 0.1317, 0.3142, 0.1186, 0.2004, 0.2883], device='cuda:6'), in_proj_covar=tensor([0.0285, 0.0302, 0.0214, 0.0277, 0.0314, 0.0256, 0.0249, 0.0267], device='cuda:6'), out_proj_covar=tensor([1.1386e-04, 1.1909e-04, 8.4265e-05, 1.0900e-04, 1.2638e-04, 1.0098e-04, 1.0048e-04, 1.0535e-04], device='cuda:6') 2023-04-27 22:59:27,655 INFO [finetune.py:976] (6/7) Epoch 25, batch 5650, loss[loss=0.1789, simple_loss=0.2584, pruned_loss=0.04971, over 4807.00 frames. ], tot_loss[loss=0.1689, simple_loss=0.2431, pruned_loss=0.04732, over 955440.19 frames. ], batch size: 51, lr: 2.99e-03, grad_scale: 32.0 2023-04-27 22:59:29,528 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.4088, 3.0065, 0.9676, 1.8888, 2.3440, 1.6507, 4.4324, 2.4619], device='cuda:6'), covar=tensor([0.0721, 0.0744, 0.0935, 0.1277, 0.0565, 0.1023, 0.0192, 0.0524], device='cuda:6'), in_proj_covar=tensor([0.0050, 0.0064, 0.0047, 0.0046, 0.0049, 0.0051, 0.0073, 0.0050], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:6') 2023-04-27 23:00:02,743 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=2.10 vs. limit=2.0 2023-04-27 23:00:12,817 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5749, 3.2707, 1.1217, 1.9148, 1.8467, 2.2968, 2.0493, 1.1880], device='cuda:6'), covar=tensor([0.1547, 0.1518, 0.2204, 0.1453, 0.1208, 0.1313, 0.1668, 0.1907], device='cuda:6'), in_proj_covar=tensor([0.0118, 0.0239, 0.0136, 0.0121, 0.0132, 0.0152, 0.0117, 0.0119], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-27 23:00:24,805 INFO [finetune.py:976] (6/7) Epoch 25, batch 5700, loss[loss=0.1238, simple_loss=0.1971, pruned_loss=0.02528, over 3929.00 frames. ], tot_loss[loss=0.1654, simple_loss=0.2384, pruned_loss=0.04619, over 934214.86 frames. ], batch size: 17, lr: 2.98e-03, grad_scale: 32.0 2023-04-27 23:00:27,756 INFO [optim.py:369] (6/7) 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] (6/7) Epoch 26, batch 0, loss[loss=0.1653, simple_loss=0.2485, pruned_loss=0.04104, over 4894.00 frames. ], tot_loss[loss=0.1653, simple_loss=0.2485, pruned_loss=0.04104, over 4894.00 frames. ], batch size: 46, lr: 2.98e-03, grad_scale: 32.0 2023-04-27 23:01:04,602 INFO [finetune.py:1001] (6/7) Computing validation loss 2023-04-27 23:01:12,704 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8751, 2.1741, 1.8995, 2.1319, 1.7314, 1.8102, 1.7690, 1.4703], device='cuda:6'), covar=tensor([0.1554, 0.1227, 0.0767, 0.1075, 0.3240, 0.1093, 0.1778, 0.2272], device='cuda:6'), in_proj_covar=tensor([0.0284, 0.0302, 0.0213, 0.0276, 0.0313, 0.0256, 0.0248, 0.0266], device='cuda:6'), out_proj_covar=tensor([1.1349e-04, 1.1920e-04, 8.3963e-05, 1.0865e-04, 1.2629e-04, 1.0072e-04, 1.0005e-04, 1.0505e-04], device='cuda:6') 2023-04-27 23:01:21,431 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.3787, 1.2711, 1.6204, 1.6005, 1.2832, 1.2293, 1.3520, 0.8580], device='cuda:6'), covar=tensor([0.0489, 0.0582, 0.0386, 0.0548, 0.0693, 0.0994, 0.0537, 0.0542], device='cuda:6'), in_proj_covar=tensor([0.0069, 0.0067, 0.0066, 0.0068, 0.0073, 0.0094, 0.0072, 0.0063], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-27 23:01:24,405 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.2686, 1.5766, 1.8519, 1.9464, 1.9061, 2.0292, 1.8519, 1.8695], device='cuda:6'), covar=tensor([0.3804, 0.5495, 0.4501, 0.5150, 0.5796, 0.6929, 0.5393, 0.4914], device='cuda:6'), in_proj_covar=tensor([0.0341, 0.0375, 0.0328, 0.0341, 0.0349, 0.0395, 0.0360, 0.0332], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 23:01:26,498 INFO [finetune.py:1010] (6/7) 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] (6/7) Maximum memory allocated so far is 6435MB 2023-04-27 23:02:03,872 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7605, 1.1136, 1.8168, 2.1864, 1.8193, 1.7193, 1.8095, 1.7486], device='cuda:6'), covar=tensor([0.4722, 0.7084, 0.6610, 0.6110, 0.5937, 0.7988, 0.8293, 0.9144], device='cuda:6'), in_proj_covar=tensor([0.0440, 0.0420, 0.0514, 0.0507, 0.0468, 0.0502, 0.0505, 0.0517], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 23:02:16,871 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=143229.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:02:34,936 INFO [finetune.py:976] (6/7) Epoch 26, batch 50, loss[loss=0.1538, simple_loss=0.2362, pruned_loss=0.03566, over 4776.00 frames. ], tot_loss[loss=0.1705, simple_loss=0.244, pruned_loss=0.04851, over 215338.47 frames. ], batch size: 51, lr: 2.98e-03, grad_scale: 32.0 2023-04-27 23:03:09,818 INFO [optim.py:369] (6/7) 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 23:03:19,709 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=143277.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:03:40,347 INFO [finetune.py:976] (6/7) Epoch 26, batch 100, loss[loss=0.1674, simple_loss=0.242, pruned_loss=0.04636, over 4735.00 frames. ], tot_loss[loss=0.1653, simple_loss=0.2382, pruned_loss=0.04623, over 379856.59 frames. ], batch size: 54, lr: 2.98e-03, grad_scale: 32.0 2023-04-27 23:03:54,530 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.1067, 2.5015, 1.1137, 1.4859, 1.9155, 1.2143, 3.1568, 1.6387], device='cuda:6'), covar=tensor([0.0633, 0.0622, 0.0732, 0.1092, 0.0459, 0.0951, 0.0231, 0.0574], device='cuda:6'), in_proj_covar=tensor([0.0050, 0.0064, 0.0047, 0.0046, 0.0049, 0.0051, 0.0073, 0.0050], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:6') 2023-04-27 23:04:19,088 INFO [finetune.py:976] (6/7) Epoch 26, batch 150, loss[loss=0.1721, simple_loss=0.2334, pruned_loss=0.05536, over 4833.00 frames. ], tot_loss[loss=0.1653, simple_loss=0.2359, pruned_loss=0.04732, over 509039.14 frames. ], batch size: 33, lr: 2.98e-03, grad_scale: 32.0 2023-04-27 23:04:26,572 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.89 vs. limit=5.0 2023-04-27 23:04:37,629 INFO [optim.py:369] (6/7) 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 23:04:43,892 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 26, batch 200, loss[loss=0.174, simple_loss=0.2426, pruned_loss=0.05268, over 4819.00 frames. ], tot_loss[loss=0.163, simple_loss=0.233, pruned_loss=0.04645, over 608741.12 frames. ], batch size: 39, lr: 2.98e-03, grad_scale: 32.0 2023-04-27 23:04:59,041 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=143400.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 23:05:16,580 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=143428.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:05:16,661 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.4434, 1.9101, 2.3621, 2.9807, 2.3545, 1.8160, 1.8068, 2.3123], device='cuda:6'), covar=tensor([0.3155, 0.3117, 0.1682, 0.2201, 0.2729, 0.2864, 0.3533, 0.1906], device='cuda:6'), in_proj_covar=tensor([0.0293, 0.0247, 0.0229, 0.0315, 0.0221, 0.0235, 0.0227, 0.0187], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:6') 2023-04-27 23:05:31,682 INFO [finetune.py:976] (6/7) Epoch 26, batch 250, loss[loss=0.131, simple_loss=0.2066, pruned_loss=0.02768, over 4786.00 frames. ], tot_loss[loss=0.1658, simple_loss=0.237, pruned_loss=0.04725, over 686840.21 frames. ], batch size: 26, lr: 2.98e-03, grad_scale: 32.0 2023-04-27 23:05:44,908 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=143461.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 23:05:47,384 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.0370, 2.0293, 1.7607, 1.7773, 2.2079, 1.7965, 2.6928, 1.6298], device='cuda:6'), covar=tensor([0.3469, 0.2080, 0.4401, 0.2796, 0.1658, 0.2453, 0.1335, 0.4026], device='cuda:6'), in_proj_covar=tensor([0.0337, 0.0351, 0.0423, 0.0348, 0.0380, 0.0374, 0.0367, 0.0422], device='cuda:6'), out_proj_covar=tensor([9.9711e-05, 1.0461e-04, 1.2805e-04, 1.0466e-04, 1.1269e-04, 1.1139e-04, 1.0755e-04, 1.2697e-04], device='cuda:6') 2023-04-27 23:05:50,285 INFO [optim.py:369] (6/7) 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 23:05:58,099 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=143473.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 23:06:15,188 INFO [finetune.py:976] (6/7) Epoch 26, batch 300, loss[loss=0.1552, simple_loss=0.2417, pruned_loss=0.03437, over 4840.00 frames. ], tot_loss[loss=0.1697, simple_loss=0.242, pruned_loss=0.04874, over 746878.76 frames. ], batch size: 49, lr: 2.98e-03, grad_scale: 32.0 2023-04-27 23:06:30,776 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.6377, 3.2241, 2.6574, 2.9315, 2.0394, 2.8512, 2.9508, 2.2813], device='cuda:6'), covar=tensor([0.1904, 0.1431, 0.0910, 0.1385, 0.3632, 0.1111, 0.1666, 0.2667], device='cuda:6'), in_proj_covar=tensor([0.0286, 0.0302, 0.0214, 0.0277, 0.0315, 0.0256, 0.0250, 0.0267], device='cuda:6'), out_proj_covar=tensor([1.1415e-04, 1.1927e-04, 8.4152e-05, 1.0908e-04, 1.2682e-04, 1.0058e-04, 1.0063e-04, 1.0539e-04], device='cuda:6') 2023-04-27 23:06:43,069 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=143534.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 23:06:48,238 INFO [finetune.py:976] (6/7) Epoch 26, batch 350, loss[loss=0.2033, simple_loss=0.2682, pruned_loss=0.06921, over 4795.00 frames. ], tot_loss[loss=0.1717, simple_loss=0.2446, pruned_loss=0.04941, over 794911.03 frames. ], batch size: 51, lr: 2.98e-03, grad_scale: 32.0 2023-04-27 23:07:08,175 INFO [optim.py:369] (6/7) 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 23:07:15,562 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([4.7943, 4.7804, 3.1794, 5.4965, 4.8438, 4.7891, 2.1483, 4.7531], device='cuda:6'), covar=tensor([0.1730, 0.0875, 0.2840, 0.0845, 0.3478, 0.1567, 0.5493, 0.2095], device='cuda:6'), in_proj_covar=tensor([0.0245, 0.0218, 0.0252, 0.0305, 0.0298, 0.0247, 0.0273, 0.0274], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 23:07:22,102 INFO [finetune.py:976] (6/7) Epoch 26, batch 400, loss[loss=0.1653, simple_loss=0.2498, pruned_loss=0.04046, over 4816.00 frames. ], tot_loss[loss=0.1715, simple_loss=0.2453, pruned_loss=0.04889, over 831464.14 frames. ], batch size: 38, lr: 2.98e-03, grad_scale: 32.0 2023-04-27 23:07:55,516 INFO [finetune.py:976] (6/7) Epoch 26, batch 450, loss[loss=0.1328, simple_loss=0.2061, pruned_loss=0.02979, over 4796.00 frames. ], tot_loss[loss=0.1696, simple_loss=0.2431, pruned_loss=0.04802, over 858480.72 frames. ], batch size: 29, lr: 2.98e-03, grad_scale: 32.0 2023-04-27 23:08:20,553 INFO [optim.py:369] (6/7) 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 23:08:43,966 INFO [finetune.py:976] (6/7) Epoch 26, batch 500, loss[loss=0.1275, simple_loss=0.2024, pruned_loss=0.02636, over 4785.00 frames. ], tot_loss[loss=0.1676, simple_loss=0.2406, pruned_loss=0.04734, over 881858.58 frames. ], batch size: 26, lr: 2.98e-03, grad_scale: 32.0 2023-04-27 23:09:17,408 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7118, 2.0472, 1.6368, 1.4944, 1.2837, 1.2682, 1.7502, 1.2338], device='cuda:6'), covar=tensor([0.1577, 0.1325, 0.1435, 0.1727, 0.2237, 0.1937, 0.0910, 0.1977], device='cuda:6'), in_proj_covar=tensor([0.0197, 0.0209, 0.0168, 0.0203, 0.0200, 0.0184, 0.0156, 0.0186], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:6') 2023-04-27 23:09:27,563 INFO [finetune.py:976] (6/7) Epoch 26, batch 550, loss[loss=0.1656, simple_loss=0.2393, pruned_loss=0.04601, over 4823.00 frames. ], tot_loss[loss=0.1658, simple_loss=0.2383, pruned_loss=0.0467, over 897947.01 frames. ], batch size: 39, lr: 2.98e-03, grad_scale: 32.0 2023-04-27 23:09:37,711 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=143756.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 23:09:47,114 INFO [optim.py:369] (6/7) 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 23:10:00,518 INFO [finetune.py:976] (6/7) Epoch 26, batch 600, loss[loss=0.169, simple_loss=0.2627, pruned_loss=0.03761, over 4836.00 frames. ], tot_loss[loss=0.1672, simple_loss=0.2396, pruned_loss=0.04737, over 911323.80 frames. ], batch size: 47, lr: 2.98e-03, grad_scale: 32.0 2023-04-27 23:10:03,635 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.9063, 2.8887, 2.2518, 3.3343, 2.9783, 2.9226, 1.1960, 2.8378], device='cuda:6'), covar=tensor([0.2261, 0.1726, 0.3291, 0.2862, 0.4296, 0.2199, 0.6188, 0.2950], device='cuda:6'), in_proj_covar=tensor([0.0245, 0.0218, 0.0251, 0.0305, 0.0297, 0.0246, 0.0273, 0.0273], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 23:10:25,723 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=143829.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 23:10:33,616 INFO [finetune.py:976] (6/7) Epoch 26, batch 650, loss[loss=0.1997, simple_loss=0.2808, pruned_loss=0.05932, over 4826.00 frames. ], tot_loss[loss=0.1692, simple_loss=0.2422, pruned_loss=0.04812, over 919312.97 frames. ], batch size: 38, lr: 2.98e-03, grad_scale: 32.0 2023-04-27 23:11:14,500 INFO [optim.py:369] (6/7) 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 23:11:39,524 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6939, 1.7915, 0.8042, 1.3347, 1.7725, 1.5472, 1.3922, 1.5074], device='cuda:6'), covar=tensor([0.0507, 0.0371, 0.0347, 0.0569, 0.0281, 0.0539, 0.0540, 0.0578], device='cuda:6'), in_proj_covar=tensor([0.0028, 0.0023, 0.0022, 0.0028, 0.0019, 0.0027, 0.0027, 0.0028], device='cuda:6'), out_proj_covar=tensor([0.0051, 0.0046, 0.0038, 0.0053, 0.0039, 0.0050, 0.0050, 0.0052], device='cuda:6') 2023-04-27 23:11:40,024 INFO [finetune.py:976] (6/7) Epoch 26, batch 700, loss[loss=0.1564, simple_loss=0.2445, pruned_loss=0.03418, over 4781.00 frames. ], tot_loss[loss=0.1698, simple_loss=0.2436, pruned_loss=0.04802, over 927606.10 frames. ], batch size: 25, lr: 2.98e-03, grad_scale: 64.0 2023-04-27 23:12:45,799 INFO [finetune.py:976] (6/7) Epoch 26, batch 750, loss[loss=0.1775, simple_loss=0.2502, pruned_loss=0.05241, over 4823.00 frames. ], tot_loss[loss=0.1698, simple_loss=0.2439, pruned_loss=0.04778, over 933898.05 frames. ], batch size: 38, lr: 2.98e-03, grad_scale: 64.0 2023-04-27 23:13:02,620 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6763, 1.3402, 1.8392, 2.2063, 1.8143, 1.6720, 1.8089, 1.7166], device='cuda:6'), covar=tensor([0.4708, 0.6782, 0.6068, 0.5747, 0.5681, 0.7986, 0.7866, 0.9449], device='cuda:6'), in_proj_covar=tensor([0.0443, 0.0424, 0.0518, 0.0511, 0.0471, 0.0507, 0.0509, 0.0521], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 23:13:26,583 INFO [optim.py:369] (6/7) 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 23:13:56,907 INFO [finetune.py:976] (6/7) Epoch 26, batch 800, loss[loss=0.1436, simple_loss=0.2231, pruned_loss=0.032, over 4735.00 frames. ], tot_loss[loss=0.1687, simple_loss=0.2431, pruned_loss=0.04711, over 937091.02 frames. ], batch size: 23, lr: 2.98e-03, grad_scale: 64.0 2023-04-27 23:13:58,266 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9520, 2.5088, 1.8724, 1.8648, 1.4118, 1.4446, 2.0133, 1.3090], device='cuda:6'), covar=tensor([0.1645, 0.1333, 0.1387, 0.1658, 0.2266, 0.1870, 0.0955, 0.2075], device='cuda:6'), in_proj_covar=tensor([0.0198, 0.0209, 0.0168, 0.0204, 0.0201, 0.0185, 0.0157, 0.0187], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:6') 2023-04-27 23:14:51,267 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6771, 1.4024, 0.6056, 1.2955, 1.4336, 1.5249, 1.4093, 1.4159], device='cuda:6'), covar=tensor([0.0480, 0.0382, 0.0362, 0.0562, 0.0287, 0.0501, 0.0508, 0.0551], device='cuda:6'), in_proj_covar=tensor([0.0028, 0.0023, 0.0022, 0.0028, 0.0019, 0.0027, 0.0027, 0.0028], device='cuda:6'), out_proj_covar=tensor([0.0051, 0.0046, 0.0038, 0.0053, 0.0038, 0.0050, 0.0050, 0.0052], device='cuda:6') 2023-04-27 23:15:01,727 INFO [finetune.py:976] (6/7) Epoch 26, batch 850, loss[loss=0.1415, simple_loss=0.2154, pruned_loss=0.03376, over 4824.00 frames. ], tot_loss[loss=0.1672, simple_loss=0.2411, pruned_loss=0.0467, over 942483.85 frames. ], batch size: 33, lr: 2.98e-03, grad_scale: 64.0 2023-04-27 23:15:15,465 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=144056.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 23:15:35,342 INFO [optim.py:369] (6/7) 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 23:16:05,583 INFO [finetune.py:976] (6/7) Epoch 26, batch 900, loss[loss=0.1713, simple_loss=0.2417, pruned_loss=0.05049, over 4821.00 frames. ], tot_loss[loss=0.1655, simple_loss=0.2385, pruned_loss=0.04626, over 946409.71 frames. ], batch size: 38, lr: 2.97e-03, grad_scale: 32.0 2023-04-27 23:16:06,941 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9245, 2.2774, 1.8417, 1.6840, 1.3917, 1.4493, 1.9518, 1.3145], device='cuda:6'), covar=tensor([0.1697, 0.1397, 0.1471, 0.1751, 0.2350, 0.1972, 0.1000, 0.2114], device='cuda:6'), in_proj_covar=tensor([0.0197, 0.0209, 0.0168, 0.0204, 0.0200, 0.0184, 0.0157, 0.0186], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:6') 2023-04-27 23:16:12,932 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=144104.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 23:16:16,903 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.49 vs. limit=2.0 2023-04-27 23:16:17,170 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7031, 1.5642, 0.6229, 1.3445, 1.7159, 1.5445, 1.4315, 1.4790], device='cuda:6'), covar=tensor([0.0491, 0.0374, 0.0367, 0.0550, 0.0256, 0.0522, 0.0492, 0.0578], device='cuda:6'), in_proj_covar=tensor([0.0028, 0.0023, 0.0022, 0.0028, 0.0019, 0.0027, 0.0027, 0.0028], device='cuda:6'), out_proj_covar=tensor([0.0052, 0.0046, 0.0038, 0.0053, 0.0039, 0.0050, 0.0050, 0.0052], device='cuda:6') 2023-04-27 23:16:34,963 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=144129.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 23:16:53,905 INFO [finetune.py:976] (6/7) Epoch 26, batch 950, loss[loss=0.1659, simple_loss=0.2277, pruned_loss=0.05204, over 4909.00 frames. ], tot_loss[loss=0.1646, simple_loss=0.2371, pruned_loss=0.04611, over 949179.06 frames. ], batch size: 37, lr: 2.97e-03, grad_scale: 32.0 2023-04-27 23:17:06,496 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.70 vs. limit=2.0 2023-04-27 23:17:28,112 INFO [optim.py:369] (6/7) 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 23:17:37,855 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=144177.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 23:17:50,085 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0835, 1.5425, 1.9210, 2.2015, 1.9499, 1.5444, 1.2100, 1.7288], device='cuda:6'), covar=tensor([0.2661, 0.2869, 0.1457, 0.1778, 0.2134, 0.2422, 0.3967, 0.1763], device='cuda:6'), in_proj_covar=tensor([0.0291, 0.0245, 0.0228, 0.0312, 0.0220, 0.0234, 0.0226, 0.0186], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:6') 2023-04-27 23:17:59,697 INFO [finetune.py:976] (6/7) Epoch 26, batch 1000, loss[loss=0.1833, simple_loss=0.2635, pruned_loss=0.05154, over 4854.00 frames. ], tot_loss[loss=0.1663, simple_loss=0.2389, pruned_loss=0.04682, over 949104.90 frames. ], batch size: 44, lr: 2.97e-03, grad_scale: 32.0 2023-04-27 23:18:10,338 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.2980, 1.9142, 2.2426, 2.5919, 2.6417, 2.0469, 1.8430, 2.2751], device='cuda:6'), covar=tensor([0.0872, 0.1158, 0.0732, 0.0603, 0.0617, 0.0968, 0.0812, 0.0601], device='cuda:6'), in_proj_covar=tensor([0.0186, 0.0204, 0.0186, 0.0172, 0.0178, 0.0178, 0.0152, 0.0179], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 23:18:16,691 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.76 vs. limit=2.0 2023-04-27 23:18:32,531 INFO [finetune.py:976] (6/7) Epoch 26, batch 1050, loss[loss=0.2132, simple_loss=0.2844, pruned_loss=0.07096, over 4874.00 frames. ], tot_loss[loss=0.1683, simple_loss=0.2418, pruned_loss=0.04741, over 951262.37 frames. ], batch size: 34, lr: 2.97e-03, grad_scale: 32.0 2023-04-27 23:18:51,248 INFO [optim.py:369] (6/7) 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 23:19:06,554 INFO [finetune.py:976] (6/7) Epoch 26, batch 1100, loss[loss=0.1793, simple_loss=0.2542, pruned_loss=0.05224, over 4735.00 frames. ], tot_loss[loss=0.17, simple_loss=0.2438, pruned_loss=0.04806, over 949463.31 frames. ], batch size: 59, lr: 2.97e-03, grad_scale: 32.0 2023-04-27 23:19:39,804 INFO [finetune.py:976] (6/7) Epoch 26, batch 1150, loss[loss=0.1746, simple_loss=0.2463, pruned_loss=0.05146, over 4820.00 frames. ], tot_loss[loss=0.1699, simple_loss=0.2443, pruned_loss=0.04776, over 951651.90 frames. ], batch size: 33, lr: 2.97e-03, grad_scale: 32.0 2023-04-27 23:19:59,726 INFO [optim.py:369] (6/7) 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 23:20:14,235 INFO [finetune.py:976] (6/7) Epoch 26, batch 1200, loss[loss=0.1402, simple_loss=0.2228, pruned_loss=0.02883, over 4921.00 frames. ], tot_loss[loss=0.1683, simple_loss=0.2426, pruned_loss=0.04694, over 952815.63 frames. ], batch size: 38, lr: 2.97e-03, grad_scale: 32.0 2023-04-27 23:20:32,361 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=144411.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:21:13,635 INFO [finetune.py:976] (6/7) Epoch 26, batch 1250, loss[loss=0.1442, simple_loss=0.2109, pruned_loss=0.03876, over 4795.00 frames. ], tot_loss[loss=0.1671, simple_loss=0.2407, pruned_loss=0.04676, over 953671.22 frames. ], batch size: 51, lr: 2.97e-03, grad_scale: 32.0 2023-04-27 23:21:34,350 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6306, 1.4130, 4.2972, 4.0249, 3.7063, 4.0856, 3.9894, 3.7676], device='cuda:6'), covar=tensor([0.7384, 0.5873, 0.1078, 0.1711, 0.1216, 0.1722, 0.1843, 0.1501], device='cuda:6'), in_proj_covar=tensor([0.0312, 0.0308, 0.0410, 0.0410, 0.0350, 0.0416, 0.0319, 0.0366], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-27 23:21:45,825 INFO [optim.py:369] (6/7) 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 23:21:51,783 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=144472.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:22:16,354 INFO [finetune.py:976] (6/7) Epoch 26, batch 1300, loss[loss=0.1344, simple_loss=0.2097, pruned_loss=0.0296, over 4757.00 frames. ], tot_loss[loss=0.1665, simple_loss=0.2396, pruned_loss=0.04673, over 955668.52 frames. ], batch size: 27, lr: 2.97e-03, grad_scale: 32.0 2023-04-27 23:23:06,078 INFO [finetune.py:976] (6/7) Epoch 26, batch 1350, loss[loss=0.1905, simple_loss=0.2815, pruned_loss=0.04969, over 4842.00 frames. ], tot_loss[loss=0.1666, simple_loss=0.2394, pruned_loss=0.04694, over 956919.39 frames. ], batch size: 49, lr: 2.97e-03, grad_scale: 32.0 2023-04-27 23:23:09,086 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.30 vs. limit=5.0 2023-04-27 23:23:26,215 INFO [optim.py:369] (6/7) 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 23:23:39,998 INFO [finetune.py:976] (6/7) Epoch 26, batch 1400, loss[loss=0.2031, simple_loss=0.2751, pruned_loss=0.06556, over 4826.00 frames. ], tot_loss[loss=0.1678, simple_loss=0.2413, pruned_loss=0.04709, over 957374.12 frames. ], batch size: 33, lr: 2.97e-03, grad_scale: 32.0 2023-04-27 23:23:49,551 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([3.4094, 3.3314, 2.4664, 3.8868, 3.4161, 3.3350, 1.4345, 3.2560], device='cuda:6'), covar=tensor([0.1937, 0.1431, 0.3446, 0.2614, 0.4283, 0.2215, 0.6215, 0.2871], device='cuda:6'), in_proj_covar=tensor([0.0246, 0.0219, 0.0253, 0.0307, 0.0300, 0.0248, 0.0274, 0.0274], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 23:24:12,864 INFO [finetune.py:976] (6/7) Epoch 26, batch 1450, loss[loss=0.2023, simple_loss=0.2708, pruned_loss=0.06689, over 4817.00 frames. ], tot_loss[loss=0.1683, simple_loss=0.2424, pruned_loss=0.04713, over 955783.78 frames. ], batch size: 33, lr: 2.97e-03, grad_scale: 32.0 2023-04-27 23:24:14,691 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.2565, 1.6877, 2.2142, 2.7255, 2.1984, 1.6666, 1.5273, 2.0321], device='cuda:6'), covar=tensor([0.2925, 0.3013, 0.1510, 0.1912, 0.2502, 0.2580, 0.3804, 0.1898], device='cuda:6'), in_proj_covar=tensor([0.0291, 0.0245, 0.0227, 0.0312, 0.0221, 0.0234, 0.0226, 0.0187], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:6') 2023-04-27 23:24:19,336 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7159, 1.5642, 1.7234, 2.0680, 2.1191, 1.6452, 1.4282, 1.8369], device='cuda:6'), covar=tensor([0.0857, 0.1226, 0.0819, 0.0512, 0.0571, 0.0883, 0.0722, 0.0580], device='cuda:6'), in_proj_covar=tensor([0.0184, 0.0203, 0.0185, 0.0171, 0.0176, 0.0178, 0.0151, 0.0178], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 23:24:30,601 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.5727, 1.9144, 2.0123, 2.1215, 1.9757, 2.0169, 2.1027, 2.0648], device='cuda:6'), covar=tensor([0.3909, 0.5287, 0.4692, 0.4502, 0.5430, 0.6969, 0.5224, 0.4741], device='cuda:6'), in_proj_covar=tensor([0.0339, 0.0375, 0.0329, 0.0340, 0.0350, 0.0395, 0.0360, 0.0333], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 23:24:33,312 INFO [optim.py:369] (6/7) 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] (6/7) Epoch 26, batch 1500, loss[loss=0.1516, simple_loss=0.217, pruned_loss=0.04307, over 4250.00 frames. ], tot_loss[loss=0.1689, simple_loss=0.2429, pruned_loss=0.04745, over 956311.06 frames. ], batch size: 66, lr: 2.97e-03, grad_scale: 32.0 2023-04-27 23:24:47,731 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.63 vs. limit=5.0 2023-04-27 23:24:48,880 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.39 vs. limit=5.0 2023-04-27 23:25:20,121 INFO [finetune.py:976] (6/7) Epoch 26, batch 1550, loss[loss=0.1402, simple_loss=0.2174, pruned_loss=0.03154, over 4713.00 frames. ], tot_loss[loss=0.1685, simple_loss=0.2427, pruned_loss=0.0472, over 955068.32 frames. ], batch size: 23, lr: 2.97e-03, grad_scale: 32.0 2023-04-27 23:25:20,208 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.4334, 1.3871, 4.2106, 3.9703, 3.6622, 4.0269, 3.9973, 3.7130], device='cuda:6'), covar=tensor([0.7081, 0.5904, 0.0998, 0.1508, 0.1093, 0.1705, 0.0915, 0.1295], device='cuda:6'), in_proj_covar=tensor([0.0311, 0.0307, 0.0408, 0.0409, 0.0349, 0.0415, 0.0318, 0.0365], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-27 23:25:42,634 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=144767.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:25:45,419 INFO [optim.py:369] (6/7) 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 23:26:14,645 INFO [finetune.py:976] (6/7) Epoch 26, batch 1600, loss[loss=0.1943, simple_loss=0.253, pruned_loss=0.06779, over 4828.00 frames. ], tot_loss[loss=0.1676, simple_loss=0.2411, pruned_loss=0.04706, over 955120.00 frames. ], batch size: 33, lr: 2.97e-03, grad_scale: 32.0 2023-04-27 23:27:00,876 INFO [finetune.py:976] (6/7) Epoch 26, batch 1650, loss[loss=0.1738, simple_loss=0.2432, pruned_loss=0.05213, over 4937.00 frames. ], tot_loss[loss=0.1664, simple_loss=0.239, pruned_loss=0.04686, over 954844.63 frames. ], batch size: 38, lr: 2.97e-03, grad_scale: 32.0 2023-04-27 23:27:20,465 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.4776, 1.7463, 1.8987, 1.9673, 1.8403, 1.8279, 1.9357, 1.8699], device='cuda:6'), covar=tensor([0.3751, 0.5215, 0.4207, 0.3987, 0.5182, 0.7317, 0.4991, 0.4813], device='cuda:6'), in_proj_covar=tensor([0.0342, 0.0378, 0.0330, 0.0342, 0.0350, 0.0397, 0.0362, 0.0335], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 23:27:20,928 INFO [optim.py:369] (6/7) 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 23:27:31,385 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-04-27 23:27:39,745 INFO [finetune.py:976] (6/7) Epoch 26, batch 1700, loss[loss=0.1099, simple_loss=0.1738, pruned_loss=0.02303, over 4089.00 frames. ], tot_loss[loss=0.1633, simple_loss=0.2355, pruned_loss=0.04558, over 957172.34 frames. ], batch size: 17, lr: 2.97e-03, grad_scale: 32.0 2023-04-27 23:28:44,465 INFO [finetune.py:976] (6/7) Epoch 26, batch 1750, loss[loss=0.1821, simple_loss=0.2565, pruned_loss=0.05385, over 3804.00 frames. ], tot_loss[loss=0.1667, simple_loss=0.239, pruned_loss=0.04717, over 955265.46 frames. ], batch size: 16, lr: 2.97e-03, grad_scale: 32.0 2023-04-27 23:28:55,657 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.78 vs. limit=2.0 2023-04-27 23:29:25,331 INFO [optim.py:369] (6/7) 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 23:29:50,039 INFO [finetune.py:976] (6/7) Epoch 26, batch 1800, loss[loss=0.1656, simple_loss=0.2575, pruned_loss=0.03689, over 4908.00 frames. ], tot_loss[loss=0.1691, simple_loss=0.2421, pruned_loss=0.04802, over 955005.23 frames. ], batch size: 37, lr: 2.97e-03, grad_scale: 32.0 2023-04-27 23:29:53,359 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=144996.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:30:15,084 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=145027.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:30:24,558 INFO [finetune.py:976] (6/7) Epoch 26, batch 1850, loss[loss=0.1728, simple_loss=0.2477, pruned_loss=0.04893, over 4820.00 frames. ], tot_loss[loss=0.1681, simple_loss=0.2416, pruned_loss=0.04736, over 952655.07 frames. ], batch size: 25, lr: 2.97e-03, grad_scale: 32.0 2023-04-27 23:30:25,788 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.3203, 1.2301, 1.5704, 1.5217, 1.1890, 1.1559, 1.2426, 0.7700], device='cuda:6'), covar=tensor([0.0496, 0.0568, 0.0319, 0.0541, 0.0683, 0.0958, 0.0547, 0.0538], device='cuda:6'), in_proj_covar=tensor([0.0070, 0.0067, 0.0066, 0.0069, 0.0075, 0.0095, 0.0073, 0.0063], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-27 23:30:26,988 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.4194, 3.2054, 2.7030, 2.9902, 2.2189, 2.7181, 2.8074, 2.1436], device='cuda:6'), covar=tensor([0.2089, 0.1166, 0.0726, 0.0999, 0.2789, 0.1055, 0.1643, 0.2392], device='cuda:6'), in_proj_covar=tensor([0.0286, 0.0301, 0.0214, 0.0277, 0.0314, 0.0256, 0.0249, 0.0265], device='cuda:6'), out_proj_covar=tensor([1.1399e-04, 1.1847e-04, 8.4338e-05, 1.0905e-04, 1.2668e-04, 1.0085e-04, 1.0031e-04, 1.0468e-04], device='cuda:6') 2023-04-27 23:30:27,664 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.59 vs. limit=2.0 2023-04-27 23:30:34,404 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=145057.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 23:30:41,229 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=145066.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:30:41,821 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=145067.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:30:44,629 INFO [optim.py:369] (6/7) 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 23:30:48,832 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6159, 1.6140, 0.7744, 1.2311, 1.7164, 1.4319, 1.3265, 1.3989], device='cuda:6'), covar=tensor([0.0472, 0.0376, 0.0336, 0.0561, 0.0265, 0.0497, 0.0489, 0.0548], device='cuda:6'), in_proj_covar=tensor([0.0028, 0.0023, 0.0022, 0.0028, 0.0019, 0.0028, 0.0027, 0.0028], device='cuda:6'), out_proj_covar=tensor([0.0052, 0.0046, 0.0038, 0.0053, 0.0039, 0.0051, 0.0050, 0.0052], device='cuda:6') 2023-04-27 23:30:51,936 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-04-27 23:30:55,907 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 26, batch 1900, loss[loss=0.1854, simple_loss=0.259, pruned_loss=0.05589, over 4796.00 frames. ], tot_loss[loss=0.1702, simple_loss=0.2436, pruned_loss=0.04839, over 951651.12 frames. ], batch size: 51, lr: 2.97e-03, grad_scale: 32.0 2023-04-27 23:31:13,801 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=145115.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:31:22,158 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 26, batch 1950, loss[loss=0.1565, simple_loss=0.2194, pruned_loss=0.04681, over 4903.00 frames. ], tot_loss[loss=0.1693, simple_loss=0.2424, pruned_loss=0.04815, over 952329.82 frames. ], batch size: 36, lr: 2.97e-03, grad_scale: 32.0 2023-04-27 23:32:03,280 INFO [optim.py:369] (6/7) 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] (6/7) Epoch 26, batch 2000, loss[loss=0.1748, simple_loss=0.2474, pruned_loss=0.05106, over 4795.00 frames. ], tot_loss[loss=0.1674, simple_loss=0.2402, pruned_loss=0.04731, over 954678.11 frames. ], batch size: 45, lr: 2.97e-03, grad_scale: 32.0 2023-04-27 23:32:43,697 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.6146, 2.2626, 2.3881, 3.0697, 2.8851, 2.3509, 2.0618, 2.6657], device='cuda:6'), covar=tensor([0.0790, 0.1085, 0.0832, 0.0520, 0.0574, 0.0839, 0.0736, 0.0573], device='cuda:6'), in_proj_covar=tensor([0.0185, 0.0203, 0.0185, 0.0171, 0.0177, 0.0178, 0.0151, 0.0179], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 23:33:00,472 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0210, 2.4031, 1.0043, 1.3321, 1.7759, 1.1291, 2.9955, 1.5781], device='cuda:6'), covar=tensor([0.0669, 0.0549, 0.0831, 0.1282, 0.0505, 0.1013, 0.0262, 0.0609], device='cuda:6'), in_proj_covar=tensor([0.0050, 0.0064, 0.0047, 0.0046, 0.0049, 0.0051, 0.0073, 0.0051], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:6') 2023-04-27 23:33:03,265 INFO [finetune.py:976] (6/7) Epoch 26, batch 2050, loss[loss=0.1521, simple_loss=0.218, pruned_loss=0.04314, over 4140.00 frames. ], tot_loss[loss=0.1653, simple_loss=0.2377, pruned_loss=0.04644, over 955390.02 frames. ], batch size: 65, lr: 2.97e-03, grad_scale: 32.0 2023-04-27 23:33:31,605 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.3613, 2.1823, 2.3131, 2.7865, 2.7017, 2.1969, 1.8902, 2.4229], device='cuda:6'), covar=tensor([0.0805, 0.1029, 0.0724, 0.0531, 0.0589, 0.0910, 0.0758, 0.0565], device='cuda:6'), in_proj_covar=tensor([0.0186, 0.0203, 0.0186, 0.0172, 0.0177, 0.0178, 0.0151, 0.0179], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 23:33:43,224 INFO [optim.py:369] (6/7) 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 23:33:43,814 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([3.4748, 3.3216, 2.6052, 3.9279, 3.3964, 3.3990, 1.4751, 3.3588], device='cuda:6'), covar=tensor([0.1862, 0.1395, 0.3245, 0.2141, 0.3803, 0.2065, 0.6002, 0.2437], device='cuda:6'), in_proj_covar=tensor([0.0245, 0.0218, 0.0251, 0.0302, 0.0299, 0.0246, 0.0273, 0.0272], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 23:33:52,633 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.2460, 2.5628, 2.1866, 2.6312, 1.8256, 2.2852, 2.2120, 1.7699], device='cuda:6'), covar=tensor([0.1805, 0.1485, 0.0735, 0.0839, 0.3162, 0.1002, 0.1604, 0.2294], device='cuda:6'), in_proj_covar=tensor([0.0284, 0.0298, 0.0213, 0.0275, 0.0312, 0.0255, 0.0247, 0.0264], device='cuda:6'), out_proj_covar=tensor([1.1324e-04, 1.1749e-04, 8.4059e-05, 1.0845e-04, 1.2567e-04, 1.0024e-04, 9.9770e-05, 1.0403e-04], device='cuda:6') 2023-04-27 23:33:58,466 INFO [finetune.py:976] (6/7) Epoch 26, batch 2100, loss[loss=0.1748, simple_loss=0.2464, pruned_loss=0.05159, over 4828.00 frames. ], tot_loss[loss=0.1654, simple_loss=0.2378, pruned_loss=0.04649, over 956382.22 frames. ], batch size: 30, lr: 2.97e-03, grad_scale: 32.0 2023-04-27 23:34:32,516 INFO [finetune.py:976] (6/7) Epoch 26, batch 2150, loss[loss=0.1676, simple_loss=0.2577, pruned_loss=0.03874, over 4836.00 frames. ], tot_loss[loss=0.1672, simple_loss=0.2402, pruned_loss=0.04715, over 956205.92 frames. ], batch size: 49, lr: 2.97e-03, grad_scale: 32.0 2023-04-27 23:34:39,184 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=145352.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 23:34:49,511 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2023-04-27 23:34:51,070 INFO [optim.py:369] (6/7) 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 23:34:51,203 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=145371.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:34:51,272 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-04-27 23:34:52,444 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0498, 1.6613, 2.1767, 2.4138, 2.0866, 1.9881, 2.0870, 2.0233], device='cuda:6'), covar=tensor([0.4590, 0.7235, 0.7058, 0.5347, 0.5931, 0.8425, 0.9201, 0.9968], device='cuda:6'), in_proj_covar=tensor([0.0437, 0.0419, 0.0513, 0.0504, 0.0467, 0.0500, 0.0503, 0.0516], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 23:34:59,919 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=145383.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:35:10,687 INFO [finetune.py:976] (6/7) Epoch 26, batch 2200, loss[loss=0.16, simple_loss=0.2358, pruned_loss=0.04207, over 4812.00 frames. ], tot_loss[loss=0.1684, simple_loss=0.242, pruned_loss=0.04744, over 956598.19 frames. ], batch size: 39, lr: 2.97e-03, grad_scale: 32.0 2023-04-27 23:35:29,172 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-04-27 23:35:30,052 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=145422.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:35:37,123 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=145432.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:35:44,058 INFO [finetune.py:976] (6/7) Epoch 26, batch 2250, loss[loss=0.1763, simple_loss=0.251, pruned_loss=0.05077, over 4826.00 frames. ], tot_loss[loss=0.1705, simple_loss=0.2438, pruned_loss=0.04867, over 957077.04 frames. ], batch size: 38, lr: 2.97e-03, grad_scale: 32.0 2023-04-27 23:35:45,825 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7199, 1.6717, 2.0053, 2.0915, 1.5849, 1.4691, 1.6720, 0.9846], device='cuda:6'), covar=tensor([0.0659, 0.0658, 0.0454, 0.0769, 0.0716, 0.0981, 0.0660, 0.0676], device='cuda:6'), in_proj_covar=tensor([0.0070, 0.0067, 0.0066, 0.0069, 0.0074, 0.0095, 0.0073, 0.0063], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-27 23:35:52,447 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.2402, 1.7090, 1.5142, 1.8795, 1.9106, 2.0232, 1.5927, 4.0529], device='cuda:6'), covar=tensor([0.0551, 0.0771, 0.0768, 0.1173, 0.0593, 0.0518, 0.0688, 0.0115], device='cuda:6'), in_proj_covar=tensor([0.0037, 0.0038, 0.0039, 0.0042, 0.0040, 0.0037, 0.0037, 0.0054], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0012, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:6') 2023-04-27 23:36:03,179 INFO [optim.py:369] (6/7) 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 23:36:17,802 INFO [finetune.py:976] (6/7) Epoch 26, batch 2300, loss[loss=0.16, simple_loss=0.2329, pruned_loss=0.0435, over 4776.00 frames. ], tot_loss[loss=0.1679, simple_loss=0.2416, pruned_loss=0.0471, over 953975.11 frames. ], batch size: 28, lr: 2.97e-03, grad_scale: 32.0 2023-04-27 23:36:49,858 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([3.7402, 3.6629, 2.6155, 4.2975, 3.7503, 3.7167, 1.5988, 3.6925], device='cuda:6'), covar=tensor([0.1935, 0.1282, 0.3170, 0.1750, 0.3984, 0.1912, 0.5893, 0.2502], device='cuda:6'), in_proj_covar=tensor([0.0246, 0.0219, 0.0254, 0.0306, 0.0301, 0.0249, 0.0275, 0.0275], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 23:36:51,032 INFO [finetune.py:976] (6/7) Epoch 26, batch 2350, loss[loss=0.1421, simple_loss=0.2104, pruned_loss=0.03695, over 4734.00 frames. ], tot_loss[loss=0.1665, simple_loss=0.2401, pruned_loss=0.04641, over 955902.78 frames. ], batch size: 23, lr: 2.96e-03, grad_scale: 32.0 2023-04-27 23:36:59,434 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.7159, 2.3689, 2.5243, 3.0662, 2.9795, 2.3267, 2.1478, 2.7073], device='cuda:6'), covar=tensor([0.0736, 0.0932, 0.0671, 0.0560, 0.0548, 0.0907, 0.0714, 0.0518], device='cuda:6'), in_proj_covar=tensor([0.0185, 0.0203, 0.0185, 0.0171, 0.0177, 0.0178, 0.0151, 0.0179], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 23:37:10,060 INFO [optim.py:369] (6/7) 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 23:37:40,484 INFO [finetune.py:976] (6/7) Epoch 26, batch 2400, loss[loss=0.1777, simple_loss=0.2474, pruned_loss=0.05398, over 4822.00 frames. ], tot_loss[loss=0.1652, simple_loss=0.2379, pruned_loss=0.04623, over 955829.95 frames. ], batch size: 51, lr: 2.96e-03, grad_scale: 32.0 2023-04-27 23:37:41,820 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=145594.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:38:12,116 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5790, 1.3482, 4.3646, 4.0972, 3.7632, 4.2239, 4.1481, 3.8406], device='cuda:6'), covar=tensor([0.7320, 0.6358, 0.1075, 0.1737, 0.1282, 0.2013, 0.1332, 0.1495], device='cuda:6'), in_proj_covar=tensor([0.0313, 0.0310, 0.0411, 0.0411, 0.0351, 0.0419, 0.0321, 0.0366], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-27 23:38:12,258 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.17 vs. limit=5.0 2023-04-27 23:38:39,058 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([3.8583, 3.8012, 2.7741, 4.4536, 3.9169, 3.8176, 1.6434, 3.7662], device='cuda:6'), covar=tensor([0.1792, 0.1422, 0.3121, 0.1758, 0.2819, 0.2072, 0.6081, 0.2372], device='cuda:6'), in_proj_covar=tensor([0.0246, 0.0220, 0.0254, 0.0306, 0.0300, 0.0249, 0.0275, 0.0274], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 23:38:46,676 INFO [finetune.py:976] (6/7) Epoch 26, batch 2450, loss[loss=0.1368, simple_loss=0.2176, pruned_loss=0.02799, over 4914.00 frames. ], tot_loss[loss=0.1627, simple_loss=0.2349, pruned_loss=0.04519, over 954219.50 frames. ], batch size: 32, lr: 2.96e-03, grad_scale: 32.0 2023-04-27 23:38:47,935 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=145643.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:39:00,543 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=145652.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 23:39:08,345 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=145655.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 23:39:30,139 INFO [optim.py:369] (6/7) 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 23:39:42,996 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=145683.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:39:48,402 INFO [finetune.py:976] (6/7) Epoch 26, batch 2500, loss[loss=0.1922, simple_loss=0.257, pruned_loss=0.06365, over 4867.00 frames. ], tot_loss[loss=0.1646, simple_loss=0.237, pruned_loss=0.04609, over 954712.41 frames. ], batch size: 31, lr: 2.96e-03, grad_scale: 32.0 2023-04-27 23:39:54,790 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=145700.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:39:57,792 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=145704.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:40:08,485 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.5358, 1.8265, 1.9918, 2.0714, 1.9153, 1.9139, 2.0596, 1.9798], device='cuda:6'), covar=tensor([0.3806, 0.5882, 0.4682, 0.4616, 0.5731, 0.6999, 0.5178, 0.4835], device='cuda:6'), in_proj_covar=tensor([0.0339, 0.0374, 0.0328, 0.0339, 0.0351, 0.0393, 0.0360, 0.0332], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 23:40:09,643 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=145722.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:40:12,708 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=145727.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:40:15,135 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=145731.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:40:22,236 INFO [finetune.py:976] (6/7) Epoch 26, batch 2550, loss[loss=0.1935, simple_loss=0.2634, pruned_loss=0.06183, over 4891.00 frames. ], tot_loss[loss=0.1668, simple_loss=0.24, pruned_loss=0.04679, over 953493.70 frames. ], batch size: 32, lr: 2.96e-03, grad_scale: 32.0 2023-04-27 23:40:41,741 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=145770.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:40:42,280 INFO [optim.py:369] (6/7) 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 23:40:47,838 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7839, 2.1160, 1.7292, 1.3671, 1.3351, 1.3521, 1.8074, 1.2820], device='cuda:6'), covar=tensor([0.1576, 0.1288, 0.1399, 0.1824, 0.2288, 0.1976, 0.1000, 0.2060], device='cuda:6'), in_proj_covar=tensor([0.0198, 0.0211, 0.0170, 0.0204, 0.0201, 0.0187, 0.0157, 0.0188], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:6') 2023-04-27 23:40:56,106 INFO [finetune.py:976] (6/7) Epoch 26, batch 2600, loss[loss=0.2041, simple_loss=0.2856, pruned_loss=0.06133, over 4819.00 frames. ], tot_loss[loss=0.1684, simple_loss=0.2422, pruned_loss=0.0473, over 956528.33 frames. ], batch size: 33, lr: 2.96e-03, grad_scale: 32.0 2023-04-27 23:41:16,307 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=145821.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 23:41:29,915 INFO [finetune.py:976] (6/7) Epoch 26, batch 2650, loss[loss=0.1747, simple_loss=0.2479, pruned_loss=0.05074, over 4195.00 frames. ], tot_loss[loss=0.1694, simple_loss=0.2433, pruned_loss=0.04774, over 955590.34 frames. ], batch size: 65, lr: 2.96e-03, grad_scale: 32.0 2023-04-27 23:41:43,985 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.0963, 1.4793, 1.2945, 1.6877, 1.5321, 1.6087, 1.3484, 3.0200], device='cuda:6'), covar=tensor([0.0654, 0.0822, 0.0835, 0.1247, 0.0648, 0.0518, 0.0761, 0.0169], device='cuda:6'), in_proj_covar=tensor([0.0037, 0.0038, 0.0039, 0.0043, 0.0040, 0.0037, 0.0038, 0.0054], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0013, 0.0013, 0.0015, 0.0013, 0.0013, 0.0013, 0.0016], device='cuda:6') 2023-04-27 23:41:49,917 INFO [optim.py:369] (6/7) 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 23:41:57,251 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=145882.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 23:42:03,155 INFO [finetune.py:976] (6/7) Epoch 26, batch 2700, loss[loss=0.1472, simple_loss=0.226, pruned_loss=0.03422, over 4813.00 frames. ], tot_loss[loss=0.1681, simple_loss=0.2423, pruned_loss=0.04699, over 954787.31 frames. ], batch size: 38, lr: 2.96e-03, grad_scale: 32.0 2023-04-27 23:42:14,014 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.84 vs. limit=5.0 2023-04-27 23:42:17,613 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=3.96 vs. limit=5.0 2023-04-27 23:42:36,792 INFO [finetune.py:976] (6/7) Epoch 26, batch 2750, loss[loss=0.1751, simple_loss=0.2401, pruned_loss=0.05501, over 4895.00 frames. ], tot_loss[loss=0.1675, simple_loss=0.2403, pruned_loss=0.04736, over 953886.95 frames. ], batch size: 43, lr: 2.96e-03, grad_scale: 32.0 2023-04-27 23:42:42,209 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=145950.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 23:42:56,788 INFO [optim.py:369] (6/7) 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] (6/7) Epoch 26, batch 2800, loss[loss=0.1614, simple_loss=0.2312, pruned_loss=0.04581, over 4851.00 frames. ], tot_loss[loss=0.165, simple_loss=0.2371, pruned_loss=0.04644, over 954919.10 frames. ], batch size: 44, lr: 2.96e-03, grad_scale: 32.0 2023-04-27 23:43:14,391 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=145999.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:43:19,471 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 2023-04-27 23:43:24,681 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([3.6063, 3.4802, 2.9137, 4.1398, 3.3522, 3.5861, 1.8466, 3.5660], device='cuda:6'), covar=tensor([0.1844, 0.1257, 0.4146, 0.1380, 0.3220, 0.1909, 0.4763, 0.2483], device='cuda:6'), in_proj_covar=tensor([0.0244, 0.0218, 0.0252, 0.0305, 0.0299, 0.0247, 0.0273, 0.0273], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 23:43:35,204 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=146027.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:43:35,215 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.4301, 1.6528, 1.8457, 2.7278, 2.8016, 2.0610, 1.7626, 2.4209], device='cuda:6'), covar=tensor([0.0770, 0.1634, 0.1098, 0.0522, 0.0542, 0.0989, 0.0936, 0.0578], device='cuda:6'), in_proj_covar=tensor([0.0184, 0.0202, 0.0184, 0.0170, 0.0176, 0.0177, 0.0150, 0.0177], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 23:43:44,744 INFO [finetune.py:976] (6/7) Epoch 26, batch 2850, loss[loss=0.1284, simple_loss=0.203, pruned_loss=0.02692, over 4764.00 frames. ], tot_loss[loss=0.1636, simple_loss=0.2354, pruned_loss=0.04591, over 955016.67 frames. ], batch size: 26, lr: 2.96e-03, grad_scale: 32.0 2023-04-27 23:44:22,216 INFO [optim.py:369] (6/7) 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 23:44:22,352 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.0620, 1.2836, 1.1444, 1.4937, 1.3798, 1.4662, 1.2518, 2.4389], device='cuda:6'), covar=tensor([0.0604, 0.0870, 0.0835, 0.1305, 0.0703, 0.0498, 0.0769, 0.0197], device='cuda:6'), in_proj_covar=tensor([0.0037, 0.0039, 0.0039, 0.0043, 0.0040, 0.0037, 0.0038, 0.0055], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0013, 0.0013, 0.0015, 0.0014, 0.0013, 0.0013, 0.0016], device='cuda:6') 2023-04-27 23:44:25,242 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=146075.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:44:43,961 INFO [finetune.py:976] (6/7) Epoch 26, batch 2900, loss[loss=0.1324, simple_loss=0.2211, pruned_loss=0.02184, over 4697.00 frames. ], tot_loss[loss=0.1656, simple_loss=0.2375, pruned_loss=0.04685, over 953864.22 frames. ], batch size: 23, lr: 2.96e-03, grad_scale: 64.0 2023-04-27 23:45:49,358 INFO [finetune.py:976] (6/7) Epoch 26, batch 2950, loss[loss=0.1541, simple_loss=0.2206, pruned_loss=0.04377, over 4116.00 frames. ], tot_loss[loss=0.1664, simple_loss=0.2395, pruned_loss=0.04664, over 954814.32 frames. ], batch size: 17, lr: 2.96e-03, grad_scale: 64.0 2023-04-27 23:45:59,514 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=146149.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:45:59,819 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.35 vs. limit=5.0 2023-04-27 23:46:06,696 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.5087, 1.8109, 1.7353, 2.3591, 2.4945, 2.0825, 2.0123, 1.8044], device='cuda:6'), covar=tensor([0.1690, 0.1661, 0.1728, 0.1362, 0.1199, 0.1637, 0.2076, 0.2140], device='cuda:6'), in_proj_covar=tensor([0.0311, 0.0307, 0.0348, 0.0283, 0.0325, 0.0301, 0.0296, 0.0371], device='cuda:6'), out_proj_covar=tensor([6.3431e-05, 6.3254e-05, 7.3218e-05, 5.6700e-05, 6.6610e-05, 6.2871e-05, 6.1295e-05, 7.8730e-05], device='cuda:6') 2023-04-27 23:46:13,680 INFO [optim.py:369] (6/7) 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 23:46:18,904 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=146177.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 23:46:28,438 INFO [finetune.py:976] (6/7) Epoch 26, batch 3000, loss[loss=0.2148, simple_loss=0.2784, pruned_loss=0.07561, over 4262.00 frames. ], tot_loss[loss=0.1692, simple_loss=0.2421, pruned_loss=0.04815, over 954453.27 frames. ], batch size: 65, lr: 2.96e-03, grad_scale: 64.0 2023-04-27 23:46:28,438 INFO [finetune.py:1001] (6/7) Computing validation loss 2023-04-27 23:46:31,770 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.2792, 2.5641, 1.1033, 1.4659, 2.1145, 1.3267, 3.1263, 1.6950], device='cuda:6'), covar=tensor([0.0589, 0.0634, 0.0675, 0.1137, 0.0376, 0.0852, 0.0261, 0.0552], device='cuda:6'), in_proj_covar=tensor([0.0050, 0.0064, 0.0047, 0.0046, 0.0049, 0.0051, 0.0073, 0.0051], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:6') 2023-04-27 23:46:38,926 INFO [finetune.py:1010] (6/7) Epoch 26, validation: loss=0.1526, simple_loss=0.2216, pruned_loss=0.04183, over 2265189.00 frames. 2023-04-27 23:46:38,926 INFO [finetune.py:1011] (6/7) Maximum memory allocated so far is 6435MB 2023-04-27 23:46:50,697 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=146210.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:47:11,572 INFO [finetune.py:976] (6/7) Epoch 26, batch 3050, loss[loss=0.1555, simple_loss=0.2065, pruned_loss=0.05226, over 4171.00 frames. ], tot_loss[loss=0.1697, simple_loss=0.2436, pruned_loss=0.04792, over 954738.82 frames. ], batch size: 18, lr: 2.96e-03, grad_scale: 64.0 2023-04-27 23:47:17,858 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=146250.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:47:31,030 INFO [optim.py:369] (6/7) 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 23:47:42,165 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([3.7880, 3.7990, 2.8352, 4.4523, 3.8849, 3.8241, 1.6076, 3.7986], device='cuda:6'), covar=tensor([0.2126, 0.1326, 0.3190, 0.1694, 0.3119, 0.2005, 0.6207, 0.2276], device='cuda:6'), in_proj_covar=tensor([0.0244, 0.0219, 0.0252, 0.0305, 0.0299, 0.0246, 0.0274, 0.0273], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 23:47:45,142 INFO [finetune.py:976] (6/7) Epoch 26, batch 3100, loss[loss=0.1464, simple_loss=0.2248, pruned_loss=0.03402, over 4930.00 frames. ], tot_loss[loss=0.1694, simple_loss=0.2425, pruned_loss=0.04816, over 953206.32 frames. ], batch size: 38, lr: 2.96e-03, grad_scale: 32.0 2023-04-27 23:47:49,796 INFO [zipformer.py:1188] (6/7) 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] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=146299.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:47:55,255 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8948, 1.9635, 2.3071, 2.5073, 1.7139, 1.5827, 1.9186, 1.1292], device='cuda:6'), covar=tensor([0.0607, 0.0572, 0.0394, 0.0648, 0.0695, 0.1095, 0.0602, 0.0679], device='cuda:6'), in_proj_covar=tensor([0.0070, 0.0067, 0.0066, 0.0068, 0.0074, 0.0094, 0.0072, 0.0063], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-27 23:48:18,965 INFO [finetune.py:976] (6/7) Epoch 26, batch 3150, loss[loss=0.1383, simple_loss=0.2164, pruned_loss=0.03003, over 4927.00 frames. ], tot_loss[loss=0.1666, simple_loss=0.2389, pruned_loss=0.04711, over 953221.84 frames. ], batch size: 37, lr: 2.96e-03, grad_scale: 32.0 2023-04-27 23:48:22,567 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=146347.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:48:38,470 INFO [optim.py:369] (6/7) 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 23:48:51,065 INFO [finetune.py:976] (6/7) Epoch 26, batch 3200, loss[loss=0.141, simple_loss=0.2129, pruned_loss=0.03457, over 4697.00 frames. ], tot_loss[loss=0.1661, simple_loss=0.2375, pruned_loss=0.04736, over 953330.88 frames. ], batch size: 23, lr: 2.96e-03, grad_scale: 32.0 2023-04-27 23:49:24,698 INFO [finetune.py:976] (6/7) Epoch 26, batch 3250, loss[loss=0.1978, simple_loss=0.2786, pruned_loss=0.05846, over 4910.00 frames. ], tot_loss[loss=0.1654, simple_loss=0.2373, pruned_loss=0.04673, over 953943.83 frames. ], batch size: 37, lr: 2.96e-03, grad_scale: 32.0 2023-04-27 23:49:45,352 INFO [optim.py:369] (6/7) 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 23:49:53,886 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=146477.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 23:50:12,999 INFO [finetune.py:976] (6/7) Epoch 26, batch 3300, loss[loss=0.1494, simple_loss=0.2246, pruned_loss=0.03706, over 4758.00 frames. ], tot_loss[loss=0.1684, simple_loss=0.2411, pruned_loss=0.04789, over 953578.74 frames. ], batch size: 26, lr: 2.96e-03, grad_scale: 32.0 2023-04-27 23:50:28,282 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=146505.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:50:57,305 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=146525.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 23:51:19,415 INFO [finetune.py:976] (6/7) Epoch 26, batch 3350, loss[loss=0.1779, simple_loss=0.2347, pruned_loss=0.06056, over 4774.00 frames. ], tot_loss[loss=0.169, simple_loss=0.2422, pruned_loss=0.04791, over 955218.69 frames. ], batch size: 26, lr: 2.96e-03, grad_scale: 32.0 2023-04-27 23:51:39,291 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.4373, 1.3452, 1.6073, 1.6699, 1.3236, 1.1608, 1.1909, 0.6812], device='cuda:6'), covar=tensor([0.0571, 0.0620, 0.0452, 0.0486, 0.0759, 0.1614, 0.0621, 0.0723], device='cuda:6'), in_proj_covar=tensor([0.0070, 0.0067, 0.0066, 0.0069, 0.0074, 0.0095, 0.0072, 0.0063], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-27 23:51:45,685 INFO [optim.py:369] (6/7) 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 23:51:57,904 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=146591.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:51:58,397 INFO [finetune.py:976] (6/7) Epoch 26, batch 3400, loss[loss=0.1827, simple_loss=0.243, pruned_loss=0.06117, over 4243.00 frames. ], tot_loss[loss=0.1711, simple_loss=0.2446, pruned_loss=0.04886, over 955204.02 frames. ], batch size: 66, lr: 2.96e-03, grad_scale: 32.0 2023-04-27 23:51:59,159 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0152, 1.7252, 2.2180, 2.4819, 2.0887, 2.0449, 2.1925, 2.0441], device='cuda:6'), covar=tensor([0.5007, 0.7552, 0.6865, 0.5582, 0.6262, 0.8527, 0.8769, 1.0774], device='cuda:6'), in_proj_covar=tensor([0.0441, 0.0421, 0.0515, 0.0507, 0.0470, 0.0506, 0.0507, 0.0519], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-27 23:52:15,072 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.4280, 1.7964, 1.8947, 1.9528, 1.8166, 1.8691, 1.9442, 1.8859], device='cuda:6'), covar=tensor([0.3824, 0.5469, 0.4378, 0.4208, 0.5872, 0.7337, 0.5297, 0.4915], device='cuda:6'), in_proj_covar=tensor([0.0340, 0.0376, 0.0329, 0.0340, 0.0350, 0.0395, 0.0359, 0.0332], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-27 23:52:27,939 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=146626.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:52:28,492 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.1310, 1.0865, 4.8357, 4.2637, 4.1580, 4.4838, 4.1473, 3.9421], device='cuda:6'), covar=tensor([0.8985, 0.9148, 0.1141, 0.2808, 0.1871, 0.2667, 0.2926, 0.2823], device='cuda:6'), in_proj_covar=tensor([0.0313, 0.0310, 0.0410, 0.0411, 0.0350, 0.0416, 0.0320, 0.0366], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-27 23:52:43,201 INFO [finetune.py:976] (6/7) Epoch 26, batch 3450, loss[loss=0.1861, simple_loss=0.2556, pruned_loss=0.05834, over 4910.00 frames. ], tot_loss[loss=0.1712, simple_loss=0.2445, pruned_loss=0.04897, over 954823.65 frames. ], batch size: 36, lr: 2.96e-03, grad_scale: 32.0 2023-04-27 23:52:49,827 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=146652.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:53:04,801 INFO [optim.py:369] (6/7) 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 23:53:13,984 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=146687.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 23:53:16,856 INFO [finetune.py:976] (6/7) Epoch 26, batch 3500, loss[loss=0.1876, simple_loss=0.2541, pruned_loss=0.0606, over 4906.00 frames. ], tot_loss[loss=0.169, simple_loss=0.2419, pruned_loss=0.04804, over 956557.44 frames. ], batch size: 46, lr: 2.96e-03, grad_scale: 32.0 2023-04-27 23:53:39,790 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.1668, 2.8198, 2.2575, 2.3614, 1.5643, 1.5702, 2.4201, 1.5675], device='cuda:6'), covar=tensor([0.1835, 0.1557, 0.1435, 0.1700, 0.2520, 0.2137, 0.0963, 0.2135], device='cuda:6'), in_proj_covar=tensor([0.0198, 0.0211, 0.0170, 0.0204, 0.0200, 0.0187, 0.0157, 0.0187], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:6') 2023-04-27 23:53:47,064 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7022, 1.3072, 4.4939, 4.2231, 3.8912, 4.2764, 4.1600, 3.9574], device='cuda:6'), covar=tensor([0.6744, 0.6052, 0.1012, 0.1784, 0.1102, 0.1486, 0.1343, 0.1411], device='cuda:6'), in_proj_covar=tensor([0.0311, 0.0309, 0.0409, 0.0409, 0.0348, 0.0413, 0.0319, 0.0364], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-27 23:53:50,646 INFO [finetune.py:976] (6/7) Epoch 26, batch 3550, loss[loss=0.146, simple_loss=0.2209, pruned_loss=0.03552, over 4816.00 frames. ], tot_loss[loss=0.1658, simple_loss=0.2383, pruned_loss=0.04665, over 958272.54 frames. ], batch size: 25, lr: 2.96e-03, grad_scale: 32.0 2023-04-27 23:54:11,377 INFO [optim.py:369] (6/7) 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 23:54:19,277 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0956, 1.6091, 1.9547, 1.7299, 1.9067, 1.5886, 0.8954, 1.5297], device='cuda:6'), covar=tensor([0.3063, 0.3124, 0.1583, 0.2076, 0.2369, 0.2505, 0.3853, 0.1969], device='cuda:6'), in_proj_covar=tensor([0.0292, 0.0245, 0.0227, 0.0312, 0.0220, 0.0234, 0.0226, 0.0186], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:6') 2023-04-27 23:54:24,573 INFO [finetune.py:976] (6/7) Epoch 26, batch 3600, loss[loss=0.1583, simple_loss=0.2418, pruned_loss=0.03739, over 4822.00 frames. ], tot_loss[loss=0.1631, simple_loss=0.2356, pruned_loss=0.04534, over 958323.50 frames. ], batch size: 40, lr: 2.96e-03, grad_scale: 32.0 2023-04-27 23:54:32,794 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=146805.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:54:59,553 INFO [finetune.py:976] (6/7) Epoch 26, batch 3650, loss[loss=0.169, simple_loss=0.2491, pruned_loss=0.04444, over 4780.00 frames. ], tot_loss[loss=0.1651, simple_loss=0.2384, pruned_loss=0.04592, over 957950.65 frames. ], batch size: 29, lr: 2.96e-03, grad_scale: 32.0 2023-04-27 23:55:00,928 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=146844.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:55:10,835 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-04-27 23:55:11,629 INFO [zipformer.py:1188] (6/7) 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] (6/7) 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 23:55:37,111 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.1799, 1.6807, 1.4908, 2.0198, 2.1734, 1.7892, 1.8147, 1.5260], device='cuda:6'), covar=tensor([0.1986, 0.1902, 0.2027, 0.1821, 0.1339, 0.2189, 0.2531, 0.2579], device='cuda:6'), in_proj_covar=tensor([0.0319, 0.0315, 0.0357, 0.0291, 0.0332, 0.0309, 0.0303, 0.0380], device='cuda:6'), out_proj_covar=tensor([6.5351e-05, 6.4725e-05, 7.5155e-05, 5.8202e-05, 6.8079e-05, 6.4438e-05, 6.2824e-05, 8.0553e-05], device='cuda:6') 2023-04-27 23:55:40,316 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=3.99 vs. limit=5.0 2023-04-27 23:55:48,369 INFO [finetune.py:976] (6/7) Epoch 26, batch 3700, loss[loss=0.1813, simple_loss=0.2493, pruned_loss=0.05665, over 4892.00 frames. ], tot_loss[loss=0.1673, simple_loss=0.2419, pruned_loss=0.04631, over 957759.58 frames. ], batch size: 35, lr: 2.96e-03, grad_scale: 32.0 2023-04-27 23:56:06,851 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=146905.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:56:36,154 INFO [finetune.py:976] (6/7) Epoch 26, batch 3750, loss[loss=0.1647, simple_loss=0.2404, pruned_loss=0.04445, over 4810.00 frames. ], tot_loss[loss=0.1685, simple_loss=0.2429, pruned_loss=0.04705, over 958962.18 frames. ], batch size: 39, lr: 2.96e-03, grad_scale: 32.0 2023-04-27 23:56:44,464 INFO [zipformer.py:1188] (6/7) 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] (6/7) 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] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=146982.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 23:57:19,077 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7155, 3.3759, 0.8348, 1.8271, 1.9778, 2.3011, 1.8797, 0.9654], device='cuda:6'), covar=tensor([0.1209, 0.0844, 0.2026, 0.1201, 0.1066, 0.1018, 0.1587, 0.2035], device='cuda:6'), in_proj_covar=tensor([0.0119, 0.0240, 0.0137, 0.0122, 0.0133, 0.0154, 0.0118, 0.0120], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-27 23:57:20,823 INFO [finetune.py:976] (6/7) Epoch 26, batch 3800, loss[loss=0.2117, simple_loss=0.281, pruned_loss=0.07119, over 4813.00 frames. ], tot_loss[loss=0.169, simple_loss=0.2439, pruned_loss=0.0471, over 958825.94 frames. ], batch size: 38, lr: 2.95e-03, grad_scale: 32.0 2023-04-27 23:57:24,868 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-04-27 23:57:47,526 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.48 vs. limit=2.0 2023-04-27 23:58:05,507 INFO [finetune.py:976] (6/7) Epoch 26, batch 3850, loss[loss=0.1558, simple_loss=0.2397, pruned_loss=0.03591, over 4800.00 frames. ], tot_loss[loss=0.1683, simple_loss=0.2429, pruned_loss=0.04685, over 958642.14 frames. ], batch size: 51, lr: 2.95e-03, grad_scale: 32.0 2023-04-27 23:58:05,795 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.79 vs. limit=2.0 2023-04-27 23:58:06,307 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=3.76 vs. limit=5.0 2023-04-27 23:58:23,783 INFO [optim.py:369] (6/7) 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 23:58:39,181 INFO [finetune.py:976] (6/7) Epoch 26, batch 3900, loss[loss=0.1279, simple_loss=0.2096, pruned_loss=0.0231, over 4755.00 frames. ], tot_loss[loss=0.1675, simple_loss=0.2412, pruned_loss=0.04693, over 958730.99 frames. ], batch size: 28, lr: 2.95e-03, grad_scale: 32.0 2023-04-27 23:59:03,847 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.69 vs. limit=2.0 2023-04-27 23:59:12,137 INFO [finetune.py:976] (6/7) Epoch 26, batch 3950, loss[loss=0.1522, simple_loss=0.2251, pruned_loss=0.03964, over 4770.00 frames. ], tot_loss[loss=0.1638, simple_loss=0.237, pruned_loss=0.04531, over 956864.80 frames. ], batch size: 28, lr: 2.95e-03, grad_scale: 32.0 2023-04-27 23:59:31,441 INFO [optim.py:369] (6/7) 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 23:59:45,576 INFO [finetune.py:976] (6/7) Epoch 26, batch 4000, loss[loss=0.18, simple_loss=0.2603, pruned_loss=0.04985, over 4871.00 frames. ], tot_loss[loss=0.1639, simple_loss=0.2367, pruned_loss=0.04553, over 956616.58 frames. ], batch size: 44, lr: 2.95e-03, grad_scale: 32.0 2023-04-27 23:59:51,631 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=147200.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:59:54,701 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([3.6853, 3.6412, 2.7211, 4.2575, 3.7318, 3.6833, 1.5273, 3.6727], device='cuda:6'), covar=tensor([0.1733, 0.1303, 0.2937, 0.1770, 0.4425, 0.1778, 0.6155, 0.2273], device='cuda:6'), in_proj_covar=tensor([0.0245, 0.0220, 0.0252, 0.0304, 0.0299, 0.0247, 0.0273, 0.0275], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-28 00:00:18,669 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 26, batch 4050, loss[loss=0.1192, simple_loss=0.2002, pruned_loss=0.01911, over 4822.00 frames. ], tot_loss[loss=0.1653, simple_loss=0.2388, pruned_loss=0.04596, over 954545.04 frames. ], batch size: 25, lr: 2.95e-03, grad_scale: 32.0 2023-04-28 00:00:23,245 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=147247.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:00:27,931 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7191, 1.5835, 4.5662, 4.2965, 3.9409, 4.3935, 4.2634, 4.0548], device='cuda:6'), covar=tensor([0.6879, 0.5718, 0.0885, 0.1554, 0.1023, 0.1823, 0.1157, 0.1559], device='cuda:6'), in_proj_covar=tensor([0.0310, 0.0308, 0.0407, 0.0409, 0.0348, 0.0414, 0.0318, 0.0364], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-28 00:00:44,478 INFO [optim.py:369] (6/7) 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 00:00:55,651 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=147282.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:01:10,421 INFO [finetune.py:976] (6/7) Epoch 26, batch 4100, loss[loss=0.1604, simple_loss=0.2495, pruned_loss=0.03565, over 4906.00 frames. ], tot_loss[loss=0.1658, simple_loss=0.2396, pruned_loss=0.04597, over 954065.10 frames. ], batch size: 36, lr: 2.95e-03, grad_scale: 32.0 2023-04-28 00:01:12,321 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=147295.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:01:23,297 INFO [zipformer.py:1188] (6/7) 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] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=147310.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:01:54,627 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 26, batch 4150, loss[loss=0.1959, simple_loss=0.259, pruned_loss=0.06641, over 4860.00 frames. ], tot_loss[loss=0.1671, simple_loss=0.2414, pruned_loss=0.04637, over 953716.23 frames. ], batch size: 31, lr: 2.95e-03, grad_scale: 32.0 2023-04-28 00:02:56,045 INFO [zipformer.py:1188] (6/7) 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] (6/7) 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 00:03:09,330 INFO [finetune.py:976] (6/7) Epoch 26, batch 4200, loss[loss=0.1652, simple_loss=0.2359, pruned_loss=0.0472, over 4913.00 frames. ], tot_loss[loss=0.1668, simple_loss=0.2414, pruned_loss=0.04608, over 953429.02 frames. ], batch size: 38, lr: 2.95e-03, grad_scale: 32.0 2023-04-28 00:03:21,141 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=147407.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:03:37,869 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=147434.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:03:42,687 INFO [finetune.py:976] (6/7) Epoch 26, batch 4250, loss[loss=0.1547, simple_loss=0.2228, pruned_loss=0.0433, over 4769.00 frames. ], tot_loss[loss=0.1655, simple_loss=0.2393, pruned_loss=0.04586, over 953667.92 frames. ], batch size: 27, lr: 2.95e-03, grad_scale: 32.0 2023-04-28 00:03:48,890 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([3.9123, 3.7851, 2.8597, 4.4980, 3.8332, 3.9377, 1.5993, 3.8859], device='cuda:6'), covar=tensor([0.1746, 0.1248, 0.3469, 0.1371, 0.3166, 0.1897, 0.5979, 0.2133], device='cuda:6'), in_proj_covar=tensor([0.0244, 0.0219, 0.0251, 0.0304, 0.0297, 0.0246, 0.0272, 0.0272], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-28 00:03:50,181 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8736, 2.2455, 1.8379, 1.6719, 1.3878, 1.4281, 1.9020, 1.3310], device='cuda:6'), covar=tensor([0.1682, 0.1321, 0.1430, 0.1649, 0.2311, 0.1957, 0.0952, 0.2062], device='cuda:6'), in_proj_covar=tensor([0.0199, 0.0211, 0.0170, 0.0204, 0.0200, 0.0187, 0.0157, 0.0187], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:6') 2023-04-28 00:03:56,815 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-04-28 00:03:58,812 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.2477, 1.5023, 1.9818, 2.2906, 2.1103, 1.6239, 1.1435, 1.7981], device='cuda:6'), covar=tensor([0.3500, 0.3883, 0.2035, 0.2601, 0.2686, 0.2865, 0.4609, 0.2122], device='cuda:6'), in_proj_covar=tensor([0.0294, 0.0247, 0.0229, 0.0315, 0.0222, 0.0236, 0.0229, 0.0187], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:6') 2023-04-28 00:04:01,770 INFO [zipformer.py:1188] (6/7) 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] (6/7) 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 00:04:07,343 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7767, 2.1024, 1.7940, 2.0361, 1.6254, 1.7929, 1.6774, 1.3393], device='cuda:6'), covar=tensor([0.1609, 0.1275, 0.0797, 0.1053, 0.3310, 0.1151, 0.2064, 0.2502], device='cuda:6'), in_proj_covar=tensor([0.0285, 0.0300, 0.0217, 0.0278, 0.0316, 0.0255, 0.0250, 0.0264], device='cuda:6'), out_proj_covar=tensor([1.1393e-04, 1.1803e-04, 8.5265e-05, 1.0927e-04, 1.2724e-04, 1.0025e-04, 1.0097e-04, 1.0404e-04], device='cuda:6') 2023-04-28 00:04:13,392 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.1106, 1.7831, 2.2779, 2.4637, 2.1549, 2.0132, 2.1376, 2.0961], device='cuda:6'), covar=tensor([0.4469, 0.7280, 0.6855, 0.5593, 0.5993, 0.8223, 0.8963, 0.9834], device='cuda:6'), in_proj_covar=tensor([0.0441, 0.0421, 0.0513, 0.0506, 0.0468, 0.0503, 0.0506, 0.0517], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-28 00:04:16,306 INFO [finetune.py:976] (6/7) Epoch 26, batch 4300, loss[loss=0.1925, simple_loss=0.2603, pruned_loss=0.06233, over 4864.00 frames. ], tot_loss[loss=0.1649, simple_loss=0.2377, pruned_loss=0.04611, over 955218.52 frames. ], batch size: 49, lr: 2.95e-03, grad_scale: 32.0 2023-04-28 00:04:18,264 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=147495.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:04:22,205 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=147500.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:04:49,579 INFO [finetune.py:976] (6/7) Epoch 26, batch 4350, loss[loss=0.1249, simple_loss=0.2018, pruned_loss=0.02403, over 4766.00 frames. ], tot_loss[loss=0.1615, simple_loss=0.2336, pruned_loss=0.04466, over 952544.85 frames. ], batch size: 27, lr: 2.95e-03, grad_scale: 32.0 2023-04-28 00:04:53,342 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=147548.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:05:10,813 INFO [optim.py:369] (6/7) 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 00:05:13,679 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.77 vs. limit=2.0 2023-04-28 00:05:16,120 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-04-28 00:05:23,015 INFO [finetune.py:976] (6/7) Epoch 26, batch 4400, loss[loss=0.1872, simple_loss=0.2572, pruned_loss=0.05861, over 4824.00 frames. ], tot_loss[loss=0.1633, simple_loss=0.2358, pruned_loss=0.04541, over 953351.21 frames. ], batch size: 38, lr: 2.95e-03, grad_scale: 32.0 2023-04-28 00:05:26,138 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=147597.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:05:37,169 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.3692, 1.7026, 1.9101, 1.9768, 1.8659, 1.9403, 2.0031, 1.9503], device='cuda:6'), covar=tensor([0.3596, 0.5431, 0.4488, 0.4620, 0.5396, 0.6568, 0.4804, 0.4622], device='cuda:6'), in_proj_covar=tensor([0.0341, 0.0375, 0.0329, 0.0340, 0.0351, 0.0395, 0.0361, 0.0331], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-28 00:05:56,084 INFO [finetune.py:976] (6/7) Epoch 26, batch 4450, loss[loss=0.1568, simple_loss=0.2416, pruned_loss=0.036, over 4897.00 frames. ], tot_loss[loss=0.1652, simple_loss=0.2384, pruned_loss=0.04603, over 951150.37 frames. ], batch size: 36, lr: 2.95e-03, grad_scale: 32.0 2023-04-28 00:06:01,113 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6548, 1.0512, 1.6721, 2.1200, 1.7284, 1.5992, 1.6328, 1.6035], device='cuda:6'), covar=tensor([0.4147, 0.6895, 0.5649, 0.5437, 0.5626, 0.7541, 0.7632, 0.8716], device='cuda:6'), in_proj_covar=tensor([0.0442, 0.0421, 0.0515, 0.0507, 0.0469, 0.0505, 0.0507, 0.0519], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-28 00:06:17,564 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=147659.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:06:27,432 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=147666.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:06:29,366 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7216, 2.5678, 2.0276, 2.3868, 2.7136, 2.1746, 3.3749, 1.9322], device='cuda:6'), covar=tensor([0.3678, 0.2527, 0.4559, 0.3446, 0.1785, 0.2679, 0.1786, 0.4725], device='cuda:6'), in_proj_covar=tensor([0.0339, 0.0354, 0.0424, 0.0349, 0.0382, 0.0373, 0.0365, 0.0423], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-28 00:06:31,477 INFO [optim.py:369] (6/7) 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 00:06:44,115 INFO [finetune.py:976] (6/7) Epoch 26, batch 4500, loss[loss=0.1639, simple_loss=0.2318, pruned_loss=0.048, over 4816.00 frames. ], tot_loss[loss=0.1675, simple_loss=0.241, pruned_loss=0.04702, over 951461.20 frames. ], batch size: 25, lr: 2.95e-03, grad_scale: 32.0 2023-04-28 00:06:51,735 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([3.0885, 2.4384, 2.3285, 2.4418, 2.2680, 2.4161, 2.3151, 2.3100], device='cuda:6'), covar=tensor([0.3229, 0.6059, 0.4444, 0.4403, 0.5783, 0.6647, 0.6391, 0.5667], device='cuda:6'), in_proj_covar=tensor([0.0340, 0.0373, 0.0328, 0.0339, 0.0350, 0.0395, 0.0360, 0.0331], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-28 00:07:22,603 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=147720.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:07:37,059 INFO [finetune.py:976] (6/7) Epoch 26, batch 4550, loss[loss=0.1832, simple_loss=0.2645, pruned_loss=0.05091, over 4914.00 frames. ], tot_loss[loss=0.169, simple_loss=0.2424, pruned_loss=0.04778, over 951151.51 frames. ], batch size: 38, lr: 2.95e-03, grad_scale: 32.0 2023-04-28 00:08:01,007 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=147763.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:08:02,623 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6578, 1.5911, 0.8200, 1.3606, 1.6203, 1.5112, 1.4251, 1.4533], device='cuda:6'), covar=tensor([0.0488, 0.0361, 0.0337, 0.0564, 0.0282, 0.0504, 0.0482, 0.0562], device='cuda:6'), in_proj_covar=tensor([0.0028, 0.0023, 0.0022, 0.0028, 0.0019, 0.0027, 0.0027, 0.0028], device='cuda:6'), out_proj_covar=tensor([0.0052, 0.0046, 0.0038, 0.0053, 0.0039, 0.0050, 0.0050, 0.0052], device='cuda:6') 2023-04-28 00:08:14,011 INFO [optim.py:369] (6/7) 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 00:08:24,996 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8206, 1.6936, 2.1134, 2.3355, 1.5861, 1.5287, 1.8457, 1.0546], device='cuda:6'), covar=tensor([0.0705, 0.0734, 0.0525, 0.0671, 0.0739, 0.1031, 0.0598, 0.0740], device='cuda:6'), in_proj_covar=tensor([0.0070, 0.0067, 0.0066, 0.0069, 0.0074, 0.0094, 0.0072, 0.0063], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-28 00:08:31,440 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=147790.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:08:32,584 INFO [finetune.py:976] (6/7) Epoch 26, batch 4600, loss[loss=0.1256, simple_loss=0.2034, pruned_loss=0.02394, over 4786.00 frames. ], tot_loss[loss=0.1681, simple_loss=0.2411, pruned_loss=0.04754, over 949831.63 frames. ], batch size: 29, lr: 2.95e-03, grad_scale: 32.0 2023-04-28 00:08:47,333 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 26, batch 4650, loss[loss=0.1608, simple_loss=0.2332, pruned_loss=0.04417, over 4907.00 frames. ], tot_loss[loss=0.1664, simple_loss=0.2385, pruned_loss=0.04718, over 949561.10 frames. ], batch size: 46, lr: 2.95e-03, grad_scale: 32.0 2023-04-28 00:09:15,443 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=147857.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:09:25,963 INFO [optim.py:369] (6/7) 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 00:09:29,047 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=147876.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:09:40,052 INFO [finetune.py:976] (6/7) Epoch 26, batch 4700, loss[loss=0.1476, simple_loss=0.2199, pruned_loss=0.03765, over 4695.00 frames. ], tot_loss[loss=0.1647, simple_loss=0.2366, pruned_loss=0.04641, over 951316.61 frames. ], batch size: 23, lr: 2.95e-03, grad_scale: 32.0 2023-04-28 00:09:43,202 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=147897.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:09:51,639 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=147911.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:09:55,896 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=147918.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:10:06,677 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.80 vs. limit=5.0 2023-04-28 00:10:12,814 INFO [finetune.py:976] (6/7) Epoch 26, batch 4750, loss[loss=0.2268, simple_loss=0.2811, pruned_loss=0.08624, over 4817.00 frames. ], tot_loss[loss=0.1627, simple_loss=0.2341, pruned_loss=0.04566, over 951495.95 frames. ], batch size: 38, lr: 2.95e-03, grad_scale: 32.0 2023-04-28 00:10:15,215 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=147945.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:10:28,067 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=147966.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:10:31,606 INFO [optim.py:369] (6/7) 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 00:10:31,737 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 26, batch 4800, loss[loss=0.1776, simple_loss=0.2649, pruned_loss=0.04515, over 4815.00 frames. ], tot_loss[loss=0.1648, simple_loss=0.237, pruned_loss=0.0463, over 951182.25 frames. ], batch size: 41, lr: 2.95e-03, grad_scale: 32.0 2023-04-28 00:10:50,276 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8783, 1.4416, 1.5722, 1.6557, 2.1162, 1.6683, 1.4887, 1.4651], device='cuda:6'), covar=tensor([0.1712, 0.1594, 0.2055, 0.1483, 0.0859, 0.1648, 0.1985, 0.2378], device='cuda:6'), in_proj_covar=tensor([0.0315, 0.0310, 0.0353, 0.0287, 0.0328, 0.0305, 0.0299, 0.0376], device='cuda:6'), out_proj_covar=tensor([6.4301e-05, 6.3716e-05, 7.4088e-05, 5.7491e-05, 6.7034e-05, 6.3777e-05, 6.1834e-05, 7.9755e-05], device='cuda:6') 2023-04-28 00:10:55,859 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0247, 2.5893, 1.0475, 1.3515, 2.0329, 1.2612, 3.3988, 1.6764], device='cuda:6'), covar=tensor([0.0693, 0.0707, 0.0818, 0.1330, 0.0531, 0.1007, 0.0244, 0.0660], device='cuda:6'), in_proj_covar=tensor([0.0051, 0.0064, 0.0047, 0.0046, 0.0050, 0.0051, 0.0073, 0.0051], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:6') 2023-04-28 00:11:01,951 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=148014.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:11:02,583 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=148015.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:11:11,464 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.2550, 1.5089, 1.2906, 1.4886, 1.2848, 1.2541, 1.1633, 1.0458], device='cuda:6'), covar=tensor([0.1603, 0.1137, 0.0903, 0.1033, 0.3541, 0.1141, 0.1735, 0.2098], device='cuda:6'), in_proj_covar=tensor([0.0281, 0.0295, 0.0213, 0.0274, 0.0311, 0.0251, 0.0246, 0.0260], device='cuda:6'), out_proj_covar=tensor([1.1200e-04, 1.1620e-04, 8.3822e-05, 1.0780e-04, 1.2546e-04, 9.8802e-05, 9.9343e-05, 1.0242e-04], device='cuda:6') 2023-04-28 00:11:18,290 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([4.7263, 4.6450, 3.1361, 5.3587, 4.7617, 4.6184, 2.0001, 4.6251], device='cuda:6'), covar=tensor([0.1562, 0.1099, 0.3141, 0.0930, 0.3028, 0.1814, 0.5587, 0.2039], device='cuda:6'), in_proj_covar=tensor([0.0247, 0.0221, 0.0253, 0.0305, 0.0300, 0.0249, 0.0275, 0.0276], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-28 00:11:18,369 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.6173, 2.1072, 2.4871, 3.1910, 2.4764, 1.9530, 2.0682, 2.4096], device='cuda:6'), covar=tensor([0.2888, 0.2882, 0.1516, 0.1923, 0.2605, 0.2512, 0.3393, 0.1831], device='cuda:6'), in_proj_covar=tensor([0.0295, 0.0247, 0.0228, 0.0314, 0.0222, 0.0236, 0.0229, 0.0187], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:6') 2023-04-28 00:11:21,740 INFO [finetune.py:976] (6/7) Epoch 26, batch 4850, loss[loss=0.1702, simple_loss=0.2392, pruned_loss=0.05058, over 4888.00 frames. ], tot_loss[loss=0.1671, simple_loss=0.2402, pruned_loss=0.04694, over 951969.85 frames. ], batch size: 32, lr: 2.95e-03, grad_scale: 32.0 2023-04-28 00:11:22,293 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 2023-04-28 00:11:44,643 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=148063.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:11:55,267 INFO [optim.py:369] (6/7) 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 00:11:56,638 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9622, 2.4754, 1.9270, 1.7633, 1.4518, 1.4359, 1.9939, 1.4246], device='cuda:6'), covar=tensor([0.1627, 0.1328, 0.1402, 0.1674, 0.2273, 0.1995, 0.0989, 0.2004], device='cuda:6'), in_proj_covar=tensor([0.0198, 0.0211, 0.0170, 0.0204, 0.0200, 0.0186, 0.0157, 0.0188], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:6') 2023-04-28 00:12:18,070 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=148090.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:12:24,892 INFO [finetune.py:976] (6/7) Epoch 26, batch 4900, loss[loss=0.1723, simple_loss=0.2469, pruned_loss=0.04882, over 4814.00 frames. ], tot_loss[loss=0.1675, simple_loss=0.241, pruned_loss=0.04696, over 952634.73 frames. ], batch size: 40, lr: 2.95e-03, grad_scale: 32.0 2023-04-28 00:12:47,901 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=148111.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:13:22,194 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=148138.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:13:22,837 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([3.9539, 3.8761, 2.6988, 4.5038, 3.9787, 3.9489, 1.6665, 3.8778], device='cuda:6'), covar=tensor([0.1669, 0.1110, 0.3292, 0.1478, 0.2354, 0.1631, 0.5768, 0.2147], device='cuda:6'), in_proj_covar=tensor([0.0247, 0.0221, 0.0253, 0.0305, 0.0300, 0.0248, 0.0274, 0.0275], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-28 00:13:30,213 INFO [finetune.py:976] (6/7) Epoch 26, batch 4950, loss[loss=0.2007, simple_loss=0.2742, pruned_loss=0.06356, over 4839.00 frames. ], tot_loss[loss=0.1677, simple_loss=0.2416, pruned_loss=0.04687, over 952906.19 frames. ], batch size: 49, lr: 2.95e-03, grad_scale: 32.0 2023-04-28 00:13:52,220 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=148171.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:13:52,738 INFO [optim.py:369] (6/7) 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 00:14:06,561 INFO [finetune.py:976] (6/7) Epoch 26, batch 5000, loss[loss=0.1601, simple_loss=0.2291, pruned_loss=0.04555, over 4887.00 frames. ], tot_loss[loss=0.1663, simple_loss=0.2401, pruned_loss=0.04626, over 954803.41 frames. ], batch size: 35, lr: 2.95e-03, grad_scale: 32.0 2023-04-28 00:14:21,041 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=148213.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:14:39,766 INFO [finetune.py:976] (6/7) Epoch 26, batch 5050, loss[loss=0.164, simple_loss=0.2437, pruned_loss=0.04218, over 4914.00 frames. ], tot_loss[loss=0.1644, simple_loss=0.2375, pruned_loss=0.04563, over 955424.18 frames. ], batch size: 37, lr: 2.95e-03, grad_scale: 32.0 2023-04-28 00:14:56,596 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=148267.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:14:59,523 INFO [optim.py:369] (6/7) 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 00:15:01,593 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.54 vs. limit=2.0 2023-04-28 00:15:08,811 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=148287.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:15:10,041 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.3143, 1.7130, 1.5744, 2.1855, 2.3409, 1.9385, 1.9471, 1.5688], device='cuda:6'), covar=tensor([0.1543, 0.1675, 0.1885, 0.1497, 0.1119, 0.1691, 0.1704, 0.2096], device='cuda:6'), in_proj_covar=tensor([0.0313, 0.0309, 0.0351, 0.0285, 0.0326, 0.0304, 0.0297, 0.0374], device='cuda:6'), out_proj_covar=tensor([6.3894e-05, 6.3508e-05, 7.3800e-05, 5.7147e-05, 6.6724e-05, 6.3381e-05, 6.1483e-05, 7.9355e-05], device='cuda:6') 2023-04-28 00:15:11,732 INFO [finetune.py:976] (6/7) Epoch 26, batch 5100, loss[loss=0.18, simple_loss=0.2407, pruned_loss=0.05966, over 4813.00 frames. ], tot_loss[loss=0.1616, simple_loss=0.234, pruned_loss=0.04456, over 954377.46 frames. ], batch size: 38, lr: 2.95e-03, grad_scale: 64.0 2023-04-28 00:15:28,682 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=148315.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:15:33,791 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 2023-04-28 00:15:45,038 INFO [finetune.py:976] (6/7) Epoch 26, batch 5150, loss[loss=0.1814, simple_loss=0.2541, pruned_loss=0.05436, over 4874.00 frames. ], tot_loss[loss=0.1623, simple_loss=0.2347, pruned_loss=0.04493, over 954587.10 frames. ], batch size: 34, lr: 2.95e-03, grad_scale: 64.0 2023-04-28 00:15:47,681 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.30 vs. limit=5.0 2023-04-28 00:15:49,798 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=148348.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:16:00,255 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=148363.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:16:06,185 INFO [optim.py:369] (6/7) 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 00:16:18,408 INFO [finetune.py:976] (6/7) Epoch 26, batch 5200, loss[loss=0.179, simple_loss=0.2515, pruned_loss=0.0533, over 4914.00 frames. ], tot_loss[loss=0.1663, simple_loss=0.2391, pruned_loss=0.04681, over 954088.57 frames. ], batch size: 37, lr: 2.95e-03, grad_scale: 64.0 2023-04-28 00:16:19,760 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([0.6013, 1.4005, 1.4153, 1.0178, 1.3528, 1.1822, 1.7443, 1.3440], device='cuda:6'), covar=tensor([0.3126, 0.1854, 0.4623, 0.2520, 0.1574, 0.2143, 0.1400, 0.4223], device='cuda:6'), in_proj_covar=tensor([0.0336, 0.0353, 0.0423, 0.0348, 0.0380, 0.0373, 0.0363, 0.0421], device='cuda:6'), out_proj_covar=tensor([9.9413e-05, 1.0510e-04, 1.2802e-04, 1.0437e-04, 1.1278e-04, 1.1086e-04, 1.0601e-04, 1.2646e-04], device='cuda:6') 2023-04-28 00:16:44,055 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=148429.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:16:51,854 INFO [finetune.py:976] (6/7) Epoch 26, batch 5250, loss[loss=0.1686, simple_loss=0.2351, pruned_loss=0.05107, over 4918.00 frames. ], tot_loss[loss=0.1671, simple_loss=0.2411, pruned_loss=0.04656, over 952647.60 frames. ], batch size: 36, lr: 2.94e-03, grad_scale: 64.0 2023-04-28 00:17:12,097 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=148471.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:17:12,618 INFO [optim.py:369] (6/7) 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 00:17:23,744 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=148490.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:17:24,854 INFO [finetune.py:976] (6/7) Epoch 26, batch 5300, loss[loss=0.1585, simple_loss=0.2195, pruned_loss=0.04876, over 4381.00 frames. ], tot_loss[loss=0.1677, simple_loss=0.242, pruned_loss=0.04667, over 949790.84 frames. ], batch size: 19, lr: 2.94e-03, grad_scale: 64.0 2023-04-28 00:17:28,555 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=148498.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:17:52,382 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=148513.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:17:56,055 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=148519.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:18:02,880 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5902, 3.5580, 1.0338, 1.8404, 1.9020, 2.6132, 1.9054, 1.1413], device='cuda:6'), covar=tensor([0.1437, 0.1033, 0.2047, 0.1367, 0.1151, 0.0976, 0.1731, 0.2112], device='cuda:6'), in_proj_covar=tensor([0.0117, 0.0236, 0.0135, 0.0120, 0.0130, 0.0151, 0.0116, 0.0118], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-28 00:18:24,818 INFO [finetune.py:976] (6/7) Epoch 26, batch 5350, loss[loss=0.1864, simple_loss=0.2486, pruned_loss=0.0621, over 4926.00 frames. ], tot_loss[loss=0.1682, simple_loss=0.2429, pruned_loss=0.04679, over 951411.50 frames. ], batch size: 33, lr: 2.94e-03, grad_scale: 64.0 2023-04-28 00:18:28,056 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6973, 1.3033, 1.3410, 1.4854, 1.9232, 1.5342, 1.3351, 1.2770], device='cuda:6'), covar=tensor([0.1457, 0.1387, 0.1519, 0.1291, 0.0767, 0.1420, 0.1828, 0.2103], device='cuda:6'), in_proj_covar=tensor([0.0318, 0.0313, 0.0355, 0.0290, 0.0330, 0.0308, 0.0302, 0.0379], device='cuda:6'), out_proj_covar=tensor([6.5058e-05, 6.4302e-05, 7.4650e-05, 5.8020e-05, 6.7474e-05, 6.4277e-05, 6.2542e-05, 8.0217e-05], device='cuda:6') 2023-04-28 00:18:36,695 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.6451, 2.2471, 2.5065, 3.1730, 2.4862, 1.9458, 2.1058, 2.4500], device='cuda:6'), covar=tensor([0.3100, 0.2889, 0.1661, 0.2131, 0.2788, 0.2673, 0.3445, 0.1904], device='cuda:6'), in_proj_covar=tensor([0.0292, 0.0245, 0.0228, 0.0314, 0.0221, 0.0235, 0.0228, 0.0187], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:6') 2023-04-28 00:18:46,017 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=148559.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:18:47,666 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=148561.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:18:56,342 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=148567.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:18:59,260 INFO [optim.py:369] (6/7) 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 00:19:22,270 INFO [finetune.py:976] (6/7) Epoch 26, batch 5400, loss[loss=0.1739, simple_loss=0.2534, pruned_loss=0.04721, over 4779.00 frames. ], tot_loss[loss=0.1684, simple_loss=0.2418, pruned_loss=0.04747, over 952682.56 frames. ], batch size: 29, lr: 2.94e-03, grad_scale: 64.0 2023-04-28 00:19:42,040 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0245, 2.6577, 1.0769, 1.3964, 1.9812, 1.1537, 3.2204, 1.6594], device='cuda:6'), covar=tensor([0.0699, 0.0610, 0.0808, 0.1125, 0.0485, 0.1026, 0.0199, 0.0584], device='cuda:6'), in_proj_covar=tensor([0.0051, 0.0064, 0.0047, 0.0046, 0.0050, 0.0052, 0.0074, 0.0051], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:6') 2023-04-28 00:19:42,666 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([0.5535, 1.4056, 1.4060, 0.9605, 1.3233, 1.1853, 1.6866, 1.3549], device='cuda:6'), covar=tensor([0.3271, 0.1644, 0.4604, 0.2548, 0.1637, 0.1977, 0.1562, 0.4429], device='cuda:6'), in_proj_covar=tensor([0.0337, 0.0352, 0.0423, 0.0348, 0.0382, 0.0374, 0.0364, 0.0421], device='cuda:6'), out_proj_covar=tensor([9.9664e-05, 1.0498e-04, 1.2808e-04, 1.0448e-04, 1.1318e-04, 1.1119e-04, 1.0626e-04, 1.2664e-04], device='cuda:6') 2023-04-28 00:19:54,122 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=148615.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:20:11,714 INFO [finetune.py:976] (6/7) Epoch 26, batch 5450, loss[loss=0.1676, simple_loss=0.2439, pruned_loss=0.04563, over 4904.00 frames. ], tot_loss[loss=0.1658, simple_loss=0.2385, pruned_loss=0.04649, over 952711.56 frames. ], batch size: 35, lr: 2.94e-03, grad_scale: 64.0 2023-04-28 00:20:12,389 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=148643.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:20:17,247 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=148651.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:20:31,293 INFO [optim.py:369] (6/7) 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] (6/7) Epoch 26, batch 5500, loss[loss=0.1393, simple_loss=0.2092, pruned_loss=0.03468, over 4830.00 frames. ], tot_loss[loss=0.1645, simple_loss=0.2363, pruned_loss=0.04635, over 951475.72 frames. ], batch size: 30, lr: 2.94e-03, grad_scale: 64.0 2023-04-28 00:20:55,470 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-04-28 00:20:57,700 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=148712.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:21:00,114 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=148716.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:21:18,757 INFO [finetune.py:976] (6/7) Epoch 26, batch 5550, loss[loss=0.1937, simple_loss=0.2749, pruned_loss=0.05622, over 4916.00 frames. ], tot_loss[loss=0.1659, simple_loss=0.2377, pruned_loss=0.04711, over 951737.02 frames. ], batch size: 38, lr: 2.94e-03, grad_scale: 64.0 2023-04-28 00:21:38,043 INFO [optim.py:369] (6/7) 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 00:21:38,812 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-04-28 00:21:41,049 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=148777.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:21:45,568 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=148785.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:21:49,603 INFO [finetune.py:976] (6/7) Epoch 26, batch 5600, loss[loss=0.1668, simple_loss=0.2467, pruned_loss=0.04346, over 4893.00 frames. ], tot_loss[loss=0.169, simple_loss=0.2416, pruned_loss=0.04825, over 950279.33 frames. ], batch size: 35, lr: 2.94e-03, grad_scale: 64.0 2023-04-28 00:22:19,037 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7855, 1.1816, 1.7880, 2.2145, 1.8155, 1.7126, 1.7580, 1.7251], device='cuda:6'), covar=tensor([0.4817, 0.7279, 0.6603, 0.5958, 0.6579, 0.8445, 0.8191, 1.0044], device='cuda:6'), in_proj_covar=tensor([0.0442, 0.0421, 0.0515, 0.0504, 0.0468, 0.0506, 0.0507, 0.0518], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-28 00:22:20,258 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-04-28 00:22:20,665 INFO [finetune.py:976] (6/7) Epoch 26, batch 5650, loss[loss=0.1992, simple_loss=0.2812, pruned_loss=0.05856, over 4807.00 frames. ], tot_loss[loss=0.1695, simple_loss=0.2439, pruned_loss=0.04758, over 953470.56 frames. ], batch size: 40, lr: 2.94e-03, grad_scale: 32.0 2023-04-28 00:22:27,778 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=148854.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:22:39,029 INFO [optim.py:369] (6/7) 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 00:22:50,264 INFO [finetune.py:976] (6/7) Epoch 26, batch 5700, loss[loss=0.1579, simple_loss=0.2231, pruned_loss=0.04635, over 4068.00 frames. ], tot_loss[loss=0.1671, simple_loss=0.2397, pruned_loss=0.04723, over 929107.05 frames. ], batch size: 17, lr: 2.94e-03, grad_scale: 32.0 2023-04-28 00:23:20,153 INFO [finetune.py:976] (6/7) Epoch 27, batch 0, loss[loss=0.1849, simple_loss=0.2554, pruned_loss=0.05714, over 4920.00 frames. ], tot_loss[loss=0.1849, simple_loss=0.2554, pruned_loss=0.05714, over 4920.00 frames. ], batch size: 33, lr: 2.94e-03, grad_scale: 32.0 2023-04-28 00:23:20,153 INFO [finetune.py:1001] (6/7) Computing validation loss 2023-04-28 00:23:31,100 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.1629, 2.5363, 1.0951, 1.4218, 2.0453, 1.2815, 3.0425, 1.7628], device='cuda:6'), covar=tensor([0.0619, 0.0567, 0.0666, 0.1203, 0.0391, 0.0921, 0.0239, 0.0562], device='cuda:6'), in_proj_covar=tensor([0.0051, 0.0064, 0.0047, 0.0046, 0.0050, 0.0051, 0.0073, 0.0051], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:6') 2023-04-28 00:23:31,557 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.2825, 1.2537, 1.5202, 1.5206, 1.2337, 1.1429, 1.3228, 0.8626], device='cuda:6'), covar=tensor([0.0476, 0.0501, 0.0488, 0.0479, 0.0604, 0.0938, 0.0462, 0.0520], device='cuda:6'), in_proj_covar=tensor([0.0070, 0.0067, 0.0066, 0.0068, 0.0074, 0.0094, 0.0072, 0.0063], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-28 00:23:41,724 INFO [finetune.py:1010] (6/7) 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] (6/7) Maximum memory allocated so far is 6435MB 2023-04-28 00:24:11,140 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=148943.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:24:41,643 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-04-28 00:24:43,146 INFO [finetune.py:976] (6/7) Epoch 27, batch 50, loss[loss=0.1889, simple_loss=0.2684, pruned_loss=0.05473, over 4892.00 frames. ], tot_loss[loss=0.1656, simple_loss=0.2391, pruned_loss=0.04609, over 215592.46 frames. ], batch size: 35, lr: 2.94e-03, grad_scale: 32.0 2023-04-28 00:24:45,493 INFO [optim.py:369] (6/7) 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 00:24:53,724 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0764, 1.2568, 4.6509, 4.4202, 4.0373, 4.2973, 4.1217, 4.1369], device='cuda:6'), covar=tensor([0.6679, 0.6063, 0.1076, 0.1630, 0.1088, 0.1209, 0.2687, 0.1619], device='cuda:6'), in_proj_covar=tensor([0.0312, 0.0310, 0.0410, 0.0411, 0.0349, 0.0415, 0.0320, 0.0367], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-28 00:25:07,492 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=148991.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:25:28,985 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 27, batch 100, loss[loss=0.1886, simple_loss=0.2513, pruned_loss=0.06295, over 4783.00 frames. ], tot_loss[loss=0.1653, simple_loss=0.238, pruned_loss=0.04627, over 380382.88 frames. ], batch size: 51, lr: 2.94e-03, grad_scale: 32.0 2023-04-28 00:25:48,516 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.2568, 1.2708, 1.3970, 1.6325, 1.6079, 1.3004, 1.0656, 1.5424], device='cuda:6'), covar=tensor([0.0765, 0.1240, 0.0822, 0.0541, 0.0650, 0.0716, 0.0755, 0.0544], device='cuda:6'), in_proj_covar=tensor([0.0184, 0.0200, 0.0182, 0.0169, 0.0176, 0.0175, 0.0150, 0.0177], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-28 00:25:58,870 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([3.3365, 3.3137, 2.4397, 3.8562, 3.3461, 3.3145, 1.3790, 3.2640], device='cuda:6'), covar=tensor([0.1937, 0.1291, 0.3578, 0.2653, 0.3031, 0.2116, 0.6074, 0.2692], device='cuda:6'), in_proj_covar=tensor([0.0246, 0.0222, 0.0255, 0.0306, 0.0301, 0.0250, 0.0276, 0.0276], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-28 00:26:01,175 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.39 vs. limit=5.0 2023-04-28 00:26:11,903 INFO [finetune.py:976] (6/7) Epoch 27, batch 150, loss[loss=0.1212, simple_loss=0.1987, pruned_loss=0.02183, over 4805.00 frames. ], tot_loss[loss=0.1631, simple_loss=0.2345, pruned_loss=0.04582, over 510091.69 frames. ], batch size: 25, lr: 2.94e-03, grad_scale: 32.0 2023-04-28 00:26:13,188 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=149072.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:26:13,727 INFO [optim.py:369] (6/7) 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 00:26:18,534 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=3.98 vs. limit=5.0 2023-04-28 00:26:22,652 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=149085.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:26:45,181 INFO [finetune.py:976] (6/7) Epoch 27, batch 200, loss[loss=0.1201, simple_loss=0.2011, pruned_loss=0.01962, over 4787.00 frames. ], tot_loss[loss=0.1626, simple_loss=0.2335, pruned_loss=0.04581, over 609600.00 frames. ], batch size: 29, lr: 2.94e-03, grad_scale: 32.0 2023-04-28 00:26:55,614 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=149133.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:26:56,265 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.3593, 3.0089, 0.9750, 1.7174, 2.2384, 1.2818, 3.7792, 1.8767], device='cuda:6'), covar=tensor([0.0622, 0.0753, 0.0881, 0.1169, 0.0499, 0.0981, 0.0204, 0.0638], device='cuda:6'), in_proj_covar=tensor([0.0051, 0.0064, 0.0047, 0.0046, 0.0049, 0.0051, 0.0073, 0.0051], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:6') 2023-04-28 00:27:08,564 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=149154.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:27:18,795 INFO [finetune.py:976] (6/7) Epoch 27, batch 250, loss[loss=0.2036, simple_loss=0.2798, pruned_loss=0.06369, over 4810.00 frames. ], tot_loss[loss=0.1642, simple_loss=0.236, pruned_loss=0.0462, over 684489.55 frames. ], batch size: 40, lr: 2.94e-03, grad_scale: 32.0 2023-04-28 00:27:21,650 INFO [optim.py:369] (6/7) 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 00:27:24,253 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.0020, 1.9043, 1.6519, 1.4958, 1.9953, 1.5552, 2.4456, 1.5185], device='cuda:6'), covar=tensor([0.3896, 0.1982, 0.5210, 0.3039, 0.1741, 0.2603, 0.1404, 0.4538], device='cuda:6'), in_proj_covar=tensor([0.0336, 0.0352, 0.0421, 0.0347, 0.0380, 0.0373, 0.0364, 0.0419], device='cuda:6'), out_proj_covar=tensor([9.9406e-05, 1.0488e-04, 1.2760e-04, 1.0401e-04, 1.1259e-04, 1.1072e-04, 1.0642e-04, 1.2590e-04], device='cuda:6') 2023-04-28 00:27:41,050 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=149202.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:27:48,400 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8935, 1.1830, 1.5678, 1.6530, 1.6259, 1.6809, 1.5749, 1.5548], device='cuda:6'), covar=tensor([0.3821, 0.4871, 0.3920, 0.3921, 0.5142, 0.6706, 0.4368, 0.4141], device='cuda:6'), in_proj_covar=tensor([0.0341, 0.0376, 0.0330, 0.0341, 0.0351, 0.0396, 0.0361, 0.0334], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-28 00:27:52,340 INFO [finetune.py:976] (6/7) Epoch 27, batch 300, loss[loss=0.1578, simple_loss=0.2481, pruned_loss=0.03371, over 4812.00 frames. ], tot_loss[loss=0.1655, simple_loss=0.2387, pruned_loss=0.0462, over 744854.30 frames. ], batch size: 38, lr: 2.94e-03, grad_scale: 32.0 2023-04-28 00:28:25,827 INFO [finetune.py:976] (6/7) Epoch 27, batch 350, loss[loss=0.1728, simple_loss=0.2445, pruned_loss=0.05053, over 4746.00 frames. ], tot_loss[loss=0.1669, simple_loss=0.2401, pruned_loss=0.04682, over 790904.40 frames. ], batch size: 54, lr: 2.94e-03, grad_scale: 32.0 2023-04-28 00:28:28,123 INFO [optim.py:369] (6/7) 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 00:28:51,978 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=149307.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:28:59,852 INFO [finetune.py:976] (6/7) Epoch 27, batch 400, loss[loss=0.1332, simple_loss=0.2149, pruned_loss=0.02582, over 4889.00 frames. ], tot_loss[loss=0.1666, simple_loss=0.241, pruned_loss=0.04617, over 828626.72 frames. ], batch size: 43, lr: 2.94e-03, grad_scale: 32.0 2023-04-28 00:29:20,121 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.8612, 2.2286, 2.1355, 2.2502, 1.9976, 2.1208, 2.1898, 2.1202], device='cuda:6'), covar=tensor([0.3476, 0.5963, 0.4922, 0.4278, 0.5615, 0.6622, 0.5884, 0.5609], device='cuda:6'), in_proj_covar=tensor([0.0340, 0.0375, 0.0330, 0.0340, 0.0351, 0.0395, 0.0360, 0.0333], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-28 00:29:39,547 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=149355.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:29:42,067 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8852, 1.7504, 2.0309, 2.2492, 2.2846, 1.8581, 1.5343, 2.1589], device='cuda:6'), covar=tensor([0.0796, 0.1072, 0.0635, 0.0562, 0.0558, 0.0828, 0.0751, 0.0540], device='cuda:6'), in_proj_covar=tensor([0.0183, 0.0200, 0.0182, 0.0169, 0.0176, 0.0175, 0.0151, 0.0177], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-28 00:29:48,685 INFO [finetune.py:976] (6/7) Epoch 27, batch 450, loss[loss=0.1775, simple_loss=0.2539, pruned_loss=0.05049, over 4905.00 frames. ], tot_loss[loss=0.1652, simple_loss=0.2395, pruned_loss=0.04543, over 858577.92 frames. ], batch size: 37, lr: 2.94e-03, grad_scale: 32.0 2023-04-28 00:29:49,393 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6362, 1.2482, 4.3457, 4.0650, 3.7831, 4.1450, 4.0197, 3.7836], device='cuda:6'), covar=tensor([0.7387, 0.6484, 0.1122, 0.1755, 0.1201, 0.1740, 0.1798, 0.1529], device='cuda:6'), in_proj_covar=tensor([0.0311, 0.0310, 0.0410, 0.0411, 0.0348, 0.0414, 0.0320, 0.0365], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-28 00:29:50,498 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=149372.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:29:50,986 INFO [optim.py:369] (6/7) 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 00:30:18,667 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.3572, 2.6967, 2.3215, 2.7165, 1.8578, 2.3025, 2.5504, 1.8803], device='cuda:6'), covar=tensor([0.1870, 0.1176, 0.0769, 0.1148, 0.3868, 0.1112, 0.1824, 0.2604], device='cuda:6'), in_proj_covar=tensor([0.0285, 0.0303, 0.0215, 0.0279, 0.0317, 0.0254, 0.0250, 0.0265], device='cuda:6'), out_proj_covar=tensor([1.1374e-04, 1.1917e-04, 8.4696e-05, 1.0966e-04, 1.2764e-04, 1.0013e-04, 1.0084e-04, 1.0467e-04], device='cuda:6') 2023-04-28 00:30:41,612 INFO [finetune.py:976] (6/7) Epoch 27, batch 500, loss[loss=0.1318, simple_loss=0.2066, pruned_loss=0.02854, over 4755.00 frames. ], tot_loss[loss=0.1643, simple_loss=0.2376, pruned_loss=0.04548, over 882310.72 frames. ], batch size: 27, lr: 2.93e-03, grad_scale: 32.0 2023-04-28 00:30:41,676 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=149420.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:31:00,201 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=3.60 vs. limit=5.0 2023-04-28 00:31:43,469 INFO [finetune.py:976] (6/7) Epoch 27, batch 550, loss[loss=0.1337, simple_loss=0.2115, pruned_loss=0.02801, over 4781.00 frames. ], tot_loss[loss=0.1621, simple_loss=0.2347, pruned_loss=0.0448, over 897587.46 frames. ], batch size: 26, lr: 2.93e-03, grad_scale: 32.0 2023-04-28 00:31:45,298 INFO [optim.py:369] (6/7) 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 00:32:04,624 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.3928, 1.3274, 1.8013, 1.7413, 1.2577, 1.1657, 1.4693, 0.8963], device='cuda:6'), covar=tensor([0.0520, 0.0655, 0.0351, 0.0654, 0.0762, 0.1078, 0.0532, 0.0568], device='cuda:6'), in_proj_covar=tensor([0.0070, 0.0067, 0.0066, 0.0069, 0.0074, 0.0095, 0.0072, 0.0063], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-28 00:32:06,948 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.4765, 2.9685, 0.9022, 1.8441, 2.3490, 1.5285, 4.2389, 2.1529], device='cuda:6'), covar=tensor([0.0635, 0.0817, 0.0906, 0.1200, 0.0541, 0.0972, 0.0181, 0.0609], device='cuda:6'), in_proj_covar=tensor([0.0050, 0.0064, 0.0047, 0.0046, 0.0049, 0.0051, 0.0072, 0.0050], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:6') 2023-04-28 00:32:12,539 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7976, 1.4388, 1.9373, 2.3443, 1.9090, 1.8358, 1.8911, 1.8590], device='cuda:6'), covar=tensor([0.4424, 0.6855, 0.6071, 0.5143, 0.5920, 0.8345, 0.7686, 0.8667], device='cuda:6'), in_proj_covar=tensor([0.0444, 0.0423, 0.0519, 0.0508, 0.0471, 0.0508, 0.0510, 0.0522], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-28 00:32:13,120 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 27, batch 600, loss[loss=0.1865, simple_loss=0.2577, pruned_loss=0.05772, over 4898.00 frames. ], tot_loss[loss=0.1649, simple_loss=0.2372, pruned_loss=0.04636, over 910858.67 frames. ], batch size: 37, lr: 2.93e-03, grad_scale: 32.0 2023-04-28 00:32:18,500 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.2236, 2.2007, 1.9454, 1.8320, 2.3205, 1.8271, 2.8136, 1.7378], device='cuda:6'), covar=tensor([0.3495, 0.1863, 0.4390, 0.2759, 0.1471, 0.2495, 0.1202, 0.3892], device='cuda:6'), in_proj_covar=tensor([0.0338, 0.0353, 0.0424, 0.0348, 0.0380, 0.0373, 0.0365, 0.0420], device='cuda:6'), out_proj_covar=tensor([9.9829e-05, 1.0507e-04, 1.2840e-04, 1.0441e-04, 1.1256e-04, 1.1063e-04, 1.0654e-04, 1.2628e-04], device='cuda:6') 2023-04-28 00:32:47,536 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=149565.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:32:50,410 INFO [finetune.py:976] (6/7) Epoch 27, batch 650, loss[loss=0.183, simple_loss=0.25, pruned_loss=0.058, over 4830.00 frames. ], tot_loss[loss=0.1678, simple_loss=0.2402, pruned_loss=0.04766, over 920823.89 frames. ], batch size: 30, lr: 2.93e-03, grad_scale: 32.0 2023-04-28 00:32:52,255 INFO [optim.py:369] (6/7) 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 00:32:53,626 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=149575.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:33:23,678 INFO [finetune.py:976] (6/7) Epoch 27, batch 700, loss[loss=0.1656, simple_loss=0.2319, pruned_loss=0.04968, over 4915.00 frames. ], tot_loss[loss=0.1684, simple_loss=0.2416, pruned_loss=0.04758, over 928551.30 frames. ], batch size: 38, lr: 2.93e-03, grad_scale: 32.0 2023-04-28 00:33:27,509 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=149626.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 00:33:44,020 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-04-28 00:33:56,991 INFO [finetune.py:976] (6/7) Epoch 27, batch 750, loss[loss=0.1698, simple_loss=0.2524, pruned_loss=0.04361, over 4862.00 frames. ], tot_loss[loss=0.1685, simple_loss=0.2417, pruned_loss=0.0477, over 933095.05 frames. ], batch size: 34, lr: 2.93e-03, grad_scale: 32.0 2023-04-28 00:33:58,788 INFO [optim.py:369] (6/7) 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 00:34:35,665 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.4571, 2.9442, 2.4042, 2.8262, 2.2601, 2.5139, 2.6972, 1.9717], device='cuda:6'), covar=tensor([0.1882, 0.1193, 0.0848, 0.1250, 0.2922, 0.1248, 0.1939, 0.2674], device='cuda:6'), in_proj_covar=tensor([0.0285, 0.0301, 0.0216, 0.0278, 0.0316, 0.0255, 0.0250, 0.0266], device='cuda:6'), out_proj_covar=tensor([1.1391e-04, 1.1861e-04, 8.4803e-05, 1.0960e-04, 1.2746e-04, 1.0018e-04, 1.0092e-04, 1.0475e-04], device='cuda:6') 2023-04-28 00:34:42,998 INFO [finetune.py:976] (6/7) Epoch 27, batch 800, loss[loss=0.186, simple_loss=0.2587, pruned_loss=0.05664, over 4894.00 frames. ], tot_loss[loss=0.1684, simple_loss=0.2422, pruned_loss=0.04733, over 938491.51 frames. ], batch size: 36, lr: 2.93e-03, grad_scale: 32.0 2023-04-28 00:34:44,901 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.1953, 1.5408, 1.3863, 1.7709, 1.6105, 1.8898, 1.4693, 3.4590], device='cuda:6'), covar=tensor([0.0585, 0.0787, 0.0763, 0.1144, 0.0611, 0.0519, 0.0694, 0.0131], device='cuda:6'), in_proj_covar=tensor([0.0037, 0.0039, 0.0039, 0.0043, 0.0040, 0.0038, 0.0038, 0.0055], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0013, 0.0013, 0.0015, 0.0014, 0.0013, 0.0013, 0.0016], device='cuda:6') 2023-04-28 00:35:14,880 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=149746.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 00:35:48,633 INFO [finetune.py:976] (6/7) Epoch 27, batch 850, loss[loss=0.1251, simple_loss=0.1995, pruned_loss=0.02538, over 4762.00 frames. ], tot_loss[loss=0.1661, simple_loss=0.2397, pruned_loss=0.04631, over 942042.74 frames. ], batch size: 27, lr: 2.93e-03, grad_scale: 32.0 2023-04-28 00:35:55,545 INFO [optim.py:369] (6/7) 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 00:35:55,671 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.2247, 1.5305, 1.3552, 1.9029, 1.6033, 1.9052, 1.4834, 3.3928], device='cuda:6'), covar=tensor([0.0600, 0.0782, 0.0785, 0.1103, 0.0624, 0.0461, 0.0677, 0.0132], device='cuda:6'), in_proj_covar=tensor([0.0037, 0.0039, 0.0039, 0.0043, 0.0040, 0.0038, 0.0038, 0.0055], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0013, 0.0013, 0.0015, 0.0014, 0.0013, 0.0013, 0.0016], device='cuda:6') 2023-04-28 00:36:07,910 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-04-28 00:36:18,679 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6591, 1.5403, 1.9649, 2.0425, 1.4273, 1.3655, 1.6836, 0.9785], device='cuda:6'), covar=tensor([0.0481, 0.0656, 0.0368, 0.0568, 0.0756, 0.1125, 0.0592, 0.0636], device='cuda:6'), in_proj_covar=tensor([0.0070, 0.0067, 0.0066, 0.0069, 0.0074, 0.0094, 0.0072, 0.0063], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-28 00:36:24,438 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=149807.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 00:36:33,260 INFO [finetune.py:976] (6/7) Epoch 27, batch 900, loss[loss=0.1676, simple_loss=0.2347, pruned_loss=0.05024, over 4798.00 frames. ], tot_loss[loss=0.1638, simple_loss=0.2367, pruned_loss=0.04544, over 947124.89 frames. ], batch size: 51, lr: 2.93e-03, grad_scale: 32.0 2023-04-28 00:36:54,808 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8320, 1.3010, 1.8440, 2.2757, 1.8880, 1.7630, 1.7977, 1.7798], device='cuda:6'), covar=tensor([0.4134, 0.6328, 0.5940, 0.4962, 0.5217, 0.6996, 0.7202, 0.8226], device='cuda:6'), in_proj_covar=tensor([0.0442, 0.0421, 0.0517, 0.0506, 0.0469, 0.0506, 0.0508, 0.0521], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-28 00:37:06,215 INFO [finetune.py:976] (6/7) Epoch 27, batch 950, loss[loss=0.1751, simple_loss=0.2354, pruned_loss=0.05738, over 4783.00 frames. ], tot_loss[loss=0.1642, simple_loss=0.2362, pruned_loss=0.0461, over 950440.39 frames. ], batch size: 29, lr: 2.93e-03, grad_scale: 32.0 2023-04-28 00:37:06,278 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=149870.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:37:08,535 INFO [optim.py:369] (6/7) 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 00:37:40,043 INFO [finetune.py:976] (6/7) Epoch 27, batch 1000, loss[loss=0.1957, simple_loss=0.2539, pruned_loss=0.06873, over 4145.00 frames. ], tot_loss[loss=0.1661, simple_loss=0.2385, pruned_loss=0.04688, over 950464.54 frames. ], batch size: 65, lr: 2.93e-03, grad_scale: 32.0 2023-04-28 00:37:40,701 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=149921.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 00:37:55,445 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.59 vs. limit=5.0 2023-04-28 00:38:13,693 INFO [finetune.py:976] (6/7) Epoch 27, batch 1050, loss[loss=0.1818, simple_loss=0.249, pruned_loss=0.05728, over 4927.00 frames. ], tot_loss[loss=0.1677, simple_loss=0.241, pruned_loss=0.04722, over 951708.32 frames. ], batch size: 33, lr: 2.93e-03, grad_scale: 32.0 2023-04-28 00:38:15,497 INFO [optim.py:369] (6/7) 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 00:38:30,537 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-04-28 00:38:48,335 INFO [finetune.py:976] (6/7) Epoch 27, batch 1100, loss[loss=0.2048, simple_loss=0.2885, pruned_loss=0.0605, over 4838.00 frames. ], tot_loss[loss=0.1691, simple_loss=0.2428, pruned_loss=0.04768, over 950984.31 frames. ], batch size: 49, lr: 2.93e-03, grad_scale: 32.0 2023-04-28 00:39:22,079 INFO [finetune.py:976] (6/7) Epoch 27, batch 1150, loss[loss=0.1799, simple_loss=0.2623, pruned_loss=0.04874, over 4918.00 frames. ], tot_loss[loss=0.1695, simple_loss=0.2435, pruned_loss=0.04777, over 952827.57 frames. ], batch size: 33, lr: 2.93e-03, grad_scale: 32.0 2023-04-28 00:39:23,883 INFO [optim.py:369] (6/7) 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 00:39:28,117 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=150079.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:39:43,694 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=150102.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 00:39:56,130 INFO [finetune.py:976] (6/7) Epoch 27, batch 1200, loss[loss=0.1677, simple_loss=0.2443, pruned_loss=0.04552, over 4816.00 frames. ], tot_loss[loss=0.1688, simple_loss=0.2424, pruned_loss=0.04757, over 954347.89 frames. ], batch size: 41, lr: 2.93e-03, grad_scale: 32.0 2023-04-28 00:40:10,335 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=150140.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:40:45,076 INFO [finetune.py:976] (6/7) Epoch 27, batch 1250, loss[loss=0.1392, simple_loss=0.2091, pruned_loss=0.03464, over 4870.00 frames. ], tot_loss[loss=0.1678, simple_loss=0.241, pruned_loss=0.04732, over 955611.71 frames. ], batch size: 31, lr: 2.93e-03, grad_scale: 32.0 2023-04-28 00:40:45,685 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=150170.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:40:47,940 INFO [optim.py:369] (6/7) 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 00:41:49,588 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=150218.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:41:50,800 INFO [finetune.py:976] (6/7) Epoch 27, batch 1300, loss[loss=0.1482, simple_loss=0.2167, pruned_loss=0.03981, over 4789.00 frames. ], tot_loss[loss=0.1643, simple_loss=0.2368, pruned_loss=0.04591, over 955419.68 frames. ], batch size: 29, lr: 2.93e-03, grad_scale: 32.0 2023-04-28 00:41:51,514 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=150221.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:42:56,734 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 27, batch 1350, loss[loss=0.1629, simple_loss=0.2419, pruned_loss=0.04194, over 4822.00 frames. ], tot_loss[loss=0.1635, simple_loss=0.236, pruned_loss=0.0455, over 954008.87 frames. ], batch size: 30, lr: 2.93e-03, grad_scale: 32.0 2023-04-28 00:42:57,447 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6795, 1.0705, 1.7453, 2.1410, 1.7523, 1.6853, 1.7365, 1.6719], device='cuda:6'), covar=tensor([0.4097, 0.6086, 0.5483, 0.5067, 0.5094, 0.6484, 0.6378, 0.8002], device='cuda:6'), in_proj_covar=tensor([0.0441, 0.0422, 0.0516, 0.0505, 0.0470, 0.0506, 0.0507, 0.0521], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-28 00:42:59,132 INFO [optim.py:369] (6/7) 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 00:42:59,366 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.15 vs. limit=5.0 2023-04-28 00:43:05,207 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=150274.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:43:32,307 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5024, 1.4292, 1.3487, 1.7595, 1.5615, 1.6780, 1.3881, 3.0970], device='cuda:6'), covar=tensor([0.0577, 0.0849, 0.0803, 0.1172, 0.0679, 0.0545, 0.0773, 0.0184], device='cuda:6'), in_proj_covar=tensor([0.0037, 0.0039, 0.0039, 0.0042, 0.0040, 0.0038, 0.0038, 0.0055], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0013, 0.0013, 0.0015, 0.0014, 0.0013, 0.0013, 0.0016], device='cuda:6') 2023-04-28 00:43:49,650 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.3385, 1.5853, 1.4215, 1.5546, 1.3389, 1.3207, 1.3338, 1.1236], device='cuda:6'), covar=tensor([0.1510, 0.1231, 0.0893, 0.1071, 0.3356, 0.1186, 0.1584, 0.1970], device='cuda:6'), in_proj_covar=tensor([0.0286, 0.0303, 0.0217, 0.0279, 0.0317, 0.0255, 0.0251, 0.0267], device='cuda:6'), out_proj_covar=tensor([1.1440e-04, 1.1943e-04, 8.5365e-05, 1.1000e-04, 1.2749e-04, 1.0037e-04, 1.0118e-04, 1.0543e-04], device='cuda:6') 2023-04-28 00:44:01,100 INFO [finetune.py:976] (6/7) Epoch 27, batch 1400, loss[loss=0.1512, simple_loss=0.2308, pruned_loss=0.03585, over 4173.00 frames. ], tot_loss[loss=0.1659, simple_loss=0.2389, pruned_loss=0.04642, over 953618.79 frames. ], batch size: 66, lr: 2.93e-03, grad_scale: 16.0 2023-04-28 00:44:22,644 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=150335.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:44:42,912 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.1693, 1.3795, 1.2566, 1.6781, 1.4997, 1.5785, 1.3506, 2.5004], device='cuda:6'), covar=tensor([0.0615, 0.0823, 0.0804, 0.1182, 0.0682, 0.0515, 0.0767, 0.0214], device='cuda:6'), in_proj_covar=tensor([0.0037, 0.0039, 0.0039, 0.0043, 0.0040, 0.0038, 0.0038, 0.0055], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0013, 0.0013, 0.0015, 0.0014, 0.0013, 0.0013, 0.0016], device='cuda:6') 2023-04-28 00:45:00,622 INFO [finetune.py:976] (6/7) Epoch 27, batch 1450, loss[loss=0.1538, simple_loss=0.2284, pruned_loss=0.03958, over 4762.00 frames. ], tot_loss[loss=0.1656, simple_loss=0.2392, pruned_loss=0.04598, over 954137.89 frames. ], batch size: 28, lr: 2.93e-03, grad_scale: 16.0 2023-04-28 00:45:03,057 INFO [optim.py:369] (6/7) 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 00:45:09,746 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6633, 2.0492, 1.7664, 2.0055, 1.4156, 1.6575, 1.7016, 1.4193], device='cuda:6'), covar=tensor([0.1797, 0.1255, 0.0884, 0.1041, 0.3692, 0.1243, 0.1857, 0.2312], device='cuda:6'), in_proj_covar=tensor([0.0286, 0.0302, 0.0216, 0.0278, 0.0315, 0.0254, 0.0250, 0.0266], device='cuda:6'), out_proj_covar=tensor([1.1407e-04, 1.1892e-04, 8.5012e-05, 1.0964e-04, 1.2683e-04, 9.9929e-05, 1.0075e-04, 1.0487e-04], device='cuda:6') 2023-04-28 00:45:22,906 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=150402.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 00:45:29,589 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9818, 2.4695, 1.0649, 1.3180, 1.9535, 1.1438, 3.1990, 1.5951], device='cuda:6'), covar=tensor([0.0733, 0.0597, 0.0773, 0.1259, 0.0501, 0.1038, 0.0271, 0.0652], device='cuda:6'), in_proj_covar=tensor([0.0050, 0.0063, 0.0046, 0.0045, 0.0049, 0.0051, 0.0072, 0.0050], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0007, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:6') 2023-04-28 00:45:31,724 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.81 vs. limit=2.0 2023-04-28 00:45:33,782 INFO [finetune.py:976] (6/7) Epoch 27, batch 1500, loss[loss=0.187, simple_loss=0.2616, pruned_loss=0.0562, over 4875.00 frames. ], tot_loss[loss=0.1669, simple_loss=0.241, pruned_loss=0.04646, over 956408.03 frames. ], batch size: 34, lr: 2.93e-03, grad_scale: 16.0 2023-04-28 00:45:44,405 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=150435.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:45:54,545 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=150450.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 00:45:57,590 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2023-04-28 00:46:06,957 INFO [finetune.py:976] (6/7) Epoch 27, batch 1550, loss[loss=0.1655, simple_loss=0.2328, pruned_loss=0.04906, over 4818.00 frames. ], tot_loss[loss=0.1673, simple_loss=0.2415, pruned_loss=0.04659, over 954805.14 frames. ], batch size: 30, lr: 2.93e-03, grad_scale: 16.0 2023-04-28 00:46:09,360 INFO [optim.py:369] (6/7) 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 00:46:22,384 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7687, 1.5158, 1.9748, 2.1956, 1.5744, 1.2617, 1.5983, 0.9384], device='cuda:6'), covar=tensor([0.0533, 0.0918, 0.0407, 0.0613, 0.0683, 0.1523, 0.0696, 0.0746], device='cuda:6'), in_proj_covar=tensor([0.0070, 0.0067, 0.0066, 0.0069, 0.0074, 0.0094, 0.0072, 0.0063], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-28 00:46:29,504 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0344, 1.8304, 2.4826, 2.5039, 1.7987, 1.7061, 1.9565, 1.1203], device='cuda:6'), covar=tensor([0.0527, 0.0846, 0.0397, 0.0901, 0.0819, 0.1032, 0.0744, 0.0735], device='cuda:6'), in_proj_covar=tensor([0.0070, 0.0067, 0.0066, 0.0069, 0.0074, 0.0094, 0.0072, 0.0063], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-28 00:46:32,482 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.5004, 3.1168, 1.0032, 1.7756, 2.4736, 1.4615, 4.2409, 1.9697], device='cuda:6'), covar=tensor([0.0669, 0.0743, 0.0886, 0.1243, 0.0510, 0.1029, 0.0168, 0.0618], device='cuda:6'), in_proj_covar=tensor([0.0050, 0.0063, 0.0046, 0.0045, 0.0049, 0.0051, 0.0072, 0.0050], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0007, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:6') 2023-04-28 00:46:40,768 INFO [finetune.py:976] (6/7) Epoch 27, batch 1600, loss[loss=0.1606, simple_loss=0.2338, pruned_loss=0.0437, over 4834.00 frames. ], tot_loss[loss=0.1663, simple_loss=0.2396, pruned_loss=0.04652, over 955301.56 frames. ], batch size: 30, lr: 2.93e-03, grad_scale: 16.0 2023-04-28 00:47:08,129 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=150537.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:47:12,164 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-04-28 00:47:30,609 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.57 vs. limit=2.0 2023-04-28 00:47:41,253 INFO [finetune.py:976] (6/7) Epoch 27, batch 1650, loss[loss=0.1781, simple_loss=0.2378, pruned_loss=0.05922, over 4832.00 frames. ], tot_loss[loss=0.1655, simple_loss=0.2379, pruned_loss=0.04658, over 955853.53 frames. ], batch size: 30, lr: 2.93e-03, grad_scale: 16.0 2023-04-28 00:47:43,702 INFO [optim.py:369] (6/7) 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 00:47:51,045 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6146, 1.4037, 0.3690, 1.2719, 1.3868, 1.4601, 1.3391, 1.4085], device='cuda:6'), covar=tensor([0.0452, 0.0366, 0.0387, 0.0539, 0.0281, 0.0497, 0.0494, 0.0533], device='cuda:6'), in_proj_covar=tensor([0.0028, 0.0023, 0.0022, 0.0028, 0.0019, 0.0027, 0.0027, 0.0028], device='cuda:6'), out_proj_covar=tensor([0.0052, 0.0046, 0.0038, 0.0053, 0.0039, 0.0051, 0.0051, 0.0053], device='cuda:6') 2023-04-28 00:47:54,593 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.1552, 2.5915, 1.1090, 1.5827, 2.0080, 1.3961, 3.5360, 1.8786], device='cuda:6'), covar=tensor([0.0638, 0.0643, 0.0815, 0.1130, 0.0511, 0.0897, 0.0288, 0.0558], device='cuda:6'), in_proj_covar=tensor([0.0050, 0.0064, 0.0047, 0.0046, 0.0049, 0.0051, 0.0072, 0.0050], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:6') 2023-04-28 00:47:54,609 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7636, 1.7058, 0.8707, 1.4109, 1.7894, 1.5952, 1.4897, 1.5744], device='cuda:6'), covar=tensor([0.0447, 0.0340, 0.0307, 0.0519, 0.0261, 0.0484, 0.0460, 0.0514], device='cuda:6'), in_proj_covar=tensor([0.0028, 0.0023, 0.0022, 0.0028, 0.0019, 0.0027, 0.0027, 0.0028], device='cuda:6'), out_proj_covar=tensor([0.0052, 0.0046, 0.0038, 0.0053, 0.0039, 0.0051, 0.0051, 0.0053], device='cuda:6') 2023-04-28 00:48:00,528 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=150598.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:48:14,794 INFO [finetune.py:976] (6/7) Epoch 27, batch 1700, loss[loss=0.1897, simple_loss=0.2582, pruned_loss=0.06061, over 4734.00 frames. ], tot_loss[loss=0.1641, simple_loss=0.2364, pruned_loss=0.04595, over 956220.03 frames. ], batch size: 54, lr: 2.93e-03, grad_scale: 16.0 2023-04-28 00:48:20,949 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=150630.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:48:48,163 INFO [finetune.py:976] (6/7) Epoch 27, batch 1750, loss[loss=0.1615, simple_loss=0.2386, pruned_loss=0.04225, over 4862.00 frames. ], tot_loss[loss=0.1655, simple_loss=0.2377, pruned_loss=0.04668, over 954766.23 frames. ], batch size: 31, lr: 2.93e-03, grad_scale: 16.0 2023-04-28 00:48:50,602 INFO [optim.py:369] (6/7) 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 00:49:21,909 INFO [finetune.py:976] (6/7) Epoch 27, batch 1800, loss[loss=0.1285, simple_loss=0.1875, pruned_loss=0.03477, over 4405.00 frames. ], tot_loss[loss=0.1663, simple_loss=0.2394, pruned_loss=0.04659, over 955177.03 frames. ], batch size: 19, lr: 2.93e-03, grad_scale: 16.0 2023-04-28 00:49:29,224 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=150732.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:49:31,013 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=150735.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:49:54,549 INFO [finetune.py:976] (6/7) Epoch 27, batch 1850, loss[loss=0.1545, simple_loss=0.2326, pruned_loss=0.03821, over 4770.00 frames. ], tot_loss[loss=0.1669, simple_loss=0.24, pruned_loss=0.04686, over 955568.24 frames. ], batch size: 28, lr: 2.93e-03, grad_scale: 16.0 2023-04-28 00:50:02,279 INFO [optim.py:369] (6/7) 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 00:50:13,365 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=150783.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:50:25,756 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=150793.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 00:50:56,513 INFO [finetune.py:976] (6/7) Epoch 27, batch 1900, loss[loss=0.1599, simple_loss=0.248, pruned_loss=0.03588, over 4892.00 frames. ], tot_loss[loss=0.1682, simple_loss=0.2415, pruned_loss=0.04744, over 956385.14 frames. ], batch size: 43, lr: 2.93e-03, grad_scale: 16.0 2023-04-28 00:51:18,110 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8729, 1.1618, 4.9504, 4.6061, 4.2748, 4.6443, 4.4001, 4.3851], device='cuda:6'), covar=tensor([0.7148, 0.6306, 0.0899, 0.1670, 0.0944, 0.1421, 0.1770, 0.1505], device='cuda:6'), in_proj_covar=tensor([0.0313, 0.0310, 0.0412, 0.0411, 0.0349, 0.0417, 0.0321, 0.0365], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-28 00:51:29,793 INFO [finetune.py:976] (6/7) Epoch 27, batch 1950, loss[loss=0.1825, simple_loss=0.2349, pruned_loss=0.065, over 4716.00 frames. ], tot_loss[loss=0.1682, simple_loss=0.2415, pruned_loss=0.04747, over 955018.44 frames. ], batch size: 59, lr: 2.92e-03, grad_scale: 16.0 2023-04-28 00:51:32,194 INFO [optim.py:369] (6/7) 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 00:51:44,423 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=150893.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:51:46,884 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([3.7776, 3.7900, 2.8647, 4.4358, 3.7756, 3.7776, 1.6180, 3.7404], device='cuda:6'), covar=tensor([0.1611, 0.1159, 0.3350, 0.1408, 0.3135, 0.1660, 0.5613, 0.2243], device='cuda:6'), in_proj_covar=tensor([0.0242, 0.0217, 0.0249, 0.0299, 0.0295, 0.0245, 0.0271, 0.0270], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-28 00:51:49,216 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=150900.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:51:55,098 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.12 vs. limit=2.0 2023-04-28 00:52:03,055 INFO [finetune.py:976] (6/7) Epoch 27, batch 2000, loss[loss=0.1364, simple_loss=0.217, pruned_loss=0.02787, over 4829.00 frames. ], tot_loss[loss=0.1669, simple_loss=0.2392, pruned_loss=0.04735, over 953315.40 frames. ], batch size: 30, lr: 2.92e-03, grad_scale: 16.0 2023-04-28 00:52:09,166 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=150930.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:52:51,913 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=150961.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:53:02,955 INFO [finetune.py:976] (6/7) Epoch 27, batch 2050, loss[loss=0.1396, simple_loss=0.2158, pruned_loss=0.03169, over 4824.00 frames. ], tot_loss[loss=0.1634, simple_loss=0.2355, pruned_loss=0.04567, over 954861.91 frames. ], batch size: 38, lr: 2.92e-03, grad_scale: 16.0 2023-04-28 00:53:05,398 INFO [optim.py:369] (6/7) 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 00:53:12,614 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=150978.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:54:07,739 INFO [finetune.py:976] (6/7) Epoch 27, batch 2100, loss[loss=0.1924, simple_loss=0.2635, pruned_loss=0.06067, over 4908.00 frames. ], tot_loss[loss=0.163, simple_loss=0.2349, pruned_loss=0.04556, over 954527.10 frames. ], batch size: 36, lr: 2.92e-03, grad_scale: 16.0 2023-04-28 00:55:11,248 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0423, 2.6277, 2.2968, 2.4701, 1.8938, 2.2627, 2.1352, 1.8267], device='cuda:6'), covar=tensor([0.1884, 0.1098, 0.0749, 0.1081, 0.3111, 0.1075, 0.1743, 0.2481], device='cuda:6'), in_proj_covar=tensor([0.0284, 0.0301, 0.0216, 0.0278, 0.0316, 0.0253, 0.0249, 0.0265], device='cuda:6'), out_proj_covar=tensor([1.1339e-04, 1.1855e-04, 8.4833e-05, 1.0936e-04, 1.2735e-04, 9.9623e-05, 1.0036e-04, 1.0437e-04], device='cuda:6') 2023-04-28 00:55:11,731 INFO [finetune.py:976] (6/7) Epoch 27, batch 2150, loss[loss=0.1315, simple_loss=0.2045, pruned_loss=0.0293, over 4683.00 frames. ], tot_loss[loss=0.1649, simple_loss=0.2376, pruned_loss=0.04607, over 953775.90 frames. ], batch size: 23, lr: 2.92e-03, grad_scale: 16.0 2023-04-28 00:55:19,343 INFO [optim.py:369] (6/7) 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 00:55:33,252 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=151088.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 00:56:16,163 INFO [finetune.py:976] (6/7) Epoch 27, batch 2200, loss[loss=0.1878, simple_loss=0.2617, pruned_loss=0.05693, over 4716.00 frames. ], tot_loss[loss=0.1678, simple_loss=0.241, pruned_loss=0.04735, over 952736.28 frames. ], batch size: 59, lr: 2.92e-03, grad_scale: 16.0 2023-04-28 00:57:19,655 INFO [finetune.py:976] (6/7) Epoch 27, batch 2250, loss[loss=0.1401, simple_loss=0.219, pruned_loss=0.0306, over 4824.00 frames. ], tot_loss[loss=0.1701, simple_loss=0.2432, pruned_loss=0.04845, over 952851.23 frames. ], batch size: 47, lr: 2.92e-03, grad_scale: 16.0 2023-04-28 00:57:25,680 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.4519, 1.9927, 2.3517, 2.9489, 2.3262, 1.8963, 1.9056, 2.1431], device='cuda:6'), covar=tensor([0.2935, 0.2955, 0.1503, 0.2133, 0.2632, 0.2472, 0.3485, 0.2067], device='cuda:6'), in_proj_covar=tensor([0.0294, 0.0247, 0.0228, 0.0314, 0.0221, 0.0234, 0.0228, 0.0186], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:6') 2023-04-28 00:57:27,226 INFO [optim.py:369] (6/7) 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 00:57:38,554 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=151184.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:57:49,145 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=151193.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:58:22,086 INFO [finetune.py:976] (6/7) Epoch 27, batch 2300, loss[loss=0.2001, simple_loss=0.2741, pruned_loss=0.06306, over 4803.00 frames. ], tot_loss[loss=0.1696, simple_loss=0.2429, pruned_loss=0.04812, over 953142.07 frames. ], batch size: 40, lr: 2.92e-03, grad_scale: 16.0 2023-04-28 00:58:42,721 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=151241.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:58:45,182 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=151245.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:58:52,343 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=151256.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:58:58,325 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0018, 1.5086, 1.5531, 1.6666, 2.1269, 1.7451, 1.5269, 1.4733], device='cuda:6'), covar=tensor([0.1605, 0.1591, 0.1846, 0.1417, 0.0941, 0.1669, 0.1887, 0.2288], device='cuda:6'), in_proj_covar=tensor([0.0318, 0.0310, 0.0354, 0.0289, 0.0330, 0.0309, 0.0302, 0.0379], device='cuda:6'), out_proj_covar=tensor([6.4889e-05, 6.3651e-05, 7.4314e-05, 5.7966e-05, 6.7453e-05, 6.4369e-05, 6.2503e-05, 8.0214e-05], device='cuda:6') 2023-04-28 00:59:01,206 INFO [finetune.py:976] (6/7) Epoch 27, batch 2350, loss[loss=0.1244, simple_loss=0.1991, pruned_loss=0.02491, over 4788.00 frames. ], tot_loss[loss=0.1671, simple_loss=0.2399, pruned_loss=0.04721, over 952554.85 frames. ], batch size: 29, lr: 2.92e-03, grad_scale: 16.0 2023-04-28 00:59:04,083 INFO [optim.py:369] (6/7) 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 00:59:10,175 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=151282.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:59:12,541 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.4583, 1.8595, 1.9628, 2.0379, 1.8950, 1.9007, 1.9794, 1.9590], device='cuda:6'), covar=tensor([0.3952, 0.5394, 0.4325, 0.4511, 0.5465, 0.7270, 0.5170, 0.4771], device='cuda:6'), in_proj_covar=tensor([0.0341, 0.0376, 0.0330, 0.0341, 0.0350, 0.0394, 0.0360, 0.0334], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-28 00:59:20,484 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=151298.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:59:34,710 INFO [finetune.py:976] (6/7) Epoch 27, batch 2400, loss[loss=0.1622, simple_loss=0.228, pruned_loss=0.0482, over 4823.00 frames. ], tot_loss[loss=0.1655, simple_loss=0.2376, pruned_loss=0.04667, over 952209.29 frames. ], batch size: 51, lr: 2.92e-03, grad_scale: 16.0 2023-04-28 00:59:50,911 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=151343.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:00:01,135 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=151359.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:00:01,376 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=5.56 vs. limit=5.0 2023-04-28 01:00:07,713 INFO [finetune.py:976] (6/7) Epoch 27, batch 2450, loss[loss=0.1858, simple_loss=0.2434, pruned_loss=0.06415, over 4868.00 frames. ], tot_loss[loss=0.1643, simple_loss=0.2356, pruned_loss=0.04649, over 950358.37 frames. ], batch size: 34, lr: 2.92e-03, grad_scale: 16.0 2023-04-28 01:00:10,591 INFO [optim.py:369] (6/7) 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 01:00:21,211 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=151388.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:00:41,810 INFO [finetune.py:976] (6/7) Epoch 27, batch 2500, loss[loss=0.2089, simple_loss=0.2721, pruned_loss=0.07282, over 4102.00 frames. ], tot_loss[loss=0.1665, simple_loss=0.2378, pruned_loss=0.0476, over 952837.07 frames. ], batch size: 65, lr: 2.92e-03, grad_scale: 16.0 2023-04-28 01:00:53,958 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=151436.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:00:59,313 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.2930, 1.5283, 1.3632, 1.4802, 1.2628, 1.3068, 1.2730, 1.1201], device='cuda:6'), covar=tensor([0.1779, 0.1234, 0.0870, 0.1204, 0.4058, 0.1145, 0.1830, 0.2275], device='cuda:6'), in_proj_covar=tensor([0.0285, 0.0302, 0.0216, 0.0279, 0.0318, 0.0254, 0.0249, 0.0265], device='cuda:6'), out_proj_covar=tensor([1.1373e-04, 1.1867e-04, 8.4900e-05, 1.0977e-04, 1.2797e-04, 9.9776e-05, 1.0041e-04, 1.0467e-04], device='cuda:6') 2023-04-28 01:01:15,578 INFO [finetune.py:976] (6/7) Epoch 27, batch 2550, loss[loss=0.1497, simple_loss=0.2313, pruned_loss=0.03403, over 4803.00 frames. ], tot_loss[loss=0.1697, simple_loss=0.2419, pruned_loss=0.04876, over 954395.81 frames. ], batch size: 51, lr: 2.92e-03, grad_scale: 16.0 2023-04-28 01:01:17,943 INFO [optim.py:369] (6/7) 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 01:01:48,925 INFO [finetune.py:976] (6/7) Epoch 27, batch 2600, loss[loss=0.1509, simple_loss=0.2316, pruned_loss=0.0351, over 4729.00 frames. ], tot_loss[loss=0.1696, simple_loss=0.2421, pruned_loss=0.04856, over 955096.31 frames. ], batch size: 54, lr: 2.92e-03, grad_scale: 16.0 2023-04-28 01:02:02,586 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=151540.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:02:19,395 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=151556.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:02:22,506 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-04-28 01:02:33,623 INFO [finetune.py:976] (6/7) Epoch 27, batch 2650, loss[loss=0.1827, simple_loss=0.2597, pruned_loss=0.05286, over 4889.00 frames. ], tot_loss[loss=0.1704, simple_loss=0.2437, pruned_loss=0.04852, over 955072.69 frames. ], batch size: 35, lr: 2.92e-03, grad_scale: 16.0 2023-04-28 01:02:41,499 INFO [optim.py:369] (6/7) 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 01:03:23,466 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=151604.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:03:39,057 INFO [finetune.py:976] (6/7) Epoch 27, batch 2700, loss[loss=0.1481, simple_loss=0.2257, pruned_loss=0.0353, over 4857.00 frames. ], tot_loss[loss=0.1696, simple_loss=0.243, pruned_loss=0.04814, over 956727.40 frames. ], batch size: 31, lr: 2.92e-03, grad_scale: 16.0 2023-04-28 01:04:05,032 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=151638.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:04:07,476 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9520, 1.8219, 2.4663, 2.5321, 1.6869, 1.5929, 1.8877, 1.0138], device='cuda:6'), covar=tensor([0.0600, 0.0720, 0.0359, 0.0820, 0.0764, 0.1113, 0.0633, 0.0734], device='cuda:6'), in_proj_covar=tensor([0.0069, 0.0067, 0.0065, 0.0068, 0.0074, 0.0094, 0.0072, 0.0063], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-28 01:04:26,282 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=151654.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:04:40,567 INFO [finetune.py:976] (6/7) Epoch 27, batch 2750, loss[loss=0.1264, simple_loss=0.2035, pruned_loss=0.02467, over 4745.00 frames. ], tot_loss[loss=0.1684, simple_loss=0.2411, pruned_loss=0.04786, over 958131.39 frames. ], batch size: 27, lr: 2.92e-03, grad_scale: 16.0 2023-04-28 01:04:48,521 INFO [optim.py:369] (6/7) 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 01:05:07,213 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=151688.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:05:07,842 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0882, 2.5543, 1.0850, 1.4924, 1.9498, 1.1755, 3.2623, 1.7383], device='cuda:6'), covar=tensor([0.0633, 0.0597, 0.0774, 0.1108, 0.0451, 0.0929, 0.0214, 0.0555], device='cuda:6'), in_proj_covar=tensor([0.0050, 0.0064, 0.0047, 0.0046, 0.0049, 0.0051, 0.0072, 0.0050], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:6') 2023-04-28 01:05:44,742 INFO [finetune.py:976] (6/7) Epoch 27, batch 2800, loss[loss=0.1391, simple_loss=0.2068, pruned_loss=0.03566, over 4769.00 frames. ], tot_loss[loss=0.1657, simple_loss=0.2379, pruned_loss=0.04679, over 952563.93 frames. ], batch size: 28, lr: 2.92e-03, grad_scale: 16.0 2023-04-28 01:06:01,216 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([0.9011, 1.1641, 3.3049, 3.0618, 2.9751, 3.2227, 3.2633, 2.9200], device='cuda:6'), covar=tensor([0.7317, 0.5651, 0.1485, 0.2153, 0.1505, 0.2078, 0.1593, 0.1774], device='cuda:6'), in_proj_covar=tensor([0.0314, 0.0310, 0.0412, 0.0411, 0.0351, 0.0418, 0.0322, 0.0366], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-28 01:06:24,878 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=151749.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:06:53,983 INFO [finetune.py:976] (6/7) Epoch 27, batch 2850, loss[loss=0.1567, simple_loss=0.2371, pruned_loss=0.03813, over 4872.00 frames. ], tot_loss[loss=0.1638, simple_loss=0.236, pruned_loss=0.04581, over 954044.85 frames. ], batch size: 31, lr: 2.92e-03, grad_scale: 16.0 2023-04-28 01:06:56,484 INFO [optim.py:369] (6/7) 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 01:07:18,030 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-04-28 01:07:29,973 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=151801.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:07:58,280 INFO [finetune.py:976] (6/7) Epoch 27, batch 2900, loss[loss=0.1878, simple_loss=0.2657, pruned_loss=0.05498, over 4854.00 frames. ], tot_loss[loss=0.1652, simple_loss=0.238, pruned_loss=0.04622, over 952863.12 frames. ], batch size: 44, lr: 2.92e-03, grad_scale: 16.0 2023-04-28 01:08:11,756 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=151840.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:08:16,654 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.40 vs. limit=5.0 2023-04-28 01:08:19,572 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.4333, 1.3811, 0.5379, 1.1562, 1.4455, 1.3222, 1.2579, 1.2876], device='cuda:6'), covar=tensor([0.0513, 0.0374, 0.0404, 0.0545, 0.0320, 0.0483, 0.0482, 0.0552], device='cuda:6'), in_proj_covar=tensor([0.0028, 0.0023, 0.0021, 0.0028, 0.0019, 0.0027, 0.0027, 0.0028], device='cuda:6'), out_proj_covar=tensor([0.0052, 0.0046, 0.0038, 0.0053, 0.0039, 0.0051, 0.0050, 0.0052], device='cuda:6') 2023-04-28 01:08:32,189 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=151862.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:08:40,598 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7702, 2.1784, 0.9485, 1.1041, 1.4834, 1.0301, 2.4379, 1.2187], device='cuda:6'), covar=tensor([0.0761, 0.0557, 0.0662, 0.1365, 0.0527, 0.1168, 0.0314, 0.0812], device='cuda:6'), in_proj_covar=tensor([0.0051, 0.0064, 0.0047, 0.0046, 0.0049, 0.0051, 0.0072, 0.0051], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:6') 2023-04-28 01:08:42,907 INFO [finetune.py:976] (6/7) Epoch 27, batch 2950, loss[loss=0.1671, simple_loss=0.2355, pruned_loss=0.04932, over 4871.00 frames. ], tot_loss[loss=0.1684, simple_loss=0.2415, pruned_loss=0.04766, over 954966.73 frames. ], batch size: 31, lr: 2.92e-03, grad_scale: 16.0 2023-04-28 01:08:44,223 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.1711, 1.5717, 1.3691, 1.6567, 1.6113, 2.0605, 1.3722, 3.6893], device='cuda:6'), covar=tensor([0.0582, 0.0814, 0.0799, 0.1256, 0.0625, 0.0476, 0.0737, 0.0122], device='cuda:6'), in_proj_covar=tensor([0.0038, 0.0039, 0.0039, 0.0043, 0.0040, 0.0038, 0.0038, 0.0055], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0013, 0.0013, 0.0015, 0.0014, 0.0013, 0.0013, 0.0016], device='cuda:6') 2023-04-28 01:08:50,554 INFO [optim.py:369] (6/7) 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 01:09:04,990 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=151888.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:09:48,000 INFO [finetune.py:976] (6/7) Epoch 27, batch 3000, loss[loss=0.1535, simple_loss=0.2351, pruned_loss=0.03593, over 4870.00 frames. ], tot_loss[loss=0.1682, simple_loss=0.2419, pruned_loss=0.0473, over 954323.11 frames. ], batch size: 31, lr: 2.92e-03, grad_scale: 16.0 2023-04-28 01:09:48,000 INFO [finetune.py:1001] (6/7) Computing validation loss 2023-04-28 01:10:08,676 INFO [finetune.py:1010] (6/7) Epoch 27, validation: loss=0.1539, simple_loss=0.2224, pruned_loss=0.04268, over 2265189.00 frames. 2023-04-28 01:10:08,677 INFO [finetune.py:1011] (6/7) Maximum memory allocated so far is 6435MB 2023-04-28 01:10:24,219 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.4656, 1.7717, 1.8954, 2.0207, 1.8581, 1.8300, 1.8388, 1.8420], device='cuda:6'), covar=tensor([0.4029, 0.5794, 0.4519, 0.4299, 0.5541, 0.7269, 0.5545, 0.5030], device='cuda:6'), in_proj_covar=tensor([0.0342, 0.0376, 0.0332, 0.0341, 0.0351, 0.0395, 0.0362, 0.0334], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-28 01:10:24,784 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=151938.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:10:35,079 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=151954.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:10:45,720 INFO [finetune.py:976] (6/7) Epoch 27, batch 3050, loss[loss=0.1582, simple_loss=0.2335, pruned_loss=0.04148, over 4816.00 frames. ], tot_loss[loss=0.168, simple_loss=0.2422, pruned_loss=0.04693, over 956844.83 frames. ], batch size: 39, lr: 2.92e-03, grad_scale: 16.0 2023-04-28 01:10:48,113 INFO [optim.py:369] (6/7) 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] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=151986.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:11:08,297 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([4.4883, 4.5004, 3.0222, 5.1284, 4.4395, 4.4728, 1.9785, 4.2928], device='cuda:6'), covar=tensor([0.1458, 0.0961, 0.3327, 0.1002, 0.3230, 0.1433, 0.5460, 0.2176], device='cuda:6'), in_proj_covar=tensor([0.0244, 0.0218, 0.0251, 0.0302, 0.0298, 0.0247, 0.0274, 0.0274], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-28 01:11:08,926 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=152002.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:11:20,460 INFO [finetune.py:976] (6/7) Epoch 27, batch 3100, loss[loss=0.1586, simple_loss=0.2293, pruned_loss=0.04395, over 4916.00 frames. ], tot_loss[loss=0.1657, simple_loss=0.2398, pruned_loss=0.04575, over 955986.39 frames. ], batch size: 36, lr: 2.92e-03, grad_scale: 16.0 2023-04-28 01:11:37,606 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=152044.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:11:43,180 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-04-28 01:11:46,316 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-04-28 01:11:54,347 INFO [finetune.py:976] (6/7) Epoch 27, batch 3150, loss[loss=0.1262, simple_loss=0.2023, pruned_loss=0.02502, over 4827.00 frames. ], tot_loss[loss=0.1634, simple_loss=0.237, pruned_loss=0.04496, over 954665.72 frames. ], batch size: 39, lr: 2.92e-03, grad_scale: 16.0 2023-04-28 01:11:56,742 INFO [optim.py:369] (6/7) 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 01:12:27,069 INFO [finetune.py:976] (6/7) Epoch 27, batch 3200, loss[loss=0.1611, simple_loss=0.2294, pruned_loss=0.04635, over 4935.00 frames. ], tot_loss[loss=0.1612, simple_loss=0.2339, pruned_loss=0.04424, over 956885.38 frames. ], batch size: 38, lr: 2.92e-03, grad_scale: 16.0 2023-04-28 01:12:39,404 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.2649, 2.0894, 1.8194, 1.8548, 2.3048, 1.8017, 2.7468, 1.7144], device='cuda:6'), covar=tensor([0.3516, 0.1956, 0.3835, 0.2669, 0.1567, 0.2413, 0.1272, 0.3803], device='cuda:6'), in_proj_covar=tensor([0.0336, 0.0353, 0.0421, 0.0349, 0.0379, 0.0373, 0.0364, 0.0419], device='cuda:6'), out_proj_covar=tensor([9.9230e-05, 1.0504e-04, 1.2744e-04, 1.0434e-04, 1.1209e-04, 1.1080e-04, 1.0634e-04, 1.2600e-04], device='cuda:6') 2023-04-28 01:12:51,246 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-04-28 01:12:52,028 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=152157.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:13:00,434 INFO [finetune.py:976] (6/7) Epoch 27, batch 3250, loss[loss=0.1627, simple_loss=0.2462, pruned_loss=0.03965, over 4829.00 frames. ], tot_loss[loss=0.1618, simple_loss=0.2342, pruned_loss=0.04469, over 955088.09 frames. ], batch size: 39, lr: 2.92e-03, grad_scale: 16.0 2023-04-28 01:13:02,808 INFO [optim.py:369] (6/7) 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 01:13:33,550 INFO [finetune.py:976] (6/7) Epoch 27, batch 3300, loss[loss=0.2044, simple_loss=0.2732, pruned_loss=0.06776, over 4858.00 frames. ], tot_loss[loss=0.1645, simple_loss=0.2375, pruned_loss=0.04577, over 955684.91 frames. ], batch size: 31, lr: 2.92e-03, grad_scale: 16.0 2023-04-28 01:13:36,715 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=152225.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:13:53,260 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9148, 1.4728, 2.0611, 2.3002, 2.0447, 1.9128, 1.9923, 1.9545], device='cuda:6'), covar=tensor([0.5114, 0.7245, 0.6922, 0.6632, 0.6362, 0.9104, 0.8312, 0.9701], device='cuda:6'), in_proj_covar=tensor([0.0441, 0.0420, 0.0516, 0.0505, 0.0468, 0.0504, 0.0504, 0.0520], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-28 01:14:13,061 INFO [finetune.py:976] (6/7) Epoch 27, batch 3350, loss[loss=0.1579, simple_loss=0.2332, pruned_loss=0.04127, over 4828.00 frames. ], tot_loss[loss=0.1663, simple_loss=0.2395, pruned_loss=0.04651, over 955703.69 frames. ], batch size: 33, lr: 2.92e-03, grad_scale: 16.0 2023-04-28 01:14:14,908 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=152273.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:14:15,408 INFO [optim.py:369] (6/7) 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 01:14:22,813 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7932, 4.0816, 0.6658, 2.1640, 2.1831, 2.6287, 2.3598, 0.9505], device='cuda:6'), covar=tensor([0.1304, 0.0866, 0.2147, 0.1185, 0.1053, 0.1035, 0.1383, 0.2158], device='cuda:6'), in_proj_covar=tensor([0.0118, 0.0238, 0.0135, 0.0121, 0.0131, 0.0152, 0.0117, 0.0118], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-28 01:14:35,114 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=152286.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:15:17,722 INFO [finetune.py:976] (6/7) Epoch 27, batch 3400, loss[loss=0.1816, simple_loss=0.2521, pruned_loss=0.05553, over 4280.00 frames. ], tot_loss[loss=0.1672, simple_loss=0.2404, pruned_loss=0.04702, over 954374.77 frames. ], batch size: 65, lr: 2.92e-03, grad_scale: 32.0 2023-04-28 01:15:37,356 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=152334.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:15:50,199 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=152344.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:16:22,889 INFO [finetune.py:976] (6/7) Epoch 27, batch 3450, loss[loss=0.1459, simple_loss=0.225, pruned_loss=0.03339, over 4871.00 frames. ], tot_loss[loss=0.1657, simple_loss=0.2394, pruned_loss=0.04601, over 953198.34 frames. ], batch size: 34, lr: 2.91e-03, grad_scale: 32.0 2023-04-28 01:16:31,452 INFO [optim.py:369] (6/7) 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 01:16:53,823 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=152392.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:17:27,396 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7842, 1.9480, 1.6795, 2.1430, 2.0999, 2.2915, 1.8503, 4.5949], device='cuda:6'), covar=tensor([0.0456, 0.0730, 0.0741, 0.1113, 0.0580, 0.0406, 0.0651, 0.0090], device='cuda:6'), in_proj_covar=tensor([0.0037, 0.0039, 0.0039, 0.0043, 0.0040, 0.0038, 0.0038, 0.0055], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0013, 0.0013, 0.0015, 0.0014, 0.0013, 0.0013, 0.0016], device='cuda:6') 2023-04-28 01:17:28,481 INFO [finetune.py:976] (6/7) Epoch 27, batch 3500, loss[loss=0.2057, simple_loss=0.2623, pruned_loss=0.07456, over 3964.00 frames. ], tot_loss[loss=0.1636, simple_loss=0.2369, pruned_loss=0.04511, over 953656.38 frames. ], batch size: 17, lr: 2.91e-03, grad_scale: 32.0 2023-04-28 01:18:00,041 INFO [zipformer.py:1188] (6/7) 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] (6/7) 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] (6/7) Epoch 27, batch 3550, loss[loss=0.1371, simple_loss=0.2168, pruned_loss=0.0287, over 4823.00 frames. ], tot_loss[loss=0.1616, simple_loss=0.2344, pruned_loss=0.04437, over 954624.19 frames. ], batch size: 39, lr: 2.91e-03, grad_scale: 32.0 2023-04-28 01:18:34,714 INFO [optim.py:369] (6/7) 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 01:19:25,048 INFO [zipformer.py:1188] (6/7) 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] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=152506.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:19:39,436 INFO [finetune.py:976] (6/7) Epoch 27, batch 3600, loss[loss=0.1527, simple_loss=0.2249, pruned_loss=0.04019, over 4829.00 frames. ], tot_loss[loss=0.1613, simple_loss=0.2336, pruned_loss=0.04453, over 953976.98 frames. ], batch size: 33, lr: 2.91e-03, grad_scale: 32.0 2023-04-28 01:20:18,894 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-04-28 01:20:42,431 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9826, 2.4002, 2.0505, 2.3663, 1.7041, 2.0185, 2.0230, 1.5638], device='cuda:6'), covar=tensor([0.2057, 0.1316, 0.0851, 0.1217, 0.3398, 0.1181, 0.1882, 0.2743], device='cuda:6'), in_proj_covar=tensor([0.0282, 0.0300, 0.0215, 0.0276, 0.0314, 0.0254, 0.0246, 0.0263], device='cuda:6'), out_proj_covar=tensor([1.1265e-04, 1.1816e-04, 8.4623e-05, 1.0851e-04, 1.2664e-04, 9.9684e-05, 9.9389e-05, 1.0382e-04], device='cuda:6') 2023-04-28 01:20:44,726 INFO [finetune.py:976] (6/7) Epoch 27, batch 3650, loss[loss=0.119, simple_loss=0.1931, pruned_loss=0.02241, over 4767.00 frames. ], tot_loss[loss=0.164, simple_loss=0.2365, pruned_loss=0.04578, over 951588.71 frames. ], batch size: 26, lr: 2.91e-03, grad_scale: 32.0 2023-04-28 01:20:51,638 INFO [optim.py:369] (6/7) 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 01:21:01,749 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=152581.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:21:02,416 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=152582.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:21:12,807 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=152589.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:21:49,540 INFO [finetune.py:976] (6/7) Epoch 27, batch 3700, loss[loss=0.1782, simple_loss=0.2544, pruned_loss=0.05103, over 4886.00 frames. ], tot_loss[loss=0.1659, simple_loss=0.239, pruned_loss=0.04633, over 949140.89 frames. ], batch size: 32, lr: 2.91e-03, grad_scale: 32.0 2023-04-28 01:22:00,790 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=152629.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:22:21,230 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=152643.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:22:31,103 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=152650.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:22:54,450 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7960, 2.0495, 1.8355, 2.0519, 1.5585, 1.7256, 1.7295, 1.4102], device='cuda:6'), covar=tensor([0.1475, 0.1354, 0.0859, 0.1016, 0.3811, 0.1129, 0.1794, 0.2303], device='cuda:6'), in_proj_covar=tensor([0.0282, 0.0301, 0.0215, 0.0275, 0.0314, 0.0253, 0.0246, 0.0263], device='cuda:6'), out_proj_covar=tensor([1.1262e-04, 1.1825e-04, 8.4725e-05, 1.0843e-04, 1.2653e-04, 9.9541e-05, 9.9257e-05, 1.0385e-04], device='cuda:6') 2023-04-28 01:22:56,765 INFO [finetune.py:976] (6/7) Epoch 27, batch 3750, loss[loss=0.1927, simple_loss=0.2747, pruned_loss=0.05537, over 4858.00 frames. ], tot_loss[loss=0.1677, simple_loss=0.2414, pruned_loss=0.04701, over 952372.42 frames. ], batch size: 44, lr: 2.91e-03, grad_scale: 32.0 2023-04-28 01:23:04,996 INFO [optim.py:369] (6/7) 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 01:23:45,783 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9180, 2.3377, 0.8299, 1.1587, 1.6871, 1.0756, 2.9505, 1.4858], device='cuda:6'), covar=tensor([0.0899, 0.0703, 0.0917, 0.1712, 0.0658, 0.1367, 0.0455, 0.0908], device='cuda:6'), in_proj_covar=tensor([0.0050, 0.0064, 0.0047, 0.0046, 0.0049, 0.0051, 0.0072, 0.0050], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:6') 2023-04-28 01:23:59,705 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.2725, 2.1433, 1.8974, 1.8601, 2.2958, 1.8861, 2.7494, 1.6518], device='cuda:6'), covar=tensor([0.3733, 0.1933, 0.4825, 0.2934, 0.1809, 0.2511, 0.1270, 0.4829], device='cuda:6'), in_proj_covar=tensor([0.0336, 0.0353, 0.0419, 0.0349, 0.0378, 0.0373, 0.0365, 0.0418], device='cuda:6'), out_proj_covar=tensor([9.9208e-05, 1.0507e-04, 1.2677e-04, 1.0454e-04, 1.1187e-04, 1.1064e-04, 1.0684e-04, 1.2587e-04], device='cuda:6') 2023-04-28 01:24:01,458 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.1100, 2.7141, 2.0923, 1.9922, 1.5225, 1.5493, 2.0921, 1.4105], device='cuda:6'), covar=tensor([0.1718, 0.1210, 0.1325, 0.1608, 0.2197, 0.2005, 0.0996, 0.2002], device='cuda:6'), in_proj_covar=tensor([0.0200, 0.0211, 0.0171, 0.0206, 0.0201, 0.0188, 0.0157, 0.0188], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:6') 2023-04-28 01:24:08,005 INFO [finetune.py:976] (6/7) Epoch 27, batch 3800, loss[loss=0.1824, simple_loss=0.2543, pruned_loss=0.05528, over 4912.00 frames. ], tot_loss[loss=0.1678, simple_loss=0.2415, pruned_loss=0.047, over 951855.32 frames. ], batch size: 37, lr: 2.91e-03, grad_scale: 32.0 2023-04-28 01:24:11,826 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=152726.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:25:13,180 INFO [finetune.py:976] (6/7) Epoch 27, batch 3850, loss[loss=0.1799, simple_loss=0.2469, pruned_loss=0.05646, over 4925.00 frames. ], tot_loss[loss=0.1672, simple_loss=0.2409, pruned_loss=0.0467, over 953183.44 frames. ], batch size: 38, lr: 2.91e-03, grad_scale: 32.0 2023-04-28 01:25:15,593 INFO [optim.py:369] (6/7) 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 01:25:15,745 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.2617, 1.7163, 2.1397, 2.5956, 2.1426, 1.7084, 1.5619, 2.0273], device='cuda:6'), covar=tensor([0.3160, 0.3180, 0.1615, 0.2067, 0.2560, 0.2660, 0.4340, 0.2075], device='cuda:6'), in_proj_covar=tensor([0.0295, 0.0248, 0.0229, 0.0316, 0.0223, 0.0236, 0.0230, 0.0187], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:6') 2023-04-28 01:25:33,581 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=152787.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:25:47,902 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=152801.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:25:54,536 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7816, 2.7607, 2.1559, 2.4416, 2.8284, 2.3679, 3.6899, 1.9155], device='cuda:6'), covar=tensor([0.4009, 0.2360, 0.4784, 0.3679, 0.1902, 0.2838, 0.1746, 0.4751], device='cuda:6'), in_proj_covar=tensor([0.0337, 0.0353, 0.0421, 0.0350, 0.0379, 0.0374, 0.0366, 0.0420], device='cuda:6'), out_proj_covar=tensor([9.9365e-05, 1.0524e-04, 1.2747e-04, 1.0489e-04, 1.1218e-04, 1.1104e-04, 1.0705e-04, 1.2621e-04], device='cuda:6') 2023-04-28 01:26:17,504 INFO [finetune.py:976] (6/7) Epoch 27, batch 3900, loss[loss=0.1185, simple_loss=0.1849, pruned_loss=0.02608, over 4754.00 frames. ], tot_loss[loss=0.1655, simple_loss=0.2384, pruned_loss=0.04625, over 953532.27 frames. ], batch size: 54, lr: 2.91e-03, grad_scale: 32.0 2023-04-28 01:26:27,462 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0394, 1.6385, 1.6001, 1.7932, 2.2290, 1.7864, 1.6177, 1.5582], device='cuda:6'), covar=tensor([0.1522, 0.1349, 0.1928, 0.1359, 0.0760, 0.1413, 0.1675, 0.2328], device='cuda:6'), in_proj_covar=tensor([0.0315, 0.0308, 0.0352, 0.0285, 0.0327, 0.0305, 0.0300, 0.0375], device='cuda:6'), out_proj_covar=tensor([6.4144e-05, 6.3162e-05, 7.3785e-05, 5.7103e-05, 6.6882e-05, 6.3627e-05, 6.2030e-05, 7.9387e-05], device='cuda:6') 2023-04-28 01:27:10,994 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9678, 1.4703, 1.8156, 1.7887, 1.7865, 1.4706, 0.8287, 1.4883], device='cuda:6'), covar=tensor([0.3255, 0.3002, 0.1709, 0.2176, 0.2492, 0.2592, 0.4228, 0.2024], device='cuda:6'), in_proj_covar=tensor([0.0295, 0.0248, 0.0229, 0.0316, 0.0223, 0.0236, 0.0229, 0.0187], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:6') 2023-04-28 01:27:22,907 INFO [finetune.py:976] (6/7) Epoch 27, batch 3950, loss[loss=0.1385, simple_loss=0.2079, pruned_loss=0.03458, over 4824.00 frames. ], tot_loss[loss=0.1624, simple_loss=0.2349, pruned_loss=0.04499, over 955499.34 frames. ], batch size: 38, lr: 2.91e-03, grad_scale: 32.0 2023-04-28 01:27:26,326 INFO [optim.py:369] (6/7) 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 01:27:36,690 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=152881.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:28:27,784 INFO [finetune.py:976] (6/7) Epoch 27, batch 4000, loss[loss=0.2022, simple_loss=0.2703, pruned_loss=0.06706, over 4826.00 frames. ], tot_loss[loss=0.1627, simple_loss=0.2351, pruned_loss=0.04515, over 955096.40 frames. ], batch size: 33, lr: 2.91e-03, grad_scale: 32.0 2023-04-28 01:28:39,702 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=152929.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:28:39,728 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=152929.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:28:50,572 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=152938.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:29:00,135 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=152945.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:29:32,519 INFO [finetune.py:976] (6/7) Epoch 27, batch 4050, loss[loss=0.1705, simple_loss=0.2459, pruned_loss=0.04753, over 4768.00 frames. ], tot_loss[loss=0.1641, simple_loss=0.2372, pruned_loss=0.04549, over 955502.38 frames. ], batch size: 28, lr: 2.91e-03, grad_scale: 32.0 2023-04-28 01:29:35,451 INFO [optim.py:369] (6/7) 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] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=152977.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:30:36,405 INFO [finetune.py:976] (6/7) Epoch 27, batch 4100, loss[loss=0.159, simple_loss=0.2259, pruned_loss=0.04608, over 4871.00 frames. ], tot_loss[loss=0.166, simple_loss=0.2398, pruned_loss=0.04616, over 956088.42 frames. ], batch size: 31, lr: 2.91e-03, grad_scale: 32.0 2023-04-28 01:31:29,766 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-04-28 01:31:40,933 INFO [finetune.py:976] (6/7) Epoch 27, batch 4150, loss[loss=0.2089, simple_loss=0.2831, pruned_loss=0.06736, over 4913.00 frames. ], tot_loss[loss=0.1664, simple_loss=0.2407, pruned_loss=0.04611, over 957444.77 frames. ], batch size: 36, lr: 2.91e-03, grad_scale: 32.0 2023-04-28 01:31:43,381 INFO [optim.py:369] (6/7) 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 01:31:59,387 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=153082.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:32:22,880 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=153101.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:32:43,346 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=153117.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:32:45,075 INFO [finetune.py:976] (6/7) Epoch 27, batch 4200, loss[loss=0.1632, simple_loss=0.2378, pruned_loss=0.04425, over 4856.00 frames. ], tot_loss[loss=0.1664, simple_loss=0.2411, pruned_loss=0.04582, over 957374.14 frames. ], batch size: 31, lr: 2.91e-03, grad_scale: 32.0 2023-04-28 01:33:26,261 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 27, batch 4250, loss[loss=0.155, simple_loss=0.224, pruned_loss=0.04295, over 4866.00 frames. ], tot_loss[loss=0.1656, simple_loss=0.24, pruned_loss=0.04563, over 957417.81 frames. ], batch size: 31, lr: 2.91e-03, grad_scale: 32.0 2023-04-28 01:33:57,253 INFO [optim.py:369] (6/7) 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 01:33:59,835 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=153178.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:34:34,599 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.5685, 1.8773, 2.0211, 2.1087, 1.9046, 1.9412, 2.0610, 2.0640], device='cuda:6'), covar=tensor([0.3985, 0.5315, 0.4357, 0.4516, 0.5756, 0.6999, 0.5300, 0.4744], device='cuda:6'), in_proj_covar=tensor([0.0342, 0.0374, 0.0330, 0.0341, 0.0350, 0.0394, 0.0361, 0.0334], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-28 01:34:54,745 INFO [finetune.py:976] (6/7) Epoch 27, batch 4300, loss[loss=0.1396, simple_loss=0.2104, pruned_loss=0.03443, over 4845.00 frames. ], tot_loss[loss=0.1647, simple_loss=0.2382, pruned_loss=0.04563, over 957515.38 frames. ], batch size: 47, lr: 2.91e-03, grad_scale: 32.0 2023-04-28 01:34:55,996 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7088, 2.0412, 1.8619, 2.0732, 1.5775, 1.7262, 1.6621, 1.3955], device='cuda:6'), covar=tensor([0.1582, 0.1188, 0.0685, 0.0923, 0.3218, 0.1023, 0.1695, 0.2176], device='cuda:6'), in_proj_covar=tensor([0.0284, 0.0301, 0.0217, 0.0276, 0.0316, 0.0255, 0.0248, 0.0265], device='cuda:6'), out_proj_covar=tensor([1.1335e-04, 1.1855e-04, 8.5185e-05, 1.0882e-04, 1.2722e-04, 1.0014e-04, 1.0018e-04, 1.0447e-04], device='cuda:6') 2023-04-28 01:35:17,881 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=153238.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:35:23,567 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 27, batch 4350, loss[loss=0.1337, simple_loss=0.209, pruned_loss=0.02918, over 4820.00 frames. ], tot_loss[loss=0.1623, simple_loss=0.2351, pruned_loss=0.04477, over 957683.09 frames. ], batch size: 40, lr: 2.91e-03, grad_scale: 32.0 2023-04-28 01:35:41,100 INFO [optim.py:369] (6/7) 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 01:35:44,146 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=153278.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:35:48,972 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=153286.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:35:51,277 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6025, 1.5001, 4.1576, 3.8764, 3.5840, 3.8623, 3.8131, 3.6183], device='cuda:6'), covar=tensor([0.6998, 0.5366, 0.1060, 0.1687, 0.1174, 0.1940, 0.2100, 0.1587], device='cuda:6'), in_proj_covar=tensor([0.0314, 0.0309, 0.0409, 0.0409, 0.0350, 0.0418, 0.0320, 0.0366], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-28 01:35:54,193 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=153293.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:36:12,554 INFO [finetune.py:976] (6/7) Epoch 27, batch 4400, loss[loss=0.1686, simple_loss=0.2322, pruned_loss=0.05245, over 4900.00 frames. ], tot_loss[loss=0.1644, simple_loss=0.237, pruned_loss=0.04587, over 958821.00 frames. ], batch size: 35, lr: 2.91e-03, grad_scale: 32.0 2023-04-28 01:36:15,154 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0225, 1.5083, 1.9348, 2.2336, 1.8898, 1.4928, 1.1446, 1.6816], device='cuda:6'), covar=tensor([0.3286, 0.3041, 0.1695, 0.2106, 0.2305, 0.2799, 0.4119, 0.1806], device='cuda:6'), in_proj_covar=tensor([0.0294, 0.0246, 0.0229, 0.0315, 0.0222, 0.0235, 0.0229, 0.0186], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:6') 2023-04-28 01:36:33,331 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=153339.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:37:16,505 INFO [finetune.py:976] (6/7) Epoch 27, batch 4450, loss[loss=0.169, simple_loss=0.2463, pruned_loss=0.04586, over 4917.00 frames. ], tot_loss[loss=0.1676, simple_loss=0.2414, pruned_loss=0.04688, over 959199.16 frames. ], batch size: 37, lr: 2.91e-03, grad_scale: 32.0 2023-04-28 01:37:18,878 INFO [optim.py:369] (6/7) 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 01:37:30,040 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=153382.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:38:10,739 INFO [finetune.py:976] (6/7) Epoch 27, batch 4500, loss[loss=0.2489, simple_loss=0.3039, pruned_loss=0.09696, over 4847.00 frames. ], tot_loss[loss=0.1694, simple_loss=0.2433, pruned_loss=0.04778, over 956754.79 frames. ], batch size: 47, lr: 2.91e-03, grad_scale: 32.0 2023-04-28 01:38:11,457 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=153421.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:38:16,907 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=153430.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:38:26,320 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.32 vs. limit=5.0 2023-04-28 01:38:43,734 INFO [finetune.py:976] (6/7) Epoch 27, batch 4550, loss[loss=0.1477, simple_loss=0.2175, pruned_loss=0.03896, over 4867.00 frames. ], tot_loss[loss=0.17, simple_loss=0.2446, pruned_loss=0.04774, over 956580.93 frames. ], batch size: 34, lr: 2.91e-03, grad_scale: 32.0 2023-04-28 01:38:45,570 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=153473.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:38:46,110 INFO [optim.py:369] (6/7) 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 01:38:51,093 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=153482.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:39:15,748 INFO [finetune.py:976] (6/7) Epoch 27, batch 4600, loss[loss=0.1881, simple_loss=0.2424, pruned_loss=0.06687, over 4889.00 frames. ], tot_loss[loss=0.1695, simple_loss=0.2435, pruned_loss=0.0477, over 955903.52 frames. ], batch size: 32, lr: 2.91e-03, grad_scale: 32.0 2023-04-28 01:39:25,605 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6103, 1.5056, 1.8764, 1.9327, 1.3537, 1.2937, 1.5689, 0.9559], device='cuda:6'), covar=tensor([0.0515, 0.0621, 0.0383, 0.0641, 0.0681, 0.1106, 0.0579, 0.0652], device='cuda:6'), in_proj_covar=tensor([0.0070, 0.0068, 0.0066, 0.0069, 0.0075, 0.0095, 0.0072, 0.0063], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-28 01:39:48,982 INFO [finetune.py:976] (6/7) Epoch 27, batch 4650, loss[loss=0.15, simple_loss=0.2232, pruned_loss=0.03843, over 4902.00 frames. ], tot_loss[loss=0.1679, simple_loss=0.2409, pruned_loss=0.04741, over 954899.08 frames. ], batch size: 43, lr: 2.91e-03, grad_scale: 32.0 2023-04-28 01:39:51,381 INFO [optim.py:369] (6/7) 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 01:40:26,360 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5405, 2.5987, 2.1194, 2.2286, 2.5593, 2.1436, 3.4512, 1.9547], device='cuda:6'), covar=tensor([0.3556, 0.2260, 0.3842, 0.3147, 0.1803, 0.2463, 0.1619, 0.4037], device='cuda:6'), in_proj_covar=tensor([0.0340, 0.0357, 0.0426, 0.0353, 0.0382, 0.0376, 0.0371, 0.0423], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-28 01:40:27,423 INFO [finetune.py:976] (6/7) Epoch 27, batch 4700, loss[loss=0.2039, simple_loss=0.2494, pruned_loss=0.0792, over 4872.00 frames. ], tot_loss[loss=0.1656, simple_loss=0.2377, pruned_loss=0.04674, over 956402.17 frames. ], batch size: 34, lr: 2.91e-03, grad_scale: 32.0 2023-04-28 01:40:41,153 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=153634.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:40:46,073 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5757, 1.2845, 0.5572, 1.3139, 1.4392, 1.4556, 1.3754, 1.4217], device='cuda:6'), covar=tensor([0.0481, 0.0386, 0.0388, 0.0552, 0.0294, 0.0490, 0.0479, 0.0571], device='cuda:6'), in_proj_covar=tensor([0.0028, 0.0023, 0.0022, 0.0028, 0.0019, 0.0028, 0.0027, 0.0028], device='cuda:6'), out_proj_covar=tensor([0.0052, 0.0046, 0.0039, 0.0053, 0.0039, 0.0051, 0.0051, 0.0053], device='cuda:6') 2023-04-28 01:41:06,296 INFO [finetune.py:976] (6/7) Epoch 27, batch 4750, loss[loss=0.1952, simple_loss=0.2657, pruned_loss=0.06237, over 4914.00 frames. ], tot_loss[loss=0.1638, simple_loss=0.236, pruned_loss=0.04576, over 956946.61 frames. ], batch size: 32, lr: 2.91e-03, grad_scale: 32.0 2023-04-28 01:41:08,702 INFO [optim.py:369] (6/7) 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 01:41:20,996 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8758, 2.4389, 1.8622, 1.8228, 1.3632, 1.4148, 1.9639, 1.2891], device='cuda:6'), covar=tensor([0.1595, 0.1283, 0.1363, 0.1569, 0.2232, 0.1817, 0.0917, 0.2006], device='cuda:6'), in_proj_covar=tensor([0.0197, 0.0209, 0.0169, 0.0204, 0.0200, 0.0185, 0.0155, 0.0187], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:6') 2023-04-28 01:41:39,700 INFO [finetune.py:976] (6/7) Epoch 27, batch 4800, loss[loss=0.1949, simple_loss=0.2779, pruned_loss=0.05591, over 4931.00 frames. ], tot_loss[loss=0.1658, simple_loss=0.2384, pruned_loss=0.04663, over 957227.52 frames. ], batch size: 42, lr: 2.91e-03, grad_scale: 32.0 2023-04-28 01:42:13,209 INFO [finetune.py:976] (6/7) Epoch 27, batch 4850, loss[loss=0.2027, simple_loss=0.2749, pruned_loss=0.06528, over 4895.00 frames. ], tot_loss[loss=0.1667, simple_loss=0.2404, pruned_loss=0.04655, over 954313.24 frames. ], batch size: 35, lr: 2.91e-03, grad_scale: 32.0 2023-04-28 01:42:15,624 INFO [zipformer.py:1188] (6/7) 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] (6/7) 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 01:42:17,994 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=153777.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:42:21,650 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([0.7452, 1.7244, 1.6067, 1.3322, 1.6947, 1.4367, 2.1743, 1.2691], device='cuda:6'), covar=tensor([0.3321, 0.1591, 0.4723, 0.2658, 0.1673, 0.2304, 0.1495, 0.5000], device='cuda:6'), in_proj_covar=tensor([0.0339, 0.0355, 0.0425, 0.0351, 0.0380, 0.0375, 0.0370, 0.0422], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-28 01:43:05,705 INFO [finetune.py:976] (6/7) Epoch 27, batch 4900, loss[loss=0.2031, simple_loss=0.273, pruned_loss=0.06658, over 4839.00 frames. ], tot_loss[loss=0.1687, simple_loss=0.2423, pruned_loss=0.04758, over 952586.71 frames. ], batch size: 47, lr: 2.91e-03, grad_scale: 32.0 2023-04-28 01:43:06,908 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=153821.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:43:45,234 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-04-28 01:44:09,301 INFO [finetune.py:976] (6/7) Epoch 27, batch 4950, loss[loss=0.1818, simple_loss=0.2539, pruned_loss=0.05485, over 4817.00 frames. ], tot_loss[loss=0.1687, simple_loss=0.2422, pruned_loss=0.04761, over 952326.18 frames. ], batch size: 33, lr: 2.90e-03, grad_scale: 32.0 2023-04-28 01:44:18,028 INFO [optim.py:369] (6/7) 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 01:45:13,177 INFO [finetune.py:976] (6/7) Epoch 27, batch 5000, loss[loss=0.1355, simple_loss=0.1985, pruned_loss=0.03619, over 4726.00 frames. ], tot_loss[loss=0.1679, simple_loss=0.241, pruned_loss=0.04736, over 952009.64 frames. ], batch size: 23, lr: 2.90e-03, grad_scale: 32.0 2023-04-28 01:45:21,169 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=153923.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:45:31,672 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7524, 2.2036, 1.7968, 1.5613, 1.3531, 1.3602, 1.8616, 1.2898], device='cuda:6'), covar=tensor([0.1792, 0.1293, 0.1428, 0.1677, 0.2335, 0.2019, 0.1010, 0.2107], device='cuda:6'), in_proj_covar=tensor([0.0197, 0.0209, 0.0169, 0.0204, 0.0200, 0.0186, 0.0156, 0.0187], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:6') 2023-04-28 01:45:34,550 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=153934.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:46:17,424 INFO [finetune.py:976] (6/7) Epoch 27, batch 5050, loss[loss=0.1672, simple_loss=0.2385, pruned_loss=0.04788, over 4898.00 frames. ], tot_loss[loss=0.1647, simple_loss=0.2378, pruned_loss=0.04578, over 952523.04 frames. ], batch size: 37, lr: 2.90e-03, grad_scale: 32.0 2023-04-28 01:46:25,376 INFO [optim.py:369] (6/7) 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] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=153982.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:46:34,091 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=153984.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:46:34,161 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.62 vs. limit=2.0 2023-04-28 01:46:57,619 INFO [finetune.py:976] (6/7) Epoch 27, batch 5100, loss[loss=0.1592, simple_loss=0.2349, pruned_loss=0.04178, over 4823.00 frames. ], tot_loss[loss=0.162, simple_loss=0.235, pruned_loss=0.04454, over 954505.29 frames. ], batch size: 41, lr: 2.90e-03, grad_scale: 32.0 2023-04-28 01:47:18,523 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=154049.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 01:47:31,250 INFO [finetune.py:976] (6/7) Epoch 27, batch 5150, loss[loss=0.1584, simple_loss=0.2459, pruned_loss=0.03542, over 4815.00 frames. ], tot_loss[loss=0.1618, simple_loss=0.2351, pruned_loss=0.04425, over 954714.08 frames. ], batch size: 40, lr: 2.90e-03, grad_scale: 32.0 2023-04-28 01:47:33,692 INFO [optim.py:369] (6/7) 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 01:47:35,068 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=154076.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:47:35,623 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=154077.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:47:41,346 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=154084.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:48:15,445 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=154110.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 01:48:26,896 INFO [finetune.py:976] (6/7) Epoch 27, batch 5200, loss[loss=0.1555, simple_loss=0.2348, pruned_loss=0.03807, over 4831.00 frames. ], tot_loss[loss=0.1638, simple_loss=0.2377, pruned_loss=0.04492, over 953006.55 frames. ], batch size: 33, lr: 2.90e-03, grad_scale: 32.0 2023-04-28 01:48:35,436 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=154125.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:48:38,228 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.59 vs. limit=2.0 2023-04-28 01:48:49,929 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=154137.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 01:49:00,251 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=154145.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:49:31,680 INFO [finetune.py:976] (6/7) Epoch 27, batch 5250, loss[loss=0.1923, simple_loss=0.2628, pruned_loss=0.06086, over 4887.00 frames. ], tot_loss[loss=0.1666, simple_loss=0.2411, pruned_loss=0.04607, over 956212.22 frames. ], batch size: 32, lr: 2.90e-03, grad_scale: 32.0 2023-04-28 01:49:34,132 INFO [optim.py:369] (6/7) 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 01:50:35,789 INFO [finetune.py:976] (6/7) Epoch 27, batch 5300, loss[loss=0.1715, simple_loss=0.2425, pruned_loss=0.05026, over 4751.00 frames. ], tot_loss[loss=0.1686, simple_loss=0.2431, pruned_loss=0.04707, over 951076.48 frames. ], batch size: 54, lr: 2.90e-03, grad_scale: 32.0 2023-04-28 01:50:36,530 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=154221.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:51:41,568 INFO [finetune.py:976] (6/7) Epoch 27, batch 5350, loss[loss=0.1356, simple_loss=0.2128, pruned_loss=0.0292, over 4815.00 frames. ], tot_loss[loss=0.1682, simple_loss=0.2428, pruned_loss=0.04686, over 952045.73 frames. ], batch size: 47, lr: 2.90e-03, grad_scale: 32.0 2023-04-28 01:51:49,115 INFO [optim.py:369] (6/7) 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 01:51:52,262 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=154279.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:51:54,128 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=154282.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 01:51:56,621 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=154286.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:51:59,600 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.75 vs. limit=2.0 2023-04-28 01:52:00,587 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=154291.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:52:19,626 INFO [finetune.py:976] (6/7) Epoch 27, batch 5400, loss[loss=0.1506, simple_loss=0.2319, pruned_loss=0.03467, over 4753.00 frames. ], tot_loss[loss=0.1672, simple_loss=0.2405, pruned_loss=0.0469, over 949865.79 frames. ], batch size: 27, lr: 2.90e-03, grad_scale: 64.0 2023-04-28 01:52:30,049 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=154337.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 01:52:37,711 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=154347.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:52:41,233 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=154352.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:52:50,433 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.8597, 4.6940, 1.3475, 2.8695, 3.2393, 3.4508, 3.0101, 1.5229], device='cuda:6'), covar=tensor([0.0936, 0.0761, 0.1822, 0.1022, 0.0715, 0.0891, 0.1075, 0.1753], device='cuda:6'), in_proj_covar=tensor([0.0117, 0.0238, 0.0135, 0.0120, 0.0131, 0.0152, 0.0117, 0.0117], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-28 01:52:52,201 INFO [finetune.py:976] (6/7) Epoch 27, batch 5450, loss[loss=0.1426, simple_loss=0.2284, pruned_loss=0.02843, over 4913.00 frames. ], tot_loss[loss=0.164, simple_loss=0.2366, pruned_loss=0.0457, over 951242.91 frames. ], batch size: 37, lr: 2.90e-03, grad_scale: 64.0 2023-04-28 01:52:54,629 INFO [optim.py:369] (6/7) 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 01:53:09,974 INFO [zipformer.py:1188] (6/7) 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] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=154405.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 01:53:31,463 INFO [finetune.py:976] (6/7) Epoch 27, batch 5500, loss[loss=0.1451, simple_loss=0.2186, pruned_loss=0.0358, over 4747.00 frames. ], tot_loss[loss=0.1616, simple_loss=0.2335, pruned_loss=0.04488, over 950099.95 frames. ], batch size: 27, lr: 2.90e-03, grad_scale: 64.0 2023-04-28 01:53:44,349 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=154432.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 01:53:54,062 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=154440.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:54:16,382 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6771, 3.9203, 0.9166, 1.8407, 1.9841, 2.5508, 2.2556, 1.0185], device='cuda:6'), covar=tensor([0.1770, 0.1378, 0.2390, 0.1887, 0.1328, 0.1366, 0.1777, 0.2343], device='cuda:6'), in_proj_covar=tensor([0.0117, 0.0239, 0.0135, 0.0121, 0.0131, 0.0152, 0.0117, 0.0117], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-28 01:54:35,187 INFO [finetune.py:976] (6/7) Epoch 27, batch 5550, loss[loss=0.223, simple_loss=0.3046, pruned_loss=0.07066, over 4836.00 frames. ], tot_loss[loss=0.1623, simple_loss=0.2348, pruned_loss=0.04489, over 951876.39 frames. ], batch size: 47, lr: 2.90e-03, grad_scale: 64.0 2023-04-28 01:54:37,628 INFO [optim.py:369] (6/7) 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 01:54:37,749 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=154474.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:55:31,959 INFO [finetune.py:976] (6/7) Epoch 27, batch 5600, loss[loss=0.1585, simple_loss=0.2368, pruned_loss=0.04008, over 4688.00 frames. ], tot_loss[loss=0.1654, simple_loss=0.2386, pruned_loss=0.04608, over 949624.90 frames. ], batch size: 23, lr: 2.90e-03, grad_scale: 64.0 2023-04-28 01:55:51,026 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=154535.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:55:52,720 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.61 vs. limit=2.0 2023-04-28 01:56:35,597 INFO [finetune.py:976] (6/7) Epoch 27, batch 5650, loss[loss=0.2771, simple_loss=0.3251, pruned_loss=0.1146, over 4062.00 frames. ], tot_loss[loss=0.1681, simple_loss=0.242, pruned_loss=0.04709, over 947426.95 frames. ], batch size: 65, lr: 2.90e-03, grad_scale: 64.0 2023-04-28 01:56:36,037 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-04-28 01:56:42,718 INFO [optim.py:369] (6/7) 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 01:56:44,489 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=154577.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 01:56:45,676 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=154579.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:57:36,096 INFO [finetune.py:976] (6/7) Epoch 27, batch 5700, loss[loss=0.1304, simple_loss=0.19, pruned_loss=0.03538, over 4469.00 frames. ], tot_loss[loss=0.1652, simple_loss=0.2375, pruned_loss=0.04651, over 925104.52 frames. ], batch size: 19, lr: 2.90e-03, grad_scale: 32.0 2023-04-28 01:57:45,549 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=154627.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:57:46,188 INFO [zipformer.py:1188] (6/7) 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] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=154642.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:58:12,011 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 28, batch 0, loss[loss=0.1501, simple_loss=0.2253, pruned_loss=0.03749, over 4716.00 frames. ], tot_loss[loss=0.1501, simple_loss=0.2253, pruned_loss=0.03749, over 4716.00 frames. ], batch size: 54, lr: 2.90e-03, grad_scale: 32.0 2023-04-28 01:58:12,561 INFO [finetune.py:1001] (6/7) Computing validation loss 2023-04-28 01:58:17,172 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5311, 1.2443, 1.3422, 1.3216, 1.6806, 1.4322, 1.2043, 1.3080], device='cuda:6'), covar=tensor([0.1947, 0.1403, 0.1909, 0.1359, 0.0801, 0.1366, 0.2006, 0.2531], device='cuda:6'), in_proj_covar=tensor([0.0313, 0.0306, 0.0350, 0.0286, 0.0325, 0.0304, 0.0300, 0.0375], device='cuda:6'), 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:6') 2023-04-28 01:58:29,428 INFO [finetune.py:1010] (6/7) 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] (6/7) Maximum memory allocated so far is 6435MB 2023-04-28 01:58:33,234 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.1333, 2.5602, 1.0846, 1.3760, 1.8979, 1.1607, 3.3155, 1.8099], device='cuda:6'), covar=tensor([0.0651, 0.0654, 0.0805, 0.1153, 0.0500, 0.1002, 0.0226, 0.0575], device='cuda:6'), in_proj_covar=tensor([0.0051, 0.0064, 0.0047, 0.0046, 0.0050, 0.0051, 0.0073, 0.0051], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:6') 2023-04-28 01:58:43,441 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.4954, 3.0984, 0.8163, 1.7857, 1.7209, 2.3036, 1.8152, 1.0418], device='cuda:6'), covar=tensor([0.1709, 0.1392, 0.2386, 0.1547, 0.1355, 0.1185, 0.1749, 0.2211], device='cuda:6'), in_proj_covar=tensor([0.0118, 0.0240, 0.0136, 0.0121, 0.0132, 0.0152, 0.0118, 0.0118], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-28 01:58:55,770 INFO [optim.py:369] (6/7) 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 01:59:03,273 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7090, 1.5431, 1.7658, 1.9980, 2.0680, 1.6234, 1.4114, 1.8069], device='cuda:6'), covar=tensor([0.0805, 0.1303, 0.0825, 0.0651, 0.0650, 0.0813, 0.0724, 0.0587], device='cuda:6'), in_proj_covar=tensor([0.0186, 0.0204, 0.0186, 0.0173, 0.0180, 0.0179, 0.0153, 0.0180], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-28 01:59:04,489 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=154689.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:59:06,858 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=154693.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 01:59:09,839 INFO [finetune.py:976] (6/7) Epoch 28, batch 50, loss[loss=0.1792, simple_loss=0.2519, pruned_loss=0.05328, over 4877.00 frames. ], tot_loss[loss=0.1685, simple_loss=0.2435, pruned_loss=0.04679, over 218166.90 frames. ], batch size: 32, lr: 2.90e-03, grad_scale: 32.0 2023-04-28 01:59:17,126 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=154705.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 01:59:20,243 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.4826, 1.8425, 1.9319, 2.0348, 1.9086, 1.9362, 2.0116, 1.9988], device='cuda:6'), covar=tensor([0.3484, 0.4966, 0.4362, 0.4347, 0.5342, 0.7230, 0.5125, 0.4652], device='cuda:6'), in_proj_covar=tensor([0.0341, 0.0374, 0.0329, 0.0341, 0.0350, 0.0394, 0.0360, 0.0333], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-28 01:59:26,323 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=154720.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:59:33,613 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=154732.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:59:38,585 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=154740.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:59:43,316 INFO [finetune.py:976] (6/7) Epoch 28, batch 100, loss[loss=0.1287, simple_loss=0.2094, pruned_loss=0.02394, over 4780.00 frames. ], tot_loss[loss=0.1627, simple_loss=0.2358, pruned_loss=0.0448, over 379997.87 frames. ], batch size: 26, lr: 2.90e-03, grad_scale: 32.0 2023-04-28 01:59:46,231 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.8666, 2.8359, 2.2145, 3.3006, 2.8853, 2.8816, 1.1736, 2.8687], device='cuda:6'), covar=tensor([0.2378, 0.1751, 0.3431, 0.2987, 0.2972, 0.2438, 0.5856, 0.2989], device='cuda:6'), in_proj_covar=tensor([0.0244, 0.0218, 0.0251, 0.0304, 0.0297, 0.0248, 0.0273, 0.0273], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-28 01:59:48,389 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=154753.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 01:59:50,132 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8582, 1.6556, 1.8645, 2.2140, 2.2929, 1.8260, 1.5702, 1.8753], device='cuda:6'), covar=tensor([0.0766, 0.1251, 0.0789, 0.0553, 0.0570, 0.0702, 0.0688, 0.0612], device='cuda:6'), in_proj_covar=tensor([0.0186, 0.0203, 0.0185, 0.0172, 0.0179, 0.0179, 0.0152, 0.0179], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-28 02:00:02,205 INFO [optim.py:369] (6/7) 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 02:00:05,349 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=154780.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:00:05,998 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=154781.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:00:10,233 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 28, batch 150, loss[loss=0.1333, simple_loss=0.1946, pruned_loss=0.03601, over 4833.00 frames. ], tot_loss[loss=0.1592, simple_loss=0.2317, pruned_loss=0.0433, over 509266.62 frames. ], batch size: 40, lr: 2.90e-03, grad_scale: 32.0 2023-04-28 02:00:22,397 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.53 vs. limit=2.0 2023-04-28 02:00:29,329 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.1200, 2.6851, 2.0174, 2.1853, 1.5509, 1.5721, 2.0628, 1.4338], device='cuda:6'), covar=tensor([0.1436, 0.1392, 0.1288, 0.1537, 0.2015, 0.1788, 0.0899, 0.1870], device='cuda:6'), in_proj_covar=tensor([0.0199, 0.0210, 0.0171, 0.0205, 0.0201, 0.0187, 0.0157, 0.0189], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:6') 2023-04-28 02:00:30,203 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.11 vs. limit=2.0 2023-04-28 02:00:31,777 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.13 vs. limit=2.0 2023-04-28 02:00:38,345 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=154830.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:00:49,255 INFO [finetune.py:976] (6/7) Epoch 28, batch 200, loss[loss=0.1796, simple_loss=0.2495, pruned_loss=0.05478, over 4900.00 frames. ], tot_loss[loss=0.1603, simple_loss=0.2321, pruned_loss=0.0442, over 607521.98 frames. ], batch size: 43, lr: 2.89e-03, grad_scale: 32.0 2023-04-28 02:01:08,235 INFO [optim.py:369] (6/7) 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 02:01:09,543 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=154877.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 02:01:22,149 INFO [finetune.py:976] (6/7) Epoch 28, batch 250, loss[loss=0.2184, simple_loss=0.2907, pruned_loss=0.07307, over 4803.00 frames. ], tot_loss[loss=0.1628, simple_loss=0.2352, pruned_loss=0.04521, over 685210.76 frames. ], batch size: 41, lr: 2.89e-03, grad_scale: 32.0 2023-04-28 02:01:41,472 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=154925.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:01:51,924 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=154942.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:02:00,128 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 28, batch 300, loss[loss=0.1492, simple_loss=0.2302, pruned_loss=0.03414, over 4855.00 frames. ], tot_loss[loss=0.1653, simple_loss=0.2391, pruned_loss=0.04573, over 746574.96 frames. ], batch size: 31, lr: 2.89e-03, grad_scale: 32.0 2023-04-28 02:02:36,612 INFO [optim.py:369] (6/7) 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 02:02:37,361 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=154976.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:02:47,502 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=154984.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:02:56,120 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=154990.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:02:57,998 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=154993.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 02:02:59,177 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=154995.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:03:06,552 INFO [finetune.py:976] (6/7) Epoch 28, batch 350, loss[loss=0.1672, simple_loss=0.2459, pruned_loss=0.04428, over 4921.00 frames. ], tot_loss[loss=0.1683, simple_loss=0.2421, pruned_loss=0.04731, over 792379.14 frames. ], batch size: 33, lr: 2.89e-03, grad_scale: 32.0 2023-04-28 02:03:29,258 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.1725, 1.6147, 2.0475, 2.5091, 2.0202, 1.6157, 1.4005, 1.9009], device='cuda:6'), covar=tensor([0.2847, 0.3092, 0.1598, 0.2037, 0.2371, 0.2477, 0.4074, 0.1791], device='cuda:6'), in_proj_covar=tensor([0.0295, 0.0248, 0.0230, 0.0316, 0.0224, 0.0237, 0.0230, 0.0187], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:6') 2023-04-28 02:03:33,791 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5157, 1.3310, 4.1297, 3.8732, 3.5787, 3.9421, 3.8738, 3.6317], device='cuda:6'), covar=tensor([0.7185, 0.5903, 0.1273, 0.1780, 0.1231, 0.1624, 0.1790, 0.1577], device='cuda:6'), in_proj_covar=tensor([0.0312, 0.0310, 0.0408, 0.0410, 0.0350, 0.0418, 0.0319, 0.0366], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-28 02:03:45,312 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=155037.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:03:47,612 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=155041.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 02:03:51,698 INFO [finetune.py:976] (6/7) Epoch 28, batch 400, loss[loss=0.1584, simple_loss=0.2196, pruned_loss=0.04857, over 4782.00 frames. ], tot_loss[loss=0.1682, simple_loss=0.2428, pruned_loss=0.04685, over 830485.67 frames. ], batch size: 26, lr: 2.89e-03, grad_scale: 32.0 2023-04-28 02:04:06,198 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=155069.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:04:11,281 INFO [optim.py:369] (6/7) 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] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=155076.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:04:13,373 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.02 vs. limit=5.0 2023-04-28 02:04:25,391 INFO [finetune.py:976] (6/7) Epoch 28, batch 450, loss[loss=0.1376, simple_loss=0.2186, pruned_loss=0.02826, over 4697.00 frames. ], tot_loss[loss=0.1671, simple_loss=0.2413, pruned_loss=0.04645, over 857595.04 frames. ], batch size: 23, lr: 2.89e-03, grad_scale: 32.0 2023-04-28 02:04:48,042 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=155130.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:04:48,086 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=155130.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:04:57,545 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.68 vs. limit=5.0 2023-04-28 02:04:58,938 INFO [finetune.py:976] (6/7) Epoch 28, batch 500, loss[loss=0.1655, simple_loss=0.2277, pruned_loss=0.05163, over 4833.00 frames. ], tot_loss[loss=0.1651, simple_loss=0.2384, pruned_loss=0.0459, over 880110.60 frames. ], batch size: 33, lr: 2.89e-03, grad_scale: 32.0 2023-04-28 02:05:17,834 INFO [optim.py:369] (6/7) 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 02:05:20,241 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=155178.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:05:32,276 INFO [finetune.py:976] (6/7) Epoch 28, batch 550, loss[loss=0.1443, simple_loss=0.2192, pruned_loss=0.03469, over 4821.00 frames. ], tot_loss[loss=0.1615, simple_loss=0.2345, pruned_loss=0.04429, over 895834.25 frames. ], batch size: 41, lr: 2.89e-03, grad_scale: 32.0 2023-04-28 02:05:42,333 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.2594, 2.0492, 1.8556, 1.7430, 2.2061, 1.7503, 2.5987, 1.7021], device='cuda:6'), covar=tensor([0.3571, 0.1870, 0.4141, 0.3000, 0.1666, 0.2478, 0.1639, 0.4105], device='cuda:6'), in_proj_covar=tensor([0.0342, 0.0358, 0.0425, 0.0353, 0.0382, 0.0378, 0.0373, 0.0424], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-28 02:05:53,395 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.1891, 1.6796, 2.0233, 2.2069, 2.0693, 1.6287, 1.1131, 1.7485], device='cuda:6'), covar=tensor([0.2839, 0.2852, 0.1514, 0.1847, 0.2212, 0.2354, 0.3891, 0.1744], device='cuda:6'), in_proj_covar=tensor([0.0294, 0.0247, 0.0230, 0.0316, 0.0223, 0.0236, 0.0229, 0.0186], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:6') 2023-04-28 02:05:56,877 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6154, 1.9457, 1.7650, 1.9516, 1.5161, 1.5938, 1.7055, 1.4478], device='cuda:6'), covar=tensor([0.1659, 0.1057, 0.0740, 0.0952, 0.2715, 0.1161, 0.1689, 0.1778], device='cuda:6'), in_proj_covar=tensor([0.0284, 0.0304, 0.0217, 0.0276, 0.0314, 0.0255, 0.0249, 0.0263], device='cuda:6'), out_proj_covar=tensor([1.1316e-04, 1.1979e-04, 8.5319e-05, 1.0878e-04, 1.2647e-04, 1.0021e-04, 1.0045e-04, 1.0378e-04], device='cuda:6') 2023-04-28 02:06:05,934 INFO [finetune.py:976] (6/7) Epoch 28, batch 600, loss[loss=0.2308, simple_loss=0.2869, pruned_loss=0.0873, over 4125.00 frames. ], tot_loss[loss=0.1634, simple_loss=0.236, pruned_loss=0.04544, over 907610.97 frames. ], batch size: 65, lr: 2.89e-03, grad_scale: 32.0 2023-04-28 02:06:23,749 INFO [optim.py:369] (6/7) 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 02:06:30,720 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=155284.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:06:39,097 INFO [finetune.py:976] (6/7) Epoch 28, batch 650, loss[loss=0.164, simple_loss=0.2487, pruned_loss=0.03963, over 4913.00 frames. ], tot_loss[loss=0.1654, simple_loss=0.2384, pruned_loss=0.04617, over 919043.41 frames. ], batch size: 43, lr: 2.89e-03, grad_scale: 32.0 2023-04-28 02:07:01,135 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.3001, 1.4664, 1.7036, 1.7975, 1.7364, 1.7460, 1.7506, 1.8057], device='cuda:6'), covar=tensor([0.3681, 0.4872, 0.4349, 0.4147, 0.5061, 0.6413, 0.4726, 0.4464], device='cuda:6'), in_proj_covar=tensor([0.0343, 0.0376, 0.0331, 0.0342, 0.0351, 0.0394, 0.0362, 0.0334], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-28 02:07:02,746 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=155332.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:07:02,755 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 28, batch 700, loss[loss=0.1848, simple_loss=0.2648, pruned_loss=0.0524, over 4806.00 frames. ], tot_loss[loss=0.1661, simple_loss=0.2393, pruned_loss=0.04649, over 926307.82 frames. ], batch size: 41, lr: 2.89e-03, grad_scale: 32.0 2023-04-28 02:07:34,633 INFO [optim.py:369] (6/7) 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 02:07:40,181 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=155376.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:07:51,755 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0221, 1.2621, 5.2005, 4.8990, 4.4594, 5.0226, 4.6791, 4.5930], device='cuda:6'), covar=tensor([0.6868, 0.6352, 0.1025, 0.1879, 0.1227, 0.1291, 0.0991, 0.1740], device='cuda:6'), in_proj_covar=tensor([0.0311, 0.0309, 0.0407, 0.0408, 0.0349, 0.0417, 0.0317, 0.0366], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-28 02:08:09,447 INFO [finetune.py:976] (6/7) Epoch 28, batch 750, loss[loss=0.1367, simple_loss=0.2122, pruned_loss=0.03065, over 4684.00 frames. ], tot_loss[loss=0.1673, simple_loss=0.2409, pruned_loss=0.04682, over 934226.78 frames. ], batch size: 23, lr: 2.89e-03, grad_scale: 32.0 2023-04-28 02:08:41,179 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=155424.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:08:41,840 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=155425.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:09:01,854 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7030, 1.7159, 0.8991, 1.3128, 1.8499, 1.5707, 1.4060, 1.4625], device='cuda:6'), covar=tensor([0.0505, 0.0361, 0.0336, 0.0549, 0.0265, 0.0504, 0.0493, 0.0590], device='cuda:6'), in_proj_covar=tensor([0.0027, 0.0023, 0.0022, 0.0028, 0.0019, 0.0027, 0.0027, 0.0028], device='cuda:6'), out_proj_covar=tensor([0.0052, 0.0046, 0.0039, 0.0053, 0.0039, 0.0051, 0.0051, 0.0052], device='cuda:6') 2023-04-28 02:09:12,223 INFO [finetune.py:976] (6/7) Epoch 28, batch 800, loss[loss=0.1515, simple_loss=0.2354, pruned_loss=0.03385, over 4715.00 frames. ], tot_loss[loss=0.1654, simple_loss=0.2394, pruned_loss=0.0457, over 938629.53 frames. ], batch size: 59, lr: 2.89e-03, grad_scale: 32.0 2023-04-28 02:09:23,007 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=155457.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:09:36,894 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.11 vs. limit=2.0 2023-04-28 02:09:42,709 INFO [optim.py:369] (6/7) 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 02:09:59,774 INFO [finetune.py:976] (6/7) Epoch 28, batch 850, loss[loss=0.1302, simple_loss=0.2027, pruned_loss=0.02885, over 4864.00 frames. ], tot_loss[loss=0.1648, simple_loss=0.2384, pruned_loss=0.04562, over 942397.56 frames. ], batch size: 31, lr: 2.89e-03, grad_scale: 32.0 2023-04-28 02:10:12,004 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=155518.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:10:27,088 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9963, 1.5012, 1.8354, 1.8129, 1.7488, 1.4937, 0.8380, 1.5200], device='cuda:6'), covar=tensor([0.3113, 0.3000, 0.1700, 0.2081, 0.2564, 0.2578, 0.3987, 0.1805], device='cuda:6'), in_proj_covar=tensor([0.0294, 0.0247, 0.0230, 0.0316, 0.0223, 0.0236, 0.0229, 0.0186], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:6') 2023-04-28 02:10:28,293 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.2839, 1.7797, 2.1370, 2.4997, 2.0871, 1.6933, 1.3730, 1.9438], device='cuda:6'), covar=tensor([0.2619, 0.2701, 0.1389, 0.1673, 0.2201, 0.2235, 0.3686, 0.1619], device='cuda:6'), in_proj_covar=tensor([0.0294, 0.0247, 0.0230, 0.0316, 0.0223, 0.0236, 0.0229, 0.0186], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:6') 2023-04-28 02:10:32,817 INFO [finetune.py:976] (6/7) Epoch 28, batch 900, loss[loss=0.1321, simple_loss=0.204, pruned_loss=0.03014, over 4908.00 frames. ], tot_loss[loss=0.1622, simple_loss=0.2355, pruned_loss=0.04447, over 946719.44 frames. ], batch size: 36, lr: 2.89e-03, grad_scale: 32.0 2023-04-28 02:10:49,542 INFO [optim.py:369] (6/7) 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 02:11:05,996 INFO [finetune.py:976] (6/7) Epoch 28, batch 950, loss[loss=0.1533, simple_loss=0.2045, pruned_loss=0.05108, over 4059.00 frames. ], tot_loss[loss=0.1608, simple_loss=0.2336, pruned_loss=0.04401, over 949500.12 frames. ], batch size: 17, lr: 2.89e-03, grad_scale: 32.0 2023-04-28 02:11:10,959 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=155605.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:11:27,348 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 28, batch 1000, loss[loss=0.1629, simple_loss=0.2296, pruned_loss=0.04811, over 4766.00 frames. ], tot_loss[loss=0.1647, simple_loss=0.2378, pruned_loss=0.04583, over 949918.84 frames. ], batch size: 27, lr: 2.89e-03, grad_scale: 32.0 2023-04-28 02:11:42,089 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.61 vs. limit=2.0 2023-04-28 02:11:45,566 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=155657.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:11:51,063 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=155666.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:11:56,468 INFO [optim.py:369] (6/7) 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 02:11:59,553 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=155680.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:12:12,448 INFO [finetune.py:976] (6/7) Epoch 28, batch 1050, loss[loss=0.1753, simple_loss=0.2495, pruned_loss=0.05051, over 4761.00 frames. ], tot_loss[loss=0.1656, simple_loss=0.2398, pruned_loss=0.04572, over 951668.26 frames. ], batch size: 28, lr: 2.89e-03, grad_scale: 32.0 2023-04-28 02:12:25,695 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=155718.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:12:29,975 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=155725.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:12:43,380 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 2023-04-28 02:12:45,201 INFO [finetune.py:976] (6/7) Epoch 28, batch 1100, loss[loss=0.1103, simple_loss=0.1916, pruned_loss=0.01452, over 4773.00 frames. ], tot_loss[loss=0.167, simple_loss=0.2413, pruned_loss=0.04641, over 953249.55 frames. ], batch size: 26, lr: 2.89e-03, grad_scale: 32.0 2023-04-28 02:13:13,292 INFO [zipformer.py:1188] (6/7) 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] (6/7) 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 02:13:30,939 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=155784.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:13:46,059 INFO [finetune.py:976] (6/7) Epoch 28, batch 1150, loss[loss=0.1723, simple_loss=0.2456, pruned_loss=0.04953, over 4755.00 frames. ], tot_loss[loss=0.1691, simple_loss=0.2436, pruned_loss=0.0473, over 953032.86 frames. ], batch size: 28, lr: 2.89e-03, grad_scale: 32.0 2023-04-28 02:13:58,498 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=155813.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:14:18,098 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=155845.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 02:14:19,789 INFO [finetune.py:976] (6/7) Epoch 28, batch 1200, loss[loss=0.1588, simple_loss=0.2291, pruned_loss=0.04421, over 4753.00 frames. ], tot_loss[loss=0.1676, simple_loss=0.2416, pruned_loss=0.04678, over 952826.76 frames. ], batch size: 28, lr: 2.89e-03, grad_scale: 32.0 2023-04-28 02:14:33,713 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.58 vs. limit=2.0 2023-04-28 02:14:54,753 INFO [optim.py:369] (6/7) 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 02:15:15,262 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.3569, 2.6699, 1.3706, 1.7557, 2.2728, 1.6097, 3.5988, 2.1356], device='cuda:6'), covar=tensor([0.0640, 0.0678, 0.0707, 0.1077, 0.0455, 0.0873, 0.0306, 0.0535], device='cuda:6'), in_proj_covar=tensor([0.0051, 0.0065, 0.0047, 0.0046, 0.0050, 0.0051, 0.0073, 0.0051], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:6') 2023-04-28 02:15:25,826 INFO [finetune.py:976] (6/7) Epoch 28, batch 1250, loss[loss=0.172, simple_loss=0.2401, pruned_loss=0.05197, over 4934.00 frames. ], tot_loss[loss=0.1666, simple_loss=0.2396, pruned_loss=0.04681, over 953377.65 frames. ], batch size: 38, lr: 2.89e-03, grad_scale: 32.0 2023-04-28 02:16:00,970 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-04-28 02:16:20,268 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 2023-04-28 02:16:22,988 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7088, 1.5730, 1.7843, 2.0316, 2.1049, 1.6515, 1.4099, 1.8739], device='cuda:6'), covar=tensor([0.0760, 0.1266, 0.0807, 0.0604, 0.0626, 0.0767, 0.0720, 0.0560], device='cuda:6'), in_proj_covar=tensor([0.0183, 0.0200, 0.0183, 0.0168, 0.0177, 0.0176, 0.0149, 0.0176], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-28 02:16:33,390 INFO [finetune.py:976] (6/7) Epoch 28, batch 1300, loss[loss=0.2311, simple_loss=0.2912, pruned_loss=0.08553, over 4936.00 frames. ], tot_loss[loss=0.165, simple_loss=0.2378, pruned_loss=0.04615, over 953664.98 frames. ], batch size: 38, lr: 2.89e-03, grad_scale: 32.0 2023-04-28 02:16:53,068 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=155961.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:16:53,773 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9394, 1.4595, 2.0069, 2.3357, 2.0205, 1.8894, 1.9454, 1.9404], device='cuda:6'), covar=tensor([0.4807, 0.7447, 0.6928, 0.6217, 0.6280, 0.8979, 0.8800, 0.9927], device='cuda:6'), in_proj_covar=tensor([0.0443, 0.0424, 0.0520, 0.0508, 0.0473, 0.0510, 0.0511, 0.0525], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-28 02:17:07,932 INFO [optim.py:369] (6/7) 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] (6/7) Epoch 28, batch 1350, loss[loss=0.145, simple_loss=0.212, pruned_loss=0.03896, over 4795.00 frames. ], tot_loss[loss=0.1661, simple_loss=0.2381, pruned_loss=0.04705, over 952468.44 frames. ], batch size: 26, lr: 2.89e-03, grad_scale: 32.0 2023-04-28 02:17:56,776 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 28, batch 1400, loss[loss=0.1753, simple_loss=0.2578, pruned_loss=0.04644, over 4825.00 frames. ], tot_loss[loss=0.1661, simple_loss=0.2393, pruned_loss=0.04641, over 952631.66 frames. ], batch size: 39, lr: 2.89e-03, grad_scale: 32.0 2023-04-28 02:19:12,212 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=156070.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:19:20,859 INFO [optim.py:369] (6/7) 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 02:19:36,469 INFO [finetune.py:976] (6/7) Epoch 28, batch 1450, loss[loss=0.1573, simple_loss=0.2318, pruned_loss=0.04139, over 4928.00 frames. ], tot_loss[loss=0.167, simple_loss=0.2407, pruned_loss=0.04664, over 952566.40 frames. ], batch size: 38, lr: 2.89e-03, grad_scale: 32.0 2023-04-28 02:19:46,653 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=156113.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:19:59,753 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=156131.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 02:20:05,159 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=156140.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 02:20:09,928 INFO [finetune.py:976] (6/7) Epoch 28, batch 1500, loss[loss=0.1982, simple_loss=0.2849, pruned_loss=0.05568, over 4722.00 frames. ], tot_loss[loss=0.1685, simple_loss=0.2425, pruned_loss=0.04722, over 951629.40 frames. ], batch size: 54, lr: 2.89e-03, grad_scale: 32.0 2023-04-28 02:20:23,250 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=156161.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:20:32,705 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.3063, 1.3370, 1.4066, 1.6190, 1.6465, 1.3260, 1.0329, 1.5098], device='cuda:6'), covar=tensor([0.0899, 0.1284, 0.1006, 0.0644, 0.0750, 0.0829, 0.0856, 0.0644], device='cuda:6'), in_proj_covar=tensor([0.0184, 0.0202, 0.0184, 0.0170, 0.0178, 0.0177, 0.0150, 0.0178], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-28 02:20:43,690 INFO [optim.py:369] (6/7) 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 02:21:08,781 INFO [finetune.py:976] (6/7) Epoch 28, batch 1550, loss[loss=0.1851, simple_loss=0.2631, pruned_loss=0.05355, over 4815.00 frames. ], tot_loss[loss=0.1673, simple_loss=0.2415, pruned_loss=0.04652, over 953321.65 frames. ], batch size: 33, lr: 2.89e-03, grad_scale: 32.0 2023-04-28 02:21:15,296 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5361, 1.1523, 1.2862, 1.1764, 1.6752, 1.3649, 1.1747, 1.2662], device='cuda:6'), covar=tensor([0.1389, 0.1279, 0.1693, 0.1330, 0.0790, 0.1297, 0.1495, 0.2091], device='cuda:6'), in_proj_covar=tensor([0.0315, 0.0307, 0.0351, 0.0285, 0.0325, 0.0306, 0.0301, 0.0376], device='cuda:6'), out_proj_covar=tensor([6.4283e-05, 6.2959e-05, 7.3680e-05, 5.6984e-05, 6.6261e-05, 6.3808e-05, 6.2190e-05, 7.9722e-05], device='cuda:6') 2023-04-28 02:21:31,135 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 28, batch 1600, loss[loss=0.2208, simple_loss=0.2782, pruned_loss=0.08169, over 4838.00 frames. ], tot_loss[loss=0.1662, simple_loss=0.2396, pruned_loss=0.04636, over 951177.79 frames. ], batch size: 49, lr: 2.89e-03, grad_scale: 32.0 2023-04-28 02:22:12,821 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=156261.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:22:24,230 INFO [optim.py:369] (6/7) 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 02:22:25,633 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=156277.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:22:26,637 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-04-28 02:22:34,034 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=156290.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:22:38,825 INFO [finetune.py:976] (6/7) Epoch 28, batch 1650, loss[loss=0.1395, simple_loss=0.2214, pruned_loss=0.02878, over 4827.00 frames. ], tot_loss[loss=0.1638, simple_loss=0.2368, pruned_loss=0.0454, over 953455.12 frames. ], batch size: 41, lr: 2.89e-03, grad_scale: 32.0 2023-04-28 02:22:45,575 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=156309.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:22:48,053 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=156313.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:23:00,993 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0325, 1.7473, 2.2220, 2.3351, 2.0830, 1.9388, 2.1127, 2.1359], device='cuda:6'), covar=tensor([0.4996, 0.6885, 0.6900, 0.5909, 0.5689, 0.9114, 0.8287, 0.9483], device='cuda:6'), in_proj_covar=tensor([0.0443, 0.0424, 0.0520, 0.0507, 0.0471, 0.0509, 0.0510, 0.0525], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-28 02:23:12,279 INFO [finetune.py:976] (6/7) Epoch 28, batch 1700, loss[loss=0.1818, simple_loss=0.2512, pruned_loss=0.05617, over 4916.00 frames. ], tot_loss[loss=0.1623, simple_loss=0.2346, pruned_loss=0.04497, over 952040.65 frames. ], batch size: 37, lr: 2.88e-03, grad_scale: 32.0 2023-04-28 02:23:14,237 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=156351.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:23:20,267 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=156361.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:23:21,603 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.1695, 2.9156, 2.3222, 2.4667, 1.6042, 1.6270, 2.5732, 1.6696], device='cuda:6'), covar=tensor([0.1798, 0.1414, 0.1346, 0.1625, 0.2318, 0.1923, 0.0865, 0.2117], device='cuda:6'), in_proj_covar=tensor([0.0198, 0.0209, 0.0169, 0.0204, 0.0200, 0.0185, 0.0156, 0.0188], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:6') 2023-04-28 02:23:28,666 INFO [optim.py:369] (6/7) 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 02:23:45,144 INFO [finetune.py:976] (6/7) Epoch 28, batch 1750, loss[loss=0.1644, simple_loss=0.2458, pruned_loss=0.04147, over 4759.00 frames. ], tot_loss[loss=0.1641, simple_loss=0.2368, pruned_loss=0.04568, over 952655.02 frames. ], batch size: 59, lr: 2.88e-03, grad_scale: 32.0 2023-04-28 02:24:08,540 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=156426.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 02:24:29,097 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=156440.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 02:24:39,111 INFO [finetune.py:976] (6/7) Epoch 28, batch 1800, loss[loss=0.1697, simple_loss=0.2509, pruned_loss=0.04418, over 4932.00 frames. ], tot_loss[loss=0.1674, simple_loss=0.2412, pruned_loss=0.04678, over 953221.38 frames. ], batch size: 38, lr: 2.88e-03, grad_scale: 32.0 2023-04-28 02:25:11,111 INFO [optim.py:369] (6/7) 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 02:25:31,042 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=156488.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:25:42,455 INFO [finetune.py:976] (6/7) Epoch 28, batch 1850, loss[loss=0.2053, simple_loss=0.2664, pruned_loss=0.07206, over 4882.00 frames. ], tot_loss[loss=0.1695, simple_loss=0.2434, pruned_loss=0.04779, over 953175.42 frames. ], batch size: 35, lr: 2.88e-03, grad_scale: 32.0 2023-04-28 02:26:48,597 INFO [finetune.py:976] (6/7) Epoch 28, batch 1900, loss[loss=0.1327, simple_loss=0.2105, pruned_loss=0.02746, over 4774.00 frames. ], tot_loss[loss=0.1691, simple_loss=0.2434, pruned_loss=0.0474, over 954682.22 frames. ], batch size: 28, lr: 2.88e-03, grad_scale: 32.0 2023-04-28 02:27:20,595 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=156572.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:27:22,345 INFO [optim.py:369] (6/7) 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 02:27:30,987 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.3042, 1.7201, 2.1148, 2.3586, 2.1038, 1.7044, 1.2970, 1.8643], device='cuda:6'), covar=tensor([0.2997, 0.2980, 0.1626, 0.2282, 0.2526, 0.2583, 0.4139, 0.1977], device='cuda:6'), in_proj_covar=tensor([0.0291, 0.0245, 0.0228, 0.0314, 0.0221, 0.0234, 0.0228, 0.0185], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:6') 2023-04-28 02:27:54,788 INFO [finetune.py:976] (6/7) Epoch 28, batch 1950, loss[loss=0.2183, simple_loss=0.277, pruned_loss=0.07986, over 4819.00 frames. ], tot_loss[loss=0.1694, simple_loss=0.2436, pruned_loss=0.04765, over 954150.99 frames. ], batch size: 39, lr: 2.88e-03, grad_scale: 32.0 2023-04-28 02:28:06,415 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5041, 1.0485, 0.4586, 1.2561, 1.0879, 1.4141, 1.3304, 1.3223], device='cuda:6'), covar=tensor([0.0434, 0.0385, 0.0390, 0.0510, 0.0297, 0.0445, 0.0461, 0.0513], device='cuda:6'), in_proj_covar=tensor([0.0028, 0.0023, 0.0022, 0.0028, 0.0019, 0.0028, 0.0027, 0.0028], device='cuda:6'), out_proj_covar=tensor([0.0052, 0.0046, 0.0039, 0.0053, 0.0039, 0.0051, 0.0051, 0.0053], device='cuda:6') 2023-04-28 02:29:00,506 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=156646.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:29:01,644 INFO [finetune.py:976] (6/7) Epoch 28, batch 2000, loss[loss=0.1489, simple_loss=0.2024, pruned_loss=0.04773, over 4762.00 frames. ], tot_loss[loss=0.1677, simple_loss=0.2408, pruned_loss=0.04725, over 955472.96 frames. ], batch size: 27, lr: 2.88e-03, grad_scale: 64.0 2023-04-28 02:29:34,744 INFO [optim.py:369] (6/7) 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 02:29:42,014 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8465, 2.5644, 1.7240, 1.9127, 1.3363, 1.3796, 1.6898, 1.3249], device='cuda:6'), covar=tensor([0.1977, 0.1289, 0.1692, 0.1740, 0.2562, 0.2206, 0.1186, 0.2242], device='cuda:6'), in_proj_covar=tensor([0.0197, 0.0208, 0.0169, 0.0204, 0.0199, 0.0185, 0.0155, 0.0187], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:6') 2023-04-28 02:30:06,050 INFO [finetune.py:976] (6/7) Epoch 28, batch 2050, loss[loss=0.1687, simple_loss=0.2419, pruned_loss=0.04775, over 4914.00 frames. ], tot_loss[loss=0.1647, simple_loss=0.2371, pruned_loss=0.04611, over 955171.18 frames. ], batch size: 43, lr: 2.88e-03, grad_scale: 64.0 2023-04-28 02:30:14,191 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7036, 2.5453, 2.1062, 2.3042, 2.5436, 2.0792, 3.3396, 2.0636], device='cuda:6'), covar=tensor([0.3529, 0.2388, 0.4085, 0.3118, 0.1924, 0.2734, 0.1965, 0.3995], device='cuda:6'), in_proj_covar=tensor([0.0342, 0.0357, 0.0426, 0.0353, 0.0384, 0.0379, 0.0372, 0.0425], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-28 02:30:43,829 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=156726.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 02:31:05,849 INFO [finetune.py:976] (6/7) Epoch 28, batch 2100, loss[loss=0.1672, simple_loss=0.2571, pruned_loss=0.03861, over 4860.00 frames. ], tot_loss[loss=0.1645, simple_loss=0.2364, pruned_loss=0.04636, over 953434.42 frames. ], batch size: 47, lr: 2.88e-03, grad_scale: 64.0 2023-04-28 02:31:07,078 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=156749.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:31:39,406 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=156774.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:31:39,958 INFO [optim.py:369] (6/7) 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 02:31:47,217 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.4637, 1.2343, 0.4456, 1.1591, 1.3151, 1.3210, 1.2405, 1.2379], device='cuda:6'), covar=tensor([0.0614, 0.0388, 0.0420, 0.0619, 0.0300, 0.0651, 0.0648, 0.0650], device='cuda:6'), in_proj_covar=tensor([0.0028, 0.0023, 0.0022, 0.0028, 0.0019, 0.0028, 0.0027, 0.0028], device='cuda:6'), out_proj_covar=tensor([0.0052, 0.0046, 0.0039, 0.0053, 0.0039, 0.0051, 0.0051, 0.0053], device='cuda:6') 2023-04-28 02:32:10,052 INFO [finetune.py:976] (6/7) Epoch 28, batch 2150, loss[loss=0.2002, simple_loss=0.2695, pruned_loss=0.06544, over 4277.00 frames. ], tot_loss[loss=0.1679, simple_loss=0.2407, pruned_loss=0.04754, over 952179.30 frames. ], batch size: 65, lr: 2.88e-03, grad_scale: 32.0 2023-04-28 02:32:17,440 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-04-28 02:32:27,895 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=156810.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:32:27,902 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5956, 1.7787, 1.4550, 1.2560, 1.2192, 1.2031, 1.4773, 1.1302], device='cuda:6'), covar=tensor([0.1575, 0.1148, 0.1431, 0.1523, 0.2204, 0.1899, 0.1021, 0.2033], device='cuda:6'), in_proj_covar=tensor([0.0198, 0.0209, 0.0169, 0.0204, 0.0200, 0.0186, 0.0156, 0.0187], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:6') 2023-04-28 02:32:52,174 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-04-28 02:32:57,253 INFO [finetune.py:976] (6/7) Epoch 28, batch 2200, loss[loss=0.1932, simple_loss=0.263, pruned_loss=0.06169, over 4869.00 frames. ], tot_loss[loss=0.1684, simple_loss=0.2416, pruned_loss=0.04763, over 951533.41 frames. ], batch size: 31, lr: 2.88e-03, grad_scale: 32.0 2023-04-28 02:33:15,064 INFO [zipformer.py:1188] (6/7) 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] (6/7) 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 02:33:23,527 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=156885.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:33:25,395 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7473, 1.2541, 1.7544, 2.2023, 1.8172, 1.6799, 1.7297, 1.7141], device='cuda:6'), covar=tensor([0.4561, 0.6931, 0.6668, 0.5531, 0.5880, 0.8030, 0.8174, 0.9084], device='cuda:6'), in_proj_covar=tensor([0.0445, 0.0424, 0.0521, 0.0508, 0.0474, 0.0511, 0.0511, 0.0528], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-28 02:33:31,446 INFO [finetune.py:976] (6/7) Epoch 28, batch 2250, loss[loss=0.1472, simple_loss=0.219, pruned_loss=0.0377, over 4790.00 frames. ], tot_loss[loss=0.1694, simple_loss=0.2428, pruned_loss=0.04804, over 950674.17 frames. ], batch size: 25, lr: 2.88e-03, grad_scale: 32.0 2023-04-28 02:33:39,563 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8864, 2.5402, 1.0827, 1.3051, 1.8273, 1.1302, 2.9487, 1.6045], device='cuda:6'), covar=tensor([0.0695, 0.0507, 0.0710, 0.1106, 0.0452, 0.0946, 0.0207, 0.0587], device='cuda:6'), in_proj_covar=tensor([0.0050, 0.0064, 0.0047, 0.0046, 0.0049, 0.0051, 0.0072, 0.0050], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:6') 2023-04-28 02:33:47,981 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=156920.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:33:53,034 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.50 vs. limit=5.0 2023-04-28 02:34:04,151 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=156946.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:34:04,177 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=156946.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:34:05,296 INFO [finetune.py:976] (6/7) Epoch 28, batch 2300, loss[loss=0.1583, simple_loss=0.2375, pruned_loss=0.03957, over 4771.00 frames. ], tot_loss[loss=0.1679, simple_loss=0.2419, pruned_loss=0.047, over 951938.77 frames. ], batch size: 28, lr: 2.88e-03, grad_scale: 32.0 2023-04-28 02:34:25,204 INFO [optim.py:369] (6/7) 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 02:34:36,692 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=156994.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:34:39,084 INFO [finetune.py:976] (6/7) Epoch 28, batch 2350, loss[loss=0.1363, simple_loss=0.221, pruned_loss=0.02579, over 4819.00 frames. ], tot_loss[loss=0.1672, simple_loss=0.2409, pruned_loss=0.04675, over 955156.63 frames. ], batch size: 40, lr: 2.88e-03, grad_scale: 32.0 2023-04-28 02:34:48,501 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=157010.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:34:55,506 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7061, 1.6534, 1.8081, 2.0905, 2.0564, 1.6481, 1.4078, 1.8902], device='cuda:6'), covar=tensor([0.0815, 0.1119, 0.0776, 0.0517, 0.0581, 0.0818, 0.0746, 0.0541], device='cuda:6'), in_proj_covar=tensor([0.0186, 0.0204, 0.0186, 0.0172, 0.0180, 0.0179, 0.0152, 0.0179], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-28 02:35:13,521 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0014, 2.3768, 2.0883, 2.3702, 1.9196, 2.0357, 2.0153, 1.5487], device='cuda:6'), covar=tensor([0.1576, 0.1322, 0.0840, 0.1066, 0.2959, 0.1311, 0.1767, 0.2476], device='cuda:6'), in_proj_covar=tensor([0.0284, 0.0303, 0.0218, 0.0277, 0.0317, 0.0255, 0.0249, 0.0264], device='cuda:6'), out_proj_covar=tensor([1.1338e-04, 1.1936e-04, 8.5716e-05, 1.0917e-04, 1.2773e-04, 1.0037e-04, 1.0023e-04, 1.0421e-04], device='cuda:6') 2023-04-28 02:35:23,596 INFO [finetune.py:976] (6/7) Epoch 28, batch 2400, loss[loss=0.1877, simple_loss=0.2433, pruned_loss=0.06602, over 4083.00 frames. ], tot_loss[loss=0.1657, simple_loss=0.2384, pruned_loss=0.04646, over 954654.30 frames. ], batch size: 18, lr: 2.88e-03, grad_scale: 32.0 2023-04-28 02:35:50,231 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=157071.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:35:53,119 INFO [optim.py:369] (6/7) 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 02:36:07,062 INFO [finetune.py:976] (6/7) Epoch 28, batch 2450, loss[loss=0.1636, simple_loss=0.2367, pruned_loss=0.04519, over 4820.00 frames. ], tot_loss[loss=0.1628, simple_loss=0.2353, pruned_loss=0.04511, over 955026.15 frames. ], batch size: 40, lr: 2.88e-03, grad_scale: 32.0 2023-04-28 02:36:11,409 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=157105.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:36:58,724 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=157142.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:37:07,639 INFO [finetune.py:976] (6/7) Epoch 28, batch 2500, loss[loss=0.1994, simple_loss=0.2878, pruned_loss=0.05547, over 4858.00 frames. ], tot_loss[loss=0.1645, simple_loss=0.2374, pruned_loss=0.04578, over 956634.13 frames. ], batch size: 44, lr: 2.88e-03, grad_scale: 32.0 2023-04-28 02:37:44,331 INFO [optim.py:369] (6/7) 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] (6/7) Epoch 28, batch 2550, loss[loss=0.1582, simple_loss=0.24, pruned_loss=0.0382, over 4815.00 frames. ], tot_loss[loss=0.1671, simple_loss=0.2409, pruned_loss=0.04661, over 954459.28 frames. ], batch size: 41, lr: 2.88e-03, grad_scale: 32.0 2023-04-28 02:38:22,305 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.2259, 1.8682, 2.1275, 2.6867, 2.1961, 1.8056, 1.8014, 2.0202], device='cuda:6'), covar=tensor([0.2640, 0.2693, 0.1544, 0.1899, 0.2429, 0.2239, 0.3631, 0.1961], device='cuda:6'), in_proj_covar=tensor([0.0292, 0.0245, 0.0228, 0.0313, 0.0221, 0.0235, 0.0228, 0.0186], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:6') 2023-04-28 02:38:22,915 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=157203.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:39:11,686 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=157241.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:39:16,446 INFO [finetune.py:976] (6/7) Epoch 28, batch 2600, loss[loss=0.158, simple_loss=0.2375, pruned_loss=0.0393, over 4906.00 frames. ], tot_loss[loss=0.1679, simple_loss=0.2417, pruned_loss=0.04711, over 952550.23 frames. ], batch size: 37, lr: 2.88e-03, grad_scale: 32.0 2023-04-28 02:39:50,655 INFO [optim.py:369] (6/7) 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 02:40:05,061 INFO [finetune.py:976] (6/7) Epoch 28, batch 2650, loss[loss=0.1314, simple_loss=0.2118, pruned_loss=0.02549, over 4749.00 frames. ], tot_loss[loss=0.1689, simple_loss=0.2425, pruned_loss=0.04767, over 952421.32 frames. ], batch size: 27, lr: 2.88e-03, grad_scale: 32.0 2023-04-28 02:40:06,353 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-04-28 02:40:33,013 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.6922, 2.5807, 2.6698, 3.1316, 3.0479, 2.3578, 2.2213, 2.7276], device='cuda:6'), covar=tensor([0.0753, 0.0875, 0.0620, 0.0553, 0.0526, 0.0865, 0.0714, 0.0524], device='cuda:6'), in_proj_covar=tensor([0.0185, 0.0202, 0.0185, 0.0170, 0.0179, 0.0178, 0.0151, 0.0177], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-28 02:40:53,194 INFO [finetune.py:976] (6/7) Epoch 28, batch 2700, loss[loss=0.1453, simple_loss=0.2146, pruned_loss=0.03805, over 4905.00 frames. ], tot_loss[loss=0.167, simple_loss=0.2407, pruned_loss=0.04659, over 954914.62 frames. ], batch size: 43, lr: 2.88e-03, grad_scale: 32.0 2023-04-28 02:40:58,717 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=157356.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:41:04,717 INFO [zipformer.py:1188] (6/7) 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] (6/7) 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 02:41:15,930 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=157382.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:41:26,096 INFO [finetune.py:976] (6/7) Epoch 28, batch 2750, loss[loss=0.1601, simple_loss=0.2317, pruned_loss=0.04422, over 4764.00 frames. ], tot_loss[loss=0.1646, simple_loss=0.2376, pruned_loss=0.04576, over 954085.76 frames. ], batch size: 26, lr: 2.88e-03, grad_scale: 32.0 2023-04-28 02:41:30,008 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=157404.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 02:41:31,103 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=157405.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:41:38,355 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=157417.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:42:02,272 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 28, batch 2800, loss[loss=0.1479, simple_loss=0.2225, pruned_loss=0.03669, over 4765.00 frames. ], tot_loss[loss=0.1612, simple_loss=0.2339, pruned_loss=0.04427, over 953344.91 frames. ], batch size: 26, lr: 2.88e-03, grad_scale: 32.0 2023-04-28 02:42:13,772 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=157453.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:42:23,525 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.2477, 1.5498, 1.4592, 1.7891, 1.6293, 1.7537, 1.4381, 3.0750], device='cuda:6'), covar=tensor([0.0627, 0.0807, 0.0808, 0.1188, 0.0650, 0.0486, 0.0718, 0.0166], device='cuda:6'), in_proj_covar=tensor([0.0037, 0.0038, 0.0039, 0.0042, 0.0040, 0.0037, 0.0038, 0.0054], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0013, 0.0013, 0.0015, 0.0014, 0.0013, 0.0013, 0.0016], device='cuda:6') 2023-04-28 02:42:26,579 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=157465.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 02:42:38,118 INFO [optim.py:369] (6/7) 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] (6/7) Epoch 28, batch 2850, loss[loss=0.1635, simple_loss=0.2268, pruned_loss=0.05006, over 4927.00 frames. ], tot_loss[loss=0.1599, simple_loss=0.2327, pruned_loss=0.04353, over 955127.51 frames. ], batch size: 33, lr: 2.88e-03, grad_scale: 32.0 2023-04-28 02:43:07,362 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=157498.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:43:07,482 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=5.06 vs. limit=5.0 2023-04-28 02:43:17,198 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=157506.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 02:44:00,025 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=157541.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:44:09,941 INFO [finetune.py:976] (6/7) Epoch 28, batch 2900, loss[loss=0.1997, simple_loss=0.2701, pruned_loss=0.06462, over 4895.00 frames. ], tot_loss[loss=0.1635, simple_loss=0.2369, pruned_loss=0.04505, over 954661.23 frames. ], batch size: 35, lr: 2.88e-03, grad_scale: 32.0 2023-04-28 02:44:33,726 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=157567.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 02:44:34,325 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=157568.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:44:43,278 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7216, 2.3551, 1.8200, 1.7479, 1.3066, 1.3579, 1.9061, 1.2596], device='cuda:6'), covar=tensor([0.1768, 0.1388, 0.1542, 0.1686, 0.2389, 0.2161, 0.1000, 0.2191], device='cuda:6'), in_proj_covar=tensor([0.0199, 0.0209, 0.0170, 0.0204, 0.0201, 0.0187, 0.0157, 0.0188], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:6') 2023-04-28 02:44:44,345 INFO [optim.py:369] (6/7) 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 02:45:03,635 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 28, batch 2950, loss[loss=0.1787, simple_loss=0.2442, pruned_loss=0.05665, over 4131.00 frames. ], tot_loss[loss=0.1664, simple_loss=0.2401, pruned_loss=0.04633, over 952681.84 frames. ], batch size: 65, lr: 2.88e-03, grad_scale: 32.0 2023-04-28 02:45:50,102 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=157629.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:46:09,062 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6145, 1.6000, 0.9507, 1.3364, 1.7689, 1.4679, 1.3716, 1.4142], device='cuda:6'), covar=tensor([0.0456, 0.0344, 0.0340, 0.0506, 0.0290, 0.0469, 0.0432, 0.0521], device='cuda:6'), in_proj_covar=tensor([0.0028, 0.0023, 0.0022, 0.0028, 0.0019, 0.0027, 0.0027, 0.0028], device='cuda:6'), out_proj_covar=tensor([0.0052, 0.0046, 0.0039, 0.0053, 0.0039, 0.0051, 0.0051, 0.0053], device='cuda:6') 2023-04-28 02:46:18,335 INFO [finetune.py:976] (6/7) Epoch 28, batch 3000, loss[loss=0.2021, simple_loss=0.2687, pruned_loss=0.06772, over 4882.00 frames. ], tot_loss[loss=0.1671, simple_loss=0.2412, pruned_loss=0.04655, over 954387.50 frames. ], batch size: 32, lr: 2.88e-03, grad_scale: 32.0 2023-04-28 02:46:18,335 INFO [finetune.py:1001] (6/7) Computing validation loss 2023-04-28 02:46:32,841 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8875, 1.1183, 1.7726, 2.3343, 1.9702, 1.7943, 1.7809, 1.7227], device='cuda:6'), covar=tensor([0.4334, 0.7217, 0.6410, 0.5375, 0.5962, 0.7839, 0.8471, 0.9101], device='cuda:6'), in_proj_covar=tensor([0.0447, 0.0425, 0.0522, 0.0509, 0.0475, 0.0512, 0.0513, 0.0527], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-28 02:46:34,884 INFO [finetune.py:1010] (6/7) 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] (6/7) Maximum memory allocated so far is 6435MB 2023-04-28 02:46:43,629 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6061, 1.2087, 1.3450, 1.9327, 2.0292, 1.4553, 1.2444, 1.6930], device='cuda:6'), covar=tensor([0.0916, 0.1876, 0.1343, 0.0619, 0.0672, 0.1058, 0.0928, 0.0681], device='cuda:6'), in_proj_covar=tensor([0.0185, 0.0202, 0.0185, 0.0170, 0.0178, 0.0178, 0.0151, 0.0177], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-28 02:46:49,103 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=157666.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:46:55,061 INFO [optim.py:369] (6/7) 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 02:47:09,007 INFO [finetune.py:976] (6/7) Epoch 28, batch 3050, loss[loss=0.1414, simple_loss=0.2127, pruned_loss=0.03508, over 4886.00 frames. ], tot_loss[loss=0.1675, simple_loss=0.2422, pruned_loss=0.0464, over 956134.12 frames. ], batch size: 32, lr: 2.88e-03, grad_scale: 32.0 2023-04-28 02:47:27,241 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=157712.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:47:28,449 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=157714.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:47:58,456 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=157738.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:48:09,327 INFO [finetune.py:976] (6/7) Epoch 28, batch 3100, loss[loss=0.1517, simple_loss=0.2233, pruned_loss=0.04003, over 4760.00 frames. ], tot_loss[loss=0.1656, simple_loss=0.2398, pruned_loss=0.04568, over 955203.80 frames. ], batch size: 27, lr: 2.88e-03, grad_scale: 32.0 2023-04-28 02:48:22,593 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=157760.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 02:48:32,273 INFO [optim.py:369] (6/7) 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 02:48:45,236 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-04-28 02:48:46,290 INFO [finetune.py:976] (6/7) Epoch 28, batch 3150, loss[loss=0.1798, simple_loss=0.2398, pruned_loss=0.05987, over 4749.00 frames. ], tot_loss[loss=0.1642, simple_loss=0.2375, pruned_loss=0.04546, over 954239.54 frames. ], batch size: 27, lr: 2.88e-03, grad_scale: 32.0 2023-04-28 02:48:46,382 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=157798.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:49:04,252 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0730, 1.2567, 5.1950, 4.9047, 4.5468, 5.0736, 4.6153, 4.6039], device='cuda:6'), covar=tensor([0.6785, 0.6287, 0.0951, 0.1676, 0.0990, 0.1470, 0.1320, 0.1601], device='cuda:6'), in_proj_covar=tensor([0.0313, 0.0312, 0.0410, 0.0410, 0.0353, 0.0420, 0.0321, 0.0368], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-28 02:49:05,860 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2023-04-28 02:49:18,158 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=157846.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:49:19,358 INFO [finetune.py:976] (6/7) Epoch 28, batch 3200, loss[loss=0.1243, simple_loss=0.2004, pruned_loss=0.02406, over 4924.00 frames. ], tot_loss[loss=0.1616, simple_loss=0.2343, pruned_loss=0.04448, over 955596.98 frames. ], batch size: 38, lr: 2.88e-03, grad_scale: 32.0 2023-04-28 02:49:22,545 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=157853.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:49:27,985 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=157862.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 02:49:38,914 INFO [optim.py:369] (6/7) 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 02:50:07,747 INFO [finetune.py:976] (6/7) Epoch 28, batch 3250, loss[loss=0.1399, simple_loss=0.2148, pruned_loss=0.03246, over 4755.00 frames. ], tot_loss[loss=0.1622, simple_loss=0.2352, pruned_loss=0.04462, over 955671.65 frames. ], batch size: 27, lr: 2.87e-03, grad_scale: 32.0 2023-04-28 02:50:29,707 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=157914.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:50:43,200 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=157924.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:51:14,743 INFO [finetune.py:976] (6/7) Epoch 28, batch 3300, loss[loss=0.1302, simple_loss=0.2116, pruned_loss=0.0244, over 4755.00 frames. ], tot_loss[loss=0.1656, simple_loss=0.2395, pruned_loss=0.04586, over 956622.82 frames. ], batch size: 27, lr: 2.87e-03, grad_scale: 32.0 2023-04-28 02:51:28,476 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7618, 2.2049, 1.7713, 1.7364, 1.2243, 1.3353, 1.9594, 1.1691], device='cuda:6'), covar=tensor([0.1738, 0.1463, 0.1423, 0.1707, 0.2342, 0.1970, 0.0909, 0.2136], device='cuda:6'), in_proj_covar=tensor([0.0199, 0.0210, 0.0170, 0.0205, 0.0201, 0.0187, 0.0157, 0.0189], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:6') 2023-04-28 02:51:51,135 INFO [optim.py:369] (6/7) 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 02:52:20,397 INFO [finetune.py:976] (6/7) Epoch 28, batch 3350, loss[loss=0.159, simple_loss=0.2435, pruned_loss=0.03721, over 4798.00 frames. ], tot_loss[loss=0.1669, simple_loss=0.2408, pruned_loss=0.04647, over 954993.45 frames. ], batch size: 51, lr: 2.87e-03, grad_scale: 32.0 2023-04-28 02:52:33,265 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.2184, 1.4920, 1.7096, 1.8257, 1.7521, 1.8270, 1.7135, 1.7673], device='cuda:6'), covar=tensor([0.3534, 0.4878, 0.3889, 0.3535, 0.4846, 0.6138, 0.4695, 0.4086], device='cuda:6'), in_proj_covar=tensor([0.0341, 0.0375, 0.0329, 0.0341, 0.0350, 0.0391, 0.0360, 0.0334], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-28 02:52:41,459 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=158012.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:53:04,785 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5100, 1.3605, 1.8486, 1.8201, 1.3283, 1.2523, 1.4663, 0.9611], device='cuda:6'), covar=tensor([0.0499, 0.0728, 0.0376, 0.0607, 0.0742, 0.1141, 0.0557, 0.0556], device='cuda:6'), in_proj_covar=tensor([0.0070, 0.0067, 0.0065, 0.0068, 0.0075, 0.0094, 0.0072, 0.0062], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-28 02:53:12,139 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=158038.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:53:23,378 INFO [finetune.py:976] (6/7) Epoch 28, batch 3400, loss[loss=0.2022, simple_loss=0.2794, pruned_loss=0.06252, over 4714.00 frames. ], tot_loss[loss=0.1678, simple_loss=0.2419, pruned_loss=0.04691, over 955999.29 frames. ], batch size: 59, lr: 2.87e-03, grad_scale: 32.0 2023-04-28 02:53:41,204 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=158060.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:53:41,252 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=158060.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 02:53:56,200 INFO [optim.py:369] (6/7) 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 02:54:10,639 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 28, batch 3450, loss[loss=0.1585, simple_loss=0.2333, pruned_loss=0.04189, over 4905.00 frames. ], tot_loss[loss=0.168, simple_loss=0.2424, pruned_loss=0.04675, over 955964.80 frames. ], batch size: 46, lr: 2.87e-03, grad_scale: 32.0 2023-04-28 02:54:34,533 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=158108.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 02:54:34,596 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.5402, 1.8779, 1.7605, 2.3156, 2.5199, 2.0072, 1.9972, 1.8606], device='cuda:6'), covar=tensor([0.1620, 0.1743, 0.1684, 0.1637, 0.1014, 0.1796, 0.1752, 0.2205], device='cuda:6'), in_proj_covar=tensor([0.0316, 0.0307, 0.0350, 0.0286, 0.0324, 0.0306, 0.0300, 0.0375], device='cuda:6'), out_proj_covar=tensor([6.4377e-05, 6.2954e-05, 7.3346e-05, 5.7326e-05, 6.6120e-05, 6.3759e-05, 6.1895e-05, 7.9368e-05], device='cuda:6') 2023-04-28 02:54:54,872 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=158122.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:55:27,636 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-04-28 02:55:28,043 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.7064, 4.6779, 1.2638, 2.9524, 3.2134, 3.4480, 3.0615, 1.5733], device='cuda:6'), covar=tensor([0.1020, 0.1010, 0.1901, 0.0975, 0.0802, 0.0932, 0.1153, 0.1792], device='cuda:6'), in_proj_covar=tensor([0.0118, 0.0240, 0.0136, 0.0121, 0.0132, 0.0153, 0.0117, 0.0119], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-28 02:55:29,271 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([4.1404, 4.0502, 3.2353, 4.7856, 4.0699, 4.1819, 2.3278, 4.1804], device='cuda:6'), covar=tensor([0.1477, 0.1080, 0.2767, 0.1514, 0.3050, 0.1838, 0.5030, 0.2103], device='cuda:6'), in_proj_covar=tensor([0.0245, 0.0219, 0.0251, 0.0303, 0.0298, 0.0247, 0.0274, 0.0272], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-28 02:55:29,825 INFO [finetune.py:976] (6/7) Epoch 28, batch 3500, loss[loss=0.1541, simple_loss=0.215, pruned_loss=0.04664, over 3952.00 frames. ], tot_loss[loss=0.1664, simple_loss=0.24, pruned_loss=0.04645, over 955955.10 frames. ], batch size: 17, lr: 2.87e-03, grad_scale: 32.0 2023-04-28 02:55:46,119 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.3184, 2.9256, 1.1126, 1.6420, 2.3515, 1.2928, 3.7636, 2.0149], device='cuda:6'), covar=tensor([0.0643, 0.0724, 0.0884, 0.1144, 0.0473, 0.0986, 0.0193, 0.0547], device='cuda:6'), in_proj_covar=tensor([0.0050, 0.0064, 0.0047, 0.0046, 0.0049, 0.0051, 0.0073, 0.0050], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:6') 2023-04-28 02:55:48,565 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=158162.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 02:56:07,656 INFO [optim.py:369] (6/7) 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 02:56:11,954 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.4337, 1.9070, 1.5571, 1.8758, 1.4163, 1.5286, 1.5376, 1.3675], device='cuda:6'), covar=tensor([0.2316, 0.1800, 0.1256, 0.1449, 0.3552, 0.1493, 0.2048, 0.2518], device='cuda:6'), in_proj_covar=tensor([0.0282, 0.0302, 0.0217, 0.0276, 0.0314, 0.0253, 0.0247, 0.0263], device='cuda:6'), out_proj_covar=tensor([1.1228e-04, 1.1886e-04, 8.5457e-05, 1.0862e-04, 1.2672e-04, 9.9523e-05, 9.9579e-05, 1.0350e-04], device='cuda:6') 2023-04-28 02:56:13,082 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=158183.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:56:23,135 INFO [finetune.py:976] (6/7) Epoch 28, batch 3550, loss[loss=0.1573, simple_loss=0.2267, pruned_loss=0.04394, over 4820.00 frames. ], tot_loss[loss=0.1645, simple_loss=0.2369, pruned_loss=0.04603, over 956379.26 frames. ], batch size: 41, lr: 2.87e-03, grad_scale: 32.0 2023-04-28 02:56:27,787 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.7559, 2.0429, 1.9159, 2.6268, 2.7654, 2.3174, 2.1893, 2.0164], device='cuda:6'), covar=tensor([0.1864, 0.2052, 0.2168, 0.1841, 0.1068, 0.1773, 0.2156, 0.2239], device='cuda:6'), in_proj_covar=tensor([0.0315, 0.0307, 0.0350, 0.0286, 0.0324, 0.0306, 0.0300, 0.0375], device='cuda:6'), out_proj_covar=tensor([6.4214e-05, 6.2984e-05, 7.3369e-05, 5.7160e-05, 6.6151e-05, 6.3634e-05, 6.2003e-05, 7.9269e-05], device='cuda:6') 2023-04-28 02:56:30,197 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=158209.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:56:30,776 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=158210.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 02:56:38,624 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.4588, 3.0682, 2.5999, 2.9230, 2.1598, 2.6466, 2.6531, 2.0682], device='cuda:6'), covar=tensor([0.1808, 0.1181, 0.0753, 0.1003, 0.3138, 0.0978, 0.1916, 0.2546], device='cuda:6'), in_proj_covar=tensor([0.0282, 0.0302, 0.0217, 0.0276, 0.0314, 0.0253, 0.0247, 0.0263], device='cuda:6'), out_proj_covar=tensor([1.1222e-04, 1.1892e-04, 8.5473e-05, 1.0869e-04, 1.2667e-04, 9.9594e-05, 9.9622e-05, 1.0344e-04], device='cuda:6') 2023-04-28 02:56:39,837 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=158224.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:56:42,235 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.1924, 1.4782, 1.3513, 1.7311, 1.5791, 1.8938, 1.3483, 3.4987], device='cuda:6'), covar=tensor([0.0722, 0.1104, 0.1064, 0.1429, 0.0832, 0.0646, 0.0998, 0.0210], device='cuda:6'), in_proj_covar=tensor([0.0038, 0.0039, 0.0039, 0.0043, 0.0040, 0.0038, 0.0038, 0.0054], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0013, 0.0013, 0.0015, 0.0014, 0.0013, 0.0013, 0.0016], device='cuda:6') 2023-04-28 02:56:56,634 INFO [finetune.py:976] (6/7) Epoch 28, batch 3600, loss[loss=0.1765, simple_loss=0.2468, pruned_loss=0.05307, over 4876.00 frames. ], tot_loss[loss=0.1629, simple_loss=0.2352, pruned_loss=0.04534, over 956553.32 frames. ], batch size: 34, lr: 2.87e-03, grad_scale: 32.0 2023-04-28 02:57:11,763 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=158272.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:57:14,134 INFO [optim.py:369] (6/7) 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 02:57:29,419 INFO [finetune.py:976] (6/7) Epoch 28, batch 3650, loss[loss=0.2379, simple_loss=0.3038, pruned_loss=0.086, over 4808.00 frames. ], tot_loss[loss=0.1661, simple_loss=0.2384, pruned_loss=0.04689, over 954324.08 frames. ], batch size: 45, lr: 2.87e-03, grad_scale: 32.0 2023-04-28 02:57:52,632 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.5394, 3.1191, 1.1333, 1.7733, 2.7705, 1.7378, 4.5261, 2.4815], device='cuda:6'), covar=tensor([0.0593, 0.0778, 0.0783, 0.1247, 0.0442, 0.0922, 0.0250, 0.0498], device='cuda:6'), in_proj_covar=tensor([0.0050, 0.0064, 0.0047, 0.0046, 0.0049, 0.0051, 0.0073, 0.0050], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:6') 2023-04-28 02:58:02,903 INFO [finetune.py:976] (6/7) Epoch 28, batch 3700, loss[loss=0.2024, simple_loss=0.269, pruned_loss=0.06791, over 4815.00 frames. ], tot_loss[loss=0.1683, simple_loss=0.2414, pruned_loss=0.04756, over 954672.35 frames. ], batch size: 38, lr: 2.87e-03, grad_scale: 32.0 2023-04-28 02:58:19,865 INFO [optim.py:369] (6/7) 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 02:58:25,408 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=158384.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:58:31,758 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.2480, 2.9625, 1.0721, 1.5286, 2.2649, 1.3245, 3.9933, 1.9376], device='cuda:6'), covar=tensor([0.0651, 0.0688, 0.0823, 0.1219, 0.0467, 0.1020, 0.0185, 0.0565], device='cuda:6'), in_proj_covar=tensor([0.0050, 0.0064, 0.0047, 0.0046, 0.0049, 0.0051, 0.0073, 0.0050], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:6') 2023-04-28 02:58:35,209 INFO [finetune.py:976] (6/7) Epoch 28, batch 3750, loss[loss=0.2048, simple_loss=0.2786, pruned_loss=0.0655, over 4808.00 frames. ], tot_loss[loss=0.1695, simple_loss=0.2432, pruned_loss=0.04791, over 955300.68 frames. ], batch size: 51, lr: 2.87e-03, grad_scale: 32.0 2023-04-28 02:59:05,538 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=158445.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:59:07,671 INFO [finetune.py:976] (6/7) Epoch 28, batch 3800, loss[loss=0.1067, simple_loss=0.1727, pruned_loss=0.0203, over 4112.00 frames. ], tot_loss[loss=0.1707, simple_loss=0.2444, pruned_loss=0.04844, over 951742.25 frames. ], batch size: 17, lr: 2.87e-03, grad_scale: 32.0 2023-04-28 02:59:42,094 INFO [optim.py:369] (6/7) 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 02:59:43,424 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=158478.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:00:07,897 INFO [finetune.py:976] (6/7) Epoch 28, batch 3850, loss[loss=0.1519, simple_loss=0.2245, pruned_loss=0.03965, over 4742.00 frames. ], tot_loss[loss=0.1695, simple_loss=0.2431, pruned_loss=0.04795, over 952558.33 frames. ], batch size: 54, lr: 2.87e-03, grad_scale: 32.0 2023-04-28 03:00:25,639 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0087, 2.4635, 1.0985, 1.4257, 1.8288, 1.2352, 3.0803, 1.7650], device='cuda:6'), covar=tensor([0.0654, 0.0467, 0.0681, 0.1138, 0.0482, 0.1013, 0.0221, 0.0564], device='cuda:6'), in_proj_covar=tensor([0.0050, 0.0064, 0.0047, 0.0046, 0.0049, 0.0051, 0.0073, 0.0050], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:6') 2023-04-28 03:00:25,741 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.00 vs. limit=5.0 2023-04-28 03:00:26,826 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=158509.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:01:10,982 INFO [finetune.py:976] (6/7) Epoch 28, batch 3900, loss[loss=0.1228, simple_loss=0.1924, pruned_loss=0.02664, over 4750.00 frames. ], tot_loss[loss=0.1672, simple_loss=0.2402, pruned_loss=0.04716, over 953845.66 frames. ], batch size: 54, lr: 2.87e-03, grad_scale: 32.0 2023-04-28 03:01:28,078 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=158557.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:01:51,822 INFO [optim.py:369] (6/7) 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 03:02:23,134 INFO [finetune.py:976] (6/7) Epoch 28, batch 3950, loss[loss=0.1932, simple_loss=0.265, pruned_loss=0.06075, over 4756.00 frames. ], tot_loss[loss=0.1651, simple_loss=0.2373, pruned_loss=0.04644, over 952844.84 frames. ], batch size: 27, lr: 2.87e-03, grad_scale: 32.0 2023-04-28 03:02:36,849 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.5720, 1.9033, 2.0493, 2.1185, 2.0310, 2.0851, 2.1158, 2.0755], device='cuda:6'), covar=tensor([0.3645, 0.5067, 0.4228, 0.4454, 0.5316, 0.6435, 0.4705, 0.4364], device='cuda:6'), in_proj_covar=tensor([0.0342, 0.0376, 0.0330, 0.0342, 0.0350, 0.0392, 0.0361, 0.0334], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-28 03:02:48,082 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0208, 2.6365, 1.0784, 1.5270, 1.9123, 1.2418, 3.4328, 1.8680], device='cuda:6'), covar=tensor([0.0683, 0.0615, 0.0742, 0.1077, 0.0482, 0.0971, 0.0238, 0.0526], device='cuda:6'), in_proj_covar=tensor([0.0050, 0.0064, 0.0047, 0.0046, 0.0049, 0.0051, 0.0073, 0.0050], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:6') 2023-04-28 03:02:58,288 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5540, 1.3215, 4.3927, 4.1017, 3.8308, 4.2053, 4.1642, 3.8909], device='cuda:6'), covar=tensor([0.7172, 0.6128, 0.1177, 0.1954, 0.1224, 0.1514, 0.1161, 0.1532], device='cuda:6'), in_proj_covar=tensor([0.0310, 0.0310, 0.0406, 0.0407, 0.0350, 0.0416, 0.0318, 0.0363], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-28 03:03:31,309 INFO [finetune.py:976] (6/7) Epoch 28, batch 4000, loss[loss=0.1763, simple_loss=0.2454, pruned_loss=0.05366, over 4896.00 frames. ], tot_loss[loss=0.165, simple_loss=0.2368, pruned_loss=0.04659, over 952538.07 frames. ], batch size: 32, lr: 2.87e-03, grad_scale: 32.0 2023-04-28 03:04:13,588 INFO [optim.py:369] (6/7) 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 03:04:28,416 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=3.12 vs. limit=5.0 2023-04-28 03:04:38,151 INFO [finetune.py:976] (6/7) Epoch 28, batch 4050, loss[loss=0.1818, simple_loss=0.2545, pruned_loss=0.05455, over 4818.00 frames. ], tot_loss[loss=0.1687, simple_loss=0.2413, pruned_loss=0.04807, over 955157.06 frames. ], batch size: 33, lr: 2.87e-03, grad_scale: 32.0 2023-04-28 03:05:33,917 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=158740.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:05:43,858 INFO [finetune.py:976] (6/7) Epoch 28, batch 4100, loss[loss=0.1486, simple_loss=0.2318, pruned_loss=0.03273, over 4823.00 frames. ], tot_loss[loss=0.1697, simple_loss=0.2431, pruned_loss=0.04812, over 955525.41 frames. ], batch size: 33, lr: 2.87e-03, grad_scale: 32.0 2023-04-28 03:06:05,229 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([3.6856, 3.7594, 2.7122, 4.2938, 3.7728, 3.7099, 1.9110, 3.6854], device='cuda:6'), covar=tensor([0.1586, 0.1141, 0.3317, 0.1529, 0.2429, 0.1764, 0.5095, 0.2404], device='cuda:6'), in_proj_covar=tensor([0.0247, 0.0221, 0.0253, 0.0305, 0.0302, 0.0251, 0.0278, 0.0276], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-28 03:06:25,665 INFO [optim.py:369] (6/7) 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 03:06:26,428 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=158777.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:06:26,779 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.79 vs. limit=2.0 2023-04-28 03:06:27,018 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=158778.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:06:35,236 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=158783.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:06:37,078 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0203, 1.4690, 1.6085, 1.6510, 2.0906, 1.8028, 1.4403, 1.5344], device='cuda:6'), covar=tensor([0.1379, 0.1487, 0.1835, 0.1266, 0.0819, 0.1231, 0.1663, 0.2085], device='cuda:6'), in_proj_covar=tensor([0.0315, 0.0306, 0.0349, 0.0285, 0.0324, 0.0304, 0.0299, 0.0374], device='cuda:6'), out_proj_covar=tensor([6.4284e-05, 6.2707e-05, 7.3116e-05, 5.6954e-05, 6.6088e-05, 6.3236e-05, 6.1769e-05, 7.9164e-05], device='cuda:6') 2023-04-28 03:06:49,757 INFO [finetune.py:976] (6/7) Epoch 28, batch 4150, loss[loss=0.1817, simple_loss=0.263, pruned_loss=0.05021, over 4828.00 frames. ], tot_loss[loss=0.169, simple_loss=0.2429, pruned_loss=0.04759, over 954922.90 frames. ], batch size: 30, lr: 2.87e-03, grad_scale: 64.0 2023-04-28 03:07:30,856 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=158826.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:07:41,987 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=158835.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 03:07:43,877 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=158838.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:07:52,637 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=158844.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:07:54,975 INFO [finetune.py:976] (6/7) Epoch 28, batch 4200, loss[loss=0.1381, simple_loss=0.2113, pruned_loss=0.03245, over 4760.00 frames. ], tot_loss[loss=0.168, simple_loss=0.2423, pruned_loss=0.04686, over 955077.43 frames. ], batch size: 23, lr: 2.87e-03, grad_scale: 32.0 2023-04-28 03:08:11,159 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([3.6924, 3.7052, 2.8259, 4.2915, 3.6333, 3.6835, 1.7987, 3.6245], device='cuda:6'), covar=tensor([0.1803, 0.1183, 0.3527, 0.1410, 0.2754, 0.1820, 0.5518, 0.2422], device='cuda:6'), in_proj_covar=tensor([0.0246, 0.0220, 0.0252, 0.0304, 0.0301, 0.0251, 0.0277, 0.0274], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-28 03:08:37,326 INFO [optim.py:369] (6/7) 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 03:08:46,717 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-04-28 03:09:04,973 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9800, 2.4734, 2.1189, 2.3836, 1.8218, 2.2256, 2.0627, 1.6204], device='cuda:6'), covar=tensor([0.1932, 0.1250, 0.0840, 0.1208, 0.3106, 0.1032, 0.1935, 0.2657], device='cuda:6'), in_proj_covar=tensor([0.0281, 0.0300, 0.0217, 0.0274, 0.0311, 0.0251, 0.0246, 0.0261], device='cuda:6'), out_proj_covar=tensor([1.1181e-04, 1.1816e-04, 8.5263e-05, 1.0768e-04, 1.2557e-04, 9.8745e-05, 9.8892e-05, 1.0293e-04], device='cuda:6') 2023-04-28 03:09:04,985 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=158896.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 03:09:06,085 INFO [finetune.py:976] (6/7) Epoch 28, batch 4250, loss[loss=0.1344, simple_loss=0.2048, pruned_loss=0.03198, over 4680.00 frames. ], tot_loss[loss=0.1662, simple_loss=0.2401, pruned_loss=0.04611, over 955227.00 frames. ], batch size: 23, lr: 2.87e-03, grad_scale: 32.0 2023-04-28 03:10:12,872 INFO [finetune.py:976] (6/7) Epoch 28, batch 4300, loss[loss=0.169, simple_loss=0.2311, pruned_loss=0.05351, over 4827.00 frames. ], tot_loss[loss=0.164, simple_loss=0.2374, pruned_loss=0.04529, over 955621.51 frames. ], batch size: 30, lr: 2.87e-03, grad_scale: 32.0 2023-04-28 03:10:48,763 INFO [optim.py:369] (6/7) 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 03:11:18,967 INFO [finetune.py:976] (6/7) Epoch 28, batch 4350, loss[loss=0.1832, simple_loss=0.2422, pruned_loss=0.06207, over 4813.00 frames. ], tot_loss[loss=0.1615, simple_loss=0.2339, pruned_loss=0.04458, over 954120.93 frames. ], batch size: 45, lr: 2.87e-03, grad_scale: 32.0 2023-04-28 03:12:15,674 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 28, batch 4400, loss[loss=0.1897, simple_loss=0.2545, pruned_loss=0.06249, over 4287.00 frames. ], tot_loss[loss=0.1622, simple_loss=0.2347, pruned_loss=0.04484, over 954759.37 frames. ], batch size: 65, lr: 2.87e-03, grad_scale: 32.0 2023-04-28 03:12:27,298 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-04-28 03:12:54,453 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=159069.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:13:05,081 INFO [optim.py:369] (6/7) 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 03:13:19,042 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 28, batch 4450, loss[loss=0.1254, simple_loss=0.2079, pruned_loss=0.02144, over 4747.00 frames. ], tot_loss[loss=0.1634, simple_loss=0.2367, pruned_loss=0.04501, over 953273.56 frames. ], batch size: 28, lr: 2.87e-03, grad_scale: 32.0 2023-04-28 03:13:29,883 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.7854, 2.1640, 2.0419, 2.2338, 1.9906, 2.0224, 2.1448, 2.0520], device='cuda:6'), covar=tensor([0.4351, 0.5883, 0.5249, 0.4332, 0.5769, 0.6810, 0.6023, 0.5110], device='cuda:6'), in_proj_covar=tensor([0.0343, 0.0377, 0.0330, 0.0342, 0.0352, 0.0393, 0.0363, 0.0334], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-28 03:14:09,972 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=159130.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:14:17,182 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=159133.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:14:17,270 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7898, 1.4190, 1.8652, 2.2884, 1.9230, 1.8063, 1.8672, 1.7835], device='cuda:6'), covar=tensor([0.4517, 0.6738, 0.6499, 0.5596, 0.5889, 0.7672, 0.8207, 0.9663], device='cuda:6'), in_proj_covar=tensor([0.0447, 0.0425, 0.0523, 0.0511, 0.0476, 0.0512, 0.0514, 0.0528], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-28 03:14:21,841 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=159139.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:14:38,434 INFO [finetune.py:976] (6/7) Epoch 28, batch 4500, loss[loss=0.1665, simple_loss=0.2406, pruned_loss=0.04618, over 4810.00 frames. ], tot_loss[loss=0.1655, simple_loss=0.239, pruned_loss=0.04596, over 952022.55 frames. ], batch size: 33, lr: 2.87e-03, grad_scale: 32.0 2023-04-28 03:14:49,788 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=159159.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:15:11,104 INFO [optim.py:369] (6/7) 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 03:15:31,026 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=159191.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 03:15:41,720 INFO [finetune.py:976] (6/7) Epoch 28, batch 4550, loss[loss=0.1758, simple_loss=0.2445, pruned_loss=0.05351, over 4915.00 frames. ], tot_loss[loss=0.1673, simple_loss=0.241, pruned_loss=0.04677, over 951482.56 frames. ], batch size: 33, lr: 2.87e-03, grad_scale: 32.0 2023-04-28 03:15:51,544 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2023-04-28 03:16:05,592 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=159220.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:16:45,689 INFO [finetune.py:976] (6/7) Epoch 28, batch 4600, loss[loss=0.143, simple_loss=0.2243, pruned_loss=0.03084, over 4815.00 frames. ], tot_loss[loss=0.1665, simple_loss=0.2403, pruned_loss=0.04636, over 951341.76 frames. ], batch size: 41, lr: 2.87e-03, grad_scale: 32.0 2023-04-28 03:17:18,840 INFO [optim.py:369] (6/7) 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 03:17:29,705 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8941, 1.3707, 1.9165, 2.3668, 2.0178, 1.8596, 1.9489, 1.8295], device='cuda:6'), covar=tensor([0.4108, 0.6778, 0.6187, 0.4930, 0.5273, 0.7379, 0.7580, 0.8643], device='cuda:6'), in_proj_covar=tensor([0.0446, 0.0425, 0.0523, 0.0510, 0.0475, 0.0512, 0.0513, 0.0528], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-28 03:17:38,888 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=159290.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:17:49,022 INFO [finetune.py:976] (6/7) Epoch 28, batch 4650, loss[loss=0.1235, simple_loss=0.1983, pruned_loss=0.02439, over 4899.00 frames. ], tot_loss[loss=0.1646, simple_loss=0.2379, pruned_loss=0.04563, over 951320.59 frames. ], batch size: 43, lr: 2.87e-03, grad_scale: 32.0 2023-04-28 03:18:30,188 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5723, 1.3481, 0.7588, 1.2865, 1.3744, 1.4895, 1.3531, 1.3848], device='cuda:6'), covar=tensor([0.0476, 0.0383, 0.0338, 0.0538, 0.0290, 0.0487, 0.0479, 0.0566], device='cuda:6'), in_proj_covar=tensor([0.0028, 0.0023, 0.0021, 0.0028, 0.0019, 0.0027, 0.0027, 0.0028], device='cuda:6'), out_proj_covar=tensor([0.0052, 0.0046, 0.0039, 0.0053, 0.0039, 0.0051, 0.0051, 0.0053], device='cuda:6') 2023-04-28 03:18:54,507 INFO [finetune.py:976] (6/7) Epoch 28, batch 4700, loss[loss=0.1391, simple_loss=0.2154, pruned_loss=0.03137, over 4799.00 frames. ], tot_loss[loss=0.163, simple_loss=0.2361, pruned_loss=0.045, over 953496.96 frames. ], batch size: 26, lr: 2.87e-03, grad_scale: 32.0 2023-04-28 03:19:03,221 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=159351.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:19:36,246 INFO [optim.py:369] (6/7) 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 03:20:00,614 INFO [finetune.py:976] (6/7) Epoch 28, batch 4750, loss[loss=0.1879, simple_loss=0.2602, pruned_loss=0.05785, over 4851.00 frames. ], tot_loss[loss=0.1633, simple_loss=0.2358, pruned_loss=0.04545, over 953511.87 frames. ], batch size: 44, lr: 2.87e-03, grad_scale: 32.0 2023-04-28 03:20:33,976 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.2746, 1.5880, 1.7543, 1.8863, 1.7748, 1.7907, 1.7864, 1.7876], device='cuda:6'), covar=tensor([0.3678, 0.5197, 0.4133, 0.3973, 0.5295, 0.6367, 0.4842, 0.4360], device='cuda:6'), in_proj_covar=tensor([0.0344, 0.0378, 0.0331, 0.0344, 0.0353, 0.0394, 0.0364, 0.0335], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-28 03:20:41,301 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=159425.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:20:51,584 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=159433.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:20:55,233 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=159439.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:21:06,690 INFO [finetune.py:976] (6/7) Epoch 28, batch 4800, loss[loss=0.1493, simple_loss=0.2135, pruned_loss=0.04262, over 4526.00 frames. ], tot_loss[loss=0.1643, simple_loss=0.2371, pruned_loss=0.04571, over 953471.92 frames. ], batch size: 20, lr: 2.86e-03, grad_scale: 32.0 2023-04-28 03:21:48,625 INFO [optim.py:369] (6/7) 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 03:21:51,196 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=159481.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:21:59,478 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=159487.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:22:07,516 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=159491.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 03:22:12,298 INFO [finetune.py:976] (6/7) Epoch 28, batch 4850, loss[loss=0.1483, simple_loss=0.2279, pruned_loss=0.0344, over 4762.00 frames. ], tot_loss[loss=0.1668, simple_loss=0.2408, pruned_loss=0.04644, over 954985.08 frames. ], batch size: 26, lr: 2.86e-03, grad_scale: 32.0 2023-04-28 03:22:40,602 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=159515.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:22:40,627 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5701, 1.7050, 1.6281, 1.9517, 1.9635, 2.1467, 1.7021, 4.1590], device='cuda:6'), covar=tensor([0.0512, 0.0804, 0.0739, 0.1162, 0.0599, 0.0528, 0.0698, 0.0114], device='cuda:6'), in_proj_covar=tensor([0.0037, 0.0039, 0.0039, 0.0043, 0.0040, 0.0038, 0.0038, 0.0055], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0013, 0.0014, 0.0015, 0.0014, 0.0013, 0.0013, 0.0016], device='cuda:6') 2023-04-28 03:23:05,855 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=159539.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 03:23:16,523 INFO [finetune.py:976] (6/7) Epoch 28, batch 4900, loss[loss=0.1641, simple_loss=0.2446, pruned_loss=0.04182, over 4886.00 frames. ], tot_loss[loss=0.1683, simple_loss=0.2426, pruned_loss=0.04701, over 955647.80 frames. ], batch size: 35, lr: 2.86e-03, grad_scale: 32.0 2023-04-28 03:23:23,410 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.67 vs. limit=2.0 2023-04-28 03:23:56,944 INFO [optim.py:369] (6/7) 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 03:24:20,244 INFO [finetune.py:976] (6/7) Epoch 28, batch 4950, loss[loss=0.1211, simple_loss=0.1988, pruned_loss=0.02171, over 4801.00 frames. ], tot_loss[loss=0.168, simple_loss=0.2425, pruned_loss=0.04676, over 954415.80 frames. ], batch size: 25, lr: 2.86e-03, grad_scale: 32.0 2023-04-28 03:25:08,567 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=159631.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:25:23,012 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=159646.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:25:24,159 INFO [finetune.py:976] (6/7) Epoch 28, batch 5000, loss[loss=0.2002, simple_loss=0.2572, pruned_loss=0.07158, over 4728.00 frames. ], tot_loss[loss=0.1663, simple_loss=0.2402, pruned_loss=0.04621, over 952648.16 frames. ], batch size: 23, lr: 2.86e-03, grad_scale: 32.0 2023-04-28 03:26:05,240 INFO [optim.py:369] (6/7) 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 03:26:24,442 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=159692.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:26:32,514 INFO [finetune.py:976] (6/7) Epoch 28, batch 5050, loss[loss=0.1611, simple_loss=0.2243, pruned_loss=0.04901, over 4241.00 frames. ], tot_loss[loss=0.1645, simple_loss=0.238, pruned_loss=0.0455, over 954747.89 frames. ], batch size: 65, lr: 2.86e-03, grad_scale: 32.0 2023-04-28 03:26:45,062 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.2362, 3.1948, 2.6675, 2.9693, 3.3614, 3.0499, 4.0966, 2.5584], device='cuda:6'), covar=tensor([0.3146, 0.2246, 0.4092, 0.2717, 0.1452, 0.2329, 0.0949, 0.3540], device='cuda:6'), in_proj_covar=tensor([0.0337, 0.0353, 0.0421, 0.0350, 0.0381, 0.0373, 0.0370, 0.0422], device='cuda:6'), out_proj_covar=tensor([9.9569e-05, 1.0499e-04, 1.2733e-04, 1.0469e-04, 1.1267e-04, 1.1074e-04, 1.0820e-04, 1.2672e-04], device='cuda:6') 2023-04-28 03:26:45,085 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([3.0794, 2.7577, 3.0624, 3.6434, 2.9079, 2.5405, 2.6511, 3.0130], device='cuda:6'), covar=tensor([0.3083, 0.2430, 0.1317, 0.1912, 0.2373, 0.2332, 0.2990, 0.1643], device='cuda:6'), in_proj_covar=tensor([0.0290, 0.0244, 0.0227, 0.0312, 0.0222, 0.0234, 0.0227, 0.0185], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:6') 2023-04-28 03:27:06,801 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=159725.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:27:35,596 INFO [finetune.py:976] (6/7) Epoch 28, batch 5100, loss[loss=0.2182, simple_loss=0.2812, pruned_loss=0.07763, over 4906.00 frames. ], tot_loss[loss=0.1618, simple_loss=0.2346, pruned_loss=0.04449, over 953548.93 frames. ], batch size: 37, lr: 2.86e-03, grad_scale: 32.0 2023-04-28 03:28:08,033 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-04-28 03:28:09,670 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=159773.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:28:13,014 INFO [optim.py:369] (6/7) 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 03:28:34,297 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([0.8745, 1.1986, 3.3022, 3.0695, 2.9891, 3.1940, 3.2189, 2.9745], device='cuda:6'), covar=tensor([0.7628, 0.5649, 0.1728, 0.2475, 0.1493, 0.2351, 0.1698, 0.1842], device='cuda:6'), in_proj_covar=tensor([0.0311, 0.0311, 0.0407, 0.0410, 0.0352, 0.0417, 0.0320, 0.0365], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-28 03:28:41,782 INFO [finetune.py:976] (6/7) Epoch 28, batch 5150, loss[loss=0.1842, simple_loss=0.2662, pruned_loss=0.05116, over 4813.00 frames. ], tot_loss[loss=0.1637, simple_loss=0.236, pruned_loss=0.04565, over 953901.04 frames. ], batch size: 38, lr: 2.86e-03, grad_scale: 32.0 2023-04-28 03:28:54,090 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.4007, 1.6680, 1.8326, 1.9394, 1.7586, 1.8188, 1.8568, 1.8522], device='cuda:6'), covar=tensor([0.4052, 0.5211, 0.4201, 0.3822, 0.5351, 0.6760, 0.4872, 0.4388], device='cuda:6'), in_proj_covar=tensor([0.0343, 0.0377, 0.0331, 0.0344, 0.0353, 0.0394, 0.0364, 0.0336], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-28 03:29:03,555 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=159815.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:29:13,508 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.9267, 2.8892, 2.9188, 3.5019, 3.2838, 3.0468, 2.6263, 3.3577], device='cuda:6'), covar=tensor([0.0752, 0.0749, 0.0552, 0.0467, 0.0486, 0.0672, 0.0550, 0.0400], device='cuda:6'), in_proj_covar=tensor([0.0184, 0.0204, 0.0183, 0.0171, 0.0179, 0.0177, 0.0151, 0.0176], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-28 03:29:21,708 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.4989, 1.7571, 1.8994, 2.0081, 1.8150, 1.8431, 1.9802, 1.8936], device='cuda:6'), covar=tensor([0.3968, 0.5431, 0.4238, 0.4061, 0.5531, 0.7075, 0.4912, 0.4850], device='cuda:6'), in_proj_covar=tensor([0.0343, 0.0377, 0.0331, 0.0344, 0.0353, 0.0394, 0.0364, 0.0336], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-28 03:29:45,764 INFO [finetune.py:976] (6/7) Epoch 28, batch 5200, loss[loss=0.1787, simple_loss=0.2547, pruned_loss=0.05134, over 4914.00 frames. ], tot_loss[loss=0.1664, simple_loss=0.2393, pruned_loss=0.04679, over 955201.73 frames. ], batch size: 36, lr: 2.86e-03, grad_scale: 32.0 2023-04-28 03:29:55,494 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=159855.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:30:05,392 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=159863.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:30:26,722 INFO [optim.py:369] (6/7) 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] (6/7) Epoch 28, batch 5250, loss[loss=0.1883, simple_loss=0.2654, pruned_loss=0.05559, over 4915.00 frames. ], tot_loss[loss=0.1674, simple_loss=0.2409, pruned_loss=0.04693, over 955789.52 frames. ], batch size: 42, lr: 2.86e-03, grad_scale: 32.0 2023-04-28 03:31:00,362 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.96 vs. limit=2.0 2023-04-28 03:31:02,419 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0688, 2.5919, 1.0453, 1.4185, 2.1794, 1.2607, 3.3762, 1.6569], device='cuda:6'), covar=tensor([0.0648, 0.0795, 0.0863, 0.1180, 0.0457, 0.0969, 0.0189, 0.0594], device='cuda:6'), in_proj_covar=tensor([0.0051, 0.0065, 0.0048, 0.0046, 0.0049, 0.0051, 0.0073, 0.0051], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:6') 2023-04-28 03:31:09,938 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=159910.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:31:19,001 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=159916.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:31:55,451 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=159946.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:32:02,064 INFO [finetune.py:976] (6/7) Epoch 28, batch 5300, loss[loss=0.1866, simple_loss=0.2609, pruned_loss=0.05612, over 4856.00 frames. ], tot_loss[loss=0.169, simple_loss=0.2428, pruned_loss=0.04759, over 957098.36 frames. ], batch size: 44, lr: 2.86e-03, grad_scale: 32.0 2023-04-28 03:32:26,770 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=159971.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:32:35,448 INFO [optim.py:369] (6/7) 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 03:32:48,739 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=159987.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:32:58,790 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=159994.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:33:06,691 INFO [finetune.py:976] (6/7) Epoch 28, batch 5350, loss[loss=0.1527, simple_loss=0.225, pruned_loss=0.04017, over 4815.00 frames. ], tot_loss[loss=0.167, simple_loss=0.2415, pruned_loss=0.04625, over 956026.63 frames. ], batch size: 39, lr: 2.86e-03, grad_scale: 32.0 2023-04-28 03:33:17,043 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.4197, 1.6442, 1.8728, 1.9978, 1.8914, 1.9086, 1.8978, 1.9259], device='cuda:6'), covar=tensor([0.3671, 0.4990, 0.4068, 0.4069, 0.5136, 0.6669, 0.4623, 0.4253], device='cuda:6'), in_proj_covar=tensor([0.0343, 0.0377, 0.0331, 0.0344, 0.0353, 0.0394, 0.0364, 0.0335], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-28 03:33:32,228 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.4895, 1.3165, 4.0938, 3.8575, 3.6004, 3.9023, 3.8865, 3.5949], device='cuda:6'), covar=tensor([0.7058, 0.5665, 0.1206, 0.1693, 0.1149, 0.1754, 0.1565, 0.1695], device='cuda:6'), in_proj_covar=tensor([0.0309, 0.0310, 0.0405, 0.0408, 0.0351, 0.0416, 0.0319, 0.0365], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-28 03:34:13,400 INFO [finetune.py:976] (6/7) Epoch 28, batch 5400, loss[loss=0.1517, simple_loss=0.2301, pruned_loss=0.03672, over 4755.00 frames. ], tot_loss[loss=0.1654, simple_loss=0.2395, pruned_loss=0.04569, over 956793.92 frames. ], batch size: 27, lr: 2.86e-03, grad_scale: 32.0 2023-04-28 03:34:46,552 INFO [optim.py:369] (6/7) 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 03:35:03,332 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9008, 1.3941, 1.9347, 2.3519, 2.0109, 1.8411, 1.9313, 1.8913], device='cuda:6'), covar=tensor([0.4224, 0.6393, 0.6120, 0.5539, 0.5546, 0.7767, 0.7633, 0.8329], device='cuda:6'), in_proj_covar=tensor([0.0445, 0.0425, 0.0520, 0.0509, 0.0474, 0.0511, 0.0513, 0.0527], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-28 03:35:18,040 INFO [finetune.py:976] (6/7) Epoch 28, batch 5450, loss[loss=0.1493, simple_loss=0.2205, pruned_loss=0.03907, over 4732.00 frames. ], tot_loss[loss=0.1638, simple_loss=0.2373, pruned_loss=0.04512, over 955210.27 frames. ], batch size: 54, lr: 2.86e-03, grad_scale: 32.0 2023-04-28 03:35:27,507 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.1224, 1.8080, 2.0934, 2.5665, 2.5854, 2.1001, 1.9487, 2.4240], device='cuda:6'), covar=tensor([0.0810, 0.1227, 0.0723, 0.0534, 0.0536, 0.0753, 0.0619, 0.0440], device='cuda:6'), in_proj_covar=tensor([0.0184, 0.0203, 0.0183, 0.0170, 0.0177, 0.0177, 0.0150, 0.0175], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-28 03:35:27,515 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0844, 2.6198, 2.1361, 2.4916, 1.7812, 2.1532, 2.2389, 1.7040], device='cuda:6'), covar=tensor([0.2053, 0.1237, 0.0860, 0.1097, 0.3338, 0.1058, 0.2004, 0.2780], device='cuda:6'), in_proj_covar=tensor([0.0282, 0.0299, 0.0216, 0.0275, 0.0312, 0.0252, 0.0247, 0.0263], device='cuda:6'), out_proj_covar=tensor([1.1218e-04, 1.1773e-04, 8.4813e-05, 1.0793e-04, 1.2571e-04, 9.8819e-05, 9.9400e-05, 1.0336e-04], device='cuda:6') 2023-04-28 03:36:22,773 INFO [finetune.py:976] (6/7) Epoch 28, batch 5500, loss[loss=0.1839, simple_loss=0.2597, pruned_loss=0.05404, over 4914.00 frames. ], tot_loss[loss=0.1612, simple_loss=0.2343, pruned_loss=0.04408, over 956753.05 frames. ], batch size: 36, lr: 2.86e-03, grad_scale: 32.0 2023-04-28 03:37:00,973 INFO [optim.py:369] (6/7) 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 03:37:27,650 INFO [finetune.py:976] (6/7) Epoch 28, batch 5550, loss[loss=0.2222, simple_loss=0.2859, pruned_loss=0.07924, over 4830.00 frames. ], tot_loss[loss=0.164, simple_loss=0.2369, pruned_loss=0.04549, over 956651.70 frames. ], batch size: 49, lr: 2.86e-03, grad_scale: 32.0 2023-04-28 03:37:41,113 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=160211.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:38:03,245 INFO [finetune.py:976] (6/7) Epoch 28, batch 5600, loss[loss=0.1795, simple_loss=0.2493, pruned_loss=0.05483, over 4756.00 frames. ], tot_loss[loss=0.1669, simple_loss=0.2405, pruned_loss=0.04662, over 956250.33 frames. ], batch size: 26, lr: 2.86e-03, grad_scale: 32.0 2023-04-28 03:38:13,712 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=160266.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:38:20,149 INFO [optim.py:369] (6/7) 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 03:38:26,531 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=160287.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:38:32,857 INFO [finetune.py:976] (6/7) Epoch 28, batch 5650, loss[loss=0.2046, simple_loss=0.2701, pruned_loss=0.06953, over 4885.00 frames. ], tot_loss[loss=0.1677, simple_loss=0.2426, pruned_loss=0.04644, over 957152.31 frames. ], batch size: 32, lr: 2.86e-03, grad_scale: 32.0 2023-04-28 03:38:38,246 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=160307.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 03:38:49,406 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([0.7790, 1.6684, 1.8196, 1.3935, 1.8271, 1.5508, 2.2324, 1.5767], device='cuda:6'), covar=tensor([0.3029, 0.1768, 0.4111, 0.2506, 0.1288, 0.2252, 0.1331, 0.4140], device='cuda:6'), in_proj_covar=tensor([0.0339, 0.0355, 0.0422, 0.0353, 0.0384, 0.0377, 0.0372, 0.0426], device='cuda:6'), out_proj_covar=tensor([9.9887e-05, 1.0553e-04, 1.2756e-04, 1.0555e-04, 1.1363e-04, 1.1181e-04, 1.0847e-04, 1.2786e-04], device='cuda:6') 2023-04-28 03:38:54,998 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 28, batch 5700, loss[loss=0.1508, simple_loss=0.2214, pruned_loss=0.04015, over 4130.00 frames. ], tot_loss[loss=0.1644, simple_loss=0.2379, pruned_loss=0.0455, over 938114.17 frames. ], batch size: 18, lr: 2.86e-03, grad_scale: 32.0 2023-04-28 03:39:12,843 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.6995, 2.3406, 2.6193, 3.0362, 2.5863, 2.1800, 1.9884, 2.4093], device='cuda:6'), covar=tensor([0.2897, 0.2575, 0.1456, 0.1953, 0.2343, 0.2370, 0.3434, 0.1768], device='cuda:6'), in_proj_covar=tensor([0.0292, 0.0246, 0.0228, 0.0314, 0.0223, 0.0236, 0.0229, 0.0186], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:6') 2023-04-28 03:39:14,599 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=160368.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 03:39:31,532 INFO [finetune.py:976] (6/7) Epoch 29, batch 0, loss[loss=0.1645, simple_loss=0.2397, pruned_loss=0.04465, over 4833.00 frames. ], tot_loss[loss=0.1645, simple_loss=0.2397, pruned_loss=0.04465, over 4833.00 frames. ], batch size: 47, lr: 2.86e-03, grad_scale: 32.0 2023-04-28 03:39:31,532 INFO [finetune.py:1001] (6/7) Computing validation loss 2023-04-28 03:39:42,779 INFO [finetune.py:1010] (6/7) 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] (6/7) Maximum memory allocated so far is 6435MB 2023-04-28 03:39:44,398 INFO [optim.py:369] (6/7) 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 03:39:45,706 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5526, 3.5934, 0.8720, 2.0170, 1.9903, 2.6135, 2.1061, 1.0806], device='cuda:6'), covar=tensor([0.1349, 0.0925, 0.1965, 0.1090, 0.0975, 0.0901, 0.1365, 0.1918], device='cuda:6'), in_proj_covar=tensor([0.0117, 0.0238, 0.0136, 0.0120, 0.0131, 0.0152, 0.0117, 0.0119], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-28 03:40:16,114 INFO [finetune.py:976] (6/7) Epoch 29, batch 50, loss[loss=0.1865, simple_loss=0.2593, pruned_loss=0.0569, over 4862.00 frames. ], tot_loss[loss=0.1689, simple_loss=0.2442, pruned_loss=0.04679, over 215009.69 frames. ], batch size: 34, lr: 2.85e-03, grad_scale: 32.0 2023-04-28 03:40:21,814 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-04-28 03:40:33,652 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=3.99 vs. limit=5.0 2023-04-28 03:40:50,236 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-04-28 03:41:16,376 INFO [finetune.py:976] (6/7) Epoch 29, batch 100, loss[loss=0.1693, simple_loss=0.2394, pruned_loss=0.04955, over 4713.00 frames. ], tot_loss[loss=0.1653, simple_loss=0.2386, pruned_loss=0.04606, over 379485.56 frames. ], batch size: 54, lr: 2.85e-03, grad_scale: 16.0 2023-04-28 03:41:24,290 INFO [optim.py:369] (6/7) 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 03:42:07,350 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 29, batch 150, loss[loss=0.1668, simple_loss=0.2441, pruned_loss=0.04476, over 4817.00 frames. ], tot_loss[loss=0.1599, simple_loss=0.2319, pruned_loss=0.04393, over 505685.02 frames. ], batch size: 30, lr: 2.85e-03, grad_scale: 16.0 2023-04-28 03:42:43,469 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.3810, 1.8347, 2.2669, 2.6016, 2.1977, 1.7151, 1.5039, 2.0127], device='cuda:6'), covar=tensor([0.3277, 0.3075, 0.1615, 0.2212, 0.2609, 0.2692, 0.3853, 0.1854], device='cuda:6'), in_proj_covar=tensor([0.0293, 0.0247, 0.0228, 0.0315, 0.0223, 0.0236, 0.0229, 0.0186], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:6') 2023-04-28 03:43:05,317 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=160559.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:43:14,855 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 29, batch 200, loss[loss=0.1652, simple_loss=0.2323, pruned_loss=0.0491, over 4869.00 frames. ], tot_loss[loss=0.1576, simple_loss=0.2296, pruned_loss=0.04282, over 607175.26 frames. ], batch size: 31, lr: 2.85e-03, grad_scale: 16.0 2023-04-28 03:43:34,236 INFO [optim.py:369] (6/7) 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 03:43:37,296 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.4551, 3.2523, 0.9816, 1.8357, 1.8444, 2.3030, 1.8394, 1.0226], device='cuda:6'), covar=tensor([0.1418, 0.0893, 0.1881, 0.1186, 0.1047, 0.0990, 0.1590, 0.2013], device='cuda:6'), in_proj_covar=tensor([0.0117, 0.0238, 0.0135, 0.0119, 0.0131, 0.0152, 0.0117, 0.0119], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-28 03:44:17,869 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=160614.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:44:30,033 INFO [finetune.py:976] (6/7) Epoch 29, batch 250, loss[loss=0.1355, simple_loss=0.1965, pruned_loss=0.03719, over 4389.00 frames. ], tot_loss[loss=0.1615, simple_loss=0.2343, pruned_loss=0.04433, over 686491.82 frames. ], batch size: 19, lr: 2.85e-03, grad_scale: 16.0 2023-04-28 03:45:20,093 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=160663.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 03:45:32,747 INFO [finetune.py:976] (6/7) Epoch 29, batch 300, loss[loss=0.131, simple_loss=0.2005, pruned_loss=0.03077, over 4722.00 frames. ], tot_loss[loss=0.1657, simple_loss=0.2391, pruned_loss=0.04612, over 747070.62 frames. ], batch size: 23, lr: 2.85e-03, grad_scale: 16.0 2023-04-28 03:45:38,924 INFO [optim.py:369] (6/7) 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 03:46:01,458 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-04-28 03:46:15,442 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-04-28 03:46:37,392 INFO [finetune.py:976] (6/7) Epoch 29, batch 350, loss[loss=0.174, simple_loss=0.2492, pruned_loss=0.04939, over 4875.00 frames. ], tot_loss[loss=0.1688, simple_loss=0.2428, pruned_loss=0.04744, over 793422.85 frames. ], batch size: 34, lr: 2.85e-03, grad_scale: 16.0 2023-04-28 03:47:13,497 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9708, 1.6167, 1.8523, 2.2822, 2.2750, 1.8661, 1.6729, 2.1460], device='cuda:6'), covar=tensor([0.0753, 0.1193, 0.0769, 0.0552, 0.0591, 0.0825, 0.0659, 0.0476], device='cuda:6'), in_proj_covar=tensor([0.0184, 0.0205, 0.0184, 0.0171, 0.0178, 0.0177, 0.0151, 0.0176], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-28 03:47:35,744 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-04-28 03:47:45,907 INFO [finetune.py:976] (6/7) Epoch 29, batch 400, loss[loss=0.1816, simple_loss=0.2657, pruned_loss=0.04872, over 4812.00 frames. ], tot_loss[loss=0.1707, simple_loss=0.245, pruned_loss=0.04816, over 830139.25 frames. ], batch size: 33, lr: 2.85e-03, grad_scale: 16.0 2023-04-28 03:47:47,780 INFO [optim.py:369] (6/7) 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 03:48:25,746 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5418, 1.0110, 1.2955, 1.2158, 1.6420, 1.3936, 1.1477, 1.2185], device='cuda:6'), covar=tensor([0.1693, 0.1540, 0.1837, 0.1362, 0.0832, 0.1447, 0.1955, 0.2524], device='cuda:6'), in_proj_covar=tensor([0.0312, 0.0305, 0.0346, 0.0284, 0.0322, 0.0302, 0.0297, 0.0372], device='cuda:6'), out_proj_covar=tensor([6.3550e-05, 6.2497e-05, 7.2385e-05, 5.6854e-05, 6.5639e-05, 6.2742e-05, 6.1197e-05, 7.8507e-05], device='cuda:6') 2023-04-28 03:48:36,958 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8339, 2.0898, 1.2756, 1.5398, 2.1089, 1.6877, 1.5734, 1.6864], device='cuda:6'), covar=tensor([0.0448, 0.0331, 0.0260, 0.0509, 0.0226, 0.0464, 0.0473, 0.0521], device='cuda:6'), in_proj_covar=tensor([0.0028, 0.0023, 0.0022, 0.0028, 0.0019, 0.0028, 0.0027, 0.0028], device='cuda:6'), out_proj_covar=tensor([0.0052, 0.0046, 0.0039, 0.0053, 0.0039, 0.0051, 0.0051, 0.0053], device='cuda:6') 2023-04-28 03:48:48,103 INFO [finetune.py:976] (6/7) Epoch 29, batch 450, loss[loss=0.162, simple_loss=0.2366, pruned_loss=0.04368, over 4866.00 frames. ], tot_loss[loss=0.1689, simple_loss=0.2429, pruned_loss=0.04747, over 858656.79 frames. ], batch size: 34, lr: 2.85e-03, grad_scale: 16.0 2023-04-28 03:49:19,874 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([3.8554, 3.8317, 2.9585, 4.4031, 3.8558, 3.8900, 2.0017, 3.7434], device='cuda:6'), covar=tensor([0.1899, 0.1177, 0.3177, 0.1753, 0.4264, 0.1954, 0.5727, 0.2826], device='cuda:6'), in_proj_covar=tensor([0.0250, 0.0223, 0.0254, 0.0308, 0.0303, 0.0254, 0.0280, 0.0278], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-28 03:49:51,962 INFO [finetune.py:976] (6/7) Epoch 29, batch 500, loss[loss=0.1662, simple_loss=0.2274, pruned_loss=0.05247, over 4867.00 frames. ], tot_loss[loss=0.167, simple_loss=0.2401, pruned_loss=0.04691, over 878073.55 frames. ], batch size: 31, lr: 2.85e-03, grad_scale: 16.0 2023-04-28 03:49:53,843 INFO [optim.py:369] (6/7) 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 03:50:02,890 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-04-28 03:50:25,300 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=160901.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:50:56,031 INFO [finetune.py:976] (6/7) Epoch 29, batch 550, loss[loss=0.16, simple_loss=0.225, pruned_loss=0.04748, over 4819.00 frames. ], tot_loss[loss=0.1638, simple_loss=0.2363, pruned_loss=0.04569, over 896011.33 frames. ], batch size: 41, lr: 2.85e-03, grad_scale: 16.0 2023-04-28 03:51:48,580 INFO [zipformer.py:1188] (6/7) 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] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=160963.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 03:51:50,432 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=160965.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:51:51,650 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7226, 1.6989, 2.1385, 2.2051, 1.6150, 1.4489, 1.7162, 1.0630], device='cuda:6'), covar=tensor([0.0590, 0.0613, 0.0394, 0.0550, 0.0692, 0.1044, 0.0681, 0.0604], device='cuda:6'), in_proj_covar=tensor([0.0070, 0.0066, 0.0065, 0.0068, 0.0074, 0.0093, 0.0072, 0.0062], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-28 03:52:01,360 INFO [finetune.py:976] (6/7) Epoch 29, batch 600, loss[loss=0.1305, simple_loss=0.201, pruned_loss=0.03, over 4907.00 frames. ], tot_loss[loss=0.1633, simple_loss=0.2361, pruned_loss=0.04532, over 910259.54 frames. ], batch size: 36, lr: 2.85e-03, grad_scale: 16.0 2023-04-28 03:52:03,144 INFO [optim.py:369] (6/7) 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 03:52:35,726 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=161011.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 03:52:36,356 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9583, 2.3959, 2.0361, 2.3275, 1.7204, 2.0704, 2.0166, 1.6537], device='cuda:6'), covar=tensor([0.1766, 0.1003, 0.0706, 0.1088, 0.2969, 0.1051, 0.1885, 0.2454], device='cuda:6'), in_proj_covar=tensor([0.0282, 0.0299, 0.0216, 0.0275, 0.0313, 0.0253, 0.0248, 0.0263], device='cuda:6'), out_proj_covar=tensor([1.1216e-04, 1.1769e-04, 8.4773e-05, 1.0791e-04, 1.2607e-04, 9.9356e-05, 9.9928e-05, 1.0375e-04], device='cuda:6') 2023-04-28 03:52:38,791 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7482, 1.5019, 1.6571, 1.9893, 1.9906, 1.6539, 1.5058, 1.8909], device='cuda:6'), covar=tensor([0.0658, 0.1067, 0.0662, 0.0500, 0.0496, 0.0711, 0.0633, 0.0468], device='cuda:6'), in_proj_covar=tensor([0.0184, 0.0203, 0.0183, 0.0171, 0.0177, 0.0177, 0.0150, 0.0176], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-28 03:52:44,112 INFO [finetune.py:976] (6/7) Epoch 29, batch 650, loss[loss=0.1775, simple_loss=0.2549, pruned_loss=0.05004, over 4903.00 frames. ], tot_loss[loss=0.1666, simple_loss=0.2399, pruned_loss=0.04661, over 920082.18 frames. ], batch size: 43, lr: 2.85e-03, grad_scale: 16.0 2023-04-28 03:52:44,869 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=161026.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:52:51,936 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-04-28 03:53:17,105 INFO [finetune.py:976] (6/7) Epoch 29, batch 700, loss[loss=0.2251, simple_loss=0.3043, pruned_loss=0.07299, over 4885.00 frames. ], tot_loss[loss=0.1687, simple_loss=0.2432, pruned_loss=0.04712, over 929325.17 frames. ], batch size: 35, lr: 2.85e-03, grad_scale: 16.0 2023-04-28 03:53:18,920 INFO [optim.py:369] (6/7) 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] (6/7) Epoch 29, batch 750, loss[loss=0.1958, simple_loss=0.2724, pruned_loss=0.05962, over 4816.00 frames. ], tot_loss[loss=0.1704, simple_loss=0.2451, pruned_loss=0.04788, over 936149.93 frames. ], batch size: 38, lr: 2.85e-03, grad_scale: 16.0 2023-04-28 03:53:55,705 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.2114, 1.6522, 1.4554, 2.0505, 2.2020, 1.8745, 1.8421, 1.4905], device='cuda:6'), covar=tensor([0.1710, 0.1937, 0.1795, 0.1714, 0.1395, 0.1862, 0.2276, 0.2239], device='cuda:6'), in_proj_covar=tensor([0.0314, 0.0308, 0.0350, 0.0287, 0.0325, 0.0305, 0.0300, 0.0376], device='cuda:6'), out_proj_covar=tensor([6.3982e-05, 6.3123e-05, 7.3327e-05, 5.7366e-05, 6.6170e-05, 6.3397e-05, 6.1804e-05, 7.9502e-05], device='cuda:6') 2023-04-28 03:53:56,332 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6407, 1.6664, 1.5117, 1.0785, 1.2316, 1.2129, 1.5009, 1.1443], device='cuda:6'), covar=tensor([0.1868, 0.1281, 0.1541, 0.1889, 0.2453, 0.2042, 0.1101, 0.2142], device='cuda:6'), in_proj_covar=tensor([0.0198, 0.0210, 0.0170, 0.0204, 0.0202, 0.0186, 0.0156, 0.0187], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:6') 2023-04-28 03:54:24,404 INFO [finetune.py:976] (6/7) Epoch 29, batch 800, loss[loss=0.2286, simple_loss=0.2789, pruned_loss=0.0891, over 4892.00 frames. ], tot_loss[loss=0.1696, simple_loss=0.2443, pruned_loss=0.04748, over 940370.63 frames. ], batch size: 35, lr: 2.85e-03, grad_scale: 16.0 2023-04-28 03:54:25,695 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0009, 1.4055, 4.7161, 4.4529, 4.1086, 4.5692, 4.1914, 4.1418], device='cuda:6'), covar=tensor([0.6667, 0.5687, 0.1006, 0.1637, 0.1067, 0.1487, 0.2360, 0.1413], device='cuda:6'), in_proj_covar=tensor([0.0307, 0.0307, 0.0403, 0.0406, 0.0348, 0.0413, 0.0316, 0.0361], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-28 03:54:26,204 INFO [optim.py:369] (6/7) 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 03:54:52,592 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=161216.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:54:56,247 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7983, 1.0400, 1.7729, 2.1833, 1.8852, 1.7495, 1.7878, 1.7668], device='cuda:6'), covar=tensor([0.4005, 0.6095, 0.5641, 0.5544, 0.5110, 0.6423, 0.6807, 0.7727], device='cuda:6'), in_proj_covar=tensor([0.0448, 0.0427, 0.0523, 0.0513, 0.0477, 0.0513, 0.0516, 0.0531], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-28 03:54:57,908 INFO [finetune.py:976] (6/7) Epoch 29, batch 850, loss[loss=0.1636, simple_loss=0.2483, pruned_loss=0.03945, over 4867.00 frames. ], tot_loss[loss=0.1682, simple_loss=0.2421, pruned_loss=0.04709, over 941997.74 frames. ], batch size: 34, lr: 2.85e-03, grad_scale: 16.0 2023-04-28 03:55:01,669 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=161231.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:55:30,043 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=161257.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:55:49,079 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9758, 2.4925, 2.2632, 1.8710, 1.4126, 1.6063, 2.2800, 1.4443], device='cuda:6'), covar=tensor([0.1577, 0.1287, 0.1099, 0.1582, 0.2125, 0.1759, 0.0778, 0.1890], device='cuda:6'), in_proj_covar=tensor([0.0199, 0.0210, 0.0171, 0.0205, 0.0202, 0.0187, 0.0157, 0.0188], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:6') 2023-04-28 03:55:49,588 INFO [finetune.py:976] (6/7) Epoch 29, batch 900, loss[loss=0.1749, simple_loss=0.2348, pruned_loss=0.05748, over 4903.00 frames. ], tot_loss[loss=0.1658, simple_loss=0.2395, pruned_loss=0.04608, over 944490.60 frames. ], batch size: 35, lr: 2.85e-03, grad_scale: 16.0 2023-04-28 03:55:50,935 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=161277.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:55:51,422 INFO [optim.py:369] (6/7) 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 03:56:00,173 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=161292.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:56:09,231 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.1889, 1.4243, 1.3452, 1.6369, 1.5352, 1.7844, 1.3737, 3.0070], device='cuda:6'), covar=tensor([0.0638, 0.0785, 0.0790, 0.1220, 0.0633, 0.0483, 0.0726, 0.0163], device='cuda:6'), in_proj_covar=tensor([0.0037, 0.0038, 0.0039, 0.0043, 0.0040, 0.0037, 0.0038, 0.0054], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0013, 0.0014, 0.0015, 0.0014, 0.0013, 0.0013, 0.0016], device='cuda:6') 2023-04-28 03:56:21,349 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=161321.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:56:22,008 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8426, 1.8319, 2.2636, 2.3189, 1.6197, 1.5346, 1.9005, 1.1076], device='cuda:6'), covar=tensor([0.0560, 0.0584, 0.0395, 0.0609, 0.0704, 0.0961, 0.0537, 0.0620], device='cuda:6'), in_proj_covar=tensor([0.0070, 0.0067, 0.0065, 0.0069, 0.0075, 0.0094, 0.0072, 0.0062], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-28 03:56:23,752 INFO [finetune.py:976] (6/7) Epoch 29, batch 950, loss[loss=0.1477, simple_loss=0.2184, pruned_loss=0.03853, over 4818.00 frames. ], tot_loss[loss=0.165, simple_loss=0.2385, pruned_loss=0.04577, over 947488.92 frames. ], batch size: 25, lr: 2.85e-03, grad_scale: 16.0 2023-04-28 03:56:28,739 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5516, 3.5771, 0.8012, 1.9403, 1.9568, 2.4601, 2.0740, 1.0050], device='cuda:6'), covar=tensor([0.1426, 0.0936, 0.2083, 0.1148, 0.1052, 0.1046, 0.1379, 0.2132], device='cuda:6'), in_proj_covar=tensor([0.0117, 0.0237, 0.0135, 0.0119, 0.0131, 0.0152, 0.0117, 0.0118], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-28 03:57:07,675 INFO [finetune.py:976] (6/7) Epoch 29, batch 1000, loss[loss=0.1716, simple_loss=0.2467, pruned_loss=0.0482, over 4828.00 frames. ], tot_loss[loss=0.1653, simple_loss=0.2391, pruned_loss=0.04578, over 951249.11 frames. ], batch size: 30, lr: 2.85e-03, grad_scale: 16.0 2023-04-28 03:57:09,495 INFO [optim.py:369] (6/7) 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 03:57:38,277 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.4777, 1.8813, 1.7024, 2.3229, 2.4542, 2.0935, 1.9747, 1.7903], device='cuda:6'), covar=tensor([0.1521, 0.1611, 0.1896, 0.1360, 0.1184, 0.1574, 0.1927, 0.2143], device='cuda:6'), in_proj_covar=tensor([0.0316, 0.0309, 0.0351, 0.0288, 0.0328, 0.0307, 0.0301, 0.0377], device='cuda:6'), out_proj_covar=tensor([6.4496e-05, 6.3351e-05, 7.3511e-05, 5.7688e-05, 6.6889e-05, 6.3702e-05, 6.2145e-05, 7.9699e-05], device='cuda:6') 2023-04-28 03:57:39,575 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=161401.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:58:12,304 INFO [finetune.py:976] (6/7) Epoch 29, batch 1050, loss[loss=0.1645, simple_loss=0.2422, pruned_loss=0.04346, over 4745.00 frames. ], tot_loss[loss=0.1661, simple_loss=0.2404, pruned_loss=0.04596, over 952900.57 frames. ], batch size: 54, lr: 2.85e-03, grad_scale: 16.0 2023-04-28 03:58:24,036 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.1276, 1.9231, 2.5857, 2.7496, 1.8570, 1.6993, 1.9638, 0.9973], device='cuda:6'), covar=tensor([0.0611, 0.0697, 0.0375, 0.0542, 0.0691, 0.1040, 0.0692, 0.0716], device='cuda:6'), in_proj_covar=tensor([0.0071, 0.0067, 0.0065, 0.0069, 0.0075, 0.0094, 0.0073, 0.0062], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-28 03:58:55,488 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=161460.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:58:56,749 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=161462.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:59:17,655 INFO [finetune.py:976] (6/7) Epoch 29, batch 1100, loss[loss=0.1424, simple_loss=0.2177, pruned_loss=0.0336, over 4873.00 frames. ], tot_loss[loss=0.167, simple_loss=0.2414, pruned_loss=0.0463, over 954084.43 frames. ], batch size: 34, lr: 2.85e-03, grad_scale: 16.0 2023-04-28 03:59:20,410 INFO [optim.py:369] (6/7) 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 04:00:20,263 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=161521.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:00:23,054 INFO [finetune.py:976] (6/7) Epoch 29, batch 1150, loss[loss=0.1526, simple_loss=0.2392, pruned_loss=0.03303, over 4831.00 frames. ], tot_loss[loss=0.1676, simple_loss=0.2422, pruned_loss=0.04655, over 952541.50 frames. ], batch size: 30, lr: 2.85e-03, grad_scale: 16.0 2023-04-28 04:00:23,907 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-04-28 04:00:47,507 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-04-28 04:01:04,585 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=161557.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:01:25,596 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=161572.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:01:27,866 INFO [finetune.py:976] (6/7) Epoch 29, batch 1200, loss[loss=0.1356, simple_loss=0.2039, pruned_loss=0.03361, over 4092.00 frames. ], tot_loss[loss=0.1662, simple_loss=0.2403, pruned_loss=0.04608, over 952754.09 frames. ], batch size: 17, lr: 2.85e-03, grad_scale: 16.0 2023-04-28 04:01:29,684 INFO [optim.py:369] (6/7) 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 04:01:47,188 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=161587.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:02:09,296 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=161605.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:02:29,599 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=161621.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:02:33,466 INFO [finetune.py:976] (6/7) Epoch 29, batch 1250, loss[loss=0.1962, simple_loss=0.2598, pruned_loss=0.06634, over 4826.00 frames. ], tot_loss[loss=0.1636, simple_loss=0.2372, pruned_loss=0.04505, over 951472.79 frames. ], batch size: 41, lr: 2.85e-03, grad_scale: 16.0 2023-04-28 04:02:42,557 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.52 vs. limit=2.0 2023-04-28 04:03:14,113 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.21 vs. limit=5.0 2023-04-28 04:03:28,505 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=161669.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:03:38,165 INFO [finetune.py:976] (6/7) Epoch 29, batch 1300, loss[loss=0.1632, simple_loss=0.2403, pruned_loss=0.0431, over 4848.00 frames. ], tot_loss[loss=0.1616, simple_loss=0.2346, pruned_loss=0.04429, over 953097.83 frames. ], batch size: 47, lr: 2.85e-03, grad_scale: 16.0 2023-04-28 04:03:46,266 INFO [optim.py:369] (6/7) 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 04:04:09,210 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5563, 1.4828, 1.8308, 1.9151, 1.4375, 1.3050, 1.5791, 0.9762], device='cuda:6'), covar=tensor([0.0521, 0.0596, 0.0408, 0.0643, 0.0664, 0.1021, 0.0662, 0.0598], device='cuda:6'), in_proj_covar=tensor([0.0071, 0.0067, 0.0065, 0.0069, 0.0075, 0.0094, 0.0072, 0.0062], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-28 04:04:42,648 INFO [finetune.py:976] (6/7) Epoch 29, batch 1350, loss[loss=0.1985, simple_loss=0.2801, pruned_loss=0.05847, over 4923.00 frames. ], tot_loss[loss=0.1621, simple_loss=0.2345, pruned_loss=0.04486, over 952630.68 frames. ], batch size: 38, lr: 2.85e-03, grad_scale: 16.0 2023-04-28 04:04:54,592 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6003, 1.9590, 1.7406, 1.9431, 1.4953, 1.6027, 1.5562, 1.2646], device='cuda:6'), covar=tensor([0.1829, 0.1288, 0.0873, 0.1127, 0.3427, 0.1080, 0.2012, 0.2481], device='cuda:6'), in_proj_covar=tensor([0.0280, 0.0298, 0.0215, 0.0274, 0.0311, 0.0252, 0.0246, 0.0262], device='cuda:6'), out_proj_covar=tensor([1.1164e-04, 1.1718e-04, 8.4384e-05, 1.0784e-04, 1.2517e-04, 9.8873e-05, 9.9056e-05, 1.0340e-04], device='cuda:6') 2023-04-28 04:05:27,901 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=161757.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:05:37,224 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([3.2831, 3.2478, 2.5234, 3.8285, 3.2946, 3.2876, 1.3200, 3.2383], device='cuda:6'), covar=tensor([0.2025, 0.1514, 0.3696, 0.2321, 0.3677, 0.1921, 0.6165, 0.2710], device='cuda:6'), in_proj_covar=tensor([0.0251, 0.0223, 0.0254, 0.0308, 0.0305, 0.0254, 0.0279, 0.0278], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-28 04:05:50,109 INFO [finetune.py:976] (6/7) Epoch 29, batch 1400, loss[loss=0.1922, simple_loss=0.2645, pruned_loss=0.05999, over 4792.00 frames. ], tot_loss[loss=0.1654, simple_loss=0.2386, pruned_loss=0.04613, over 952548.78 frames. ], batch size: 51, lr: 2.85e-03, grad_scale: 16.0 2023-04-28 04:05:58,006 INFO [optim.py:369] (6/7) 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 04:06:51,801 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=161815.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:06:52,365 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=161816.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:07:03,299 INFO [finetune.py:976] (6/7) Epoch 29, batch 1450, loss[loss=0.1591, simple_loss=0.228, pruned_loss=0.04513, over 4164.00 frames. ], tot_loss[loss=0.1667, simple_loss=0.2404, pruned_loss=0.0465, over 952604.69 frames. ], batch size: 18, lr: 2.85e-03, grad_scale: 16.0 2023-04-28 04:07:36,444 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8317, 2.2103, 1.9161, 2.1365, 1.6085, 1.8482, 1.7799, 1.5389], device='cuda:6'), covar=tensor([0.1739, 0.1113, 0.0740, 0.1233, 0.3101, 0.0926, 0.1720, 0.2341], device='cuda:6'), in_proj_covar=tensor([0.0280, 0.0298, 0.0215, 0.0274, 0.0311, 0.0251, 0.0246, 0.0262], device='cuda:6'), out_proj_covar=tensor([1.1162e-04, 1.1712e-04, 8.4266e-05, 1.0779e-04, 1.2530e-04, 9.8668e-05, 9.9089e-05, 1.0318e-04], device='cuda:6') 2023-04-28 04:08:00,376 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=161872.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:08:02,913 INFO [finetune.py:976] (6/7) Epoch 29, batch 1500, loss[loss=0.1597, simple_loss=0.2355, pruned_loss=0.04197, over 4772.00 frames. ], tot_loss[loss=0.1659, simple_loss=0.2399, pruned_loss=0.04594, over 954105.21 frames. ], batch size: 28, lr: 2.85e-03, grad_scale: 16.0 2023-04-28 04:08:03,649 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=161876.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:08:04,724 INFO [optim.py:369] (6/7) 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 04:08:16,474 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=161887.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:08:38,547 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.78 vs. limit=2.0 2023-04-28 04:08:58,050 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.2808, 2.8388, 2.4217, 2.7470, 2.1429, 2.3134, 2.5829, 1.8929], device='cuda:6'), covar=tensor([0.1989, 0.1369, 0.0797, 0.1281, 0.3244, 0.1185, 0.1779, 0.2886], device='cuda:6'), in_proj_covar=tensor([0.0281, 0.0298, 0.0215, 0.0274, 0.0312, 0.0251, 0.0246, 0.0262], device='cuda:6'), out_proj_covar=tensor([1.1175e-04, 1.1708e-04, 8.4266e-05, 1.0785e-04, 1.2564e-04, 9.8771e-05, 9.9199e-05, 1.0326e-04], device='cuda:6') 2023-04-28 04:09:01,027 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=161920.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:09:09,619 INFO [finetune.py:976] (6/7) Epoch 29, batch 1550, loss[loss=0.1533, simple_loss=0.2218, pruned_loss=0.04238, over 4825.00 frames. ], tot_loss[loss=0.166, simple_loss=0.24, pruned_loss=0.04593, over 954995.73 frames. ], batch size: 33, lr: 2.85e-03, grad_scale: 16.0 2023-04-28 04:09:21,323 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=161935.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:10:14,712 INFO [finetune.py:976] (6/7) Epoch 29, batch 1600, loss[loss=0.1611, simple_loss=0.2411, pruned_loss=0.04061, over 4844.00 frames. ], tot_loss[loss=0.1636, simple_loss=0.2373, pruned_loss=0.04499, over 956053.97 frames. ], batch size: 49, lr: 2.84e-03, grad_scale: 16.0 2023-04-28 04:10:16,469 INFO [optim.py:369] (6/7) 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 04:11:22,117 INFO [finetune.py:976] (6/7) Epoch 29, batch 1650, loss[loss=0.1435, simple_loss=0.2155, pruned_loss=0.0357, over 4736.00 frames. ], tot_loss[loss=0.1623, simple_loss=0.2354, pruned_loss=0.04457, over 958589.64 frames. ], batch size: 54, lr: 2.84e-03, grad_scale: 16.0 2023-04-28 04:12:07,539 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=162057.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:12:29,332 INFO [finetune.py:976] (6/7) Epoch 29, batch 1700, loss[loss=0.189, simple_loss=0.2502, pruned_loss=0.06393, over 4859.00 frames. ], tot_loss[loss=0.1605, simple_loss=0.2328, pruned_loss=0.04409, over 956405.00 frames. ], batch size: 31, lr: 2.84e-03, grad_scale: 16.0 2023-04-28 04:12:36,101 INFO [zipformer.py:1188] (6/7) 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] (6/7) 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 04:13:11,491 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=162105.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:13:20,970 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8556, 1.6156, 1.7714, 2.1713, 2.1734, 1.7473, 1.5659, 1.8953], device='cuda:6'), covar=tensor([0.0805, 0.1146, 0.0884, 0.0636, 0.0655, 0.0851, 0.0764, 0.0594], device='cuda:6'), in_proj_covar=tensor([0.0184, 0.0203, 0.0185, 0.0171, 0.0178, 0.0178, 0.0151, 0.0177], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-28 04:13:28,610 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=162116.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:13:34,113 INFO [finetune.py:976] (6/7) Epoch 29, batch 1750, loss[loss=0.147, simple_loss=0.2268, pruned_loss=0.03362, over 4793.00 frames. ], tot_loss[loss=0.164, simple_loss=0.2368, pruned_loss=0.04556, over 958107.73 frames. ], batch size: 29, lr: 2.84e-03, grad_scale: 16.0 2023-04-28 04:13:51,961 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=162138.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:14:27,313 INFO [zipformer.py:1188] (6/7) 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] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=162171.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:14:45,209 INFO [finetune.py:976] (6/7) Epoch 29, batch 1800, loss[loss=0.1283, simple_loss=0.186, pruned_loss=0.03526, over 4152.00 frames. ], tot_loss[loss=0.1656, simple_loss=0.2393, pruned_loss=0.04598, over 957145.60 frames. ], batch size: 18, lr: 2.84e-03, grad_scale: 16.0 2023-04-28 04:14:47,023 INFO [optim.py:369] (6/7) 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 04:15:49,786 INFO [finetune.py:976] (6/7) Epoch 29, batch 1850, loss[loss=0.1243, simple_loss=0.1944, pruned_loss=0.02714, over 3991.00 frames. ], tot_loss[loss=0.1657, simple_loss=0.2396, pruned_loss=0.04587, over 956904.90 frames. ], batch size: 17, lr: 2.84e-03, grad_scale: 16.0 2023-04-28 04:15:53,542 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=162231.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:16:53,969 INFO [finetune.py:976] (6/7) Epoch 29, batch 1900, loss[loss=0.1843, simple_loss=0.2596, pruned_loss=0.05446, over 4890.00 frames. ], tot_loss[loss=0.1659, simple_loss=0.2402, pruned_loss=0.04576, over 955625.34 frames. ], batch size: 32, lr: 2.84e-03, grad_scale: 16.0 2023-04-28 04:16:55,790 INFO [optim.py:369] (6/7) 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 04:16:56,551 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.4751, 1.4575, 1.7768, 1.7643, 1.3948, 1.2427, 1.5174, 0.9849], device='cuda:6'), covar=tensor([0.0534, 0.0466, 0.0364, 0.0477, 0.0647, 0.0794, 0.0508, 0.0536], device='cuda:6'), in_proj_covar=tensor([0.0070, 0.0067, 0.0065, 0.0069, 0.0075, 0.0094, 0.0073, 0.0062], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-28 04:17:14,636 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=162292.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:17:58,314 INFO [finetune.py:976] (6/7) Epoch 29, batch 1950, loss[loss=0.1706, simple_loss=0.2429, pruned_loss=0.04915, over 4847.00 frames. ], tot_loss[loss=0.1646, simple_loss=0.2386, pruned_loss=0.04527, over 955362.95 frames. ], batch size: 44, lr: 2.84e-03, grad_scale: 16.0 2023-04-28 04:19:03,739 INFO [finetune.py:976] (6/7) Epoch 29, batch 2000, loss[loss=0.1751, simple_loss=0.2478, pruned_loss=0.05126, over 4940.00 frames. ], tot_loss[loss=0.1628, simple_loss=0.2362, pruned_loss=0.04473, over 955378.27 frames. ], batch size: 33, lr: 2.84e-03, grad_scale: 16.0 2023-04-28 04:19:05,553 INFO [optim.py:369] (6/7) 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 04:19:43,204 INFO [finetune.py:976] (6/7) Epoch 29, batch 2050, loss[loss=0.1391, simple_loss=0.2085, pruned_loss=0.03489, over 4810.00 frames. ], tot_loss[loss=0.1615, simple_loss=0.2342, pruned_loss=0.04446, over 956688.91 frames. ], batch size: 45, lr: 2.84e-03, grad_scale: 16.0 2023-04-28 04:19:48,154 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=162433.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:19:58,582 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.2063, 1.6322, 2.0447, 2.1768, 2.0045, 1.5708, 1.1257, 1.7577], device='cuda:6'), covar=tensor([0.2943, 0.3032, 0.1643, 0.2003, 0.2350, 0.2734, 0.4245, 0.1841], device='cuda:6'), in_proj_covar=tensor([0.0290, 0.0246, 0.0227, 0.0312, 0.0222, 0.0235, 0.0227, 0.0185], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:6') 2023-04-28 04:20:13,305 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 29, batch 2100, loss[loss=0.2095, simple_loss=0.2834, pruned_loss=0.06778, over 4840.00 frames. ], tot_loss[loss=0.1615, simple_loss=0.2342, pruned_loss=0.04441, over 956225.35 frames. ], batch size: 47, lr: 2.84e-03, grad_scale: 32.0 2023-04-28 04:20:19,020 INFO [optim.py:369] (6/7) 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 04:20:39,349 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=162511.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:20:46,234 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=162519.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:20:50,372 INFO [finetune.py:976] (6/7) Epoch 29, batch 2150, loss[loss=0.2171, simple_loss=0.2877, pruned_loss=0.07329, over 4258.00 frames. ], tot_loss[loss=0.163, simple_loss=0.2366, pruned_loss=0.04471, over 955710.01 frames. ], batch size: 65, lr: 2.84e-03, grad_scale: 32.0 2023-04-28 04:21:01,282 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-04-28 04:21:22,077 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=162572.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 04:21:22,996 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.79 vs. limit=5.0 2023-04-28 04:21:23,767 INFO [finetune.py:976] (6/7) Epoch 29, batch 2200, loss[loss=0.1611, simple_loss=0.2313, pruned_loss=0.04544, over 4766.00 frames. ], tot_loss[loss=0.1652, simple_loss=0.2393, pruned_loss=0.04552, over 955087.07 frames. ], batch size: 26, lr: 2.84e-03, grad_scale: 32.0 2023-04-28 04:21:26,030 INFO [optim.py:369] (6/7) 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 04:21:37,954 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=162587.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:22:23,020 INFO [finetune.py:976] (6/7) Epoch 29, batch 2250, loss[loss=0.1499, simple_loss=0.2388, pruned_loss=0.03051, over 4787.00 frames. ], tot_loss[loss=0.1667, simple_loss=0.2408, pruned_loss=0.04631, over 952707.90 frames. ], batch size: 29, lr: 2.84e-03, grad_scale: 32.0 2023-04-28 04:22:32,732 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.59 vs. limit=2.0 2023-04-28 04:22:57,162 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-04-28 04:23:03,592 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.4483, 2.2169, 1.8012, 1.8583, 2.3178, 1.7895, 2.5662, 1.5751], device='cuda:6'), covar=tensor([0.3248, 0.2079, 0.4576, 0.3054, 0.1576, 0.2617, 0.1832, 0.4600], device='cuda:6'), in_proj_covar=tensor([0.0336, 0.0352, 0.0421, 0.0352, 0.0383, 0.0374, 0.0369, 0.0424], device='cuda:6'), out_proj_covar=tensor([9.9234e-05, 1.0462e-04, 1.2725e-04, 1.0519e-04, 1.1316e-04, 1.1102e-04, 1.0760e-04, 1.2726e-04], device='cuda:6') 2023-04-28 04:23:28,886 INFO [finetune.py:976] (6/7) Epoch 29, batch 2300, loss[loss=0.1445, simple_loss=0.2267, pruned_loss=0.03117, over 4816.00 frames. ], tot_loss[loss=0.1655, simple_loss=0.24, pruned_loss=0.04556, over 952840.96 frames. ], batch size: 33, lr: 2.84e-03, grad_scale: 32.0 2023-04-28 04:23:36,410 INFO [optim.py:369] (6/7) 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 04:23:47,269 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5255, 3.4784, 0.9210, 1.8942, 1.9476, 2.5468, 1.9686, 1.0500], device='cuda:6'), covar=tensor([0.1348, 0.0856, 0.1912, 0.1187, 0.0998, 0.0922, 0.1454, 0.1992], device='cuda:6'), in_proj_covar=tensor([0.0117, 0.0237, 0.0136, 0.0120, 0.0131, 0.0153, 0.0117, 0.0118], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-28 04:23:49,724 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7281, 2.6880, 2.2076, 2.4165, 2.8229, 2.3212, 3.5654, 2.1288], device='cuda:6'), covar=tensor([0.3481, 0.2385, 0.4370, 0.3734, 0.1677, 0.2655, 0.1255, 0.3963], device='cuda:6'), in_proj_covar=tensor([0.0336, 0.0351, 0.0420, 0.0352, 0.0382, 0.0374, 0.0369, 0.0423], device='cuda:6'), out_proj_covar=tensor([9.9150e-05, 1.0449e-04, 1.2711e-04, 1.0509e-04, 1.1293e-04, 1.1086e-04, 1.0752e-04, 1.2707e-04], device='cuda:6') 2023-04-28 04:24:31,809 INFO [finetune.py:976] (6/7) Epoch 29, batch 2350, loss[loss=0.1407, simple_loss=0.2032, pruned_loss=0.03912, over 4830.00 frames. ], tot_loss[loss=0.1641, simple_loss=0.238, pruned_loss=0.04513, over 953342.07 frames. ], batch size: 33, lr: 2.84e-03, grad_scale: 32.0 2023-04-28 04:24:41,996 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=162733.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:24:51,446 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.73 vs. limit=2.0 2023-04-28 04:25:30,907 INFO [finetune.py:976] (6/7) Epoch 29, batch 2400, loss[loss=0.1858, simple_loss=0.25, pruned_loss=0.0608, over 4138.00 frames. ], tot_loss[loss=0.1616, simple_loss=0.2351, pruned_loss=0.04407, over 954386.08 frames. ], batch size: 65, lr: 2.84e-03, grad_scale: 32.0 2023-04-28 04:25:31,034 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.3734, 2.2443, 1.9534, 1.9432, 2.5263, 1.8101, 2.8445, 1.7959], device='cuda:6'), covar=tensor([0.3291, 0.2250, 0.4165, 0.2994, 0.1519, 0.2489, 0.1361, 0.3912], device='cuda:6'), in_proj_covar=tensor([0.0338, 0.0354, 0.0423, 0.0354, 0.0384, 0.0376, 0.0372, 0.0426], device='cuda:6'), out_proj_covar=tensor([9.9795e-05, 1.0529e-04, 1.2789e-04, 1.0575e-04, 1.1363e-04, 1.1161e-04, 1.0832e-04, 1.2786e-04], device='cuda:6') 2023-04-28 04:25:33,152 INFO [optim.py:369] (6/7) 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 04:25:36,110 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=162781.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:25:46,238 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.5554, 1.9624, 1.7156, 2.3545, 2.5385, 2.1038, 2.0106, 1.7176], device='cuda:6'), covar=tensor([0.2113, 0.1893, 0.2174, 0.1981, 0.1216, 0.2023, 0.2548, 0.2787], device='cuda:6'), in_proj_covar=tensor([0.0316, 0.0309, 0.0351, 0.0287, 0.0326, 0.0306, 0.0302, 0.0378], device='cuda:6'), out_proj_covar=tensor([6.4498e-05, 6.3371e-05, 7.3409e-05, 5.7384e-05, 6.6475e-05, 6.3534e-05, 6.2266e-05, 7.9806e-05], device='cuda:6') 2023-04-28 04:25:48,082 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.5749, 2.0284, 1.8044, 2.4668, 2.6574, 2.1782, 2.1020, 1.8274], device='cuda:6'), covar=tensor([0.1904, 0.1720, 0.2064, 0.1716, 0.1150, 0.1761, 0.2084, 0.2357], device='cuda:6'), in_proj_covar=tensor([0.0316, 0.0309, 0.0351, 0.0287, 0.0326, 0.0306, 0.0302, 0.0378], device='cuda:6'), out_proj_covar=tensor([6.4459e-05, 6.3338e-05, 7.3369e-05, 5.7347e-05, 6.6442e-05, 6.3500e-05, 6.2244e-05, 7.9773e-05], device='cuda:6') 2023-04-28 04:26:04,585 INFO [finetune.py:976] (6/7) Epoch 29, batch 2450, loss[loss=0.1594, simple_loss=0.2301, pruned_loss=0.04439, over 4904.00 frames. ], tot_loss[loss=0.1594, simple_loss=0.2322, pruned_loss=0.04331, over 953066.40 frames. ], batch size: 43, lr: 2.84e-03, grad_scale: 32.0 2023-04-28 04:26:05,925 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.4755, 3.3554, 1.1652, 1.8567, 1.8104, 2.4898, 2.0130, 0.9778], device='cuda:6'), covar=tensor([0.1396, 0.0877, 0.1621, 0.1213, 0.1108, 0.0935, 0.1327, 0.2056], device='cuda:6'), in_proj_covar=tensor([0.0117, 0.0237, 0.0136, 0.0120, 0.0131, 0.0153, 0.0116, 0.0118], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-28 04:26:05,954 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=162827.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:26:33,772 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=162867.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 04:26:38,578 INFO [finetune.py:976] (6/7) Epoch 29, batch 2500, loss[loss=0.2083, simple_loss=0.2783, pruned_loss=0.06918, over 4819.00 frames. ], tot_loss[loss=0.1618, simple_loss=0.2346, pruned_loss=0.04453, over 952620.88 frames. ], batch size: 40, lr: 2.84e-03, grad_scale: 32.0 2023-04-28 04:26:40,368 INFO [optim.py:369] (6/7) 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 04:26:47,977 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=162887.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:26:48,603 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=162888.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:27:03,475 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-04-28 04:27:12,362 INFO [finetune.py:976] (6/7) Epoch 29, batch 2550, loss[loss=0.1691, simple_loss=0.2459, pruned_loss=0.04614, over 4831.00 frames. ], tot_loss[loss=0.1642, simple_loss=0.2379, pruned_loss=0.04521, over 952246.32 frames. ], batch size: 47, lr: 2.84e-03, grad_scale: 32.0 2023-04-28 04:27:18,486 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.3432, 1.3303, 1.6578, 1.5906, 1.2067, 1.1866, 1.3387, 0.9057], device='cuda:6'), covar=tensor([0.0616, 0.0515, 0.0384, 0.0532, 0.0699, 0.0912, 0.0486, 0.0508], device='cuda:6'), in_proj_covar=tensor([0.0072, 0.0067, 0.0065, 0.0070, 0.0076, 0.0095, 0.0073, 0.0062], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-28 04:27:19,030 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=162935.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:27:19,466 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.80 vs. limit=2.0 2023-04-28 04:27:22,390 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.2592, 2.1968, 2.1653, 1.9368, 2.4344, 1.9528, 2.9238, 1.8522], device='cuda:6'), covar=tensor([0.3213, 0.1843, 0.4027, 0.2811, 0.1417, 0.2261, 0.1156, 0.3953], device='cuda:6'), in_proj_covar=tensor([0.0338, 0.0353, 0.0423, 0.0354, 0.0384, 0.0375, 0.0371, 0.0425], device='cuda:6'), out_proj_covar=tensor([9.9734e-05, 1.0505e-04, 1.2780e-04, 1.0572e-04, 1.1363e-04, 1.1132e-04, 1.0802e-04, 1.2745e-04], device='cuda:6') 2023-04-28 04:27:46,026 INFO [finetune.py:976] (6/7) Epoch 29, batch 2600, loss[loss=0.1878, simple_loss=0.2705, pruned_loss=0.05259, over 4803.00 frames. ], tot_loss[loss=0.1665, simple_loss=0.2405, pruned_loss=0.04628, over 953139.32 frames. ], batch size: 45, lr: 2.84e-03, grad_scale: 32.0 2023-04-28 04:27:47,807 INFO [optim.py:369] (6/7) 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 04:28:11,109 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5815, 1.2448, 0.6117, 1.2695, 1.5478, 1.4681, 1.3805, 1.3662], device='cuda:6'), covar=tensor([0.0497, 0.0417, 0.0378, 0.0552, 0.0287, 0.0518, 0.0502, 0.0596], device='cuda:6'), in_proj_covar=tensor([0.0028, 0.0023, 0.0022, 0.0028, 0.0019, 0.0027, 0.0027, 0.0028], device='cuda:6'), out_proj_covar=tensor([0.0052, 0.0046, 0.0039, 0.0053, 0.0039, 0.0051, 0.0051, 0.0053], device='cuda:6') 2023-04-28 04:28:19,511 INFO [finetune.py:976] (6/7) Epoch 29, batch 2650, loss[loss=0.1616, simple_loss=0.254, pruned_loss=0.03463, over 4815.00 frames. ], tot_loss[loss=0.1661, simple_loss=0.24, pruned_loss=0.04613, over 951939.53 frames. ], batch size: 38, lr: 2.84e-03, grad_scale: 32.0 2023-04-28 04:28:30,394 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8872, 1.3928, 1.6912, 1.7184, 1.6643, 1.3804, 0.8495, 1.4092], device='cuda:6'), covar=tensor([0.2789, 0.2905, 0.1515, 0.1821, 0.2140, 0.2339, 0.4176, 0.1786], device='cuda:6'), in_proj_covar=tensor([0.0291, 0.0245, 0.0226, 0.0312, 0.0221, 0.0234, 0.0226, 0.0184], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:6') 2023-04-28 04:28:57,567 INFO [finetune.py:976] (6/7) Epoch 29, batch 2700, loss[loss=0.1734, simple_loss=0.249, pruned_loss=0.04885, over 4814.00 frames. ], tot_loss[loss=0.1659, simple_loss=0.2397, pruned_loss=0.04603, over 951648.29 frames. ], batch size: 38, lr: 2.84e-03, grad_scale: 32.0 2023-04-28 04:29:04,330 INFO [optim.py:369] (6/7) 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 04:29:17,271 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=163090.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:29:46,789 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9675, 1.4764, 2.0143, 2.4057, 2.0565, 1.9430, 1.9873, 1.9088], device='cuda:6'), covar=tensor([0.4250, 0.6531, 0.6127, 0.5199, 0.5475, 0.7636, 0.7707, 0.8441], device='cuda:6'), in_proj_covar=tensor([0.0448, 0.0425, 0.0520, 0.0508, 0.0475, 0.0511, 0.0515, 0.0529], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-28 04:30:01,168 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=3.30 vs. limit=5.0 2023-04-28 04:30:02,213 INFO [finetune.py:976] (6/7) Epoch 29, batch 2750, loss[loss=0.1305, simple_loss=0.2046, pruned_loss=0.02817, over 4767.00 frames. ], tot_loss[loss=0.1635, simple_loss=0.2363, pruned_loss=0.04529, over 951672.85 frames. ], batch size: 23, lr: 2.84e-03, grad_scale: 32.0 2023-04-28 04:30:22,073 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.4352, 2.9181, 1.0990, 1.5981, 2.3520, 1.3857, 4.0019, 2.0304], device='cuda:6'), covar=tensor([0.0571, 0.0769, 0.0851, 0.1195, 0.0455, 0.0898, 0.0255, 0.0558], device='cuda:6'), in_proj_covar=tensor([0.0050, 0.0064, 0.0047, 0.0046, 0.0049, 0.0051, 0.0072, 0.0050], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:6') 2023-04-28 04:30:39,011 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=163151.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:30:55,831 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=163167.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 04:30:58,154 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.1687, 2.5402, 1.0536, 1.4370, 1.8722, 1.2676, 3.1255, 1.7629], device='cuda:6'), covar=tensor([0.0563, 0.0456, 0.0626, 0.1104, 0.0428, 0.0915, 0.0256, 0.0550], device='cuda:6'), in_proj_covar=tensor([0.0050, 0.0064, 0.0047, 0.0046, 0.0049, 0.0051, 0.0072, 0.0050], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:6') 2023-04-28 04:31:01,094 INFO [finetune.py:976] (6/7) Epoch 29, batch 2800, loss[loss=0.1751, simple_loss=0.2458, pruned_loss=0.05217, over 4912.00 frames. ], tot_loss[loss=0.1596, simple_loss=0.2321, pruned_loss=0.04351, over 949393.15 frames. ], batch size: 36, lr: 2.84e-03, grad_scale: 32.0 2023-04-28 04:31:08,100 INFO [optim.py:369] (6/7) 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 04:31:11,257 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=163183.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:31:50,482 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.12 vs. limit=2.0 2023-04-28 04:31:52,469 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 29, batch 2850, loss[loss=0.1769, simple_loss=0.2442, pruned_loss=0.05486, over 4815.00 frames. ], tot_loss[loss=0.1587, simple_loss=0.2315, pruned_loss=0.04293, over 950451.98 frames. ], batch size: 38, lr: 2.84e-03, grad_scale: 32.0 2023-04-28 04:32:06,417 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.8963, 2.7971, 2.2824, 3.2819, 2.8793, 2.8489, 1.1736, 2.8239], device='cuda:6'), covar=tensor([0.2309, 0.1894, 0.3758, 0.3080, 0.4856, 0.2347, 0.6250, 0.3229], device='cuda:6'), in_proj_covar=tensor([0.0250, 0.0223, 0.0256, 0.0308, 0.0305, 0.0255, 0.0280, 0.0279], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-28 04:32:08,304 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=163240.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:32:08,932 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6841, 2.3690, 1.6613, 1.7965, 1.2414, 1.2771, 1.7489, 1.2473], device='cuda:6'), covar=tensor([0.1920, 0.1233, 0.1561, 0.1681, 0.2485, 0.2352, 0.0987, 0.2162], device='cuda:6'), in_proj_covar=tensor([0.0197, 0.0209, 0.0170, 0.0203, 0.0200, 0.0186, 0.0155, 0.0186], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:6') 2023-04-28 04:32:30,351 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=3.87 vs. limit=5.0 2023-04-28 04:32:31,405 INFO [finetune.py:976] (6/7) Epoch 29, batch 2900, loss[loss=0.1333, simple_loss=0.2121, pruned_loss=0.02722, over 4710.00 frames. ], tot_loss[loss=0.1604, simple_loss=0.2335, pruned_loss=0.04365, over 950068.17 frames. ], batch size: 23, lr: 2.84e-03, grad_scale: 32.0 2023-04-28 04:32:33,719 INFO [optim.py:369] (6/7) 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 04:32:47,875 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 29, batch 2950, loss[loss=0.1589, simple_loss=0.2427, pruned_loss=0.03749, over 4895.00 frames. ], tot_loss[loss=0.1641, simple_loss=0.2381, pruned_loss=0.04511, over 952934.59 frames. ], batch size: 32, lr: 2.84e-03, grad_scale: 32.0 2023-04-28 04:33:12,083 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6046, 3.5064, 1.1283, 2.0386, 2.1264, 2.5904, 2.2214, 1.1586], device='cuda:6'), covar=tensor([0.1358, 0.0948, 0.1755, 0.1206, 0.1012, 0.0973, 0.1423, 0.2063], device='cuda:6'), in_proj_covar=tensor([0.0118, 0.0238, 0.0136, 0.0120, 0.0131, 0.0152, 0.0117, 0.0119], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-28 04:33:31,405 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.58 vs. limit=2.0 2023-04-28 04:33:37,002 INFO [finetune.py:976] (6/7) Epoch 29, batch 3000, loss[loss=0.1638, simple_loss=0.2399, pruned_loss=0.04382, over 4094.00 frames. ], tot_loss[loss=0.1642, simple_loss=0.2386, pruned_loss=0.04487, over 952580.72 frames. ], batch size: 65, lr: 2.84e-03, grad_scale: 32.0 2023-04-28 04:33:37,003 INFO [finetune.py:1001] (6/7) Computing validation loss 2023-04-28 04:33:44,560 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7482, 2.1921, 1.8373, 2.1317, 1.6840, 1.8737, 1.8367, 1.4695], device='cuda:6'), covar=tensor([0.1800, 0.1065, 0.0797, 0.0982, 0.3343, 0.0950, 0.1648, 0.2292], device='cuda:6'), in_proj_covar=tensor([0.0284, 0.0299, 0.0217, 0.0276, 0.0313, 0.0253, 0.0247, 0.0264], device='cuda:6'), out_proj_covar=tensor([1.1301e-04, 1.1745e-04, 8.5116e-05, 1.0839e-04, 1.2617e-04, 9.9184e-05, 9.9584e-05, 1.0383e-04], device='cuda:6') 2023-04-28 04:33:47,841 INFO [finetune.py:1010] (6/7) 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] (6/7) Maximum memory allocated so far is 6435MB 2023-04-28 04:33:49,648 INFO [optim.py:369] (6/7) 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 04:33:58,144 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.2551, 1.1932, 1.4160, 1.4570, 1.1938, 1.0865, 1.2371, 0.8573], device='cuda:6'), covar=tensor([0.0515, 0.0468, 0.0379, 0.0375, 0.0595, 0.0872, 0.0428, 0.0465], device='cuda:6'), in_proj_covar=tensor([0.0071, 0.0067, 0.0065, 0.0069, 0.0076, 0.0095, 0.0073, 0.0062], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-28 04:34:19,347 INFO [finetune.py:976] (6/7) Epoch 29, batch 3050, loss[loss=0.1532, simple_loss=0.2365, pruned_loss=0.03496, over 4763.00 frames. ], tot_loss[loss=0.1664, simple_loss=0.2411, pruned_loss=0.04586, over 952163.69 frames. ], batch size: 54, lr: 2.84e-03, grad_scale: 32.0 2023-04-28 04:34:23,677 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.26 vs. limit=5.0 2023-04-28 04:34:34,889 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=163446.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:34:50,165 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.35 vs. limit=5.0 2023-04-28 04:34:53,038 INFO [finetune.py:976] (6/7) Epoch 29, batch 3100, loss[loss=0.1743, simple_loss=0.2432, pruned_loss=0.05271, over 4906.00 frames. ], tot_loss[loss=0.1642, simple_loss=0.2388, pruned_loss=0.04476, over 953319.90 frames. ], batch size: 43, lr: 2.84e-03, grad_scale: 32.0 2023-04-28 04:34:55,820 INFO [optim.py:369] (6/7) 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 04:34:59,436 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=163483.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:35:27,055 INFO [finetune.py:976] (6/7) Epoch 29, batch 3150, loss[loss=0.1564, simple_loss=0.2239, pruned_loss=0.04442, over 4881.00 frames. ], tot_loss[loss=0.1608, simple_loss=0.2349, pruned_loss=0.04334, over 954331.64 frames. ], batch size: 31, lr: 2.84e-03, grad_scale: 32.0 2023-04-28 04:35:32,222 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=163531.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:35:35,249 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5148, 1.5789, 1.8635, 1.8833, 1.4113, 1.3595, 1.5638, 0.9892], device='cuda:6'), covar=tensor([0.0572, 0.0556, 0.0357, 0.0569, 0.0681, 0.1009, 0.0520, 0.0524], device='cuda:6'), in_proj_covar=tensor([0.0071, 0.0067, 0.0065, 0.0069, 0.0076, 0.0095, 0.0073, 0.0062], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-28 04:35:41,164 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.1332, 2.6688, 1.1265, 1.3990, 1.8879, 1.2555, 3.4990, 1.7944], device='cuda:6'), covar=tensor([0.0628, 0.0595, 0.0779, 0.1225, 0.0546, 0.1006, 0.0223, 0.0578], device='cuda:6'), in_proj_covar=tensor([0.0050, 0.0064, 0.0047, 0.0046, 0.0049, 0.0051, 0.0073, 0.0050], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:6') 2023-04-28 04:36:10,395 INFO [finetune.py:976] (6/7) Epoch 29, batch 3200, loss[loss=0.1672, simple_loss=0.2357, pruned_loss=0.04938, over 4823.00 frames. ], tot_loss[loss=0.1587, simple_loss=0.2319, pruned_loss=0.04275, over 953530.03 frames. ], batch size: 39, lr: 2.83e-03, grad_scale: 32.0 2023-04-28 04:36:12,739 INFO [optim.py:369] (6/7) 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 04:36:42,365 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-04-28 04:36:42,520 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=163596.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:37:05,386 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9194, 2.4464, 1.1607, 1.2704, 1.7833, 1.1521, 3.2087, 1.6333], device='cuda:6'), covar=tensor([0.0732, 0.0608, 0.0769, 0.1210, 0.0530, 0.1091, 0.0243, 0.0621], device='cuda:6'), in_proj_covar=tensor([0.0050, 0.0064, 0.0047, 0.0046, 0.0049, 0.0051, 0.0073, 0.0050], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:6') 2023-04-28 04:37:17,130 INFO [finetune.py:976] (6/7) Epoch 29, batch 3250, loss[loss=0.1181, simple_loss=0.188, pruned_loss=0.02412, over 4160.00 frames. ], tot_loss[loss=0.1615, simple_loss=0.2346, pruned_loss=0.0442, over 952946.17 frames. ], batch size: 17, lr: 2.83e-03, grad_scale: 32.0 2023-04-28 04:38:20,407 INFO [finetune.py:976] (6/7) Epoch 29, batch 3300, loss[loss=0.1536, simple_loss=0.2422, pruned_loss=0.03257, over 4760.00 frames. ], tot_loss[loss=0.1634, simple_loss=0.2369, pruned_loss=0.04497, over 952349.98 frames. ], batch size: 28, lr: 2.83e-03, grad_scale: 32.0 2023-04-28 04:38:22,208 INFO [optim.py:369] (6/7) 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 04:38:22,974 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7398, 2.1917, 1.9878, 1.7859, 1.3194, 1.3747, 2.0973, 1.3086], device='cuda:6'), covar=tensor([0.1823, 0.1622, 0.1344, 0.1764, 0.2455, 0.2099, 0.0865, 0.2106], device='cuda:6'), in_proj_covar=tensor([0.0200, 0.0211, 0.0171, 0.0205, 0.0202, 0.0188, 0.0157, 0.0188], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:6') 2023-04-28 04:39:02,777 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.4070, 3.2184, 0.9354, 1.6614, 1.8291, 2.3647, 1.7800, 1.0792], device='cuda:6'), covar=tensor([0.1511, 0.0922, 0.2074, 0.1338, 0.1157, 0.0998, 0.1602, 0.1947], device='cuda:6'), in_proj_covar=tensor([0.0118, 0.0239, 0.0136, 0.0121, 0.0132, 0.0153, 0.0117, 0.0119], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-28 04:39:22,574 INFO [finetune.py:976] (6/7) Epoch 29, batch 3350, loss[loss=0.1197, simple_loss=0.1879, pruned_loss=0.02576, over 3818.00 frames. ], tot_loss[loss=0.1661, simple_loss=0.2401, pruned_loss=0.04605, over 949504.41 frames. ], batch size: 16, lr: 2.83e-03, grad_scale: 32.0 2023-04-28 04:39:46,998 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=163746.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:39:57,962 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.75 vs. limit=2.0 2023-04-28 04:40:27,100 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9563, 2.5456, 1.7466, 2.1495, 1.4814, 1.4677, 1.8962, 1.3764], device='cuda:6'), covar=tensor([0.1934, 0.1585, 0.1719, 0.1713, 0.2593, 0.2389, 0.1099, 0.2274], device='cuda:6'), in_proj_covar=tensor([0.0198, 0.0210, 0.0170, 0.0204, 0.0201, 0.0187, 0.0156, 0.0187], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:6') 2023-04-28 04:40:27,591 INFO [finetune.py:976] (6/7) Epoch 29, batch 3400, loss[loss=0.1532, simple_loss=0.2409, pruned_loss=0.03276, over 4814.00 frames. ], tot_loss[loss=0.1675, simple_loss=0.2415, pruned_loss=0.04677, over 952735.33 frames. ], batch size: 40, lr: 2.83e-03, grad_scale: 32.0 2023-04-28 04:40:29,472 INFO [optim.py:369] (6/7) 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 04:40:50,564 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=163794.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:41:32,561 INFO [finetune.py:976] (6/7) Epoch 29, batch 3450, loss[loss=0.1534, simple_loss=0.2283, pruned_loss=0.03927, over 4764.00 frames. ], tot_loss[loss=0.1657, simple_loss=0.2399, pruned_loss=0.04573, over 952152.44 frames. ], batch size: 23, lr: 2.83e-03, grad_scale: 32.0 2023-04-28 04:41:43,590 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-04-28 04:41:55,043 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.3354, 1.6296, 1.3954, 1.5580, 1.3204, 1.4114, 1.3747, 1.1045], device='cuda:6'), covar=tensor([0.1649, 0.1053, 0.0837, 0.1053, 0.3333, 0.0999, 0.1551, 0.2050], device='cuda:6'), in_proj_covar=tensor([0.0283, 0.0299, 0.0217, 0.0276, 0.0313, 0.0253, 0.0247, 0.0264], device='cuda:6'), out_proj_covar=tensor([1.1297e-04, 1.1743e-04, 8.5099e-05, 1.0858e-04, 1.2613e-04, 9.9532e-05, 9.9396e-05, 1.0414e-04], device='cuda:6') 2023-04-28 04:42:04,542 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-04-28 04:42:36,322 INFO [finetune.py:976] (6/7) Epoch 29, batch 3500, loss[loss=0.1367, simple_loss=0.2149, pruned_loss=0.02931, over 4823.00 frames. ], tot_loss[loss=0.1642, simple_loss=0.2379, pruned_loss=0.04529, over 951846.64 frames. ], batch size: 40, lr: 2.83e-03, grad_scale: 32.0 2023-04-28 04:42:38,151 INFO [optim.py:369] (6/7) 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] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=163896.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:43:28,395 INFO [finetune.py:976] (6/7) Epoch 29, batch 3550, loss[loss=0.1487, simple_loss=0.2153, pruned_loss=0.04112, over 4875.00 frames. ], tot_loss[loss=0.1625, simple_loss=0.2356, pruned_loss=0.04476, over 954632.72 frames. ], batch size: 34, lr: 2.83e-03, grad_scale: 32.0 2023-04-28 04:43:30,929 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0534, 4.2725, 0.7162, 2.1329, 2.5775, 2.6947, 2.4929, 1.0218], device='cuda:6'), covar=tensor([0.1289, 0.0960, 0.2108, 0.1242, 0.0930, 0.1126, 0.1317, 0.2105], device='cuda:6'), in_proj_covar=tensor([0.0117, 0.0236, 0.0135, 0.0120, 0.0131, 0.0152, 0.0116, 0.0118], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-28 04:43:37,072 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.67 vs. limit=2.0 2023-04-28 04:43:39,968 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=163944.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:43:44,889 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.60 vs. limit=2.0 2023-04-28 04:43:54,914 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.73 vs. limit=2.0 2023-04-28 04:44:01,290 INFO [finetune.py:976] (6/7) Epoch 29, batch 3600, loss[loss=0.1602, simple_loss=0.2285, pruned_loss=0.04592, over 4715.00 frames. ], tot_loss[loss=0.1614, simple_loss=0.2338, pruned_loss=0.04451, over 956239.28 frames. ], batch size: 23, lr: 2.83e-03, grad_scale: 32.0 2023-04-28 04:44:03,547 INFO [optim.py:369] (6/7) 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 04:44:36,731 INFO [finetune.py:976] (6/7) Epoch 29, batch 3650, loss[loss=0.1602, simple_loss=0.2212, pruned_loss=0.0496, over 4166.00 frames. ], tot_loss[loss=0.1663, simple_loss=0.2383, pruned_loss=0.04711, over 953938.64 frames. ], batch size: 18, lr: 2.83e-03, grad_scale: 32.0 2023-04-28 04:44:52,629 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.1588, 2.0303, 2.0571, 1.7585, 2.2302, 1.8661, 2.6880, 1.8056], device='cuda:6'), covar=tensor([0.2768, 0.1535, 0.3454, 0.2125, 0.1292, 0.1955, 0.1143, 0.3746], device='cuda:6'), in_proj_covar=tensor([0.0340, 0.0355, 0.0424, 0.0354, 0.0386, 0.0377, 0.0373, 0.0427], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-28 04:45:10,041 INFO [finetune.py:976] (6/7) Epoch 29, batch 3700, loss[loss=0.1462, simple_loss=0.2138, pruned_loss=0.03932, over 4716.00 frames. ], tot_loss[loss=0.1673, simple_loss=0.2403, pruned_loss=0.04716, over 951403.24 frames. ], batch size: 23, lr: 2.83e-03, grad_scale: 32.0 2023-04-28 04:45:11,858 INFO [optim.py:369] (6/7) 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 04:45:16,036 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.6376, 3.2061, 2.6266, 3.0158, 2.2482, 2.6742, 2.8266, 2.0136], device='cuda:6'), covar=tensor([0.1714, 0.0878, 0.0731, 0.1003, 0.2718, 0.0917, 0.1577, 0.2256], device='cuda:6'), in_proj_covar=tensor([0.0283, 0.0299, 0.0216, 0.0276, 0.0313, 0.0253, 0.0246, 0.0263], device='cuda:6'), out_proj_covar=tensor([1.1258e-04, 1.1732e-04, 8.4971e-05, 1.0839e-04, 1.2602e-04, 9.9295e-05, 9.9127e-05, 1.0359e-04], device='cuda:6') 2023-04-28 04:45:19,692 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=164090.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:45:43,379 INFO [finetune.py:976] (6/7) Epoch 29, batch 3750, loss[loss=0.1635, simple_loss=0.2503, pruned_loss=0.03833, over 4832.00 frames. ], tot_loss[loss=0.1663, simple_loss=0.24, pruned_loss=0.04633, over 951784.83 frames. ], batch size: 47, lr: 2.83e-03, grad_scale: 32.0 2023-04-28 04:45:57,605 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8428, 1.4046, 1.9370, 2.2951, 1.9333, 1.7237, 1.8178, 1.8127], device='cuda:6'), covar=tensor([0.4569, 0.6954, 0.6431, 0.5371, 0.5821, 0.8379, 0.8184, 0.9525], device='cuda:6'), in_proj_covar=tensor([0.0451, 0.0427, 0.0524, 0.0510, 0.0476, 0.0514, 0.0517, 0.0532], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-28 04:46:11,056 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=164151.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:46:18,676 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=164155.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:46:43,098 INFO [finetune.py:976] (6/7) Epoch 29, batch 3800, loss[loss=0.1984, simple_loss=0.2816, pruned_loss=0.05758, over 4879.00 frames. ], tot_loss[loss=0.1665, simple_loss=0.2406, pruned_loss=0.04614, over 951869.49 frames. ], batch size: 35, lr: 2.83e-03, grad_scale: 32.0 2023-04-28 04:46:47,404 INFO [optim.py:369] (6/7) 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 04:46:54,800 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9417, 2.5201, 0.9964, 1.3487, 1.9391, 1.2587, 3.3063, 1.7438], device='cuda:6'), covar=tensor([0.0698, 0.0531, 0.0782, 0.1248, 0.0483, 0.1007, 0.0204, 0.0580], device='cuda:6'), in_proj_covar=tensor([0.0050, 0.0063, 0.0047, 0.0046, 0.0049, 0.0051, 0.0072, 0.0050], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0007], device='cuda:6') 2023-04-28 04:47:23,356 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=164216.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:47:34,746 INFO [finetune.py:976] (6/7) Epoch 29, batch 3850, loss[loss=0.1408, simple_loss=0.2144, pruned_loss=0.03365, over 4305.00 frames. ], tot_loss[loss=0.1647, simple_loss=0.239, pruned_loss=0.04525, over 953431.67 frames. ], batch size: 65, lr: 2.83e-03, grad_scale: 32.0 2023-04-28 04:48:38,392 INFO [finetune.py:976] (6/7) Epoch 29, batch 3900, loss[loss=0.1749, simple_loss=0.2445, pruned_loss=0.05263, over 4849.00 frames. ], tot_loss[loss=0.1632, simple_loss=0.2367, pruned_loss=0.04481, over 955360.91 frames. ], batch size: 49, lr: 2.83e-03, grad_scale: 32.0 2023-04-28 04:48:40,200 INFO [optim.py:369] (6/7) 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 04:48:48,789 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6969, 1.0016, 1.6335, 2.2366, 1.8009, 1.6371, 1.6876, 1.6541], device='cuda:6'), covar=tensor([0.3968, 0.6322, 0.5616, 0.4879, 0.5135, 0.6866, 0.6921, 0.8041], device='cuda:6'), in_proj_covar=tensor([0.0450, 0.0427, 0.0524, 0.0512, 0.0477, 0.0516, 0.0518, 0.0532], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-28 04:49:39,856 INFO [finetune.py:976] (6/7) Epoch 29, batch 3950, loss[loss=0.159, simple_loss=0.2291, pruned_loss=0.04444, over 4784.00 frames. ], tot_loss[loss=0.1606, simple_loss=0.2335, pruned_loss=0.04383, over 956610.59 frames. ], batch size: 25, lr: 2.83e-03, grad_scale: 32.0 2023-04-28 04:50:24,355 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=164360.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:50:24,662 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=5.08 vs. limit=5.0 2023-04-28 04:50:25,010 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.4593, 1.5630, 1.4238, 1.0759, 1.1523, 1.0981, 1.4476, 1.1302], device='cuda:6'), covar=tensor([0.1730, 0.1552, 0.1551, 0.1827, 0.2379, 0.2054, 0.0991, 0.2164], device='cuda:6'), in_proj_covar=tensor([0.0199, 0.0210, 0.0171, 0.0204, 0.0201, 0.0187, 0.0157, 0.0187], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:6') 2023-04-28 04:50:45,027 INFO [finetune.py:976] (6/7) Epoch 29, batch 4000, loss[loss=0.1791, simple_loss=0.2553, pruned_loss=0.05145, over 4850.00 frames. ], tot_loss[loss=0.1603, simple_loss=0.2334, pruned_loss=0.04365, over 957880.70 frames. ], batch size: 49, lr: 2.83e-03, grad_scale: 32.0 2023-04-28 04:50:47,324 INFO [optim.py:369] (6/7) 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 04:51:31,284 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.76 vs. limit=2.0 2023-04-28 04:51:41,781 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 29, batch 4050, loss[loss=0.1439, simple_loss=0.2248, pruned_loss=0.03148, over 4753.00 frames. ], tot_loss[loss=0.1637, simple_loss=0.2374, pruned_loss=0.04497, over 957870.41 frames. ], batch size: 27, lr: 2.83e-03, grad_scale: 32.0 2023-04-28 04:52:09,022 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=3.80 vs. limit=5.0 2023-04-28 04:52:11,696 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.2353, 1.0692, 3.7524, 3.5240, 3.3455, 3.6341, 3.6047, 3.3351], device='cuda:6'), covar=tensor([0.7913, 0.6385, 0.1323, 0.1849, 0.1355, 0.2103, 0.2344, 0.1761], device='cuda:6'), in_proj_covar=tensor([0.0312, 0.0312, 0.0410, 0.0412, 0.0353, 0.0416, 0.0320, 0.0365], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-28 04:52:20,696 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=164446.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:52:52,713 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.56 vs. limit=5.0 2023-04-28 04:52:55,386 INFO [finetune.py:976] (6/7) Epoch 29, batch 4100, loss[loss=0.187, simple_loss=0.2601, pruned_loss=0.05695, over 4890.00 frames. ], tot_loss[loss=0.164, simple_loss=0.2384, pruned_loss=0.04484, over 955904.85 frames. ], batch size: 35, lr: 2.83e-03, grad_scale: 32.0 2023-04-28 04:53:02,650 INFO [optim.py:369] (6/7) 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 04:53:16,440 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9113, 1.1076, 1.6241, 1.7337, 1.6646, 1.7092, 1.6481, 1.6422], device='cuda:6'), covar=tensor([0.3905, 0.5002, 0.4326, 0.4212, 0.4903, 0.6582, 0.4544, 0.4234], device='cuda:6'), in_proj_covar=tensor([0.0346, 0.0376, 0.0334, 0.0344, 0.0354, 0.0396, 0.0365, 0.0336], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-28 04:53:41,168 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=164511.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:53:54,829 INFO [finetune.py:976] (6/7) Epoch 29, batch 4150, loss[loss=0.1627, simple_loss=0.2394, pruned_loss=0.04299, over 4921.00 frames. ], tot_loss[loss=0.1657, simple_loss=0.2401, pruned_loss=0.0457, over 956758.49 frames. ], batch size: 33, lr: 2.83e-03, grad_scale: 32.0 2023-04-28 04:54:05,127 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5038, 3.4035, 0.8709, 1.8008, 1.9868, 2.3662, 1.9549, 0.9904], device='cuda:6'), covar=tensor([0.1344, 0.0870, 0.1905, 0.1221, 0.0976, 0.1094, 0.1366, 0.1955], device='cuda:6'), in_proj_covar=tensor([0.0117, 0.0236, 0.0135, 0.0119, 0.0131, 0.0152, 0.0116, 0.0118], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-28 04:54:35,908 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=164568.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:54:40,038 INFO [finetune.py:976] (6/7) Epoch 29, batch 4200, loss[loss=0.1362, simple_loss=0.2171, pruned_loss=0.02763, over 4763.00 frames. ], tot_loss[loss=0.1651, simple_loss=0.2398, pruned_loss=0.04516, over 954366.08 frames. ], batch size: 26, lr: 2.83e-03, grad_scale: 32.0 2023-04-28 04:54:41,949 INFO [zipformer.py:1188] (6/7) 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] (6/7) 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 04:54:47,298 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=3.49 vs. limit=5.0 2023-04-28 04:55:14,300 INFO [finetune.py:976] (6/7) Epoch 29, batch 4250, loss[loss=0.1774, simple_loss=0.2563, pruned_loss=0.04922, over 4808.00 frames. ], tot_loss[loss=0.1637, simple_loss=0.2382, pruned_loss=0.04461, over 955808.51 frames. ], batch size: 25, lr: 2.83e-03, grad_scale: 32.0 2023-04-28 04:55:16,892 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=164629.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:55:24,096 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=164639.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 04:55:37,631 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=164658.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:55:48,426 INFO [finetune.py:976] (6/7) Epoch 29, batch 4300, loss[loss=0.1799, simple_loss=0.2466, pruned_loss=0.05662, over 4828.00 frames. ], tot_loss[loss=0.162, simple_loss=0.2356, pruned_loss=0.04422, over 955747.42 frames. ], batch size: 40, lr: 2.83e-03, grad_scale: 32.0 2023-04-28 04:55:50,846 INFO [optim.py:369] (6/7) 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 04:56:16,871 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=164716.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:56:18,759 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=164719.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:56:22,291 INFO [finetune.py:976] (6/7) Epoch 29, batch 4350, loss[loss=0.1785, simple_loss=0.2398, pruned_loss=0.05858, over 4908.00 frames. ], tot_loss[loss=0.1597, simple_loss=0.2326, pruned_loss=0.04345, over 954540.14 frames. ], batch size: 36, lr: 2.83e-03, grad_scale: 32.0 2023-04-28 04:56:34,001 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8039, 1.1478, 1.7450, 2.2190, 1.8599, 1.7078, 1.7385, 1.7170], device='cuda:6'), covar=tensor([0.4317, 0.6964, 0.6217, 0.5669, 0.5913, 0.7601, 0.8234, 0.8808], device='cuda:6'), in_proj_covar=tensor([0.0448, 0.0425, 0.0522, 0.0511, 0.0476, 0.0514, 0.0517, 0.0530], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-28 04:56:36,304 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=164746.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:56:36,317 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=164746.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:57:17,725 INFO [finetune.py:976] (6/7) Epoch 29, batch 4400, loss[loss=0.197, simple_loss=0.2744, pruned_loss=0.05983, over 4932.00 frames. ], tot_loss[loss=0.1602, simple_loss=0.2333, pruned_loss=0.0436, over 953979.08 frames. ], batch size: 38, lr: 2.83e-03, grad_scale: 32.0 2023-04-28 04:57:20,181 INFO [optim.py:369] (6/7) 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 04:57:41,277 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=164794.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:57:42,577 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7075, 2.8119, 2.2414, 2.4796, 2.8155, 2.5202, 3.6369, 2.3485], device='cuda:6'), covar=tensor([0.3456, 0.2265, 0.4156, 0.3000, 0.1776, 0.2260, 0.1466, 0.3650], device='cuda:6'), in_proj_covar=tensor([0.0340, 0.0356, 0.0424, 0.0354, 0.0385, 0.0375, 0.0373, 0.0424], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-28 04:58:01,584 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=164807.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:58:04,464 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=164811.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:58:24,165 INFO [finetune.py:976] (6/7) Epoch 29, batch 4450, loss[loss=0.189, simple_loss=0.2715, pruned_loss=0.05322, over 4937.00 frames. ], tot_loss[loss=0.1631, simple_loss=0.2374, pruned_loss=0.04444, over 953212.97 frames. ], batch size: 33, lr: 2.83e-03, grad_scale: 32.0 2023-04-28 04:58:47,918 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.97 vs. limit=2.0 2023-04-28 04:58:59,956 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=164859.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:59:10,502 INFO [finetune.py:976] (6/7) Epoch 29, batch 4500, loss[loss=0.1705, simple_loss=0.2436, pruned_loss=0.04868, over 4903.00 frames. ], tot_loss[loss=0.1643, simple_loss=0.2389, pruned_loss=0.04483, over 954431.26 frames. ], batch size: 43, lr: 2.83e-03, grad_scale: 32.0 2023-04-28 04:59:18,302 INFO [optim.py:369] (6/7) 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 04:59:58,679 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7666, 4.2114, 1.2751, 2.0313, 2.3499, 2.9551, 2.4922, 1.0692], device='cuda:6'), covar=tensor([0.1344, 0.0795, 0.1712, 0.1245, 0.0970, 0.0905, 0.1253, 0.1911], device='cuda:6'), in_proj_covar=tensor([0.0116, 0.0235, 0.0134, 0.0119, 0.0130, 0.0151, 0.0116, 0.0117], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-28 05:00:20,418 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=164924.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:00:20,954 INFO [finetune.py:976] (6/7) Epoch 29, batch 4550, loss[loss=0.1589, simple_loss=0.239, pruned_loss=0.03937, over 4849.00 frames. ], tot_loss[loss=0.1651, simple_loss=0.2397, pruned_loss=0.0452, over 955008.56 frames. ], batch size: 44, lr: 2.83e-03, grad_scale: 32.0 2023-04-28 05:00:31,459 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.3524, 1.5871, 1.8320, 1.9731, 1.8235, 1.7832, 1.8441, 1.8404], device='cuda:6'), covar=tensor([0.3694, 0.5351, 0.4242, 0.4060, 0.5352, 0.6922, 0.5111, 0.4538], device='cuda:6'), in_proj_covar=tensor([0.0344, 0.0374, 0.0331, 0.0342, 0.0352, 0.0395, 0.0363, 0.0335], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-28 05:00:32,584 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=164934.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 05:01:05,796 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9140, 2.5633, 1.9388, 1.8747, 1.3841, 1.4593, 2.0624, 1.3396], device='cuda:6'), covar=tensor([0.1656, 0.1269, 0.1483, 0.1651, 0.2365, 0.1959, 0.0917, 0.2074], device='cuda:6'), in_proj_covar=tensor([0.0198, 0.0209, 0.0170, 0.0204, 0.0201, 0.0187, 0.0156, 0.0187], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:6') 2023-04-28 05:01:26,333 INFO [finetune.py:976] (6/7) Epoch 29, batch 4600, loss[loss=0.1717, simple_loss=0.2585, pruned_loss=0.04242, over 4874.00 frames. ], tot_loss[loss=0.1637, simple_loss=0.2385, pruned_loss=0.04442, over 955323.28 frames. ], batch size: 34, lr: 2.83e-03, grad_scale: 32.0 2023-04-28 05:01:29,217 INFO [optim.py:369] (6/7) 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 05:02:09,392 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.46 vs. limit=5.0 2023-04-28 05:02:19,692 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=165014.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:02:20,898 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 29, batch 4650, loss[loss=0.1467, simple_loss=0.2179, pruned_loss=0.03772, over 4814.00 frames. ], tot_loss[loss=0.1649, simple_loss=0.2385, pruned_loss=0.04564, over 953632.87 frames. ], batch size: 45, lr: 2.83e-03, grad_scale: 32.0 2023-04-28 05:03:18,782 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=165064.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:03:36,534 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7905, 2.1352, 1.7319, 1.4935, 1.3510, 1.3646, 1.8104, 1.2904], device='cuda:6'), covar=tensor([0.1552, 0.1211, 0.1387, 0.1656, 0.2274, 0.1874, 0.0942, 0.2075], device='cuda:6'), in_proj_covar=tensor([0.0199, 0.0210, 0.0171, 0.0204, 0.0202, 0.0188, 0.0157, 0.0188], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:6') 2023-04-28 05:03:37,009 INFO [finetune.py:976] (6/7) Epoch 29, batch 4700, loss[loss=0.1532, simple_loss=0.2314, pruned_loss=0.03752, over 4859.00 frames. ], tot_loss[loss=0.1629, simple_loss=0.2358, pruned_loss=0.04499, over 954232.59 frames. ], batch size: 49, lr: 2.83e-03, grad_scale: 16.0 2023-04-28 05:03:40,049 INFO [optim.py:369] (6/7) 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 05:03:58,097 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.4037, 3.2431, 0.9688, 1.7163, 1.9197, 2.2559, 1.9585, 0.9651], device='cuda:6'), covar=tensor([0.1701, 0.1385, 0.2258, 0.1444, 0.1220, 0.1269, 0.1520, 0.2063], device='cuda:6'), in_proj_covar=tensor([0.0116, 0.0236, 0.0135, 0.0120, 0.0131, 0.0152, 0.0116, 0.0117], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-28 05:04:09,168 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=165102.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:04:09,376 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-04-28 05:04:18,524 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-04-28 05:04:41,330 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=165123.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:04:42,454 INFO [finetune.py:976] (6/7) Epoch 29, batch 4750, loss[loss=0.153, simple_loss=0.2039, pruned_loss=0.05106, over 4250.00 frames. ], tot_loss[loss=0.161, simple_loss=0.2334, pruned_loss=0.04436, over 953404.33 frames. ], batch size: 18, lr: 2.83e-03, grad_scale: 16.0 2023-04-28 05:04:43,245 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7109, 0.9941, 1.6548, 1.9617, 1.7331, 1.6334, 1.6766, 1.6967], device='cuda:6'), covar=tensor([0.5134, 0.8004, 0.6967, 0.7981, 0.6689, 0.9631, 0.9166, 1.0512], device='cuda:6'), in_proj_covar=tensor([0.0445, 0.0424, 0.0519, 0.0509, 0.0474, 0.0514, 0.0515, 0.0530], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-28 05:05:25,306 INFO [finetune.py:976] (6/7) Epoch 29, batch 4800, loss[loss=0.2164, simple_loss=0.293, pruned_loss=0.06994, over 4842.00 frames. ], tot_loss[loss=0.1637, simple_loss=0.2365, pruned_loss=0.04547, over 955587.67 frames. ], batch size: 47, lr: 2.82e-03, grad_scale: 16.0 2023-04-28 05:05:28,757 INFO [optim.py:369] (6/7) 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 05:05:31,322 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=165184.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:05:48,804 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0516, 2.6137, 1.1067, 1.3755, 1.8878, 1.2631, 3.4194, 1.7029], device='cuda:6'), covar=tensor([0.0680, 0.0690, 0.0747, 0.1251, 0.0520, 0.0984, 0.0210, 0.0602], device='cuda:6'), in_proj_covar=tensor([0.0050, 0.0064, 0.0047, 0.0046, 0.0049, 0.0051, 0.0073, 0.0051], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:6') 2023-04-28 05:05:57,703 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=165224.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:05:58,221 INFO [finetune.py:976] (6/7) Epoch 29, batch 4850, loss[loss=0.1559, simple_loss=0.2209, pruned_loss=0.04545, over 4435.00 frames. ], tot_loss[loss=0.1663, simple_loss=0.2399, pruned_loss=0.04638, over 954261.20 frames. ], batch size: 19, lr: 2.82e-03, grad_scale: 16.0 2023-04-28 05:06:05,341 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=165234.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 05:06:15,007 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.1029, 0.6677, 0.9140, 0.8412, 1.2408, 0.9845, 0.8394, 0.9769], device='cuda:6'), covar=tensor([0.1820, 0.2021, 0.2438, 0.1869, 0.1252, 0.1782, 0.1959, 0.2708], device='cuda:6'), in_proj_covar=tensor([0.0312, 0.0307, 0.0348, 0.0286, 0.0324, 0.0303, 0.0299, 0.0375], device='cuda:6'), out_proj_covar=tensor([6.3530e-05, 6.2858e-05, 7.2719e-05, 5.7166e-05, 6.6027e-05, 6.2844e-05, 6.1752e-05, 7.9230e-05], device='cuda:6') 2023-04-28 05:06:29,992 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 29, batch 4900, loss[loss=0.1843, simple_loss=0.2653, pruned_loss=0.05169, over 4798.00 frames. ], tot_loss[loss=0.1679, simple_loss=0.2421, pruned_loss=0.04685, over 955496.18 frames. ], batch size: 51, lr: 2.82e-03, grad_scale: 16.0 2023-04-28 05:06:35,317 INFO [optim.py:369] (6/7) 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 05:06:37,066 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=165282.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:06:57,503 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=165314.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:07:04,968 INFO [finetune.py:976] (6/7) Epoch 29, batch 4950, loss[loss=0.1617, simple_loss=0.2411, pruned_loss=0.04113, over 4863.00 frames. ], tot_loss[loss=0.1677, simple_loss=0.2423, pruned_loss=0.04658, over 956151.91 frames. ], batch size: 34, lr: 2.82e-03, grad_scale: 16.0 2023-04-28 05:07:29,753 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=165362.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:07:38,074 INFO [finetune.py:976] (6/7) Epoch 29, batch 5000, loss[loss=0.1404, simple_loss=0.2075, pruned_loss=0.03667, over 4363.00 frames. ], tot_loss[loss=0.1652, simple_loss=0.2395, pruned_loss=0.04549, over 954600.99 frames. ], batch size: 19, lr: 2.82e-03, grad_scale: 16.0 2023-04-28 05:07:41,608 INFO [optim.py:369] (6/7) 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 05:07:57,200 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=165402.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:08:08,235 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.64 vs. limit=2.0 2023-04-28 05:08:12,132 INFO [finetune.py:976] (6/7) Epoch 29, batch 5050, loss[loss=0.1588, simple_loss=0.2319, pruned_loss=0.04287, over 4933.00 frames. ], tot_loss[loss=0.1625, simple_loss=0.2364, pruned_loss=0.04429, over 954659.28 frames. ], batch size: 38, lr: 2.82e-03, grad_scale: 16.0 2023-04-28 05:08:29,766 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=165450.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:08:29,793 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.2610, 1.2010, 3.8481, 3.6186, 3.4549, 3.7675, 3.7923, 3.4266], device='cuda:6'), covar=tensor([0.7823, 0.6147, 0.1344, 0.2025, 0.1329, 0.1715, 0.1418, 0.1637], device='cuda:6'), in_proj_covar=tensor([0.0312, 0.0312, 0.0410, 0.0414, 0.0354, 0.0417, 0.0320, 0.0366], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-28 05:08:40,114 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5993, 1.7393, 1.4646, 1.1115, 1.1863, 1.2241, 1.5030, 1.0791], device='cuda:6'), covar=tensor([0.1813, 0.1236, 0.1498, 0.1698, 0.2345, 0.1909, 0.1041, 0.2086], device='cuda:6'), in_proj_covar=tensor([0.0199, 0.0210, 0.0171, 0.0204, 0.0202, 0.0187, 0.0157, 0.0188], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:6') 2023-04-28 05:08:45,057 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.2336, 2.7078, 2.2599, 2.6358, 1.9357, 2.3323, 2.2520, 1.7184], device='cuda:6'), covar=tensor([0.1887, 0.1132, 0.0817, 0.1075, 0.3146, 0.1104, 0.1848, 0.2752], device='cuda:6'), in_proj_covar=tensor([0.0280, 0.0299, 0.0215, 0.0274, 0.0311, 0.0252, 0.0245, 0.0261], device='cuda:6'), out_proj_covar=tensor([1.1156e-04, 1.1750e-04, 8.4254e-05, 1.0739e-04, 1.2507e-04, 9.8960e-05, 9.8693e-05, 1.0268e-04], device='cuda:6') 2023-04-28 05:08:45,560 INFO [finetune.py:976] (6/7) Epoch 29, batch 5100, loss[loss=0.1323, simple_loss=0.2107, pruned_loss=0.02695, over 4834.00 frames. ], tot_loss[loss=0.1593, simple_loss=0.233, pruned_loss=0.04286, over 957478.72 frames. ], batch size: 30, lr: 2.82e-03, grad_scale: 16.0 2023-04-28 05:08:48,563 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=165479.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:08:49,076 INFO [optim.py:369] (6/7) 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 05:09:36,570 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6763, 2.0553, 1.7943, 1.9815, 1.5033, 1.8208, 1.7573, 1.4195], device='cuda:6'), covar=tensor([0.1515, 0.0925, 0.0717, 0.0899, 0.3060, 0.0933, 0.1431, 0.1914], device='cuda:6'), in_proj_covar=tensor([0.0281, 0.0300, 0.0215, 0.0274, 0.0311, 0.0253, 0.0246, 0.0262], device='cuda:6'), out_proj_covar=tensor([1.1189e-04, 1.1777e-04, 8.4529e-05, 1.0777e-04, 1.2522e-04, 9.9178e-05, 9.9053e-05, 1.0299e-04], device='cuda:6') 2023-04-28 05:09:48,072 INFO [finetune.py:976] (6/7) Epoch 29, batch 5150, loss[loss=0.1467, simple_loss=0.2311, pruned_loss=0.03117, over 4813.00 frames. ], tot_loss[loss=0.1609, simple_loss=0.2341, pruned_loss=0.04382, over 955282.96 frames. ], batch size: 40, lr: 2.82e-03, grad_scale: 16.0 2023-04-28 05:10:21,722 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=165551.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:10:52,971 INFO [finetune.py:976] (6/7) Epoch 29, batch 5200, loss[loss=0.2101, simple_loss=0.2686, pruned_loss=0.07581, over 4842.00 frames. ], tot_loss[loss=0.1637, simple_loss=0.2376, pruned_loss=0.04487, over 955129.77 frames. ], batch size: 49, lr: 2.82e-03, grad_scale: 16.0 2023-04-28 05:10:55,998 INFO [optim.py:369] (6/7) 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 05:11:45,803 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=165612.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:11:59,708 INFO [finetune.py:976] (6/7) Epoch 29, batch 5250, loss[loss=0.1677, simple_loss=0.2404, pruned_loss=0.04747, over 4779.00 frames. ], tot_loss[loss=0.1649, simple_loss=0.2392, pruned_loss=0.04531, over 955107.60 frames. ], batch size: 26, lr: 2.82e-03, grad_scale: 16.0 2023-04-28 05:12:12,050 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5415, 1.3083, 3.9464, 3.7348, 3.4611, 3.7178, 3.6935, 3.5306], device='cuda:6'), covar=tensor([0.7263, 0.5713, 0.1143, 0.1631, 0.1166, 0.1697, 0.2252, 0.1491], device='cuda:6'), in_proj_covar=tensor([0.0309, 0.0309, 0.0408, 0.0412, 0.0351, 0.0414, 0.0318, 0.0363], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-28 05:13:07,359 INFO [finetune.py:976] (6/7) Epoch 29, batch 5300, loss[loss=0.1659, simple_loss=0.248, pruned_loss=0.04184, over 4738.00 frames. ], tot_loss[loss=0.1655, simple_loss=0.2398, pruned_loss=0.04564, over 953340.12 frames. ], batch size: 27, lr: 2.82e-03, grad_scale: 16.0 2023-04-28 05:13:16,015 INFO [optim.py:369] (6/7) 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 05:13:47,646 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5823, 0.6579, 1.4530, 1.8988, 1.6244, 1.4855, 1.5122, 1.5228], device='cuda:6'), covar=tensor([0.4219, 0.6761, 0.6188, 0.6280, 0.5663, 0.7716, 0.7683, 0.8847], device='cuda:6'), in_proj_covar=tensor([0.0447, 0.0425, 0.0520, 0.0509, 0.0474, 0.0514, 0.0515, 0.0530], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-28 05:14:12,924 INFO [finetune.py:976] (6/7) Epoch 29, batch 5350, loss[loss=0.1502, simple_loss=0.2278, pruned_loss=0.03632, over 4822.00 frames. ], tot_loss[loss=0.1659, simple_loss=0.2405, pruned_loss=0.04568, over 952040.24 frames. ], batch size: 33, lr: 2.82e-03, grad_scale: 16.0 2023-04-28 05:14:48,690 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=165750.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:15:21,308 INFO [finetune.py:976] (6/7) Epoch 29, batch 5400, loss[loss=0.1621, simple_loss=0.2192, pruned_loss=0.05249, over 4001.00 frames. ], tot_loss[loss=0.1646, simple_loss=0.2385, pruned_loss=0.04532, over 952331.14 frames. ], batch size: 17, lr: 2.82e-03, grad_scale: 16.0 2023-04-28 05:15:24,330 INFO [zipformer.py:1188] (6/7) 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] (6/7) 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 05:15:55,109 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=165803.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:16:07,213 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=165811.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:16:21,414 INFO [finetune.py:976] (6/7) Epoch 29, batch 5450, loss[loss=0.123, simple_loss=0.2006, pruned_loss=0.02272, over 4789.00 frames. ], tot_loss[loss=0.1626, simple_loss=0.2357, pruned_loss=0.04474, over 953552.03 frames. ], batch size: 29, lr: 2.82e-03, grad_scale: 16.0 2023-04-28 05:16:22,707 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=165827.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:16:48,205 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=165864.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:16:55,273 INFO [finetune.py:976] (6/7) Epoch 29, batch 5500, loss[loss=0.137, simple_loss=0.2151, pruned_loss=0.02943, over 4868.00 frames. ], tot_loss[loss=0.1602, simple_loss=0.2326, pruned_loss=0.04387, over 954507.41 frames. ], batch size: 34, lr: 2.82e-03, grad_scale: 16.0 2023-04-28 05:16:58,257 INFO [optim.py:369] (6/7) 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 05:17:16,290 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=165907.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:17:28,020 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.3552, 1.3263, 1.6912, 1.5943, 1.2183, 1.1495, 1.3770, 0.8700], device='cuda:6'), covar=tensor([0.0460, 0.0537, 0.0357, 0.0504, 0.0760, 0.1008, 0.0488, 0.0512], device='cuda:6'), in_proj_covar=tensor([0.0070, 0.0066, 0.0064, 0.0068, 0.0075, 0.0094, 0.0071, 0.0061], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-28 05:17:29,078 INFO [finetune.py:976] (6/7) Epoch 29, batch 5550, loss[loss=0.1819, simple_loss=0.2624, pruned_loss=0.05075, over 4900.00 frames. ], tot_loss[loss=0.1632, simple_loss=0.2356, pruned_loss=0.04534, over 953766.50 frames. ], batch size: 43, lr: 2.82e-03, grad_scale: 16.0 2023-04-28 05:18:17,438 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7268, 2.3401, 1.8037, 1.8409, 1.2993, 1.3352, 1.9549, 1.1935], device='cuda:6'), covar=tensor([0.1587, 0.1302, 0.1410, 0.1608, 0.2253, 0.1984, 0.0898, 0.2245], device='cuda:6'), in_proj_covar=tensor([0.0199, 0.0211, 0.0171, 0.0205, 0.0201, 0.0187, 0.0157, 0.0188], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:6') 2023-04-28 05:18:28,708 INFO [finetune.py:976] (6/7) Epoch 29, batch 5600, loss[loss=0.1779, simple_loss=0.2546, pruned_loss=0.05058, over 4750.00 frames. ], tot_loss[loss=0.1635, simple_loss=0.2368, pruned_loss=0.04506, over 952946.60 frames. ], batch size: 27, lr: 2.82e-03, grad_scale: 16.0 2023-04-28 05:18:35,762 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.49 vs. limit=2.0 2023-04-28 05:18:37,368 INFO [optim.py:369] (6/7) 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 05:19:30,412 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.4882, 1.3296, 4.1958, 3.9641, 3.7197, 4.0259, 4.0429, 3.7129], device='cuda:6'), covar=tensor([0.7260, 0.6009, 0.1143, 0.1760, 0.1138, 0.2312, 0.1337, 0.1662], device='cuda:6'), in_proj_covar=tensor([0.0309, 0.0309, 0.0407, 0.0410, 0.0351, 0.0414, 0.0316, 0.0363], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-28 05:19:32,709 INFO [finetune.py:976] (6/7) Epoch 29, batch 5650, loss[loss=0.1311, simple_loss=0.2169, pruned_loss=0.02265, over 4768.00 frames. ], tot_loss[loss=0.1655, simple_loss=0.2395, pruned_loss=0.04573, over 953505.19 frames. ], batch size: 26, lr: 2.82e-03, grad_scale: 16.0 2023-04-28 05:20:11,624 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=166054.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 05:20:33,889 INFO [finetune.py:976] (6/7) Epoch 29, batch 5700, loss[loss=0.1476, simple_loss=0.2045, pruned_loss=0.04534, over 4422.00 frames. ], tot_loss[loss=0.1625, simple_loss=0.2358, pruned_loss=0.04455, over 938204.26 frames. ], batch size: 19, lr: 2.82e-03, grad_scale: 16.0 2023-04-28 05:20:42,338 INFO [optim.py:369] (6/7) 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 05:21:19,172 INFO [finetune.py:976] (6/7) Epoch 30, batch 0, loss[loss=0.2139, simple_loss=0.2921, pruned_loss=0.06785, over 4926.00 frames. ], tot_loss[loss=0.2139, simple_loss=0.2921, pruned_loss=0.06785, over 4926.00 frames. ], batch size: 42, lr: 2.82e-03, grad_scale: 16.0 2023-04-28 05:21:19,172 INFO [finetune.py:1001] (6/7) Computing validation loss 2023-04-28 05:21:25,980 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7994, 1.0907, 1.7545, 2.2977, 1.8706, 1.6976, 1.7129, 1.7400], device='cuda:6'), covar=tensor([0.4375, 0.7151, 0.6377, 0.5354, 0.6026, 0.7866, 0.8177, 0.9038], device='cuda:6'), in_proj_covar=tensor([0.0448, 0.0427, 0.0523, 0.0511, 0.0477, 0.0517, 0.0519, 0.0533], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-28 05:21:31,238 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.3230, 1.5790, 1.8843, 2.0327, 1.9836, 2.0360, 1.8866, 1.9213], device='cuda:6'), covar=tensor([0.4106, 0.5715, 0.4630, 0.4408, 0.5236, 0.6836, 0.5331, 0.4927], device='cuda:6'), in_proj_covar=tensor([0.0347, 0.0377, 0.0333, 0.0346, 0.0354, 0.0396, 0.0365, 0.0336], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-28 05:21:34,681 INFO [finetune.py:1010] (6/7) Epoch 30, validation: loss=0.1551, simple_loss=0.2236, pruned_loss=0.04334, over 2265189.00 frames. 2023-04-28 05:21:34,681 INFO [finetune.py:1011] (6/7) Maximum memory allocated so far is 6435MB 2023-04-28 05:21:37,611 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=166106.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:21:48,579 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=166115.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 05:22:23,784 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.59 vs. limit=2.0 2023-04-28 05:22:31,246 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 30, batch 50, loss[loss=0.1644, simple_loss=0.244, pruned_loss=0.04243, over 4859.00 frames. ], tot_loss[loss=0.172, simple_loss=0.2462, pruned_loss=0.04896, over 215994.04 frames. ], batch size: 34, lr: 2.81e-03, grad_scale: 16.0 2023-04-28 05:22:42,115 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-04-28 05:22:42,575 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.66 vs. limit=2.0 2023-04-28 05:22:51,879 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=166159.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:23:14,384 INFO [optim.py:369] (6/7) 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 05:23:44,637 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9516, 1.6527, 1.8150, 2.1917, 2.2652, 1.8134, 1.5340, 1.9902], device='cuda:6'), covar=tensor([0.0645, 0.1096, 0.0649, 0.0477, 0.0519, 0.0734, 0.0706, 0.0477], device='cuda:6'), in_proj_covar=tensor([0.0183, 0.0202, 0.0183, 0.0170, 0.0177, 0.0177, 0.0149, 0.0176], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-28 05:23:47,362 INFO [finetune.py:976] (6/7) Epoch 30, batch 100, loss[loss=0.1593, simple_loss=0.2213, pruned_loss=0.04867, over 4739.00 frames. ], tot_loss[loss=0.1628, simple_loss=0.2359, pruned_loss=0.04485, over 380580.83 frames. ], batch size: 59, lr: 2.81e-03, grad_scale: 16.0 2023-04-28 05:23:50,920 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=166207.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:23:56,824 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=166208.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:24:41,850 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.11 vs. limit=2.0 2023-04-28 05:24:53,828 INFO [finetune.py:976] (6/7) Epoch 30, batch 150, loss[loss=0.1687, simple_loss=0.245, pruned_loss=0.04616, over 4800.00 frames. ], tot_loss[loss=0.1603, simple_loss=0.2325, pruned_loss=0.04402, over 507315.15 frames. ], batch size: 29, lr: 2.81e-03, grad_scale: 16.0 2023-04-28 05:24:55,629 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=166255.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:25:28,829 INFO [optim.py:369] (6/7) 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] (6/7) Epoch 30, batch 200, loss[loss=0.1591, simple_loss=0.2388, pruned_loss=0.03975, over 4897.00 frames. ], tot_loss[loss=0.1603, simple_loss=0.2316, pruned_loss=0.04449, over 605477.26 frames. ], batch size: 43, lr: 2.81e-03, grad_scale: 16.0 2023-04-28 05:26:07,266 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=166307.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:26:40,554 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.53 vs. limit=2.0 2023-04-28 05:27:02,969 INFO [finetune.py:976] (6/7) Epoch 30, batch 250, loss[loss=0.1624, simple_loss=0.2413, pruned_loss=0.04179, over 4803.00 frames. ], tot_loss[loss=0.1632, simple_loss=0.2352, pruned_loss=0.04557, over 684001.82 frames. ], batch size: 41, lr: 2.81e-03, grad_scale: 16.0 2023-04-28 05:27:08,585 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7506, 1.3261, 1.4786, 1.5610, 1.8850, 1.5233, 1.3385, 1.4568], device='cuda:6'), covar=tensor([0.1785, 0.1774, 0.1931, 0.1384, 0.0963, 0.1610, 0.2101, 0.2394], device='cuda:6'), in_proj_covar=tensor([0.0316, 0.0310, 0.0350, 0.0288, 0.0327, 0.0307, 0.0302, 0.0379], device='cuda:6'), out_proj_covar=tensor([6.4388e-05, 6.3513e-05, 7.3268e-05, 5.7472e-05, 6.6625e-05, 6.3545e-05, 6.2383e-05, 8.0131e-05], device='cuda:6') 2023-04-28 05:27:19,844 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=166368.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:27:20,489 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8006, 1.7150, 2.1888, 2.3232, 1.5873, 1.4946, 1.7952, 0.9269], device='cuda:6'), covar=tensor([0.0679, 0.0674, 0.0385, 0.0728, 0.0726, 0.1030, 0.0680, 0.0726], device='cuda:6'), in_proj_covar=tensor([0.0071, 0.0067, 0.0065, 0.0069, 0.0076, 0.0095, 0.0072, 0.0062], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-28 05:27:38,573 INFO [optim.py:369] (6/7) 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 05:28:02,278 INFO [finetune.py:976] (6/7) Epoch 30, batch 300, loss[loss=0.1148, simple_loss=0.1916, pruned_loss=0.01906, over 4702.00 frames. ], tot_loss[loss=0.1662, simple_loss=0.2393, pruned_loss=0.04652, over 742985.85 frames. ], batch size: 23, lr: 2.81e-03, grad_scale: 16.0 2023-04-28 05:28:07,539 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=166406.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:28:09,972 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=166410.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 05:29:01,976 INFO [finetune.py:976] (6/7) Epoch 30, batch 350, loss[loss=0.1486, simple_loss=0.2321, pruned_loss=0.03252, over 4861.00 frames. ], tot_loss[loss=0.1667, simple_loss=0.241, pruned_loss=0.04623, over 792415.24 frames. ], batch size: 31, lr: 2.81e-03, grad_scale: 16.0 2023-04-28 05:29:08,237 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=166454.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:29:11,890 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=166459.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:29:31,635 INFO [optim.py:369] (6/7) 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 05:29:32,383 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=166481.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:29:46,147 INFO [finetune.py:976] (6/7) Epoch 30, batch 400, loss[loss=0.1477, simple_loss=0.2427, pruned_loss=0.02631, over 4754.00 frames. ], tot_loss[loss=0.1661, simple_loss=0.2411, pruned_loss=0.04553, over 828252.98 frames. ], batch size: 28, lr: 2.81e-03, grad_scale: 16.0 2023-04-28 05:29:46,224 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=166503.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:29:49,557 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=166507.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:29:50,822 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=166509.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:29:50,898 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-04-28 05:30:06,153 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=166531.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:30:13,251 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 30, batch 450, loss[loss=0.1579, simple_loss=0.2371, pruned_loss=0.03936, over 4750.00 frames. ], tot_loss[loss=0.1659, simple_loss=0.2408, pruned_loss=0.04547, over 856702.28 frames. ], batch size: 54, lr: 2.81e-03, grad_scale: 16.0 2023-04-28 05:30:31,311 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=166570.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:30:38,210 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.1724, 1.6462, 2.0316, 2.1566, 1.9763, 1.6124, 1.1328, 1.8135], device='cuda:6'), covar=tensor([0.2452, 0.2627, 0.1474, 0.1790, 0.2170, 0.2267, 0.3843, 0.1441], device='cuda:6'), in_proj_covar=tensor([0.0293, 0.0248, 0.0229, 0.0316, 0.0223, 0.0238, 0.0230, 0.0186], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:6') 2023-04-28 05:30:39,258 INFO [optim.py:369] (6/7) 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 05:30:47,088 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=166592.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:30:53,782 INFO [finetune.py:976] (6/7) Epoch 30, batch 500, loss[loss=0.1548, simple_loss=0.2326, pruned_loss=0.03854, over 4791.00 frames. ], tot_loss[loss=0.1643, simple_loss=0.2383, pruned_loss=0.04515, over 880808.30 frames. ], batch size: 29, lr: 2.81e-03, grad_scale: 16.0 2023-04-28 05:31:07,087 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5786, 1.2033, 4.2467, 4.0144, 3.6744, 4.0120, 3.9074, 3.7907], device='cuda:6'), covar=tensor([0.7302, 0.5929, 0.0978, 0.1523, 0.1101, 0.1941, 0.2017, 0.1502], device='cuda:6'), in_proj_covar=tensor([0.0312, 0.0311, 0.0409, 0.0412, 0.0353, 0.0416, 0.0319, 0.0365], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-28 05:31:27,593 INFO [finetune.py:976] (6/7) Epoch 30, batch 550, loss[loss=0.1455, simple_loss=0.2064, pruned_loss=0.04231, over 4695.00 frames. ], tot_loss[loss=0.1612, simple_loss=0.2345, pruned_loss=0.04388, over 896995.75 frames. ], batch size: 23, lr: 2.81e-03, grad_scale: 16.0 2023-04-28 05:31:34,204 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=166663.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:31:45,968 INFO [optim.py:369] (6/7) 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 05:32:01,359 INFO [finetune.py:976] (6/7) Epoch 30, batch 600, loss[loss=0.1908, simple_loss=0.259, pruned_loss=0.06135, over 4797.00 frames. ], tot_loss[loss=0.1636, simple_loss=0.2366, pruned_loss=0.04528, over 910252.80 frames. ], batch size: 29, lr: 2.81e-03, grad_scale: 16.0 2023-04-28 05:32:05,747 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=166710.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 05:32:07,141 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.62 vs. limit=2.0 2023-04-28 05:32:15,306 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5374, 1.8612, 1.7396, 2.1131, 2.0814, 2.2283, 1.8319, 3.5977], device='cuda:6'), covar=tensor([0.0518, 0.0644, 0.0658, 0.0947, 0.0493, 0.0548, 0.0616, 0.0177], device='cuda:6'), in_proj_covar=tensor([0.0037, 0.0038, 0.0039, 0.0043, 0.0040, 0.0037, 0.0038, 0.0054], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0013, 0.0013, 0.0015, 0.0014, 0.0013, 0.0013, 0.0016], device='cuda:6') 2023-04-28 05:32:23,555 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 30, batch 650, loss[loss=0.1825, simple_loss=0.2606, pruned_loss=0.05222, over 4809.00 frames. ], tot_loss[loss=0.165, simple_loss=0.2387, pruned_loss=0.04561, over 916313.08 frames. ], batch size: 45, lr: 2.81e-03, grad_scale: 16.0 2023-04-28 05:32:37,943 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=166758.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 05:32:47,098 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=166764.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:32:55,015 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6186, 0.7522, 1.5345, 1.9691, 1.6761, 1.5030, 1.5289, 1.5630], device='cuda:6'), covar=tensor([0.4326, 0.6582, 0.5772, 0.5714, 0.5909, 0.7718, 0.7532, 0.8748], device='cuda:6'), in_proj_covar=tensor([0.0449, 0.0426, 0.0523, 0.0510, 0.0476, 0.0516, 0.0519, 0.0532], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-28 05:33:07,156 INFO [optim.py:369] (6/7) 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 05:33:29,724 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=166796.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:33:39,479 INFO [finetune.py:976] (6/7) Epoch 30, batch 700, loss[loss=0.1316, simple_loss=0.2095, pruned_loss=0.02682, over 4746.00 frames. ], tot_loss[loss=0.1646, simple_loss=0.2392, pruned_loss=0.04499, over 924908.61 frames. ], batch size: 27, lr: 2.81e-03, grad_scale: 16.0 2023-04-28 05:33:39,605 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=166803.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:34:03,940 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=166825.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:34:22,692 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=166837.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:34:41,941 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=166851.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:34:43,122 INFO [finetune.py:976] (6/7) Epoch 30, batch 750, loss[loss=0.1718, simple_loss=0.2509, pruned_loss=0.04634, over 4824.00 frames. ], tot_loss[loss=0.1653, simple_loss=0.2404, pruned_loss=0.04504, over 932835.26 frames. ], batch size: 33, lr: 2.81e-03, grad_scale: 16.0 2023-04-28 05:34:51,932 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=166865.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:34:52,607 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([0.8702, 1.7746, 1.5439, 1.4494, 1.6543, 1.3480, 2.1161, 1.2664], device='cuda:6'), covar=tensor([0.3531, 0.1722, 0.4686, 0.2697, 0.2106, 0.2685, 0.1706, 0.5290], device='cuda:6'), in_proj_covar=tensor([0.0342, 0.0356, 0.0428, 0.0354, 0.0388, 0.0377, 0.0373, 0.0428], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-28 05:35:11,847 INFO [optim.py:369] (6/7) 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 05:35:14,629 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2023-04-28 05:35:22,365 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=166887.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:35:44,638 INFO [finetune.py:976] (6/7) Epoch 30, batch 800, loss[loss=0.154, simple_loss=0.2334, pruned_loss=0.03733, over 4868.00 frames. ], tot_loss[loss=0.1636, simple_loss=0.239, pruned_loss=0.04405, over 937790.39 frames. ], batch size: 34, lr: 2.81e-03, grad_scale: 16.0 2023-04-28 05:36:17,524 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=166929.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:36:49,760 INFO [finetune.py:976] (6/7) Epoch 30, batch 850, loss[loss=0.1437, simple_loss=0.2161, pruned_loss=0.0356, over 4795.00 frames. ], tot_loss[loss=0.1626, simple_loss=0.2375, pruned_loss=0.0439, over 941825.51 frames. ], batch size: 29, lr: 2.81e-03, grad_scale: 16.0 2023-04-28 05:36:59,944 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=166963.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:37:20,969 INFO [optim.py:369] (6/7) 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 05:37:37,952 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=166990.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:37:47,450 INFO [finetune.py:976] (6/7) Epoch 30, batch 900, loss[loss=0.1645, simple_loss=0.2211, pruned_loss=0.05391, over 4819.00 frames. ], tot_loss[loss=0.161, simple_loss=0.2351, pruned_loss=0.04347, over 945276.38 frames. ], batch size: 25, lr: 2.81e-03, grad_scale: 16.0 2023-04-28 05:37:52,367 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=167011.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:37:55,971 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6047, 1.0549, 1.3724, 1.2974, 1.6785, 1.4250, 1.1772, 1.3373], device='cuda:6'), covar=tensor([0.1845, 0.1709, 0.2085, 0.1536, 0.1101, 0.1508, 0.1918, 0.2491], device='cuda:6'), in_proj_covar=tensor([0.0315, 0.0309, 0.0350, 0.0287, 0.0325, 0.0305, 0.0300, 0.0378], device='cuda:6'), out_proj_covar=tensor([6.4159e-05, 6.3124e-05, 7.3263e-05, 5.7212e-05, 6.6129e-05, 6.3228e-05, 6.1918e-05, 7.9760e-05], device='cuda:6') 2023-04-28 05:38:21,334 INFO [finetune.py:976] (6/7) Epoch 30, batch 950, loss[loss=0.1743, simple_loss=0.2355, pruned_loss=0.05657, over 4783.00 frames. ], tot_loss[loss=0.1607, simple_loss=0.2343, pruned_loss=0.04356, over 948127.12 frames. ], batch size: 26, lr: 2.81e-03, grad_scale: 32.0 2023-04-28 05:38:31,162 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=3.99 vs. limit=5.0 2023-04-28 05:38:38,159 INFO [optim.py:369] (6/7) 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 05:38:45,980 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=167091.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:38:52,430 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=3.85 vs. limit=5.0 2023-04-28 05:38:52,933 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 30, batch 1000, loss[loss=0.1948, simple_loss=0.2712, pruned_loss=0.05913, over 4903.00 frames. ], tot_loss[loss=0.1632, simple_loss=0.2368, pruned_loss=0.0448, over 950873.46 frames. ], batch size: 43, lr: 2.81e-03, grad_scale: 32.0 2023-04-28 05:38:55,894 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=167104.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:39:06,124 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=167120.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:39:14,889 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.68 vs. limit=5.0 2023-04-28 05:39:17,578 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=167137.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:39:23,092 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-04-28 05:39:28,663 INFO [finetune.py:976] (6/7) Epoch 30, batch 1050, loss[loss=0.1394, simple_loss=0.2165, pruned_loss=0.03116, over 4911.00 frames. ], tot_loss[loss=0.164, simple_loss=0.2383, pruned_loss=0.0449, over 952953.47 frames. ], batch size: 36, lr: 2.81e-03, grad_scale: 32.0 2023-04-28 05:39:31,677 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6262, 3.5973, 0.9467, 1.9532, 2.1654, 2.5864, 2.0114, 0.9146], device='cuda:6'), covar=tensor([0.1381, 0.0915, 0.2071, 0.1245, 0.1019, 0.1065, 0.1628, 0.2272], device='cuda:6'), in_proj_covar=tensor([0.0118, 0.0239, 0.0137, 0.0121, 0.0132, 0.0153, 0.0118, 0.0119], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-28 05:39:34,165 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=167161.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:39:36,567 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=167165.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:39:36,597 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=167165.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:39:46,135 INFO [optim.py:369] (6/7) 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] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=167185.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:39:50,853 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=167187.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:40:01,671 INFO [finetune.py:976] (6/7) Epoch 30, batch 1100, loss[loss=0.2233, simple_loss=0.2951, pruned_loss=0.07572, over 4853.00 frames. ], tot_loss[loss=0.1651, simple_loss=0.2398, pruned_loss=0.04525, over 953784.11 frames. ], batch size: 44, lr: 2.81e-03, grad_scale: 32.0 2023-04-28 05:40:07,660 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.3662, 2.8562, 0.9490, 1.5183, 2.3292, 1.4617, 4.0930, 1.9375], device='cuda:6'), covar=tensor([0.0677, 0.0828, 0.0887, 0.1254, 0.0491, 0.0991, 0.0189, 0.0604], device='cuda:6'), in_proj_covar=tensor([0.0051, 0.0064, 0.0047, 0.0046, 0.0049, 0.0051, 0.0073, 0.0051], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:6') 2023-04-28 05:40:08,809 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=167213.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:40:22,665 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 30, batch 1150, loss[loss=0.1846, simple_loss=0.2689, pruned_loss=0.05019, over 4849.00 frames. ], tot_loss[loss=0.1664, simple_loss=0.2409, pruned_loss=0.046, over 953631.06 frames. ], batch size: 44, lr: 2.81e-03, grad_scale: 32.0 2023-04-28 05:41:03,287 INFO [optim.py:369] (6/7) 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 05:41:11,488 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=167285.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:41:33,521 INFO [finetune.py:976] (6/7) Epoch 30, batch 1200, loss[loss=0.1631, simple_loss=0.2373, pruned_loss=0.04448, over 4935.00 frames. ], tot_loss[loss=0.1663, simple_loss=0.2405, pruned_loss=0.04603, over 953844.25 frames. ], batch size: 33, lr: 2.81e-03, grad_scale: 32.0 2023-04-28 05:42:37,938 INFO [finetune.py:976] (6/7) Epoch 30, batch 1250, loss[loss=0.154, simple_loss=0.2199, pruned_loss=0.04407, over 4734.00 frames. ], tot_loss[loss=0.1641, simple_loss=0.2378, pruned_loss=0.04517, over 954951.12 frames. ], batch size: 54, lr: 2.81e-03, grad_scale: 32.0 2023-04-28 05:42:56,511 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.0391, 2.5349, 0.8375, 1.4967, 1.5340, 1.8585, 1.6706, 0.8722], device='cuda:6'), covar=tensor([0.1632, 0.1255, 0.1826, 0.1308, 0.1217, 0.1009, 0.1654, 0.1694], device='cuda:6'), in_proj_covar=tensor([0.0118, 0.0240, 0.0137, 0.0121, 0.0133, 0.0154, 0.0119, 0.0119], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-28 05:43:19,229 INFO [optim.py:369] (6/7) 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 05:43:32,137 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=167391.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:43:42,858 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.1091, 2.5941, 1.0268, 1.4052, 1.9312, 1.2220, 3.3790, 1.5642], device='cuda:6'), covar=tensor([0.0699, 0.0724, 0.0827, 0.1157, 0.0476, 0.0979, 0.0210, 0.0630], device='cuda:6'), in_proj_covar=tensor([0.0051, 0.0064, 0.0047, 0.0046, 0.0049, 0.0051, 0.0073, 0.0050], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:6') 2023-04-28 05:43:45,346 INFO [finetune.py:976] (6/7) Epoch 30, batch 1300, loss[loss=0.1342, simple_loss=0.2039, pruned_loss=0.03226, over 4784.00 frames. ], tot_loss[loss=0.1615, simple_loss=0.2346, pruned_loss=0.04423, over 954848.44 frames. ], batch size: 29, lr: 2.81e-03, grad_scale: 32.0 2023-04-28 05:44:12,871 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=167420.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:44:25,764 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=167432.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:44:35,810 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=167439.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:44:45,748 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.3244, 1.2603, 1.5926, 1.5903, 1.1819, 1.1170, 1.3436, 0.9171], device='cuda:6'), covar=tensor([0.0564, 0.0530, 0.0412, 0.0510, 0.0640, 0.0988, 0.0520, 0.0609], device='cuda:6'), in_proj_covar=tensor([0.0071, 0.0067, 0.0065, 0.0068, 0.0075, 0.0094, 0.0072, 0.0062], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-28 05:44:56,431 INFO [finetune.py:976] (6/7) Epoch 30, batch 1350, loss[loss=0.1604, simple_loss=0.2342, pruned_loss=0.04326, over 4891.00 frames. ], tot_loss[loss=0.162, simple_loss=0.2351, pruned_loss=0.04441, over 955207.32 frames. ], batch size: 32, lr: 2.81e-03, grad_scale: 32.0 2023-04-28 05:44:58,349 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=167456.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:45:00,779 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=167460.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:45:17,748 INFO [zipformer.py:1188] (6/7) 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] (6/7) 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 05:45:34,217 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=167493.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:45:41,217 INFO [finetune.py:976] (6/7) Epoch 30, batch 1400, loss[loss=0.1327, simple_loss=0.2023, pruned_loss=0.03157, over 4830.00 frames. ], tot_loss[loss=0.1631, simple_loss=0.2368, pruned_loss=0.04474, over 954166.88 frames. ], batch size: 25, lr: 2.81e-03, grad_scale: 32.0 2023-04-28 05:46:14,263 INFO [finetune.py:976] (6/7) Epoch 30, batch 1450, loss[loss=0.1572, simple_loss=0.2306, pruned_loss=0.04189, over 4844.00 frames. ], tot_loss[loss=0.1635, simple_loss=0.2378, pruned_loss=0.04459, over 953134.77 frames. ], batch size: 49, lr: 2.81e-03, grad_scale: 32.0 2023-04-28 05:46:34,049 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7574, 2.9198, 2.3646, 2.6401, 2.9559, 2.5080, 3.8173, 2.1710], device='cuda:6'), covar=tensor([0.3860, 0.2431, 0.4098, 0.3423, 0.1804, 0.2698, 0.1428, 0.4357], device='cuda:6'), in_proj_covar=tensor([0.0337, 0.0352, 0.0420, 0.0349, 0.0382, 0.0371, 0.0369, 0.0421], device='cuda:6'), 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:6') 2023-04-28 05:46:52,053 INFO [optim.py:369] (6/7) 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 05:47:00,605 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=167585.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:47:21,140 INFO [finetune.py:976] (6/7) Epoch 30, batch 1500, loss[loss=0.179, simple_loss=0.2517, pruned_loss=0.05313, over 4884.00 frames. ], tot_loss[loss=0.165, simple_loss=0.2401, pruned_loss=0.04496, over 954668.69 frames. ], batch size: 35, lr: 2.81e-03, grad_scale: 32.0 2023-04-28 05:47:52,095 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.60 vs. limit=2.0 2023-04-28 05:47:56,733 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=167633.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:48:20,797 INFO [finetune.py:976] (6/7) Epoch 30, batch 1550, loss[loss=0.1396, simple_loss=0.2022, pruned_loss=0.03853, over 4415.00 frames. ], tot_loss[loss=0.1656, simple_loss=0.2406, pruned_loss=0.04534, over 954633.13 frames. ], batch size: 19, lr: 2.81e-03, grad_scale: 32.0 2023-04-28 05:48:40,247 INFO [optim.py:369] (6/7) 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] (6/7) Epoch 30, batch 1600, loss[loss=0.1402, simple_loss=0.2139, pruned_loss=0.03328, over 4903.00 frames. ], tot_loss[loss=0.1641, simple_loss=0.2386, pruned_loss=0.04476, over 955340.58 frames. ], batch size: 36, lr: 2.81e-03, grad_scale: 32.0 2023-04-28 05:49:05,593 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 30, batch 1650, loss[loss=0.1256, simple_loss=0.2044, pruned_loss=0.02341, over 4818.00 frames. ], tot_loss[loss=0.1628, simple_loss=0.2365, pruned_loss=0.04453, over 955600.16 frames. ], batch size: 41, lr: 2.81e-03, grad_scale: 32.0 2023-04-28 05:49:30,084 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=167756.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:49:32,993 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=167760.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:49:38,368 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=167767.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:49:47,069 INFO [optim.py:369] (6/7) 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 05:49:47,204 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=167780.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:49:51,122 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9514, 1.7332, 1.9099, 2.2843, 2.3225, 1.7833, 1.5250, 2.0734], device='cuda:6'), covar=tensor([0.0854, 0.1187, 0.0826, 0.0607, 0.0618, 0.0873, 0.0766, 0.0557], device='cuda:6'), in_proj_covar=tensor([0.0182, 0.0201, 0.0182, 0.0170, 0.0177, 0.0177, 0.0148, 0.0175], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-28 05:49:52,949 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 30, batch 1700, loss[loss=0.1362, simple_loss=0.2114, pruned_loss=0.03047, over 4813.00 frames. ], tot_loss[loss=0.1624, simple_loss=0.2357, pruned_loss=0.04453, over 955740.79 frames. ], batch size: 38, lr: 2.80e-03, grad_scale: 32.0 2023-04-28 05:50:02,853 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=167804.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:50:05,289 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=167808.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:50:18,863 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=167828.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:50:35,261 INFO [finetune.py:976] (6/7) Epoch 30, batch 1750, loss[loss=0.2131, simple_loss=0.2868, pruned_loss=0.06969, over 4276.00 frames. ], tot_loss[loss=0.1632, simple_loss=0.2369, pruned_loss=0.04478, over 954870.56 frames. ], batch size: 66, lr: 2.80e-03, grad_scale: 32.0 2023-04-28 05:50:40,303 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0030, 1.6017, 1.7954, 2.2676, 2.3326, 1.7969, 1.5434, 2.0054], device='cuda:6'), covar=tensor([0.0715, 0.1198, 0.0746, 0.0520, 0.0532, 0.0846, 0.0707, 0.0538], device='cuda:6'), in_proj_covar=tensor([0.0182, 0.0202, 0.0182, 0.0171, 0.0178, 0.0177, 0.0149, 0.0176], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-28 05:50:48,446 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.0929, 2.3473, 1.0350, 1.4618, 1.4023, 1.9125, 1.4993, 0.9853], device='cuda:6'), covar=tensor([0.1279, 0.0869, 0.1357, 0.1113, 0.1093, 0.0772, 0.1322, 0.1616], device='cuda:6'), in_proj_covar=tensor([0.0119, 0.0241, 0.0138, 0.0122, 0.0134, 0.0155, 0.0120, 0.0119], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-28 05:50:50,305 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5314, 1.4585, 1.8825, 1.9109, 1.3803, 1.2905, 1.5391, 0.9275], device='cuda:6'), covar=tensor([0.0535, 0.0642, 0.0321, 0.0577, 0.0751, 0.1105, 0.0527, 0.0604], device='cuda:6'), in_proj_covar=tensor([0.0071, 0.0067, 0.0065, 0.0069, 0.0075, 0.0094, 0.0072, 0.0062], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-28 05:50:53,218 INFO [optim.py:369] (6/7) 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 05:51:08,164 INFO [finetune.py:976] (6/7) Epoch 30, batch 1800, loss[loss=0.1755, simple_loss=0.2487, pruned_loss=0.05112, over 4887.00 frames. ], tot_loss[loss=0.1658, simple_loss=0.2403, pruned_loss=0.04569, over 955285.07 frames. ], batch size: 35, lr: 2.80e-03, grad_scale: 32.0 2023-04-28 05:51:18,262 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.53 vs. limit=5.0 2023-04-28 05:51:26,695 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.5185, 1.7600, 1.9815, 2.0827, 1.9899, 1.9678, 2.0599, 2.0038], device='cuda:6'), covar=tensor([0.3424, 0.5167, 0.4373, 0.4253, 0.5165, 0.6576, 0.5220, 0.4617], device='cuda:6'), in_proj_covar=tensor([0.0343, 0.0373, 0.0331, 0.0342, 0.0351, 0.0392, 0.0361, 0.0334], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-28 05:51:41,398 INFO [finetune.py:976] (6/7) Epoch 30, batch 1850, loss[loss=0.1998, simple_loss=0.2754, pruned_loss=0.0621, over 4900.00 frames. ], tot_loss[loss=0.1679, simple_loss=0.2423, pruned_loss=0.04678, over 954679.38 frames. ], batch size: 37, lr: 2.80e-03, grad_scale: 32.0 2023-04-28 05:52:04,616 INFO [optim.py:369] (6/7) 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 05:52:26,315 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.3701, 1.2612, 1.8098, 1.7048, 1.2571, 1.1998, 1.3620, 0.8581], device='cuda:6'), covar=tensor([0.0619, 0.0656, 0.0323, 0.0686, 0.0812, 0.1113, 0.0651, 0.0579], device='cuda:6'), in_proj_covar=tensor([0.0070, 0.0067, 0.0064, 0.0068, 0.0075, 0.0094, 0.0072, 0.0062], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-28 05:52:26,788 INFO [finetune.py:976] (6/7) Epoch 30, batch 1900, loss[loss=0.1718, simple_loss=0.2463, pruned_loss=0.04869, over 4859.00 frames. ], tot_loss[loss=0.1684, simple_loss=0.2426, pruned_loss=0.0471, over 956158.93 frames. ], batch size: 44, lr: 2.80e-03, grad_scale: 32.0 2023-04-28 05:53:09,304 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=168035.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:53:21,020 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.7006, 2.0560, 2.1283, 2.2480, 2.0979, 2.1588, 2.1607, 2.1279], device='cuda:6'), covar=tensor([0.3574, 0.5206, 0.4464, 0.4218, 0.5475, 0.6926, 0.5161, 0.4492], device='cuda:6'), in_proj_covar=tensor([0.0343, 0.0372, 0.0331, 0.0341, 0.0351, 0.0393, 0.0360, 0.0333], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-28 05:53:31,845 INFO [finetune.py:976] (6/7) Epoch 30, batch 1950, loss[loss=0.1516, simple_loss=0.2285, pruned_loss=0.03735, over 4921.00 frames. ], tot_loss[loss=0.1665, simple_loss=0.2406, pruned_loss=0.04626, over 955760.70 frames. ], batch size: 38, lr: 2.80e-03, grad_scale: 32.0 2023-04-28 05:53:32,601 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.4404, 1.7566, 2.3184, 2.6503, 2.2821, 1.8191, 1.5565, 1.9940], device='cuda:6'), covar=tensor([0.3036, 0.3039, 0.1538, 0.2174, 0.2391, 0.2749, 0.4017, 0.1981], device='cuda:6'), in_proj_covar=tensor([0.0293, 0.0247, 0.0230, 0.0316, 0.0224, 0.0237, 0.0230, 0.0187], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:6') 2023-04-28 05:53:56,262 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=168075.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:54:04,672 INFO [optim.py:369] (6/7) 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 05:54:13,537 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=168088.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:54:24,493 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=168096.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 05:54:25,946 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.56 vs. limit=2.0 2023-04-28 05:54:32,404 INFO [finetune.py:976] (6/7) Epoch 30, batch 2000, loss[loss=0.1764, simple_loss=0.2387, pruned_loss=0.05704, over 4813.00 frames. ], tot_loss[loss=0.1644, simple_loss=0.2381, pruned_loss=0.04538, over 955849.85 frames. ], batch size: 39, lr: 2.80e-03, grad_scale: 32.0 2023-04-28 05:54:55,109 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=168123.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:55:14,211 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 30, batch 2050, loss[loss=0.164, simple_loss=0.2432, pruned_loss=0.04243, over 4820.00 frames. ], tot_loss[loss=0.1617, simple_loss=0.2345, pruned_loss=0.04444, over 952063.01 frames. ], batch size: 33, lr: 2.80e-03, grad_scale: 32.0 2023-04-28 05:55:40,689 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0154, 2.4642, 2.0675, 2.3952, 1.7345, 2.2217, 2.1473, 1.6165], device='cuda:6'), covar=tensor([0.1746, 0.0965, 0.0708, 0.1015, 0.3265, 0.0898, 0.1855, 0.2456], device='cuda:6'), in_proj_covar=tensor([0.0279, 0.0298, 0.0216, 0.0271, 0.0306, 0.0250, 0.0245, 0.0258], device='cuda:6'), 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:6') 2023-04-28 05:55:43,777 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0375, 2.3654, 2.0560, 1.9212, 1.5587, 1.5726, 2.0302, 1.4705], device='cuda:6'), covar=tensor([0.1516, 0.1200, 0.1351, 0.1446, 0.2097, 0.1763, 0.0945, 0.1920], device='cuda:6'), in_proj_covar=tensor([0.0198, 0.0210, 0.0170, 0.0205, 0.0201, 0.0188, 0.0157, 0.0188], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:6') 2023-04-28 05:55:44,978 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.6982, 2.4832, 2.5775, 3.2366, 3.1790, 2.5196, 2.3072, 2.7941], device='cuda:6'), covar=tensor([0.0754, 0.0990, 0.0655, 0.0506, 0.0471, 0.0844, 0.0669, 0.0500], device='cuda:6'), in_proj_covar=tensor([0.0181, 0.0201, 0.0181, 0.0169, 0.0176, 0.0176, 0.0148, 0.0174], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-28 05:55:47,312 INFO [optim.py:369] (6/7) 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 05:56:04,272 INFO [finetune.py:976] (6/7) Epoch 30, batch 2100, loss[loss=0.1812, simple_loss=0.2596, pruned_loss=0.05142, over 4839.00 frames. ], tot_loss[loss=0.1631, simple_loss=0.2355, pruned_loss=0.04533, over 952774.95 frames. ], batch size: 47, lr: 2.80e-03, grad_scale: 32.0 2023-04-28 05:56:23,572 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=168233.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:56:37,574 INFO [finetune.py:976] (6/7) Epoch 30, batch 2150, loss[loss=0.1597, simple_loss=0.228, pruned_loss=0.04566, over 4762.00 frames. ], tot_loss[loss=0.1664, simple_loss=0.2394, pruned_loss=0.04672, over 952089.46 frames. ], batch size: 28, lr: 2.80e-03, grad_scale: 32.0 2023-04-28 05:56:54,473 INFO [optim.py:369] (6/7) 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 05:57:04,549 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=168294.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:57:09,877 INFO [finetune.py:976] (6/7) Epoch 30, batch 2200, loss[loss=0.1576, simple_loss=0.227, pruned_loss=0.0441, over 4709.00 frames. ], tot_loss[loss=0.1677, simple_loss=0.2412, pruned_loss=0.04715, over 952093.75 frames. ], batch size: 23, lr: 2.80e-03, grad_scale: 32.0 2023-04-28 05:57:18,063 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.0121, 1.0316, 1.1872, 1.1713, 0.9907, 0.9321, 0.9778, 0.4840], device='cuda:6'), covar=tensor([0.0517, 0.0452, 0.0399, 0.0439, 0.0627, 0.1033, 0.0430, 0.0575], device='cuda:6'), in_proj_covar=tensor([0.0071, 0.0067, 0.0065, 0.0069, 0.0076, 0.0095, 0.0072, 0.0062], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-28 05:58:09,636 INFO [finetune.py:976] (6/7) Epoch 30, batch 2250, loss[loss=0.1838, simple_loss=0.2612, pruned_loss=0.05327, over 4128.00 frames. ], tot_loss[loss=0.1692, simple_loss=0.2431, pruned_loss=0.04765, over 952753.29 frames. ], batch size: 65, lr: 2.80e-03, grad_scale: 32.0 2023-04-28 05:58:40,754 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=168375.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:58:43,711 INFO [optim.py:369] (6/7) 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 05:58:55,537 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 30, batch 2300, loss[loss=0.1339, simple_loss=0.2106, pruned_loss=0.02864, over 4847.00 frames. ], tot_loss[loss=0.1682, simple_loss=0.2426, pruned_loss=0.04684, over 952617.22 frames. ], batch size: 49, lr: 2.80e-03, grad_scale: 32.0 2023-04-28 05:59:23,256 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9428, 1.7852, 2.0369, 2.2471, 2.2617, 1.8636, 1.5317, 2.0470], device='cuda:6'), covar=tensor([0.0855, 0.1176, 0.0689, 0.0591, 0.0681, 0.0921, 0.0788, 0.0592], device='cuda:6'), in_proj_covar=tensor([0.0181, 0.0201, 0.0182, 0.0169, 0.0176, 0.0176, 0.0148, 0.0175], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-28 05:59:38,057 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=168423.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:59:38,080 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=168423.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:00:16,727 INFO [finetune.py:976] (6/7) Epoch 30, batch 2350, loss[loss=0.146, simple_loss=0.219, pruned_loss=0.03644, over 4857.00 frames. ], tot_loss[loss=0.1665, simple_loss=0.2405, pruned_loss=0.04621, over 953588.90 frames. ], batch size: 31, lr: 2.80e-03, grad_scale: 32.0 2023-04-28 06:00:30,241 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.4376, 1.9018, 1.7251, 2.4931, 2.5930, 2.1362, 2.0898, 1.8001], device='cuda:6'), covar=tensor([0.1721, 0.1660, 0.2101, 0.1478, 0.1018, 0.1762, 0.2025, 0.2339], device='cuda:6'), in_proj_covar=tensor([0.0313, 0.0307, 0.0349, 0.0285, 0.0323, 0.0304, 0.0299, 0.0376], device='cuda:6'), out_proj_covar=tensor([6.3720e-05, 6.2615e-05, 7.2932e-05, 5.6748e-05, 6.5819e-05, 6.2901e-05, 6.1614e-05, 7.9413e-05], device='cuda:6') 2023-04-28 06:00:37,861 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=168471.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:00:39,115 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.4846, 2.9629, 0.9616, 1.8244, 1.8241, 2.2345, 1.8464, 1.0505], device='cuda:6'), covar=tensor([0.1258, 0.1018, 0.1715, 0.1103, 0.0980, 0.0922, 0.1304, 0.1801], device='cuda:6'), in_proj_covar=tensor([0.0118, 0.0239, 0.0137, 0.0121, 0.0132, 0.0154, 0.0118, 0.0118], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-28 06:00:48,857 INFO [optim.py:369] (6/7) 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 06:01:09,850 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8972, 1.4458, 1.5573, 1.7549, 2.0376, 1.6891, 1.4669, 1.4864], device='cuda:6'), covar=tensor([0.1475, 0.1513, 0.1620, 0.1033, 0.0818, 0.1422, 0.1907, 0.2144], device='cuda:6'), in_proj_covar=tensor([0.0315, 0.0309, 0.0351, 0.0286, 0.0325, 0.0305, 0.0301, 0.0378], device='cuda:6'), out_proj_covar=tensor([6.4069e-05, 6.2988e-05, 7.3386e-05, 5.7072e-05, 6.6259e-05, 6.3256e-05, 6.1999e-05, 7.9792e-05], device='cuda:6') 2023-04-28 06:01:21,061 INFO [finetune.py:976] (6/7) Epoch 30, batch 2400, loss[loss=0.1552, simple_loss=0.2251, pruned_loss=0.04259, over 4907.00 frames. ], tot_loss[loss=0.1638, simple_loss=0.2371, pruned_loss=0.04522, over 953837.49 frames. ], batch size: 36, lr: 2.80e-03, grad_scale: 32.0 2023-04-28 06:02:28,241 INFO [finetune.py:976] (6/7) Epoch 30, batch 2450, loss[loss=0.1676, simple_loss=0.2388, pruned_loss=0.04824, over 4088.00 frames. ], tot_loss[loss=0.1624, simple_loss=0.2351, pruned_loss=0.04483, over 952751.94 frames. ], batch size: 65, lr: 2.80e-03, grad_scale: 32.0 2023-04-28 06:02:38,763 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.60 vs. limit=2.0 2023-04-28 06:02:49,820 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7215, 1.6963, 2.1318, 2.2942, 1.6390, 1.4265, 1.7525, 0.9215], device='cuda:6'), covar=tensor([0.0694, 0.0530, 0.0412, 0.0549, 0.0733, 0.1130, 0.0556, 0.0730], device='cuda:6'), in_proj_covar=tensor([0.0070, 0.0067, 0.0065, 0.0068, 0.0075, 0.0094, 0.0072, 0.0062], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-28 06:03:10,382 INFO [optim.py:369] (6/7) 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 06:03:21,582 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=168589.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:03:36,318 INFO [finetune.py:976] (6/7) Epoch 30, batch 2500, loss[loss=0.1736, simple_loss=0.2391, pruned_loss=0.05404, over 4123.00 frames. ], tot_loss[loss=0.1629, simple_loss=0.2364, pruned_loss=0.04468, over 952350.63 frames. ], batch size: 65, lr: 2.80e-03, grad_scale: 32.0 2023-04-28 06:03:57,639 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5582, 1.6561, 1.4318, 1.1226, 1.1891, 1.1606, 1.3794, 1.0994], device='cuda:6'), covar=tensor([0.1831, 0.1244, 0.1531, 0.1734, 0.2340, 0.2040, 0.1091, 0.2109], device='cuda:6'), in_proj_covar=tensor([0.0200, 0.0210, 0.0171, 0.0205, 0.0202, 0.0189, 0.0158, 0.0188], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:6') 2023-04-28 06:04:49,840 INFO [finetune.py:976] (6/7) Epoch 30, batch 2550, loss[loss=0.17, simple_loss=0.2423, pruned_loss=0.04884, over 4925.00 frames. ], tot_loss[loss=0.1651, simple_loss=0.2395, pruned_loss=0.04534, over 954807.22 frames. ], batch size: 33, lr: 2.80e-03, grad_scale: 32.0 2023-04-28 06:05:22,401 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6687, 1.7230, 0.9279, 1.3731, 1.7878, 1.5210, 1.4463, 1.5137], device='cuda:6'), covar=tensor([0.0468, 0.0353, 0.0338, 0.0539, 0.0253, 0.0497, 0.0509, 0.0554], device='cuda:6'), in_proj_covar=tensor([0.0027, 0.0023, 0.0021, 0.0028, 0.0019, 0.0027, 0.0027, 0.0028], device='cuda:6'), out_proj_covar=tensor([0.0052, 0.0046, 0.0039, 0.0053, 0.0039, 0.0051, 0.0051, 0.0053], device='cuda:6') 2023-04-28 06:05:23,461 INFO [optim.py:369] (6/7) 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 06:05:31,388 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=168691.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:05:38,673 INFO [finetune.py:976] (6/7) Epoch 30, batch 2600, loss[loss=0.1707, simple_loss=0.2513, pruned_loss=0.04507, over 4892.00 frames. ], tot_loss[loss=0.1664, simple_loss=0.2407, pruned_loss=0.04605, over 954850.74 frames. ], batch size: 35, lr: 2.80e-03, grad_scale: 32.0 2023-04-28 06:06:26,240 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 30, batch 2650, loss[loss=0.1497, simple_loss=0.2369, pruned_loss=0.03127, over 4807.00 frames. ], tot_loss[loss=0.1665, simple_loss=0.2411, pruned_loss=0.04592, over 952891.19 frames. ], batch size: 39, lr: 2.80e-03, grad_scale: 32.0 2023-04-28 06:07:17,593 INFO [optim.py:369] (6/7) 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 06:07:44,074 INFO [finetune.py:976] (6/7) Epoch 30, batch 2700, loss[loss=0.1507, simple_loss=0.219, pruned_loss=0.04125, over 4804.00 frames. ], tot_loss[loss=0.1663, simple_loss=0.2405, pruned_loss=0.04602, over 952716.46 frames. ], batch size: 40, lr: 2.80e-03, grad_scale: 32.0 2023-04-28 06:08:10,493 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8119, 1.4486, 1.9136, 2.0893, 1.8456, 1.7806, 1.8341, 1.8875], device='cuda:6'), covar=tensor([0.5803, 0.8280, 0.7892, 0.8070, 0.7334, 1.0440, 1.0023, 1.0505], device='cuda:6'), in_proj_covar=tensor([0.0446, 0.0424, 0.0518, 0.0507, 0.0472, 0.0512, 0.0513, 0.0526], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-28 06:08:13,255 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-04-28 06:08:17,673 INFO [finetune.py:976] (6/7) Epoch 30, batch 2750, loss[loss=0.1613, simple_loss=0.2356, pruned_loss=0.0435, over 4891.00 frames. ], tot_loss[loss=0.163, simple_loss=0.2367, pruned_loss=0.04466, over 952102.50 frames. ], batch size: 32, lr: 2.80e-03, grad_scale: 32.0 2023-04-28 06:08:35,310 INFO [optim.py:369] (6/7) 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 06:08:42,281 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=168889.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:08:47,500 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.28 vs. limit=5.0 2023-04-28 06:08:50,352 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-04-28 06:08:50,765 INFO [finetune.py:976] (6/7) Epoch 30, batch 2800, loss[loss=0.1373, simple_loss=0.2097, pruned_loss=0.03242, over 4795.00 frames. ], tot_loss[loss=0.1612, simple_loss=0.234, pruned_loss=0.04425, over 954617.27 frames. ], batch size: 29, lr: 2.80e-03, grad_scale: 32.0 2023-04-28 06:09:05,861 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.1029, 1.3585, 1.2477, 1.6771, 1.4758, 1.4317, 1.3528, 2.4814], device='cuda:6'), covar=tensor([0.0637, 0.0848, 0.0859, 0.1218, 0.0684, 0.0497, 0.0753, 0.0208], device='cuda:6'), in_proj_covar=tensor([0.0037, 0.0038, 0.0039, 0.0043, 0.0040, 0.0037, 0.0038, 0.0054], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0013, 0.0013, 0.0015, 0.0014, 0.0013, 0.0013, 0.0016], device='cuda:6') 2023-04-28 06:09:12,958 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=168937.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:09:20,967 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.6335, 1.9723, 2.1024, 2.1629, 2.1039, 2.0685, 2.1440, 2.1275], device='cuda:6'), covar=tensor([0.4084, 0.5692, 0.4525, 0.4469, 0.5622, 0.7043, 0.5170, 0.4563], device='cuda:6'), in_proj_covar=tensor([0.0343, 0.0374, 0.0330, 0.0342, 0.0351, 0.0394, 0.0361, 0.0334], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-28 06:09:24,482 INFO [finetune.py:976] (6/7) Epoch 30, batch 2850, loss[loss=0.2366, simple_loss=0.2886, pruned_loss=0.09232, over 4936.00 frames. ], tot_loss[loss=0.1615, simple_loss=0.2337, pruned_loss=0.04461, over 956246.72 frames. ], batch size: 33, lr: 2.80e-03, grad_scale: 32.0 2023-04-28 06:09:33,164 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=168967.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:09:34,375 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=168969.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 06:09:39,648 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.0988, 1.3404, 1.2160, 1.5873, 1.4419, 1.3906, 1.3094, 2.4865], device='cuda:6'), covar=tensor([0.0571, 0.0821, 0.0793, 0.1181, 0.0635, 0.0513, 0.0750, 0.0196], device='cuda:6'), in_proj_covar=tensor([0.0037, 0.0039, 0.0039, 0.0043, 0.0040, 0.0038, 0.0038, 0.0055], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0013, 0.0014, 0.0015, 0.0014, 0.0013, 0.0013, 0.0016], device='cuda:6') 2023-04-28 06:09:41,981 INFO [optim.py:369] (6/7) 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 06:09:44,217 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-04-28 06:09:47,205 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.66 vs. limit=2.0 2023-04-28 06:09:58,542 INFO [finetune.py:976] (6/7) Epoch 30, batch 2900, loss[loss=0.1674, simple_loss=0.2419, pruned_loss=0.04647, over 4877.00 frames. ], tot_loss[loss=0.1635, simple_loss=0.2366, pruned_loss=0.04517, over 956152.12 frames. ], batch size: 31, lr: 2.80e-03, grad_scale: 32.0 2023-04-28 06:10:00,528 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0632, 1.5893, 1.8657, 2.2375, 1.9883, 1.5817, 1.3158, 1.7358], device='cuda:6'), covar=tensor([0.2433, 0.2613, 0.1384, 0.1620, 0.2106, 0.2231, 0.4062, 0.1745], device='cuda:6'), in_proj_covar=tensor([0.0292, 0.0246, 0.0229, 0.0315, 0.0223, 0.0235, 0.0230, 0.0186], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:6') 2023-04-28 06:10:14,814 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=169028.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:10:15,392 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.1423, 1.4369, 1.2948, 1.6636, 1.5185, 1.5669, 1.3649, 2.9428], device='cuda:6'), covar=tensor([0.0661, 0.0805, 0.0777, 0.1159, 0.0617, 0.0604, 0.0742, 0.0166], device='cuda:6'), in_proj_covar=tensor([0.0037, 0.0039, 0.0039, 0.0043, 0.0040, 0.0038, 0.0038, 0.0055], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0013, 0.0014, 0.0015, 0.0014, 0.0013, 0.0013, 0.0016], device='cuda:6') 2023-04-28 06:10:16,027 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=169030.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 06:10:31,850 INFO [finetune.py:976] (6/7) Epoch 30, batch 2950, loss[loss=0.1964, simple_loss=0.2674, pruned_loss=0.06271, over 4815.00 frames. ], tot_loss[loss=0.1648, simple_loss=0.2388, pruned_loss=0.04544, over 954191.84 frames. ], batch size: 38, lr: 2.80e-03, grad_scale: 32.0 2023-04-28 06:10:49,918 INFO [optim.py:369] (6/7) 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 06:10:50,645 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 30, batch 3000, loss[loss=0.1818, simple_loss=0.2589, pruned_loss=0.05232, over 4750.00 frames. ], tot_loss[loss=0.167, simple_loss=0.2414, pruned_loss=0.04631, over 956017.56 frames. ], batch size: 27, lr: 2.80e-03, grad_scale: 32.0 2023-04-28 06:11:05,768 INFO [finetune.py:1001] (6/7) Computing validation loss 2023-04-28 06:11:10,368 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.4464, 1.2265, 3.8021, 3.5193, 3.4331, 3.6824, 3.7385, 3.3829], device='cuda:6'), covar=tensor([0.7155, 0.5313, 0.1249, 0.2120, 0.1329, 0.1391, 0.0876, 0.1829], device='cuda:6'), in_proj_covar=tensor([0.0309, 0.0308, 0.0406, 0.0409, 0.0349, 0.0415, 0.0314, 0.0363], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-28 06:11:12,759 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0567, 2.5328, 1.0987, 1.3712, 1.9206, 1.2843, 3.0419, 1.6825], device='cuda:6'), covar=tensor([0.0666, 0.0567, 0.0698, 0.1204, 0.0426, 0.0935, 0.0244, 0.0551], device='cuda:6'), in_proj_covar=tensor([0.0051, 0.0064, 0.0046, 0.0046, 0.0049, 0.0051, 0.0072, 0.0050], device='cuda:6'), out_proj_covar=tensor([0.0008, 0.0010, 0.0007, 0.0008, 0.0008, 0.0008, 0.0010, 0.0008], device='cuda:6') 2023-04-28 06:11:16,542 INFO [finetune.py:1010] (6/7) 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] (6/7) Maximum memory allocated so far is 6435MB 2023-04-28 06:11:16,652 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8646, 2.2488, 1.8752, 2.3232, 1.4644, 1.9050, 2.0362, 1.4361], device='cuda:6'), covar=tensor([0.1969, 0.1468, 0.0927, 0.1072, 0.3645, 0.1236, 0.1855, 0.2675], device='cuda:6'), in_proj_covar=tensor([0.0283, 0.0302, 0.0218, 0.0275, 0.0310, 0.0256, 0.0249, 0.0262], device='cuda:6'), out_proj_covar=tensor([1.1272e-04, 1.1861e-04, 8.5661e-05, 1.0777e-04, 1.2485e-04, 1.0040e-04, 1.0019e-04, 1.0290e-04], device='cuda:6') 2023-04-28 06:11:16,861 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.78 vs. limit=2.0 2023-04-28 06:11:24,699 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=169116.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:11:27,739 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([0.9081, 1.7046, 1.6431, 1.4382, 1.8455, 1.5515, 2.2376, 1.4684], device='cuda:6'), covar=tensor([0.3081, 0.1842, 0.4541, 0.2482, 0.1415, 0.2164, 0.1261, 0.4284], device='cuda:6'), in_proj_covar=tensor([0.0341, 0.0358, 0.0429, 0.0354, 0.0387, 0.0378, 0.0373, 0.0427], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-28 06:11:52,085 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=169143.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:12:04,396 INFO [finetune.py:976] (6/7) Epoch 30, batch 3050, loss[loss=0.1426, simple_loss=0.2244, pruned_loss=0.0304, over 4916.00 frames. ], tot_loss[loss=0.1656, simple_loss=0.2409, pruned_loss=0.04517, over 956195.83 frames. ], batch size: 42, lr: 2.80e-03, grad_scale: 32.0 2023-04-28 06:12:13,911 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=169159.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:12:36,609 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=169177.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:12:42,083 INFO [optim.py:369] (6/7) 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] (6/7) Epoch 30, batch 3100, loss[loss=0.1362, simple_loss=0.2146, pruned_loss=0.02887, over 4812.00 frames. ], tot_loss[loss=0.1634, simple_loss=0.238, pruned_loss=0.04443, over 955709.05 frames. ], batch size: 41, lr: 2.80e-03, grad_scale: 32.0 2023-04-28 06:13:29,656 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=169220.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:13:38,187 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.3694, 1.8252, 2.3026, 2.6400, 2.2765, 1.8153, 1.4703, 2.0962], device='cuda:6'), covar=tensor([0.2970, 0.2954, 0.1428, 0.2148, 0.2532, 0.2562, 0.3990, 0.1863], device='cuda:6'), in_proj_covar=tensor([0.0293, 0.0247, 0.0229, 0.0316, 0.0224, 0.0236, 0.0230, 0.0187], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:6') 2023-04-28 06:13:50,603 INFO [finetune.py:976] (6/7) Epoch 30, batch 3150, loss[loss=0.1808, simple_loss=0.2501, pruned_loss=0.05572, over 4781.00 frames. ], tot_loss[loss=0.1611, simple_loss=0.235, pruned_loss=0.04359, over 955984.84 frames. ], batch size: 51, lr: 2.80e-03, grad_scale: 32.0 2023-04-28 06:14:10,081 INFO [optim.py:369] (6/7) 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 06:14:23,998 INFO [finetune.py:976] (6/7) Epoch 30, batch 3200, loss[loss=0.1477, simple_loss=0.2221, pruned_loss=0.03658, over 4867.00 frames. ], tot_loss[loss=0.1582, simple_loss=0.2316, pruned_loss=0.04246, over 955998.98 frames. ], batch size: 34, lr: 2.80e-03, grad_scale: 32.0 2023-04-28 06:14:26,789 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-04-28 06:14:32,278 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-04-28 06:14:39,323 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=169323.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:14:40,576 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=169325.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 06:14:43,053 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.8535, 2.5721, 2.8704, 3.4195, 3.1421, 2.7170, 2.4206, 3.1874], device='cuda:6'), covar=tensor([0.0710, 0.1057, 0.0594, 0.0474, 0.0533, 0.0846, 0.0638, 0.0428], device='cuda:6'), in_proj_covar=tensor([0.0185, 0.0205, 0.0185, 0.0172, 0.0179, 0.0180, 0.0151, 0.0178], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-28 06:14:48,503 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.3390, 2.9761, 0.8981, 1.7405, 1.5970, 2.1015, 1.6482, 0.9990], device='cuda:6'), covar=tensor([0.1354, 0.0867, 0.1824, 0.1167, 0.1145, 0.0990, 0.1531, 0.1774], device='cuda:6'), in_proj_covar=tensor([0.0117, 0.0237, 0.0135, 0.0120, 0.0131, 0.0153, 0.0117, 0.0117], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-28 06:14:57,967 INFO [finetune.py:976] (6/7) Epoch 30, batch 3250, loss[loss=0.2131, simple_loss=0.2682, pruned_loss=0.07899, over 4112.00 frames. ], tot_loss[loss=0.1586, simple_loss=0.2315, pruned_loss=0.04285, over 955564.89 frames. ], batch size: 65, lr: 2.80e-03, grad_scale: 32.0 2023-04-28 06:15:18,031 INFO [optim.py:369] (6/7) 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 06:15:32,128 INFO [finetune.py:976] (6/7) Epoch 30, batch 3300, loss[loss=0.1842, simple_loss=0.2687, pruned_loss=0.04981, over 4914.00 frames. ], tot_loss[loss=0.1614, simple_loss=0.2355, pruned_loss=0.04369, over 955401.73 frames. ], batch size: 37, lr: 2.79e-03, grad_scale: 32.0 2023-04-28 06:15:56,492 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=169438.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:15:58,404 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.2920, 1.6516, 1.8307, 1.8559, 1.7562, 1.7537, 1.9202, 1.8309], device='cuda:6'), covar=tensor([0.3939, 0.5293, 0.4464, 0.4272, 0.5316, 0.7183, 0.5110, 0.4655], device='cuda:6'), in_proj_covar=tensor([0.0344, 0.0375, 0.0331, 0.0343, 0.0352, 0.0396, 0.0363, 0.0336], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-28 06:16:05,662 INFO [finetune.py:976] (6/7) Epoch 30, batch 3350, loss[loss=0.1923, simple_loss=0.2601, pruned_loss=0.06228, over 4836.00 frames. ], tot_loss[loss=0.1626, simple_loss=0.2373, pruned_loss=0.04399, over 955673.79 frames. ], batch size: 30, lr: 2.79e-03, grad_scale: 32.0 2023-04-28 06:16:18,732 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=169472.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:16:26,058 INFO [optim.py:369] (6/7) 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 06:16:33,510 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([0.8329, 1.7073, 1.6731, 1.3998, 1.8173, 1.6365, 2.2653, 1.4882], device='cuda:6'), covar=tensor([0.3506, 0.1994, 0.4599, 0.2909, 0.1451, 0.2090, 0.1540, 0.4582], device='cuda:6'), in_proj_covar=tensor([0.0340, 0.0357, 0.0427, 0.0353, 0.0386, 0.0376, 0.0372, 0.0427], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-28 06:16:39,415 INFO [finetune.py:976] (6/7) Epoch 30, batch 3400, loss[loss=0.1389, simple_loss=0.2154, pruned_loss=0.03122, over 4791.00 frames. ], tot_loss[loss=0.165, simple_loss=0.2395, pruned_loss=0.04522, over 952433.97 frames. ], batch size: 29, lr: 2.79e-03, grad_scale: 32.0 2023-04-28 06:16:42,075 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=3.60 vs. limit=5.0 2023-04-28 06:16:47,267 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=169515.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:17:07,994 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=3.02 vs. limit=5.0 2023-04-28 06:17:12,718 INFO [finetune.py:976] (6/7) Epoch 30, batch 3450, loss[loss=0.1708, simple_loss=0.253, pruned_loss=0.04432, over 4885.00 frames. ], tot_loss[loss=0.1642, simple_loss=0.2389, pruned_loss=0.04469, over 953705.48 frames. ], batch size: 35, lr: 2.79e-03, grad_scale: 32.0 2023-04-28 06:17:15,232 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5058, 1.5094, 1.8635, 1.9535, 1.4239, 1.1737, 1.4197, 0.8189], device='cuda:6'), covar=tensor([0.0574, 0.0523, 0.0384, 0.0472, 0.0628, 0.1305, 0.0575, 0.0703], device='cuda:6'), in_proj_covar=tensor([0.0071, 0.0067, 0.0065, 0.0068, 0.0075, 0.0094, 0.0072, 0.0062], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-28 06:17:21,142 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([0.9864, 0.9855, 1.1433, 1.1580, 0.9851, 0.8924, 0.9838, 0.5586], device='cuda:6'), covar=tensor([0.0513, 0.0499, 0.0452, 0.0483, 0.0678, 0.0967, 0.0429, 0.0582], device='cuda:6'), in_proj_covar=tensor([0.0071, 0.0067, 0.0065, 0.0068, 0.0075, 0.0094, 0.0072, 0.0062], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-28 06:17:36,960 INFO [optim.py:369] (6/7) 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 06:18:07,737 INFO [finetune.py:976] (6/7) Epoch 30, batch 3500, loss[loss=0.1515, simple_loss=0.2229, pruned_loss=0.04008, over 4826.00 frames. ], tot_loss[loss=0.1619, simple_loss=0.2363, pruned_loss=0.04376, over 956166.45 frames. ], batch size: 39, lr: 2.79e-03, grad_scale: 32.0 2023-04-28 06:18:29,812 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=169621.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:18:31,009 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=169623.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:18:32,267 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=169625.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 06:19:13,503 INFO [finetune.py:976] (6/7) Epoch 30, batch 3550, loss[loss=0.1379, simple_loss=0.2082, pruned_loss=0.03386, over 4771.00 frames. ], tot_loss[loss=0.16, simple_loss=0.2336, pruned_loss=0.04316, over 957439.91 frames. ], batch size: 26, lr: 2.79e-03, grad_scale: 32.0 2023-04-28 06:19:25,256 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0557, 4.0022, 1.0071, 2.3127, 2.4738, 2.7480, 2.3362, 1.0081], device='cuda:6'), covar=tensor([0.1154, 0.0870, 0.1887, 0.1108, 0.0924, 0.1001, 0.1325, 0.2034], device='cuda:6'), in_proj_covar=tensor([0.0118, 0.0238, 0.0136, 0.0121, 0.0131, 0.0154, 0.0118, 0.0118], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0005, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:6') 2023-04-28 06:19:30,802 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=169671.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:19:32,047 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=169673.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 06:19:42,310 INFO [optim.py:369] (6/7) 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 06:19:43,561 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 30, batch 3600, loss[loss=0.1519, simple_loss=0.2258, pruned_loss=0.03901, over 4904.00 frames. ], tot_loss[loss=0.158, simple_loss=0.2314, pruned_loss=0.04232, over 956015.40 frames. ], batch size: 36, lr: 2.79e-03, grad_scale: 32.0 2023-04-28 06:20:45,840 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=169729.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:20:57,514 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=169738.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:21:10,904 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=169747.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:21:20,088 INFO [finetune.py:976] (6/7) Epoch 30, batch 3650, loss[loss=0.162, simple_loss=0.238, pruned_loss=0.043, over 4797.00 frames. ], tot_loss[loss=0.1619, simple_loss=0.2359, pruned_loss=0.04397, over 956112.38 frames. ], batch size: 29, lr: 2.79e-03, grad_scale: 32.0 2023-04-28 06:21:43,301 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=169772.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:21:53,883 INFO [optim.py:369] (6/7) 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 06:22:03,118 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=169786.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:22:03,158 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=169786.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:22:11,322 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=169790.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:22:15,999 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6735, 2.6142, 2.0359, 2.3056, 2.7212, 2.2973, 3.5428, 1.9885], device='cuda:6'), covar=tensor([0.3832, 0.2275, 0.4743, 0.3325, 0.1797, 0.2672, 0.1744, 0.4403], device='cuda:6'), in_proj_covar=tensor([0.0342, 0.0360, 0.0430, 0.0356, 0.0388, 0.0379, 0.0373, 0.0428], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-28 06:22:25,954 INFO [finetune.py:976] (6/7) Epoch 30, batch 3700, loss[loss=0.1825, simple_loss=0.2558, pruned_loss=0.05462, over 4796.00 frames. ], tot_loss[loss=0.164, simple_loss=0.2388, pruned_loss=0.04453, over 954728.21 frames. ], batch size: 51, lr: 2.79e-03, grad_scale: 32.0 2023-04-28 06:22:35,275 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=169808.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:22:44,959 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=169815.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:22:47,944 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=169820.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:23:22,782 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=169847.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:23:31,475 INFO [finetune.py:976] (6/7) Epoch 30, batch 3750, loss[loss=0.2041, simple_loss=0.2853, pruned_loss=0.06148, over 4270.00 frames. ], tot_loss[loss=0.1657, simple_loss=0.2408, pruned_loss=0.04528, over 954961.87 frames. ], batch size: 66, lr: 2.79e-03, grad_scale: 32.0 2023-04-28 06:23:42,232 INFO [zipformer.py:1188] (6/7) 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] (6/7) 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 06:24:14,130 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.61 vs. limit=2.0 2023-04-28 06:24:14,495 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8668, 2.3919, 1.9320, 1.9570, 1.3788, 1.4269, 2.0405, 1.3387], device='cuda:6'), covar=tensor([0.1561, 0.1351, 0.1262, 0.1459, 0.2100, 0.1771, 0.0920, 0.1887], device='cuda:6'), in_proj_covar=tensor([0.0198, 0.0208, 0.0169, 0.0203, 0.0200, 0.0187, 0.0156, 0.0187], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:6') 2023-04-28 06:24:22,139 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.3573, 1.2602, 1.5788, 1.5309, 1.2556, 1.1832, 1.2224, 0.8283], device='cuda:6'), covar=tensor([0.0491, 0.0557, 0.0332, 0.0511, 0.0740, 0.0987, 0.0508, 0.0563], device='cuda:6'), in_proj_covar=tensor([0.0070, 0.0067, 0.0064, 0.0068, 0.0075, 0.0094, 0.0072, 0.0061], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-28 06:24:36,629 INFO [finetune.py:976] (6/7) Epoch 30, batch 3800, loss[loss=0.1468, simple_loss=0.2273, pruned_loss=0.03314, over 4735.00 frames. ], tot_loss[loss=0.1663, simple_loss=0.2413, pruned_loss=0.04567, over 954542.49 frames. ], batch size: 27, lr: 2.79e-03, grad_scale: 32.0 2023-04-28 06:25:05,039 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7881, 1.0726, 1.7904, 2.1892, 1.8233, 1.6683, 1.7181, 1.6971], device='cuda:6'), covar=tensor([0.4374, 0.6949, 0.6421, 0.5969, 0.6157, 0.8055, 0.7899, 0.9250], device='cuda:6'), in_proj_covar=tensor([0.0448, 0.0429, 0.0524, 0.0509, 0.0477, 0.0517, 0.0517, 0.0533], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-28 06:25:18,397 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.1051, 2.0599, 2.2809, 2.4533, 2.6623, 1.9900, 1.8648, 2.1918], device='cuda:6'), covar=tensor([0.0826, 0.1096, 0.0652, 0.0676, 0.0626, 0.0897, 0.0771, 0.0635], device='cuda:6'), in_proj_covar=tensor([0.0184, 0.0205, 0.0185, 0.0173, 0.0179, 0.0180, 0.0151, 0.0178], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-28 06:25:40,866 INFO [finetune.py:976] (6/7) Epoch 30, batch 3850, loss[loss=0.1385, simple_loss=0.2052, pruned_loss=0.0359, over 4706.00 frames. ], tot_loss[loss=0.1649, simple_loss=0.2397, pruned_loss=0.04511, over 954365.70 frames. ], batch size: 23, lr: 2.79e-03, grad_scale: 32.0 2023-04-28 06:26:12,319 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=169977.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:26:14,680 INFO [optim.py:369] (6/7) 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 06:26:45,748 INFO [finetune.py:976] (6/7) Epoch 30, batch 3900, loss[loss=0.1361, simple_loss=0.2151, pruned_loss=0.02856, over 4766.00 frames. ], tot_loss[loss=0.1641, simple_loss=0.2378, pruned_loss=0.04521, over 953494.82 frames. ], batch size: 26, lr: 2.79e-03, grad_scale: 32.0 2023-04-28 06:27:14,560 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-04-28 06:27:26,114 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.78 vs. limit=5.0 2023-04-28 06:27:43,927 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 30, batch 3950, loss[loss=0.1502, simple_loss=0.2295, pruned_loss=0.03549, over 4778.00 frames. ], tot_loss[loss=0.1618, simple_loss=0.2349, pruned_loss=0.04437, over 955075.50 frames. ], batch size: 29, lr: 2.79e-03, grad_scale: 32.0 2023-04-28 06:28:24,769 INFO [optim.py:369] (6/7) 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 06:28:27,249 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=170085.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:28:46,918 INFO [finetune.py:976] (6/7) Epoch 30, batch 4000, loss[loss=0.1794, simple_loss=0.2578, pruned_loss=0.05052, over 4841.00 frames. ], tot_loss[loss=0.1622, simple_loss=0.235, pruned_loss=0.04467, over 954860.17 frames. ], batch size: 47, lr: 2.79e-03, grad_scale: 32.0 2023-04-28 06:28:47,001 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=170103.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:28:53,826 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=170106.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:28:54,775 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([3.7264, 3.5951, 2.8152, 4.3155, 3.6928, 3.6992, 1.6646, 3.7023], device='cuda:6'), covar=tensor([0.1743, 0.1346, 0.3774, 0.1627, 0.4488, 0.1921, 0.5764, 0.2372], device='cuda:6'), in_proj_covar=tensor([0.0249, 0.0219, 0.0253, 0.0302, 0.0300, 0.0250, 0.0274, 0.0276], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-28 06:29:39,695 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=170142.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:29:50,075 INFO [finetune.py:976] (6/7) Epoch 30, batch 4050, loss[loss=0.169, simple_loss=0.2236, pruned_loss=0.05722, over 4140.00 frames. ], tot_loss[loss=0.1646, simple_loss=0.2379, pruned_loss=0.04563, over 955800.35 frames. ], batch size: 17, lr: 2.79e-03, grad_scale: 32.0 2023-04-28 06:30:22,184 INFO [optim.py:369] (6/7) 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 06:30:36,790 INFO [finetune.py:976] (6/7) Epoch 30, batch 4100, loss[loss=0.1495, simple_loss=0.2359, pruned_loss=0.03157, over 4753.00 frames. ], tot_loss[loss=0.1654, simple_loss=0.2395, pruned_loss=0.04565, over 956958.45 frames. ], batch size: 28, lr: 2.79e-03, grad_scale: 32.0 2023-04-28 06:30:47,306 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.6542, 1.0059, 1.6514, 2.1415, 1.7222, 1.6026, 1.6578, 1.6419], device='cuda:6'), covar=tensor([0.4221, 0.6549, 0.5713, 0.5104, 0.5334, 0.6971, 0.7313, 0.8869], device='cuda:6'), in_proj_covar=tensor([0.0447, 0.0427, 0.0521, 0.0507, 0.0475, 0.0515, 0.0516, 0.0531], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-28 06:31:08,389 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7911, 1.9888, 1.1245, 1.4643, 2.2942, 1.5939, 1.5083, 1.5938], device='cuda:6'), covar=tensor([0.0479, 0.0343, 0.0273, 0.0523, 0.0208, 0.0475, 0.0489, 0.0553], device='cuda:6'), in_proj_covar=tensor([0.0027, 0.0023, 0.0021, 0.0028, 0.0019, 0.0027, 0.0027, 0.0027], device='cuda:6'), out_proj_covar=tensor([0.0052, 0.0046, 0.0039, 0.0053, 0.0039, 0.0051, 0.0051, 0.0053], device='cuda:6') 2023-04-28 06:31:10,086 INFO [finetune.py:976] (6/7) Epoch 30, batch 4150, loss[loss=0.1547, simple_loss=0.2331, pruned_loss=0.03809, over 4862.00 frames. ], tot_loss[loss=0.166, simple_loss=0.2404, pruned_loss=0.04576, over 957067.13 frames. ], batch size: 34, lr: 2.79e-03, grad_scale: 32.0 2023-04-28 06:31:26,668 INFO [zipformer.py:1188] (6/7) 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] (6/7) 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 06:31:40,682 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.5723, 1.3589, 0.7646, 1.2683, 1.4092, 1.4108, 1.3313, 1.3844], device='cuda:6'), covar=tensor([0.0508, 0.0400, 0.0369, 0.0584, 0.0313, 0.0535, 0.0555, 0.0607], device='cuda:6'), in_proj_covar=tensor([0.0027, 0.0023, 0.0021, 0.0028, 0.0019, 0.0027, 0.0027, 0.0027], device='cuda:6'), out_proj_covar=tensor([0.0052, 0.0046, 0.0039, 0.0053, 0.0039, 0.0051, 0.0051, 0.0053], device='cuda:6') 2023-04-28 06:31:42,966 INFO [finetune.py:976] (6/7) Epoch 30, batch 4200, loss[loss=0.1605, simple_loss=0.2378, pruned_loss=0.04155, over 4906.00 frames. ], tot_loss[loss=0.1667, simple_loss=0.2416, pruned_loss=0.04591, over 957826.72 frames. ], batch size: 46, lr: 2.79e-03, grad_scale: 32.0 2023-04-28 06:31:49,497 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.97 vs. limit=2.0 2023-04-28 06:31:58,825 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=170325.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:32:16,769 INFO [finetune.py:976] (6/7) Epoch 30, batch 4250, loss[loss=0.1492, simple_loss=0.2267, pruned_loss=0.03588, over 4823.00 frames. ], tot_loss[loss=0.1656, simple_loss=0.2397, pruned_loss=0.04571, over 957618.73 frames. ], batch size: 30, lr: 2.79e-03, grad_scale: 32.0 2023-04-28 06:32:36,243 INFO [optim.py:369] (6/7) 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 06:32:38,780 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=170385.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:32:49,052 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 30, batch 4300, loss[loss=0.1601, simple_loss=0.2337, pruned_loss=0.04329, over 4835.00 frames. ], tot_loss[loss=0.163, simple_loss=0.2371, pruned_loss=0.04443, over 958409.47 frames. ], batch size: 38, lr: 2.79e-03, grad_scale: 32.0 2023-04-28 06:32:50,260 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=170403.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:32:54,322 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.84 vs. limit=5.0 2023-04-28 06:32:54,838 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-04-28 06:33:09,732 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9371, 1.1924, 1.7010, 1.7981, 1.7391, 1.7966, 1.6903, 1.6305], device='cuda:6'), covar=tensor([0.3778, 0.4991, 0.3909, 0.4221, 0.5140, 0.6824, 0.4291, 0.4189], device='cuda:6'), in_proj_covar=tensor([0.0345, 0.0376, 0.0334, 0.0346, 0.0354, 0.0397, 0.0364, 0.0338], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-28 06:33:10,857 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=170433.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:33:16,840 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=170442.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:33:22,276 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=170451.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:33:23,433 INFO [finetune.py:976] (6/7) Epoch 30, batch 4350, loss[loss=0.1246, simple_loss=0.2048, pruned_loss=0.02221, over 4797.00 frames. ], tot_loss[loss=0.1597, simple_loss=0.2334, pruned_loss=0.04303, over 957258.29 frames. ], batch size: 51, lr: 2.79e-03, grad_scale: 32.0 2023-04-28 06:33:28,877 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=170461.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:33:30,146 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-04-28 06:33:52,314 INFO [optim.py:369] (6/7) 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 06:34:04,623 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=170490.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:34:23,439 INFO [finetune.py:976] (6/7) Epoch 30, batch 4400, loss[loss=0.14, simple_loss=0.2141, pruned_loss=0.03295, over 4861.00 frames. ], tot_loss[loss=0.1612, simple_loss=0.2346, pruned_loss=0.04384, over 957277.29 frames. ], batch size: 31, lr: 2.79e-03, grad_scale: 32.0 2023-04-28 06:34:34,464 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.14 vs. limit=2.0 2023-04-28 06:34:47,088 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=170522.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 06:35:08,310 INFO [finetune.py:976] (6/7) Epoch 30, batch 4450, loss[loss=0.1298, simple_loss=0.1926, pruned_loss=0.03347, over 4694.00 frames. ], tot_loss[loss=0.1634, simple_loss=0.2377, pruned_loss=0.04458, over 957058.26 frames. ], batch size: 23, lr: 2.79e-03, grad_scale: 32.0 2023-04-28 06:35:19,134 INFO [zipformer.py:1188] (6/7) 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] (6/7) 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 06:35:41,960 INFO [finetune.py:976] (6/7) Epoch 30, batch 4500, loss[loss=0.2416, simple_loss=0.2903, pruned_loss=0.09641, over 4154.00 frames. ], tot_loss[loss=0.1651, simple_loss=0.2394, pruned_loss=0.04534, over 955544.97 frames. ], batch size: 65, lr: 2.79e-03, grad_scale: 32.0 2023-04-28 06:35:45,463 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-04-28 06:35:46,300 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7565, 1.3688, 4.6387, 4.3765, 4.0683, 4.4728, 4.3151, 4.0611], device='cuda:6'), covar=tensor([0.7145, 0.6015, 0.1006, 0.1744, 0.1061, 0.1819, 0.1261, 0.1550], device='cuda:6'), in_proj_covar=tensor([0.0307, 0.0306, 0.0402, 0.0406, 0.0345, 0.0411, 0.0314, 0.0361], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-28 06:35:59,495 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=170630.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:36:15,354 INFO [finetune.py:976] (6/7) Epoch 30, batch 4550, loss[loss=0.1352, simple_loss=0.2202, pruned_loss=0.02508, over 4927.00 frames. ], tot_loss[loss=0.1659, simple_loss=0.2407, pruned_loss=0.04557, over 957140.05 frames. ], batch size: 33, lr: 2.79e-03, grad_scale: 32.0 2023-04-28 06:36:33,303 INFO [optim.py:369] (6/7) 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 06:36:47,828 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 30, batch 4600, loss[loss=0.1647, simple_loss=0.2498, pruned_loss=0.03979, over 4910.00 frames. ], tot_loss[loss=0.1649, simple_loss=0.2398, pruned_loss=0.04499, over 956750.85 frames. ], batch size: 38, lr: 2.79e-03, grad_scale: 32.0 2023-04-28 06:37:10,399 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.34 vs. limit=5.0 2023-04-28 06:37:10,674 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.3963, 1.3897, 4.1209, 3.8584, 3.5793, 3.9546, 3.9334, 3.5732], device='cuda:6'), covar=tensor([0.7245, 0.5367, 0.1202, 0.1956, 0.1253, 0.1488, 0.1330, 0.1680], device='cuda:6'), in_proj_covar=tensor([0.0310, 0.0308, 0.0405, 0.0409, 0.0348, 0.0414, 0.0316, 0.0363], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], device='cuda:6') 2023-04-28 06:37:19,720 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 30, batch 4650, loss[loss=0.1539, simple_loss=0.228, pruned_loss=0.03991, over 4821.00 frames. ], tot_loss[loss=0.1637, simple_loss=0.2377, pruned_loss=0.04486, over 956991.27 frames. ], batch size: 38, lr: 2.79e-03, grad_scale: 32.0 2023-04-28 06:37:33,022 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.8153, 2.9451, 2.3684, 2.6466, 3.0511, 2.6133, 3.8954, 2.4792], device='cuda:6'), covar=tensor([0.3324, 0.2129, 0.3844, 0.2977, 0.1507, 0.2527, 0.1067, 0.3384], device='cuda:6'), in_proj_covar=tensor([0.0340, 0.0357, 0.0428, 0.0353, 0.0386, 0.0376, 0.0371, 0.0424], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-28 06:37:39,986 INFO [optim.py:369] (6/7) 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 06:37:41,582 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-04-28 06:37:49,662 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7565, 1.2192, 1.7390, 2.2742, 1.8275, 1.6585, 1.6988, 1.6627], device='cuda:6'), covar=tensor([0.4099, 0.6577, 0.5711, 0.4922, 0.4882, 0.7006, 0.6496, 0.8750], device='cuda:6'), in_proj_covar=tensor([0.0450, 0.0429, 0.0523, 0.0510, 0.0476, 0.0518, 0.0518, 0.0534], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-28 06:37:55,527 INFO [finetune.py:976] (6/7) Epoch 30, batch 4700, loss[loss=0.1405, simple_loss=0.2115, pruned_loss=0.0348, over 4917.00 frames. ], tot_loss[loss=0.1603, simple_loss=0.234, pruned_loss=0.04334, over 958189.44 frames. ], batch size: 37, lr: 2.79e-03, grad_scale: 32.0 2023-04-28 06:38:05,050 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=170817.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 06:38:07,745 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=192, metric=1.79 vs. limit=2.0 2023-04-28 06:38:08,136 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.9747, 1.5408, 2.0546, 2.4465, 2.0073, 1.9089, 2.0215, 1.9436], device='cuda:6'), covar=tensor([0.4678, 0.6935, 0.6526, 0.5991, 0.5997, 0.7849, 0.7898, 0.9082], device='cuda:6'), in_proj_covar=tensor([0.0449, 0.0428, 0.0522, 0.0510, 0.0475, 0.0517, 0.0517, 0.0533], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-28 06:38:29,291 INFO [finetune.py:976] (6/7) Epoch 30, batch 4750, loss[loss=0.1885, simple_loss=0.2672, pruned_loss=0.05495, over 4854.00 frames. ], tot_loss[loss=0.1597, simple_loss=0.2329, pruned_loss=0.04323, over 958309.46 frames. ], batch size: 49, lr: 2.79e-03, grad_scale: 32.0 2023-04-28 06:38:34,559 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.1819, 1.7445, 1.9440, 2.4680, 2.0341, 1.7083, 1.5328, 1.9123], device='cuda:6'), covar=tensor([0.2472, 0.2938, 0.1628, 0.1772, 0.2324, 0.2323, 0.4245, 0.1974], device='cuda:6'), in_proj_covar=tensor([0.0296, 0.0248, 0.0231, 0.0317, 0.0227, 0.0237, 0.0232, 0.0187], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:6') 2023-04-28 06:38:47,637 INFO [optim.py:369] (6/7) 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 06:38:55,123 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.09 vs. limit=2.0 2023-04-28 06:39:16,798 INFO [finetune.py:976] (6/7) Epoch 30, batch 4800, loss[loss=0.1627, simple_loss=0.2408, pruned_loss=0.04231, over 4905.00 frames. ], tot_loss[loss=0.1618, simple_loss=0.2358, pruned_loss=0.04392, over 956765.81 frames. ], batch size: 37, lr: 2.79e-03, grad_scale: 32.0 2023-04-28 06:39:45,898 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=170925.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:39:47,781 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=170928.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:40:11,225 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([4.6722, 4.6454, 3.1842, 5.2767, 4.7292, 4.5697, 2.1135, 4.6660], device='cuda:6'), covar=tensor([0.1485, 0.0958, 0.3238, 0.0955, 0.3640, 0.1602, 0.5357, 0.1799], device='cuda:6'), in_proj_covar=tensor([0.0249, 0.0219, 0.0253, 0.0301, 0.0299, 0.0250, 0.0275, 0.0275], device='cuda:6'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:6') 2023-04-28 06:40:21,163 INFO [finetune.py:976] (6/7) Epoch 30, batch 4850, loss[loss=0.1577, simple_loss=0.2375, pruned_loss=0.03893, over 4832.00 frames. ], tot_loss[loss=0.1633, simple_loss=0.2378, pruned_loss=0.04436, over 955228.65 frames. ], batch size: 49, lr: 2.79e-03, grad_scale: 32.0 2023-04-28 06:40:38,868 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.3380, 2.0970, 1.7824, 1.8115, 2.2139, 1.8156, 2.6087, 1.5804], device='cuda:6'), covar=tensor([0.3597, 0.2220, 0.4038, 0.3163, 0.1863, 0.2693, 0.1649, 0.4209], device='cuda:6'), in_proj_covar=tensor([0.0341, 0.0359, 0.0429, 0.0354, 0.0388, 0.0377, 0.0373, 0.0425], device='cuda:6'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:6') 2023-04-28 06:40:52,180 INFO [optim.py:369] (6/7) 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 06:40:59,542 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=170989.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:41:06,675 INFO [scaling.py:679] (6/7) Whitening: num_groups=1, num_channels=384, metric=4.51 vs. limit=5.0 2023-04-28 06:41:19,850 INFO [finetune.py:976] (6/7) Epoch 30, batch 4900, loss[loss=0.1656, simple_loss=0.2332, pruned_loss=0.049, over 4899.00 frames. ], tot_loss[loss=0.1639, simple_loss=0.2389, pruned_loss=0.04449, over 955558.46 frames. ], batch size: 37, lr: 2.79e-03, grad_scale: 32.0 2023-04-28 06:41:21,686 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7966, 1.6171, 1.9760, 2.0671, 1.5765, 1.3888, 1.6733, 1.0349], device='cuda:6'), covar=tensor([0.0428, 0.0579, 0.0358, 0.0600, 0.0736, 0.1085, 0.0531, 0.0517], device='cuda:6'), in_proj_covar=tensor([0.0070, 0.0067, 0.0065, 0.0069, 0.0076, 0.0095, 0.0072, 0.0062], device='cuda:6'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:6') 2023-04-28 06:42:25,535 INFO [finetune.py:976] (6/7) Epoch 30, batch 4950, loss[loss=0.1572, simple_loss=0.249, pruned_loss=0.03274, over 4917.00 frames. ], tot_loss[loss=0.1653, simple_loss=0.2406, pruned_loss=0.04501, over 955751.69 frames. ], batch size: 33, lr: 2.78e-03, grad_scale: 64.0 2023-04-28 06:42:45,071 INFO [optim.py:369] (6/7) 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 06:42:52,362 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=171092.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:42:59,425 INFO [finetune.py:976] (6/7) Epoch 30, batch 5000, loss[loss=0.1655, simple_loss=0.2394, pruned_loss=0.04585, over 4865.00 frames. ], tot_loss[loss=0.1644, simple_loss=0.2391, pruned_loss=0.04483, over 957133.34 frames. ], batch size: 34, lr: 2.78e-03, grad_scale: 64.0 2023-04-28 06:43:09,449 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=171117.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 06:43:33,018 INFO [finetune.py:976] (6/7) Epoch 30, batch 5050, loss[loss=0.1317, simple_loss=0.1905, pruned_loss=0.0364, over 4174.00 frames. ], tot_loss[loss=0.1627, simple_loss=0.2369, pruned_loss=0.04431, over 957235.60 frames. ], batch size: 18, lr: 2.78e-03, grad_scale: 64.0 2023-04-28 06:43:33,128 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=171153.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:43:41,327 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=171165.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:43:43,251 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.7332, 2.3787, 1.9630, 2.3073, 1.6741, 2.0585, 1.9632, 1.5042], device='cuda:6'), covar=tensor([0.2020, 0.1118, 0.0860, 0.1110, 0.3177, 0.1036, 0.1762, 0.2629], device='cuda:6'), in_proj_covar=tensor([0.0283, 0.0300, 0.0218, 0.0274, 0.0309, 0.0256, 0.0250, 0.0261], device='cuda:6'), out_proj_covar=tensor([1.1264e-04, 1.1776e-04, 8.5372e-05, 1.0726e-04, 1.2442e-04, 1.0054e-04, 1.0027e-04, 1.0260e-04], device='cuda:6') 2023-04-28 06:43:52,459 INFO [optim.py:369] (6/7) 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 06:44:02,207 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=171196.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:44:06,504 INFO [finetune.py:976] (6/7) Epoch 30, batch 5100, loss[loss=0.1459, simple_loss=0.2185, pruned_loss=0.03666, over 4763.00 frames. ], tot_loss[loss=0.1608, simple_loss=0.234, pruned_loss=0.04381, over 956186.69 frames. ], batch size: 27, lr: 2.78e-03, grad_scale: 64.0 2023-04-28 06:44:21,438 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=171225.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:44:40,214 INFO [finetune.py:976] (6/7) Epoch 30, batch 5150, loss[loss=0.1961, simple_loss=0.2752, pruned_loss=0.05846, over 4831.00 frames. ], tot_loss[loss=0.1615, simple_loss=0.2345, pruned_loss=0.04429, over 956408.53 frames. ], batch size: 47, lr: 2.78e-03, grad_scale: 64.0 2023-04-28 06:44:43,266 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=171257.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:44:53,819 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=171273.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:45:05,608 INFO [optim.py:369] (6/7) 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 06:45:07,336 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=171284.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:45:35,543 INFO [finetune.py:976] (6/7) Epoch 30, batch 5200, loss[loss=0.235, simple_loss=0.3153, pruned_loss=0.07732, over 4263.00 frames. ], tot_loss[loss=0.1636, simple_loss=0.2374, pruned_loss=0.04489, over 956225.59 frames. ], batch size: 65, lr: 2.78e-03, grad_scale: 32.0 2023-04-28 06:45:39,799 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.2402, 1.6567, 1.4673, 1.8551, 1.8740, 1.9783, 1.6033, 3.9344], device='cuda:6'), covar=tensor([0.0577, 0.0742, 0.0751, 0.1115, 0.0588, 0.0613, 0.0687, 0.0124], device='cuda:6'), in_proj_covar=tensor([0.0037, 0.0038, 0.0039, 0.0043, 0.0040, 0.0037, 0.0038, 0.0054], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0013, 0.0014, 0.0015, 0.0014, 0.0013, 0.0013, 0.0016], device='cuda:6') 2023-04-28 06:46:31,214 INFO [zipformer.py:1188] (6/7) warmup_begin=2666.7, warmup_end=3333.3, batch_count=171345.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:46:36,592 INFO [finetune.py:976] (6/7) Epoch 30, batch 5250, loss[loss=0.1955, simple_loss=0.2908, pruned_loss=0.05009, over 4802.00 frames. ], tot_loss[loss=0.1644, simple_loss=0.2389, pruned_loss=0.04502, over 954491.86 frames. ], batch size: 45, lr: 2.78e-03, grad_scale: 32.0 2023-04-28 06:47:18,740 INFO [optim.py:369] (6/7) 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 06:47:43,265 INFO [finetune.py:976] (6/7) Epoch 30, batch 5300, loss[loss=0.1321, simple_loss=0.2063, pruned_loss=0.02894, over 4810.00 frames. ], tot_loss[loss=0.1649, simple_loss=0.2394, pruned_loss=0.04525, over 954090.11 frames. ], batch size: 25, lr: 2.78e-03, grad_scale: 32.0 2023-04-28 06:47:50,523 INFO [zipformer.py:1188] (6/7) warmup_begin=3333.3, warmup_end=4000.0, batch_count=171406.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:48:02,217 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.50 vs. limit=2.0 2023-04-28 06:48:41,623 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 30, batch 5350, loss[loss=0.1401, simple_loss=0.2138, pruned_loss=0.03317, over 4785.00 frames. ], tot_loss[loss=0.1647, simple_loss=0.2391, pruned_loss=0.04515, over 953701.89 frames. ], batch size: 26, lr: 2.78e-03, grad_scale: 32.0 2023-04-28 06:49:03,281 INFO [optim.py:369] (6/7) 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 06:49:10,274 INFO [scaling.py:679] (6/7) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-04-28 06:49:15,631 INFO [zipformer.py:1188] (6/7) 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] (6/7) Epoch 30, batch 5400, loss[loss=0.1756, simple_loss=0.2462, pruned_loss=0.05253, over 4853.00 frames. ], tot_loss[loss=0.1648, simple_loss=0.2386, pruned_loss=0.0455, over 952343.27 frames. ], batch size: 49, lr: 2.78e-03, grad_scale: 32.0 2023-04-28 06:49:51,782 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=171552.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:49:52,350 INFO [finetune.py:976] (6/7) Epoch 30, batch 5450, loss[loss=0.1604, simple_loss=0.2331, pruned_loss=0.0439, over 4876.00 frames. ], tot_loss[loss=0.1624, simple_loss=0.2359, pruned_loss=0.04445, over 955675.70 frames. ], batch size: 31, lr: 2.78e-03, grad_scale: 32.0 2023-04-28 06:49:56,124 INFO [zipformer.py:1188] (6/7) 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] (6/7) 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 06:50:12,291 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=171584.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:50:25,850 INFO [finetune.py:976] (6/7) Epoch 30, batch 5500, loss[loss=0.1583, simple_loss=0.2384, pruned_loss=0.03912, over 4834.00 frames. ], tot_loss[loss=0.1594, simple_loss=0.2325, pruned_loss=0.04315, over 955595.97 frames. ], batch size: 33, lr: 2.78e-03, grad_scale: 32.0 2023-04-28 06:50:25,983 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([2.0009, 1.3292, 1.4551, 1.7297, 2.0134, 1.6533, 1.4595, 1.4082], device='cuda:6'), covar=tensor([0.1669, 0.2187, 0.2314, 0.1466, 0.1133, 0.2201, 0.2510, 0.2937], device='cuda:6'), in_proj_covar=tensor([0.0314, 0.0308, 0.0349, 0.0285, 0.0324, 0.0306, 0.0300, 0.0376], device='cuda:6'), out_proj_covar=tensor([6.3650e-05, 6.2773e-05, 7.2912e-05, 5.6819e-05, 6.5944e-05, 6.3343e-05, 6.1795e-05, 7.9403e-05], device='cuda:6') 2023-04-28 06:50:45,013 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=171632.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:50:59,648 INFO [finetune.py:976] (6/7) Epoch 30, batch 5550, loss[loss=0.2247, simple_loss=0.3024, pruned_loss=0.0735, over 4740.00 frames. ], tot_loss[loss=0.1623, simple_loss=0.2358, pruned_loss=0.04444, over 954848.61 frames. ], batch size: 59, lr: 2.78e-03, grad_scale: 32.0 2023-04-28 06:51:38,774 INFO [optim.py:369] (6/7) 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 06:52:00,722 INFO [zipformer.py:1188] (6/7) warmup_begin=1333.3, warmup_end=2000.0, batch_count=171701.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:52:01,826 INFO [finetune.py:976] (6/7) Epoch 30, batch 5600, loss[loss=0.1659, simple_loss=0.2467, pruned_loss=0.0425, over 4751.00 frames. ], tot_loss[loss=0.1633, simple_loss=0.2378, pruned_loss=0.04444, over 953105.67 frames. ], batch size: 27, lr: 2.78e-03, grad_scale: 32.0 2023-04-28 06:52:55,573 INFO [zipformer.py:1188] (6/7) warmup_begin=2000.0, warmup_end=2666.7, batch_count=171748.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:52:58,356 INFO [finetune.py:976] (6/7) Epoch 30, batch 5650, loss[loss=0.1783, simple_loss=0.2514, pruned_loss=0.05258, over 4805.00 frames. ], tot_loss[loss=0.1639, simple_loss=0.2392, pruned_loss=0.04434, over 954888.26 frames. ], batch size: 38, lr: 2.78e-03, grad_scale: 32.0 2023-04-28 06:53:37,490 INFO [optim.py:369] (6/7) 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 06:53:39,351 INFO [zipformer.py:2441] (6/7) attn_weights_entropy = tensor([1.4207, 1.8198, 1.6571, 1.9979, 1.9224, 1.9624, 1.6905, 3.4024], device='cuda:6'), covar=tensor([0.0620, 0.0730, 0.0767, 0.1129, 0.0588, 0.0472, 0.0707, 0.0201], device='cuda:6'), in_proj_covar=tensor([0.0038, 0.0039, 0.0039, 0.0043, 0.0040, 0.0038, 0.0038, 0.0055], device='cuda:6'), out_proj_covar=tensor([0.0012, 0.0013, 0.0014, 0.0015, 0.0014, 0.0013, 0.0013, 0.0016], device='cuda:6') 2023-04-28 06:53:50,971 INFO [zipformer.py:1188] (6/7) warmup_begin=666.7, warmup_end=1333.3, batch_count=171796.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:54:00,720 INFO [finetune.py:976] (6/7) Epoch 30, batch 5700, loss[loss=0.1623, simple_loss=0.2199, pruned_loss=0.05241, over 4266.00 frames. ], tot_loss[loss=0.1628, simple_loss=0.2367, pruned_loss=0.04449, over 936450.59 frames. ], batch size: 18, lr: 2.78e-03, grad_scale: 32.0 2023-04-28 06:54:33,249 INFO [finetune.py:1241] (6/7) Done!